Missing data and drought analysis in the susurluk basin

Drought is one of the most complex and destructive natural disasters. Drought effects can be minimized through sustainable management of water resources. Reliable results from hydrological analyses should be obtained for optimal water resources management. Reliable results are closely related to the completeness of the data inventory to be used in hydrological and statistical analyses. This study consists of two main sections: missing data and drought analysis in the Susurluk Basin. In the first section of the study, it is aimed to determine the most appropriate imputation method for the statistical structure of the basin by developing a missing data methodology for the hydrology literature. In the second section, it is aimed to determine the drought risk status of the basin in detail with traditional and innovative methods. In the missing data methodology, simulated data sets are created by considering the amount of missing data, missing data pattern and missing data mechanisms of real data sets. Simple imputation techniques, Expectation Maximization (EM) and k-Nearest Neighbor (kNN) algorithm are used to estimate missing data in simulated data sets. The most appropriate missing data imputation method for simulated data sets that are similar to the statistical structure of real data sets is determined according to different evaluation criteria. In this way, the most effective method is used to estimate missing data in real data sets. Within the scope of the study, the effect of station selection and normality assumption on the performance of the EM is examined with various scenarios. Missing values in the data are completed in accordance with the specified methodology. Before conducting drought analyses, inconsistencies and breaks in the data are evaluated with homogeneity tests. Then, Standard Precipitation Index (SPI), Standard Precipitation Evapotranspiration Index (SPEI), Z-Score Index (ZSI) and Actual Precipitation Index (API) are calculated for short (1-3 months), medium (6-9 months) and long (12-24 months) time scales and extreme droughts are determined. The serial dependencies of drought index values are examined and the nonparametric Modified Mann-Kendall (mMK) test, the Innovative Trend Analysis (ITA) method for graphical trend analysis and the Innovative Trend Significance Test (ITST) are used to determine drought trends. In the study, it is determined that EM is the most effective method to complete the missing data, and the generalized extreme value (GEV) distribution represents the basin rainfall better than the Gamma distribution.