| Soil moisture is one of the key factors in the climate system and is a basic element of the water cycle,energy cycle and the biogeochemical cycle.It plays a crucial role in the hydrological and climate processes such as precipitation,runoff,evapotranspiration and drought.A variety of passive and active microwave sensors or satellites have been designed and launched since the 70s of last century for surface soil moisture monitoring.However,the spatial resolution of the soil moisture products is?25km,which is not suitable for the accurate applications at regional scale.The SMAP(Soil Moisture Active and Passive),launched in January 2015,is the first satellite carrying the L-band microwave radiometer and radar sensors at the same platform.The global 9km soil moisture(AP9)is retrieved from SMAP joint observation data,which satisfies the applications at regional scale.However,the damage of SMAP radar sensor causes the less than three months of effective AP9 and only SMAP 36km soil moisture(P36)released as usual.Therefore,aiming at the coarse spatial resolution of current microwave soil moisture produtcts and the absence of AP9,this paper fouces on microwave soil moisture spatial resolution enhancement from three aspects:spatial domain statistics,time-spatial information complementarity and multi sensors fusion.Taking AP9 extension as the starting point and microwave 9km soil moisture estimation as the main line,the global long time series(1978-2017)9km soil moisture product is estimated in the paper.The main contents are as follows:(1)The multi factor statistics-residual correction soil moisture downscaling modelThe traditional downscaling methods with the assistance of optical/thermal infrared data can improve the coarse microwave soil moisture to high spatial resolution.The core of the methods is to establish the downscaling relation between microwave soil moisture and optical/thermal infrared prediction factors.The location and land cover information are added into downscaling model and one kind of multi factor statistics-residual correction downscling model(MFSRC)is proposed using the the spatial domain information between soil moisture and prediction factors fully.Then,MFSRC is used to enhance the spatial resolution of P36 to 9km,and discuss the feasibility of AP9 extension using MFSRC.The evaluation results show that the performance of MFSRC is better than the traditional downscaling method.The estimation accuracy of MFSRC 9km soil moisture is close to P36 and better than AP9.Compared with the AP9 substitution product EP9(Ehanced SMAP passive 9km soil moisture),adopted by SMAP team using the BG(Backus-Gilbert)interpolation inversion algorithm,the accuracy of the two kinds of 9km soil moisture is comparable.Moreover,the spatial detailed information of MFSRC 9km soil moisture is better than that of EP9.Therefroe,it suggests that MFSRC can be used to extend the missing AP9.However,the spatial details richness of estimated 9km soil moisture by the above methods(including MFSRC)can not reach the level of AP9.(2)Extending SMAP 9km soil moisture using a spatial-temporal fusion modelIn addition,the SMAP team also proposed a data replacement model for Sentinel-1 radar C band in view of AP9 absence.However,the above methods only use spatial domain information between soil moisture and the auxiliary data,and does not take temporal domain information into account.Considering the strong temporal and spatial correlation of soil moisture and the existing three month of AP9,a soil moisture spatio-temporal fusion model(STFM)is proposed in this paper.Then the 9km soil moisture is estimated by STFM with the aid of SMAP 36km and 9km soil moisture before the SMAP radar was damaged so as to extend AP9.The estimated 9km soil moisture using STFM is called as STF9 and two years of STF9 is estimated from 2015-04 to 2017-04 in the paper.Then,a series of evaluations are carried for STF9.The results show that the performance of STFM is superior to the BG interpolation inversion algorithm and the proposed MFSRC for 9km soil moisture estimation.The spatial details richness of STF9 is obviously better than that of the estimated 9km soil moisture by the above two methods.More importantly,it can achieve the level of AP9 in spatial detailed information.It reveals that the advantages of the accuracy of P36 and the spatial resolution of AP9 are integrated by STF9.The accuracy and spatial resolution are mutually restricted for P36 and AP9,to a certain extent,the advantage integration of STF9 alleviates the contradiction between SMAP two kinds of soil moisture.Therefore,it suggests that STFM is an effective way to extend AP9 in the case of the malfunctioned SMAP radar.(3)Multi-sensor fusion for 9km soil moisture estimationThe soil moisture STFM is quite good for extending AP9,however,the previous study is implenmented only two years.The ability of the model to estimate 9km soil moisture over longer time series remains to be verified.In this paper,a global long time series 9km soil moisture generation using multi sensors fusion is established based on STFM so as to intergrate SMAP and the CCI(1978-2015)0.25° spatial resolution soil moisture products(CCI0.25).Thus,the CCI time series 9km soil moisture products are generated,and the data blanks are filled at the scale.The results show that the fused 9km soil moisture enhances the spatial resolution of CCI0 25 with the detailed information and maintains its temporal variation well.Anomalous time series analysis demostrates that fused 9km soil moisture can capture the anomalous changes of CCI0.25 very well,suggesting that it can reveal the climate change of CCI0.25 on a finer scale.Evaluations against in-situ soil moisture show that the accuracy of the fused results is close to that of CCI0.25 with a lower ubRMSE,because the spatial neighborhood information is added to the fusion process,which inhibits the abnormal value information of soil moisture.In addition,it is found that the low spatial resolution soil moisture in the baseline data required by STFM has a great influence on the spatial distribution of fused 9km soil moisture and it has little influence on the detailed information,time series variation and the temporal accuracy.It is found that the STFM is an effective method to improve the spatial resolution of microwave soil moisture,and its downscaling performance is obviously better than the previous methods.Therefore,this paper estimates the global long time series 9km soil moisture products from 1978 to 2017 by using STFM.The 9km product with rich spatial detailed information inherits the temporal accuracy of the original microwave soil moisture.Although microwave remote sensing technology is suitable to monitor surface soil moisture,there is a dilemma for the retrieval soil moisture:the higher of the accuracy,the coarser of the spatial resolution.The STFM initially solves the problem of mutual restriction between the spatial resolution and accuracy of the existing microwave soil moisture products.Therefore,the estimated long time series(1978-2017)9km soil moisture will play an important role in researches and applications at regional scale. |
PDF Full Text Request