Dataset and Results for: Multimodal Fusion of Remote Sensing and Agricultural Data for High-Resolution Life Expectancy Prediction

This repository contains the data and machine learning outputs accompanying the manuscript: "Multimodal Fusion of Remote Sensing and Agricultural Data for High-Resolution Life Expectancy Prediction: A County-Level Analysis Across the Continental United States." Repository Contents: full_clean_engineered_dataset_with_LE.csv.zip: The primary multimodal dataset containing 62,160 county-year observations spanning 3,108 CONUS counties from 2000–2019. It integrates 450 physically sensed and agriculturally derived features from 11 data streams (MODIS, Sentinel-1/2, Landsat, USDA CDL, ESA CCI soil moisture, JRC, Copernicus DEM, and FAO livestock densities) alongside IHME life expectancy targets. Archive.zip: Contains the final production machine learning outputs, including the 5-fold cross-validation predictions, spatial error metrics, and exact SHAP (SHapley Additive exPlanations) feature importance scores used to generate the manuscript's figures and tables. Code Availability: The Google Earth Engine extraction scripts and Python machine learning pipelines used to generate and process this data are available on GitHub at: https://github.com/albertfaiz/Multimodal_geo_fusion_FM本仓库附带对应学术手稿《多模态融合遥感与农业数据实现高分辨率预期寿命预测：美国大陆县级尺度分析》（Multimodal Fusion of Remote Sensing and Agricultural Data for High-Resolution Life Expectancy Prediction: A County-Level Analysis Across the Continental United States）的数据与机器学习实验产出。 仓库内容如下： full_clean_engineered_dataset_with_LE.csv.zip：核心多模态数据集，涵盖2000年至2019年间美国大陆3108个县级行政区的62160条县级-年度观测样本。该数据集整合了来自11个数据源的450项物理传感获取与农业衍生特征，包括MODIS、Sentinel-1/2、Landsat、美国农业部作物数据层（United States Department of Agriculture Crop Data Layer, USDA CDL）、欧洲空间局气候变化倡议土壤湿度数据集（European Space Agency Climate Change Initiative soil moisture, ESA CCI soil moisture）、欧盟联合研究中心（Joint Research Centre, JRC）、哥白尼数字高程模型（Copernicus DEM）以及联合国粮食及农业组织（Food and Agriculture Organization of the United Nations, FAO）畜牧密度数据，同时附带全球卫生度量与评估研究所（Institute for Health Metrics and Evaluation, IHME）发布的预期寿命目标值。 Archive.zip：包含最终投产的机器学习实验结果，涵盖用于生成论文图表与表格的5折交叉验证预测结果、空间误差指标以及精确SHAP（SHapley Additive exPlanations）特征重要性得分。 代码可用性：用于生成与处理本数据集的谷歌地球引擎（Google Earth Engine）提取脚本与Python机器学习流水线，已开源至GitHub：https://github.com/albertfaiz/Multimodal_geo_fusion_FM