Improvement and applicability of the soil moisture based agricultural drought index

Agricultural droughts have frequently occurred in the arid and semi-arid areas in China. Frequent drought and delayed sowing have also posed a huge impact on crop yield and economic losses in the tens of billions of every year. There is a high demand to clarify the uneven annual distribution of precipitation. In this study, the translation invariance method was used to improve the applicability of the soil moisture agricultural drought index (SMADI), in order to accurately monitor the occurrence of agricultural drought for the sustainable development of the social economy. Then, the drought grading criteria of the modified soil moisture agricultural drought index(SMADI_M) were defined by comprehensive analysis of the objective drought index. The reliability of SMADI_M was finally verified from the regional, and temporal scales, as well as the timeliness of drought response. The results show that the SMADI_M changed significantly the situation as the overestimation of drought in the low vegetation coverage areas, indicating a wider scope than the SMADI. The correlation between SMADI_M and soil moisture data of 0~10cm depth was −0.77 (P<0.01). A better correlation of SMADI_M was achieved with the damaged area, affected area and crop failure area. Specifically, the SMADI_M correlation with the crop failure area was much better than SPEI (0.43) and VCI (0.07). Therefore, the SMADI_M can be expected to accurately measure the major droughts in the typical regions. For instance, moderate and severe drought accounted for 49.53%, and 26.58%, respectively in Zhangye City in 2001. The mild and moderate drought were accounted for 26.47%, and 74.44%, respectively, in Xilinguole League in 2018. Meanwhile, the seasonal drought was ranked in descending order in the whole year: spring, autumn, winter and summer drought. The percentage of occurrence of no, mild, moderate, severe and extreme drought in the arid and semi-arid regions during spring were 7.82%, 19.91%, 32.78%, and 25.99%, respectively, while in summer were 20.69%, 26.73%, 20.78%, 18.84%, and 12.98%, respectively. Among them, the correlation coefficients between SMADI_M and VCI that lagged by 0, 30 and 60 d in spring were 0.60, 0.69 and 0.70, respectively. Once the agricultural drought occurs in spring, the SMADI_M can be expected to capture the agricultural drought up to 30 d ahead of VCI. The relationship between VCI and SMADI_M that lagged by 30d was marginally significant in summer and winter. In autumn, the correlation coefficient between VCI and SMADI_M with a lag of 60d was significantly higher than that with no-lag and a lag of 30 d. It infers that the SMADI_M captured the drought information up to 60 d earlier than VCI in autumn drought. In winter, the lagged 60 d correlation coefficient relationship between SMADI_M and SPEI was 0.65, which was higher than that no-lag and lagged 30 d , indicating the more sensitive SMADI_M to agricultural droughts. Consequently, the SMADI_M was more sensitive to the agricultural drought, where the response was 30-60 d earlier than that of VCI and SPEI. The finding can also provide a new approach to accurately monitor the agriculture drought.