Evaluating variogram models and kriging approaches for analyzing spatial trends in precipitation simulations from global climate models

Climate change has heightened the irregularity and unpredictability of weather patterns, influencing precipitation patterns. Accurate geographical projections of precipitation and other climatic variables are critical to sustainable water resource management and disaster preparedness. Variogram models are geostatistical techniques used to examine spatial correlation. Therefore, selecting the optimum variogram model for spatial interpolation is challenging. This study used six variogram models to assess spatial trends. Leave-one-out cross-validation (LOOCV) and K-fold cross-validation approaches are used to find the best variogram model based on metrics such as mean absolute error (MAE), root mean square error (RMSE), and mean bias. In this study, correlation data of 22 GCMs within observed data are predicted over 94 locations in Pakistan from 1950 to 2014. For evaluation, ordinary kriging (OK) and universal kriging (UK) are utilized as geostatistical approaches. The study highlights the suitability of the variogram models. Pentaspherical variogram (Pen) model is suggested as suitable model due to its minimum error metrics as well as the Hol effect (Hol) model has been considered beneficial for dealing with complicated data. From the geostatistical approaches, ordinary kriging (OK) yields the best prediction. Moreover, ordinary kriging (OK) and universal kriging (UK) both yield similar outcomes across some correlation-based data of 22 GCMs within observed data. Consequently, the implication of correlation analysis, optimum variogram models, and interpolation techniques enables the precise and accurate approach in the prediction of GCM performance. The efficiency of variogram models and interpolation approaches in managing data variability helps to enhance the consistency and interpretability of climate data.