Comparison of different interpolation methods in the spatial distribution of hydrological data: A case study of Konya closed basin

Today, global warming and climate change events, which affect the whole world, have started to have a serious impact on water resources. For this reason, studies on climate and hydrological issues are of great importance. Hydrological models are frequently used to understand the functioning of the hydrological process and hydrological structure in the basins, to determine the effect of climate change on water resources, to complete the missing data and to esimate the data in ungauged basin, to protect and plan the water resources, to manage the water policies correctly. Knowing the spatial distribution of hydrological data is very important in terms of protecting, planning and managing water resources. In this thesis, the applicability of deterministic and geostatistical interpolation methods in realizing the spatial distribution of hydrological data in Konya Closed Basin (KKH) was investigated. The data sets used in the study were obtained from the General Directorate of Meteorology (MGM) and the General Directorate of State Hydraulic Works (DSI). The data consists of monthly total precipitation and monthly average temperature values of 11 meteorological observation stations between 1971 and 2019 and 34 meteorological observation stations between 2014 and 2019, and water level measurements of 39 wells between 2002 and 2019. While precipitation data and well water level data were used as raw data in interpolation methods, temperature data were used in potential evapotranspiration (PET) calculation with Oudin method. In spatial estimations, Inverse Distance-Weighted interpolation method (IDW), Regularized Spline (Sp-R) and Tension Spline (Sp-T) method were used from deterministic methods, while Ordinary Kriging (OK) and Universal Kriging (UK) methods from geostatistical methods were used. In the OK method, Spherical (OK-K), Gaussian (OK-G), Circular (OK-D) and Exponential (OK-E), which are the most frequently used in the literature, were used as the semivariogram method. The accuracy and success of the prediction values obtained as a result of deterministic and geostatistical methods were evaluated with the performance criteria of mean absolute error (MAE), root mean square error (RMSE), coefficient of determination (R²) and Nash-Sutcliffe efficiency coefficient (NSE). When the performance of the most successful interpolation models of the 3 test stations for monthly total precipitation between 1971 and 2019 was evaluated, the highest prediction success was obtained at Cihanbeyli station, and the lowest prediction success was obtained at Seydişehir station according to the NSE metric. For Cihanbeyli station, the Sp-T precipitation model was the most successful with R²=0.741, NSE=0.721, MAE=9.11 mm and RMSE=12.42 mm values. For the Seydişehir station, the Sp-R precipitation model was the most successful with R²=0.762, NSE=0.561, MAE=27.78 mm and RMSE=42.55 mm. OK-G was the most successful interpolation method at Karapınar station, which was the other test station in the 1971-2019 period. At this station, R²=0.725, NSE=0.704, MAE=8.24 mm and RMSE=11.55 mm values were determined by OK-G method. When the performance of precipitation models between 2014-2019 is evaluated; Among the test stations, the station with the highest prediction success was Seydişehir station, while Sultanhanı station showed the lowest prediction success. The IDW method for the Seydişehir station was the most successful model in the monthly total precipitation estimation, with the values of R²=0.848, NSE=0.843, MAE=15.29 mm and RMSE=22.52 mm. For Sultanhanı station, R²=0.576, NSE=0.533, MAE=9.30 mm and RMSE=15.61 mm were obtained in OK-G method. Considering the performance of the interpolation models used in PET estimation was evaluated, it was determined that the NSE values of the most successful interpolation methods at the test stations were above the values of 0.996 and 0.986 in the 1971-2019 and 2014-2019 periods, respectively, and a "very good" level of success was achieved. When the estimation performance of the interpolation methods was evaluated in the estimation of the groundwater level (YASS) for the 2002-2019 period, successful results were obtained in 4 wells, while the methods used in the estimation of YASS in the remaining wells were unsuccessful. The highest estimation success among test wells was obtained with OK-G interpolation method in 52267 (13312) observation wells. R²=0.915, NSE=0.896, MAE=0.031 m and RMSE=2.457 m were obtained by OK-G method in 52267 (13312) observation well.