Upscaling flux tower data into maps of regional forest carbon removals
Siddhanth Hegde, Indiana University, siduheg@iu.edu (Presenter)
Mallory Barnes, Indiana University, malbarn@indiana.edu
Kim Novick, Indiana University, knovick@iu.edu (Presenter)
Eddy-covariance flux towers provide robust, high-frequency information about the exchange of carbon between ecosystems and the atmosphere. When lateral losses of carbon (e.g. to harvest or runoff) are small or accounted for, data from flux tower networks like AmeriFlux and Fluxnet are able to quantify atmospheric carbon removals directly, integrating over all sources and sink and providing an important complement to bottom-up approaches based on plant and soil inventory data. However, because towers are expensive to install and operate, flux tower networks tend to be relatively sparse, necessitating the use of machine-learning upscaling approaches to transform tower time series into carbon removal maps that are most useful for policy- and management-decisionmaking.
Here, we present an approach to map the net ecosystem exchange (or NEE) of Eastern US forests using machine learning approaches driven by data from forested AmeriFlux sites across the study region. By leveraging key products from remote sensing satellites, such as MODIS, we upscale NEE using vegetation indices like NDVI and meteorological variables including day length and temperature. Our method employs robust feature selection algorithms such as permutation importance, to effectively explain NEE while preserving model performance. We apply a 10 fold cross-validation technique similar to those used in FLUXCOM and achieve better results by focusing specifically on the Eastern US. The upscaled NEE maps are generated at a resolution of 500m x 500m, providing detailed visualizations of forest carbon dynamics across the region. This approach offers enhanced accuracy in NEE estimation and contributes valuable insights for understanding ecosystem carbon fluxes in Eastern US forests.
Associated Project(s):
Poster Location ID: 13
Presentation Type: Poster
Theme: Land Biomass