基于NDVI分区的全国草地生态系统覆盖区土壤水分遥感反演

Research for Remote Sensing Retrieval of Soil Moisture in Grassland Ecosystem in China Based on NDVI Partition

为了建立土壤水分遥感反演模型,反演2016年全国草地生态系统土壤水分并分析其时空变化特征,通过结合表观热惯量(ATI)和温度植被干旱指数(TVDI)的混合模型反演2016年全国草地生态系统的土壤水分,并在实测数据与反演结果精度验证的基础上确定归一化植被指数(NDVI)阈值。结果表明:(1)NDVI≤0.2的像元区域,采用ATI模型反演精度较高;NDVI≥0.78的像元上,基于增强型植被指数(EVI)的TVDI反演精度较高;0.2<NDVI<0.78的区域,基于NDVI的TVDI反演效果较好。(2)全国草地10 cm土壤水分存在明显时空差异。空间分布上由西北向东南增加,最低值(0~10%)出现在青海、西藏、新疆和内蒙古,最高值分布在南方各省。时间分布上,全年草地10 cm土壤水分集中在0~40%。春冬季节,土壤水分较少(0~60%),夏季最高,大部分像元集中在30%~80%,秋季次之,集中在0~70%。建立的混合模型反演的土壤水分精度相对较高,以期为不同植被覆盖情况的土壤水分遥感反演模型选择提供参考,对于准确监控大面积范围土壤水分将有重要的实用价值,也是目前研究中迫切需要解决的问题所在。

The soil moisture of grassland ecosystem in China in 2016 was retrieved by using hybrid model with the combination of the apparent thermal inertia(ATI)and temperature vegetation drought index(TVDI),and the NDVI threshold was determined based on the accuracy verification of the measured data and inversion results.The remote sensing inversion model of soil moisture was established to retrieve the soil moisture of grassland ecosystem in China in 2016 and analyze its spatial-temporal variation characteristics.The results show that:(1)for NDVI≤0.2 pixels,the inversion accuracy of ATI is higher;for NDVI≥0.78 pixels,the inversion accuracy of EVI based TVDI is higher;0.2<NDVI<0.78 pixels,the inversion effect of NDVI based TVDI is better;(2)the soil moisture in 10 cm layer of grassland in China was significantly different on temporal and spatial scale;the lowest value(0~10%)appeared in Qinghai,Tibet,Xinjiang and Inner Mongolia,and the highest value distributed in southern provinces;in terms of temporal distribution,the soil moisture in 10 cm layer of grassland concentrated in 0~40%;in spring and winter,the soil moisture content is less(0~60%),the highest was observed in summer,most of the pixels concentrated in 30%~80%,followed by autumn,the pixels concentrated in 0~70%.The precision of soil moisture retrieved by the established hybrid model is relatively high in this paper,we provide reference for the selection of soil moisture remote sensing retrieval model under different vegetation cover,will have important practical value for accurately monitoring soil moisture in a large area.The precision of soil moisture retrieval is an urgent problem to be solved in the current research.