Simulating land-use changes by incorporating spatial autocorrelation and self-organization in CLUE-S modeling： a case study in Zengcheng District, Guangzhou, China

摘要

The Conversion of Land Use and its Effects atSmall regional extent （CLUE-S）, which is a widely usedmodel for land-use simulation, utilizes logistic regressionto estimate the relationships between land use and itsdrivers, and thus, predict land-use change probabilities.However, logistic regression disregards possible spatialautocorrelation and self-organization in land-use data.Autologistic regression can depict spatial autocorrelationbut cannot address self-organization, while logisticregression by considering only self-organization （NE-logistic regression） fails to capture spatial autocorrelation.Therefore, this study developed a regression （NE-auto-logistic regression） method, which incorporated bothspatial autocorrelation and self-organization, to improveCLUE-S. The Zengcheng District of Guangzhou, Chinawas selected as the study area. The land-use data of 2001,2005, and 2009, as well as 10 typical driving factors, wereused to validate the proposed regression method and theimproved CLUE-S model. Then, three future land-usescenarios in 2020： the natural growth scenario, ecologicalprotection scenario, and economic development scenario,were simulated using the improved model. Validationresults showed that NE-autologistic regression performedbetter than logistic regression, autologistic regression, andNE-logistic regression in predicting land-use changeprobabilities. The spatial allocation accuracy and kappavalues of NE-autologistic-CLUE-S were higher than thoseof logistic-CLUE-S, autologistic-CLUE-S, and NE-logis-tic-CLUE-S for the simulations of two periods, 2001-2009and 2005-2009, which proved that the improved CLUE-Smodel achieved the best simulation and was therebyeffective to a certain extent. The scenario simulation resultsindicated that under all three scenarios, traffic land andresidential/industrial land would increase, whereas arableland and unused land would decrease during 2009-2020.Apparent differences also existed in the simulated changesizes and locations of each land-use type under differentscenarios. The results not only demonstrate the validity ofthe improved model but also provide a valuable referencefor relevant policy-makers.