Using Remote Sensing Data To Estimate Surface Evapotranspiration Of Plastic Mulched Land-Cover

Surface evapotranspiration(ET)is an important process of water and energy trans-fer in soil-vegetation-atmosphere system,and it is also a key link between land surface ecosystem and hydrological process.It is closely relevant to meteorological conditions,types of underlying surface and vegetation coverage.China is facing a serious shortage of water resources,while 70%the overall water resource are used in agriculture with low efficiency.Therefore,it is very important to promote the efficiency of agricultural water usage.Since plastic film mulching(PFM)can effectively reduce surface ET and enhance water utilization efficiency,it has been widely used in farming.Plastic mulched landcover(PML)has currently become an important agricultural landscape.Such a large-scale PML has an impact on surface energy balance(SEB)and ET.Quantitative estimating the impact of PML on SEB and ET is an important basis for rational deploy-ment and utilization of agricultural water resources.Based on the analysis of the surface energy balance algorithm(SEBAL),the sur-face energy balance system(SEBS)and the two source energy balance(TSEB)models,and PFM's influence on the emissivity of the test soil surface,this study focused on parameter adjustment in the aforementioned models by introducing PFM factor.Taking the oasis area in the middle reaches of the Heihe River as study area,utilizing Landsat images from April to August,2013 and April,2015,contemporaneously hydrological and the meteorological data as well as the retrieved surface parameters as main data source,the latent heat fluxes(LHF)were retrieved from SEBAL,SEBS and TSEB,and their models with parameter adjustment.Comparing LHF of the original models with those of the adjusted models,as well as with those from the observed stations,the va-lidity of the adjusted models were executed.Comparing the accuracies of modeled LHF under different underlying surface conditions,the models' applicability was determined.Combing with land-use types and NDVI values,the differences of LHF of land-use types and their relationship with NDVI were analyzed.The research results indicate:(1)As for Daman farm station,at the early stage of crop growth(NDVI<0.1),the accuracies of HLF from the adjusted models are higher than those from the originals.When the three adjusted models,the SEBAL model inversion is the highest among the three adjusted models,which are on April 16,April 22,April 24,and May 5 and the relative error of inversion values decreased to 3.27%,18.57%,15.32%,and 4.15%,respectively,it indicates that PML has a significant impact on soil moisture evaporation.While at the middle and late stages of crop growth(NDVI?0.1),the accuracies of HLF from the adjusted models are slightly lower than those from the originals because veg-etation transpiration is the main component of evapotranspiration during this period,which shows that PML has weak impact on blocking evaporation and the adjusted model is not suitable for Daman farm stations at the middle and late stages of crop growth.(2)The regression coefficients(R2)between HLF of three models-the adjusted ones for Daman and the original ones for other sites and that of the observed values are 0.8706,0.8115 and 0.8747 respectively,which indicates that the adjusted models are suitable for retrieving ET in PML area.(3)In the case that NDVI is less than 0.15,SEBAL model has obtained the highest inversion accuracy;in the case that NDVI is greater than 0.15,TSEB model has achieved the highest inversion accuracy.LHF accuracy from the SEBS model is always lowest.(4)Daman farm station:the inversion accuracies of three models have unobvious difference during the entire vegetation growth period.The regression coefficients(R2)of the model inversion and actual values are 0.9495,0.9206,and 0.8965,respectively.Bajitan Gobi station:the inversion accuracies of three models are low,the regression coefficients(R2)values of the inversion and actual values are 0.318,0.2574,and 0.5521,respectively.Zhangye wetland station:the regression coefficients(R2)of the model in-version and actual values are 0.8903,0.8204,and 0.9309,respectively.When vegeta-tion coverage is low,the inversion accuracies of three models are relatively low;when the vegetation coverage is high,the inversion accuracy of the three models are well and TSEB model obtains the best inversion accuracy.