Global Drought Hazard Monitoring in Rainfed Areas

Droughts are a significant threat to the agricultural sector in general, and rainfed farming in particular. Therefore, effective and timely responses to manage droughts and their impacts are required so that farming systems can limit the negative effects of droughts on food production. We developed a crop drought index (CDI) by integrating drought hazard and exposure and applied this index at the global scale to evaluate the influence of drought on the exposed rainfed areas for different crops. In an attempt to develop an operational, multisectoral global drought hazard forecasting system, we computed and analyzed CDI for historical periods. We further used bias-corrected seasonal climate forecasts to project the drought development in a 7-month period. The CDI was calculated by using the Global Crop Water Model (GCWM) at a global extent (5 arc-minute resolution) from 1980 to 2020. We compared the drought conditions in specific years to the CDI in the 30-year reference period 1986 to 2015. The CDI was computed for 25 specific crops or crop groups based on the relative deviation of the ratio between actual evapotranspiration (ETa) and potential evapotranspiration (ETp) in a specific year from the long-term mean ratio of ETa/ETp during the crop growing season. To test the skill of the seasonal drought forecasts, CDI computed with bias-corrected ensemble forecasts was compared to simulations with standard ERA5-reanalysis data for the year 2018 when severe drought conditions were observed across Europe and other regions. The skill of the CDI to detect drought impacts was tested for historical years by comparing the time series of the harvested area weighted CDI to detrended yield anomalies for crops and countries with predominantly rainfed production. The results of the comparison with historical yield anomalies showed that the CDI is a good indicator for negative yield anomalies, in particular in regions known to be affected regularly by droughts. The model simulations employing the bias-corrected ensemble forecasts reproduced well the reference drought condition in the year 2018 in countries such as Argentina, Australia, Italy, and Spain but showed little skill to reproduce the severe drought in Western Europe. Data availability constraints also had an impact on the accuracy of historical reconstructions and forecasts. For instance, the hazard and exposure analysis rely on static input data for crop shares and crop calendars, which can impact the results (i.e., as cropping patterns are dynamic and often can change over time). The findings suggest that bias-corrected seasonal ensemble forecasts have a significant potential to enhance seasonal drought forecasts, although the skill of the forecasts varies considerably for specific regions. Further research is needed to analyze this potential across different periods and geographies systematically to increase forecasting system efficiency and minimize processing time before this system can be run operationally. In our study, we hence want to demonstrate the current status of the CDI-based forecasting system and discuss the potential, limitations, and uncertainties of such CDI forecasts for agricultural applications.