Author(s): Sefa Nur Yesilyurt; Eyyup Yildiz; Gulay Onusluel Gul; Huseyin Yildirim Dalkilic
Linked Author(s):
Keywords: No Keywords
Abstract: Data availability is vital for hydrology and water resources engineering studies. Nevertheless, the measurement of hydrological variables is frequently subject to both systematic and random errors, which can result in missing or incomplete records. Among the various hydrological variables, precipitation plays a pivotal role as a primary input in most hydrological analyses. Therefore, the availability of complete and accurate precipitation data is crucial for generating robust and reliable research outcomes. To this end, this paper proposes integrating the KNN (K-Nearest Neighbors Classification) algorithm with the K-Means algorithm. This innovative approach effectively augments precipitation data for the Konya Closed Basin using a method not explored in this geographical context. Using the K-Means method, the model first clusters stations based on location, altitude, annual mean, and maximum and minimum temperature values. It then applies the KNN method for the clustered data sets and completes the missing data. The resulting model shows a performance reaching a maximum value of 0.775. This study's findings confirmed the proposed model's utility in imputing missing data and demonstrated its potential as a viable alternative to traditional imputation methods in hydrological research. The comprehensive data obtained through this innovative model was used to assess the drought status of the basin using the Standard Precipitation Index (SPI) based on the severity, duration, and magnitude of droughts. For the 1-month SPI, the most severe drought occurred in 1988-1998; for the 3-month SPI in 2010-2019; for the 6-month and 12-month SPI in 1966-1976.
Year: 2024