A District-Level CAMS Forecast–ERA5 Reanalysis PM-Based Air-Quality Proxy and Meteorological Dataset for Rajshahi Division, Bangladesh (July 2015–December 2025)

This dataset provides hourly district-level estimates of particulate-matter air quality and meteorological conditions for the eight ADM2 districts of Rajshahi Division, Bangladesh: Bogura, Chapainawabganj, Joypurhat, Naogaon, Natore, Pabna, Rajshahi and Sirajganj. It covers 1 July 2015 00:00 UTC through 31 December 2025 23:00 UTC and contains 736,704 district-hour records, with 92,088 hourly timestamps per district. Surface PM2.5 and PM10 concentrations are derived from CAMS global atmospheric-composition operational forecasts. Meteorological variables are obtained from ERA5 hourly single-level reanalysis and include temperature, dew point, relative humidity, 10 m wind components, wind speed and wind direction. Native CAMS and ERA5 grid-cell footprints were intersected with a pinned 2020 geoBoundaries BGD ADM2 boundary. Intersection areas were measured in EPSG:6933 and used as fractional weights for district-level aggregation. Values with less than 95% valid district-polygon coverage were withheld. The release contains analysis-ready hourly core and extended datasets, Bangladesh-local daily products and a convenience Excel workbook. Detailed fields record source provenance, spatial coverage, quality control, CAMS forecast initialization, lead time, source-archive identifiers and mapped IFS-COMPO cycles. PM-based indicators include trailing 24-hour AQI proxies, strict-completeness companion fields, a 12-hour PM NowCast proxy and daily PM AQI products. They apply the US EPA 2024 PM breakpoints with documented completeness rules. These values are model-derived district estimates, not ground-monitoring observations, population-weighted exposures or official regulatory AQI. Total-column CO, NO2, SO2 and O3 were excluded because they are not surface concentrations and cannot be used as station-equivalent AQI inputs. No Bangladesh ground-monitoring dataset was available for calibration or independent error validation. The complete district-time index is retained when source data are unavailable. Three documented CAMS companion-tile gaps total 2,232 source hours, and the first CAMS endpoint is unavailable. Affected PM and AQI values remain missing with explicit coverage flags and were not interpolated. CAMS is a stitched operational forecast record, not a fixed-system reanalysis. Model upgrades, particularly CY49R1 in November 2024, may introduce structural discontinuities in longitudinal analyses. The package includes CSV, Parquet and Excel products, data dictionaries, methods, quality reports, provenance records, source archives, request manifests, licences, reconstruction and validation source code, tests, file descriptions and SHA-256 checksums. These materials support transparent verification and responsible reuse.本数据集为孟加拉国拉杰沙希专区下辖的8个二级行政分区（ADM2）提供逐小时的颗粒物空气质量与气象条件估算值，涉及的分区包括博格拉、查帕伊诺瓦布甘杰、焦伊布尔哈特、诺阿冈、纳托尔、巴布纳、拉杰沙希以及希拉杰甘杰。数据集的时间覆盖范围为2015年7月1日00:00协调世界时（UTC）至2025年12月31日23:00协调世界时（UTC），总计包含736704条分区-小时记录，每个分区对应92088个小时时间戳。 地表PM2.5与PM10浓度数据源自哥白尼大气监测服务（CAMS）的全球大气组分业务预报产品。气象变量则来自欧洲中期天气预报中心第五代再分析数据集（ERA5）的逐小时单级再分析资料，涵盖气温、露点温度、相对湿度、10米风分量、风速以及风向。将CAMS与ERA5原始格网的空间覆盖范围与2020年固定版孟加拉国二级行政分区边界（geoBoundaries）进行空间相交运算。相交区域采用EPSG:6933坐标系进行量算，并以此作为分区聚合的权重系数。分区多边形有效覆盖占比低于95%的估算值将被剔除。 本次发布的数据包含可直接用于分析的逐小时核心与扩展数据集、孟加拉国本地每日产品以及便捷的Excel工作簿。详细字段记录了数据来源溯源、空间覆盖范围、质量控制、CAMS预报初始化信息、预报提前时长、源档案标识符以及映射得到的集成预报系统组分循环（IFS-COMPO）。基于颗粒物的指标包括连续24小时空气质量指数（AQI）代理值、严格完整性配套字段、12小时PM NowCast代理值以及每日PM空气质量指数产品。这些指标采用美国环境保护署（US EPA）2024年版颗粒物浓度断点阈值，并附带已文档化的完整性规则。 本数据集的估算值均为模式推演得到的分区结果，并非地面监测观测数据、人口加权暴露量或官方监管级空气质量指数。总柱含量的一氧化碳（CO）、二氧化氮（NO2）、二氧化硫（SO2）以及臭氧（O3）数据未被纳入，原因是其并非地表浓度数据，无法作为等效于地面监测站的空气质量指数输入参数。本次数据集未获取到孟加拉国的地面监测数据集用于校准或独立误差验证。 当源数据不可用时，完整的分区-时间索引仍会被保留。已记录的CAMS配套瓦片数据缺口共计3处，总计2232个源数据小时，且首个CAMS数据端点不可用。受影响的PM与AQI值将保留缺失状态，并附带明确的覆盖标记，未进行插值补全。CAMS数据为拼接得到的业务预报记录，并非固定系统的再分析资料。模式升级（尤其是2024年11月的CY49R1版本升级）可能会在纵向分析中引入结构不连续性。 该数据包包含CSV、Parquet及Excel格式的数据产品、数据字典、分析方法、质量报告、溯源记录、源档案、申请清单、授权协议、数据重建与验证源代码、测试脚本、文件说明以及SHA-256校验值。这些材料可支持数据的透明验证与负责任复用。