云南省景观格局的广义回归神经网络模型

A Generalized Regression Neural Network Model of the Landscape Pattern in Yunnan Province, China

摘要: 以云南省129个县（区、市）域为研究对象，基于2000、2005、2010、2015、2018年共5年土地利用（LUCC）及经济社会、自然因子数据，建立了云南省景观格局的广义回归神经网络模型（GRNN）。研究结果表明：GRNN模型能较好定量描述景观指数PD、LSI、CONTAG、SPLIT、SHDI、SHEI、AI与驱动因子之间的高度非线性关系，检验样本误差R²均﹥0.8，模拟效果较好，建立的驱动因子与景观格局指数之间的响应模型能为地区景观生态安全格局预警提供基础支撑。

Abstract: Based on the data of land use(LUCC), economic and social factors, and natural factors in 2000, 2005,2010, 2015 and 2018 for 129 counties(districts and cities) in Yunnan Province, a generalized regression neural network(GRNN) model of landscape pattern was established. The result showed that the GRNN model could better quantitatively describe the highly nonlinear relationship between the landscape indexes(PD、LSI、CONTAG、SPLIT、SHDI、SHEI、AI) and driving factors. The determinable coefficients(R₂) of test samples were greater than 0.8. The GRNN model of the landscape pattern could provide basic support for regional early warning of landscape ecological security pattern.