FVC Estimation And Its Effects On The Simulation Of Soil Temperature And Sensible/latent Heat Flux Of Noah-MP

Vegetation coverage(FVC)is an important parameter to describe the surface vegetation cover,as well as an important ecological climate parameter.It also affects the water and energy exchange between the ground and the atmosphere and is one of the important input parameters for many land surface models.NDVI(normalized vegetation index)is a widely used data for estimating vegetation coverage.Due to the influence of various factors such as atmospheric conditions,terrain,sensor degradation and other factors,the data of multi-source sensors are inconsistent in time and space,and have poor comparability,which affect the comprehensive application.In order to obtain more accurate simulation results,this study proposed the local kernel regression method based on unsupervised classification to normalize multi-source NDVI products,and then integrate multiple NDVI products to estimate the vegetation coverage in China(NFVC).And it was applied to the Noah-MP land surface model to carry out simulation research on soil temperature,sensible heat flux and latent heat flux,and compared the effects of different vegetation coverage data on model simulation.The main conclusions were as follows:(1)There were large differences between NDVI products.The difference between the METOP AVHRR 10-day composite NDVI data and the MODIS NDVI data(MOD13A3product)was-0.2-0.2.The difference between SPOT VEGETATION/PROBA-V 10-day composite NDVI with MODIS NDVI data was-0.15-0.15.After the normalization of NDVI products by using the local kernel regression method based on unsupervised classification,the difference between the data significantly reduced to-0.05-0.05.In addition,analysis of different time and land cover types showed that the differences among multi-source NDVI data were reduced after normalization.And the comparison with the observation data of typical area showed that the error of the newly produced vegetation coverage data(NFVC)was small and the data quality was good.(2)The simulation results of soil temperature in 2014 of two groups of tests with different vegetation coverage data,CTL?ST(FVC data from the Noah-MP model)and NEW?ST(newly produced NFVC data),showed that both groups of tests could well simulate the changes of soil temperature in one year.The annual mean bias and root mean square error of NEW?ST reduced by 0.41?and 0.16?,respectively.The two groups of tests had poor ability to simulate soil temperature in winter.The improvement in summer was the most obvious,and the bias and root mean square error of NEW?ST reduced by 0.67?and 0.53?,respectively.From the perspective of spatial distribution,the two groups of tests can reflect the characteristic that the annual average soil temperature in China gradually increased from the northwest to the southeast.The simulation results of NEW?ST were higher than CTL?ST about 0.5?,so the NFVC improved simulation of soil temperature.(3)Analysis of the simulated results of sensible heat flux and latent heat flux of the two groups of tests CTL?FLUX(FVC data from the Noah-MP model)and NEW?FLUX(newly produced NFVC data)showed that the two tests could reflect the change characteristics at the observation sites in the Heihe and Qinghai-Tibet Plateau regions.And the simulation of sensible heat flux and latent heat flux was better in autumn,and worse in summer.The simulation of sparse area was better than that of tall and dense vegetation area.These conditions can be improved by using NFVC data.For sensible heat flux,compared with the test CTL?FLUX,the most obvious improvement was found in the Bajitan Gobi station.The annual mean bias and root mean square error of the test NEW?FLUX reduced by 4.70W/m~2and 5.49W/m~2,respectively.In terms of spatial distribution,the sensible heat flux simulated by NEW?FLUX increased by about 1W/m~2 in most areas.For the latent heat flux,the root mean square error of each station was basically winter<autumn<spring<summer.The difference at Naqu Station was the most obvious,with annual mean bias and root mean square error reduced by 3.48W/m~2and 2.90W/m~2,respectively.In terms of spatial distribution,the simulated latent heat flux of the test NEW?FLUX in most areas decreased by 0.5-2W/m~2.