Using deep learning to predict future threats to Africa's protected areas from an expansion of tea production

Abstract

Agricultural expansion is a growing threat to biodiversity in Africa's protected areas, yet its extent and future dynamics remain unquantified. Tea cultivation, a major global cash crop and key livelihood for many low-income countries, has expanded rapidly across the continent, including into biodiversity hotspots. Here, we develop a state-of-the-art deep learning model, trained with a novel dataset across the African continent and reveal that 12% of Africa's tea plantations are located within designated protected areas in their totality, irrespective of protection categories, which are associated with lower biodiversity intactness index values by up to 18% compared to protected areas. For the future, we predict overall declines in suitable tea growing areas and a further expansion into protected areas, with an estimated negative impact on the biodiversity. Because of the model's unprecedented high spatial resolution, we are able to provide highly precise spatial information where tea production currently threatens biodiversity in Africa's protected areas and where this will most likely occur in the future. We predict the highest threat for Mt. Kenya (Kenya), Gebre Dima (Ethiopia), and Rwenzori (Uganda) informing the need for targeted conservation policy recommendations.