Parallel investigations of remote sensing and ground-truth Lake Chad's level data using statistical and machine learning methods

Lake Chad is facing critical situations since the 1960s due to the effects of climate change and anthropogenic activities. The statistical analyses of remote sensing climate variables (i.e., evapotranspiration, specific humidity, soil temperature, air temperature, precipitation, soil moisture) and remote sensing and ground-truth lake level applied to the period 1993–2012 reveal that remote sensing data has a skewed distribution; ground-truth data has a symmetrical distribution. Linear Regression (LR), Support Vector Regression (SVR), Regression Tree (RT), Random Forest Regression (RF), and Deep Learning (DL) methods show that (i) RF and LR, with the highest R² and EVS and least MAE, MSE, $\text{RMSE}$ and, ${\text{CV}}_{\text{MSE}}$ values seem the best models to further investigate remote sensing and ground-truth lake level data and (ii) the remote sensing data based models outperform the ground-truth data based models based on their $\text{MAE}$, $\text{MSE}$, $\text{RMSE}$, and ${\text{CV}}_{\text{MSE}}$ values. The most useful variables to predict lake level are precipitation and air temperature. The data analysis methodology reported here is of fundamental importance for the perspectives of an integrated and forward-looking water management system for connecting climate change, vulnerability, and human activities in the Lake Chad human-environment system. Corroboration studies are needed when more ground-truth data eventually are obtainable.