Innovative drought analysis via groundwater information

Drought hazard has complicated features related to climatic and spatio-temporal characteristics, making it challenging to accurately identify and track. Contemporary approaches to drought monitoring generally use standardized drought indices due to their practical utility. Despite the availability of a various array of drought indices, their application introduces complexities in data mining and decision-making processes, potentially resulting in confused outcomes. However, this research developed a new hybrid drought index Multivariate Cluster Ensemble Drought Evaluation Index (MCEDEI) based on machine learning technique cluster analysis using groundwater data of the KB region of T & uuml;rkiye to assess the groundwater drought. For the development of MCEDEI, this study used 540-time series observations (range: 1978-2022) of groundwater data from five stations to evaluate drought characteristics. Furthermore, this study used steady-state probability to determine the trend and long-term probabilities of the drought index in the KB region of T & uuml;rkiye. The results show that the NN (near normal) class was found to be dominant with a probability of 70.41% on a 1-month time scale, while NN was found to be dominant with a high probability of 65.94% on a 3-month time scale. The probability of the NN class was found to be equally high when the time scale was extended to 6, 9 and even 48 months. MD (moderate drought) remains important, and SD (severe drought) increases compared to SW (severe wet) classes. Findings shpw that there are significant changes in groundwater behaviour at different time scales. Short-term stability is characterized by the dominance of the NN class, while long-term scales show a trend towards extreme dry and wet conditions with a decrease in neutrality. As a result, T & uuml;rkiye may face drought challenges in the future based on the findings.