Journal of Oil Palm Research Vol. 35 (3) September 2023, p. 517-527 NARISSARA NUTHAMMACHOT1* and DIMITRIS STRATOULIAS2
Received: 25 January 2022 Accepted: 8 August 2022 Published Online: 19 October 2022
The cultivation of oil palm (Elaeis guineensis Jacq.) trees is one of the most important agricultural activities and a major sector of economic development in Thailand. However, oil palm trees are susceptible to diseases that can decrease the profitability of the business. Decreasing productivity sometimes triggers an expansion of the cultivated area, which is often negatively affecting surrounding natural habitats. Remote sensing technology has increasingly been used for investigating, detecting and mapping plant related traits. This study aims to use concurrently acquired Sentinel-2 satellite imagery, Unmanned Aerial Vehicle (UAV) field survey and ground observation data to identify the characteristics of oil palm trees based on three controlled sites (namely healthy, diseased and mixed oil palm tree areas). The GNDVI, NDVI, NDI45, RVI, MSAVI and MTCI vegetation indices (VI) were used as a predictor of plant biomass and indicator of oil palm tree disturbance. A linear regression model was applied to each of the derived VIs to determine the index with the strongest relationship to biomass for each of the three sites. The outcome of this study showed; (1) that the most effective indicators were NDVI for the healthy oil palm area and RVI index for the diseased oil palm area (R2 = 0.48 and 0.68, respectively), and (2) the MSAVI provided the best R2 value in patterns correlated to the greenness of vegetation for the mixed oil palm tree areas (R2 = 0.44). Moreover, the results show that the overall Support Vector Machine (SVM) classification accuracy is 72.97%, with the kappa coefficient is 0.56 for the healthy oil palm area, 64.16% and 0.40 for the diseased oil palm area and 50.00% and 0.37 for the mixed oil palm area. A concurrent UAV survey based on the visible and Visible Atmospherically Resistant Index (VARI) bands and SVM classification provided higher overall accuracy compared to the Sentinel-2 SVM classification.KEYWORDS:
1 Faculty of Environmental Management,
Prince of Songkla University,
Hatyai, Songkhla 90110, Thailand.
2 Asian Disaster Preparedness Center, 24th Floor,
SM Tower, 979/69-70 Paholyothin Road,
Phyathai, Bangkok 10400, Thailand.
* Corresponding author e-mail: firstname.lastname@example.org