Sensitivity of Flash Drought Prediction Based on Inter- and Intra-Annual Modes of Variability to Drought Index and Region

Flash droughts rapidly evolve over the course of a few weeks to months, as opposed to traditional droughts which may develop over the course of many months to years. This poses an enhanced challenge to drought planning by reducing the time decision makers have in order to respond through the implementation of mitigation and adaptation strategies. Large scale modes of variability such as the El Niño- Southern Oscillation (ENSO) and Madden-Julian Oscillation (MJO) have been shown to influence temperature and precipitation patterns, which may result in extreme rainfall or drying episodes when they are in constructive or destructive phases. We evaluate the predictability of rapid drought changes such as flash drought or flash amelioration by using machine learning algorithms to link measures of rapid drought change to phases of the ENSO and MJO that enhance or diminish precipitation and temperature patterns over the continental United States, at a variety of lead times from two to five weeks. Both the U.S. Drought Monitor and the Standardized Precipitation Index are used to define periods of rapid drought change. These indices have different input variables and lag periods between the first signs of developing drought and the identification of drought by the index, allowing for the investigation of model sensitivity to the choice of drought index. The spatial variation in predictability of flash drought is further evaluated by NOAA climate division to determine if there are regions where the model may perform better due to a stronger regional influence from large-scale patterns of variability.