深圳市分阶段人口转移空间分布数据（2022）

深圳市热带气旋灾中和灾后的人口转移安置路线。利用地理加权回归（GWR）将两个基于树的回归模型结合起来，以建立协变量与县级人口普查人口密度数据集之间的空间关系。经过训练后，该关系用于基于网格级协变量生成初步的精细人口地图。其次，基于数字足迹协变量，在指定的时间尺度上估计了县级人口密度，并根据县级估计值调整了初步网格级估计值。第三，利用深度残差网络进一步细化了网格估计值，以识别调整后的估计值与网格级协变量之间的非线性关系。为了获得不同时间尺度的网格化人口地图，我们改变了时间粒度，计算了不同时间尺度上的平均数字足迹协变量，并按照相同的过程开发了一个新模型，进而得到不同阶段的社区受灾人口转移安置路线。

Population transfer and resettlement routes during and after tropical cyclone disasters in Shenzhen. We combined two tree-based regression models using Geographically Weighted Regression (GWR) to establish the spatial relationship between covariates and county-level census population density datasets. After training, this established relationship was used to generate preliminary fine-grained population maps based on grid-level covariates. Secondly, county-level population densities were estimated at specified time scales based on digital footprint covariates, and the preliminary grid-level estimates were adjusted according to the county-level estimation results. Thirdly, Deep Residual Networks (ResNet) were used to further refine the grid-level estimates to identify the nonlinear relationships between the adjusted estimates and grid-level covariates. To obtain gridded population maps at different time scales, we adjusted the time granularity, calculated the average digital footprint covariates across various time scales, and developed a new model following the same procedure, thereby deriving population transfer and resettlement routes for disaster-affected communities at different stages.

搜集汇总

背景与挑战

背景概述

该数据集提供了2022年深圳市热带气旋灾中和灾后的人口转移安置路线空间分布信息，基于地理加权回归和深度残差网络等技术生成精细人口地图。数据以申请获取方式共享，大小为626.29 KB，适用于人口、自然灾害和基础设施等相关研究。