Key Techniques For Microwave Remote Sensing Soil Moisture Retrieval Under Forest

Soil moisture(SM)is a vital variable in the global land surface water cycle,controlling the exchange of water and heat between land and atmosphere,and thus affecting climate change.Accurate estimates of SM data can be directly applied to forecast and prevent natural disasters such as floods and extreme drought events.In addition,as the input parameter of numerical weather predictions and hydrological models,the accuracy of SM estimations directly influences the reliability of the simulation results of these models.Forest is the largest and most complete terrestrial ecosystem on the earth.It plays an irreplaceable role in the water cycle,carbon cycle,energy cycle and ecological balance,and is the most important maintainer of the earth ecosystem.Changes in SM moisture under forest cover have important feedback effects on the climate change.Therefore,real-time dynamic SM monitoring under forest cover is crucial for the study of forest ecosystem function,global environmental change and carbon source-sink.Forest is a complex system with distinct vertical stratification structure.Although L-band electromagnetic wave can penetrate the canopy,understory(such as shrubs and grass),litter and surface roughness will weaken the sensitivity of brightness temperature to the soil moisture.In addition,the interaction between the microwave and the scatterers with different sizes,shapes and water contents in the forest canopy is also very complex,causing interference to the quantitative decoupling of radiation energy from each layer of the forest.Although the physical model can explain the radiation and scattering mechanism in the forest canopy at L-band,it requires a large number of input parameters and huge computational resources,thus limiting the application of the physical model in the satellite operational retrieval algorithms of surface parameters.Currently,the most widely used microwave transfer model in the satellite operational retrieval algorithms is the parameterization model(such as the?-?model),which has simple form and is applicable for low and sparse vegetations.However,?-?model did not take the multiple scattering effects into account so that the microwave radiation transfer process inside the forest could not be fully explained,which increased the uncertainty of soil moisture retrieval under forest cover,and thus led to the problem that the accuracy of soil moisture estimations under forests obtained by microwave remote sensing could not meet the application requirements.As stated above,the main difficulties in obtaining accurate SM data in forest under long time series and large spatial scale by microwave remote sensing technology are caused by the insufficient understanding of the physical process of electromagnetic wave propagation in forest system,unreasonable parameterization and model state initialization.Based on this background,this study is carried out by using both theoretical model simulations and experimental observations to deepen the understanding of forest microwave radiometric features,to establish a parametric model for forest transmissivity,and to improve the accuracy of SMOS and SMAP SM products by calibrating parameters such as vegetation,roughness and soil dielectric model.The main research contents of this paper are summarized as follows:(1)Analyze the polarization and seasonal variation characteristics of effective transmissivity t and effective albedo?of typical broadleaved forests and needleleaved forests in northeast China based on ground-based L-band microwave radiometer observations at multi-angles over forests in both summer and winter;Verify the effective transmissivity t parameterization method of SMAP based on both L-band airborne and ground-based microwave remote sensing experiments.The results showed that at L-band,the t showed different polarization characteristics for the two kinds of forests.For the broadleaf forests,the measured t at h polarization was higher than that at v polarization.However,the opposite trend was shown for the needleleaf forests.This polarization difference may be attributed to the different structures of the primary branches in the two forest types according to the physical mechanism of interactions between microwave and scatterers with different structures.By comparing the experimental data in winter and summer,it was found that t in winter was significantly higher than that in summer,the difference of t between winter and summer were 0.239and 0.289 for v polarization and h polarization,respectively.As for the value of?,it was independent of polarization,incidence angle,forest type,and there was no obvious seasonal change,that is,it is reasonable to set?as a fixed value in the soil moisture retrieval algorithm.The average value of?is 0.15 for the forest sites in northeat China,which is higher than the 0.06~0.08 value of SMOS and the 0.05 value of SMAP.The value of t obtained by SMAP algorithm are significantly lower than the measurements,which indicates that the parameterization method of t by SMAP is not applicable to forest areas.(2)Establish quantitative parametric equations between sensitive variables and effective transmissivity t at different polarizations for broadleaved and needleleaved forests respectively based on the parametric sensitivity analysis of Tor Vergata model.And the above parametric equations were verified by experimental observations.The results showed that at L-band the branch water content can best describe the variation of forest t,that is,the branch water content is the best variable to parameterize the L-band transmissivity.(3)Validate and improve the SMOS and SMAP SM products over forested areas in northeastern China based on ground wireless sensor network SM measurements.Results found that both SMOS and SMAP SM values were underestimated and not meeting the target accuracy requirement of 0.04 cm~3/cm~3.For SMAP AM and PM SM products,the RMSE values were 0.16 cm~3/cm~3 and 0.17 cm~3/cm~3 respectively,which were better than that of SMOS AM and PM SM products whose RMSE values were0.30 cm~3/cm~3 and 0.31 cm~3/cm~3 respectively.Nevertheless,some improvements in SM retrieval might be achievable through refinements of the soil dielectric model with respect to high percentage of soil organic matter in the forested area.To that end,introducing Liu soil dielectric model into the retrieval algorithms of both SMOS and SMAP missions produced promising results.For SMAP,the RMSE of SM products improved from 0.16 cm~3/cm~3 to 0.07 cm~3/cm~3 for AM data,and from 0.17 cm~3/cm~3 to0.05 cm~3/cm~3 for PM data.For SMOS AM and PM data,the accuracy were improved by 56%and 45%respectively.(4)The influence of the uncertainty of t,?and surface roughness parameter H in Q-H model on the retrieval precision of SM was analyzed quantitatively,and the parameter N_p(p=v,h)in Q-H model is calibrated.The results showed that when t and?were underestimated,the retrieved SM would be overestimated,while when H were underestimated,the retrieved SM would be underestimated.Finally,the influencing factors of passive microwave SM retrieval accuracy were modified to improve the precision of SMAP SM products by taking the polarization and the seasonal variation characteristics of t and?for the typical types of forests in northeast China,the calibration results of surface roughness parameters,and Liu soil dielectric model into account.The RMSE of the revised SMAP AM SM products is 0.04 cm~3/cm~3 which have meet the target accuracy.