Land cover changes and ecosystem productivity under the changing climate in Central Asia
Wei Fang, Pace University, wfang@pace.edu (Presenter)
Sakshi Saraf, University of South Dakota, sakshi.saraf@coyotes.usd.edu
Venkatesh Kolluru, University of South Dakota, venkatesh.kolluru@coyotes.usd.edu
Maxine Uebelacker, Pace University, mu10609n@pace.edu
Ranjeet Johns, University of South Dakota, ranjeet.john@usd.edu
Jiquan Chen, Michigan State University, jqchen@msu.edu
Central Asia falls mostly in arid and semi-arid climate, with high ecological fragility and cultural significance. These countries have experienced socio-ecological, environmental and institutional shifts since 1991, and are highly vulnerable to climate change and other socioeconomic shifts. Over the past 20 years, Central Asia lost 4.8% of forests, 1.3% of grasslands (37,483 km2), and 10.5% of croplands (22,612 km2), while gained 30.8% shrublands, 50.9% open woodland, 11.6% wetlands, and 2.2% built-up area. However, the relative influences and contributions of climate and socio-economic forcing on vegetation remain unclear. In this study, we used random forest (RF) machine learning model to partition the relative contributions of climatic and social-economic forcing on gross primary productivity (GPP) for each of the seven biomes during 2001-2020 (R2 of 0.59-0.89). We employed an explainable artificial intelligence (XAI) framework to estimate GPP using a Random Forest model with Shapley values for interpretation. Among all 7 biomes, climate and ecological variables (e.g., precipitation, temperature, soil moisture, elevation, solar radiation and snow depth or cover) are consistently the top 6 predictors (with highest Shapley values) for predicting GPP, while social-economic variables (e.g., gross domestic product, population density, nighttime light) are consistently the least important predictors (with low Shapley values). Drought, i.e. scaled palmer drought severity index, was consistently the most important predictors of GPP anomaly for most biomes. The study will not only shed new light on shifts of land cover and productivity under climate change, but also provide basic land cover change information for local policy makers and resource managers.
Associated Project(s):
Poster Location ID: 2-5
Presentation Type: Poster
Session: Poster Session 2
Session Date: Wed (May 10) 5:15-7:15 PM
CCE Program: LCLUC