Detections of damage by oryctes rhinoceros L. infestation in immature oil palm using multispectral unmanned aerial vehicle imagery / Mohd Shahrizan Syukur

Oryctes rhinoceros L., known as Rhinoceros beetle (RB), is one of the major pests in oil palm. Infestation of the pest can lead to delayed maturity, reduction in oil palm trees morphology and yield loss. Integrated pest management (IPM) such as biological control and insecticide spraying are in place to suppress the infestation. However, determining infested palm trees and the exact localities relies heavily on damage assessment through manual census, which could be labour intensive, time-consuming, and prone to human error. Therefore, this study examines the level of damages of oil palm caused by RB infestation using field measurement (biophysical parameters) and multispectral data (biophysical parameters, vegetation indices and textural features). Rapid Damage Survey was performed to identify the level of damages, which were classified into three: uninfested (L1), damaged (L2), and severely damaged (L3). Field measurements including crown area, perimeter, diameter and height of oil palm were taken from immature oil palm of different levels of damage, from the age of 1 until 2 years old Subsequently, aerial imagery was acquired using Micasence RedEdge-M, a multispectral sensor mounted underneath a DJI Phantom 4, a multi-rotor unmanned aerial vehicle (UAV) with three different flying altitudes: 20 m, 60 m and 80 m above ground level (AGL). Relationships were explored for all collected fields and extracted variables towards the level of damages. Results of field data analysis show that RB infestation does reduce the morphology of oil palm (R2 = 0.53 - 0.87). Meanwhile, multispectral data analysis shows that only crown variables (crown diameter, perimeter and area) can be explained by the level of damages (R2 = 0.62 - 0.78). Vegetation indices and textural features show no relationship with the level of damage. Based on the results of the stepwise regression method, this study observed that 60 m altitude data is the best model utilizing the crown diameter as the only variable explaining the level of damages, with the highest R2 of 0.78. The model produces a substantial Cohen’s Kappa value (0.644) and the highest percentage agreement of 75.8% compared to other flying altitudes. In conclusion, the proposed remote sensing techniques are proven to be capable of providing efficient damage assessment compared to the conventional manual census. It demonstrates that extracted variables from UAV imagery are useful in detecting the level of damages in immature oil palm caused by RB infestation. The models can identify hotspots for biological control, such as artificial breeding sites and preclude blanket spraying of biological control agents. The sub-meter accuracy of the model can assist the newly introduced UAV-sprayer system, which aims explicitly to control RB infestation through insecticide spraying. This study shows that remote sensing can facilitate more targeted and effective control measures.