Publication: Modelling the infestation of coconut tree’s red palm weevil using geospatial and machine learning techniques
Loading...
Date
2023-04-01
Authors
Faradina Binti Marzukhi
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Modelling the red palm weevil infestation at a coconut plantation is really challenging. The
fundamental problem is that, except from visual observation, there is no exact method for
assessing the distribution of red palm weevils in the Northern region especially in Arau,
Perlis. Therefore, this study focuses on four (4) objectives in order to model the infestation
of coconut tree’s red palm weevil in the study area. Firstly, the remote sensing techniques were utilised to extract NDVI and categorise coconut trees into those infested with RPW and those that were not. Secondly, by using the abiotic parameters as independent
variables, the MLR forward stepwise regression methods were used to develop the red
palm weevil infestation models. The results demonstrated that the red palm weevil infestation in linear regression method has proven its efficiency and ability to predict the number of red palm weevil trapped at three trapping stations in the coconut plantation. The developed model's results for red palm weevil infestation (RPWc) were R2 (0.954) and
RMSE (1.021), with significant factors of temperature and wind. In third objective, four
groups of MLR were analysed for MLP and SVM techniques to improve the accuracy of
the RPWc prediction model. The results show that MLP performed better than SVM
methods for predictive processes, with MLP_7 producing the best results when compared
to other models. MLP_7 produced results in the train phase (R2 = 0.772, RMSE = 0.144)
and the test phase (R2 = 0.989, RMSE = 0.069). Lastly, the kriging interpolation technique was used in a calculator-like interface in GIS software (ArcGIS 10) application to map the
spatial distribution of RPW for two different months; July (2019) and December (2019) at
the catchment area to develop the regression equation for red palm weevil infestation
model. As a result, modelling the red palm weevil infestation on a coconut tree is now
possible using geospatial technologies such as GPS, GIS, and remote sensing techniques
in conjunction with machine learning techniques.