Satellite radar inversion fusion method based on NRIET machine learning

Abstract

The invention discloses a satellite radar inversion fusion method based on NRIET machine learning. The method comprises a training phase and an inversion phase. The training phase includes satellite data and processing: subjecting the disc projection data of four waveband including B08, B10, B13 and B15 of a sunflower No. 8 geostationary satellite to projection conversion and interpolating the same to a spatial grid point of 0.02*0.02 degree; lightning data and processing; radar data and radar data networking; land use type data and processing; and model training. The inversion phase includes:data preprocessing: subjecting the sunflower No. 8 geostationary satellite data obtained by real-time observation and the lightning data to the same processing as the training phase, the land use type data using the data processed in the training phase; the radar data networking: subjecting the radar-base data observed in real time to networking according to the networking steps in the training phase to calculate the combined emissivity; radar echo inversion; and satellite radar data fusion.

Description

technical field [0001] The invention relates to a radar echo inversion and fusion algorithm, in particular to an NRIET machine learning-based satellite radar inversion and fusion method. Background technique [0002] The existing radar observation range is limited. In the western region where radar deployment is relatively sparse, there is a large gap in the radar network. At the same time, the radar's observation range at sea is limited and can only cover offshore areas. In order to make up for the gaps in the radar network and the lack of maritime observations, a set of radar combined reflectivity algorithms based on satellite observation data has been developed. At present, this technology is not widely used, and the existing algorithms can be roughly divided into two categories: machine learning algorithms based on backpropagation (BP) neural network and radar echo inversion algorithms based on inversion of precipitation. However, both algorithms have certain limitatio...

AI Technical Summary

Problems solved by technology

The principle of this method is simple, but the result error is large, and there is no data for visible light at night

Microwave radiometers are currently only installed on polar-orbiting satellites, and the repetition period of polar-orbiting satellites usually takes 1 day, and precipitation data with high temporal resolution cannot be obtained.

And this method introduces precipitation intensity as an intermediate variable, and establishes the relationship between satellite observation and precipitation and radar echo and precipitation respectively, which increases the uncertainty factors and leads to the decrease of the accuracy of the results.

[0005] In addition, the above two algorithms do not consider the land and the ocean separately. There are significant differences in the concentration of aerosols and water vapor on the land and the ocean, and there are obvious differences in the concentration and diameter of cloud droplets on the land and the sea. its considered separately

In addition, the above two algorithms both regard the inverted radar echo as a separate product, and do not combine the inverted radar echo with the real-time observed radar echo for data fusion

Images