Predicting Regional Distribution Of Coffee Leaf Rust Disease Using Ensemble Machine Learning Models: A Case Study In The Philippines
Jazpher John Figueroa-Jimenez1, Francisco Geronimo-Isidro III1, Teresa Elika Joy Lacuesta-Jalotjot1, Jeffer Troy Cabangbang-Jaranilla1, Nicole Andrea Gabayno-Laguatan1, Don Enrico Buebos-Esteve2, James Eduard Limbo-Dizon2, Nikki Heherson A. Dagamac12
1Department of Biological Sciences, College of Science, University of Santo Tomas, España, Manila, 1008, Philippines
2Research Center for Natural and Applied Sciences, University of Santo Tomas, España, Manila, 1008, Philippines
Hemileia vastatrix, an obligate hemicyclic biotrophic fungi, is known to be the causative agent of coffee leaf rust (CLR) disease, a major coffee plant disease worldwide that caused detrimental effects to the Philippine coffee industry over the last decades. Occurrences of H. vastatrix and randomly generated pseudo-absences within Philippines were used to estimate its local spatial distribution. Environmental covariates influencing the development of H. vastatrix were identified and evaluated to determine CLR propagation. Ten modeling algorithms, namely, Generalized Linear Model (GLM), Generalized Boosting Model (GBM), Generalized Additive Mode (GAM), Classification Tree Analysis (CTA), Artificial Neural Network (ANN), Surface Range Envelop (SRE), Flexible Discriminant Analysis (FDA), Multiple Adaptive Regression Splines (MARS), Random Forest (RF), and Maximum Entropy (MAXENT) under the BIOMOD2 framework, were used to construct the model, weighted by true skill statistic (TSS). The ensemble model generated in this study predicted the current habitat suitability of H. vastatrix and was projected using the EC-Earth3-Veg general circulation model (GCM) to assess range shifts under two types of future Shared Socio-economic Pathways (SSP126 and SSP585) for 2041-2100 in 20 year intervals. CLR management and knowledge is considerably lacking, thus areas in parallel to the suitability of fungal proliferation are critical in the prediction of the next possible outbreak.