Uncertainties in estimating design droughts

Severity-duration-frequency (SDF) curves of periods of rainfall deficits are useful tools for drought analyses. However, accuracy of these curves are affected by uncertainties associated with a number of factors including: (i) the choice of drought index, (ii) the sampling error due to the limited length of observation data, (iii) the effects of aggregation of data with respect to drought timescales, (iv) the selection of marginal probability distribution functions of drought severity and duration, (v) and the type of copula used to approximate the dependency between severity and duration. In this paper we assess the impact of these uncertainties on estimates of Recurrence Intervals (RI) of drought events by comparing three drought indices (the Standardised Precipitation Index, Reconnaissance Drought Index, and Standardised Precipitation Evaporation Index), three timescales (three, six and twelve months), four marginal probability distribution functions (extreme value, logistic, gamma, lognormal), and two types of copulas (Gumbel and Frank). We assessed all parameterization combinations in relation to their resulting drought RIs of mild, moderate and extreme drought events for 11 sites across Eastern Australia and compared them with a selected baseline value. Across the three selected drought indices there was no difference between the RI derived with Gumbel and Frank copulas for all the sites. For mild drought events, there was no difference between the RIs derived with two drought indices (SPI and SPEI) and distribution functions with the baseline case, whereas SDF curves showed the highest uncertainty with the 12 month time scale. Design droughts or SDF curves are a critical part of any drought analysis for the management of natural systems in regions that can potentially experience water-limited conditions. Explicitly considering the uncertainties involved in developing design droughts is important when assessing the risk of ecosystem failure due to drought events.