Fall armyworm habitat analysis in Africa with multi-source earth observation data

• Published in: Computers and electronics in agriculture Vol. 225; p. 109283
• Main Authors: Qi, Tonghui, Dong, Yingying, Li, XueLing, Zhao, MingXian, Huang, Wenjiang
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.10.2024
• Subjects: African fall armyworm; Exploratory factor analysis; Habitat suitability; Random forest; Remote sensing; African fall armyworm; Habitat suitability; Exploratory factor analysis; Random forest; Remote sensing
• ISSN: 0168-1699
• DOI: 10.1016/j.compag.2024.109283

•A method for constructing habitat suitability monitoring indicators based on exploratory factor analysis was proposed.•A monthly habitat suitability monitoring model for African FAW using random forest algorithm was proposed.•The proposed model can help to realize the dynamic and high-precision monitoring of FAW in Africa. The fall armyworm (Spodoptera frugiperda, FAW) is a significant migratory agricultural pest under the global warning of the Food and Agriculture Organization of the United Nations (FAO). Habitat suitability monitoring and analysis for FAW can help support more scientific management of pest dynamics and decision-making on prevention and control. In this study, we proposed a monthly habitat suitability monitoring model for FAW that integrates multi-source earth observation data such as climate, land use, vegetation, and soil, taking Africa, which is seriously affected by FAW, as the study area. First, exploratory factor analysis (EFA) was employed to reconstruct the climate variables to obtain three factors characterizing temperature, humidity, and wind respectively. Then, habitat suitability monitoring indicators were constructed by combining the factors with other environmental variables, and a monthly habitat suitability monitoring model for FAW in Africa was developed using the random forest algorithm. Finally, based on the model, the distribution of habitat suitability of FAW in Africa by month in 2023 was analyzed, along with the temporal and spatial variation characteristics of habitat suitability. The results indicate that: (1) The exploratory factor analysis effectively extracted the information of raw climate variables and demonstrated good interpretability. The monthly habitat suitability of African FAW was closely associated with Land Use/Land Cover (LULC), Normalized Difference Vegetation Index (NDVI), and humidity factor. (2) Compared with Maxent and SVM, the monthly habitat suitability monitoring model for FAW in Africa constructed using the random forest algorithm showed better performance, with the model being greater than 0.9 in each of the metrics such as Accuracy, Sensitivity, Specificity, F1-score, AUC, Kappa and TSS. The presence points of 2023 verified the validity of the model. (3) In 2023, FAW in Africa is mainly distributed in West Africa and East Africa south of the Sahara Desert, and in the Nile Delta north of the Sahara Desert. With seasonal changes, the suitable and unsuitable areas of FAW shift, and the movement pattern of the center of the presence points is basically consistent with the movement pattern of the precipitation zone. The areas with land use types of grasslands, savannas, and croplands are the hotspots of infestation in the rainy season every year. Partial Dependence Plots (PDPs) for LULC, NDVI, humidity factor and elevation provide deeper insights into how environmental variables drive model predictions. The results of this study demonstrate that the remote sensing analysis approach of FAW habitat suitability in Africa, which integrates multi-source earth observation data, can help to realize the dynamic and high-precision monitoring of FAW, provide references for the prevention and control of insect pests, and promote the sustainable development of agriculture.