Floods are among the most frequent and destructive natural hazards in Kazakhstan, particularly in the Abai Region, Kazakhstan, where topographic, hydrological, and climatic factors strongly influence flood occurrence. This study presents a comprehensive spatial assessment of flood susceptibility in the Abai Region using a multi-criteria Geographic Information System (GIS) approach. The analysis integrates twelve flood-conditioning factors representing hydrological, topographic, environmental, and anthropogenic variables. The relative importance of these factors was determined using the Analytical Hierarchy Process (AHP). The results indicate that distance to rivers (20%) and precipitation (16%) are the most influential drivers of flood susceptibility, followed by Height Above Nearest Drainage (HAND) (11%) and drainage density (9%). The resulting flood susceptibility map classifies the study area into five susceptibility levels. Approximately 56.6% of the study area falls within the moderate susceptibility class, while 25.0% is categorized as high susceptibility, mainly concentrated in low-lying floodplains and foothill regions. Low-susceptibility areas account for 18.1% of the region, whereas the very high and very low susceptibility classes together represent less than 1% of the territory. Model performance was evaluated using Receiver Operating Characteristic (ROC) analysis, yielding an Area Under the Curve (ROC–AUC) value of 0.893, indicating strong agreement between predicted susceptibility patterns and observed flood occurrences. Additional validation metrics derived from the confusion matrix show an overall accuracy of 83.3%, precision of 0.75, recall of 1.0, and a Kappa coefficient of 0.67, confirming reliable predictive performance. Sensitivity analysis with ±10% variation in factor weights further demonstrated the spatial stability of the susceptibility results. The resulting susceptibility map provides an important spatial basis for infrastructure planning, flood mitigation, and disaster preparedness in the Abai Region and offers a transferable framework for flood-susceptibility assessment in other semi-arid regions of Central Asia.

The hydrological prediction in transboundary river basins is difficult because of their heterogeneous data distribution, restrictions of data sovereignty, and the irregular nature of the discharge pattern. In this paper, Federated Learning (FL) with an LSTM neural network is proposed for next-day predictions of discharge in the Chu–Talas transboundary basin. The basin area belongs to both Kazakhstan and Kyrgyzstan. In an FL scenario, two hydrological stations from the basin were selected as client nodes, representing two different discharge regimes. Station 15125—the Chu main channel is characterized by the highest discharge regime among stations located in the basin, while Station 15233—Merke tributary represents a small catchment with an irregular regime. The federated LSTM model is compared against a centralized LSTM and a local-only LSTM baseline model. The training process is based on nearly three decades of daily discharge measurements. The preprocessing step includes synchronization, lag calculation, and windowing operation. The models are trained using three metrics: root mean square error, mean absolute error, and Nash–Sutcliffe Efficiency, as well as using Monte Carlo Dropout for estimation of the probabilistic uncertainty. The results demonstrate that the federated model demonstrates comparable performance with the centralized one for the Chu main channel. It also improves prediction accuracy for the smaller Merke tributary compared with both centralized and local-only models. These findings show that FL can work effectively with non-IID and heterogeneous hydrological data. The study makes three main contributions: (i) it implements the FedAvg algorithm on transboundary, heterogeneous hydrological data, proving that decentralized optimization can effectively capture autoregressive temporal hydrology without data centralization; (ii) it systematically compares federated, centralized, and local-only models, demonstrating that the federated approach eliminates the scale bias that traditionally neglects smaller, high-variance catchments; and (iii) it utilizes Monte Carlo Dropout to translate deterministic AI outputs into risk-aware probabilistic bounds. Ultimately, the results of this study demonstrate the practical and scientific usefulness of FL in operational water management, as the method presents a privacy-saving means of increasing predictive capacity and enabling risk-based decision-making in transboundary river basins.

Managed Aquifer Recharge (MAR) has emerged as an effective approach for enhancing groundwater resources in water-scarce regions under increasing climate variability and growing water demand. However, most GIS-based MAR studies are limited to regional suitability mapping, whereas the hydrogeological feasibility of identified sites is rarely verified under field conditions. This study presents a multi-scale framework integrating regional GIS-based screening with detailed site-scale verification to improve the reliability of MAR suitability assessment in southeastern Kazakhstan. This research was conducted within the Zhetysu Alatau–Balkhash First-Order Hydrogeological Basin to develop and evaluate a multi-scale framework for identifying suitable MAR sites. Regional suitability assessment combined Geographic Information Systems (GIS), the Analytic Hierarchy Process (AHP), Monte Carlo simulation, Fuzzy Gamma analysis, remote sensing, and regional hydrochemical data. Based on the regional assessment, the Byzhy River Basin was selected as a representative site for detailed hydrological, hydrogeological, hydrochemical, LiDAR-based, and field investigations to verify regional suitability under actual hydrogeological conditions. The regional suitability assessment identified 21.82% (58.41 km ² ) of the representative study area as highly suitable for MAR, while Monte Carlo simulation and Fuzzy Gamma analysis confirmed the robustness of the suitability model. Hydrological assessment indicated a positive long-term mean potential water balance of 27.69 mm·year ⁻¹ , suggesting sufficient water resources for MAR. Three double-ring infiltration tests yielded a mean field-saturated hydraulic conductivity of 1.92 m·day ⁻¹ (range: 1.86–1.96 m·day ⁻¹ ), indicating high infiltration capacity of the Quaternary alluvial deposits. Groundwater mineralization remained below 1 g·L ⁻¹ , and hydrogeological investigations confirmed a 50 m thick Quaternary alluvial aquifer suitable for MAR. Long-term daily river discharge records indicated a potential annual recharge volume of 17.44 million m ³ , representing only 2.33% of the effective aquifer storage capacity (746.8 million m ³ ), demonstrating substantial storage potential. The close agreement between the regional suitability assessment and field investigations demonstrates that the multi-scale framework substantially reduces uncertainty in GIS-based MAR site identification. By integrating regional suitability mapping with hydrological, hydrogeological, hydrochemical, LiDAR, and field investigations, the framework provides a transferable methodology for identifying and verifying MAR sites in arid and semi-arid regions, supporting groundwater management and long-term water resource planning.

In data-sparse Central Asian watersheds, machine learning and Geographic Information Systems (GIS) are increasingly applied to hydrological hazard classification. This article compares Random Forest (RF), XGBoost, and Long Short-Term Memory (LSTM) models for daily elevated-discharge classification in the Koksu River basin, Zhetysu Region, Kazakhstan (1614 km², semi-arid, snowmelt- and glacier-influenced climate; 2005–2023), using verified discharge, precipitation, and temperature records from four monitoring stations. The elevated-discharge threshold (80th percentile, ~90.5 m³/s) was computed using only the training period (2005–2017) to avoid temporal data leakage into validation and test periods. Class imbalance, addressed via class weighting, was moderate under this threshold. All models were benchmarked against a naive persistence baseline. Contrary to expectation, the persistence baseline achieved the highest Critical Success Index (CSI = 0.821, 95% block-bootstrap CI [0.725, 0.891]), narrowly ahead of XGBoost (CSI = 0.813, [0.703, 0.897]), Random Forest (CSI = 0.809, [0.691, 0.897]), and LSTM (CSI = 0.763, [0.630, 0.873]); the confidence intervals overlap substantially, indicating no statistically distinguishable advantage of any trained model over simple persistence in this basin. Feature-importance and ablation analyses confirmed that lagged discharge, not precipitation, drove nearly all predictive skill: a discharge-lags-only model performed as well as or better than the complete pipeline, while a meteorology-and-seasonality-only model performed markedly worse (CSI ≈ 0.57–0.59). As a preliminary, exploratory illustration rather than a core result, Gumbel-derived return-period discharges (Q10, Q25, and Q100) were separately translated into illustrative, uncalibrated, DEM-proximity-based inundation extents, a candidate direction for future work rather than an engineering-grade flood-mapping contribution of this study. These results indicate that, in this basin and period, machine learning classifiers provide no clearly demonstrated advantage over simple discharge persistence, underscoring the necessity of routine persistence benchmarking before adopting machine learning approaches for early warning in Central Asian watersheds facing increasing flood risk under climate change.

Climate change is increasing the frequency of hydrological extremes such as floods, requiring more effective approaches to build community preparedness alongside traditional flood risk management. This study develops and evaluates a serious-game-based flood risk education intervention in Kazakhstan and examines whether post-intervention protective behavioral intention can be predicted from gameplay, hydrological-literacy, and governance-attitude data. A cross-sectional, post-intervention survey of 60 participants from flood-, GLOF-, and drought-exposed regions of Kazakhstan was analyzed using descriptive/inferential statistics and three machine-learning classifiers (logistic regression, random forest, multilayer perceptron) to predict protective behavioral intention and to identify its strongest correlates. The participants showed good knowledge about immediate responses in flood events: 91.7% were able to identify the appropriate evacuation procedures, and the mean hydrological-literacy score was 5.13 out of 7. There was relatively less knowledge about longer-term mitigation measures, especially the cost effectiveness of nature-based flood protection (56.7%). Most participants agreed with the introduction of measures for flood management using nature (58.6%) and mandatory disaster insurance (87.7%), and 81.0% indicated that they planned to take preparedness measures after playing. The predictive models with the highest discrimination were logistic regression and random forest, with an area under the receiver operating characteristic curve (AUC) of 0.81 each, while the multilayer perceptron (MLP) had an AUC of 0.75. Importantly, both machine-learning and conventional statistical analyses revealed that perceived scenario fidelity was the strongest and most consistent predictor of protective behavioral intention. These results suggest that serious-game-based flood risk education provides a useful context for examining hydrological knowledge and generating behavioral data for predictive modeling of preparedness. The framework combines experiential learning, flood-governance assessment, and machine-learning-based behavioral prediction, while the findings indicate an association between perceived scenario realism and preparedness-related behavioral intention. These results provide a basis for further investigation of data-driven disaster risk education in Kazakhstan and other settings facing similar hydrological hazard risks.