Improving the Temperature Vegetation Dryness Index (TVDI) and clarifying the agricultural drought grade threshold of
TVDI is of great significance for improving the ability of TVDI to monitor agricultural drought. Based on MODIS (Moderate Resolution Imaging Spectroradiometer) remote sensing data in the past 19 years, several feature spaces are constructed by using the single-time and multi-time methods, including NDVI (Normalized Difference Vegetation Index) -LST (Land Surface Temperature), EVI (Enhanced Vegetation Index) -LST, RVI (Ratio Vegetation Index) - LST, and SAVI (Soil Adjusted Vegetation Index) -LST. The calculation methods of TVDI are discussed, the applicability of TVDI for agricultural drought monitoring in Gansu Province is analyzed, and classification standards for summer TVDI agricultural drought in Gansu Province are clarified. The research results are as follows: 1) The TVDI calculated from the SAVI-LST feature space is more suitable for agricultural drought monitoring in Gansu Province. The root mean squared error (RMSE) and mean absolute error (MAE) of its fitting relative soil moisture (RSM) decreased by 1%–5% compared with the RMSE and MAE of RSM fitted by NDVI-LST feature space TVDI for RSM, which is used more commonly. 2) TVDI is suitable for agricultural drought monitoring at shallow depths of 10 and 20 cm in non-arid areas such as semi-arid, semi-humid and humid areas in
Gansu Province in summer. The RMSE and MAE are approximately 15.6% and 12.6%, and the fitting errors in humid areas are the least, and they are less in semi-humid areas than in semi-arid areas they are the largest. 3) Compared to TVDI drought grades divided by 0.2 intervals and TVDI with uncertain classification criteria , the TVDI agricultural drought grade determined by the linear relationship between TVDI and RSM is more conducive to improving the accuracy of TVDI monitoring agricultural drought.
