Land surface data assimilation method and device for numerical weather forecast and application

The invention provides a land surface data assimilation method and device and application, and the method comprises the following steps: setting an assimilation region, obtaining ensemble forecast data and land surface observation data in the assimilation region, and obtaining the background variance of the ensemble forecast data through coupling parameters; sparsification is carried out on the land surface observation data by using a sparsification parameter to obtain observation data to be assimilated; performing localization on the background variance by using a horizontal localization parameter and a vertical localization parameter to obtain a mode space; and assimilating each observation point in the mode space based on the observation data. Through the scheme, observation sparsification can be carried out based on the distance, multiple times of iteration updating can be carried out, and parameter adjustment is carried out on the background variance by adopting a localization strategy, so that the overa