基于元胞自动机与旋转森林模型的粤港澳大湾区未来人口空间分布精细模拟

Under the background of global warming, China's climate is complex and changeable in recent years, extreme weather and climate events are increasing, there is an urgent need for high-precision future population distribution data to assist in early warning. The existing data on future population distribution are too rough to meet the increasingly complex needs of disaster exposure and risk assessment. Therefore, it is of great theoretical significance and application value to carry out high-precision simulation of future population spatial distribution. In this project, Guangdong-Hong Kong-Macau Greater Bay Area is taken as a case study, and the rotation forest algorithm is used to construct the machine learning model of population spatial distribution. Then the cellular automaton model (CA) is constructed by using deep learning algorithm for land use simulation. then the deep-CA and rotating forest are integrated to carries out multi-scenario refinement simulations of future population distribution. The specific research contents include: (1) the machine learning model of population spatial distribution based on rotating forest; (2) the future land use simulation based on deep-CA; (3) population distribution pattern and disaster risk early warning in different future scenarios.The purpose of this project is to provide theoretical methods and technical support for the fine simulation of the spatial distribution of the future population, and to provide scientific decision-making basis for the management to deal with climate change and disaster risk early warning.

在全球气候日趋变暖的背景下，近年来中国气候复杂多变，极端天气、气候事件日益增多，极需要高精度的未来人口分布数据协助进行预警。已有的未来人口分布数据过于粗糙，难以满足越加复杂的灾害暴露与风险评估需求。因此，开展高精度未来人口空间分布模拟研究具有重要的理论意义与应用价值。本项目以粤港澳大湾区为研究案例，采用旋转森林算法构建人口空间分布的机器学习模型，然后采用深度学习算法构建土地利用模拟的元胞自动机模型（CA），接着整合深度CA与旋转森林对未来人口分布进行多情景精细化模拟。具体的研究内容包括：（1）基于旋转森林的人口空间分布机器学习模型；（2）基于深度CA的未来土地利用模拟；（3）未来不同情景下的人口分布格局及灾害风险预警。本项目旨在为未来人口的空间分布精细模拟提供理论方法与技术支撑，为管理部门应对气候变化、灾害风险预警提供科学的决策依据。

在全球气候日趋变暖的背景下，近年来中国气候复杂多变，极端天气、气候事件日益增多，极需要高精度的未来人口分布数据协助进行预警。已有的未来人口分布数据过于粗糙，难以满足越加复杂的灾害暴露与风险评估需求。因此，开展高精度未来人口空间分布模拟研究具有重要的理论意义与应用价值。本项目以粤港澳大湾区为研究案例，采用旋转森林算法构建人口空间分布的机器学习模型，然后采用深度学习算法构建土地利用模拟的元胞自动机模型（CA），接着整合深度CA与旋转森林对未来人口分布进行多情景精细化模拟。取得的成果主要包括：（1）提出基于旋转森林的人口空间分布机器学习模型；（2）提出基于深度CA的未来土地利用模拟模型；（3）进行未来不同情景下的人口分布格局模拟及灾害风险预警。项目的成果可为未来人口的空间分布精细模拟提供理论方法与技术支撑，为管理部门应对气候变化、灾害风险预警提供科学的决策依据