Anthrax is one of the most neglected tropical disease affecting humans, livestock, and wildlife worldwide. The disease is caused by soil-borne spore-forming bacteria called Bacillus anthracis. A machine learning algorithm with the biomod2 package of R software was used to develop a predictive map for the Amhara regional state of Ethiopia. One hundred twenty-eight georeferenced confirmed outbreak reports of anthrax in livestock and 11 bioclimatic, eight soil characteristics, and three livestock density variables were used to train the model. The algorithm was set to run 3-fold with a total of 27 outputs for the nine selected models. An ensemble model was developed with ROC evaluation metrics set at 0.8. The ensemble model showed an improved performance than the individual models (KAPPA, TSS, and ROC values of 0.86, 0.93, and 0.99, respectively). Variables like annual precipitation (22.51 %), precipitation of warmest quarter (14.17 %), precipitation of wettest month (11.61 %), cattle density (9.67 %), sheep density (6.6 %), annual maximum temperature (6.17 %), altitude/elevation (5.24 %), and sand content (4.83 %) contributed the highest share in the ensemble model. The predicted suitable areas were primarily in the Central and Southern parts of the region. West Gojam and South Gondar zones were found highly suitable; while parts of Waghemira, North Wollo, and South Wollo were not significantly suitable. Besides, East Gojam, North Gondar, and Awi administrative zones were also reasonably suitable to Bacillus anthracis. The study can be used as a basis in the planning of prevention and control approaches of anthrax outbreaks in the region. Administrative zones like West Gojam, South Gondar, Awi, and East Gojam have to be prioritized as a risky-areas in the planning of preventive measures of anthrax in the region.
Keywords: Amhara region; Anthrax; Bacillus anthracis; Ensemble modelling; Suitability modelling.
Copyright © 2020 Elsevier B.V. All rights reserved.