Study On Wind Retrieval Accuracy Of HY-2A Scatterometer Based On BP Neural Network

Haiyang-2A(HY-2A)is the first ocean dynamic environment satellite developed in our country,with a microwave scatterometer(HY2SCAT)which can provide a large number of real-time sea surface wind field data.The surface wind field data in the Marine environment numerical forecast is essential,but the retrieval precision of scatterometer wind field is affected by environmental factors such as precipitation.The aim of this paper is to consider the effect of environmental factors such as precipitation on the retrieval accuracy of HY-2A scatterometer,and analyze the accuracy of the HY-2A scatterometer wind field and modifies it by BP neural network to improve the quality of the HY-2A scatterometer data in the wind field.Using three years of multi-source wind vector data from 2012 to 2014,including NDBC buoys,TAO buoy observation data of wind field and ERA-Interim reanalysis wind field data,matching the corresponding CMORPH precipitation data,accuracy of HY-2A near real-time sea surface wind field data from 2012 to 2014 is analyzed,with the mean bias,root mean square error,and standard deviation,and we also analysis H2SCAT error compared with different data;By using the BP neural network to correct the deviation of the HY-2A scatterometer vector field,the residual distribution of the wind vector field of HY-2A scatterometer and the data wind field of each buoy was analyzed according to the wind speed.At the same time,sensitivity test was designed to explore the sensitivity of the input sample size of neural network and the environmental factors in each input element to the corrected results.The results of this study are as follows:retrieval wind speed of HY-2A scatterometer compared with 3 different wind speed observation data are overestimated,and the wind direction presents irregular nonlinear changes;After the deviation correction with the BP neural network,retrieval accuracy of HY-2A scatterometer wind field is improved greatly,especially the wind speed.After 19 sets of sensitivity tests were performed on the ERA-Interim reanalysis wind field data which are sufficient,HY-2A wind speed deviation from average 0.99 m/s down to 0.01 m/s average,with 0.2×10-3 m/s minimum;The corrected effect of the environmental elements in the input sample,including both the sea temperature and precipitation,is better than that of the model which only includes one of the environmental factors;the wave data is added to the input sample,and the wind speed data of HY-2A scatterometer will also have a certain correction effect.Within a certain input sample size,when the sample size gradually increases,the average deviation will gradually decrease,the correction effect will be better,but over a certain amount it will produce overfitting phenomenon,with the gradual increase of the input sample size,the mean bias does not decrease,but fluctuates and gradually increases generally.