Mongolia 30m resolution grass yield estimation dataset (2017-2021)

Deep neural network training: use the GEE cloud computing platform to obtain the NDVI, surface temperature and precipitation data corresponding to the measured grass yield coordinates. The obtained data and the measured grass yield were normalized, and a three-layer neural network model was built. Fitting the relationship between the grass yield and the other three influencing factors, build the grass yield estimation model. Due to the small number of training samples, the method of cross validation is adopted. 70% of the training samples are randomly selected as the training set and 30% of the data as the validation set. The grass yield estimation model is obtained through three cross validation.Grass yield estimation: based on the improvement of model estimation accuracy, the three grass yield estimation models trained are loaded into the GEE platform by calling the Python language API interface provided by GEE. Complete a series of steps such as calling and processing the basic remote sensing data in the GEE cloud platform, take the average of the retrieval results of the three grass yield models, and obtain the estimation data of Mongolia's domestic grass yield.Resampling and re-projection of remote sensing data are completed on the GEE platform.