On the methods for impact-based early warning of agricultural drought

<sec><p indent="0mm">Agricultural droughts, causing 9%–10% loss in global crop yield, seriously threaten food security and sustainable socio-economic development at regional and global scales. Based on impact-based early warning, the threats from agricultural drought hazards can be effectively prevented and drought loss can be significantly mitigated, and thus it is a hot and cutting-edge scientific topic presently. In this article, we aimed at the impact-based early warning of agricultural drought, and focused on its three important issues: agricultural drought impact data, agricultural drought indices, and methods for determining agricultural drought thresholds, respectively. Our objective is to provide a comprehensive review of the current research progress and future prospects in this field. </sec><sec> The key to impact-based early warning of agricultural drought is to establish the relationship between actual agricultural drought impacts and agricultural drought indicators. However, low spatiotemporal resolution of existing agricultural drought impact data cannot meet the needs of agricultural drought early warning at local scales, especially when facing diverse agricultural conditions. Thus, it is required to apply high-resolution meteorological, hydrological, and vegetation spectral data, and also generate high-resolution agricultural drought impact dataset through suitable downscaling approaches. A large number of indices have been proposed and developed for characterizing agricultural drought and describing its spatiotemporal variability. </sec><sec> These indices include single variable-based drought indices and multiple variables-based composite drought indices. Among them, vegetation indices, which exhibit a strong correlation with crop yields and are sensitive to water deficit conditions, have much better application potential for impact-based early warning of agricultural drought, than other types of widely used drought indices. Agricultural drought thresholds, linking the critical states of water deficit with actual drought impacts, are crucial for early warning of agricultural droughts. Particularly, the thresholds of vegetation indices can accurately capture the abrupt changes in crop physiological traits caused by water deficit conditions. The absolute threshold method has been widely used for the determination of agricultural drought thresholds. However, it fails to reflect the inherent spatial heterogeneity of vegetation conditions in specific regions, resulting in the biased assessment of agricultural drought impacts as well as its early warning. Alternatively, it would be more reasonable to evaluate the agricultural drought degrees using the “relative threshold method”. The core of relative threshold method is to estimate the relative changes of actual agricultural drought condition compared to its own baseline, based on which different agricultural drought degrees can be reasonably quantified. The relative threshold method enhances the spatial applicability of agricultural drought thresholds, and thus can be a more adaptive approach for the impact-based early warning of agricultural droughts. </sec><sec> Currently, the impact-based early warning of agricultural droughts is still a big research gap, and its development is challenging. In order to promote the research on impact-based early warning of agricultural droughts, we suggest that more reliable agricultural drought impact dataset with high resolution should be generated by applying both statistical and dynamic downscaling methods, and the relative threshold method should be further improved by verifying its reasonableness and applicability. Meanwhile, more efforts should focus on establishing synthetic agricultural drought early warning system by coupling physical models and artificial intelligence methods, and multi-dimensional influencing factors and their cascading impacts on socio-economic developments should be considered together. Then, it can be expected that the uncertainties of agricultural drought forecasting and early warning can be reduced, and agricultural drought hazards can be more effectively prevented. </sec>