Jean Hounkpè

Reliable streamflow forecasting is a fundamental component of flood risk management. However, the accuracy of such forecasts in data-scarce river basins remains one of the pressing challenges in most African catchments. While Long Short-Term Memory (LSTM) networks have demonstrated their effectiveness in rainfall-runoff modelling, their data-intensive requirements limit their direct applicability in poorly gauged catchments across sub-Saharan Africa. To address this issue, we propose a climate similarity-based transfer learning framework using the global Caravan hydrological dataset to improve local streamflow forecasting in Ouémé River Basin (ORB). Specifically, tropical-climate catchments are first identified from Caravan global dataset using the Köppen-Geiger climate classification. A composite climate similarity index (CI) is constructed from two normalized climate indices: annual mean precipitation and seasonality index. 200 basins were selected and ranked by CI. From the 200 ranked basins, a fixed subset of 50 basins was selected using stratified sampling across CI quartiles, including 30 basins for validation and 20 basins for independent testing. The remaining 150 basins were then used to define three cumulative pre-training subsets of increasing size (50, 100, and 150 basins), ranked by CI. Different LSTM models are pre-trained on each subset and fine-tuned on the ORB streamflow. The developed transfer learning models are evaluated against the LSTM model trained exclusively on local data as a baseline model. Pre-training improves prediction at all five stations, raising the median Kling–Gupta Efficiency from 0.65 for the local model to 0.75 for the best transfer configuration. The smallest, most climatically similar donor subset gives the best fine-tuning performance; adding less similar catchments does not help and slightly degrades it. The benefit is largest at stations with short or event-poor records and marginal at the well-gauged basin outlet, while peak flows remain underestimated across all configurations. These results show that, for data-scarce tropical basins, the climatic similarity of the transfer basins matters more than their number, and they point to similarity-guided donor selection combined with peak-weighted or hybrid formulations as a practical route to operational flood forecasting.

Reliable streamflow forecasting is a fundamental component of flood risk management. However, the accuracy of such forecasts in data-scarce river basins remains one of the pressing challenges in most African catchments. While Long Short-Term Memory (LSTM) networks have demonstrated their effectiveness in rainfall-runoff modelling, their data-intensive requirements limit their direct applicability in poorly gauged catchments across sub-Saharan Africa. To address this issue, we propose a climate similarity-based transfer learning framework using the global Caravan hydrological dataset to improve local streamflow forecasting in Ouémé River Basin (ORB). Specifically, tropical-climate catchments are first identified from Caravan global dataset using the Köppen-Geiger climate classification. A composite climate similarity index (CI) is constructed from two normalized climate indices: annual mean precipitation and seasonality index. 200 basins were selected and ranked by CI. From the 200 ranked basins, a fixed subset of 50 basins was selected using stratified sampling across CI quartiles, including 30 basins for validation and 20 basins for independent testing. The remaining 150 basins were then used to define three cumulative pre-training subsets of increasing size (50, 100, and 150 basins), ranked by CI. Different LSTM models are pre-trained on each subset and fine-tuned on the ORB streamflow. The developed transfer learning models are evaluated against the LSTM model trained exclusively on local data as a baseline model. Pre-training improves prediction at all five stations, raising the median Kling–Gupta Efficiency from 0.65 for the local model to 0.75 for the best transfer configuration. The smallest, most climatically similar donor subset gives the best fine-tuning performance; adding less similar catchments does not help and slightly degrades it. The benefit is largest at stations with short or event-poor records and marginal at the well-gauged basin outlet, while peak flows remain underestimated across all configurations. These results show that, for data-scarce tropical basins, the climatic similarity of the transfer basins matters more than their number, and they point to similarity-guided donor selection combined with peak-weighted or hybrid formulations as a practical route to operational flood forecasting.

Reliable streamflow forecasting is a fundamental component of flood risk management. However, the accuracy of such forecasts in data-scarce river basins remains one of the pressing challenges in most African catchments. While Long Short-Term Memory (LSTM) networks have demonstrated their effectiveness in rainfall-runoff modelling, their data-intensive requirements limit their direct applicability in poorly gauged catchments across sub-Saharan Africa. To address this issue, we propose a climate similarity-based transfer learning framework using the global Caravan hydrological dataset to improve local streamflow forecasting in Ouémé River Basin (ORB). Specifically, tropical-climate catchments are first identified from Caravan global dataset using the Köppen-Geiger climate classification. A composite climate similarity index (CI) is constructed from two normalized climate indices: annual mean precipitation and seasonality index. 200 basins were selected and ranked by CI. From the 200 ranked basins, a fixed subset of 50 basins was selected using stratified sampling across CI quartiles, including 30 basins for validation and 20 basins for independent testing. The remaining 150 basins were then used to define three cumulative pre-training subsets of increasing size (50, 100, and 150 basins), ranked by CI. Different LSTM models are pre-trained on each subset and fine-tuned on the ORB streamflow. The developed transfer learning models are evaluated against the LSTM model trained exclusively on local data as a baseline model. Pre-training improves prediction at all five stations, raising the median Kling–Gupta Efficiency from 0.65 for the local model to 0.75 for the best transfer configuration. The smallest, most climatically similar donor subset gives the best fine-tuning performance; adding less similar catchments does not help and slightly degrades it. The benefit is largest at stations with short or event-poor records and marginal at the well-gauged basin outlet, while peak flows remain underestimated across all configurations. These results show that, for data-scarce tropical basins, the climatic similarity of the transfer basins matters more than their number, and they point to similarity-guided donor selection combined with peak-weighted or hybrid formulations as a practical route to operational flood forecasting.

The agroecosystems of the rural continuum of the Sudano-Sahelian region of the Volta River Basin are undergoing severe degradation, resulting in serious declines in ecosystem service capacities and weakening community livelihoods, as agriculture remains the primary source of income in the region. This degradation is mainly driven by poor management of hydroclimatic risks, non-adapted agricultural practices characterized by intensive use of chemical fertilizers and pesticides, and insufficient consideration of spatial heterogeneity in decision-making processes. The objectives of this study are (i) to identify a Network of 30 m resolution Grid Cells (NGC) representative of the spatial heterogeneity of the studied agroecosystem – the Dassari headwater catchment (550 km ² ), a tributary of the Volta River in northern Benin – and (ii) to regionalize plot-scale socioeconomic data to support improved decision-making. The NGC was derived by combining iterative principal component analysis (IPCA) with a Conditioned Latin Hypercube Sampling (CLHS) approach using 37 satellite-derived variables (e.g. Normalized Difference Vegetation Index, saturation index, coloration index), resulting in the selection of 150 grid cells. Field and laboratory investigations provided soil properties (e.g. texture, carbon, field capacity, nitrogen content) and socioeconomic data (e.g. harvest quantity, input costs, total production cost, and gross income), which were standardized and analyzed for major crops (millet, sorghum, maize, and cotton). Results show that natural spatial disparities among NGC cells translate into additional labor and input costs that may be unsustainable for farmers. Multiple linear regression models were developed to relate socioeconomic indicators to soil and remote sensing variables for each crop type, and the resulting regional models proved robust, with predicted and observed values closely aligned within the 95 % confidence interval, coefficients of determination exceeding 70 %, and p -values below 0.01.

Reliable streamflow forecasting is a fundamental component of flood risk management. However, the accuracy of such forecasts in data-scarce river basins remains one of the pressing challenges in most African catchments. While Long Short-Term Memory (LSTM) networks have demonstrated their effectiveness in rainfall-runoff modelling, their data-intensive requirements limit their direct applicability in poorly gauged catchments across sub-Saharan Africa. To address this issue, we propose a climate similarity-based transfer learning framework using the global Caravan hydrological dataset to improve local streamflow forecasting in Ouémé River Basin (ORB). Specifically, tropical-climate catchments are first identified from Caravan global dataset using the Köppen-Geiger climate classification. A composite climate similarity index (CI) is constructed from two normalized climate indices: annual mean precipitation and seasonality index. 200 basins were selected and ranked by CI. From the 200 ranked basins, a fixed subset of 50 basins was selected using stratified sampling across CI quartiles, including 30 basins for validation and 20 basins for independent testing. The remaining 150 basins were then used to define three cumulative pre-training subsets of increasing size (50, 100, and 150 basins), ranked by CI. Different LSTM models are pre-trained on each subset and fine-tuned on the ORB streamflow. The developed transfer learning models are evaluated against the LSTM model trained exclusively on local data as a baseline model. Pre-training improves prediction at all five stations, raising the median Kling–Gupta Efficiency from 0.65 for the local model to 0.75 for the best transfer configuration. The smallest, most climatically similar donor subset gives the best fine-tuning performance; adding less similar catchments does not help and slightly degrades it. The benefit is largest at stations with short or event-poor records and marginal at the well-gauged basin outlet, while peak flows remain underestimated across all configurations. These results show that, for data-scarce tropical basins, the climatic similarity of the transfer basins matters more than their number, and they point to similarity-guided donor selection combined with peak-weighted or hybrid formulations as a practical route to operational flood forecasting.

Reliable streamflow forecasting is a fundamental component of flood risk management. However, the accuracy of such forecasts in data-scarce river basins remains one of the pressing challenges in most African catchments. While Long Short-Term Memory (LSTM) networks have demonstrated their effectiveness in rainfall-runoff modelling, their data-intensive requirements limit their direct applicability in poorly gauged catchments across sub-Saharan Africa. To address this issue, we propose a climate similarity-based transfer learning framework using the global Caravan hydrological dataset to improve local streamflow forecasting in Ouémé River Basin (ORB). Specifically, tropical-climate catchments are first identified from Caravan global dataset using the Köppen-Geiger climate classification. A composite climate similarity index (CI) is constructed from two normalized climate indices: annual mean precipitation and seasonality index. 200 basins were selected and ranked by CI. From the 200 ranked basins, a fixed subset of 50 basins was selected using stratified sampling across CI quartiles, including 30 basins for validation and 20 basins for independent testing. The remaining 150 basins were then used to define three cumulative pre-training subsets of increasing size (50, 100, and 150 basins), ranked by CI. Different LSTM models are pre-trained on each subset and fine-tuned on the ORB streamflow. The developed transfer learning models are evaluated against the LSTM model trained exclusively on local data as a baseline model. Pre-training improves prediction at all five stations, raising the median Kling–Gupta Efficiency from 0.65 for the local model to 0.75 for the best transfer configuration. The smallest, most climatically similar donor subset gives the best fine-tuning performance; adding less similar catchments does not help and slightly degrades it. The benefit is largest at stations with short or event-poor records and marginal at the well-gauged basin outlet, while peak flows remain underestimated across all configurations. These results show that, for data-scarce tropical basins, the climatic similarity of the transfer basins matters more than their number, and they point to similarity-guided donor selection combined with peak-weighted or hybrid formulations as a practical route to operational flood forecasting.

Reliable precipitation forecasts are critical for hydrological modelling and flood early warning in West African river basins, where rainfall is dominated by highly variable monsoon-driven convection. This study evaluates and improves the precipitation forecasting skill of six numerical weather prediction (NWP) models over the Ouémé River basin in Benin, with particular emphasis on lead-time dependence, basin-scale effects, and the added value of statistical bias correction. Daily precipitation forecasts, over the period 1985–2015 across lead times of one to seven days, are assessed across six sub-basins using complementary continuous and event-based verification metrics. The results indicate that precipitation forecast skill varies with model choice, forecast horizon, and spatial scale. Among the raw forecasts, the ECMWF and UK Met Office models consistently outperform the other systems with KGE values reaching 0.5. ECMWF exhibits the highest overall skill at short to medium lead times, while the UK Met Office model shows relatively low volumetric bias across most sub-basins (Pbias less than 25%). For some models, forecast performance improves with increasing basin size, reflecting the smoothing effect of spatial aggregation, although this relationship remains model-specific. Distribution-based methods outperform regression-based approaches, with empirical quantile mapping providing the most robust and consistent improvements across lead times and sub-basins. Following bias correction, Empirical quantile mapping achieved median Likelihood Ratio values of approximately 6 during validation, with upper-range values reaching 15–18 across sub-basins for both ECMWF and UK Met Office forecasts. This represents a substantial improvement over raw predictions whose distributions remained consistently bounded below 10 throughout the calibration and validation phases (more than 50% improvement). Overall, the combination of ECMWF or UK Met Office precipitation forecasts with empirical quantile mapping offers a reliable framework for improving precipitation inputs to hydrological models and flood early warning systems in the Ouémé basin. The findings highlight the importance of multi-criteria evaluation and appropriate bias correction when applying NWP precipitation forecasts in monsoon-influenced hydrological environments and flood forecasting.

Understanding how irrigation water management shapes crop performance is critical for improving productivity and resource-use efficiency in peri-urban agriculture. This study investigated the socio-technical factors driving sprinkler system abandonment and assessed how irrigation water variability influences vegetable yield variability at two market gardening sites (Bogdin and 14 Yaar) in Ouagadougou, Burkina Faso. Survey data from 50 farmers and field measurements of soil properties, irrigation water application, and lettuce yield were analyzed using descriptive statistics, Spearman correlations, and principal component analysis. More than 80% of farmers had ceased using the sprinkler system within two years of installation, 76% reported major equipment failures, and 70% expressed willingness to re-adopt an improved system. Irrigation dose and yield showed considerable variability across sites (CV = 20.9–42.3% and 36.4–44.0%, respectively). At 14 Yaar, irrigation dose was strongly associated with yield (r = 0.862, p = 0.006), indicating that uneven water application was a major constraint on productivity. At Bogdin, where irrigation was more uniform, no single soil or water variable dominated yield variability. Although soil fertility variables contributed to multivariate patterns, nutrient–yield correlations were not statistically significant under the available sample size, and their potential influence on yield requires confirmation with larger datasets. Overall, operational constraints, equipment failures, and inadequate support services contributed to sprinkler system abandonment, while variability in manual water application was associated with variability in crop productivity. These findings highlight the need for irrigation strategies that are both technically robust and adapted to farmers’ realities.

Accurate streamflow forecasting is essential for water resource management, yet remains challenging in data-scarce regions. This study develops and evaluates a hybrid modeling framework for the Ouémé River basin (Benin) that combines the conceptual GR4J hydrological model with machine learning (ML) via a residual-correction strategy. The approach integrates GR4J with three ML techniques Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM) networks to correct systematic errors in simulated streamflow. Models were trained on daily data from 1996–2012 and validated on an independent period (2013–2020) using performance metrics including Kling–Gupta efficiency (KGE), Nash–Sutcliffe efficiency (NSE), and percent bias (PBIAS). Results show that all hybrid models (GR4J–RF, GR4J–XGB, and GR4J–LSTM) outperform the standalone GR4J model. The tree-based hybrids (GR4J–RF and GR4J–XGB) achieved the highest performance gains, with validation KGE values of 0.80 and 0.79, respectively, compared to 0.75 for GR4J alone. While the GR4J–LSTM hybrid also improved baseline results, it exhibited comparatively lower performance, attributed to the data-intensive nature of deep learning in a limited-data context. The study demonstrates that hybrid residual-correction frameworks, particularly those employing efficient tree-based ML, offer a robust and practical pathway for enhancing streamflow prediction in data-scarce catchments.

Understanding hydroclimatic variability and climate teleconnections in a warming world is crucial for drought-prone regions like West Africa, where economies heavily depend on rain-fed agriculture. This study investigates the spatiotemporal dynamics of hydrological and meteorological droughts in the Beninese Part of the Niger River Basin (BPNRB). Using the Standardized Precipitation Index (SPI), the Standardized Precipitation Evapotranspiration Index (SPEI), the Consecutive Dry Days Index (CDD), and the Streamflow Drought Index (SDI), the study assessed drought variability on 3- and 12-month timescales. Statistical methods were employed to explore drought patterns. The findings reveal a recovery phase in the 1990s following severe droughts in the 1970s and 1980s. Significant trends include an increase in CDD values during the rainy season, with an average of 18 dry spell days, and a pronounced upward trend in SDI, indicating intensifying hydrological droughts. These changes suggest growing pressure on both surface and groundwater resources. These findings provide critical guidance for water resource management, promoting efficient irrigation practices and developing preparedness plans. This research offers a robust scientific foundation for designing proactive drought management strategies and enhancing water and food security in Northern Benin.