How uncertainties in data and drought indices affect drought identification
Drought is a slowly evolving phenomenon whose modulating mechanisms stem from complex interactions of atmospheric, land surface and oceanic processes. As drought manifests in different ways, it can be difficult to define.
Ideally, the best way to assess drought on seasonal to annual time scales and with an eye to agricultural impacts would be to base its occurrence on soil moisture measurements. However, direct soil moisture measurements are rarely employed in drought indices because in situ soil moisture observations are sparse in space and time. Therefore, many drought indices have been developed that are based on other variables that can act as a proxy for soil moisture, for example, precipitation or some combination of precipitation and evaporation.
All approaches have distinct differences in how they are calculated, how they are affected by uncertainties in the data being used and ultimately how they represent drought. What has not been examined is whether there is more or less uncertainty in the data using a complex drought index that includes evaporation compared to examining drought using a simpler metric based entirely on precipitation.
CLEX researchers examined the uncertainties of the input data of three commonly used drought indices, with the data coming from different sources, including observations and reanalysis. The ability of these indices to detect drought was assessed against soil moisture from multiple global land surface models.
The results showed that the choice of input data and the uncertainties of each data set changes the skill of the drought index to such an extent that the differences between them for drought identification purposes become small. In many cases, the uncertainty associated with input data when compared to the indices that included evaporation meant the precipitation-only index showed greater skill in identifying drought. The magnitude of the uncertainty from different input data sets varied according to coverage, quality and length of available observations.
Originally published by CLEX, 24 September 2020.
