Comparison of different machine learning method for GPP estimation using remote sensing data

Abstract

This paper selects eight sites with typical characteristics in China (Changbai Mountain, Qianyanzhou, Dinghushan, etc.). Based on remote sensing data acquired from the Google Earth Engine (GEE) big data cloud platform, four machine learning models were established to estimate GPP. Firstly, remote sensing data such as EVI, NDVI, precipitation and temperature were downloaded by GEE, and the flux tower data of 8 sites of China-FLUX was obtained. Secondly, the machine learning algorithm is used to establish the connection between the two types of data. Finally, the machine learning model is used to predict the test group data, and the results are evaluated by using R2, RMSE and other related precision indicators, and the accuracy of the MODIS data is compared. Studies have shown that machine learning models can obtain more accurate GPP predictions.