Abstract:
Drought is a severe problem in many areas of Sri Lanka, where rainfall amounts are low and extremely high due to climatic changes. To reduce the negative consequences of droughts, it is important to understand the drought characteristics (drought duration and drought severity) and their associations. Therefore, to build a drought severity-duration-frequency (SDF) relationship, a probabilistic technique is proposed using rainfall data from 1996 to 2018 in the two districts Anuradhapura and Puttalam in the dry zone of Sri Lanka. Drought characteristics were defined using 3-month standardized precipitation index (SPI) and Copulas are employed to derive the joint distribution function. Occurrences of 41 droughts from both stations were identified. The derived SDF relationship is a function of marginal distribution functions of drought characteristics linked by a copula. Log-normal distribution and Gamma distribution were identified as the best marginal distribution to represent drought duration and drought severity, respectively, using AIC, BIC and Kolmogorov-Smirnov test. Gaussian copula and Frank copula were identified as the best among the other five copulas based on AIC, BIC and Cramer-Von Mises statistics. Both Log- normal distribution and Gamma distribution along with the Gaussian copula and Frank copula combined to derive the joint distribution of Anuradhapura and Puttalam, respectively. Joint return periods in terms of recurrence intervals were calculated and derived the SDF curves. According to the derived SDF curves, drought severity in Puttalam is greater than those in Anuradhapura for a given recurrence interval and drought duration.
