Groundwater recharge and geospatial heterogeneity by hydrological modelling under climatic influence: the case of the Davo River watershed (south-east of Côte d'Ivoire)

 Corresponding author: Kouakou Noumh Dickens ATCHEREMI Copyright © 2025 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution License 4.0. Groundwater recharge and geospatial heterogeneity by hydrological modelling under climatic influence: the case of the Davo River watershed (south-east of Côte d'Ivoire) Kouakou Noumh Dickens ATCHEREMI 1, *, Gnangui Christian Rodrigue ADON 2, Kossou Yves Bertrand BEUGRE 1, Mahaman Bachir SALEY 2, Kan Jean KOUAME 2, Yao Muller YAO 3 and Patrice Jean Roger JOURDA 2 1 Department of Coastal Sciences, Marine Sciences Training and Research Unit, University of San Pedro, BP 1800 San Pedro, Côte d'Ivoire. 2 University Center for Research and Application in Remote Sensing (CURAT), Félix Houphouët-Boigny University, Abidjan, Côte d'Ivoire. 3 Department of Geospatial Science and Technology, Computer Science and Digital Sciences Training and Research Unit, Virtual University of Côte d'Ivoire, Abidjan, Côte d'Ivoire. World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 Publication history: Received on 12 September 2025; revised on 22 October 2025; accepted on 25 October 2025 Article DOI: https://doi.org/10.30574/wjarr.2025.28.1.3587 Abstract The sustainable management of groundwater resources in Côte d'Ivoire requires a reliable estimate of groundwater recharge. This study proposes a methodology for assessing the recharge of the Davo watershed (South-West Côte d'Ivoire) through distributed hydrological modelling with the HYDROTEL software. Physiographic (DEM, soils, land cover) and hydrometeorological (rainfall, discharge, evapotranspiration) data were integrated into the model via PHYSITEL. The three-layer vertical water budget (BV3C) was used to estimate vertical flows and potential aquifer recharge. The results indicate an average annual recharge of 221 mm, with spatial variability between –27 and 644 mm/year depending on the hydrological units. The performance of the model is considered satisfactory (Nash-Sutcliffe = 0.62), volume deviation = -1.11%). The sensitivity analysis shows the major role of evapotranspiration and land use on recharge, confirming the impact of deforestation and enthronization on the reduction of infiltration. This integrated approach provides a robust scientific basis for the quantitative assessment of groundwater resources and can be replicated in other similar contexts when all the required conditions are met. Keywords: Recharge; Hydrological Modelling; Geospatial Heterogeneity; Davo River; Côte d'Ivoire 1. Introduction Groundwater resources are now a major strategic issue for sustainable development. In Côte d'Ivoire, this issue has a critical dimension: the constant increase in groundwater withdrawals, combined with climatic uncertainties, requires rigorous and scientifically based management of groundwater bodies ([1], [2]). According to the National Guide for the Assessment of the Quantitative Status of Groundwater Bodies, the assessment of the quantitative status of these resources is based on a "Balance" test which compares the withdrawals made to the natural recharge of aquifers, thus determining whether a body of water is in "good status" or "bad status" ([3]). This regulatory approach immediately raises a fundamental scientific question: how can we reliably estimate groundwater recharge? Historically, this issue has crystallized two distinct disciplinary approaches that have long evolved in parallel ([4]). On the one hand, surface hydrologists have developed models focused on the prediction of river flows, often considering groundwater flows as a loss for their study system. On the other hand, hydrogeologists have focused on groundwater dynamics, paying secondary attention to surface processes. This historical disciplinary World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 1858 dichotomy has gradually revealed its limitations, highlighting the need for an integrated approach to understanding the complexity of recharge processes. Faced with the multiplicity of estimation methods available in the scientific literature, a rigorous methodological selection is necessary to identify the approaches best suited to the geographical and operational constraints of the Ivorian context. The selection criteria favour methods applicable to significant spatial units (several tens of km²), capable of providing multi-year average estimates, allowing spatial aggregation from the scale of the catchment area to that of the water body, usable with limited parameters and easily accessible data, and finally economically viable for large-scale application. The comparative analysis of the three main methodological families identified by ([5]): approaches based on flow balances, those using hydrograph analysis to determine baseflow, and those based on the analysis of piezometric fluctuations, reveals that distributed hydrological modelling is a particularly promising approach. According to these authors, surface water and unsaturated zone methods, such as surface hydrological models, estimate the potential recharge of water percolating from the surface layers to the water table, while those based on groundwater data generally provide estimates of actual recharge. In this methodological context, the choice of the HYDROTEL software for the estimation of recharge is based on a convergence of scientific and technical arguments. As a distributed hydrological model, HYDROTEL meets the design requirements identified by [6] - system geometry, inputs, process formalization laws, initial state and boundary conditions, and outputs - while satisfying the stated selection criteria. According to [7], this distributed approach takes into account the spatial variability of hydrological processes, input variables and watershed characteristics. [8] identify two fundamental advantages of this approach: it can better capture spatial variability or heterogeneity and allows the consequences of scaling up to be studied. The robustness of HYDROTEL has been demonstrated by its validation in various geographical contexts: initially validated on the Chaudière River in Quebec ([7], [9]), the model was then successfully tested in other Canadian provinces, in France on the Orb basin ([10]), in southern Quebec ([11]), and in Vietnam ([12]). Closer to our study area, HYDROTEL has proven itself in Côte d'Ivoire with successful applications in the N'zo ([1]), Bandama River ([13]) and N'zi ([4]) watersheds, demonstrating its adaptability to West African hydro-climatic conditions. Another decisive advantage is HYDROTEL's modular architecture and its capacity is particularly important for recharge estimation, as it allows the precise quantification of the different components of the water cycle ([14]). The proposed methodological project thus constitutes a model for the quantitative assessment of groundwater resources, the approaches of which can be replicated in similar geographical contexts. The aim is to define the conditions for the success of a reliable estimate of recharge and to establish a reference methodological framework contributing to the sustainable management of Ivorian groundwater bodies. 2. Materials and methods 2.1. Study area The Davo River is a tributary of the Cassandra River. It is located on the east side of this river which is located in the southwest of Côte d'Ivoire. The Davo River watershed covers an area of 7,194 km². It lies between longitudes 6°47 W and 5°69 W and latitudes 6°85 N and 5°03 N (Figure 1). World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 1859 Figure 1 Geographic location and locations of the study area With a total length of about 566.1 km, the Davo River crosses, from north to south, the various localities of Sinfra, Ouragahio, Guiberoua, Gagnoa, Gueyo, Dakpadou and Kokolopazo. This watershed has an elongated shape. 2.2. Materials The hydrological modelling was conducted with the HYDROTEL system (INRS, Canada), including PHYSITEL 3.0 for the preparation of physiographic data (DEM, soils, land cover, hydrographic network) and HYDROTEL 2.6 for the distributed and continuous simulation of flows ([7], [9], [15]). Input data included an SRTM DEM (90 m, NASA), daily rainfall series (1961-2016, 7 stations, SODEXAM), observed flows at 3 stations (ONEP), a soil map ([16]), as well as Landsat7 ETM+ (2000) and Landsat8 OLI/TIRS (2014) satellite images for land cover. Specific parameters (soil hydraulic properties, leaf area index, root depth) have been integrated into PHYSITEL. Geospatial processing mobilized ENVI 4.3, MAPINFO 8.0/Vertical Mapper 3.0, ARCGIS 10.0, ERDAS, PCI and Google Earth. The coherence and stationarity of the hydroclimatic series were verified using KHRONOSTAT 1.01 (IRD) and the EVC programme ([17]). Together, this enabled reliable distributed modelling of the Davo basin and an integrated estimation of water recharge. World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 1860 2.3. Methods The estimation of aquifer recharge was achieved by the implementation of the distributed hydrological model HYDROTEL ([9]), specifically using the conceptual sub-model of the Three-Layer Vertical Balance (3LVB/BV3C). This approach allows for the annual and monthly hydrological balance of precipitation, surface runoff, actual evapotranspiration (REE) and potential aquifer recharge. The choice of this configuration is based on its prior validation in similar hydro-climatic contexts in West Africa ([1], [4], [13]). The implementation of this approach first requires a phase, the constitution of the physiographic database, which involves: • Preparation of input data; • The spatial structuring of the watershed. Then, a hydraulic parameterization phase that takes into account the hydraulic properties of the soils such as porosity, hydraulic conductivity, field capacity, wilting point. Another phase follows the previous one and concerns the configuration of the HYDROTEL model. This is done through the choice of algorithms for calculating the model, which is based on feedback from previous studies in tropical environments ([1], [4], [9], [10]). The algorithms selected for each sub-model are • Meteorological interpolation: Weighted average of the three closest stations; • Potential evapotranspiration: Hydro-Québec algorithm based solely on maximum and minimum temperatures; • Vertical water balance: three-layer vertical balance (3LVB/BV3C); • Earth flow: Kinematic wave; • River flow: Kinematic wave. At the BV3C sub-model, the soil is divided vertically into three functional layers ([9] Fortin et al., 1995): • Layer 1 (10-20 cm): controls evaporation and surface runoff; • Layer 2 (transition): manages delayed flows in the upper part of the soil; • Layer 3 (deep): controls the base flow rate and recharge, kept close to saturation. Soil discrimination makes it possible to obtain the variation in water content of each layer, which is modelled by continuity equations, taking into account vertical flows between layers, lateral flows and evapotranspiration losses. The flow Q₃ of the deep layer, assimilated to the potential recharge, is modeled as a function of the water content θ₃, the thickness of the layer and a recession coefficient kr. Thus, a continuous simulation over 5 years (1991-1995) was carried out. It made it possible to: • the establishment of the overall hydrological balance of the catchment area; • the estimation of the spatial distribution of the potential recharge; • the daily quantification of the balance sheet terms for each Relatively Homogeneous Hydrological Unit (RHHU/UHRH). The potential recharge is directly provided by the q₂,₃ flux of the BV3C submodel, representing the transfer of water from the intermediate layer to the deep layer, assimilated to the recharge of underground reservoirs. This integrated methodology allows for a spatially distributed and temporally continuous estimation of potential recharge, which is essential for the assessment of the quantitative status of groundwater bodies. World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 1861 3. Results 3.1. Modelled physiographic characteristics of the Davo River watershed (DRW/BVRD) The automatic delimitation (PHYSITEL) indicates an area of 6,943 km² for 857,161 pixels (90 m on each side), very close to the topographic reference (6,959 km²), with a marginal underestimation of 0.23% leading to a faithful respect for the morphology of the basin (Figure 2). The dominance of occupancy classes varies between UHRHs, generating differentiated responses in runoff, infiltration, evapotranspiration and interception. The overall hydrodynamics of the basin result from the aggregation of these elementary processes within the 223 UHRHs (Figure 3). Figure 2 Actual contour of the basin (in dark black) imposed on the modeled contour Figure 3 Occurrence of land cover classes within each UHRH of the Davo basin The hydrological model is based on three hierarchical sub-basins, defined according to the available hydrometric stations, resulting from the strategic grouping of the 223 UHRHs. This approach allowed a gradual calibration of the model, from upstream to downstream (Figure 4). World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 1862 Figure 4 DRW model dunnage sub-basins 3.2. Performance of the HYDROTEL model The evaluation of the performance of the HYDROTEL model (Figures 5 to 7) was carried out in three stages: calibration (1992-1993), validation (1994-1995) and simulation over the entire period (1991-1995). Figure 5 Hydrographs observed and simulated over the calibration period (1992-1993) Figure 6 Hydrographs observed and simulated over the validation period (1994-1995) World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 1863 Figure 7 Hydrographs observed and simulated over the period of cross-simulation calibration/Validation (19911995) The results show good model fidelity, with Nash-Sutcliffe (NS) coefficients generally above 0.60 and moderate volume deviations (VE) (Figure 8). Overall, the results indicate a good agreement between the simulated and observed flows, both for calibration and for validation and overall simulation. Figure 8 Relationship between observed and simulated flows The quality of the models is considered acceptable to good by the stations to demonstrate the robustness of the model in representing the hydrological variability of the Davo River basin (Figure 9). The scatter plot shows the correlation between the two series; The dashed line is the perfect fit (1:1), while the shaded area delineates the confidence interval associated with linear regression. World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 1864 Figure 9 Performance of the HYDROTEL model by station in the Davo basin 3.3. Hydrological balance of the DRW (BVRD) The results of the annual balance sheet recorded in Table I present the interannual variations in the elements of the water balance. Rainfall varies from 1328 mm (year 1991) to 991 mm (year 1992) with an average of 1193 mm/year. The average runoff is 40 mm/year, represents 3.39% of the rainfall and varies between 45 mm (year 1995) and 35 mm (year 1992), which leads to a flow deficit of 1153 mm/year, representing 96.61% of the rainfall, which will be distributed between the REE and the recharge of the underground reservoirs. The ETR varies from 1286 mm (year 1991) to 985 mm (year 1992), with an average of 1137 mm/year, i.e. 95.25% of rainfall. As for the values of the potential recharge of aquifers, they range from 282 mm (year 1995) to 167 mm (year 1991), with an average of 221 mm/year, which is equivalent to 18.52% of precipitation. Table 1 Values of the annual components (mm) of the water balance based on the BV3C sub-model (1991-1995) 1991 1992 1993 1994 1995 Avg. Rain 1236 991 1187 1224 1328 1193 Flow 42 35 39 42 45 40 Flow deficit 1194 957 1148 1182 1283 1153 ETR 1286 985 1061 1175 1176 1137 Potential Charging 167 172 239 242 282 221 The results of the monthly report illustrated in Table II reflect the particularity of each month during the hydrological year in the Davo River watershed. Heavy rains occur over two periods; from March (139 mm) to June (149 mm) with a peak in May (184 mm) and then in October (152 mm). These heavy rains have led to waves flowing over these same periods varying from 4 mm (March and April) to 5 mm (June and October) through the peak of 6 mm (Month of May) on the catchment area. When the REE values are lower than those of rainfall, aquifer recharge begins in March (20 mm) and ends in October (47 mm) with an estimated peak in May of 67 mm. On the other hand, when the REE values are higher than the rainfall, it is found that the underground reservoirs support the baseflow of the hydrographic network. This translates into the negative values of the potential recharge that are observed from November to February. From 1991-1995, the various results of the annual and monthly hydrological balance, obtained using the BV3C submodel, show that the balance is in surplus over the entire Davo River watershed, with a recharge period of the underground reservoirs that extends from March to October. As HYDROTEL is a distributed model, it is possible to spatialize the overall results of the balance. Thus, a spatialized assessment of the potential recharge of aquifers made it possible to locate areas with high recharge rates for a more efficient management of groundwater resources. World Journal of Advanced Research and Reviews, 2025, 28(01), 1857-1870 1865 Table 2 Monthly component values (mm) of the water balance based on the BV3C sub-model (1991-1995) Jan. Feb. March Apr May June Jul August Seven. Oct. Nov. Dec. Rain 10 61 139 145 184 145 71 85 97 152 63 33 Flow 0 2 4 4 6 5 2 2 3 5 2 1 Flow deficit 10 59 135 140 178 141 69 83 94 148 61 32 ETR 92 57 85 102 93 97 79 54 85 114 135 124 Potential Charging -30 -1 20 33 57 47 10 21 22 47 -14 -16 3.4. Recharge estimation and uncertainties The Monte Carlo simulation (Figure 10) made it possible to quantify the uncertainty associated with the estimation of recharge. The resulting distribution reveals a median of 140.8 mm/year, with a P5-P95 confidence interval of between 112 and 172.7 mm/year and a coefficient of variation of 13.4%, reflecting moderate variability. Figure 10 Distribution of uncertainty on potential recharge (Monte Carlo simulation) 3.5. Sensitivity of the Nash-Sutcliffe criterion to the evapotranspiration coefficient The observation in Figure 11 highlights the importance of the parameterization of the evapotranspiration coefficient (ETP) in the performance of the model. The Nash-Sutcliffe criterion varies almost parabolicly according to this parameter, with an optimum around intermediate value (NS ≈ 0.62-0.65). This behaviour reflects the strong influence of atmospheric flows on the water balance, of which evapotranspiration is one of the major factors conditioning recharge.