基于FY-3A/VIRR和TERRA/MODIS数据藏北干旱监测对比

为获得藏北地区干旱遥感监测的方法,以及验证FY-3A/VIRR数据在藏北地区干旱监测的可行性,论文基于2015年7月25日—8月4日FY-3A/VIRR和TERRA/MODIS数据,计算NDVI和EVI反演温度植被干旱指数（TVDI）,利用实测20 cm土壤水分数据和气象站点累计降水量验证并对比了两种数据源、不同TVDI的遥感干旱监测的精度。结果表明：1）拟合干湿边时,噪点和拟合像元数影响TVDI监测的精度,去除噪点可提高干湿边拟合精度和干旱监测精度;2）MODIS-TVDI_E（下标E表示EVI反演）、MODIS-TVDI_N（下标N表示NDVI反演）、FY/VIRR-TVDI_E、FY/VIRR-TVDI_N与实测20 cm土壤水分数据和累计降水量的相关性检验都达到了0.05的水平,说明基于两种数据源采用TVDI的方法都能有效监测藏北地区干旱。其中,MODIS-TVDI_E和FY/VIRR-TVDI_E监测精度分别高于MODIS-TVDI_N和FY/VIRR-TVDI_N,表明TVDI_E的监测精度优于TVDI_N。MODIS-TVDI的监测精度高于FY/VIRR-TVDI,反映MODIS数据在藏北地区干旱监测精度优于FY/VIRR数据;3）基于MODIS-TVDI_E和FY/VIRR-TVDI_E划分的藏北地区干旱空间特征结果大体相同,说明FY/VIRR数据也可作为气象部门监测干旱的业务化产品,为藏北地区干旱监测提供数据支撑。

The Comparison of FY-3A/VIRR and TERRA/MODIS Data for Drought Monitoring

Drought is the key factor to restrict the development of agriculture and animal husbandry in northern Tibet. Temperature Vegetation Drought Index (TVDI) is one of the commonly used remote sensing methods for monitoring drought, which couples surface temperature (Ts) and vegetation index (VI ). The TERRA/MODIS L1B, MODIS-LST, FY/VIRR L1B and FY/VIRR-LST, at 1 km spatial resolution, are used for the monitoring and analysis. The monitoring period is from 25 July to 4 August 2015. The space of VI-Ts for the whole study area is typically triangular, from which a linear regression analysis is conducted to get the equations of the dry and wet line. TVDI for northern Tibet is extracted. Then, the measured soil moisture data and cumulative total precipitation data in the same period are used to verify the accuracy of TVDI to monitor drought by comparing TVDI_E (E represents EVI) and TVDI_N (N represents NDVI). The result shows that noise and number of pixels effect monitoring precision, and the precision is better after removing the noise. Small number of fitting pixels will lower their correlation with the equations of the dry and wet line, which affect the accuracy of drought monitoring. There is a significant linear correlation between TVDI and measured soil moisture (P < 0.05), and the coefficients between MODIS-TVDI_E, MODIS-TVDI_N, FY/VIRR-TVDI_E, FY/VIRR-TVDI_N and measured soil moisture were 0.611, 0.581, 0.420 and 0.386 respectively. Correlation between MODIS-TVDI and measured soil moisture is higher than that between FY/VIRR-TVDI and measured soil moisture. The correlation between TVDI_E and measured soil moisture is also higher than that of TVDI_N and measured soil moisture. The coefficients between MODIS-TVDI_E, MODIS-TVDI_N, FY/VIRR-TVDI_E, FY/VIRR-TVDI_N and cumulative total precipitation were 0.370, 0.336, 0.275 and 0.171 respectively (P < 0.05). The correlations are consistent with the correlations between TVDI and measured soil moisture. The result suggests that the TVDI based on MODIS and FY/VIRR data are both feasible for drought monitoring in the study area, and TVDI_E is better than TVDI_N to monitor drought. The monitoring precision of MODIS-TVDI is higher than that of FY/VIRR-TVDI, but FY/VIRR data is also a reliable product for monitoring drought.