Integration of Land Surface Modeling, Data Assimilation and Climate Change in Assessing Past and Future Hydroclimatic Conditions over Burkina Faso, West Africa

Abstract

Estimating climate change impacts on water resources in West Africa has been challenged by hydrological data scarcity and inconsistencies in the available climate projections. In this thesis, an integrated approach involving land surface modelling, data assimilation and multi-model ensemble of the most recent regional climate model output is used to simulate the hydroclimatic impacts of climate change over Burkina Faso. To this end, high-resolution simulations from theCO2-responsive versions of the Interactions between Soil, Biosphere, and Atmosphere (ISBA), the global Land Data Assimilation System (LDAS-Monde) and a multi-model ensemble based on the most recent version of the Regional Climate Model (RegCM4) under two Representative Concentration Pathways (RCP4.5 and RCP8.5) are used.ISBA estimates are assessed throughits forcings (ERA5 and ERA-Interim reanalyses) for precipitation and solar radiation variables. First, it is shown that both reanalyses present a good performance in representing precipitation variability and incoming solar radiation (with better score for ERA5). This highlights a good calibration and the potential of ISBA to provide good quality estimates of land surface estimates such as Leaf Area Index (LAI) and Surface Soil Moisture (SSM). Then, within LDAS-Monde,SSMandLAIobservationsfromtheCopernicus Global Land Service (CGLS) are assimilated with a simplified extended Kalman filter (SEKF) using ISBA over a long period (2001-2018). Results of four experiments are then compared: Open-loop simulation (i.e., model run with no assimilation) and analysis (i.e., joint assimilation of SSM and LAI) both forced by either ERA5 or ERA-Interim. After jointly assimilating SSM and LAI, sensitivity study of the model to the observations permits to notice that the assimilation is able to impact soil moisture in the first top soil layers (mainly up the first 20 cm), but also in deeper soil layers (from 20 cm to 60 cm and below), as reflected by the structure of the SEKF Jacobians. The benefit of using ERA5 reanalysis over ERA-Interim when used in LDAS-Monde is highlighted. The assimilation is able to improve the simulation of both SSM and LAI: the analyses add skills to both configurations, indicating the good behaviour of LDAS-Monde. For LAI in particular, the southern region of Burkina Faso (dominated by a Sudan-Guinean climate) highlights a strong impact of the assimilation compared to the other two sub-regions of Burkina Faso (dominated by Sahelian and Sudan-Sahelian climates). In the southern part of the domain, differences between the model and the observations are the largest, prior to any assimilation. These differences are linked to the model failing to represent the behaviour of some specific vegetation species, which are known to produce leaves before the first rains of v the season. The LDAS-Monde analysis is very efficient at compensating for this model weakness. Evapotranspiration estimates from the Global Land Evaporation Amsterdam Model (GLEAM) project as well as upscaled carbon uptake from the FLUXCOM project and sun-induced fluorescence from the Global Ozone Monitoring Experiment-2 (GOME-2) are used in the evaluation process, again demonstrating improvements in the representation of evapotranspiration and gross primary production from assimilation. Finally, the impact of anthropogenic climate change in the hydroclimatology of Burkina Fasofor the middle (2041– 2060) and late (2080–2099)21stcentury has been investigated with regard to the historical period (2001-2018). The results indicate that an increased warming, leading to substantial increase of atmospheric water demand, is projected over all Burkina Faso areas. In addition, mean precipitation unveils contrasting changes with wetter conditions (for all three climatic zones) by the middle of the century and drier conditions during the late twenty-first century(mostly for the Sahelian zone). Such changes cause more/less evapotranspiration and soil moisture respectively during the two future periods. Furthermore, surface runoff shows a tendency toincrease and decrease along with short spatial gradients regardless whether the region receives more or less precipitation. Finally, it is found that while dry and semi-arid conditions develop in the RCP4.5 scenario, generalized arid conditions prevail over the whole Burkina Faso for RCP8.5. It is thus evident that these future climate conditions substantially threaten water resources availability for the country as well asagricultural activities. Therefore, strong strategedies are needed to help design response options to cope with the challenges posed by the projected climate change for the country.