Multi-source satellite precipitation fusion method based on Stacking and EMOS-CSG

Abstract

The invention discloses a multi-source satellite rainfall fusion method based on Stacking and EMOS-CSG, and the method comprises the steps: data preprocessing: obtaining rainfall station, original satellite rainfall and auxiliary variable data; downscaling each original satellite product by means of a geographically weighted regression model; a plurality of base learners on the first layer of the Stacking integrated framework are adopted to carry out deviation correction on each downscaling satellite product; integrating the deviation correction results of the downscaling satellite products by using a second-layer learning device of the Stacking integration framework; and based on an EMOS-CSG method, Stacking integration results are fused, and a rainfall fusion product with high temporal-spatial resolution is obtained. According to the method, the advantages of different satellite precipitation products are effectively combined through a Stacking and EMOS-CSG fusion method, the deviation of the satellite products is corrected by means of a machine learning model, the precision of the fusion product is improved, and high-precision precipitation input can be provided for a hydrological model.

Description

technical field [0001] The invention belongs to the technical field of hydrology and meteorology, and in particular relates to a multi-source satellite precipitation fusion method based on Stacking and EMOS-CSG. Background technique [0002] High-precision precipitation data is of great significance to flood forecasting, drought monitoring and water resource management in river basins. There are usually two ways to obtain precipitation data: one is to obtain directly through ground rainfall stations, and the other is to obtain indirectly through remote sensing means (mainly radar and satellite). Both methods have their own advantages and disadvantages. Precipitation stations have high data accuracy and long data series, but often the station density is not high, and it is difficult to accurately reflect the spatial distribution of precipitation; satellite precipitation covers a wide range, but due to the influence of inversion algorithms, cloud properties and sensors, there ...

AI Technical Summary

Problems solved by technology

[0004] It is feasible to select the optimal algorithm from a variety of machine learning algorithms or use a new machine learning algorithm for fusion, but different machine learning algorithms have their own advantages and disadvantages. How to make full use of the advantages of each machine learning algorithm in precipitation fusion research In addition, traditional fusion methods are usually based on the original satellite resolution (0.25° or 0.1°), and the resulting fusion products are still too coarse for hydrological and meteorological applications

Images