Predicting potential ecological distribution of Bactrocera dorsalis in China using GARP ecological niche modeling
ZHOU Guo-Liang,Chen Chen,Jun Ye,HU Bai-Shi,LIU Feng-Quan
2007-01-01
Abstract:Oriental fruit fly, Bactrocera dorsalis (Hendel), is a polyphagous pest. In order to monitor and control B. dorsalis effectively, it is necessary to investigate the potential distributional areas of this pest. In this study, an ecological niche modeling method, machine-learning algorithm for nonparametric prediction of species distribution (the Genetic Algorithm for Rule-set Prediction, or GARP), combined with Spatial Analysis function in GIS Geographic Information System) program, was applied to predict potential geographic distribution of B. dorsalis in Asia, especially in China. On the basis of the data about current distribution of B. dorsalis in China and 14 environmental and geological covariates in Asia, the resulting GARP geographic predictions were exported as ASCII raster grid files into ArcGIS8. 3 for further analysis. Projection of the distribution of B. dorsalis in Asia indicated that the pest could be distributed to some Asian countries and regions including China, Japan, Philippines, Malaysia, northern Thailand, Vietnam, Cambodia, Laos, Burma, Nepal, Pakistan, Bangladesh and Sri Lanka, broadly across eastern and southern Asia. The results were in agreement with the known worldwide distribution regions of B dorsalis reported by the EPPO European and Mediterranean Plant Protection Organization). Using the same sets of ecological dimensions and taking the model agreement into account, the ecological niche model was projected onto the landscape of provinces, municipalities and counties in China. The results showed that the most part of Yunnan Province, eastern and southern part of Sichuan Province, most part of Guizhou Province and Chongqing Municipality, Guangxi Zhuang Autonomous Region, Hong Kong, Macao, Guangdong, Taiwan, Hainan, Fujian and Jiangxi provinces, most part of Zhejiang, Hunan and Hubei provinces, Shanghai Municipality, southern part of Jiangsu, and some parts of Henan and Anhui provinces were suitable establishment areas of B. dorsalis' invasion, surrounded by sub-suitable establishment areas, which included some scattered areas in Sichuan Province, Guizhou Province, Chongqing Municipality, northern part of Hubei province and southern part of Jiangsu and Henan provinces. Host fruit plants have been cultivated at high densities in both suitable and sub-suitable establishment areas. The remaining test points were overlaid with the distributional maps to determine the numbers falling within the areas of predicted present and test the validity of this approach, and the one-tailed goodness-of-fit χ2-statistics were employed to assess the significance of the model. All of our predictions were highly statistically significant, indicating an excellent predictive ability. The success of this test demonstrated that the ecological niche modeling approaches provide a promising tool for addressing a wide range of issues in ecology, biogeography and conservation biology, and open door to many synthetic analyses based on primary sets of present-day point occurrence information. Moreover, this methodology provides a technical platform for the suitability analysis, monitoring, prevention and control of the pest.