Analysis of 10-m Sentinel-2 imagery and a re-normalization approach reveals a declining trend in the latest rubber plantations in Xishuangbanna

Abstract

Natural rubber is one of the four main raw materials of industry and a globally important strategic material. Rubber plantations with agroforestry attributes have undergone rapid growth in the northern part of mainland Southeast Asia (including Xishuangbanna) since the 1990s, leading to a series of eco-environmental effects on biological diversity, climate change, carbon stocks, and hydrological processes. Accurate, detailed, and updated spatial information on rubber plantations, therefore, is fundamental for developing efficient management strategies. Currently, the booming rubber plantations are being experienced new changes in Xishuangbanna. The commonly used coarser spatial resolution satellite data (e.g., 250-m MODIS and/or 30-m Landsat) have limited applications because most plantations are often small and scattered, especially in the mountainous areas. Here, we developed a straightforward and effective re-normalization of red green normalized difference vegetation index (ReNDI) approach based on finer resolution (10-m) Sentinel-2 imagery, and then mapped the first annual 10-m rubber plantations in the entire Xishuangbanna during 2018–2021. Interestingly, rubber plots no longer expanded as rapidly as in past decades and even decreased slightly. The latest area of rubber plantations was 2514.7 km2, a decrease of 31.9 km2 comparing since 2018, with an average overall accuracy and kappa coefficient of four-year reached up to 95.37 % and 0.90, respectively. Among them, Jinghong city and Menghai county increased by 52.4 km2 and 14.8 km2 respectively, and Mengla county decreased by 99.1 km2. The main area of rubber expansion was around the Jinghong city, where rubber was initially planted. More importantly, owing to pursuing increased economic benefits and requirements of environmental protection, a considerable portion of the rubber encroachment has disappeared, particularly near roads and rivers. Our phenology-based ReNDI algorithm not only enriches the remotely-sensed methods for other industrial plantations mapping, but also provides a new chance to understanding the current patterns of rubber plantations (i.e., increase and decrease), which will contributed to rational planting planning and agroforestry management in the future, especially in the tropics.