Obtaining rubber plantation age information from very dense Landsat TM & ETM plus time series data and pixel-based image compositing

摘要

Determining the age of rubber tree plantations (Hevea brasiliensis) is of great interest to plantation managers and land-use decision makers as it enables, among others, reliable forecasts of resource availability. The acquisition of age information with field campaigns, however, is time consuming, laborious and expensive. We present an approach that facilitates rapid assessment of rubber plantation age at regional scales applying very dense Landsat times series (LTS) satellite data over China's second largest rubber planting area - Xishuangbanna. We aggregated 270 Landsat TM and ETM + surface reflectance images into annual best-available-pixel composites of minimum Normalized Difference Moisture Index (NDMI). The annual composites were classified into vegetated and non-vegetated pixels applying a global NDMI threshold of 0. As it is common practice in Xishuangbanna to clear the land before a new plantation is established, the last year in which a pixel located in a rubber plantation, was classified as non-vegetated, was recorded as the year of plantation establishment. The resultant plantation age map was validated using a stratified random sample of 184 data points, collected by visual interpretation of historical Landsat data. A Root Mean Square Error (RMSE) of 2.5 years was achieved. We estimate that in 2015 50% of Xishuangbanna's rubber plantations had an age suitable for latex tapping (8-25 years), 27% were too young to be tapped (<8 years) and 24% had already reached an age of reduced latex productivity (>25 years) and will be presumably harvested for wood in the near future. The proposed technique enables acquisitions of accurate rubber plantation age information from Landsat time series and NDMI over large areas with the potential of providing insights into land cover dynamics and allowing for the development of improved management strategies and local decision support activities. (C) 2017 Elsevier Inc. All rights reserved.