基于FY-3D卫星的微波与光学陆表温度融合研究  

Fusion of Microwave and Optical Land Surface Temperature Based on FY-3D Satellite

作　　者：刘勇洪 翁富忠 徐永明 韩秀珍[4] 段四波[5] 唐世浩[4] 叶成志 LIU Yonghong;WENG Fuzhong;XU Yongming;HAN Xiuzhen;DUAN Sibo;TANG Shihao;YE Chengzhi(CMA Earth System Modeling and Prediction Centre,Beijing 100081;State Key Laboratory of Severe Weather,Beijing 100081;School of Remote Sensing and Geomatics Engineering,Nanjing University of Information Science and Technology,Nanjing 210044;National Satellite Meteorological Centre,Beijing 100081;Institute of Agricultural Resources and Regional Planning,Chinese Academy of Agricultural Sciences,Beijing 100081;Hunan Meteorological Service,Changsha 410118)

机构地区：[1]中国气象局地球系统数值预报中心,北京100081 [2]灾害天气国家重点实验室,北京100081 [3]南京信息工程大学遥感与测绘工程学院,南京210044 [4]国家卫星气象中心,北京100081 [5]中国农业科学院农业资源与农业区划研究所,北京100081 [6]湖南省气象局,长沙410118

出　　处：《气象》2024年第1期1-17,共17页Meteorological Monthly

基　　金：国家自然科学基金项目(U2142212);湖南省自然科学基金重大项目(2021JC0009);风云卫星应用先行计划共同资助。

摘　　要：目前还没有基于国产卫星的1 km分辨率的全天候陆表温度(LST)产品,FY-3D卫星提供了中分辨率成像仪(MERSI)Ⅱ型1 km分辨率晴空LST产品与微波成像仪(MWRI)25 km全天候LST产品,因此可结合两者优势开展全天候1 km分辨率LST的融合研究。基于地理加权回归(GWR)方法,选择海拔、FY-3D归一化植被指数和归一化建筑指数等建立GWR模型对FY-3D/MWRI 25 km LST降尺度到1 km,并与MERSI 1 km LST进行融合;同时针对MWRI轨道间隙,利用前后1天融合后的云覆盖像元1 km LST进行补值,可以得到接近全天候下的1 km LST。基于以上融合算法,选择了中国区域多个典型日期FY-3D/MERSI和MWRI LST官网产品进行了融合试验,并利用公开发布的全天候1 km LST产品(TPDC LST)对FY-3D 1 km LST融合结果进行了评估。研究结果表明,基于GWR法的LST降尺度方法,可以有效避免传统微波LST降尺度方法中存在的“斑块”效应和局地温度偏低等问题;LST融合结果有值率从融合前的22.4%~36.9%可提高到融合后69.3%~80.7%,融合结果与TPDC LST的空间决定系数为0.503~0.787,均方根误差为3.6~5.8 K,其中晴空为2.6~4.9 K,云下为4.1~6.1 K;分析还表明目前官网产品FY-3D/MERSI和MWRI LST均存在缺值较多与精度偏低等问题,显示其存在较大改进潜力,这有利于进一步改进FY-3D LST融合质量。There is no 1 km spatial resolution all-weather land surface temperature(LST)product based on domestic satellites in China.FY-3D satellite provides the clear sky LST products with 1 km resolution from the medium resolution spectral imager(MERSI)Ⅱand the all-weather LST products with 25 km resolution from the microwave radiation imager(MWRI).Therefore,the integration research of all-weather 1 km resolution LST can be carried out by combining their advantages.By using geographical weighted regression(GWR)method,this study selects altitude,FY-3D normalized difference vegetation index and normalized difference building index to establish LST downscaling regression model in order to downscale FY-3D/MWRI 25 km LST to 1 km,and integrates them with MERSI 1 km LST.For the MWRI track gaps,1 km LST of cloud covered pixels fused in the previous one day and the next day can be used to supplement,which is close to the all-weather 1 km LST.Based on the above fusion algorithms,multiple Chinese FY-3D/MERSI and MWRI LST products on typical dates from the official website are selected for fusion test,and the existing all-weather 1 km LST products(TPDC LST)were used to evaluate the results of FY-3D 1 km LST fusion products.The results show that the LST downscaling method based on GWR method can effectively eliminate the“patches”effect and low local temperature in traditional microwave downscaling methods based on the combination of altitude with mixed pixel decomposition.The rate of FY-3D 1 km LST can be increased from 22.4%-36.9%before fusion to 69.3%-80.7%after fusion.The spatial correlation between the fusion product and TPDC LST is 0.503-0.787,and the RMSE is 3.6-5.8 K with 2.6-4.9 K in clear sky and 4.1-6.1 K in cloudy sky.The analysis also shows that the current FY-3D/MERSI LST and MWRI LST products from the official website have problems such as obvious lack of value and low accuracy,suggesting that they have a great potential to be improved.This is conducive to further improving the quality of FY-3D LST fusion.

关 键 词：FY-3D MERSI-Ⅱ FY-3D MWRI 全天候 陆表温度 地理加权回归 

分 类 号：P412[天文地球—大气科学及气象学] P423