TempEst 2 Development Data

This resource contains code and data related to the development of the TempEst 2 stream temperature remote sensing model (manuscript in review with Journal of Hydrology). The code includes the model implementation (model.R), some utility functions (valfn.R), data retrieval scripts for Google Earth Engine (eeretrieval.py and datapts.py), and a reproducible validation notebook (validation.Rmd), along with the knitted PDF of the latter (validation.pdf). The main data include stream temperature daily mean/max observations retrieved from the USGS NWIS as well as remotely-sensed and gridded observations retrieved using Google Earth Engine from NLDAS, ESA WorldCover, MODIS, ERA5, and EPA Ecoregions (using eeretrieval.py). These are contained in three files. AllData.csv includes all observations for mean temperature. ExtData.csv (\"extended data\") adds maximum temperature, at the expense of fewer total observations being included. Ecoregions.csv is not central to the analysis, but includes EPA Level I ecoregion classifications for convenience.

Model performance tests can be reproduced using validation.Rmd. To run validation.Rmd in full, there must be a Data directory with subdirectories Density and TSLen, a Figures directory, at least one of the main data files (AllData.csv, ExtData.csv) or equivalent, and Ecoregions.csv. A knitted version of the Notebook is included in this resource. The error map plots also use an EPA Level I Ecoregions (https://gaftp.epa.gov/EPADataCommons/ORD/Ecoregions/cec_na/na_cec_eco_l1.zip) shapefile, which is assumed to be in an Ecoregions subdirectory of the *parent* directory. This dependency can be removed by replacing the `plot.eco` function with ordinary ggplot plotting.

The two rda (RData) files contain different versions of a pre-trained model. model.rda contains a regular, pre-trained model function that can be used directly to generate predictions. krigs.rda contains a list of the actual fitted kriging models, which can be used for investigating model components (see demo.pdf).

These data and related items of information have not been formally disseminated by NOAA, and do not represent any agency determination, view, or policy.

本资源包含与TempEst 2溪流温度遥感模型开发相关的代码与数据（相关稿件已提交至《Journal of Hydrology》（水文学报）审稿中）。代码涵盖模型实现脚本（model.R）、若干工具函数（valfn.R）、适用于Google Earth Engine（GEE）的数据获取脚本（eeretrieval.py与datapts.py），以及可复现验证的R笔记本（validation.Rmd），同时附带该笔记本的编织输出PDF文件（validation.pdf）。
核心数据集涵盖从美国地质调查局国家水信息系统（USGS NWIS）获取的溪流日平均/最高温度观测数据，以及通过Google Earth Engine从NLDAS、ESA WorldCover、MODIS、ERA5以及美国环境保护署（EPA）生态区获取的遥感与格点观测数据（可通过eeretrieval.py完成数据获取）。上述数据存储于三个文件中：AllData.csv包含所有日平均温度观测数据；ExtData.csv（"扩展数据集"）额外包含最高温度观测数据，但总观测样本量有所减少；Ecoregions.csv并非本次分析的核心数据，仅为方便使用而附带的EPA一级生态区分类数据。
可通过validation.Rmd复现模型性能测试。若要完整运行该验证笔记本，需创建Data目录及其子目录Density与TSLen、Figures目录，且至少需包含一个主数据文件（AllData.csv或ExtData.csv，或等效替代文件）与Ecoregions.csv。本资源已附带该笔记本的编织输出版本。误差地图绘图还需使用EPA一级生态区形状文件（https://gaftp.epa.gov/EPADataCommons/ORD/Ecoregions/cec_na/na_cec_eco_l1.zip），该文件需放置于父目录下的Ecoregions子目录中。若需移除该依赖项，可将`plot.eco`函数替换为常规的ggplot绘图代码。
本资源包含两个RData（.rda）文件，分别存储不同版本的预训练模型。model.rda包含可直接用于生成预测结果的常规预训练模型函数；krigs.rda包含实际拟合的克里金模型列表，可用于探究模型组件细节（详见demo.pdf）。
本数据集及相关信息尚未由美国国家海洋和大气管理局（NOAA）正式发布，不代表任何机构的立场、观点或政策。