Hydrological modelling of Sahelian hydrological paradox: accounting for explicit land use/land cover change in the simulation of hydrological processes

Land use/land cover (LULC) change is a major factor affecting the hydrological response at the watershed scale. However, hydrological modelling, in its current practice, is usually carried using a single and static LULC layer for simulation runs over long periods. Eventually, this approach leads to failure in accounting for LULC spatial and temporal changes as well as non-linear impacts on simulated outputs. Besides, in the typical case of Sahelian hydrosystems, previous modelling attempts based on this approach failed at reproducing the well-known Sahelian hydrological paradox which occurred in the area during the period 1970-1990. This study aims at assessing the added value of dynamical integration of LULC changes in hydrological modelling of surface runoff in Sahelian hydrosystems. The Tougou watershed (37 km²), located in Northern Burkina Faso is selected as a case study. LULC maps of the watershed are produced from 1952 to 2017 from the processing of Landsat satellite images. The SWAT (Soil & Water Assessment Tool) model, using the SCS-CN method (for surface runoff estimation), is calibrated and validated using observed runoff data collected over the period 2004-2018. The calibration/validation is carried using LULC maps of the watershed in 1999, 2009 and 2017, dynamically integrated into the model using a specific land use update module. Further, the calibrated model parameters helped in the reconstitution of surface runoff over the historical period 1952-2005 and integrating dynamically LULC maps in 1952, 1973, 1986 and 1999. The results showed that between the periods 1952-1968 (P1) and 1986-2005 (P3), the average annual rainfall decreased by 87.9 mm while paradoxically, average annual runoff increased by 1 mm. Further analysis revealed that the increase in runoff is mainly attributed to LULC changes (+647%) which offsets the effect of the decrease in rainfall (-547%). From the analysis of LULC maps, it was found that from P1 to P3 periods, the decrease in natural vegetation (CN = 67.3 ± 5.7) by 40%, replaced by bare and degraded soils (CN = 83.8 ± 2.5) explained the observed increase in surface runoff potential of the watershed, as shown by their calibrated CN values. These findings are reminiscent of the Sahelian hydrological paradox reported in the literature and provide evidence of the sensitivity of surface runoff to LULC changes. Overall, the results call to hydrologists, water resources planners and managers, regarding the advantages of coupling LULC changes in hydrological modelling. Also, the study advocates for the development of integrated modelling platforms integrating both LULC changes and hydrological modelling to allow a better understanding and the more accurate long-term forecasting of water resources, in particular in the case of Sahelian hydrosystems.

Keywords: Dynamic LULC input, Hydrological modelling, Surface runoff, SWAT model, Burkina Faso, Sahelian paradox.