AUTHOR=Shi Qianwen , Gong Yanfeng , Zhao Jian , Qin Zhiqiang , Zhang Jun , Wu Jingzhi , Hu Zengyun , Li Shizhu
TITLE=Spatial and Temporal Distribution Pattern of Oncomelania hupensis Caused by Multiple Environmental Factors Using Ecological Niche Models
JOURNAL=Frontiers in Environmental Science
VOLUME=10
YEAR=2022
URL=https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2022.942183
DOI=10.3389/fenvs.2022.942183
ISSN=2296-665X
ABSTRACT=Objective: This study aimed to predict the spatial and temporal distribution pattern of Oncomelania hupensis (O. hupensis) on a fine scale based on ecological niche models, so as to provide insights into O. hupensis surveillance.
Methods: Geographic distribution and environmental variables of O. hupensis in Suzhou City were collected from 2016 to 2020. Five machine learning algorithms were used, including eXtreme gradient boosting (XGB), support vector machine (SVM), random forest (RF), generalized boosted (GBM), and C5.0 algorithms, to predict the distribution of O. hupensis and investigate the relative contribution of each environmental variable. The accuracy of the five ecological niche models was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC) with ten-fold cross-validation.
Results: Five models predicted that the potential distribution of O. hupensis was in southwestern areas of Wuzhong, Wujiang, Taichang, and Xiangcheng counties. The AUC of RF, XGB, GBM, SVM, and C5.0 algorithms were 0.8233, 0.8051, 0.7938, 0.7897, and 0.7282, respectively. Comparing the predictive results and the truth of O. hupensis distribution in 2021, XGB and GBM models were shown to be more effective. The six greatest contributors to predicting potential O. hupensis distribution included silt content (13.13%), clay content (10.21%), population density (8.16%), annual accumulated temperatures of ≥0°C (8.12%), night-time lights (7.67%), and average annual precipitation (7.23%).
Conclusions: Environmental factors play a key role in the spatial and temporal distribution pattern of O. hupensis. The XGB and GBM machine learning algorithms are effective and highly accurate for fine-scale prediction of potential O. hupensis distribution, which provides insights into the surveillance of O. hupensis.