Mixed-effects modelling of time series NDVI-rainfall relationship for detecting human-induced loss of vegetation cover in drylands

Many researchers have used time-series analysis of remotely sensed images to gain understanding of the dynamics of loss of vegetation cover in drylands. However, complex interactions between vegetation and climate still mask the potential of remote sensing signals to detect human-induced loss of vegetation cover. This paper presents mixed-effect modelling method for time-series NDVI-rainfall relationship to account for the complex interaction between vegetation and climate. Mixed-effects method is a form of statistical modelling that can simultaneously model environmental relationships for a population and for different groups within the population. In this study, it was used to model the NDVI-rainfall relationship in Somalia and for different vegetation types in the country. Its time-series application removed the interaction between vegetation and rainfall and identified areas experiencing human-induced loss of vegetation cover in the country. On average, it gave an accurate relationship between rainfall and NDVI (r2 > 60%) and detected areas with human-induced loss of vegetation cover (kappa = 75%). Although the potential of mixed-effects was shown using vegetation types, other factors such as soil types and land use can also be included in the method to improve accuracy of time-series NDVI images in detecting human-induced loss of vegetation cover in the drylands.