Winter browse by snowshoe hares, particularly during peaks in their population cycles, may influence wider treeline dynamics by killing spruce seedlings and young trees in areas where warming conditions might otherwise favor treeline advance. However, young trees escape winter browse mortality if they are sufficiently protected by deep snow. Simultaneously, hares may benefit from snow depth evolution over the course of a snow season, as deepening snow lifts them into previously unbrowsed canopy (the snow elevator). Still, it is unclear how differences in the temporal patterns of snow depth within a snow season ultimately affect either browse availability for hares or spruce vulnerability to browse death. To explore this question, we are developing a simple Python model that integrates estimated availability of spruce browse in 1 cm height bins, hare density, and a range of snow season scenarios, and then simulates hare consumption of spruce over the snow seasons of a hare population cycle. In particular, we altered recent daily time series of SnowModel snow seasons extracted from our region of interest (southern Brooks Range, Alaska) to develop sixteen snow season scenarios that vary in length (short or long season), snow onset date (early or late onset), maximum snow depth (low or high maximum snow depth), and snow accumulation rate (slow or fast accumulation). Our results could help clarify potential impacts of changing snow regimes on browse-mediated treeline dynamics.
  [poster]
Winter browse by snowshoe hares, particularly during peaks in their population cycles, may influence wider treeline dynamics by killing spruce seedlings and young trees in areas where warming conditions might otherwise favor treeline advance. However, young trees escape winter browse mortality if they are sufficiently protected by deep snow. Simultaneously, hares may benefit from snow depth evolution over the course of a snow season, as deepening snow lifts them into previously unbrowsed canopy (the snow elevator). Still, it is unclear how differences in the temporal patterns of snow depth within a snow season ultimately affect either browse availability for hares or spruce vulnerability to browse death. To explore this question, we are developing a simple Python model that integrates estimated availability of spruce browse in 1 cm height bins, hare density, and a range of snow season scenarios, and then simulates hare consumption of spruce over the snow seasons of a hare population cycle. In particular, we altered recent daily time series of SnowModel snow seasons extracted from our region of interest (southern Brooks Range, Alaska) to develop sixteen snow season scenarios that vary in length (short or long season), snow onset date (early or late onset), maximum snow depth (low or high maximum snow depth), and snow accumulation rate (slow or fast accumulation). Our results could help clarify potential impacts of changing snow regimes on browse-mediated treeline dynamics.

Snowshoe hares (Lepus americanus) are known to have significant effects on vegetation dynamics in their habitats. In this study, we investigated the potential of satellite remote sensing in combination with output from SnowModel to provide insights into snowshoe hare density. We compiled Landsat satellite data and snowpack information from SnowModel for seven site locations along the Dalton Highway in Alaska, where hare density was sampled annually every April from 2013 to 2023. Using a linear mixed-effects model analysis, we found that the average June Normalized Difference Vegetation Index (NDVI) value of the previous year, the average June NDVI value of the current year, and snow depth in May were all positive and significant predictors of hare density. These three predictors collectively explained 45% (marginal r² = 0.45) of the variability in observed hare density across all seven sites and 64% (conditional r² = 0.63) at the site level. We observed that average NDVI values from earlier (e.g., May) or later (e.g., August) months were not as strong predictors as the average NDVI value in June. This difference in predictive power could be attributed to June typically offering the most cloud- and snow-free imagery across the study sites. Our findings suggest that snow depth in May is the best predictor of snowshoe hare density, followed by the average June NDVI of the previous year and the average June NDVI of the current year. From an ecological standpoint, the significance of these predictors may suggest that satellite imagery and SnowModel outputs may be providing important surrogate information for hare browse availability and predation pressure. Overall, our preliminary results indicate that combining SnowModel outputs with greenness information (NDVI) from satellite remote sensing can provide valuable insights into snowshoe hare density dynamics.

Snowshoe hares (Lepus americanus) are known to have significant effects on vegetation dynamics in their habitats. In this study, we investigated the potential of satellite remote sensing in combination with output from SnowModel to provide insights into snowshoe hare density. We compiled Landsat satellite data and snowpack information from SnowModel for seven site locations along the Dalton Highway in Alaska, where hare density was sampled annually every April from 2013 to 2023. Using a linear mixed-effects model analysis, we found that the average June Normalized Difference Vegetation Index (NDVI) value of the previous year, the average June NDVI value of the current year, and snow depth in May were all positive and significant predictors of hare density. These three predictors collectively explained 45% (marginal r² = 0.45) of the variability in observed hare density across all seven sites and 64% (conditional r² = 0.63) at the site level. We observed that average NDVI values from earlier (e.g., May) or later (e.g., August) months were not as strong predictors as the average NDVI value in June. This difference in predictive power could be attributed to June typically offering the most cloud- and snow-free imagery across the study sites. Our findings suggest that snow depth in May is the best predictor of snowshoe hare density, followed by the average June NDVI of the previous year and the average June NDVI of the current year. From an ecological standpoint, the significance of these predictors may suggest that satellite imagery and SnowModel outputs may be providing important surrogate information for hare browse availability and predation pressure. Overall, our preliminary results indicate that combining SnowModel outputs with greenness information (NDVI) from satellite remote sensing can provide valuable insights into snowshoe hare density dynamics.

Timing of migratory waterfowl arrival at northern breeding sites is partly determined by environmental conditions at the breeding sites and on the landscapes along migration routes. Presumably, the development of open (liquid) water on the breeding landscape is particularly important to waterfowl arrival timing, as well as breeding habitat preference, but the scale of Arctic landscapes and the challenges of employing remote sensing at high latitudes complicate the detection of open water at ecologically relevant temporal and spatial scales. To address this challenge, we compiled estimates of first springtime open-water dates for the years 2019–2023 for ~80 locations across northern Alaska and northwest Yukon at which data from acoustic recording units (ARUs) allow automated identification of waterfowl arrival timing and breeding habitat preference. We developed our open-water date estimates using datasets derived from: Sentinel-2 imagery, including both human image evaluation and Dynamic World automated image classification; a water detection algorithm employing synthetic aperture radar (SAR) from Sentinel-1; and Soil Moisture Active Passive (SMAP) first seasonal soil moisture detections. We are interested in 1) relationships among estimated open water dates derived from different sources, 2) relationships between remotely derived estimates of open water dates and the arrival dates of migratory waterfowl, and also 3) identifying spatial scales at which summertime cover of open water best relates to waterfowl activity during that time (i.e. breeding season). In our preliminary analysis it appears that the frequent overpass and relatively high spatial resolution of sensors on the Sentinel-2 satellites are important for providing data sufficiently time-dense and fine-scale to identify the appearance of open water relatively precisely; Sentinel-2 first open water detections also appeared most closely related to waterfowl arrival timing, both among sites within years and across years. Perhaps unsurprisingly, detection of still water (lakes, ponds) appears more closely related to waterfowl arrival timing than detection of running water (rivers, creeks), although distinguishing between these water types can be challenging for automated landcover classification algorithms. We also found that summer water cover measured within the largest spatial buffer we sampled around each ARU (5 km) was most strongly related to overall waterfowl vocal activity, suggesting that high water cover over large landscape extents is more important in determining breeding habitat occupancy than high water cover at comparatively local scales (<1000 m). In addition to gaining further insight into waterfowl behavior, we expect that our ongoing analyses will provide insights into challenges and relevance of applying widely available remotely sensed datasets in wildlife research.
  [poster]
Timing of migratory waterfowl arrival at northern breeding sites is partly determined by environmental conditions at the breeding sites and on the landscapes along migration routes. Presumably, the development of open (liquid) water on the breeding landscape is particularly important to waterfowl arrival timing, as well as breeding habitat preference, but the scale of Arctic landscapes and the challenges of employing remote sensing at high latitudes complicate the detection of open water at ecologically relevant temporal and spatial scales. To address this challenge, we compiled estimates of first springtime open-water dates for the years 2019–2023 for ~80 locations across northern Alaska and northwest Yukon at which data from acoustic recording units (ARUs) allow automated identification of waterfowl arrival timing and breeding habitat preference. We developed our open-water date estimates using datasets derived from: Sentinel-2 imagery, including both human image evaluation and Dynamic World automated image classification; a water detection algorithm employing synthetic aperture radar (SAR) from Sentinel-1; and Soil Moisture Active Passive (SMAP) first seasonal soil moisture detections. We are interested in 1) relationships among estimated open water dates derived from different sources, 2) relationships between remotely derived estimates of open water dates and the arrival dates of migratory waterfowl, and also 3) identifying spatial scales at which summertime cover of open water best relates to waterfowl activity during that time (i.e. breeding season). In our preliminary analysis it appears that the frequent overpass and relatively high spatial resolution of sensors on the Sentinel-2 satellites are important for providing data sufficiently time-dense and fine-scale to identify the appearance of open water relatively precisely; Sentinel-2 first open water detections also appeared most closely related to waterfowl arrival timing, both among sites within years and across years. Perhaps unsurprisingly, detection of still water (lakes, ponds) appears more closely related to waterfowl arrival timing than detection of running water (rivers, creeks), although distinguishing between these water types can be challenging for automated landcover classification algorithms. We also found that summer water cover measured within the largest spatial buffer we sampled around each ARU (5 km) was most strongly related to overall waterfowl vocal activity, suggesting that high water cover over large landscape extents is more important in determining breeding habitat occupancy than high water cover at comparatively local scales (<1000 m). In addition to gaining further insight into waterfowl behavior, we expect that our ongoing analyses will provide insights into challenges and relevance of applying widely available remotely sensed datasets in wildlife research.

Wildfires have been increasing in frequency across Arctic and boreal biomes for the last several decades, with important ecological, social and economic consequences. The use of remote sensing data to map burned area is an important tool to track spatiotemporal patterns in burned area and assess potential consequences. In this work we use a convolutional neural network-based approach to map burned area across boreal and tundra biomes in North America using Landsat and Sentinel-2 imagery. We leveraged existing reference data from the United States and Canadian governments to train our model, using the entire fire history from 1985 - 2020, and achieved promising results with an Intersection Over Union score of 0.88, an F1 (harmonic mean of precision and recall) score of 0.93 and an overall accuracy of 0.98 on unseen testing data. In addition, we found that courser resolution MODIS based products tend to underestimate burned area in this region. This is the first study to our knowledge to use such an extensive historical fire record in combination with both Landsat and Sentinel-2 imagery to model burned area in arctic-boreal North America. This methodology could be leveraged to map burned area at 30 m resolution in other domains, including high latitude Eurasia.

Wildfires have been increasing in frequency across Arctic and boreal biomes for the last several decades, with important ecological, social and economic consequences. The use of remote sensing data to map burned area is an important tool to track spatiotemporal patterns in burned area and assess potential consequences. In this work we use a convolutional neural network-based approach to map burned area across boreal and tundra biomes in North America using Landsat and Sentinel-2 imagery. We leveraged existing reference data from the United States and Canadian governments to train our model, using the entire fire history from 1985 - 2020, and achieved promising results with an Intersection Over Union score of 0.88, an F1 (harmonic mean of precision and recall) score of 0.93 and an overall accuracy of 0.98 on unseen testing data. In addition, we found that courser resolution MODIS based products tend to underestimate burned area in this region. This is the first study to our knowledge to use such an extensive historical fire record in combination with both Landsat and Sentinel-2 imagery to model burned area in arctic-boreal North America. This methodology could be leveraged to map burned area at 30 m resolution in other domains, including high latitude Eurasia.

The Arctic-boreal region is warming rapidly, with consequences for northern ecosystems and global climate. Fires across the Arctic-boreal region are a major natural disturbance mechanism that initiate climate warming (positive) and cooling (negative) feedbacks. Understanding the net forcing effect from boreal fire on climate is crucial in managing and mitigating climate change impacts of boreal fires. Here we report radiative forcing estimates from boreal forest fires across Alaska and Western Canada (Arctic Boreal Vulnerability Experiment-domain). Our results integrate the effect of greenhouse gas emissions (warming) and aerosols emission (net cooling) have through direct combustion, post-fire vegetation recovery sequestering carbon (cooling), fire-induced permafrost degradation emitting CO2 and CH4 (warming), and changes in surface albedo (cooling). Alaskan fires are on average climate warming (1.34±2.95 W/m2 per burned area) – uncertainty given as spatial standard deviation, while Canadian fires show on average a climate cooling (?2.26±2.48 W/m2 per burned area) effect. The emissions from the combustion of organic soils and post-fire permafrost thaw dominate the positive feedback for Alaskan fires, whereas the cooling effect of post-fire changes in surface albedo because of prolonged spring snow cover dominate for the western Canadian fires. Our work demonstrates large-scale spatial variability in the climate feedbacks from North American boreal forest fires. Such fine-scale spatial information on the warming and cooling influences of forest fires could be useful in designing forest management and fire suppression activities informed by climate impacts
  [poster] 
The Arctic-boreal region is warming rapidly, with consequences for northern ecosystems and global climate. Fires across the Arctic-boreal region are a major natural disturbance mechanism that initiate climate warming (positive) and cooling (negative) feedbacks. Understanding the net forcing effect from boreal fire on climate is crucial in managing and mitigating climate change impacts of boreal fires. Here we report radiative forcing estimates from boreal forest fires across Alaska and Western Canada (Arctic Boreal Vulnerability Experiment-domain). Our results integrate the effect of greenhouse gas emissions (warming) and aerosols emission (net cooling) have through direct combustion, post-fire vegetation recovery sequestering carbon (cooling), fire-induced permafrost degradation emitting CO2 and CH4 (warming), and changes in surface albedo (cooling). Alaskan fires are on average climate warming (1.34±2.95 W/m2 per burned area) – uncertainty given as spatial standard deviation, while Canadian fires show on average a climate cooling (?2.26±2.48 W/m2 per burned area) effect. The emissions from the combustion of organic soils and post-fire permafrost thaw dominate the positive feedback for Alaskan fires, whereas the cooling effect of post-fire changes in surface albedo because of prolonged spring snow cover dominate for the western Canadian fires. Our work demonstrates large-scale spatial variability in the climate feedbacks from North American boreal forest fires. Such fine-scale spatial information on the warming and cooling influences of forest fires could be useful in designing forest management and fire suppression activities informed by climate impacts

Circum-boreal systems are crucial carbon sinks that are experiencing amplified warming and wildfire activity. Across the North American boreal, short-interval reburn events, where forests do not fully recover between two sequential fire events, alter fire effects and ecological response with broad implications for biophysical feedbacks to climate. Wildfires alter land surface properties, including albedo and land surface temperature, impacting regional climate and ecosystem dynamics. Yet the understanding of how short-interval reburns impact land surface temperature and albedo dynamics is unclear in boreal North American forests, and as fire activity increases, we see a boost in short-interval disturbance events. Here, we seek to address this issue by deriving albedo from Landsat 8 (30 m) for locations that experience reburn events to determine the impact on key biophysical properties. We sample fires where reburns occurred from 2013 to 2022 and where the initial burn occurred from 1984 to 2022. Reburn patches were identified by a combination of overlaps between fire perimeters and burned pixels derived from Landsat. Pixels were sampled across once burned (i.e., a single fire event), twice burned (i.e., reburned patches), and unburned patches to serve as controls. We used Landsat 5, 7, and 8 to calculate burn severity as dNBR. We used Landsat 8 to derive blue sky albedo and land surface temperature. Individual pixel values were extracted by points. We compared albedo across unburned, once-burned, and twice-burned sites. Reburn events result in higher albedo during snow-on months. Twice-burned pixels were statistically different from single-burned pixels both immediately after the fire and the 2-5 years following the fire. Short-interval disturbances can alter the anticipated ecological response post-disturbance and can impact ecosystem resilience and energy exchange with broader implications that may result in shifts from carbon sinks to sources.
  [poster] 
Circum-boreal systems are crucial carbon sinks that are experiencing amplified warming and wildfire activity. Across the North American boreal, short-interval reburn events, where forests do not fully recover between two sequential fire events, alter fire effects and ecological response with broad implications for biophysical feedbacks to climate. Wildfires alter land surface properties, including albedo and land surface temperature, impacting regional climate and ecosystem dynamics. Yet the understanding of how short-interval reburns impact land surface temperature and albedo dynamics is unclear in boreal North American forests, and as fire activity increases, we see a boost in short-interval disturbance events. Here, we seek to address this issue by deriving albedo from Landsat 8 (30 m) for locations that experience reburn events to determine the impact on key biophysical properties. We sample fires where reburns occurred from 2013 to 2022 and where the initial burn occurred from 1984 to 2022. Reburn patches were identified by a combination of overlaps between fire perimeters and burned pixels derived from Landsat. Pixels were sampled across once burned (i.e., a single fire event), twice burned (i.e., reburned patches), and unburned patches to serve as controls. We used Landsat 5, 7, and 8 to calculate burn severity as dNBR. We used Landsat 8 to derive blue sky albedo and land surface temperature. Individual pixel values were extracted by points. We compared albedo across unburned, once-burned, and twice-burned sites. Reburn events result in higher albedo during snow-on months. Twice-burned pixels were statistically different from single-burned pixels both immediately after the fire and the 2-5 years following the fire. Short-interval disturbances can alter the anticipated ecological response post-disturbance and can impact ecosystem resilience and energy exchange with broader implications that may result in shifts from carbon sinks to sources.

Rising temperatures and drier conditions associated with climate change have intensified the occurrence of wildfires, presenting a growing risk to the integrity and function of boreal forests. Fire activity is believed to be predominantly influenced by climatic factors (top-down control), yet the age and composition of forest stands can also play significant roles. An example of this bottom-up control is age-dependent burning, where young stands are less susceptible to fire until fuel levels reach a threshold conducive to ignition. However, the climate and its associated effects on the fire regime have already exceeded historical ranges of variability, which alongside considerably higher fire severity, might override age-dependent burning and result in increased instances of reburning. To assess where and when reburning occurs and determine the consequences of reburning for post-fire successional trajectories we are synthesizing foundational remote sensing data with existing field data throughout the ABoVE domain. Here we have created a new high-resolution (30m x 30m) mapping product, quantifying the annual areas burned and reburned from 2010 to 2022 using Landsat observations. During the 12 years, 629,258 hectares of the ABoVE domain (about 1.14%) had a single burn event, 9,209 hectares (0.0167%) had two burn events, and 71 hectares had a three burn events. Through the next steps of integrating forest stand age, fire weather indices, land cover, tree above ground biomass by functional group, and forest stress we will determine how interactions between environmental and vegetation characteristics impact age-dependent burning, culminating in maps predicting vulnerability to reburning. Understanding the drivers of reburning is critical to understanding the vulnerability and resilience of arctic and boreal ecosystems in this changing environment. Shorter fire return intervals could trigger the combustion of legacy carbon (C) and alter ecosystems that have historically served as a net carbon sink, potentially shifting them toward net carbon loss. Additionally, alterations in fire frequency and severity may induce different successional pathways following wildfires, transitioning these forests from the historical black spruce stands to those dominated by pine or broadleaf deciduous species.
 
Rising temperatures and drier conditions associated with climate change have intensified the occurrence of wildfires, presenting a growing risk to the integrity and function of boreal forests. Fire activity is believed to be predominantly influenced by climatic factors (top-down control), yet the age and composition of forest stands can also play significant roles. An example of this bottom-up control is age-dependent burning, where young stands are less susceptible to fire until fuel levels reach a threshold conducive to ignition. However, the climate and its associated effects on the fire regime have already exceeded historical ranges of variability, which alongside considerably higher fire severity, might override age-dependent burning and result in increased instances of reburning. To assess where and when reburning occurs and determine the consequences of reburning for post-fire successional trajectories we are synthesizing foundational remote sensing data with existing field data throughout the ABoVE domain. Here we have created a new high-resolution (30m x 30m) mapping product, quantifying the annual areas burned and reburned from 2010 to 2022 using Landsat observations. During the 12 years, 629,258 hectares of the ABoVE domain (about 1.14%) had a single burn event, 9,209 hectares (0.0167%) had two burn events, and 71 hectares had a three burn events. Through the next steps of integrating forest stand age, fire weather indices, land cover, tree above ground biomass by functional group, and forest stress we will determine how interactions between environmental and vegetation characteristics impact age-dependent burning, culminating in maps predicting vulnerability to reburning. Understanding the drivers of reburning is critical to understanding the vulnerability and resilience of arctic and boreal ecosystems in this changing environment. Shorter fire return intervals could trigger the combustion of legacy carbon (C) and alter ecosystems that have historically served as a net carbon sink, potentially shifting them toward net carbon loss. Additionally, alterations in fire frequency and severity may induce different successional pathways following wildfires, transitioning these forests from the historical black spruce stands to those dominated by pine or broadleaf deciduous species.

Boreal peatlands have provided rich reservoirs of stored carbon for millennia, harboring 30% of the earth’s soil organic C. However, climate-induced warming and drying patterns across the boreal zone combined with increased wildfire are putting these systems at risk of changing successional trajectories and shifting from a C sink to a source. To investigate the impacts and consequences of drought and wildfire on peatlands in a 14.6 million ha boreal study area, field observations were linked to remote sensing to answer 3 specific research questions: 1) What are the patterns of burned areas across the landscape and which ecotypes may be acting as fire refugia?; 2) How do the landscape patterns of burn severity vary by ecoregion, ecotype, season of burn and year of burn?; and 3) What is the vulnerability of peatlands to wildfire? To address the research questions, three geospatial remote sensing products were created for the study area: 1) pre-fire land cover type (distinguishing bogs, treed and open fens, upland conifer and broadleaf); 2) maps of burn severity to the organic soil (peat) layers based on pre- and post-fire Landsat-8; and 3) fire progression maps from MODIS hot spot data for season of burn. Since peatlands store large amounts of carbon belowground, it is the susceptibility of these soil organic layers to burning that is critical for assessment. The geospatial intersection of the surface organic layer burn severity, fire progression and ecotype maps for 136 2014-15 wildfires allowed us to study the patterns and controls of burn severity in peatlands across the ecological zones and seasons of fire, for assessment of the exposure and sensitivity of the landscape to wildfire. We found that compared with the Taiga plains, peatlands on the Taiga shield were more susceptible to wildfire with minimal fire refugia. This is likely due to the fact that they are typically spatially small on the landscape and often hydrologically isolated by the rocky landscape. Nearly every ecotype burned with greater than 90% burn area on the shield and they burned nearly uniformly with light-moderate severity. In contrast, peatlands on the Taiga plains were less susceptible to wildfire in extreme drought with many more fire refugia. Understanding the tipping point of drought conditions at which the landscape becomes connected, and peatlands are susceptible to wildfire with deeper burning of the organic soil layers is important for understanding the potential future effects of climate change and projected increases in wildfire on peatlands.

Boreal peatlands have provided rich reservoirs of stored carbon for millennia, harboring 30% of the earth’s soil organic C. However, climate-induced warming and drying patterns across the boreal zone combined with increased wildfire are putting these systems at risk of changing successional trajectories and shifting from a C sink to a source. To investigate the impacts and consequences of drought and wildfire on peatlands in a 14.6 million ha boreal study area, field observations were linked to remote sensing to answer 3 specific research questions: 1) What are the patterns of burned areas across the landscape and which ecotypes may be acting as fire refugia?; 2) How do the landscape patterns of burn severity vary by ecoregion, ecotype, season of burn and year of burn?; and 3) What is the vulnerability of peatlands to wildfire? To address the research questions, three geospatial remote sensing products were created for the study area: 1) pre-fire land cover type (distinguishing bogs, treed and open fens, upland conifer and broadleaf); 2) maps of burn severity to the organic soil (peat) layers based on pre- and post-fire Landsat-8; and 3) fire progression maps from MODIS hot spot data for season of burn. Since peatlands store large amounts of carbon belowground, it is the susceptibility of these soil organic layers to burning that is critical for assessment. The geospatial intersection of the surface organic layer burn severity, fire progression and ecotype maps for 136 2014-15 wildfires allowed us to study the patterns and controls of burn severity in peatlands across the ecological zones and seasons of fire, for assessment of the exposure and sensitivity of the landscape to wildfire. We found that compared with the Taiga plains, peatlands on the Taiga shield were more susceptible to wildfire with minimal fire refugia. This is likely due to the fact that they are typically spatially small on the landscape and often hydrologically isolated by the rocky landscape. Nearly every ecotype burned with greater than 90% burn area on the shield and they burned nearly uniformly with light-moderate severity. In contrast, peatlands on the Taiga plains were less susceptible to wildfire in extreme drought with many more fire refugia. Understanding the tipping point of drought conditions at which the landscape becomes connected, and peatlands are susceptible to wildfire with deeper burning of the organic soil layers is important for understanding the potential future effects of climate change and projected increases in wildfire on peatlands.

In the northern high latitudes, the timing and progression of seasonal thawing and freezing in the soil active layer overlying permafrost is a critical indicator of permafrost stability and other subsurface conditions, including soil moisture, water availability to plants, soil microbial activity, and greenhouse gas emissions. Remote sensing from low frequency (L-, P-band) microwave radars and radiometers have proven effective for regional mapping and monitoring of soil freeze/thaw (F/T) dynamics due to greater characteristic soil sensitivity and large contrast in surface and soil dielectric properties between predominantly frozen and non-frozen states at these lower frequencies. However, the freezing point of soil water can extend below the 0.0°C freezing point of pure water and can span a large soil dielectric gradient, which can lead to F/T retrieval error. The dielectric properties of highly organic tundra soils are also variable, which can contribute additional F/T uncertainty. Recent advances in characterizing the dielectric properties of highly organic tundra soils and associated improvements in physics-based radar remote sensing models enable a substantial new capability for detecting and characterizing F/T dynamics within the soil profile using complimentary L-band and P-band PolSAR data. In this work, we first present a physics-based modeling framework integrating radar and subsurface bio/geophysical characteristics informed from detailed soil field measurements to explain L-, P-band radar (PolSAR) backscatter sensitivity to tundra soil properties in northern Alaska. The model simulations are informed from a wealth of P-band and L-band PolSAR data collected during the ABoVE and Pre-ABoVE Airborne Campaigns in Alaska. These data include repeat-sampling of PolSAR retrievals over legacy flightlines spanning seasonal and interannual variability in surface and subsurface characteristics along regional environmental gradients. We show how the physics-based modeling that relies on detailed organic soil dielectric profile characterization can effectively capture the onset of seasonal freezing in autumn. The algorithm is then applied over five flight lines across the Alaskan North Slope. These results provide enhanced delineation of local F/T patterns and surface conditions. Our results enable more accurate modeling of tundra soil dielectric behavior for improved remote sensing of soil F/T dynamics in the Arctic.

In the northern high latitudes, the timing and progression of seasonal thawing and freezing in the soil active layer overlying permafrost is a critical indicator of permafrost stability and other subsurface conditions, including soil moisture, water availability to plants, soil microbial activity, and greenhouse gas emissions. Remote sensing from low frequency (L-, P-band) microwave radars and radiometers have proven effective for regional mapping and monitoring of soil freeze/thaw (F/T) dynamics due to greater characteristic soil sensitivity and large contrast in surface and soil dielectric properties between predominantly frozen and non-frozen states at these lower frequencies. However, the freezing point of soil water can extend below the 0.0°C freezing point of pure water and can span a large soil dielectric gradient, which can lead to F/T retrieval error. The dielectric properties of highly organic tundra soils are also variable, which can contribute additional F/T uncertainty. Recent advances in characterizing the dielectric properties of highly organic tundra soils and associated improvements in physics-based radar remote sensing models enable a substantial new capability for detecting and characterizing F/T dynamics within the soil profile using complimentary L-band and P-band PolSAR data. In this work, we first present a physics-based modeling framework integrating radar and subsurface bio/geophysical characteristics informed from detailed soil field measurements to explain L-, P-band radar (PolSAR) backscatter sensitivity to tundra soil properties in northern Alaska. The model simulations are informed from a wealth of P-band and L-band PolSAR data collected during the ABoVE and Pre-ABoVE Airborne Campaigns in Alaska. These data include repeat-sampling of PolSAR retrievals over legacy flightlines spanning seasonal and interannual variability in surface and subsurface characteristics along regional environmental gradients. We show how the physics-based modeling that relies on detailed organic soil dielectric profile characterization can effectively capture the onset of seasonal freezing in autumn. The algorithm is then applied over five flight lines across the Alaskan North Slope. These results provide enhanced delineation of local F/T patterns and surface conditions. Our results enable more accurate modeling of tundra soil dielectric behavior for improved remote sensing of soil F/T dynamics in the Arctic.

Arctic rivers transport large quantities of freshwater and constituents to coastal areas composed of a diverse array of bays, deltas and lagoons. In northern Alaska half of the Beaufort Sea coastline is skirted by discontinuous barrier islands that enclose numerous shallow lagoon ecosystems. These ecosystems host productive, biologically complex assemblages of biota that are inherently shaped by the extreme seasonal variations in physical-chemical conditions and increasingly challenged by alterations in river discharge and chemistry driven by climate change. Advances in remote sensing observations and a growing collection of measurements from river sampling provide a wealth of key data to integrate with numerical models to yield spatially and temporally extensive estimates of river exports and, in turn, advance knowledge of the dynamics driving coastal water biogeochemistry. We describe the observations, measurements and models being used to quantify freshwater and dissolved organic carbon (DOC) fluxes to the Beaufort Sea coast over the past four decades. Remote sensing and other geospatial data from the NASA ABoVE and other regional studies and monitoring networks are being leveraged for soil freeze-thaw state, surface water extent, and coastal water DOC concentration, with the latter derived from platforms including satellite retrievals, aircraft overflights, and handheld spectroradiometers. River measurements span sampling sites from headwater streams in the Brooks Range to river mainstems near the coast. Simulations with a permafrost-hydrology model for the period 1980-2022 provide estimates of freshwater and DOC exports, capture alterations occurring on land, reveal watershed spatio-temporal patterns in surface runoff and DOC leaching/loading, and point to changes in the chemistry, biology, and circulation of the lagoons and adjacent coastal waters of northern Alaska. Synthesis of observations and modeling is crucial for understanding how shifts in seasonality, permafrost thaw and water cycle intensification in the rapidly warming Arctic are affecting aquatic biogeochemistry and the food webs that inhabit coastal zones.
  [poster]
Arctic rivers transport large quantities of freshwater and constituents to coastal areas composed of a diverse array of bays, deltas and lagoons. In northern Alaska half of the Beaufort Sea coastline is skirted by discontinuous barrier islands that enclose numerous shallow lagoon ecosystems. These ecosystems host productive, biologically complex assemblages of biota that are inherently shaped by the extreme seasonal variations in physical-chemical conditions and increasingly challenged by alterations in river discharge and chemistry driven by climate change. Advances in remote sensing observations and a growing collection of measurements from river sampling provide a wealth of key data to integrate with numerical models to yield spatially and temporally extensive estimates of river exports and, in turn, advance knowledge of the dynamics driving coastal water biogeochemistry. We describe the observations, measurements and models being used to quantify freshwater and dissolved organic carbon (DOC) fluxes to the Beaufort Sea coast over the past four decades. Remote sensing and other geospatial data from the NASA ABoVE and other regional studies and monitoring networks are being leveraged for soil freeze-thaw state, surface water extent, and coastal water DOC concentration, with the latter derived from platforms including satellite retrievals, aircraft overflights, and handheld spectroradiometers. River measurements span sampling sites from headwater streams in the Brooks Range to river mainstems near the coast. Simulations with a permafrost-hydrology model for the period 1980-2022 provide estimates of freshwater and DOC exports, capture alterations occurring on land, reveal watershed spatio-temporal patterns in surface runoff and DOC leaching/loading, and point to changes in the chemistry, biology, and circulation of the lagoons and adjacent coastal waters of northern Alaska. Synthesis of observations and modeling is crucial for understanding how shifts in seasonality, permafrost thaw and water cycle intensification in the rapidly warming Arctic are affecting aquatic biogeochemistry and the food webs that inhabit coastal zones.

Wetlands in high-latitude North America is a significant natural source of atmospheric methane (CH4) emissions. As the second most long-lived greenhouse gas on earth, methane emissions amplify climate change effects, and rising temperatures in high-latitude wetlands may be increasing methane emissions due to the thawing of permafrost and the expansion of wetland areas. Bottom-up models of methane fluxes exhibit large uncertainties in boreal and arctic ecosystems; these models likely underestimate methane fluxes from Arctic tundra and overestimate of methane fluxes from the boreal zone in North America. The aim of the study is to quantify over a decade of methane fluxes in high-latitude North America using multiple sets of atmospheric methane observations, such as data from recent NASA aircraft campaigns as well as observations from regular tower and aircraft sites across the northern US and Canada. To compare the bottom-up methane estimates with existing tower and aircraft data products, we use a geostatistical inverse model (GIM) to estimate methane emissions across these high-latitude North America regions from years 2010 through 2022.
 
Wetlands in high-latitude North America is a significant natural source of atmospheric methane (CH4) emissions. As the second most long-lived greenhouse gas on earth, methane emissions amplify climate change effects, and rising temperatures in high-latitude wetlands may be increasing methane emissions due to the thawing of permafrost and the expansion of wetland areas. Bottom-up models of methane fluxes exhibit large uncertainties in boreal and arctic ecosystems; these models likely underestimate methane fluxes from Arctic tundra and overestimate of methane fluxes from the boreal zone in North America. The aim of the study is to quantify over a decade of methane fluxes in high-latitude North America using multiple sets of atmospheric methane observations, such as data from recent NASA aircraft campaigns as well as observations from regular tower and aircraft sites across the northern US and Canada. To compare the bottom-up methane estimates with existing tower and aircraft data products, we use a geostatistical inverse model (GIM) to estimate methane emissions across these high-latitude North America regions from years 2010 through 2022.

Soil moisture and active layer thickness (ALT) play critical roles in understanding the dynamics of permafrost regions, their response to climate change, and their influence on the broader Earth system. The ABoVE (Arctic-Boreal Vulnerability Experiment) project has been a significant contributor to the scientific community's understanding of these crucial environmental parameters in Alaska and Northwest Territories (NWT), Canada over the last decade. Here we present an updated version of the ABoVE Soil Moisture and Active Layer Thickness (SMALT) database, developed by Schaefer et al. (2021). By incorporating new ABoVE Field Campaign data collected through 2022, as well as additional contributions from affiliated scientists in previous years, we have increased the SMALT database’s archive of field data by more than 15,000 ALT and 13,000 soil moisture measurements.
 
Soil moisture and active layer thickness (ALT) play critical roles in understanding the dynamics of permafrost regions, their response to climate change, and their influence on the broader Earth system. The ABoVE (Arctic-Boreal Vulnerability Experiment) project has been a significant contributor to the scientific community's understanding of these crucial environmental parameters in Alaska and Northwest Territories (NWT), Canada over the last decade. Here we present an updated version of the ABoVE Soil Moisture and Active Layer Thickness (SMALT) database, developed by Schaefer et al. (2021). By incorporating new ABoVE Field Campaign data collected through 2022, as well as additional contributions from affiliated scientists in previous years, we have increased the SMALT database’s archive of field data by more than 15,000 ALT and 13,000 soil moisture measurements.

In the northern circumpolar regions, continued climate change promotes abrupt and pervasive thawing of organic-rich permafrost soils, rendering any associated carbon susceptible to mobilization and degradation. Permafrost-derived dissolved and particulate organic carbon entrained into aquatic systems is highly labile and rapidly mineralized to carbon dioxide (CO2). Consequently, signals of permafrost disturbances are short-lived and easily missed. Equally transient is the methane (CH4) released from anaerobic, water-saturated soils in permafrost regions, wetlands, and lakes and transferred to fluvial systems. Here, we explore the isotopic composition of dissolved CO2 and CH4 to fingerprint the magnitude and extent of permafrost thaw in the Sagavanirktok and Kuparuk River watersheds of North Slope Alaska. We investigate spatial and seasonal variations in the stable (?13C) and radiocarbon (F14C) isotopic signatures of dissolved gases. Isotopic evidence suggests that the CO2 compositions are homogenous along the fluvial continuum and integrate geogenic and biogenic sources across the watershed. From spring to fall, Bayesian mixing modeling predicts a systematic depletion in the stable and F14C isotopic signatures of dissolved CO2 sources. We attribute this pattern to increasing contributions of aged carbon in response to seasonal deepening of the active layer. As percolating groundwater accesses deeper soil horizons, CO2 produced by heterotrophic oxidation of organic matter is leached and transported to streams and rivers. In contrast, we observe no clear relationships between CH4 isotopic compositions and landscape properties. Stream CH4 is likely influenced by local inputs from streambeds, adjacent water saturated soils, and lake outflows. Our findings demonstrate that greenhouse gases dissolved in fluvial systems are sensitive indicators of the release of old carbon from thawing permafrost and may provide early warning signals for acceleration of the permafrost carbon feedback.
 
In the northern circumpolar regions, continued climate change promotes abrupt and pervasive thawing of organic-rich permafrost soils, rendering any associated carbon susceptible to mobilization and degradation. Permafrost-derived dissolved and particulate organic carbon entrained into aquatic systems is highly labile and rapidly mineralized to carbon dioxide (CO2). Consequently, signals of permafrost disturbances are short-lived and easily missed. Equally transient is the methane (CH4) released from anaerobic, water-saturated soils in permafrost regions, wetlands, and lakes and transferred to fluvial systems. Here, we explore the isotopic composition of dissolved CO2 and CH4 to fingerprint the magnitude and extent of permafrost thaw in the Sagavanirktok and Kuparuk River watersheds of North Slope Alaska. We investigate spatial and seasonal variations in the stable (?13C) and radiocarbon (F14C) isotopic signatures of dissolved gases. Isotopic evidence suggests that the CO2 compositions are homogenous along the fluvial continuum and integrate geogenic and biogenic sources across the watershed. From spring to fall, Bayesian mixing modeling predicts a systematic depletion in the stable and F14C isotopic signatures of dissolved CO2 sources. We attribute this pattern to increasing contributions of aged carbon in response to seasonal deepening of the active layer. As percolating groundwater accesses deeper soil horizons, CO2 produced by heterotrophic oxidation of organic matter is leached and transported to streams and rivers. In contrast, we observe no clear relationships between CH4 isotopic compositions and landscape properties. Stream CH4 is likely influenced by local inputs from streambeds, adjacent water saturated soils, and lake outflows. Our findings demonstrate that greenhouse gases dissolved in fluvial systems are sensitive indicators of the release of old carbon from thawing permafrost and may provide early warning signals for acceleration of the permafrost carbon feedback.

Complex non-linear relationships exist between the permafrost thermal state, active layer thickness, and terrestrial carbon cycle dynamics In Arctic and boreal Alaska. The rate, magnitude, and extent of permafrost degradation remain uncertain, with an increasing recognition of the importance of abrupt thaw mechanisms. Similarly, large uncertainties in the rate, magnitude, timing, location, and composition of the permafrost carbon feedback complicate this issue. The challenge of monitoring sub-surface phenomena, such as the soil temperature and soil moisture profiles, with remote sensing technology further complicates the situation. There is an urgent need to understand how and to what extent permafrost degradation is destabilizing the Alaskan carbon balance and to characterize the feedbacks involved. We employ our artificial intelligence (AI)-driven model GeoCryoAI to quantify permafrost thaw dynamics and greenhouse gas emissions in Alaska. GeoCryoAI uses a hybridized multimodal deep learning architecture of stacked convolutionally layered memory-encoded bidirectional recurrent neural networks and 12.4 million parameters to simultaneously ingest and analyze 13.1 million in situ measurements (i.e., CALM, GTNP, ABoVE ReSALT, FLUXNET, NEON), 8.06 billion remote sensing airborne observations (i.e., UAVSAR, AVIRIS-NG), and 7.48 billion process-based modeling outputs (i.e., SIBBORK-TTE, TCFM-Arctic) with disparate spatiotemporal sampling and data densities. This framework introduces ecological memory components and effectively learns subtle spatiotemporal covariate complexities in high-latitude ecosystems by emulating permafrost degradation and carbon flux dynamics across Alaska with high precision and minimal loss (RMSE: 1.007cm, 0.694nmolCH4m-2s-1, 0.213µmolCO2m-2s-1). GeoCryoAI captures abrupt and persistent changes while providing a novel methodology for assimilating contemporaneous information on scales from individual sites to the pan-Arctic. Our approach overcomes traditional model inefficiencies and seamlessly resolves spatiotemporal disparities.
  [poster]
Complex non-linear relationships exist between the permafrost thermal state, active layer thickness, and terrestrial carbon cycle dynamics In Arctic and boreal Alaska. The rate, magnitude, and extent of permafrost degradation remain uncertain, with an increasing recognition of the importance of abrupt thaw mechanisms. Similarly, large uncertainties in the rate, magnitude, timing, location, and composition of the permafrost carbon feedback complicate this issue. The challenge of monitoring sub-surface phenomena, such as the soil temperature and soil moisture profiles, with remote sensing technology further complicates the situation. There is an urgent need to understand how and to what extent permafrost degradation is destabilizing the Alaskan carbon balance and to characterize the feedbacks involved. We employ our artificial intelligence (AI)-driven model GeoCryoAI to quantify permafrost thaw dynamics and greenhouse gas emissions in Alaska. GeoCryoAI uses a hybridized multimodal deep learning architecture of stacked convolutionally layered memory-encoded bidirectional recurrent neural networks and 12.4 million parameters to simultaneously ingest and analyze 13.1 million in situ measurements (i.e., CALM, GTNP, ABoVE ReSALT, FLUXNET, NEON), 8.06 billion remote sensing airborne observations (i.e., UAVSAR, AVIRIS-NG), and 7.48 billion process-based modeling outputs (i.e., SIBBORK-TTE, TCFM-Arctic) with disparate spatiotemporal sampling and data densities. This framework introduces ecological memory components and effectively learns subtle spatiotemporal covariate complexities in high-latitude ecosystems by emulating permafrost degradation and carbon flux dynamics across Alaska with high precision and minimal loss (RMSE: 1.007cm, 0.694nmolCH4m-2s-1, 0.213µmolCO2m-2s-1). GeoCryoAI captures abrupt and persistent changes while providing a novel methodology for assimilating contemporaneous information on scales from individual sites to the pan-Arctic. Our approach overcomes traditional model inefficiencies and seamlessly resolves spatiotemporal disparities.

Background: Accurate time series maps of aboveground biomass (AGB) are crucial for characterizing how the carbon cycle is responding to rapid climate change and increasing disturbance like wildfire. Regionally consistent field-calibrated maps of annual AGB with high spatial resolution are currently unavailable for the entire North American boreal zone, but are necessary to characterize carbon dynamics and compare with estimates from models and atmospheric inversions. Therefore, our primary goal is to generate spatially exhaustive annual AGB maps for the North American Arctic and Boreal regions from 1984 to 2022. Methods: To calibrate and validate the annual AGB maps, we have compiled and analyzed 2,171 ground plots from various forest inventory programs and more than 116,000 km² of airborne laser scanning (ALS) data from multiple programs in Canada and Goddard’s Lidar, Hyperspectral and Thermal Imager (G-LiHT) program in Alaska. We used a two-step modeling approach. Using the area-based approach we first derived localized maps of AGB from the ALS data calibrated with spatially corresponding ground plots. To scale these observations across space and time, we analyzed over 100TB of Landsat Collection 2 Surface Reflectance using the Continuous Change Detection and Classification (CCDC) algorithm to create seasonally corrected synthetic reflectance values and develop spectral-temporal input features. These features and ancillary data (i.e., land cover, topographical variables, and long-term mean climate) were then used in the second modeling step as predictors in a machine learning model calibrated with the ALS-derived AGB predictions to generate preliminary wall-to-wall 30 m resolution maps of annual AGB. We then used ground-based AGB from repeated field surveys and canopy height from repeated ALS acquisitions to analyze whether Landsat-derived AGB can capture the temporal changes in forest attributes. Preliminary results: We observed high accuracy for both the model that used the ground plots to calibrate AGB model with ALS (R-square = 0.82, MAE = 30 Mg/ha), and the model that used the ALS-derived AGB to calibrate AGB model with Landsat data (R-square = 0.90, MAE = 19 Mg/ha). Validation of the preliminary model using independent ground plots (R-square > 0.40, MAE < 50 Mg/ha) showed the best accuracy based on R-square, and the second-best accuracy based on MAE compared to three existing AGB maps partially covering the study domain (R-square: 0.29-0.40, MAE: 47-54 Mg/ha). In all maps analyzed, there was a consistent underestimation of AGB at high AGB levels, with some maps also showing underestimation at relatively low AGB levels. AGB change analysis based on our preliminary map indicated that Landsat-derived AGB can capture the sign of temporal AGB and canopy height change in more than 65% and 80% of cases, respectively. Next steps: To improve AGB maps, we will integrate newly acquired ground plots (>25,000) for model calibration and synthesizing additional airborne data from LVIS and spaceborne LiDAR from ICESat-2. This comprehensive analysis is expected to significantly contribute to the advancement of AGB mapping techniques and enable robust quantification of the impacts of climate change and disturbance on the Arctic-boreal carbon cycle.
  [poster]
Background: Accurate time series maps of aboveground biomass (AGB) are crucial for characterizing how the carbon cycle is responding to rapid climate change and increasing disturbance like wildfire. Regionally consistent field-calibrated maps of annual AGB with high spatial resolution are currently unavailable for the entire North American boreal zone, but are necessary to characterize carbon dynamics and compare with estimates from models and atmospheric inversions. Therefore, our primary goal is to generate spatially exhaustive annual AGB maps for the North American Arctic and Boreal regions from 1984 to 2022. Methods: To calibrate and validate the annual AGB maps, we have compiled and analyzed 2,171 ground plots from various forest inventory programs and more than 116,000 km² of airborne laser scanning (ALS) data from multiple programs in Canada and Goddard’s Lidar, Hyperspectral and Thermal Imager (G-LiHT) program in Alaska. We used a two-step modeling approach. Using the area-based approach we first derived localized maps of AGB from the ALS data calibrated with spatially corresponding ground plots. To scale these observations across space and time, we analyzed over 100TB of Landsat Collection 2 Surface Reflectance using the Continuous Change Detection and Classification (CCDC) algorithm to create seasonally corrected synthetic reflectance values and develop spectral-temporal input features. These features and ancillary data (i.e., land cover, topographical variables, and long-term mean climate) were then used in the second modeling step as predictors in a machine learning model calibrated with the ALS-derived AGB predictions to generate preliminary wall-to-wall 30 m resolution maps of annual AGB. We then used ground-based AGB from repeated field surveys and canopy height from repeated ALS acquisitions to analyze whether Landsat-derived AGB can capture the temporal changes in forest attributes. Preliminary results: We observed high accuracy for both the model that used the ground plots to calibrate AGB model with ALS (R-square = 0.82, MAE = 30 Mg/ha), and the model that used the ALS-derived AGB to calibrate AGB model with Landsat data (R-square = 0.90, MAE = 19 Mg/ha). Validation of the preliminary model using independent ground plots (R-square > 0.40, MAE < 50 Mg/ha) showed the best accuracy based on R-square, and the second-best accuracy based on MAE compared to three existing AGB maps partially covering the study domain (R-square: 0.29-0.40, MAE: 47-54 Mg/ha). In all maps analyzed, there was a consistent underestimation of AGB at high AGB levels, with some maps also showing underestimation at relatively low AGB levels. AGB change analysis based on our preliminary map indicated that Landsat-derived AGB can capture the sign of temporal AGB and canopy height change in more than 65% and 80% of cases, respectively. Next steps: To improve AGB maps, we will integrate newly acquired ground plots (>25,000) for model calibration and synthesizing additional airborne data from LVIS and spaceborne LiDAR from ICESat-2. This comprehensive analysis is expected to significantly contribute to the advancement of AGB mapping techniques and enable robust quantification of the impacts of climate change and disturbance on the Arctic-boreal carbon cycle.

Climate change is influencing weather patterns in the Arctic region, and ultimately leading to changes in the structure and composition of vegetation. As a result, gross primary productivity (GPP) trends have been variable, showing either positive trends (i.e., ‘Arctic greening’) or negative trends (i.e., Arctic browning). Understanding the seasonal patterns of GPP, and the associated weather, will help identify the factors driving this complex change. Here, we analyzed growing season trends of GPP and environmental variables (eg., air temperature, vapor pressure deficit, air pressure, precipitation, net radiation) using multi-year eddy covariance and micrometeorological data. Measurements were made between 2017 and 2022 at Trail Valley Creek; a mineral upland tundra in the southern part of the Southern Arctic ecozone near Inuvik, Northwest Territories, Canada. Results showed that peak growing season occurred between July and August during the study years, with daily averaged GPP ranging from 0.002 to 7.54 µ mol C m^-2 s^-1 in those months. However, we observed an abrupt decrease in GPP right in the middle of the growing season (July) for the years 2017, 2021 and 2022; where daily averaged GPP decreased about 1.5 µ mol C m^-2 s^-1. Based on the statistical relationship between GPP and environmental variables, we found that the abrupt decline in GPP during 2017, 2021 and 2022 was more closely related to incoming and outgoing, short-wave and long-wave radiation. Therefore, for the study period, the factors driving temporal shifts in GPP (e.g., decreasing GPP during the growing season) seemed to be mainly related to the energy available at the site. Moreover, GPP had a greater correlation with long-wave outgoing radiation (r = -0.75), air temperature (r = -0.60) and VPD (r = -0.55) throughout the years of study. Future research will be focused on the potential impacts of vegetation distribution and composition changes on tundra productivity.
  [poster]
Climate change is influencing weather patterns in the Arctic region, and ultimately leading to changes in the structure and composition of vegetation. As a result, gross primary productivity (GPP) trends have been variable, showing either positive trends (i.e., ‘Arctic greening’) or negative trends (i.e., Arctic browning). Understanding the seasonal patterns of GPP, and the associated weather, will help identify the factors driving this complex change. Here, we analyzed growing season trends of GPP and environmental variables (eg., air temperature, vapor pressure deficit, air pressure, precipitation, net radiation) using multi-year eddy covariance and micrometeorological data. Measurements were made between 2017 and 2022 at Trail Valley Creek; a mineral upland tundra in the southern part of the Southern Arctic ecozone near Inuvik, Northwest Territories, Canada. Results showed that peak growing season occurred between July and August during the study years, with daily averaged GPP ranging from 0.002 to 7.54 µ mol C m^-2 s^-1 in those months. However, we observed an abrupt decrease in GPP right in the middle of the growing season (July) for the years 2017, 2021 and 2022; where daily averaged GPP decreased about 1.5 µ mol C m^-2 s^-1. Based on the statistical relationship between GPP and environmental variables, we found that the abrupt decline in GPP during 2017, 2021 and 2022 was more closely related to incoming and outgoing, short-wave and long-wave radiation. Therefore, for the study period, the factors driving temporal shifts in GPP (e.g., decreasing GPP during the growing season) seemed to be mainly related to the energy available at the site. Moreover, GPP had a greater correlation with long-wave outgoing radiation (r = -0.75), air temperature (r = -0.60) and VPD (r = -0.55) throughout the years of study. Future research will be focused on the potential impacts of vegetation distribution and composition changes on tundra productivity.

The circumpolar forest-tundra ecotone is experiencing rapid changes in vegetation composition and structure. Thus, altered surface properties affect how the land surface interacts with the atmosphere. Collectively, these changes modify surface-atmosphere energy exchanges and thus boundary layer characteristics (height, temperature, and humidity). Here, we characterize differences in bulk surface properties and resulting energy balance components using multi-annual eddy covariance and supporting measurements made at a mineral upland tundra (Trail Valley Creek, TVC) and a nearby subarctic woodland (Havikpak Creek, HPC) between 2013 to 2022. Together, the two sites are characteristic of the forest-tundra ecotone of northwestern Canada. A mixed-layer slab model (Chemistry Land-surface Atmosphere Soil Slab, CLASS), in combination with radiosonde observations, was used to shed light on site differences in boundary layer characteristics. Compared to TVC, the sparse tree cover at HPC led to an enhanced ability to transfer heat into the atmosphere (indicated by larger aerodynamic conductance), a higher resistance to evapotranspiration (indicated by lower surface conductance) and a higher influence of atmospheric conditions on turbulent fluxes (indicated by a lower decoupling factor). Sensible heat flux (H) at HPC was generally higher than at TVC. The largest difference in daily mean H was observed in late winter (January to March) and spring (March to May) when albedo of the snow-covered landscape at HPC was reduced by 45 % compared to TVC. During this period, HPC was characterized by more available energy. At HPC, modelled temperature in the mixed-layer was up to 10 °C higher in spring and up to 3 °C higher in summer (June to August) compared to TVC. In accordance with CLASS, radiosonde observations indicated a higher boundary layer and higher air temperature at HPC than at TVC. The presence of trees in the southern part of the forest-tundra ecotone has a warming effect on air temperatures throughout the year and a drying effect in spring. Consequently, vegetation changes in the forest-tundra ecotone can be expected to modify surface climate change pathways and need to be considered when assessing climate change impacts in the region. Next steps include examining decadal carbon flux measurements at TVC and HPC to understand how vegetation changes may affect photosynthetic carbon dioxide uptake and ecosystem respiration in the forest-tundra ecotone of northwestern Canada.

The circumpolar forest-tundra ecotone is experiencing rapid changes in vegetation composition and structure. Thus, altered surface properties affect how the land surface interacts with the atmosphere. Collectively, these changes modify surface-atmosphere energy exchanges and thus boundary layer characteristics (height, temperature, and humidity). Here, we characterize differences in bulk surface properties and resulting energy balance components using multi-annual eddy covariance and supporting measurements made at a mineral upland tundra (Trail Valley Creek, TVC) and a nearby subarctic woodland (Havikpak Creek, HPC) between 2013 to 2022. Together, the two sites are characteristic of the forest-tundra ecotone of northwestern Canada. A mixed-layer slab model (Chemistry Land-surface Atmosphere Soil Slab, CLASS), in combination with radiosonde observations, was used to shed light on site differences in boundary layer characteristics. Compared to TVC, the sparse tree cover at HPC led to an enhanced ability to transfer heat into the atmosphere (indicated by larger aerodynamic conductance), a higher resistance to evapotranspiration (indicated by lower surface conductance) and a higher influence of atmospheric conditions on turbulent fluxes (indicated by a lower decoupling factor). Sensible heat flux (H) at HPC was generally higher than at TVC. The largest difference in daily mean H was observed in late winter (January to March) and spring (March to May) when albedo of the snow-covered landscape at HPC was reduced by 45 % compared to TVC. During this period, HPC was characterized by more available energy. At HPC, modelled temperature in the mixed-layer was up to 10 °C higher in spring and up to 3 °C higher in summer (June to August) compared to TVC. In accordance with CLASS, radiosonde observations indicated a higher boundary layer and higher air temperature at HPC than at TVC. The presence of trees in the southern part of the forest-tundra ecotone has a warming effect on air temperatures throughout the year and a drying effect in spring. Consequently, vegetation changes in the forest-tundra ecotone can be expected to modify surface climate change pathways and need to be considered when assessing climate change impacts in the region. Next steps include examining decadal carbon flux measurements at TVC and HPC to understand how vegetation changes may affect photosynthetic carbon dioxide uptake and ecosystem respiration in the forest-tundra ecotone of northwestern Canada.

Boreal forest changes such as forest productivity and biomass are key factors in calculating national carbon budgets and informing regional management decisions. However, most models predicting changes in boreal forests are long-term (i.e., many decades) and relatively coarse-resolution, whereas most management decisions are fine-scale and short-term. This study aims to develop and apply a 5-20-year high resolution forecast for aboveground biomass in management-relevant regions in interior Alaska, British Columbia, and Saskatchewan. We are building on past work at the site level that forecasts biomass gains and losses using an ensemble of machine learning models and Landsat, vegetation, climate, permafrost, and soil data. This site-level model has transitioned to a 30m gridded product within stakeholder regions using gridded inputs. The gridded inputs come from published datasets and are also derived from Landsat imagery. To better expand biomass predictions at the 30m scale, we conducted experiments to determine which gridded dataset inputs yield the most accurate model results. We are also developing a visualization app with Google Earth Engine to allow for easy access of biomass predictions for resource managers.

Boreal forest changes such as forest productivity and biomass are key factors in calculating national carbon budgets and informing regional management decisions. However, most models predicting changes in boreal forests are long-term (i.e., many decades) and relatively coarse-resolution, whereas most management decisions are fine-scale and short-term. This study aims to develop and apply a 5-20-year high resolution forecast for aboveground biomass in management-relevant regions in interior Alaska, British Columbia, and Saskatchewan. We are building on past work at the site level that forecasts biomass gains and losses using an ensemble of machine learning models and Landsat, vegetation, climate, permafrost, and soil data. This site-level model has transitioned to a 30m gridded product within stakeholder regions using gridded inputs. The gridded inputs come from published datasets and are also derived from Landsat imagery. To better expand biomass predictions at the 30m scale, we conducted experiments to determine which gridded dataset inputs yield the most accurate model results. We are also developing a visualization app with Google Earth Engine to allow for easy access of biomass predictions for resource managers.

Plant biomass is a fundamental ecosystem attribute that is sensitive to the rapid climatic changes occurring in the Arctic, yet much remains unclear about current plant biomass composition and distribution, let alone recent changes. Therefore, we are developing the first moderate resolution (30 m) maps of total and woody (tree + shrub) plant aboveground biomass for the Arctic tundra biome, focusing on the years 2000 and 2020 and including pixel-level uncertainty estimates. Our project involves linking extensive field measurements with the Landsat satellite record using machine learning algorithms. To provide the necessary field measurements, we created The Arctic Plant Aboveground Biomass Synthesis Dataset that includes measurements of plant aboveground biomass by functional type for about 600 field sites across the Arctic. We are then linking these field measurements with seasonal synthetic composites created from decades of surface reflectance measurements by the Landsat satellites. Initial results show we can model plant biomass presence vs. absence with high accuracy and plant biomass quantity with reasonable accuracy. We anticipate these new maps will be available in the coming year and can be used for applications related to wildlife, carbon accounting, disturbances, climate modeling and more.
  [poster]
Plant biomass is a fundamental ecosystem attribute that is sensitive to the rapid climatic changes occurring in the Arctic, yet much remains unclear about current plant biomass composition and distribution, let alone recent changes. Therefore, we are developing the first moderate resolution (30 m) maps of total and woody (tree + shrub) plant aboveground biomass for the Arctic tundra biome, focusing on the years 2000 and 2020 and including pixel-level uncertainty estimates. Our project involves linking extensive field measurements with the Landsat satellite record using machine learning algorithms. To provide the necessary field measurements, we created The Arctic Plant Aboveground Biomass Synthesis Dataset that includes measurements of plant aboveground biomass by functional type for about 600 field sites across the Arctic. We are then linking these field measurements with seasonal synthetic composites created from decades of surface reflectance measurements by the Landsat satellites. Initial results show we can model plant biomass presence vs. absence with high accuracy and plant biomass quantity with reasonable accuracy. We anticipate these new maps will be available in the coming year and can be used for applications related to wildlife, carbon accounting, disturbances, climate modeling and more.

Climate change is likely to disproportionately impact arctic tundra ecosystems. Vegetative communities in these regions both respond to and affect the magnitude of potential ramifications from phenomena such as surface temperature and precipitation changes, permafrost melt, greenhouse gas release, and others. Plant functional traits are tools that allow us to link physiological and anatomical relationships at multiple scales to ecological processes. Traits such as leaf mass per area, leaf dry matter content, equivalent water thickness, and others concerning leaf photosynthetic economy are particularly useful and practical for teasing out relationships of concern. We sampled over 150 plant species to assess whether plant traits related to carbon and water acquisition can help differentiate plant functional types (PFTs) based on composition of plant dominance such as Alder, Birch, Lichen, Riparian shrub and Sedge-moss tundra. We found small differences in morphological traits among PFTs, except for Alder which differentiate from the other PFTs, while there is a less variation in traits on the remaining PFTs. Alder PFT is separated from the other PFTs based on leaf area and leaf water content. These analyses are supported by comparison on the Normalized Different Water Index across PFTs. Future analyses will evaluate the importance of traits related to nutrient acquisition such as foliar carbon and nitrogen.
  [poster]
Climate change is likely to disproportionately impact arctic tundra ecosystems. Vegetative communities in these regions both respond to and affect the magnitude of potential ramifications from phenomena such as surface temperature and precipitation changes, permafrost melt, greenhouse gas release, and others. Plant functional traits are tools that allow us to link physiological and anatomical relationships at multiple scales to ecological processes. Traits such as leaf mass per area, leaf dry matter content, equivalent water thickness, and others concerning leaf photosynthetic economy are particularly useful and practical for teasing out relationships of concern. We sampled over 150 plant species to assess whether plant traits related to carbon and water acquisition can help differentiate plant functional types (PFTs) based on composition of plant dominance such as Alder, Birch, Lichen, Riparian shrub and Sedge-moss tundra. We found small differences in morphological traits among PFTs, except for Alder which differentiate from the other PFTs, while there is a less variation in traits on the remaining PFTs. Alder PFT is separated from the other PFTs based on leaf area and leaf water content. These analyses are supported by comparison on the Normalized Different Water Index across PFTs. Future analyses will evaluate the importance of traits related to nutrient acquisition such as foliar carbon and nitrogen.

The initiation and expansion of Retrogressive Thaw Slumps (RTS) are rapidly increasing across the Arctic. These mass-wasting features are transforming permafrost landscapes, influencing localized patterns of plant species cover and abundance and associated carbon storage. Although imaging spectroscopy is capable of detecting fine-scale patterns of plant species and various plant functional types, due to its limited availability, coverage, and high cost, knowledge of the spatiotemporal patterns of plant community change following RTS disturbance remains limited. Here, we determine the viability of hyperspectral image simulation for detecting the fractional cover (FCover) of dominant plant species and plant functional types across RTS spanning the NASA-ABoVE core domain. We leveraged a data-rich RTS site, extensively sampled during the summer of 2023 in the Noatak National Preserve, which includes coincidental ground (i.e., vegetation surveys, hyperspectral, and LiDAR data), airborne (i.e., multispectral 5cm Red Edge-m and AVIRIS-3), and space-borne (i.e., Sentinel-2 MSI) datasets. We simulated AVIRIS-3 using a Sentinel-2 MSI image using a spectral reconstruction method known as the Universal Pattern Decomposition Method (UPDM), which produced a 10m spatial resolution simulated product. This method requires the Spectral Response Function (SRF) of both sensors to normalize the ground spectra of prominent endmembers. We mapped dominant shrub species and plant functional types across the RTS site using 5cm multispectral imagery and evaluated classification accuracy with ground observations. This high-resolution classification was used to compute the FCover of AVIRIS-3, simulated AVIRIS-3, and Sentinel-2. Preliminary results suggest FCover classification accuracies were substantially improved by hyperspectral image simulation methods, providing a pathway for scaling this analysis within RTS scars across the ABoVE core domain.
  [poster]
The initiation and expansion of Retrogressive Thaw Slumps (RTS) are rapidly increasing across the Arctic. These mass-wasting features are transforming permafrost landscapes, influencing localized patterns of plant species cover and abundance and associated carbon storage. Although imaging spectroscopy is capable of detecting fine-scale patterns of plant species and various plant functional types, due to its limited availability, coverage, and high cost, knowledge of the spatiotemporal patterns of plant community change following RTS disturbance remains limited. Here, we determine the viability of hyperspectral image simulation for detecting the fractional cover (FCover) of dominant plant species and plant functional types across RTS spanning the NASA-ABoVE core domain. We leveraged a data-rich RTS site, extensively sampled during the summer of 2023 in the Noatak National Preserve, which includes coincidental ground (i.e., vegetation surveys, hyperspectral, and LiDAR data), airborne (i.e., multispectral 5cm Red Edge-m and AVIRIS-3), and space-borne (i.e., Sentinel-2 MSI) datasets. We simulated AVIRIS-3 using a Sentinel-2 MSI image using a spectral reconstruction method known as the Universal Pattern Decomposition Method (UPDM), which produced a 10m spatial resolution simulated product. This method requires the Spectral Response Function (SRF) of both sensors to normalize the ground spectra of prominent endmembers. We mapped dominant shrub species and plant functional types across the RTS site using 5cm multispectral imagery and evaluated classification accuracy with ground observations. This high-resolution classification was used to compute the FCover of AVIRIS-3, simulated AVIRIS-3, and Sentinel-2. Preliminary results suggest FCover classification accuracies were substantially improved by hyperspectral image simulation methods, providing a pathway for scaling this analysis within RTS scars across the ABoVE core domain.

Arctic and boreal ecosystems are experiencing the most rapid and pronounced climate change on the planet. Disturbance processes such as fire are pivotal agents that shape how climate change is affecting vegetation dynamics at high latitudes. Recent studies have suggested that intensification of disturbance regimes is driving systematic changes in the distribution and abundance of key plant functional types; for example, some models project that early successional deciduous forests will expand at high latitudes due to more favorable climate conditions and more frequent disturbance events. Despite the rapid rates of change and importance of Arctic and boreal ecosystems in the global climate system, our ability to characterize and quantify trends in high latitude land cover and vegetation composition is incomplete. To address this, we based on the framework of the GLobal Land Cover and Estimation (GLanCE) product, which provides annual land cover data from satellite remote sensing data, to quantify land cover changes at moderate spatial resolution (30 m) from 1984 to 2022 across boreal Canada and Alaska (roughly 1.2 × 10 km ). Preliminary results identify a net decrease in area dominated by evergreen needleleaf forests and shrubs of 0.8 % (-1.3 × 10 km ) and 9.8 % (-2.4 ×10 km ), respectively, for the period 2001-2019. Similarly, our results indicate that deciduous forests and herbaceous land cover increased by 4.5% (8.7 × 10 km ) and 3.9% (2.2 × 10 km ), respectively, for the same period. A number of studies have shown that the early phase of post-disturbance succession determines the trajectory of vegetation regrowth, and specifically, whether or not the post-disturbance vegetation community shifts to a different composition or reverts to the same composition that it had before the disturbance. Moving forward, we plan to leverage the land cover change maps we are creating, in combination with historical information from the Alaska Large Fire Database and National Burned Area Composite database for Canada (NBAC), to analyze land cover changes in the first 5 years following disturbance events, and by extension to improve understanding of how climate change is altering the structure and composition of high-latitude ecosystems.
 
Arctic and boreal ecosystems are experiencing the most rapid and pronounced climate change on the planet. Disturbance processes such as fire are pivotal agents that shape how climate change is affecting vegetation dynamics at high latitudes. Recent studies have suggested that intensification of disturbance regimes is driving systematic changes in the distribution and abundance of key plant functional types; for example, some models project that early successional deciduous forests will expand at high latitudes due to more favorable climate conditions and more frequent disturbance events. Despite the rapid rates of change and importance of Arctic and boreal ecosystems in the global climate system, our ability to characterize and quantify trends in high latitude land cover and vegetation composition is incomplete. To address this, we based on the framework of the GLobal Land Cover and Estimation (GLanCE) product, which provides annual land cover data from satellite remote sensing data, to quantify land cover changes at moderate spatial resolution (30 m) from 1984 to 2022 across boreal Canada and Alaska (roughly 1.2 × 10 km ). Preliminary results identify a net decrease in area dominated by evergreen needleleaf forests and shrubs of 0.8 % (-1.3 × 10 km ) and 9.8 % (-2.4 ×10 km ), respectively, for the period 2001-2019. Similarly, our results indicate that deciduous forests and herbaceous land cover increased by 4.5% (8.7 × 10 km ) and 3.9% (2.2 × 10 km ), respectively, for the same period. A number of studies have shown that the early phase of post-disturbance succession determines the trajectory of vegetation regrowth, and specifically, whether or not the post-disturbance vegetation community shifts to a different composition or reverts to the same composition that it had before the disturbance. Moving forward, we plan to leverage the land cover change maps we are creating, in combination with historical information from the Alaska Large Fire Database and National Burned Area Composite database for Canada (NBAC), to analyze land cover changes in the first 5 years following disturbance events, and by extension to improve understanding of how climate change is altering the structure and composition of high-latitude ecosystems.

Arctic ecosystems have experienced rapid warming in recent decades, which is driving widespread shifts in vegetation cover and composition, along with concomitant changes in spectral greenness. These changes are heterogeneous across space, and vary both in magnitude and direction (e.g., greening and browning) across the Arctic tundra landscape. In order to interpret the ecological meaning and causes of spectral greening and browning, it is important to connect observed trends and changes to quantifiable metrics of vegetation change. Mapping vegetation with remote sensing has historically been difficult in tundra ecosystems, owing to fine-scale spatial heterogeneity and spectral similarities between different vegetation types. To address these challenges, we created a new framework for modeling annual fractional cover of vegetation types in the tundra that leverages land surface phenology (LSP) derived from Harmonized Landsat Sentinel-2 (HLS) data. Here we present initial results from maps of annual 30 m fractional cover for 2016-22 in the Arctic Foothills ecoregion of Alaska. Using LSP data from HLS, we derived a set of synthetic visible, NIR, and SWIR spectra that captures the spectral response of different cover types (trees, shrub, non-woody, barren, and water) at three points in the phenological cycle: 50% green-up, peak greenness, and 50% green-down. To estimate our models and assess our results, we created high-spatial resolution (3 m) maps of land cover using Maxar imagery, which we then up-sampled to 30 m maps. Random Forest regression models were estimated to generate 30 m maps of fractional cover over the entire study area. Our results demonstrate that LSP can be used to accurately map and capture changes in sub-pixel cover over time. Ultimately, this will allow us to tie trends in both spectral greenness and phenology to fractional changes in land cover. Moving forward, we plan to connect these different vegetation responses to climate, disturbance, and topographic drivers by analyzing how vegetation changes vary over climatic and geographic gradients as well as how successional trajectories shift in response to intensifying disturbance regimes.
  [poster] 
Arctic ecosystems have experienced rapid warming in recent decades, which is driving widespread shifts in vegetation cover and composition, along with concomitant changes in spectral greenness. These changes are heterogeneous across space, and vary both in magnitude and direction (e.g., greening and browning) across the Arctic tundra landscape. In order to interpret the ecological meaning and causes of spectral greening and browning, it is important to connect observed trends and changes to quantifiable metrics of vegetation change. Mapping vegetation with remote sensing has historically been difficult in tundra ecosystems, owing to fine-scale spatial heterogeneity and spectral similarities between different vegetation types. To address these challenges, we created a new framework for modeling annual fractional cover of vegetation types in the tundra that leverages land surface phenology (LSP) derived from Harmonized Landsat Sentinel-2 (HLS) data. Here we present initial results from maps of annual 30 m fractional cover for 2016-22 in the Arctic Foothills ecoregion of Alaska. Using LSP data from HLS, we derived a set of synthetic visible, NIR, and SWIR spectra that captures the spectral response of different cover types (trees, shrub, non-woody, barren, and water) at three points in the phenological cycle: 50% green-up, peak greenness, and 50% green-down. To estimate our models and assess our results, we created high-spatial resolution (3 m) maps of land cover using Maxar imagery, which we then up-sampled to 30 m maps. Random Forest regression models were estimated to generate 30 m maps of fractional cover over the entire study area. Our results demonstrate that LSP can be used to accurately map and capture changes in sub-pixel cover over time. Ultimately, this will allow us to tie trends in both spectral greenness and phenology to fractional changes in land cover. Moving forward, we plan to connect these different vegetation responses to climate, disturbance, and topographic drivers by analyzing how vegetation changes vary over climatic and geographic gradients as well as how successional trajectories shift in response to intensifying disturbance regimes.

Disturbances in northern forested environments are only expected to increase with climate change, resulting in greater uncertainty pertaining to how these ecosystems will evolve and change over time. This makes the development of remote sensing techniques that generate accurate, spatially explicit maps of disturbance impacts increasingly important in alleviating these uncertainties and supporting management decisions. Unlike stand-replacing disturbances such as fire and harvest, non-stand replacing disturbances caused by bark beetle infestations have been relatively overlooked in the remote sensing literature. Additionally, remote sensing studies looking at bark beetle-disturbed forests have mainly used optical data to map disturbance extents and severities while lacking in their ability to assess disturbed quantity. The purpose of this research is to assess the utility of using multisensor airborne data from ABoVE campaigns to generate detailed maps of bark beetle-disturbed forests in Kluane, Yukon. Using predictors derived from the AVIRIS-3, LVIS-F, and UAVSAR acquisitions with field training and validation, models of component live-, dead-, and total-tree volumes will be generated. Overall, this work aims to assess the utility of multisensor approaches in the context of modeling and mapping bark-beetle disturbance impacts. The outputs of these airborne models will be used to inform future, upscaled spaceborne models to generate wall-to-wall maps of bark beetle-disturbed regions across the greater Kluane region.

Disturbances in northern forested environments are only expected to increase with climate change, resulting in greater uncertainty pertaining to how these ecosystems will evolve and change over time. This makes the development of remote sensing techniques that generate accurate, spatially explicit maps of disturbance impacts increasingly important in alleviating these uncertainties and supporting management decisions. Unlike stand-replacing disturbances such as fire and harvest, non-stand replacing disturbances caused by bark beetle infestations have been relatively overlooked in the remote sensing literature. Additionally, remote sensing studies looking at bark beetle-disturbed forests have mainly used optical data to map disturbance extents and severities while lacking in their ability to assess disturbed quantity. The purpose of this research is to assess the utility of using multisensor airborne data from ABoVE campaigns to generate detailed maps of bark beetle-disturbed forests in Kluane, Yukon. Using predictors derived from the AVIRIS-3, LVIS-F, and UAVSAR acquisitions with field training and validation, models of component live-, dead-, and total-tree volumes will be generated. Overall, this work aims to assess the utility of multisensor approaches in the context of modeling and mapping bark-beetle disturbance impacts. The outputs of these airborne models will be used to inform future, upscaled spaceborne models to generate wall-to-wall maps of bark beetle-disturbed regions across the greater Kluane region.

This project employed commercial high spatial resolution satellite imagery to evaluate shrub succession and improve its mapping across various Arctic tundra locations in Northern Alaska. We analyzed 242 2-by-2 km areas using machine learning techniques, subdividing each into 100-by-100 m segments. These segments were processed using pansharpened, high-resolution QuickBird and WorldView imagery of cloud-free, summer tundra landscapes as explanatory training data. Our models—Convolutional Neural Networks (CNNs), CNNs with filters, and Spatial-Temporal CNNs (ST-CNNs)—were trained using response data from high-resolution UAS imagery and lidar from the Vegetation Community Map of the Toolik Lake Area (2013-2015). Shrub Roughness Maps, created with a standard deviation kernel, strongly related to aboveground biomass determined from 30 m Landsat imagery. The accuracies of the CNN, CNN-with-Filter, and ST-CNN models were 73%, 73%, and 58% respectively, with corresponding F1 scores of 41%, 53%, and 52%. The ST-CNN showed improved detection of shrubs, albeit at a loss of precision. The CNN-with-filter was designed to increase recall without sacrificing accuracy, processing each pixel that varied across time series to estimate probabilities at a 0.50 threshold. Estimated average shrub covers were 8%, 12%, and 21%, respectively. Additionally, we observed correlations between shrub cover, water depth, and tundra landscape types. The use of large datasets, obtained from high-resolution remote sensing can reveal patterns not observable with previous methodologies. Combined with ML, this is a novel means of approaching a complex, thawing system experiencing important hydrologic and carbon cycle responses to climate change; and a better understanding of this behavior provides insights into the likely future environmental changes as warming accelerates.
  [poster] 
This project employed commercial high spatial resolution satellite imagery to evaluate shrub succession and improve its mapping across various Arctic tundra locations in Northern Alaska. We analyzed 242 2-by-2 km areas using machine learning techniques, subdividing each into 100-by-100 m segments. These segments were processed using pansharpened, high-resolution QuickBird and WorldView imagery of cloud-free, summer tundra landscapes as explanatory training data. Our models—Convolutional Neural Networks (CNNs), CNNs with filters, and Spatial-Temporal CNNs (ST-CNNs)—were trained using response data from high-resolution UAS imagery and lidar from the Vegetation Community Map of the Toolik Lake Area (2013-2015). Shrub Roughness Maps, created with a standard deviation kernel, strongly related to aboveground biomass determined from 30 m Landsat imagery. The accuracies of the CNN, CNN-with-Filter, and ST-CNN models were 73%, 73%, and 58% respectively, with corresponding F1 scores of 41%, 53%, and 52%. The ST-CNN showed improved detection of shrubs, albeit at a loss of precision. The CNN-with-filter was designed to increase recall without sacrificing accuracy, processing each pixel that varied across time series to estimate probabilities at a 0.50 threshold. Estimated average shrub covers were 8%, 12%, and 21%, respectively. Additionally, we observed correlations between shrub cover, water depth, and tundra landscape types. The use of large datasets, obtained from high-resolution remote sensing can reveal patterns not observable with previous methodologies. Combined with ML, this is a novel means of approaching a complex, thawing system experiencing important hydrologic and carbon cycle responses to climate change; and a better understanding of this behavior provides insights into the likely future environmental changes as warming accelerates.

We are extending and refining plant functional type (PFT) time-series maps developed for Arctic and boreal Alaska and Yukon to cover the ABoVE core domain. We are also applying an extensive set of detailed Permanent Sample Plot (PSP) forest measurements to partition above-ground biomass (AGB) map models among key species groups, including black and white spruce.
  [poster]
We are extending and refining plant functional type (PFT) time-series maps developed for Arctic and boreal Alaska and Yukon to cover the ABoVE core domain. We are also applying an extensive set of detailed Permanent Sample Plot (PSP) forest measurements to partition above-ground biomass (AGB) map models among key species groups, including black and white spruce.

The unprecedented rate of warming in the Arctic has driven dramatic changes in the structure and composition of tundra vegetation, with potentially widespread effects on Arctic ecosystem functioning and global climate feedbacks. The expansion of deciduous tall shrubs is of particular concern, as increases in shrub cover, height, and density alter the region’s surface albedo, which may trigger changes to energy balance and modify land-atmosphere interactions. Despite this importance, significant uncertainties remain in our understanding of how fine-scale vegetation patterns drive large scale spatial and temporal variation in tundra surface albedo. In this study, we attempt to resolve these uncertainties by (1) quantifying the spatial and temporal variation in surface albedo and (2) modeling the relative influence of fine-scale vegetation composition, structure, and environmental conditions on albedo at a representative low-tundra site on the Seward Peninsula in western Alaska, USA. To do this, we first derived white-sky albedo across our site using Landsat 8 imagery. We then synthesized this albedo product with a fine-scale map of Arctic plant functional types (PFTs) and a canopy height model to investigate how vegetation composition and structure affects albedo and the timing of seasonal albedo transitions (i.e. from high winter albedo to low summer albedo) across our site. Finally, we used a random forest model to evaluate the influence of vegetation properties on albedo relative to that of topographical and hydrological conditions. Our results indicate that areas dominated by deciduous tall shrubs displayed lower albedo than those dominated by other PFTs, particularly in the winter months (tall shrub albedo was 26.8% lower than non-woody vegetation), and that the presence of tall vegetation advanced the end of snowmelt by an average of 5.53 days. Furthermore, variable importance metrics from the random forest model show that canopy height had the greatest effect on wintertime surface albedo, but environmental conditions generally had a larger impact on albedo variation than PFT composition. We highlight that these findings are a product of tall shrubs’ ability to protrude above winter snowpacks and lower the surface reflectance of snowy tundra, resulting in greater wintertime energy absorption, earlier completion of snowmelt, and earlier spring leaf-out. Similar studies across the circumpolar region are needed to determine the large-scale impacts of vegetation change on energy feedbacks, climate, and biogeochemical cycling, and ultimately to enable the incorporation of more sophisticated vegetation-albedo dynamics into models of global climate change.
 
The unprecedented rate of warming in the Arctic has driven dramatic changes in the structure and composition of tundra vegetation, with potentially widespread effects on Arctic ecosystem functioning and global climate feedbacks. The expansion of deciduous tall shrubs is of particular concern, as increases in shrub cover, height, and density alter the region’s surface albedo, which may trigger changes to energy balance and modify land-atmosphere interactions. Despite this importance, significant uncertainties remain in our understanding of how fine-scale vegetation patterns drive large scale spatial and temporal variation in tundra surface albedo. In this study, we attempt to resolve these uncertainties by (1) quantifying the spatial and temporal variation in surface albedo and (2) modeling the relative influence of fine-scale vegetation composition, structure, and environmental conditions on albedo at a representative low-tundra site on the Seward Peninsula in western Alaska, USA. To do this, we first derived white-sky albedo across our site using Landsat 8 imagery. We then synthesized this albedo product with a fine-scale map of Arctic plant functional types (PFTs) and a canopy height model to investigate how vegetation composition and structure affects albedo and the timing of seasonal albedo transitions (i.e. from high winter albedo to low summer albedo) across our site. Finally, we used a random forest model to evaluate the influence of vegetation properties on albedo relative to that of topographical and hydrological conditions. Our results indicate that areas dominated by deciduous tall shrubs displayed lower albedo than those dominated by other PFTs, particularly in the winter months (tall shrub albedo was 26.8% lower than non-woody vegetation), and that the presence of tall vegetation advanced the end of snowmelt by an average of 5.53 days. Furthermore, variable importance metrics from the random forest model show that canopy height had the greatest effect on wintertime surface albedo, but environmental conditions generally had a larger impact on albedo variation than PFT composition. We highlight that these findings are a product of tall shrubs’ ability to protrude above winter snowpacks and lower the surface reflectance of snowy tundra, resulting in greater wintertime energy absorption, earlier completion of snowmelt, and earlier spring leaf-out. Similar studies across the circumpolar region are needed to determine the large-scale impacts of vegetation change on energy feedbacks, climate, and biogeochemical cycling, and ultimately to enable the incorporation of more sophisticated vegetation-albedo dynamics into models of global climate change.

The Arctic is warming more than two times faster than the rest of the Earth, driving major changes in vegetation seasonality that regulates annual carbon, water, and energy fluxes. However, complex interactions among climate, permafrost, and vegetation create significant landscape-scale variability in tundra vegetation distribution and phenology that remains poorly studied. Failing to account for this variability has led to significant discrepancies in observed phenological patterns across warming experiments, long-term ecological research, and satellite records, undermining our ability to understand plant responses to climate change in the Arctic. To address this problem, we quantified plant phenology across three low-Arctic tundra landscapes on the Seward Peninsula, Alaska using a combination of in-situ phenocam observations and high-resolution PlanetScope CubeSat data. In this study, we investigated the patterns and drivers of phenological diversity across the landscape by (1) quantifying phenological diversity across dominant plant function types (PFTs) and (2) modeling the interrelation between plant phenology and fine-scale landscape characteristics, including topography, snowmelt, and vegetation. We show that both spring and fall phenology varied dramatically across Arctic PFTs, accounting for about 16 – 54% and 43 – 52%, respectively, of the landscape-scale variation in start of spring [SOS] and start of fall [SOF]. In particular, deciduous tall shrubs (e.g., alder and willow) had a later SOS (~7 days behind the mean of other PFTs), but completed leaf expansion much faster than other tundra PFTs within only about 2 weeks. We modeled the landscape-scale variation in SOS and SOF using random forest (RF) which showed that plant phenology can be well captured by a combination of vegetation composition, topographic characteristics, and snowmelt timing (variance explained: 60 - 81% for SOS and 65 - 88% for SOF). Importantly, snowmelt timing played a key role in determining SOS, which was previously ignored in most phenology models. Collectively, our study highlights a strong control of fine-scale vegetation composition, snow seasonality, and landscape characteristics on tundra phenological heterogeneity. Improved understanding of these controls, as well as phenological diversity across dominant PFTs, is critically needed to better represent tundra phenology in process models.

The Arctic is warming more than two times faster than the rest of the Earth, driving major changes in vegetation seasonality that regulates annual carbon, water, and energy fluxes. However, complex interactions among climate, permafrost, and vegetation create significant landscape-scale variability in tundra vegetation distribution and phenology that remains poorly studied. Failing to account for this variability has led to significant discrepancies in observed phenological patterns across warming experiments, long-term ecological research, and satellite records, undermining our ability to understand plant responses to climate change in the Arctic. To address this problem, we quantified plant phenology across three low-Arctic tundra landscapes on the Seward Peninsula, Alaska using a combination of in-situ phenocam observations and high-resolution PlanetScope CubeSat data. In this study, we investigated the patterns and drivers of phenological diversity across the landscape by (1) quantifying phenological diversity across dominant plant function types (PFTs) and (2) modeling the interrelation between plant phenology and fine-scale landscape characteristics, including topography, snowmelt, and vegetation. We show that both spring and fall phenology varied dramatically across Arctic PFTs, accounting for about 16 – 54% and 43 – 52%, respectively, of the landscape-scale variation in start of spring [SOS] and start of fall [SOF]. In particular, deciduous tall shrubs (e.g., alder and willow) had a later SOS (~7 days behind the mean of other PFTs), but completed leaf expansion much faster than other tundra PFTs within only about 2 weeks. We modeled the landscape-scale variation in SOS and SOF using random forest (RF) which showed that plant phenology can be well captured by a combination of vegetation composition, topographic characteristics, and snowmelt timing (variance explained: 60 - 81% for SOS and 65 - 88% for SOF). Importantly, snowmelt timing played a key role in determining SOS, which was previously ignored in most phenology models. Collectively, our study highlights a strong control of fine-scale vegetation composition, snow seasonality, and landscape characteristics on tundra phenological heterogeneity. Improved understanding of these controls, as well as phenological diversity across dominant PFTs, is critically needed to better represent tundra phenology in process models.

Optical sensing provides approaches for making repeated measurements physiological condition of boreal forest to describe detailed temporal variation with minimal interference to the sensitive plants. Through the 2023 growing season automated systems were deployed to measure diurnal and seasonal variation in leaf and canopy reflectance and solar induced fluorescence (SIF) of the boreal forest at the Caribou Creek (BONA), AK NEON flux tower. The Monitoring-PAM (MoniPAM) made active measurements of leaf level chlorophyll fluorescence (ChlF) providing relative electron transport rates and photochemical yields of photosystem II (PSII), while the FLoX (Dual FLuorescence boX) sensors provided proximal measurements of canopy Solar Induced Fluorescence (SIF) and spectral reflectance at approximately 3-minute intervals. The collection captures seasonal and diurnal variation in reflectance, ChlF, SIF in the O2A (SIFA) and O2B (SIFB) atmospheric bands. We examine the springtime period of vegetation development from snowmelt to early summer. Relationships are examined between diurnal and seasonal observations of canopy SIF and spectral vegetation indices (SVI) from FLoX and leaf level chlorophyll fluorescence (F), F yield (Fy) to PSII and electron transport rate (ETR = Fy/PAR) from MoniPAM (YII). Our findings demonstrate that canopy optical measurements including SIF and SVI metrics are able to capture the dynamics in photosynthesis at both leaf and canopy levels, and these relationships inform the use of satellite observations.
  [poster]
Optical sensing provides approaches for making repeated measurements physiological condition of boreal forest to describe detailed temporal variation with minimal interference to the sensitive plants. Through the 2023 growing season automated systems were deployed to measure diurnal and seasonal variation in leaf and canopy reflectance and solar induced fluorescence (SIF) of the boreal forest at the Caribou Creek (BONA), AK NEON flux tower. The Monitoring-PAM (MoniPAM) made active measurements of leaf level chlorophyll fluorescence (ChlF) providing relative electron transport rates and photochemical yields of photosystem II (PSII), while the FLoX (Dual FLuorescence boX) sensors provided proximal measurements of canopy Solar Induced Fluorescence (SIF) and spectral reflectance at approximately 3-minute intervals. The collection captures seasonal and diurnal variation in reflectance, ChlF, SIF in the O2A (SIFA) and O2B (SIFB) atmospheric bands. We examine the springtime period of vegetation development from snowmelt to early summer. Relationships are examined between diurnal and seasonal observations of canopy SIF and spectral vegetation indices (SVI) from FLoX and leaf level chlorophyll fluorescence (F), F yield (Fy) to PSII and electron transport rate (ETR = Fy/PAR) from MoniPAM (YII). Our findings demonstrate that canopy optical measurements including SIF and SVI metrics are able to capture the dynamics in photosynthesis at both leaf and canopy levels, and these relationships inform the use of satellite observations.

More than half of the Earth’s soil organic carbon is sequestered in northern high latitude (NHL) regions, where the most prevalent subsurface feature is permafrost. As Arctic amplification of global warming continues to unfold, better understanding of permafrost stability and the potential mobilization and degradation of soil organic matter due to gradual or abrupt permafrost thaw is essential for understanding the carbon-climate feedback. To this end, physics-based computational models linking low frequency (L-, P-band) microwave radar remote sensing with subsurface soil and associated bio/geophysical characteristics are key for effective mapping and monitoring of soil organic carbon within the permafrost active layer, along with other climate-sensitive features over vast and complex NHL landscapes. The NASA ABoVE airborne campaign (AAC) captured complex environmental gradients in Arctic ecosystem dynamics using complementary remote sensing data, including low frequency Radars. Repeated AirMOSS P-band polarimetric radar backscatter measurements were collected along regional transects spanning northern Alaska tundra during the early thaw season (June), maximum thaw season (August), and early frozen season (October). This work, for the first time, presents radar-derived high-resolution maps of the soil organic matter profile, soil moisture profile, and thaw depth for five AAC flight lines spanning multiple transects across the Alaska North Slope. These maps are created through a physics-based computational algorithm that generates superior performance by accounting for both time-variant and time-invariant behavior in subsurface features. A detailed study of static and dynamic tundra soil characteristics was conducted and informed by extensive in-situ measurements. Subsequently, coupled hydrological-electromagnetic modeling and measurements were performed to capture a realistic characterization of active layer organic soil dielectric properties, which is essential in relating field-scale processes to radar backscatter measurements. Our results are among the finest-resolution soil organic carbon maps available for the Arctic tundra region and are being used to improve understanding of the distribution and stability of permafrost and soil organic carbon.

More than half of the Earth’s soil organic carbon is sequestered in northern high latitude (NHL) regions, where the most prevalent subsurface feature is permafrost. As Arctic amplification of global warming continues to unfold, better understanding of permafrost stability and the potential mobilization and degradation of soil organic matter due to gradual or abrupt permafrost thaw is essential for understanding the carbon-climate feedback. To this end, physics-based computational models linking low frequency (L-, P-band) microwave radar remote sensing with subsurface soil and associated bio/geophysical characteristics are key for effective mapping and monitoring of soil organic carbon within the permafrost active layer, along with other climate-sensitive features over vast and complex NHL landscapes. The NASA ABoVE airborne campaign (AAC) captured complex environmental gradients in Arctic ecosystem dynamics using complementary remote sensing data, including low frequency Radars. Repeated AirMOSS P-band polarimetric radar backscatter measurements were collected along regional transects spanning northern Alaska tundra during the early thaw season (June), maximum thaw season (August), and early frozen season (October). This work, for the first time, presents radar-derived high-resolution maps of the soil organic matter profile, soil moisture profile, and thaw depth for five AAC flight lines spanning multiple transects across the Alaska North Slope. These maps are created through a physics-based computational algorithm that generates superior performance by accounting for both time-variant and time-invariant behavior in subsurface features. A detailed study of static and dynamic tundra soil characteristics was conducted and informed by extensive in-situ measurements. Subsequently, coupled hydrological-electromagnetic modeling and measurements were performed to capture a realistic characterization of active layer organic soil dielectric properties, which is essential in relating field-scale processes to radar backscatter measurements. Our results are among the finest-resolution soil organic carbon maps available for the Arctic tundra region and are being used to improve understanding of the distribution and stability of permafrost and soil organic carbon.

We present EcoPerceiver, a novel multimodal neural architecture for predicting terrestrial carbon fluxes at the hourly timescale. Modern carbon flux upscaling is typically performed with tabular machine learning techniques; while these techniques are considered optimal in the tabular problem domain, they treat their data as homogenous and Markovian in nature. Recent advances in multimodal learning now allow us to fuse geospatial data with tabular meteorological data without sacrificing textural information such as by averaging pixel values. EcoPerceiver predicts terrestrial carbon fluxes similarly to a time series model with a sliding window, conditioning on local meteorological observations at each time step. We train this model on globally available eddy covariance station data from FLUXNET, Ameriflux, and ICOS publications. It demonstrates state-of-the-art performance predicting net ecosystem exchange (NEE) on holdout data with an average Nash-Sutcliffe Efficiency (NSE) of 0.73 across all sites. The model is extremely robust to data availability and supports intermittent insertion of geospatial data such as MODIS. Present experiments are assessing the ability of EcoPerceiver to model carbon fluxes specifically in the challenging regions of boreal forests and arctic tundra.
 
We present EcoPerceiver, a novel multimodal neural architecture for predicting terrestrial carbon fluxes at the hourly timescale. Modern carbon flux upscaling is typically performed with tabular machine learning techniques; while these techniques are considered optimal in the tabular problem domain, they treat their data as homogenous and Markovian in nature. Recent advances in multimodal learning now allow us to fuse geospatial data with tabular meteorological data without sacrificing textural information such as by averaging pixel values. EcoPerceiver predicts terrestrial carbon fluxes similarly to a time series model with a sliding window, conditioning on local meteorological observations at each time step. We train this model on globally available eddy covariance station data from FLUXNET, Ameriflux, and ICOS publications. It demonstrates state-of-the-art performance predicting net ecosystem exchange (NEE) on holdout data with an average Nash-Sutcliffe Efficiency (NSE) of 0.73 across all sites. The model is extremely robust to data availability and supports intermittent insertion of geospatial data such as MODIS. Present experiments are assessing the ability of EcoPerceiver to model carbon fluxes specifically in the challenging regions of boreal forests and arctic tundra.

High-latitude terrestrial ecosystems are experiencing rapid environmental change as the warming rate in these areas is up to four times faster than the global mean. One notable feature that provides evidence of the changes in terrestrial ecosystems is the increase in the seasonal cycle amplitude of atmospheric CO2. The increasing trend in the amplitude suggests higher levels of terrestrial metabolism, with increasing uptake during summer and higher emissions during spring and fall shoulder seasons. Moreover, it is not only the magnitude of the carbon fluxes that are changing but also their timing, which can change the shape of the seasonal cycle. Here we use monthly atmospheric CO2 measurements from the National Oceanic Atmospheric Administration’s (NOAA) Global Monitoring Laboratory as well as Scripps CO2 program to quantify changes in the amplitude and shape of annual CO2 cycle across multiple stations. We focused on 11 stations located above 50°N, each with at least 20 years (240 months) of data. We removed the long-term trend and mean annual cycle using quadratic polynomials and four seasonal harmonics. We then analyzed the residuals to examine changes in the shape of the annual cycle and interannual anomalies. residuals. Our longest record at Utqia?vik, Alaska (at the Point Barrow station), has an amplitude enhancement of approximately 26% from 1971 to 2022 years) with a trend of 0.52 ppm per year. Other stations above 60°N have an amplitude trend between 0.52 ppm per year and 0.87 ppm per year. A change in the shape accompanies the increasing amplitude. The strongest negative trends (deepening of the mid-summer trough) occur during July and August, just before or during the CO2 minimum. We also find that stations above 60°N have strong temporal correlations amongst each other and even stronger seasonal correlations in spring and summer, indicating that anomalies propagate through stations. Our preliminary findings provide evidence for a forward shift in the annual cycle, indicating greater proportional uptake near the beginning of the growing season. We also find that monthly CO2 anomalies detected at one station are prevalent at other stations north of 60°N. The covariance of CO2 anomalies across stations is most robust in spring (March-May) and during the middle and end of the growing season (July and August). We further test the hypothesis that spring and summer anomalies are partially driven by the changes in temperature influencing the magnitude and sign of SIF and CO2 anomalies. To test this hypothesis, we analyzed temperature from ERA5 solar-induced fluorescence from the Orbiting Carbon Observatory-2.
 
High-latitude terrestrial ecosystems are experiencing rapid environmental change as the warming rate in these areas is up to four times faster than the global mean. One notable feature that provides evidence of the changes in terrestrial ecosystems is the increase in the seasonal cycle amplitude of atmospheric CO2. The increasing trend in the amplitude suggests higher levels of terrestrial metabolism, with increasing uptake during summer and higher emissions during spring and fall shoulder seasons. Moreover, it is not only the magnitude of the carbon fluxes that are changing but also their timing, which can change the shape of the seasonal cycle. Here we use monthly atmospheric CO2 measurements from the National Oceanic Atmospheric Administration’s (NOAA) Global Monitoring Laboratory as well as Scripps CO2 program to quantify changes in the amplitude and shape of annual CO2 cycle across multiple stations. We focused on 11 stations located above 50°N, each with at least 20 years (240 months) of data. We removed the long-term trend and mean annual cycle using quadratic polynomials and four seasonal harmonics. We then analyzed the residuals to examine changes in the shape of the annual cycle and interannual anomalies. residuals. Our longest record at Utqia?vik, Alaska (at the Point Barrow station), has an amplitude enhancement of approximately 26% from 1971 to 2022 years) with a trend of 0.52 ppm per year. Other stations above 60°N have an amplitude trend between 0.52 ppm per year and 0.87 ppm per year. A change in the shape accompanies the increasing amplitude. The strongest negative trends (deepening of the mid-summer trough) occur during July and August, just before or during the CO2 minimum. We also find that stations above 60°N have strong temporal correlations amongst each other and even stronger seasonal correlations in spring and summer, indicating that anomalies propagate through stations. Our preliminary findings provide evidence for a forward shift in the annual cycle, indicating greater proportional uptake near the beginning of the growing season. We also find that monthly CO2 anomalies detected at one station are prevalent at other stations north of 60°N. The covariance of CO2 anomalies across stations is most robust in spring (March-May) and during the middle and end of the growing season (July and August). We further test the hypothesis that spring and summer anomalies are partially driven by the changes in temperature influencing the magnitude and sign of SIF and CO2 anomalies. To test this hypothesis, we analyzed temperature from ERA5 solar-induced fluorescence from the Orbiting Carbon Observatory-2.

The Arctic has experienced the most rapid change in climate than anywhere else on earth. The wetlands therein are key stores of organic carbon (C), and play a major role in the greenhouse gas balance of high-latitude ecosystems. However, the impact of recent climate change and rising CO2 on Arctic land-atmosphere greenhouse gas exchanges remains poorly understood. To quantitatively assess the carbon feedback to recent climate trends and rising CO2, we use a terrestrial biosphere model (DALEC) and a data-model integration approach (CARbon Data Model fraMework, CARDAMOM) to produce a data-constrained analysis of environmental controls of carbon exchange and its sensitivity to climate and atmospheric CO2 trends at six-teen ABoVE domain high-latitude eddy covariance sites. The observation-optimized model accurately represents seasonal and inter-annual variability of CH4 and CO2 fluxes. Based on observation-constrained model processes, we perturb meteorological forcings to quantify the sensitivity of CH4 and CO2 fluxes to trends in precipitation, radiation, VPD, temperature, and atmospheric CO2 concentration. Our study demonstrates the need for joint observational constraints on CO2 and CH4 biogeochemical processes – and their associated climatic sensitivities – for understanding and predicting the role of climate drivers and CO2 concentration on the sign and magnitude of high-latitude ecosystem C-climate feedback in the past and coming decades.
  [poster]
The Arctic has experienced the most rapid change in climate than anywhere else on earth. The wetlands therein are key stores of organic carbon (C), and play a major role in the greenhouse gas balance of high-latitude ecosystems. However, the impact of recent climate change and rising CO2 on Arctic land-atmosphere greenhouse gas exchanges remains poorly understood. To quantitatively assess the carbon feedback to recent climate trends and rising CO2, we use a terrestrial biosphere model (DALEC) and a data-model integration approach (CARbon Data Model fraMework, CARDAMOM) to produce a data-constrained analysis of environmental controls of carbon exchange and its sensitivity to climate and atmospheric CO2 trends at six-teen ABoVE domain high-latitude eddy covariance sites. The observation-optimized model accurately represents seasonal and inter-annual variability of CH4 and CO2 fluxes. Based on observation-constrained model processes, we perturb meteorological forcings to quantify the sensitivity of CH4 and CO2 fluxes to trends in precipitation, radiation, VPD, temperature, and atmospheric CO2 concentration. Our study demonstrates the need for joint observational constraints on CO2 and CH4 biogeochemical processes – and their associated climatic sensitivities – for understanding and predicting the role of climate drivers and CO2 concentration on the sign and magnitude of high-latitude ecosystem C-climate feedback in the past and coming decades.

Wildfire is a major disturbance agent in Arctic boreal and tundra ecosystems, causing profound ecological changes with the potential to significantly impact ecosystem services critical to Arctic communities. The accelerated warming trend and changes in precipitation regime in this region have resulted in an increase in the frequency and extent of wildfires. Although there is ample evidence for ongoing modification for individual ecological components and processes in the Arctic boreal and tundra biomes, comparatively little is understood about the complex interactions and feedback loops that connect them. Additionally, there is a lack of integrated observational/modeling frameworks with high temporal and spatial resolution that can model and predict these processes that cause wildfires. Due to the inaccessibility of the Arctic boreal and tundra ecosystems, cloud-to-ground lightning (CGL) is the primary ignition source of wildfires. Thus, regular monitoring and modeling at regional to global scales is crucial to comprehend how this disturbance affects Earth system processes and human dynamics. In this study we have developed a machine learning framework capable of simulating the probability of cloud-to-ground lightning across the Alaska boreal and tundra biomes by integrating weather variables from the Weather Research & Forecasting (WRF) model and historically recorded lightning strikes from the Alaska Lightning Detection Network (ALDN). Our workflow includes an end-to-end GPU-accelerated machine learning pipeline for (1) the parallel extraction and calculation of weather features from the WRF model output, (2) training and inference of cloud-to-ground lightning probability using single and multi-GPU resources, (3) and the inclusion of learning feedback loops to continuously monitor forecasting performance. A variety of temporal windows were used to apply the trained model and evaluate how well it performed under various scenarios throughout the full study area. In addition, explainable artificial intelligence (XAI) techniques were used to further analyze the climate drivers influencing CGL conditions and their difference within boreal and tundra Alaska. Our findings demonstrate the potential of forecasting cloud-to-ground lightning probability across Alaska’s boreal forest and tundra ecosystems using WRF-simulated weather variables at 24-hour, 48-hour, and 10 days into the future.
 
Wildfire is a major disturbance agent in Arctic boreal and tundra ecosystems, causing profound ecological changes with the potential to significantly impact ecosystem services critical to Arctic communities. The accelerated warming trend and changes in precipitation regime in this region have resulted in an increase in the frequency and extent of wildfires. Although there is ample evidence for ongoing modification for individual ecological components and processes in the Arctic boreal and tundra biomes, comparatively little is understood about the complex interactions and feedback loops that connect them. Additionally, there is a lack of integrated observational/modeling frameworks with high temporal and spatial resolution that can model and predict these processes that cause wildfires. Due to the inaccessibility of the Arctic boreal and tundra ecosystems, cloud-to-ground lightning (CGL) is the primary ignition source of wildfires. Thus, regular monitoring and modeling at regional to global scales is crucial to comprehend how this disturbance affects Earth system processes and human dynamics. In this study we have developed a machine learning framework capable of simulating the probability of cloud-to-ground lightning across the Alaska boreal and tundra biomes by integrating weather variables from the Weather Research & Forecasting (WRF) model and historically recorded lightning strikes from the Alaska Lightning Detection Network (ALDN). Our workflow includes an end-to-end GPU-accelerated machine learning pipeline for (1) the parallel extraction and calculation of weather features from the WRF model output, (2) training and inference of cloud-to-ground lightning probability using single and multi-GPU resources, (3) and the inclusion of learning feedback loops to continuously monitor forecasting performance. A variety of temporal windows were used to apply the trained model and evaluate how well it performed under various scenarios throughout the full study area. In addition, explainable artificial intelligence (XAI) techniques were used to further analyze the climate drivers influencing CGL conditions and their difference within boreal and tundra Alaska. Our findings demonstrate the potential of forecasting cloud-to-ground lightning probability across Alaska’s boreal forest and tundra ecosystems using WRF-simulated weather variables at 24-hour, 48-hour, and 10 days into the future.

Climate-driven disturbances in Interior Alaska are driving an ongoing shift in forest composition and productivity. Changes to forest resources may threaten the economic security and safety of nearby communities. While adaptive forest management techniques show promise in many managed forest systems, their effectiveness should be reexamined for Interior Alaska, where the climate is warming more rapidly and unique ecological processes play an important role in future forest trajectories. Here, we implement dynamic management alternatives in an ecological process model to quantify the impacts of climate change and adaptive forest management on future availability of natural resources and ecosystem services. We used the model University of Virginia Forest Model Enhanced (UVAFME), an individual-based gap model for the boreal forest, to predict climate impacts on the Tanana Valley State Forest (TVSF) of Alaska. Then, we interviewed a group of local stakeholders to elicit probable adaptive management actions in the face of climate change and ran our model under different climate change and management scenarios. We found that climate change drives substantial decreases in carbon sequestration and high-value timber and shifts in wildlife habitat across TVSF. At the landscape scale, there is such light management under any scenario that the conservative management techniques currently under consideration are unable to moderate climate impacts. However, we also found that specific fuel treatments can reduce the risk of intense fire at the site level for several years, and may be targeted around critical infrastructure. Our yearslong collaboration with TVSF stakeholders, in which we shared several rounds of preliminary model results and received experiential feedback, improved the applicability of our ecological model and the actionable knowledge of the people currently writing the next TVSF management plan. This work underscores the value of working with local communities for improving the scientific basis of future policies and indicates that climate change mitigation - as opposed to adaptation - is critical for protecting communities in sparsely managed, climate-vulnerable forests.
  [poster] 
Climate-driven disturbances in Interior Alaska are driving an ongoing shift in forest composition and productivity. Changes to forest resources may threaten the economic security and safety of nearby communities. While adaptive forest management techniques show promise in many managed forest systems, their effectiveness should be reexamined for Interior Alaska, where the climate is warming more rapidly and unique ecological processes play an important role in future forest trajectories. Here, we implement dynamic management alternatives in an ecological process model to quantify the impacts of climate change and adaptive forest management on future availability of natural resources and ecosystem services. We used the model University of Virginia Forest Model Enhanced (UVAFME), an individual-based gap model for the boreal forest, to predict climate impacts on the Tanana Valley State Forest (TVSF) of Alaska. Then, we interviewed a group of local stakeholders to elicit probable adaptive management actions in the face of climate change and ran our model under different climate change and management scenarios. We found that climate change drives substantial decreases in carbon sequestration and high-value timber and shifts in wildlife habitat across TVSF. At the landscape scale, there is such light management under any scenario that the conservative management techniques currently under consideration are unable to moderate climate impacts. However, we also found that specific fuel treatments can reduce the risk of intense fire at the site level for several years, and may be targeted around critical infrastructure. Our yearslong collaboration with TVSF stakeholders, in which we shared several rounds of preliminary model results and received experiential feedback, improved the applicability of our ecological model and the actionable knowledge of the people currently writing the next TVSF management plan. This work underscores the value of working with local communities for improving the scientific basis of future policies and indicates that climate change mitigation - as opposed to adaptation - is critical for protecting communities in sparsely managed, climate-vulnerable forests.

The impact of aerosols remains a large source of uncertainty in climate projections. Here we present a methodology for utilizing vertically resolved NASA Langley High Spectral Resolution Lidar (HSRL) retrievals of aerosol types (i.e. smoke, urban, dust, marine, etc.; Burton et al., 2012) and abundance to estimate the aerosol direct radiative effect (DRE) and forcing (DRF) by employing aerosol type-specific optical properties. Aerosol type-specific values for the single scattering albedo (SSA) and asymmetry parameter (g) are derived using GEOS-Chem and the Creating Aerosol Types from Chemistry (CATCH; Dawson et al., 2017) algorithm. CATCH assigns HSRL-analogous aerosol types based on GEOS-Chem simulated aerosol microphysical properties across the North American domain for the SABOR and DISCOVER-AQ California campaign periods. In the absence of widespread distributions of HSRL retrievals and assigned aerosol types, we utilize CATCH assigned aerosol types along with a modified version of GEOS-Chem to model the state of the atmosphere to act as a testbed for assessing the uncertainty introduced by using type-specific optical properties in DRE and DRF calculations. Sensitivity studies are performed to establish the likely uncertainty due to 1) using specific values for the optical properties for all aerosols assigned the same type, and 2) due to the possibility of misclassification of the aerosol type. Across the North American domain average GEOS-Chem/CATCH derived DRE and DRF were estimated to be -1.98 ± 0.21 W/m2 and -0.77 ± 0.13 W/m2 respectively, during the DISCOVER-AQ California campaign (January-February 2013). During the SABOR campaign (July-August 2014) DRE was estimated to be –4.20 ± 0.42 W/m2 and DRF was estimated to be -1.41 ± 0.21 W/m2. These results indicate that using type-specific optical properties in radiative transfer calculations may be a promising step forward in the future where there is a NASA HSRL-like observing system in orbit.
  [poster] 
The impact of aerosols remains a large source of uncertainty in climate projections. Here we present a methodology for utilizing vertically resolved NASA Langley High Spectral Resolution Lidar (HSRL) retrievals of aerosol types (i.e. smoke, urban, dust, marine, etc.; Burton et al., 2012) and abundance to estimate the aerosol direct radiative effect (DRE) and forcing (DRF) by employing aerosol type-specific optical properties. Aerosol type-specific values for the single scattering albedo (SSA) and asymmetry parameter (g) are derived using GEOS-Chem and the Creating Aerosol Types from Chemistry (CATCH; Dawson et al., 2017) algorithm. CATCH assigns HSRL-analogous aerosol types based on GEOS-Chem simulated aerosol microphysical properties across the North American domain for the SABOR and DISCOVER-AQ California campaign periods. In the absence of widespread distributions of HSRL retrievals and assigned aerosol types, we utilize CATCH assigned aerosol types along with a modified version of GEOS-Chem to model the state of the atmosphere to act as a testbed for assessing the uncertainty introduced by using type-specific optical properties in DRE and DRF calculations. Sensitivity studies are performed to establish the likely uncertainty due to 1) using specific values for the optical properties for all aerosols assigned the same type, and 2) due to the possibility of misclassification of the aerosol type. Across the North American domain average GEOS-Chem/CATCH derived DRE and DRF were estimated to be -1.98 ± 0.21 W/m2 and -0.77 ± 0.13 W/m2 respectively, during the DISCOVER-AQ California campaign (January-February 2013). During the SABOR campaign (July-August 2014) DRE was estimated to be –4.20 ± 0.42 W/m2 and DRF was estimated to be -1.41 ± 0.21 W/m2. These results indicate that using type-specific optical properties in radiative transfer calculations may be a promising step forward in the future where there is a NASA HSRL-like observing system in orbit.

Permafrost soils hold a large fraction of the global terrestrial carbon store. The longterm stability of carbon stocks in permafrost is a function of climatic, ecosystem, and soil thermal processes and interactions. Climate-driven permafrost degradation is accelerated by increasing wildfire frequency and severity. Wildfire can consume both vegetation and the surface organic soil layer. The loss of organic layer thickness from consumption reduces insulation of underlying permafrost leading to thawing and subsequent increases in soil C loss from decomposition. Most ecosystem models lack the structural and functional representations necessary to reproduce these complex interactions, making regional assessment of the effect of wildfire on permafrost carbon dynamics uncertain. As part of the Permafrost Pathways program we are improving the modeling of the ecological impacts of wildfire in the Arctic Boreal Zone. We are enhancing the wildfire model in DVM-DOS-TEM, a process-based ecosystem model with a detailed representation of permafrost, to predict wildfire behavior and effects based on meteorological and ecological drivers. These enhancements will improve the capacity of the model to represent the spatiotemporal patterns of wildfire effect on permafrost and ecosystem carbon dynamics. We will use this to project the impact of climate-driven change in wildfire on the permafrost carbon climate feedback through the end of the century.
  [poster] 
Permafrost soils hold a large fraction of the global terrestrial carbon store. The longterm stability of carbon stocks in permafrost is a function of climatic, ecosystem, and soil thermal processes and interactions. Climate-driven permafrost degradation is accelerated by increasing wildfire frequency and severity. Wildfire can consume both vegetation and the surface organic soil layer. The loss of organic layer thickness from consumption reduces insulation of underlying permafrost leading to thawing and subsequent increases in soil C loss from decomposition. Most ecosystem models lack the structural and functional representations necessary to reproduce these complex interactions, making regional assessment of the effect of wildfire on permafrost carbon dynamics uncertain. As part of the Permafrost Pathways program we are improving the modeling of the ecological impacts of wildfire in the Arctic Boreal Zone. We are enhancing the wildfire model in DVM-DOS-TEM, a process-based ecosystem model with a detailed representation of permafrost, to predict wildfire behavior and effects based on meteorological and ecological drivers. These enhancements will improve the capacity of the model to represent the spatiotemporal patterns of wildfire effect on permafrost and ecosystem carbon dynamics. We will use this to project the impact of climate-driven change in wildfire on the permafrost carbon climate feedback through the end of the century.

The arctic is warming at an alarming rate causing substantial environmental and ecosystem changes not experienced at lower latitudes. Changs in temperature and precipitation are causing reduced snow cover, thawing permafrost, increased vegetation biomass, changes in phenology, and changes to species compositions and abundances. These changes are often nonlinear with respect to warming and effect the Arctic carbon balance between the atmosphere, biosphere, and soil. We use the Ecosystem Demography model (ED2) to predict these changes out to 2100 under two different warming scenarios and report their impact on the overall carbon balance of the ABoVE domain. Overall, we find that increases in vegetation biomass are being more than offset by soil carbon loss, with notable regional patterns, such that over the ABoVE domain roughly 11 tC/ha is lost to the atmosphere between 2018 and 2100.

The arctic is warming at an alarming rate causing substantial environmental and ecosystem changes not experienced at lower latitudes. Changs in temperature and precipitation are causing reduced snow cover, thawing permafrost, increased vegetation biomass, changes in phenology, and changes to species compositions and abundances. These changes are often nonlinear with respect to warming and effect the Arctic carbon balance between the atmosphere, biosphere, and soil. We use the Ecosystem Demography model (ED2) to predict these changes out to 2100 under two different warming scenarios and report their impact on the overall carbon balance of the ABoVE domain. Overall, we find that increases in vegetation biomass are being more than offset by soil carbon loss, with notable regional patterns, such that over the ABoVE domain roughly 11 tC/ha is lost to the atmosphere between 2018 and 2100.

Characterizing the three-dimensional microheterogeneity of bio/geophysical processes within the permafrost active layer has been an outstanding challenge of the NASA ABoVE. Multiple airborne acquisitions from UAVSAR L-band and AirMOSS P-band data provide the potential for indirect measurement of these processes at local scale (30m x 30m resolution) and closer to the level of detailed in-situ field measurements and biophysical models. However, a lack of sufficient understanding of sub-grid heterogeneity hinders the development of more effective linkages between the detailed ground truth measurements and remote sensing retrievals. The ABoVE Phase III provides a unique opportunity for limited field campaigns to fill remaining knowledge gaps, including improved characterizations of surface and soil microheterogeneity and upscaling properties for soil organic matter, soil moisture, active layer depth and freeze/thaw dynamics. Multi-scale airborne L-band SAR acquisitions will be acquired over Alaskan tundra at varying spatial scales ranging from a few meters to 30m footprints, and collocated with detailed ground measurements of tundra soil properties. These data will be used to inform physics-based radar remote sensing models to derive enhanced soil retrievals and characterizations of spatial heterogeneity and underlying environmental controls. A “Forward-Backward” L-band acquisition along with pixel microheterogeneity characterization will clarify local heterogeneity in surface and soil conditions affecting bio/geophysical processes and radar backscatter. These results will help to inform the design and optimization of SAR retrievals in tundra for next generation missions, including NISAR.

Characterizing the three-dimensional microheterogeneity of bio/geophysical processes within the permafrost active layer has been an outstanding challenge of the NASA ABoVE. Multiple airborne acquisitions from UAVSAR L-band and AirMOSS P-band data provide the potential for indirect measurement of these processes at local scale (30m x 30m resolution) and closer to the level of detailed in-situ field measurements and biophysical models. However, a lack of sufficient understanding of sub-grid heterogeneity hinders the development of more effective linkages between the detailed ground truth measurements and remote sensing retrievals. The ABoVE Phase III provides a unique opportunity for limited field campaigns to fill remaining knowledge gaps, including improved characterizations of surface and soil microheterogeneity and upscaling properties for soil organic matter, soil moisture, active layer depth and freeze/thaw dynamics. Multi-scale airborne L-band SAR acquisitions will be acquired over Alaskan tundra at varying spatial scales ranging from a few meters to 30m footprints, and collocated with detailed ground measurements of tundra soil properties. These data will be used to inform physics-based radar remote sensing models to derive enhanced soil retrievals and characterizations of spatial heterogeneity and underlying environmental controls. A “Forward-Backward” L-band acquisition along with pixel microheterogeneity characterization will clarify local heterogeneity in surface and soil conditions affecting bio/geophysical processes and radar backscatter. These results will help to inform the design and optimization of SAR retrievals in tundra for next generation missions, including NISAR.

The Arctic Methane and Permafrost Challenge (AMPAC), a collaborative initiative between NASA and ESA, aims to address critical scientific questions concerning methane emissions, permafrost degradation, and ecosystem change in the rapidly warming Arctic region. This transatlantic initiative integrates advanced Earth Observation (EO) products, satellite technologies, and international scientific expertise to enhance understanding and predict changes in Arctic methane dynamics and permafrost resilience. Through AMPAC, efforts such as AMPAC-Net and MethaneCAMP foster scientific exchanges and community engagement, focusing on reconciling bottom-up and top-down estimates of methane stocktakes, enhancing data accessibility, and improving land cover representations critical for methane source identification. AMPAC also emphasizes the development and validation of novel remote sensing methods, including the use of multispectral imaging and synthetic aperture radar from UAVSAR, AVIRIS, and the Copernicus Sentinel missions, to improve the spatial resolution and thematic accuracy of Arctic land cover data, which is vital for understanding methane emission heterogeneity. The initiative underscores the importance of new and upcoming satellite missions (e.g., NISAR, PREFIRE, CRISTAL) in achieving these goals, alongside promoting the integration of already existing datasets into comprehensive catalogs for broader scientific use. By addressing the scientific and technological challenges in estimating both current and future methane fluxes from the Arctic, AMPAC aims to improve the prediction of landscape and biodiversity changes due to permafrost thaw and to leverage groundbreaking science and technology to advance our understanding of Arctic systems under a changing climate. The collaborative framework of AMPAC supports a wide array of activities including workshops, training, and outreach to enhance the capabilities of the global scientific community in responding to these environmental challenges.
  [poster]
The Arctic Methane and Permafrost Challenge (AMPAC), a collaborative initiative between NASA and ESA, aims to address critical scientific questions concerning methane emissions, permafrost degradation, and ecosystem change in the rapidly warming Arctic region. This transatlantic initiative integrates advanced Earth Observation (EO) products, satellite technologies, and international scientific expertise to enhance understanding and predict changes in Arctic methane dynamics and permafrost resilience. Through AMPAC, efforts such as AMPAC-Net and MethaneCAMP foster scientific exchanges and community engagement, focusing on reconciling bottom-up and top-down estimates of methane stocktakes, enhancing data accessibility, and improving land cover representations critical for methane source identification. AMPAC also emphasizes the development and validation of novel remote sensing methods, including the use of multispectral imaging and synthetic aperture radar from UAVSAR, AVIRIS, and the Copernicus Sentinel missions, to improve the spatial resolution and thematic accuracy of Arctic land cover data, which is vital for understanding methane emission heterogeneity. The initiative underscores the importance of new and upcoming satellite missions (e.g., NISAR, PREFIRE, CRISTAL) in achieving these goals, alongside promoting the integration of already existing datasets into comprehensive catalogs for broader scientific use. By addressing the scientific and technological challenges in estimating both current and future methane fluxes from the Arctic, AMPAC aims to improve the prediction of landscape and biodiversity changes due to permafrost thaw and to leverage groundbreaking science and technology to advance our understanding of Arctic systems under a changing climate. The collaborative framework of AMPAC supports a wide array of activities including workshops, training, and outreach to enhance the capabilities of the global scientific community in responding to these environmental challenges.

The Oak Ridge National Laboratory Distributed Active Archive Center (ORNL DAAC) is the primary repository for NASA’s Terrestrial Ecology program's Arctic Boreal Vulnerability Experiment (ABoVE) project. Its mission is to assemble, publish and maintain a comprehensive archive of field observations, airborne derived data products, and models in the fields of terrestrial biogeochemistry and ecological dynamics that facilitate research, education, and decision-making in support of NASA's earth science. This poster highlights the ABoVE data published through the ORNL DAAC including the collaborative efforts of scientists, data managers, and data curation experts in advancing our understanding of the Arctic and Boreal ecosystems. Additionally, the poster offers insights into the curation challenges encountered during the publication process, including issues related to data standardization, quality assurance, and metadata documentation. By reflecting on these challenges and analyzing some of the publication metrics, we aim to promote a culture of transparency and innovation in data archival and dissemination practices.
  [poster]
The Oak Ridge National Laboratory Distributed Active Archive Center (ORNL DAAC) is the primary repository for NASA’s Terrestrial Ecology program's Arctic Boreal Vulnerability Experiment (ABoVE) project. Its mission is to assemble, publish and maintain a comprehensive archive of field observations, airborne derived data products, and models in the fields of terrestrial biogeochemistry and ecological dynamics that facilitate research, education, and decision-making in support of NASA's earth science. This poster highlights the ABoVE data published through the ORNL DAAC including the collaborative efforts of scientists, data managers, and data curation experts in advancing our understanding of the Arctic and Boreal ecosystems. Additionally, the poster offers insights into the curation challenges encountered during the publication process, including issues related to data standardization, quality assurance, and metadata documentation. By reflecting on these challenges and analyzing some of the publication metrics, we aim to promote a culture of transparency and innovation in data archival and dissemination practices.

The objective of the NASA Commercial Smallsat Data Acquisition (CSDA) Program is identification, evaluation, and acquisition of commercial remote sensing imagery and data that support the NASA Earth science research and applications community. CSDA currently provides data to the community from a number of commercial sources, is evaluating new sources of data for sustained acquisition, and periodically issues new Requests For Proposals from additional sources. In addition to providing data to the community, CSDA provides data quality assessments and data utility evaluation reports to the community for the sources investigated. CSDA is interested in supporting the NASA ABoVE Program toward meeting its goals, and to understand program successes and challenges in working with commercial remote sensing data.

The objective of the NASA Commercial Smallsat Data Acquisition (CSDA) Program is identification, evaluation, and acquisition of commercial remote sensing imagery and data that support the NASA Earth science research and applications community. CSDA currently provides data to the community from a number of commercial sources, is evaluating new sources of data for sustained acquisition, and periodically issues new Requests For Proposals from additional sources. In addition to providing data to the community, CSDA provides data quality assessments and data utility evaluation reports to the community for the sources investigated. CSDA is interested in supporting the NASA ABoVE Program toward meeting its goals, and to understand program successes and challenges in working with commercial remote sensing data.

Northern ice- and lake-rich permafrost regions are experiencing changing climate conditions, which are affecting landscape conditions (e.g., enhanced hydrological connectivity, lake drainage, thermokarst, increased shrub vegetation). Old Crow Flats (OCF), Yukon, is a 14,500-km2 headwater catchment of the Yukon River Basin where there are concerns regarding environmental change. This wetland complex, which encompasses over 8000 thermokarst lakes and ponds, is recognized internationally for its ecological and culture importance. Our long-term water isotope-based monitoring program at OCF revealed higher proportions of annual rainfall up to 2019 resulting in increasingly positive lake water balances and enhanced hydrological connectivity between lakes and rivers. This was followed by drier summers during 2020-22, when isotope results indicated that lake-river connectivity was reduced. Remote sensing analysis has also revealed annual landscape-scale water balance fluctuations including notable lake drainage events during the last decade, which coincide with downstream biogeochemistry. Many uncertainties remain about the effect of hydrological changes on biogeochemical conditions of lakes and rivers. Integration of water chemistry parameters and water-isotope tracers provides an effective means for tracking the spatial variability and magnitude of lake to river hydrologic connectivity and the associated processes. Multivariate analyses reveal strong spatial patterns in water chemistry according to catchment characteristics. Notably, we found higher nutrient concentrations associated with relatively large sub catchments with lower proportions of surface water coverage. Carbon isotopes (?13C-DOC and ?13C-DIC) indicate these larger catchments provide a greater biogenic source of carbon compared to smaller catchments where lakes likely sequester more atmospheric carbon. Our findings provide important baseline insight required for tracking continued responses of lakes and rivers to changing climate as well as associated impacts on carbon and other nutrient cycling. The dynamic thermokarst processes of northern landscapes with abundant surface water have complex biogeochemical processes that require improved understanding to effectively anticipate future responses to climate change and associated feedback. Overall, with greater hydrologic connectivity (i.e., channel development and deeper active layer) we expect greater carbon mobility from catchments to downstream lakes and rivers.
  [poster]
Northern ice- and lake-rich permafrost regions are experiencing changing climate conditions, which are affecting landscape conditions (e.g., enhanced hydrological connectivity, lake drainage, thermokarst, increased shrub vegetation). Old Crow Flats (OCF), Yukon, is a 14,500-km2 headwater catchment of the Yukon River Basin where there are concerns regarding environmental change. This wetland complex, which encompasses over 8000 thermokarst lakes and ponds, is recognized internationally for its ecological and culture importance. Our long-term water isotope-based monitoring program at OCF revealed higher proportions of annual rainfall up to 2019 resulting in increasingly positive lake water balances and enhanced hydrological connectivity between lakes and rivers. This was followed by drier summers during 2020-22, when isotope results indicated that lake-river connectivity was reduced. Remote sensing analysis has also revealed annual landscape-scale water balance fluctuations including notable lake drainage events during the last decade, which coincide with downstream biogeochemistry. Many uncertainties remain about the effect of hydrological changes on biogeochemical conditions of lakes and rivers. Integration of water chemistry parameters and water-isotope tracers provides an effective means for tracking the spatial variability and magnitude of lake to river hydrologic connectivity and the associated processes. Multivariate analyses reveal strong spatial patterns in water chemistry according to catchment characteristics. Notably, we found higher nutrient concentrations associated with relatively large sub catchments with lower proportions of surface water coverage. Carbon isotopes (?13C-DOC and ?13C-DIC) indicate these larger catchments provide a greater biogenic source of carbon compared to smaller catchments where lakes likely sequester more atmospheric carbon. Our findings provide important baseline insight required for tracking continued responses of lakes and rivers to changing climate as well as associated impacts on carbon and other nutrient cycling. The dynamic thermokarst processes of northern landscapes with abundant surface water have complex biogeochemical processes that require improved understanding to effectively anticipate future responses to climate change and associated feedback. Overall, with greater hydrologic connectivity (i.e., channel development and deeper active layer) we expect greater carbon mobility from catchments to downstream lakes and rivers.

Wetlands provide many ecosystem services, including waterfowl habitat, carbon sequestration, and water quality. Northern wetlands are threatened from both land use expansion and climate change disruptions, prompting the need for informed management strategies. We aim to assist communities in managing wetlands and set land stewardship policies by providing access to remote sensing-based information on short-term and long-term wetland dynamics. The research utilizes methods for mapping wetland type and seasonal inundation developed by our research team in Phases 1 and 2 of ABoVE. In the current phase, these techniques are being further developed and expanded to make them accessible and useful for land management and policy decisions. Investments in data collection, algorithm development, and product creation by ABoVE are being used for capacity building for decision-making in the ABoVE region, as envisioned in the ABoVE Concise Experiment Plan. This poster will be focused on the collaborations that our team has developed during this project, with the primary focus being detailing an On-The-Land camp that our team is building in collaboration with Ecology North and Northern Youth Leadership. This camp will provide 8-10 high school students with the opportunity to learn about wetlands, wildfowl, and remote sensing from the micro-scale to the macro scale - beginning with macroinvertebrate sampling and ending with satellite imagery analysis. Other collaborative efforts with communities and organizations will also be detailed, such as a Knowledge Sharing Seminar that provided information about our research to local governments, communities, and organizations, and our work to inform policies to designating and managing Indigenous Protected Conserved Area (IPCA) lands and waters. Outreach activities are essential to the success of our project, as this provides an avenue for our team to disseminate the information generated in Phases 1 and 2 of ABoVE to the communities that would benefit from it the most. Indigenous and local communities have a vested interest in the protection and long-term monitoring of the ecosystems in which they live, and it is our responsibility as outside researchers to provide information in the way that is most useful to Indigenous communities. Supporting and working in partnership with the Indigenous communities who inhabit ABoVE’s study area is one of the most impactful and enduring actions we can take to benefit the ecosystems that we are focused on protecting.
 
Wetlands provide many ecosystem services, including waterfowl habitat, carbon sequestration, and water quality. Northern wetlands are threatened from both land use expansion and climate change disruptions, prompting the need for informed management strategies. We aim to assist communities in managing wetlands and set land stewardship policies by providing access to remote sensing-based information on short-term and long-term wetland dynamics. The research utilizes methods for mapping wetland type and seasonal inundation developed by our research team in Phases 1 and 2 of ABoVE. In the current phase, these techniques are being further developed and expanded to make them accessible and useful for land management and policy decisions. Investments in data collection, algorithm development, and product creation by ABoVE are being used for capacity building for decision-making in the ABoVE region, as envisioned in the ABoVE Concise Experiment Plan. This poster will be focused on the collaborations that our team has developed during this project, with the primary focus being detailing an On-The-Land camp that our team is building in collaboration with Ecology North and Northern Youth Leadership. This camp will provide 8-10 high school students with the opportunity to learn about wetlands, wildfowl, and remote sensing from the micro-scale to the macro scale - beginning with macroinvertebrate sampling and ending with satellite imagery analysis. Other collaborative efforts with communities and organizations will also be detailed, such as a Knowledge Sharing Seminar that provided information about our research to local governments, communities, and organizations, and our work to inform policies to designating and managing Indigenous Protected Conserved Area (IPCA) lands and waters. Outreach activities are essential to the success of our project, as this provides an avenue for our team to disseminate the information generated in Phases 1 and 2 of ABoVE to the communities that would benefit from it the most. Indigenous and local communities have a vested interest in the protection and long-term monitoring of the ecosystems in which they live, and it is our responsibility as outside researchers to provide information in the way that is most useful to Indigenous communities. Supporting and working in partnership with the Indigenous communities who inhabit ABoVE’s study area is one of the most impactful and enduring actions we can take to benefit the ecosystems that we are focused on protecting.

Surface water in the northern high latitudes plays pivotal roles in hydrological, biological, and biogeochemical processes. The rapid warming of the Arctic and sub-Arctic regions induces profound changes in the hydrological dynamics of these ecosystems, with cascading impacts on local communities, wildlife, and greenhouse gas (GHG) feedbacks. Remote sensing has been instrumental in mapping surface water and monitoring changes over multi-year timescales. However, the analysis of hydrological change is often limited by spatial resolutions that are too coarse to capture small ponds (<0.01 km2), which are considerable sources of methane emissions and crucial habitats for vertebrates. Moreover, the temporal revisit rates of many satellite platforms, coupled with persistent cloud cover, result in a limited number of usable observations in any given year. In the Yukon-Kuskokwim Delta (YKD), Alaska, the dynamics of surface water, both inter and intra-annually, remain poorly understood despite more than 30% of the region being comprised of open water bodies. To address this gap, we developed comprehensive maps delineating open water and partial inundation in the YKD. Leveraging high-resolution optical satellite imagery from Planet Labs, we employed a convolutional neural network to detect and track changes in lake and pond open water extents at a 3 m resolution. Additionally, we generated maps of partial inundation using an unsupervised classification based on the normalized-difference water index. Changes in open water and partial inundation surface area were tracked over the course of four years (2020 - 2023) at a sub-weekly interval. This effort builds towards an improved understanding of high-latitude water cycles. Future work can leverage the products for assessments of community risk and waterfowl habitat, and inform GHG emissions upscaling and modeling.
  [poster] 
Surface water in the northern high latitudes plays pivotal roles in hydrological, biological, and biogeochemical processes. The rapid warming of the Arctic and sub-Arctic regions induces profound changes in the hydrological dynamics of these ecosystems, with cascading impacts on local communities, wildlife, and greenhouse gas (GHG) feedbacks. Remote sensing has been instrumental in mapping surface water and monitoring changes over multi-year timescales. However, the analysis of hydrological change is often limited by spatial resolutions that are too coarse to capture small ponds (<0.01 km2), which are considerable sources of methane emissions and crucial habitats for vertebrates. Moreover, the temporal revisit rates of many satellite platforms, coupled with persistent cloud cover, result in a limited number of usable observations in any given year. In the Yukon-Kuskokwim Delta (YKD), Alaska, the dynamics of surface water, both inter and intra-annually, remain poorly understood despite more than 30% of the region being comprised of open water bodies. To address this gap, we developed comprehensive maps delineating open water and partial inundation in the YKD. Leveraging high-resolution optical satellite imagery from Planet Labs, we employed a convolutional neural network to detect and track changes in lake and pond open water extents at a 3 m resolution. Additionally, we generated maps of partial inundation using an unsupervised classification based on the normalized-difference water index. Changes in open water and partial inundation surface area were tracked over the course of four years (2020 - 2023) at a sub-weekly interval. This effort builds towards an improved understanding of high-latitude water cycles. Future work can leverage the products for assessments of community risk and waterfowl habitat, and inform GHG emissions upscaling and modeling.

Wetlands are the largest natural source of methane (CH4) emissions globally. Northern wetlands (>45° N), accounting for 42% of global wetland area, are increasingly vulnerable to carbon loss, especially as CH4 emissions may accelerate under intensified high-latitude warming. However, the magnitude and spatial patterns of high-latitude CH4 emissions remain relatively uncertain. Here we present estimates of daily CH4 fluxes obtained using a new machine learning-based wetland CH4 upscaling framework (WetCH4) that applies the most complete database of eddy covariance (EC) observations available to date, and satellite remote sensing informed observations of environmental conditions at 10-km resolution. The most important predictor variables included near-surface soil temperatures (top 40 cm), vegetation reflectance, and soil moisture. Our results, modeled from 138 site-years across 26 sites, had relatively strong predictive skill with a mean R2 of 0.46 and 0.62 and a mean absolute error (MAE) of 23 nmol m-2 s-1 and 21 nmol m-2 s-1 for daily and monthly fluxes, respectively. Based on the model results, we estimated an annual average of 20.8 ±2.1 Tg CH4 yr-1 for the northern wetland region (2016-2022) and total budgets ranged from 13.7 - 44.1 Tg CH4 yr-1, depending on wetland map extents. Although 86% of the estimated CH4 budget occurred during the May-October period, a considerable amount (1.4 ±0.2 Tg CH4) occurred during winter. Regionally, the West Siberian wetlands accounted for a majority (51%) of the interannual variation in domain CH4 emissions. Significant issues with data coverage remain, with only 23% of the sites observing year-round and most of the data from 11 wetland sites in Alaska and 10 bog/fen sites in Canada and Fennoscandia, and in general, Western Siberian Lowlands are underrepresented by EC CH4 sites. Our results provide high spatiotemporal information on the wetland emissions in the high-latitude carbon cycle and possible responses to climate change. Continued, all-season tower observations and improved soil moisture products are needed for future improvement of CH4 upscaling. The dataset can be found at https://doi.org/10.5281/zenodo.10802154 (Ying et al., 2024).
  [poster]
Wetlands are the largest natural source of methane (CH4) emissions globally. Northern wetlands (>45° N), accounting for 42% of global wetland area, are increasingly vulnerable to carbon loss, especially as CH4 emissions may accelerate under intensified high-latitude warming. However, the magnitude and spatial patterns of high-latitude CH4 emissions remain relatively uncertain. Here we present estimates of daily CH4 fluxes obtained using a new machine learning-based wetland CH4 upscaling framework (WetCH4) that applies the most complete database of eddy covariance (EC) observations available to date, and satellite remote sensing informed observations of environmental conditions at 10-km resolution. The most important predictor variables included near-surface soil temperatures (top 40 cm), vegetation reflectance, and soil moisture. Our results, modeled from 138 site-years across 26 sites, had relatively strong predictive skill with a mean R2 of 0.46 and 0.62 and a mean absolute error (MAE) of 23 nmol m-2 s-1 and 21 nmol m-2 s-1 for daily and monthly fluxes, respectively. Based on the model results, we estimated an annual average of 20.8 ±2.1 Tg CH4 yr-1 for the northern wetland region (2016-2022) and total budgets ranged from 13.7 - 44.1 Tg CH4 yr-1, depending on wetland map extents. Although 86% of the estimated CH4 budget occurred during the May-October period, a considerable amount (1.4 ±0.2 Tg CH4) occurred during winter. Regionally, the West Siberian wetlands accounted for a majority (51%) of the interannual variation in domain CH4 emissions. Significant issues with data coverage remain, with only 23% of the sites observing year-round and most of the data from 11 wetland sites in Alaska and 10 bog/fen sites in Canada and Fennoscandia, and in general, Western Siberian Lowlands are underrepresented by EC CH4 sites. Our results provide high spatiotemporal information on the wetland emissions in the high-latitude carbon cycle and possible responses to climate change. Continued, all-season tower observations and improved soil moisture products are needed for future improvement of CH4 upscaling. The dataset can be found at https://doi.org/10.5281/zenodo.10802154 (Ying et al., 2024).

Soil moisture is an important indicator of fire danger in the boreal. Understanding soil moisture conditions using remote sensing is difficult due to the soil surface typically being obscured by vegetation cover. Synthetic Aperture Radar (SAR) sensors, which emit microwave radiation and measure the backscattered intensity signals, are well suited for assessing soil moisture conditions as the lower frequency radiation is capable of penetrating vegetation biomass and interacting with the soil. Accurate in situ soil moisture data, collected close to the SAR collections, is needed to parameterize and validate SAR derived soil moisture retrieval algorithms. These data are difficult to collect as they require specialized equipment (TDR probes) and the ability to travel to the regions of interest, which can be time consuming and costly. Therefore, robust datasets with corresponding SAR imagery and field data are lacking. To overcome this obstacle, we tested the use of Functional Data Analysis (FDA) with time-series SAR data from a variety of sensors to provide information on soil moisture patterns in various regions of the ABoVE domain where in situ soil moisture data is available. By employing FDA, data within the SAR time series is treated as a series of functions rather than discrete data points. This approach allows both for the estimation of soil properties, such as drainage characteristics, and for the estimation of moisture anywhere along the approximated functional curves derived from the input data. For this project, we analyzed data from Red Earth Creek, AB, Bonanza Creek, AK, and the Great Slave Lake region. Preliminary results show high correlation (R2 = .82) between Functional Principal Components and in situ soil moisture variance. VMC estimation at discrete time stamps is less accurate, but provides a promising avenue for future soil moisture retrievals, especially with the impending launch of new SAR sensors such as NISAR.

Soil moisture is an important indicator of fire danger in the boreal. Understanding soil moisture conditions using remote sensing is difficult due to the soil surface typically being obscured by vegetation cover. Synthetic Aperture Radar (SAR) sensors, which emit microwave radiation and measure the backscattered intensity signals, are well suited for assessing soil moisture conditions as the lower frequency radiation is capable of penetrating vegetation biomass and interacting with the soil. Accurate in situ soil moisture data, collected close to the SAR collections, is needed to parameterize and validate SAR derived soil moisture retrieval algorithms. These data are difficult to collect as they require specialized equipment (TDR probes) and the ability to travel to the regions of interest, which can be time consuming and costly. Therefore, robust datasets with corresponding SAR imagery and field data are lacking. To overcome this obstacle, we tested the use of Functional Data Analysis (FDA) with time-series SAR data from a variety of sensors to provide information on soil moisture patterns in various regions of the ABoVE domain where in situ soil moisture data is available. By employing FDA, data within the SAR time series is treated as a series of functions rather than discrete data points. This approach allows both for the estimation of soil properties, such as drainage characteristics, and for the estimation of moisture anywhere along the approximated functional curves derived from the input data. For this project, we analyzed data from Red Earth Creek, AB, Bonanza Creek, AK, and the Great Slave Lake region. Preliminary results show high correlation (R2 = .82) between Functional Principal Components and in situ soil moisture variance. VMC estimation at discrete time stamps is less accurate, but provides a promising avenue for future soil moisture retrievals, especially with the impending launch of new SAR sensors such as NISAR.

We present here results from the analysis of the concentrations, composition, and optical properties of constituents in coastal Beaufort Sea waters along the North Slope of Alaska. The data were collected over a two-year period and covers the coastal North Slope region from the Colville delta on the west to the Canning estuary on the east. The dataset includes a mix of coastal and offshore stations. The measurements collected include concentrations of total and size-fractionated chlorophyll-a, suspended particulate matter, colored dissolved organic matter, particulate organic carbon, and dissolved organic carbon, absorption and scattering of light through the water column, at-surface radiometry, and airborne hyperspectral imagery. The goal of the study was to investigate geospatial variations in optical and biophysical properties along the North Slope coast and in offshore waters. The data reveal distinct differences in optical and biophysical properties between stations north and south of the barrier islands around Prudhoe Bay, thus demonstrating the influence of the barrier islands on the biophysical properties of coastal Arctic waters. Data from offshore stations on the eastern part of the study area show the influence of the Mackenzie River plume several hundred kilometers from the source. Spatio-temporal variations in suspended particle concentrations and size characteristics and their relation to variations in river discharge are presented. The study presents a report on geospatial and temporal variations in the composition and optical properties of U.S. coastal Arctic waters in the North Slope region.

We present here results from the analysis of the concentrations, composition, and optical properties of constituents in coastal Beaufort Sea waters along the North Slope of Alaska. The data were collected over a two-year period and covers the coastal North Slope region from the Colville delta on the west to the Canning estuary on the east. The dataset includes a mix of coastal and offshore stations. The measurements collected include concentrations of total and size-fractionated chlorophyll-a, suspended particulate matter, colored dissolved organic matter, particulate organic carbon, and dissolved organic carbon, absorption and scattering of light through the water column, at-surface radiometry, and airborne hyperspectral imagery. The goal of the study was to investigate geospatial variations in optical and biophysical properties along the North Slope coast and in offshore waters. The data reveal distinct differences in optical and biophysical properties between stations north and south of the barrier islands around Prudhoe Bay, thus demonstrating the influence of the barrier islands on the biophysical properties of coastal Arctic waters. Data from offshore stations on the eastern part of the study area show the influence of the Mackenzie River plume several hundred kilometers from the source. Spatio-temporal variations in suspended particle concentrations and size characteristics and their relation to variations in river discharge are presented. The study presents a report on geospatial and temporal variations in the composition and optical properties of U.S. coastal Arctic waters in the North Slope region.

Climate-driven temperature changes in the Arctic are estimated to be increasing at a rate of up to four times that of the rest of the planet (Rantenen, 2022), triggering a cascade of impacts in the region and across the planet. Terrestrial disturbances like wildfires and permafrost thaw impact Alaskan Indigenous cultural practices like subsistence, pose repercussions for emissions targets, and consequences for infrastructure and ecosystems in the region. Effective decision-making in response to these consequences depends upon robust observing systems. Yet, the region’s observing systems present profound deficiencies compared to mid-latitudes, such as poor spatial coverage of in-situ observations. While many of these deficiencies can be attributed to polar conditions, one often overlooked consideration is that Arctic Indigenous peoples have lived in these ‘remote’ areas since time immemorial and routinely observe them in support of their subsistence, stewardship, and cultural practices. Recognizing this, Indigenous communities continue to call for increased funding for community-led observing systems supporting their self-determination in a changing climate. Therefore, policymakers, funders, and others engaged with improving Arctic observations must weigh diverse success criteria (e.g., community resilience and global indicators of change) and implementation strategies (e.g., conventional networks and community-led monitoring) to make strategic and equitable investments. In response, the United States Arctic Observing Network (US AON) has developed Benefit Assessment methods and a Benefit Tool to support informed solutions. By gathering structured input from Arctic subject matter experts, Benefit Assessments systematically link observational inputs to the data products, applications, and societal benefits they support. The Benefit Tool enables analysis and visualizations to support a variety of research planning and policy use cases for funders, community members, and researchers; Benefit examples of relevance to the objectives of NASA’s ABOVE will be highlighted to illustrate the potential of expanding this work.
  [poster]
Climate-driven temperature changes in the Arctic are estimated to be increasing at a rate of up to four times that of the rest of the planet (Rantenen, 2022), triggering a cascade of impacts in the region and across the planet. Terrestrial disturbances like wildfires and permafrost thaw impact Alaskan Indigenous cultural practices like subsistence, pose repercussions for emissions targets, and consequences for infrastructure and ecosystems in the region. Effective decision-making in response to these consequences depends upon robust observing systems. Yet, the region’s observing systems present profound deficiencies compared to mid-latitudes, such as poor spatial coverage of in-situ observations. While many of these deficiencies can be attributed to polar conditions, one often overlooked consideration is that Arctic Indigenous peoples have lived in these ‘remote’ areas since time immemorial and routinely observe them in support of their subsistence, stewardship, and cultural practices. Recognizing this, Indigenous communities continue to call for increased funding for community-led observing systems supporting their self-determination in a changing climate. Therefore, policymakers, funders, and others engaged with improving Arctic observations must weigh diverse success criteria (e.g., community resilience and global indicators of change) and implementation strategies (e.g., conventional networks and community-led monitoring) to make strategic and equitable investments. In response, the United States Arctic Observing Network (US AON) has developed Benefit Assessment methods and a Benefit Tool to support informed solutions. By gathering structured input from Arctic subject matter experts, Benefit Assessments systematically link observational inputs to the data products, applications, and societal benefits they support. The Benefit Tool enables analysis and visualizations to support a variety of research planning and policy use cases for funders, community members, and researchers; Benefit examples of relevance to the objectives of NASA’s ABOVE will be highlighted to illustrate the potential of expanding this work.