River Sediment Database (RivSed)

The River Sediment Database (RivSed) database contains surface suspended sediment concentrations (SSC) derived from Landsat 5, 7, and 8 Level 1 Collection 1 surface reflectance from all rivers in the contiguous USA that are ~60 meters wide or greater. SSC represent spatially integrated "reach" median concentrations over the footprint of NHDPlusV2 centerlines where high quality river water pixels were detected within each Landsat image from 1984-2018. This is built in the River Surface Reflectance database (RiverSR) also in Zenodo (Gardner et al,. 2020 Geophysical Research Letters). The paper associated with RivSed: Gardner, J., Pavelsky, T. M., Topp, S., Yang, X., Ross, M. R., & Cohen, S. (2023). Human activities change suspended sediment concentration along rivers. Environmental Research Letters. https://iopscience.iop.org/article/10.1088/1748-9326/acd8d8 Files: 1) Metadata (riverSed_v1.0_metadata.pdf): Description of all data files associated with this repository. 2) RiverSed (RiverSed_USA_v1.1.txt). Table of SSC and associated data that is joinable to nhdplusv2_modified_v1.0.shp based on the "ID" column and to the original NHDplusV2 flowlines with the "COMID" column. 3) Shapefile of river centerlines to which the reflectance data can be attached (nhdplusv2_modified_v1.0.shp). 4) Shapefile of the reach polygons associated with each nhdplusv2_modified reach. (nhdplusv2_polygons_v1.0.shp). 5) The look up table for reach IDs of original (COMID) and modified (ID) NHDplusV2 centerlines. (COMID_ID.csv). Short reaches were joined together to optimize for remote sensing data collection and make more consistent reach lengths. 6) SSC-Landsat matchup database with extended metadata on locations and in-situ data derived from Aquasat (Ross et al., 2019) (Aquasat_TSS_v1.1.csv) 7) The final training data used to build the xgboost machine learning model (train_clean_xgb_v1.1.csv) 8) The xgboost model that can make SSC predictions over inland waters in USA using Landsat bands/band combinations (finalmodel_xgb_v1.1.rds and .RData). The model can only be loaded in R for now.

河流沉积物数据库（River Sediment Database, RivSed）包含基于Landsat 5、7、8 Level 1 Collection 1地表反射率数据反演得到的地表悬浮沉积物浓度（Surface Suspended Sediment Concentration, SSC），覆盖美国本土宽度约60米及以上的所有河流。SSC代表1984-2018年间每幅Landsat影像中，在NHDPlusV2水文数据集中心线覆盖范围内且检测到高质量河流水体像素的区域内，经空间整合得到的河段中位浓度。该数据集基于Zenodo平台发布的河流地表反射率数据库（River Surface Reflectance, RiverSR）构建而成（Gardner等，2020，《Geophysical Research Letters》）。

与RivSed相关的研究论文为：Gardner, J., Pavelsky, T. M., Topp, S., Yang, X., Ross, M. R., & Cohen, S. (2023). Human activities change suspended sediment concentration along rivers. Environmental Research Letters. https://iopscience.iop.org/article/10.1088/1748-9326/acd8d8

本仓库包含以下文件：
1) 元数据文件（riverSed_v1.0_metadata.pdf）：详细说明本仓库所有关联数据文件的信息。
2) 河流沉积物数据文件（RiverSed_USA_v1.1.txt）：包含SSC及关联数据，可通过"ID"列与nhdplusv2_modified_v1.0.shp进行关联，亦可通过"COMID"列与原始NHDplusV2水文流线数据对接。
3) 河流中心线矢量文件（nhdplusv2_modified_v1.0.shp）：可用于附加反射率数据。
4) 河段多边形矢量文件（nhdplusv2_polygons_v1.0.shp）：对应每个经修改的nhdplusv2河段的多边形数据。
5) 河段ID对照表（COMID_ID.csv）：包含原始NHDplusV2中心线的COMID与修改后中心线的ID的对应关系。为优化遥感数据采集效率并统一河段长度，短河段已进行合并处理。
6) SSC-Landsat匹配数据集（Aquasat_TSS_v1.1.csv）：包含扩展元数据，其位置信息与原位数据源自Aquasat数据集（Ross等，2019）。
7) 最终训练数据集（train_clean_xgb_v1.1.csv）：用于构建XGBoost机器学习模型。
8) XGBoost预测模型（finalmodel_xgb_v1.1.rds 与 .RData）：可基于Landsat波段/波段组合对美国内陆水体开展SSC预测，目前仅可在R语言环境中加载。