Remotely sensed relative humidity for predicting Metisa plana's population oil palm plantations

• Main Authors: Ruslan, Siti Aisyah, Muharam, Farrah Melissa, Omar, Dzolkhifli, Zulkafli, Zed Diyana, Zambri, Muhammad Pilus
• Format: Conference or Workshop Item
• Language: English
• Published: 2018
• Online Access: http://psasir.upm.edu.my/id/eprint/67015/1/39TH%20ACRS-2.pdf; http://psasir.upm.edu.my/id/eprint/67015/

Metisa plana (Walker) is leaves defoliating insect that is able to cause a staggering loss of USD 2.32 billion within two years to Malaysian oil palm industry. Therefore, an early warning system to predict the outbreak of Metisa plana that is cost, time, and energy effective is crucial. In order to do this, the role of environmental factors such as relative humidity (RH) on the pests’ population’s fluctuations should be well understood. Hence, this study utilized the geospatial technologies to i) to construct the relationship between the geospatially derived relative humidity and Metisa plana outbreak, and ii) to predict the outbreak of Metisa plana in oil palm plantation. Metisa plana census data of larvae instar 1, 2, 3, and 4 were collected approximately biweekly over the period of 2014 and 2015. Moderate Resolution Imaging Spectroradiometer (MODIS) satellite images providing values of RH were extracted and apportioned to 6 time lags; 1 week (T1), 2 weeks (T2), 3 week (T3), 4 weeks (T4), 5 week (T5) and 6 weeks (T6) prior to census date. Pearson’s correlation, multiple linear regression (MLR) and multiple polynomial regression analysis (MPR) were carried out to analyse the linear relationship between Metisa plana and RH. Artificial neural network (ANN) was then used to develop the best prediction model of Metisa plana’s outbreak. Results show that there are correlations between the presence of Metisa plana with RH, however, the time lag effect was not prominent. MPR was able to produce model with higher R2 in comparison to MLR with the highest R2 for both analysis were 0.48 and 0.15 respectively at T4 to T6. Model with the highest accuracy was achieved by ANN that utilized the RH at T1 to T3 at 95.29%. Based on the result of this study, the prediction of Metisa plana’s landscape ecology was possible with the utilization of geospatial technology and RH as the predictor parameter.