Publications

Zhang, QY; Yao, T; Huemmrich, KF; Middleton, EM; Lyapustin, A; Wang, YJ (2020). Evaluating impacts of snow, surface water, soil and vegetation on empirical vegetation and snow indices for the Utqiagvik tundra ecosystem in Alaska with the LVS3 model. REMOTE SENSING OF ENVIRONMENT, 240, 111677.

Abstract
Satellite observations for the Arctic and boreal region may contain information of vegetation, soil, snow, snowmelt, and/or other surface water bodies. We investigated the impacts of vegetation, soil, snow and surface water on empirical vegetation/snow indices on a tundra ecosystem area located around Utqiagvik (formerly Barrow) of Alaska with the Moderate Resolution Imaging Spectrometer (MODIS) images in 2001-2014. Empirical vegetation indices, such as normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), the index of near infrared of vegetation (NIRv), and modified EVI (EVI2), have been used to monitor vegetation. Normalized difference snow index (NDSI) has been widely applied to monitor snow. The vegetation cover fraction (VGCF), the soil cover fraction (SOILCF), the snow cover fraction (SNOWCF), the surface water body cover fraction (WaterBodyCF), the fractional absorption of photosynthetically active radiation (PAR) by vegetation chlorophyll (fAPAR(chl)), the fractional absorption of PAR by non-chlorophyll components of the vegetation (fAPAR(non-chl)), and the fractional absorption of PAR by the entire canopy (fAPAR(canopy)) are retrieved with the MODIS images and a coupled Leaf-Vegetation-Soil-Snow-Surface water body radiative transfer model, LVS3. The vegetation indices (NDVI, EVI, EVI2 and NIRv) differ from VGCF, fAPAR(chl), fAPAR(non-chl), and fAPAR(canopy). In addition to vegetation, we find that soil, snow and surface water also have impacts on vegetation indices NDVI, EVI (EVI2), and NIRv. Presence of snow makes lower the observed values of NDVI, EVI2 and NIRv. After snowmelt is gone, the vegetation indices (NDVI, EVI, EVI2 and NIRv) linearly decrease with SOILCF and WaterBodyCF, and WaterBodyCF has stronger impacts on these vegetation indices than SOILCF. The relationship between EVI and snow is complicated. NDSI non-linearly increases with SNOWCF, but linearly increases with sum of SNOWCF and WaterBodyCF (sum = 0.5893 x NDSI +0.4342, R-2 = 0.976). NDSI linearly decreases with VGCF, and the relationship between NDSI and SOILCF is complex. Retrievals of VGCF, fAPAR(chl), fAPAR(non-chl) and fAPAR(canopy) with the LVS3 model provide alternatives for vegetation monitoring and ecological modeling.

DOI:
10.1016/j.rse.2020.111677

ISSN:
0034-4257