基于薄板平滑样条模型的中国大陆区域降水数据融合研究

Merging Precipitation Datasets in China Mainland Based on Thin-plate Smoothing Spline Model

高精度的降水网格数据产品的研制对于数值天气模式检验、水文陆面模拟、山洪地质灾害监测都有着重要意义。本文选取2015年中国大陆区域740个常规气象观测站的日降水观测数据以及CMORPH卫星日降水数据作为研究对象,采用将CMORPH卫星数据作为协变量的薄板平滑样条模型将两种来源的降水观测数据进行融合,并尝试应用调整的Cressman方案对薄板平滑样条的残差场进行订正。最后以740个地面常规气象站的降水观测数据为标准对融合方案以及残差场的订正效果进行评估和分析。结果表明,将CMORPH卫星降水数据作为样条模型的协变量比没有协变量的方案在中国大陆区域整体上精度提高了约7%,其中CMORPH数据在南部沿海地区的表现优于青藏高原地区以及高纬度地区;应用调整的Cressman方案对薄板平滑样条模型的残差场进行订正不能提高降水估计场的精度;利用薄板平滑样条模型产生的降水估计场的精度比CMOPRH卫星降水数据的精度提高了约12%。

Developing high-accuracy precipitation grid dataset is of great significance to verification of numerical weather models,hydrological and land surface simulation,and monitoring of geological hazard monitoring.Based on daily rainfall observations from about 740 rain gauges in China Mainland in the whole year of 2015,this paper used the thin-plate smoothing spline model to merge different kinds of precipitation datasets and applied the adjusted Cressman weight to correct the residuals of the spline model.The merging scheme the paper proposed was validated by the observations from 740 rain gauges in China Mainland.The results suggested,using CMORPH data as a covariate can improve the accuracy of the model by about 7%as a whole.The performance of CMORPH dataset as a covariate in the southern coastal areas is better than that in the Qinghai-Tibet Plateau and higher latitude areas.Adjusted Cressman weight can not improve the accuracy of the estimated field generated by thin-plate smoothing spline model.The merging precipitation dataset generated by thin plate smoothing spline model with CMORPH as a covariate was significantly better than the CMOPRH dataset by 12%.