A Century of Data Is Not Enough to Establish Reliable Drought Thresholds
To project long-term future hydrologic drought conditions, drought researchers customarily use statistics collected over a historical reference period of 100 years or less to establish a threshold for declaring when a region is—or is not—experiencing a drought. This study demonstrates the statistical uncertainty in estimating a threshold from these small datasets is sufficient to compromise many types of drought analysis. The study develops formulae for calculating the statistical uncertainties caused by limited record lengths and for estimating the record length needed to achieve a specified level of threshold accuracy. Results show that hydrologic datasets of 100 years are approximately one-tenth the length needed to achieve the level of drought threshold reliability required for many applications.
Researchers and practitioners in the hydrology community are generally aware of the effects of limited datasets on estimating thresholds, but issues surrounding threshold uncertainty still are not well-quantified or acknowledged in the scientific literature. This study raises the hydrology community’s awareness of this issue by assessing threshold uncertainties, identifying record lengths required to achieve high reliability, and summarizing options for augmenting the historical record to increase threshold reliability.
To evaluate long-term changes in future hydrologic drought conditions, drought researchers customarily use hydrologic data collected over a historical reference period to establish a threshold for declaring when a region is experiencing a drought. The length of the historical record is often no longer than 50 to 100 years. This study shows the statistical uncertainty in threshold estimates resulting from these limited datasets is sufficient to compromise many types of drought analysis. The researchers provide formulae for calculating the statistical uncertainties caused by limited record lengths and for estimating the record length needed to achieve a specified level of accuracy in a given drought analysis. Results show that datasets of 100 years (or less) are approximately one-tenth (or less) the length needed to achieve the level of reliability required for many drought applications. Despite the widespread use of thresholds in drought research, this threshold uncertainty issue has not previously been acknowledged in the hydrology literature. Finally, the study summarizes options for augmenting the historical record to improve threshold accuracy.