More Than Meets the Eye: Using Sentinel-2 to Map Small Plantations in Complex Forest Landscapes

Keiko Nomura,Edward Mitchard

doi:10.3390/rs10111693

Abstract

Many tropical forest landscapes are now complex mosaics of intact forests, recovering forests, tree crops, agroforestry, pasture, and crops. The small patch size of each land cover type contributes to making them difficult to separate using satellite remote sensing data. We used Sentinel-2 data to conduct supervised classifications covering seven classes, including oil palm, rubber, and betel nut plantations in Southern Myanmar, based on an extensive training dataset derived from expert interpretation of WorldView-3 and UAV data. We used a Random Forest classifier with all 13 Sentinel-2 bands, as well as vegetation and texture indices, over an area of 13,330 ha. The median overall accuracy of 1000 iterations was >95% (95.5%–96.0%) against independent test data, even though the tree crop classes appear visually very similar at a 20 m resolution. We conclude that the Sentinel-2 data, which are freely available with very frequent (five day) revisits, are able to differentiate these similar tree crop types. We suspect that this is due to the large number of spectral bands in Sentinel-2 data, indicating great potential for the wider application of Sentinel-2 data for the classification of small land parcels without needing to resort to object-based classification of higher resolution data.

Highlights

Land use change in the tropics has a significant impact on the carbon cycle, and global climate change, but it is poorly quantified [1,2]
While there are sufficient data on deforestation provided by systematic and free-to-use remote sensing [3], what happens to land after deforestation varies by location [4] and there are no global products providing these data, making local classification of the resulting land use necessary for both carbon accounting and policy implementation purposes
Using the reference samples from high resolution imagery as training data for a Random Forest classifier with 30 trees and four prediction variables, Sentinel-2 data were able to classify both areas at overall accuracy rates of 95% and higher for all the four images (Figure 6 and Table 6)

Summary

Introduction

Land use change in the tropics has a significant impact on the carbon cycle, and global climate change, but it is poorly quantified [1,2]. In mitigating climate change through conserving and enhancing forest carbon stocks, monitoring the changes in land cover and land use provides crucial information for policy development and enforcement in areas such as forest conservation, watershed, and environmental protection [2]. There are a number of ways in which the area of different land cover and land use types within an area, and how they are changing, can be assessed. These range from agricultural census surveys to various types of remote sensing. In Asia, this change has been especially pronounced, with the average size of agricultural holdings falling from 2.5 hectares in 1950 to one hectare in 2000, where the fragmentation of holdings driven by population growth is Remote Sens. 2018, 10, 1693; doi:10.3390/rs10111693 www.mdpi.com/journal/remotesensing

Results

Discussion

Conclusion