Comparison of One-Site vs Multi-Sites calibration/validation schemes for hydrological modelling of nested watersheds in the West African Sahel

Abstract

<p><strong>Abstract:</strong></p><p>Runoff simulation in highly anthropized watersheds is complex, but essential for water management, especially in poorly gauged and data-scarce hydrosystems of the West African Sahel. In this study, the physically-based and semi-distributed hydrological model SWAT (Soil and Water Assessment Tool) was applied to simulate daily runoff in the Nakanbé watershed at Wayen (21,178 km<sup>2</sup>) in Burkina Faso over the period 2006-2012. The data used concerned 9 water reservoirs, 5 hydrometric stations, 15 rain gauges, 3 synoptic stations, a digital elevation model (spatial resolution: 30m), a soil map (scale: 1:5,000,000), a land use and land cover map (scale: 1:200,000). The 5 hydrometric stations (Tougou, Dombré, Rambo, Ramsa and Wayen) gauge 5 nested watersheds ranging in size from 38 to 21,178 Km<sup>2</sup>. The added value of nested watersheds was assessed through the following three calibration/validation (C/V) schemes: one-site C/V (OS) at the watershed outlet (Wayen); multi-sites C/V considering nested sub-basins (MS1) and multi-sites C/V without considering nested sub-basins (MS2). The results at the watershed outlet (Wayen) indicated that the three C/V schemes (OS, MS1, MS2) are overall satisfactory (KGE above 0.7). However, MS2 scheme (KGE = 0.75) performed better as compared to the traditional one-site calibration scheme (OS: KGE = 0.73). On the other hand, the MS1 scheme (KGE =0.71) performed worse than the OS scheme. The water balance analysis at Wayen showed that 60 to 90% of the annual rainfall is lost as evapotranspiration whereas 5 to 8% contributes to surface runoff, groundwater recharge remaining very low. The comparison of the three calibration/validation schemes provides evidence that accounting for nested sub-basins does not necessarily improve the quality of rainfall-runoff simulations. Yet, multi-sites modeling remains the best scheme when the watersheds used are not nested. The availability of higher resolution data might allow a better understanding of spatial scale effects.</p><p><strong>Keywords: </strong></p><p>Nested watersheds, SWAT model, Nakanbé, West Africa Sahel, Burkina Faso.</p>