遥感降水降尺度高精度校正及不确定性分析方法

High-precision correction and uncertainty analysis method for remote sensing precipitation downscaling

为消除降水场同质部分影响,提升统计降水降尺度结果精度,提出了基于贝叶斯高精度曲面建模(Bayes-HASM)算法的遥感降水降尺度高精度校正方法。该方法通过引入模拟精度更高的高精度曲面建模方法,并结合贝叶斯优化算法,实现了模型参数自动优化选择和高精度降尺度校正,解决了现有降尺度残差校正方法存在的误差和多尺度问题。结果表明:贝叶斯优化使高精度曲面建模的不确定性显著减少;经过Bayes-HASM残差校正后,降尺度结果的散点分布更加接近1∶1线,年、季、月和旬尺度的精度指标均得到了显著的改善,CC和IA指标提高至0.9左右,RMSE下降明显,RB也显著改善。本方法能显著降低模型的不确定性并起到消除降水场同质部分影响的作用,有效提升降水降尺度结果精度。

To eliminate the influence of homogeneous parts in precipitation fields and enhance the accuracy of statistical precipitation downscaling results,a high-precision correction method is proposed for remote sensing precipitation downscaling based on the Bayesian High-Accuracy Surface Modeling(Bayes-HASM)algorithm.This method introduces a higher-precision surface modeling approach and combines it with Bayesian optimization algorithms to achieve automatic optimization of model parameters and high-precision downscaling correction.It addresses the errors and multi-scale issues present in existing downscaling residual correction methods.The results indicate that Bayesian optimization significantly reduces the uncertainty of high-precision surface modeling;after Bayes-HASM residual correction,the scatter distribution of the downscaled results is closer to the 1∶1 line.The accuracy indicators at the annual,seasonal,monthly,and ten-day scales have been significantly improved,with CC and IA indicators reaching around 0.9,RMSE has significantly decreased,and RB significantly reduced.The above results indicate that this method can significantly reduce the model’s uncertainty and effectively eliminate the impact of homogeneous parts of the precipitation field,thereby improving the accuracy of the downscaled precipitation results.