<b>SWOT PIXC enables reliable water</b><b>‑</b><b>level and volume estimation of supraglacial lakes on the Greenland Ice Sheet</b>

ICESat-2 observations provide independent validation. Catchments and drainage networks for representative lakes are derived from ArcticDEM.</p><h3 dir="ltr">Directory structure</h3><p dir="ltr"><b>01_inventory_classification</b></p><p dir="ltr">Contains the consolidated supraglacial lake inventory for northeast and southwest Greenland. The tables provide lake identifiers, regional and spatial attributes, observation summaries, WSE-curve availability, and manually interpreted behavioural classifications, including continuous filling, repeated filling–drainage, slow drainage, and rapid drainage.</p><p dir="ltr"><b>02_wse_timeseries</b></p><p dir="ltr">Contains quality-controlled SWOT PIXC WSE records and the final robust branchwise WSE time series.

The final dataset comprises 178 northeastern and 217 southwestern branch series. CSV and Parquet formats are provided. The corresponding 395 branchwise WSE figures are stored in the uncompressed TAR file <code>02_wse_timeseries_figures.tar</code>.

Historical rejected or noisy curve galleries are excluded, while point-level quality-control fields remain available in the released tables.</p><p dir="ltr"><b>03_area_wse_models</b></p><p dir="ltr">Contains paired Sentinel-2 lake-area and SWOT WSE observations used for model fitting. It includes linear and quadratic model candidates, AICc model-selection results, fitted coefficients, model statistics, evaluated area–WSE curves, and a routing table linking each fitting unit to its source data and figure.

The 83 individual fitting figures are stored in the uncompressed TAR file <code>03_area_wse_fit_figures.tar</code>. Lakes 1 and 2 include separately processed pre-merger branches and post-merger lake observations.</p><p dir="ltr"><b>04_volume_change</b></p><p dir="ltr">Contains relative lake-volume-change time series derived from the selected area–WSE relationships and reference water levels. The directory includes fitted volume curves, time-series tables, summary statistics, uncertainty estimates, representative-lake identifier crosswalks, individual volume-change figures, and representative runoff–volume comparison figures.

These products describe relative volume variations rather than absolute total lake volumes.</p><p dir="ltr"><b>05_observation_statistics</b></p><p dir="ltr">Contains the final regional and cross-region observation statistics used in the study. These include branchwise WSE retrieval rates, temporal coverage, valid and rejected SWOT observations, Sentinel-2 and SWOT observation-day complementarity, and WSE-threshold sensitivity results.

The adopted thresholds retain 395 robust branchwise WSE time series and 7,100 valid granule-level WSE observations across the two regions. Three northeastern audit tables describing an earlier 150-lake/152-branch cohort are retained for provenance and should not be interpreted as the final 178-branch results.</p><p dir="ltr"><b>06_vectors</b></p><p dir="ltr">Contains the public geospatial layers representing the supraglacial lake inventory, lake boundaries, behavioural classifications, representative catchments, and DEM-derived drainage networks.

The data are supplied as a GeoPackage and as standard Shapefile components for compatibility with different GIS software. A JSON file records the DEM-processing and catchment-delineation parameters. Large intermediate DEM rasters are not distributed.</p><p dir="ltr"><b>07_validation</b></p><p dir="ltr">Contains the comparison tables used to validate SWOT PIXC- and Raster-derived WSE against independent ICESat-2 observations.

It also includes 72 final ICESat-2 profile figures, an index linking each profile to its validation record, and summary diagnostic figures showing uncertainty and error distributions.</p><p dir="ltr"><b>08_figure_metadata</b></p><p dir="ltr">Contains the manuscript figure captions and a figure-to-data mapping table. These files identify the principal datasets and analysis products associated with each released scientific figure.</p><p dir="ltr"><b>09_quicklook_png</b></p><p dir="ltr">Contains 13 final PNG figures used in the manuscript and supplementary materials.

These include the study-area and SWOT-coverage map, research workflow, SWOT–ICESat-2 validation, lake inventory and behavioural map, northeast and southwest WSE curves, Lake 1 SWOT SLC phase panels, representative area–WSE models, DEM-derived catchments and drainage networks, representative volume changes, and regional temporal-coverage figures. The PNG files are retained individually for direct online preview.</p><p dir="ltr"><b>code</b></p><p dir="ltr">Contains the frozen public analysis-code snapshot for the northeast and southwest Greenland workflows.

The numbered scripts document the main processing sequence:</p><ol><li>Sentinel-2 preprocessing and spectral-index calculation;</li><li>supraglacial lake extraction and boundary generation;</li><li>effective lake-area retrieval;</li><li>branch-topology construction;</li><li>SWOT PIXC extraction;</li><li>WSE calculation, denoising, uncertainty estimation, and quality control;</li><li>final WSE-curve generation and behavioural classification;</li><li>AICc-based linear or quadratic area–WSE modelling;</li><li>observation-statistics generation;</li><li>relative lake-volume estimation and uncertainty analysis;</li><li>Lake 1 SWOT SLC phase-panel preparation;</li><li>manuscript-figure collection and validation;</li><li>thematic-map and study-area figure preparation;</li><li>final behavioural-product and figure-layout processing; and</li><li>WSE-threshold sensitivity analysis.</li></ol><p dir="ltr">The <code>_engine</code> directories contain the frozen underlying implementations used by the simplified numbered entry points.

Some stages are absent from the southwest workflow because area–WSE modelling, volume estimation, and SLC phase analysis were performed only for the northeast region. Public path placeholders must be replaced with local input and output directories before rerunning the workflow.</p><h3 dir="ltr">File formats and documentation</h3><p dir="ltr">CSV files are provided for broad accessibility, while Parquet versions of the largest tables support efficient programmatic analysis.

Geospatial data are supplied in GeoPackage and Shapefile formats. To remain within Figshare’s per-item file limit, the two largest collections of per-lake PNG figures are stored in uncompressed TAR containers. The TAR files preserve the original filenames and directory structure without altering or recompressing the PNG images.</p><p dir="ltr">The package-level <code>README.md</code>, <code>DATA_DICTIONARY.md</code>, <code>DATA_SOURCES.md</code>, and <code>UPLOAD_LAYOUT.md</code> files describe the directory structure, variables, data provenance, software requirements, official data-access locations, and TAR layout.

SHA-256 manifests are provided for file-integrity verification.</p><p dir="ltr">本数据集配套论文 <b>《基于 SWOT PIXC 观测估算格陵兰冰面湖水位与水量变化》</b>，提供论文研究中使用的主要衍生数据、空间数据产品、科研图件以及可复现分析代码。研究区域包括格陵兰东北部和西南部。</p><p dir="ltr">本研究结合 Sentinel-2 光学影像与 SWOT 二级高分辨率水体掩膜像元云产品（Level-2 High-Rate Water Mask Pixel Cloud，PIXC）。Sentinel-2 影像用于提取湖泊边界和湖泊面积时间序列，经过质量控制的 SWOT PIXC 观测用于获取水面高程（WSE）。对于具有足够面积–水位配对观测的湖泊，研究采用修正赤池信息准则（AICc）比较线性和二次面积–水位模型，并设置拟合优度阈值。最终选定的模型用于补充估算水位和相对湖泊水量变化。ICESat-2 观测用于独立验证，代表性湖泊的汇水区和排水网络则基于 ArcticDEM 提取。</p><h3 dir="ltr">目录结构</h3><p dir="ltr"><b>01_inventory_classification</b></p><p dir="ltr">包含格陵兰东北部和西南部冰面湖的综合清单。表格提供湖泊编号、区域与空间属性、观测统计、水位曲线可用情况以及人工判读的湖泊变化类型，包括持续填充、反复填充–排水、缓慢排水和快速排水。</p><p dir="ltr"><b>02_wse_timeseries</b></p><p dir="ltr">包含经过质量控制的 SWOT PIXC 水面高程记录以及最终稳健的分支水位时间序列。最终数据包括东北区域178条和西南区域217条分支序列。数据同时提供 CSV 和 Parquet 格式。对应的395张分支水位曲线存放在不压缩的 <code>02_wse_timeseries_figures.tar</code> 文件中。历史剔除曲线和噪声曲线图库未纳入发布包，但逐观测点的质量控制字段仍保留在发布表格中。</p><p dir="ltr"><b>03_area_wse_models</b></p><p dir="ltr">包含用于模型拟合的 Sentinel-2 湖泊面积与 SWOT 水位配对观测。数据包括线性和二次候选模型、AICc 模型选择结果、拟合系数、模型统计量、拟合后的面积–水位曲线，以及连接每个拟合单元与其源数据和图件的索引表。83张独立拟合图存放在不压缩的 <code>03_area_wse_fit_figures.tar</code> 文件中。Lake 1 和 Lake 2 分别处理了合并前的不同分支以及合并后的湖泊观测。</p><p dir="ltr"><b>04_volume_change</b></p><p dir="ltr">包含依据最终面积–水位关系和参考水位计算的相对湖泊水量变化时间序列。该目录包括拟合水量曲线、时间序列表、汇总统计、不确定性估计、代表性湖泊编号对应表、逐湖水量变化图以及代表性径流–水量对比图。这些结果表示相对水量变化，而不是湖泊的绝对总水量。</p><p dir="ltr"><b>05_observation_statistics</b></p><p dir="ltr">包含论文采用的最终区域及跨区域观测统计，包括分支水位序列提取率、时间覆盖度、有效与剔除的 SWOT 观测、Sentinel-2 与 SWOT 观测日期的互补性，以及水位阈值敏感性分析。采用最终阈值时，两个区域共保留395条稳健分支水位时间序列和7,100个有效的 granule 级水位观测。目录根部的三个东北区域审计表对应早期150个湖泊、152条分支的统计队列，仅用于来源追踪，不能作为最终178条分支结果使用。</p><p dir="ltr"><b>06_vectors</b></p><p dir="ltr">包含公开发布的冰面湖空间数据，包括湖泊清单、湖泊边界、变化类型、代表性汇水区以及基于 DEM 提取的排水网络。数据以 GeoPackage 和标准 Shapefile 组件两种形式提供，以兼容不同 GIS 软件。JSON 文件记录 DEM 处理和汇水区划分参数。体积较大的中间 DEM 栅格未包含在发布数据中。</p><p dir="ltr"><b>07_validation</b></p><p dir="ltr">包含使用独立 ICESat-2 观测验证 SWOT PIXC 和 SWOT Raster 水位结果的对比表。该目录还包括72张最终 ICESat-2 剖面图、连接剖面图与验证记录的索引表，以及展示不确定性和误差分布的汇总诊断图。</p><p dir="ltr"><b>08_figure_metadata</b></p><p dir="ltr">包含论文图件说明和图件–数据对应表，用于标明每张发布图件所对应的主要数据和分析产品。</p><p dir="ltr"><b>09_quicklook_png</b></p><p dir="ltr">包含论文正文和补充材料使用的13张最终 PNG 图件，包括研究区与 SWOT 覆盖图、研究流程图、SWOT–ICESat-2 验证图、湖泊清单与变化类型专题图、东北和西南区域水位曲线、Lake 1 SWOT SLC 相位图、代表性面积–水位模型、基于 DEM 的汇水区与排水网络、代表性湖泊水量变化图，以及两个区域的时间覆盖图。所有 PNG 均以独立文件形式保留，可在 Figshare 页面直接预览。</p><p dir="ltr"><b>code</b></p><p dir="ltr">包含格陵兰东北部和西南部分析流程的冻结公开代码。编号脚本记录了主要处理步骤：</p><ol><li>Sentinel-2 预处理与光谱指数计算；</li><li>冰面湖提取与边界生成；</li><li>有效湖泊面积提取；</li><li>湖泊分支拓扑构建；</li><li>SWOT PIXC 数据提取；</li><li>水位计算、去噪、不确定性估计与质量控制；</li><li>最终水位曲线生成与湖泊变化类型划分；</li><li>基于 AICc 的线性或二次面积–水位模型选择；</li><li>观测统计生成；</li><li>相对湖泊水量估算与不确定性分析；</li><li>Lake 1 SWOT SLC 相位图制作；</li><li>论文图件汇总与校验；</li><li>专题地图和研究区图件准备；</li><li>最终变化类型产品和图件版式处理；以及</li><li>水位阈值敏感性分析。</li></ol><p dir="ltr"><code>_engine</code> 目录保存简化编号入口实际调用的冻结算法实现。西南区域未执行面积–水位模型、水量估算和 SLC 相位分析，因此缺少相应步骤。重新运行代码前，需要将公开路径占位符替换为本地输入和输出目录。</p><h3 dir="ltr">文件格式与说明文档</h3><p dir="ltr">为便于广泛使用，主要数据以 CSV 格式提供；较大的表格同时提供 Parquet 格式，以支持高效的程序化读取。空间数据以 GeoPackage 和 Shapefile 格式发布。为满足 Figshare 单个条目的文件数量限制，两组规模最大的逐湖 PNG 图件采用不压缩 TAR 容器保存。TAR 仅用于组织文件并保留原始文件名和目录结构，不会重新压缩或改变 PNG 图像内容。</p><p dir="ltr">数据包中的 <code>README.md</code>、<code>DATA_DICTIONARY.md</code>、<code>DATA_SOURCES.md</code> 和 <code>UPLOAD_LAYOUT.md</code> 分别说明目录结构、字段含义、数据来源、软件需求、官方数据获取地址及 TAR 文件组织方式。SHA-256 清单可用于验证文件完整性。</p>