Temperature\|vegetation index feature space, which couples information of land surface temperature (Ts) and vegetation, is an important method for soil moisture estimation and agricultural drought monitoring. In this paper, Ts and enhanced vegetation index (EVI), which derived from AQUA\| MODIS (Moderate Resolution Imaging Spectroradiometer) data, were used to build triangular Ts\|EVI feature space. How to determine better parameters in dry edge and wet edge equations was discussed and temperature vegetation drought index (TVDI) was calculated. Relations between TVDI and relative soil moisture (RSM) in different depths were analyzed and the abilities of soil moisture estimation of TVDI were also compared. Results showed that: the traditional method which used the biggest inflection point as the starting one to fit linear regression equations was not perfect to get parameters of dry edge and wet edge. However, according to frequency distribution of different pixel value, the endpoint approximatioss method, which retained the maximum effective information and keep higher fitting precision, had strong ability to get better parameters. TVDI got from Ts\|EVI feature space could estimate RSM in soil surface with the depth of 10, 20cm and 50cm, and the correlations between RSM and TVDI passed t\|test at significance level α = 0.001, but the accuracy of soil moisture estimation by TVDI for the depths of 20cm and 10cm were relatively higher.