Ozupek, Ertugrul; Teke, Alihan; Celik, Nuri; Kavzoglu, Taskin

Meteorological drought, exacerbated by global climate change, is a recurrent and complex catastrophe that disrupts water resources, agriculture, and economies worldwide. Forecasting future drought conditions is therefore of utmost importance in facilitating effective intervention and adaptation strategies. Artificial intelligence and machine learning techniques have gained attention owing to their potential in forecasting drought severity. However, they are often criticized for their black-box nature and lack of explainability and interpretability. To address this issue, this study explores the potential of local and global explainable artificial intelligence (XAI) tools, including SHAP and LIME, to extract intrinsic characteristics from different drought indices, including Standardized Precipitation Index (SPI) and Standardized Precipitation Evapotranspiration Index (SPEI) and climatic parameters (temperature, relative humidity, and wind speed). These techniques were applied to understand the driving forces behind drought events under different scenarios. SHAP was used to provide global model explanations, revealing how climatic parameters varied in importance across stations and scenarios, while LIME was employed to offer local explanations for specific predictions, highlighting the role of lag time and the significance of each parameter in individual forecasting scenarios. Using data from four meteorological stations in the Konya province of Türkiye, spanning 1965 to 2023, six ensemble-based machine learning models were employed to forecast SPI and SPEI values at 6-month and 12-month scales, resulting in a total of 768 scenarios. Widely used evaluation metrics, namely MSE, RMSE, MAE, MAPE and were employed to compare model performances. The prediction accuracy ranged from 0.513 to 0.864 values, and CatBoost emerged as the best-performing method in the scenario in which SPEI-12 (t), SPEI-12 (t-1), temperature (T), and wind speed (W) were used as inputs. The global explanation using the SHAP revealed that climatic parameters had varying degrees of impact across stations and scenarios, whereas the influence of lag time was significant in the predictive models. Alongside these considerations, LIME and SHAP methods were applied to investigate the status of the parameters at local and global levels. LIME and SHAP analyses also revealed that T and W parameters played less decisive roles at all stations than SPEI(t) and SPEI(t-1) parameters. The conclusion drawn from the current investigation may offer more explainability when working with machine learning algorithms in drought prediction studies, which could be critical for predicting future drought events.

Achite, Mohammed; Simsek, Oguz; Sankaran, Adarsh; Katipoğlu, Okan Mert; Caloiero, Tommaso

Understanding the relationship between hydrological and meteorological drought in drought-prone regions is critical for proper reservoir management. This study presents a novel multiscale framework for investigating the associations between hydrological and meteorological drought based on the Time-Dependent Intrinsic Correlation (TDIC) method. Firstly, the characteristics of short, medium and long term standardized precipitation index (SPI) and the standardized runoff index (SRI) of the Wadi Mina basin (Algeria) have been analyzed based on data from 6 rainfall and hydrometric stations. Then an Improved Complete Ensemble Empirical Mode Decomposition with adaptive noise (ICEEMDAN) method is used to decompose the most correlated SPI and SRI series to different scales. A stronger association between the two types of droughts is evident in the low-frequency trend component regardless of the station, but their evolution pattern does not remain the same. Subsequently, a TDIC based running correlation analysis is performed between the modes to examine the SPI–SRI associations over the time domain and across the time scales. TDIC analysis has proven the dynamic behavior in the SPI–SRI associations bearing frequent alterations in nature and strength across the process scales and along the time domain. In general, at the intra-annual scales the SPI–SRI correlations are mostly weak positive with localized alterations to negative along the time domain, whereas the relationship is dominantly strong positive and long range at inter-annual scales up to 4 years. This dynamic behavior in the SPI–SRI association and the evolution pattern of trend decipher that the rainfall processes are not directly transferred to streamflow drought, but it also gets controlled by many other local meteorological processes.

Katipoğlu, Okan Mert; Kartal, Veysi; Pande, Chaitanya Baliram

The service life of downstream dams, river hydraulics, waterworks construction, and reservoir management is significantly affected by the amount of sediment load (SL). This study combined models such as the artificial neural network (ANN) algorithm with the Firefly algorithm (FA) and Artificial Bee Colony (ABC) optimization techniques for the estimation of monthly SL values in the Çoruh River in Northeastern Turkey. The estimation of SL values was achieved using inputs of previous SL and streamflow values provided to the models. Various statistical metrics were used to evaluate the accuracy of the established hybrid and stand-alone models. The hybrid model is a novel approach for estimating sediment load based on various input variables. The results of the analysis determined that the ABC-ANN hybrid approach outperformed others in SL estimation. In this study, two combinations, M1 and M2, with different input variables, were used to assess the model's accuracy, and the best-performing model for monthly SL estimation was identified. Two scenarios, Q(t) and Q(t − 1), were coupled with the ABC-ANN algorithm, resulting in a highly effective hybrid approach with the best accuracy results (R2 = 0.90, RMSE = 1406.730, MAE = 769.545, MAPE = 5.861, MBE = − 251.090, Bias Factor = − 4.457, and KGE = 0.737) compared to other models. Furthermore, the utilization of FA and ABC optimization techniques facilitated the optimization of the ANN model parameters. The significant results demonstrated that the optimization and hybrid techniques provided the most effective outcomes in forecasting SL for both combination scenarios. As a result, the prediction outputs achieved higher accuracy than those of a stand-alone ANN model. The findings of this study can provide essential resources to various managers and policymakers for the management of water resources.

Katipoğlu, Okan Mert; Sarıgöl, Metin

Flood routing models are vital in predicting floods and taking all necessary precautions in the region where floods occur, preventing loss of life and property in the region and protecting agricultural areas. This study aims to compare the performance of various machine learning models such as Bagged Tree, Gradient-Boosted Machine, Random Forest, K-Nearest Neighbor, Support Vector Machine and Extreme Gradient Boosting for flood routing prediction models in Ankara, Eskişehir and Sivas. In addition, the predictive success of tree-based algorithms established according to the optimized and default parameters was compared. For this purpose, the flood data of 2013, 2014 and 2015 discharge observation stations located in Ankara D12A242-D12A126, D12A170-D12A172 in Eskişehir and D15A290-E15A035 in Sivas were used. While establishing the machine learning (ML) models, the data was selected as 80% training and 20% testing. Model performances were tested according to various statistical indicators such as root mean square error, mean absolute error and determination coefficient. As a result of the study, the Gradient-Boosted Machine was chosen as the most successful model in estimating flood routing. In addition, the K-nearest neighbor model with 3-nearest neighbor achieved high-level prediction success with the lowest error rates in Ankara. The findings are important in terms of flood management and taking necessary precautions before the flood occurs.

Katipoğlu, Okan Mert; Acar, Reşat

The main purpose of this study is to map the spatio-temporal variation of hydrological drought severities in the Euphrates Basin by using Kriging, Radial Based Function (RBF), Inverse Distance Weighting (IDW), Local Polynomial Interpolation (LPI), and Global Polynomial Interpolation (GPI) methods and to determine the distribution of hydrological drought trends. For this purpose, 18 streamflow observation stations and grid-based ERA5 reanalysis data were employed in and around the Euphrates Basin. The standardized runoff index (SRI) was calculated on monthly, 3-month, and 12-month time scales to determine hydrological droughts. Hydrological drought trends were determined by Mann Kendall (MK) and Modified Mann Kendall (MMK) tests. According to SRI, spatial-temporal distributions of droughts were mapped by Kriging, RBF, IDW, GPI, and LPI methods in the Geographical Information Systems environment. The LPI and Kriging method was selected as the most effective methods to predict hydrological droughts based on the cross-validation error. As a result of the trend analysis, the predominance of decreasing SRI trends in May (61.1%), June (72.2%), July (55.6%), August (61.1%) indicates that the basin is at risk of hydrological drought, especially in summer. According to the maps of spatial droughts, the widespread distribution of droughts in the basin indicated that the basin was greatly affected by hydrological droughts. For this reason, it has been emphasized that transboundary water effective management between riparian countries, planning of water structures, and effective implementation of the drought management plan is of great importance.