Modeling net primary productivity of wetland with a satellite-based light use efficiency model

Wetland, an important carbon pool on the earth, is of great significance for human beings and the environment. In this study, we modeled the wetland NPP using Carnegie-Ames-Stanford Approach (CASA) and time series with high spatial and temporal resolution generated by Landsat 8 and Sentinel-2 images. Firstly, the downscaled Landsat 8 data (10 m) combined with Sentinel-2 data were utilized to produce time series with high spatial and temporal resolution. Subsequently, all Sentinel-1 and Sentinel-2 data within each stage (five stages.) are employed to obtain monthly wetland maps in these stages by an adaptive Stacking algorithm, respectively. Then, monthly fraction cover maps (mainly sedge, reed and poplar) were derived from the time series reflectance product using fully constrained least squares (FCLS) in these five Stages, respectively. Finally, monthly normalized difference vegetation index (NDVI), land surface water index (LSWI), temperature, solar radiation, as well as wetland maps and were combined to estimate monthly and total NPP in the Dongting Lake wetland by CASA model. The high correlation (R-2 = 0.8445) and low (RMSE = 20.30 g C/m2) between the estimated NPP using proposed method and measured NPP demonstrated a significant linear relationship and the estimated NPP based on Sentinel-2 data using the CASA model with the above-described input parameters is creditable. The NPP estimation method in this paper is expected to provide scientific data support for quantitative research of regional wetland carbon reserves and sustainable development.