AZ-Hydro — Historical and Projected Arizona Annual Water Use: Software, Input Data, Models, Raster and Well Package Predictions, and Validation at 2 km Resolution (1896–2099)
收藏资源简介：
AZ-Hydro: Historical and Projected Arizona Annual Water Use, 1896–2099 A 2 km-resolution gridded dataset of Arizona groundwater and surface-water withdrawals, irrigation consumptive use, and pumping-induced surface-water capture, spanning 204 years (1896–2099) with quadrature-combined uncertainty bands. Companion data archive for Majumdar et al. (in prep., Nature Scientific Data) and Majumdar et al. (in prep., AGU Earth's Future). Graphical Abstract: https://github.com/montimaj/az-hydro/blob/main/docs/images/Graphical_Abstract_Fig1.png What's in this deposit File Size Contents Audience az-hydro-headline.7z ~8.8 GB Focused subset: published per-pixel/per-year predictions (6-band augmented rasters with σ + CV + SNR + 95 % CI), per-well GeoParquet, SW capture, aggregated time series, CAP shortage scenario outputs, validation against USGS/ADWR/Reitz, era-mean and trend spatial figures. Most users want this. Reviewers, downstream researchers using the published product az-hydro-data.7z ~80 GB Full reproducibility archive: all of the above plus raw inputs (GEE tiles, ADWR meter records, well registry, GW basin / AMA-INA / CAP / SRP / streamflow / USBR vectors, statewide WTD), Step 2 cross-validation outputs, intermediate predictor stacks, per-component σ rasters (σ_MACA / σ_Model / σ_Irr / σ_LULC / σ_GW / σ_USBR / σ_CU). Anyone reproducing the full pipeline from scratch az-hydro-1.0.0.zip ~55 MB Source code release (Python pipeline, GEE export scripts, documentation). Same content as the GitHub repository at the tagged release. Anyone running the pipeline Inside each .7z, see Data/HEADLINE_README.md (headline archive) or Data/README.md (full archive) for a complete per-directory inventory and external-source citations. Important: how to unpack the .7z files The .7z format is not openable by macOS Archive Utility. Use one of: macOS — Keka or The Unarchiver Windows — 7-Zip, WinRAR, or Bandizip Linux — p7zip (e.g. apt install p7zip-full), then 7z x az-hydro-data.7z LZMA2 + solid-block compression yields ~35 % ratio (80 GB compressed from ~224 GB raw; 8.8 GB compressed from ~40 GB raw). Methods overview AZ-Hydro is a four-step physics-constrained ML pipeline: XGBRF prediction of total annual water-use depth per pixel, trained on per-well ADWR meter records (1984–2024) with 16 predictor bands (climate, ET, Peff, irrigation fraction, well density, canal density, water-rights density, etc.). Density-ratio partition decomposing the total prediction into Irrigation/Non-Irrigation × GW/SW × CU using era-mapped factors anchored to USGS Circulars 1950–2015 and ADWR Annual Reports 2016–2024. Six-component quadrature uncertainty quantification: σ_MACA (5 GCMs) + σ_Model (10 XGBRF seeds, t-corrected) + σ_LULC (4 USGS FORE-SCE scenarios) + σ_USBR (5 CMIP3 Upper-Colorado streamflow members, t-corrected) + σ_GW (5 recent ADWR Well Registry snapshots, t-corrected) + σ_CU (analytic propagation through Irrigation Efficiency). Per-pixel SW Capture Fraction and Volume with σ_GW propagation to quantify pumping-induced streamflow depletion. Key validations 2016 ADWR Total: model 6.72 MAF vs ADWR ~7.0 MAF (within −0.28 MAF) 2017 ADWR: 6.81 vs 7.0 MAF; GW share 44.9 % vs 41 % (within 4 pp) 2015 USGS GW pumping: 2.96 vs USGS 3.09 MAF (within −0.13 MAF) 2019–2020 ADWR irrigation share: 73.8 % vs 74 % (essentially exact) WestWater (2026) CAP shortage scenarios: AZ-Hydro Basic Coordination cumulative ΔGW = 7.24 MAF vs WestWater Fig 4 anchor 8.0 MAF (within −9 %); Extreme Shortage = 13.08 MAF vs Fig 4 anchor 8.7 MAF (gap reflects AZ-Hydro's no-regulatory-ceiling framing). CAP delivery shortage scenario sweep Eight scenarios (Baseline_900kAF, DCP Tier 0/1/2a/2b/3, WestWater Basic Coordination, Extreme Shortage) re-partitioned 2026–2099. Cumulative additional GW pumping over 2027–2060: DCP Tier 3 = 10.7 MAF, Basic Coordination = 7.24 MAF, Extreme Shortage = 13.08 MAF. Spatial maps (basin choropleth, per-pixel cumulative ΔGW, σ_cum context, basin/pixel signal-to-noise) included for both 2027–2060 (WestWater anchor) and 2027–2099 windows. External datasets required to reproduce the pipeline (not redistributed here) The pipeline reads four USGS ScienceBase data products that you must download separately: USGS NHM withdrawals (Haynes et al. 2023) — irrigation withdrawals + efficiency by HUC12, 2000–2020 USGS NHM CU / IE reanalysis (Martin et al. 2023; Martin et al. 2025) — irrigation consumptive use + Peff by HUC12 USGS public-supply reanalysis (Luukkonen et al. 2023; Alzraiee et al. 2024) — public-supply withdrawals by HUC12 USGS Reitz historical ET / Peff (Reitz et al. 2023; Reitz, Sanford & Saxe 2023) — 800 m gridded irrigation ET, 1980–2018 Bundled here directly: ADWR Well Registry, ADWR Meter Data, GW basin / AMA-INA / CAP / SRP / streamflow / USBR vectors, HarDWR water-rights shapefile (Lisk et al. 2024), GRAIN canal network (Suresh et al. 2026), and the Ma et al. 2026 statewide WTD TIFs. See Data/README.md inside each .7z for exact filename / sub-directory naming required. Citation Data archive (this deposit): Majumdar, S., Smith, R.G., ReVelle, P., Hasan, M.F., & Wogenstahl, C. (2026). AZ-Hydro — Historical and Projected Arizona Annual Water Use: Software, Input Data, Models, Raster and Well Package Predictions, and Validation at 2 km Resolution (1896–2099). Zenodo. https://doi.org/10.5281/zenodo.19057936 Companion papers: Majumdar, S., Smith, R.G., ReVelle, P., Hasan, M.F., & Wogenstahl, C. (2026). Freshwater withdrawals, irrigation consumptive use, and surface water capture for Arizona, 1896–2099. In prep. for Nature Scientific Data. Majumdar, S., Smith, R.G., ReVelle, P., Hasan, M.F., & Wogenstahl, C. (2026). Where Arizona's Water Goes: Declining Agricultural Dominance and Rising Urban Demand Drive a Two-Century Shift in Withdrawal Patterns (1896–2099). In prep. for AGU Earth's Future. License Data archives (az-hydro-data.7z, az-hydro-headline.7z): CC-BY-4.0 Source code (az-hydro-1.0.0.zip): BSD 3-Clause "Revised" (see LICENSE inside the zip) External datasets bundled here (HarDWR, GRAIN, WTD, ADWR products) retain their original upstream licenses Acknowledgments This work was supported by NASA (Grant numbers 80NSSC21K0979 and 80NSSC23K1453) and U.S. Army Corps of Engineers (Grant number W912HZ25C0016). We thank the open-source software and data communities, the OpenET consortium, and the Arizona Department of Water Resources for making their resources and datasets publicly available, and Google Earth Engine for compute and storage support. S.M. and P.R. acknowledge Dr. Justin L. Huntington, Christopher Pearson, Charles G. Morton, Blake A. Minor, Dr. Samapriya Roy at the Desert Research Institute, and Dr. David Ketchum at the University of Montana for their contributions to related projects that informed this work. We also thank Rahel Pommerenke at Colorado State University for presenting preliminary results from this work at the 2025 ESA Living Planet Symposium. The views expressed herein are those of the authors and do not necessarily reflect those of the funding agencies. Related links Live web app — AZ-Hydro Explorer: https://azhydro.projects.earthengine.app/view/azhydro-explorer (interactive GEE App: year slider 1896–2099, side-by-side category compare, click-driven pixel/basin/sub-basin/well time series with 95 % CI, CAP scenario × window dropdowns) GitHub repository: https://github.com/montimaj/az-hydro Issue tracker / bug reports: https://github.com/montimaj/az-hydro/issues Nature Scientific Data preprint (when available): (link) AGU Earth's Future preprint (when available): (link)
## AZ-Hydro：1896–2099年亚利桑那州历史与预测年度用水量数据集 本数据集为分辨率2km的格点数据，涵盖亚利桑那州1896–2099年共204年的地下水与地表水取水量、灌溉耗水量以及抽水诱发的地表水截留量，附带正交组合的不确定性区间。本数据集是Majumdar等（2026年待刊，《Nature Scientific Data》）及Majumdar等（2026年待刊，《AGU Earth's Future》）的配套数据存档。 ### 图形摘要 https://github.com/montimaj/az-hydro/blob/main/docs/images/Graphical_Abstract_Fig1.png ## 本存档包含内容如下 | 文件名称 | 大小 | 内容说明 | 适用受众 | |------------------------|------------|--------------------------------------------------------------------------|------------------------------| | az-hydro-headline.7z | ~8.8 GB | 精选子集：已发布的逐像元/逐年预测结果（含σ、变异系数CV、信噪比SNR、95%置信区间CI的6波段增强栅格）、逐井GeoPackage与Parquet文件、地表水截留指数、聚合时间序列、中央亚利桑那工程（Central Arizona Project, CAP）短缺情景输出、针对美国地质调查局（United States Geological Survey, USGS）/亚利桑那州水资源部（Arizona Department of Water Resources, ADWR）/Reitz数据的验证结果、时代均值与趋势空间图。本子集为多数用户所需。 | 审稿人、使用已发布成果的下游研究者 | | az-hydro-data.7z | ~80 GB | 完整可复现存档：包含上述全部内容，外加原始输入数据（谷歌地球引擎（Google Earth Engine, GEE）瓦片、ADWR计量记录、井位登记数据、地下水流域/亚利桑那州管理区-河道流量区（Arizona Management Areas - Instream Flow Areas, AMA-INA）/CAP/盐河项目（Salt River Project, SRP）/径流/美国垦务局（United States Bureau of Reclamation, USBR）矢量数据、全州地下水埋深（Water Table Depth, WTD）数据）、第二步交叉验证结果、中间预测因子堆叠数据、各分量不确定性σ栅格（σ_MACA、σ_Model、σ_Irr、σ_LULC、σ_GW、σ_USBR、σ_CU）。 | 需从头复现完整流程的研究者 | | az-hydro-1.0.0.zip | ~55 MB | 源代码发布包：含Python流程脚本、GEE导出脚本与文档，内容与标记版本的GitHub仓库一致。 | 需运行该流程的研究者 | 解压每个.7z文件后，可查看`Data/HEADLINE_README.md`（精选子集存档）或`Data/README.md`（完整存档）以获取完整的目录清单与外部数据源引用说明。 ## 重要说明：.7z文件解压方法 macOS系统自带的归档工具无法解压.7z格式文件，可使用以下工具之一： - macOS：Keka 或 The Unarchiver - Windows：7-Zip、WinRAR 或 Bandizip - Linux：p7zip（可通过`apt install p7zip-full`安装，执行命令`7z x az-hydro-data.7z`进行解压） 本数据集采用LZMA2+分块固体压缩算法，压缩比约为35%（原始数据约224GB，压缩后80GB；精选子集原始数据约40GB，压缩后8.8GB）。 ## 方法概述 AZ-Hydro采用四步物理约束机器学习（Machine Learning, ML）流程： 1. 极端梯度提升随机森林（Extreme Gradient Boosting Random Forest, XGBRF）模型预测逐像元年度总用水深度：基于1984–2024年逐井ADWR计量记录训练，输入包含16个预测因子波段（气候、蒸散发（Evapotranspiration, ET）、有效降水（Effective Precipitation, Peff）、灌溉占比、井密度、渠道密度、水权密度等）。 2. 密度比拆分：采用基于时代映射的因子，将总预测结果拆解为灌溉/非灌溉×地下水/地表水×耗水量（Consumptive Use, CU），因子锚定USGS通报1950–2015年及ADWR年度报告2016–2024年的数据。 3. 六分量正交不确定性量化：包含σ_MACA（5个全球气候模式（General Circulation Model, GCM）成员）、σ_Model（10个XGBRF随机种子，经t校正）、σ_LULC（4个USGS FORE-SCE情景）、σ_USBR（5个CMIP3（Coupled Model Intercomparison Project Phase 3）科罗拉多河上游径流成员，经t校正）、σ_GW（5个最新ADWR井位登记快照，经t校正）、σ_CU（通过灌溉效率的解析传播计算）。 4. 逐像元地表水截留指数：结合σ_GW的传播量，量化抽水诱发的径流衰减量。 ## 关键验证结果 - 2016年ADWR总用水量：模型估算6.72百万英亩-英尺（Million Acre-Feet, MAF），与ADWR公布的约7.0 MAF相比，误差为-0.28 MAF - 2017年ADWR数据：模型估算6.81 MAF，与7.0 MAF相比误差为0.19 MAF；地下水取占比44.9%，与ADWR公布的41%相比误差为3.9个百分点（percentage points, pp） - 2015年USGS地下水取水量：模型估算2.96 MAF，与USGS公布的3.09 MAF相比误差为-0.13 MAF - 2019–2020年ADWR灌溉用水占比：模型估算73.8%，与74%基本一致 - WestWater（2026）CAP短缺情景：AZ-Hydro基本协调情景下累计地下水变化量ΔGW为7.24 MAF，与WestWater图4基准值8.0 MAF相比误差为-9%；极端短缺情景下为13.08 MAF，与基准值8.7 MAF存在偏差（该偏差源于AZ-Hydro未设置监管上限的框架设定）。 ## CAP供水短缺情景扫描分析 共包含8种情景（Baseline_900kAF、DCP（Central Arizona Project Delivery Contract）Tier 0/1/2a/2b/3、WestWater基本协调情景、极端短缺情景），对2026–2099年用水进行重新分配。2027–2060年累计额外地下水取水量：DCP Tier 3为10.7 MAF、基本协调情景为7.24 MAF、极端短缺情景为13.08 MAF。针对2027–2060年（WestWater基准情景）与2027–2099年两个时段，均提供空间制图结果（流域等值线图、逐像元累计ΔGW、σ_cum背景图、流域/像元信噪比图）。 ## 复现流程所需的外部数据集（本存档未包含） 需单独下载以下4个USGS ScienceBase数据产品： 1. USGS水文模型（National Hydrologic Model, NHM）取水量数据（Haynes等，2023）：按HUC12水文单元划分的灌溉取水量与效率，2000–2020年 2. USGS NHM耗水量（CU）/灌溉效率（Irrigation Efficiency, IE）再分析数据（Martin等，2023；Martin等，2025）：按HUC12水文单元划分的灌溉耗水量与有效降水Peff 3. USGS公共供水再分析数据（Luukkonen等，2023；Alzraiee等，2024）：按HUC12水文单元划分的公共供水量 4. USGS Reitz历史蒸散发ET/有效降水Peff数据（Reitz等，2023；Reitz, Sanford & Saxe, 2023）：分辨率800m的格点灌溉蒸散发数据，1980–2018年 本存档直接包含以下数据：ADWR井位登记数据、ADWR计量数据、地下水流域/AMA-INA/CAP/SRP/径流/USBR矢量数据、HarDWR水权形状文件（Lisk等，2024）、GRAIN渠道网络（Suresh等，2026）以及Ma等（2026）发布的全州WTD TIF文件。 可查看每个.7z文件内的`Data/README.md`，获取所需的精确文件名与子目录命名规则。 ## 引用方式 ### 本数据存档引用 Majumdar, S., Smith, R.G., ReVelle, P., Hasan, M.F., & Wogenstahl, C. (2026). AZ-Hydro — 1896–2099年亚利桑那州历史与预测年度用水量数据集：软件、输入数据、模型、栅格与逐井GeoPackage预测结果及2km分辨率验证结果. Zenodo. https://doi.org/10.5281/zenodo.19057936 ### 配套论文引用 1. Majumdar, S., Smith, R.G., ReVelle, P., Hasan, M.F., & Wogenstahl, C. (2026). 亚利桑那州1896–2099年历史与预测地下水/地表水取水量、灌溉耗水量及抽水诱发地表水截留量. 待刊于《Nature Scientific Data》 2. Majumdar, S., Smith, R.G., ReVelle, P., Hasan, M.F., & Wogenstahl, C. (2026). 亚利桑那州水资源去向：农业占比下降与城市需求上升驱动两个世纪的取用水模式转变（1896–2099）. 待刊于《AGU Earth's Future》 ## 授权协议 - 数据存档（az-hydro-data.7z、az-hydro-headline.7z）：CC-BY-4.0 - 源代码（az-hydro-1.0.0.zip）：BSD 3-Clause“修订版”授权协议（可查看压缩包内的LICENSE文件） - 本存档包含的外部数据集（HarDWR、GRAIN、WTD、ADWR相关产品）：保留其原始上游授权协议 ## 致谢 本研究得到美国国家航空航天局（National Aeronautics and Space Administration, NASA）（项目编号80NSSC21K0979与80NSSC23K1453）及美国陆军工程兵团（项目编号W912HZ25C0016）的资助。感谢开源软件与数据社区、OpenET联盟以及ADWR公开分享其资源与数据集，同时感谢GEE提供计算与存储支持。S.M.与P.R.感谢内华达大学沙漠研究所的Justin L. Huntington博士、Christopher Pearson、Charles G. Morton、Blake A. Minor、Samapriya Roy博士，以及蒙大拿大学的David Ketchum博士，他们在相关项目中的工作为本研究提供了参考。此外感谢科罗拉多州立大学的Rahel Pommerenke在2025年ESA生活星球研讨会上展示本研究的初步成果。本文观点仅代表作者本人，不代表资助机构的立场。 ## 相关链接 1. 在线Web应用——AZ-Hydro Explorer：https://azhydro.projects.earthengine.app/view/azhydro-explorer（交互式GEE应用：支持1896–2099年滑块选择、分类对比、点击获取像元/流域/子流域/井位时间序列及95%CI、CAP情景与时段下拉选择） 2. GitHub仓库：https://github.com/montimaj/az-hydro 3. 问题追踪/ bug反馈：https://github.com/montimaj/az-hydro/issues 4. 《Nature Scientific Data》预印本（上线后可用）：(link) 5. 《AGU Earth's Future》预印本（上线后可用）：(link)
