Extraction and Spatiotemporal Analysis of Rubber Plantations on Hainan Island, China (1990 – 2023) Using Phenological and Multivariate Remote Sensing Features

Rubber is an economically significant crop, yet large–scale monitoring of its distribution remains challenging due to the limitations of traditional field surveys. Remote sensing combined with machine learning offers a scalable solution, though research integrating phenological traits and multivariate data remains limited. This study focuses on Hainan Island, the main rubber–producing region in China, and comprehensively incorporates the phenological characteristics of rubber tree growth. Leveraging Landsat time–series imagery from 1990 to 2023, we integrated spectral bands, vegetation indices, texture metrics, and terrain variables to construct classification models based on Random Forest (RF), Gradient Boosting Tree (GBT), and Support Vector Machine (SVM) models. The results show that, the RF achieves the highest accuracy, with overall accuracy exceeding 90% and Kappa coefficients above 0.81. Rubber plantation area expanded by 37.60%, from 3942.27 km² in 1990 to 5424.53 km² in 2023, with the most rapid growth between 2005 and 2010. Spatially, the distribution of rubber plantations on Hainan Island has shown a general trend of migration from the south to the northwest over the past three decades. The area of rubber plantations in the southern regions has gradually decreased, which is closely related to the frequent typhoons and the cultivation of other high–income economic crops in the region. In contrast, the northwest region has seen gradual expansion, driven by the relatively low impact of typhoons, economic incentives stemming from high international rubber price fluctuations during critical planting periods, and adjustments in industrial layout guided by local government policies. The findings provide valuable scientific and technological support for the management of rubber production and the optimization of planting patterns on Hainan Island.