Meteorological Drought - Universal Monitoring and reliable seasonal Prediction with the Standardized Precipitation Index

Abstract

Drought is arguably the most complex and least-understood natural hazard. Its understanding is obscured by irreconcilable spatiotemporal monitoring across different model realizations and observational datasets. This obscurity and our generally limited understanding adversely affect our ability to predict this hazard’s probability of occurrence. While promising developments show potential improvements for both of these shortcomings, further progress through novel approaches are still in urgent need. This dissertation addresses both shortcomings by reconciling drought monitoring across the dimensions mentioned above and demonstrating reliable skill of dynamical seasonal drought predictions at unprecedented lead times.

The emergence of standardized drought indices revolutionized drought monitoring. Their advantages reside in their probability-based interpretability and application-based flexibility. In contrast, their disadvantages concern deficits in their robustness, extendability, and tractability. A calculation algorithm that universally standardizes highly non-normally distributed precipitation time series would rectify these deficits for the most widely used drought index – the Standardized Precipitation Index (SPI). However, such a calculation algorithm proved elusive in the past because the abundance of involved dimensions seemed irreconcilable. This dissertation presents a computation algorithm that universally standardizes the index across space, time, and different realizations. The results demonstrate that the exponentiated Weibull distribution excels in the standardization of the index. Particularly notable is that this finding establishes the theoretical basis for the SPI to be applied to simulations.

This basis formally allows the evaluation of dynamical SPI predictions on seasonal timescales. On seasonal timescales, drought predictions need to merge multiple sources of information to be skillful. Previous investigations show significant drought hindcast skill up to one lead month by merging predicted and observed precipitation. In contrast, this dissertation merges the dynamical prediction with information about the observed state of the El Niño-Southern Oscillation (ENSO). In this process, the results illustrate the conditional drought hindcast skill during active ENSO years. When an active ENSO state is present at the start of the prediction in October, this investigation reveals significant and reliable winter drought hindcast skill up to lead month in equatorial South- and southern North America. Further, the area of reliable hindcast skill is largest when an active ENSO state is already present in the preceding summer. Particularly beneficial is that the analysis discloses this skill during the dry phase of ENSO. Additionally, by using ENSO as a second source of information (instead of observed precipitation), the methodology decouples the lead time of reliable predictions from SPI’s accumulation period. This decoupling enables the present methodology to demonstrate reliable skill at unprecedented lead times.

Universally monitoring and reliably predicting the SPI increase the lead time of valuable information essential for managing the risks of drought impacts. Additionally, this dissertation’s findings carry the potential to extend our general understanding of drought by dissipating obscurities that surround its early detection and timely prediction.