Geoskill: 尾矿库风险评估

Screen tailings dam bodies, reservoir areas, catchments, and downstream exposure using remote sensing change detection. Produce patrol priorities based on hazard, exposure, and evidence.

Prerequisites / 先准备 X 文件

⚠️ 必读 — 本 skill 不属于即用型，需要先准备特定文件才能跑。

本 skill 需要 设施清单 JSON（geometry + 属性）。--bbox 模式下 DEM 和 Sentinel-2 时序自动从 MPC 下载。

👉 完整教程见仓库根目录 PREREQUISITES.md 1.6 节。

先准备 X 文件：先准备一份 facilities.json（可手写小样本），其他自动下载。

快速试跑命令：

bash

python tailings_dam_risk.py --bbox 116.30,39.85,116.45,39.95 --date-range 2023-06-01,2023-09-30 --facilities facilities.json --output-dir ./tdr

Tailings Dam Risk

Screens tailings dam facilities using remote sensing and DEM analysis. Produces risk rankings and patrol priorities based on multi-temporal change detection, catchment analysis, and downstream exposure assessment.

Trigger

Use when the user wants to:

• Screen tailings dam facilities for safety risks
• Detect water surface changes in tailings reservoirs
• Analyze upstream catchment contributions
• Estimate downstream impact zones from potential dam failure
• Identify exposed objects (buildings, roads, population) downstream
• Generate risk reports for patrol prioritization
• Assess deformation patterns from InSAR data

CLI Usage

bash

# Basic screening with synthetic data (demo)
python scripts/tailings_dam_risk.py --output-dir ./tdr-output

# With custom facilities file
python scripts/tailings_dam_risk.py \
  --facilities facilities.json \
  --output-dir ./tdr-output

# With DEM and water masks
python scripts/tailings_dam_risk.py \
  --dem-file dem.tif \
  --water-masks water_2019.tif water_2021.tif water_2023.tif \
  --output-dir ./tdr-output

# Full analysis with all options
python scripts/tailings_dam_risk.py \
  --facilities facilities.json \
  --dem-file dem.tif \
  --water-masks water_2019.tif water_2021.tif \
  --deformation-data insar_deformation.tif \
  --exposure-objects downstream_objects.json \
  --rainfall-scenario 500yr \
  --runout-method simplified \
  --deformation-threshold 15.0 \
  --exposure-radius 8000.0 \
  --output-dir ./tdr-output

数据下载

本 skill 可自动从 Microsoft Planetary Computer 下载 DEM 和 Sentinel-2 (无需 API key):

bash

# 自动下载 cop-dem-glo-30 DEM + sentinel-2-l2a 水体变化快照
python scripts/tailings_dam_risk.py \
    --bbox 116.30,39.85,116.45,39.95 \
    --date-range 2023-06-01,2023-09-30 \
    --output-dir ./tdr-auto

下载的内容:

• Copernicus DEM GLO-30 (cop-dem-glo-30) — 替换 --dem-file
• Sentinel-2 L2A (sentinel-2-l2a) — 2 个时相的影像，生成水体变化，替换 --water-masks

下载元数据（data_source, fetched_at, collection, dem_path, sentinel2_paths）会写入 output-manifest.json。

Parameters

Parameter	Default	Description
--facilities	None	Path to facilities JSON file with geometry and attributes
--dem-file	None	Path to DEM GeoTIFF file
--water-masks	None	Input water mask GeoTIFF files (ordered by time)
--deformation-data	None	Path to InSAR deformation rate GeoTIFF (mm/yr)
--exposure-objects	None	Path to downstream exposure objects JSON
--rainfall-scenario	100yr	Rainfall scenario: 100yr, 500yr, 1000yr
--runout-method	simplified	Runout method: simplified, energy_line
--deformation-threshold	10.0	Deformation threshold in mm/yr
--exposure-radius	5000.0	Exposure analysis radius in meters
--risk-rules	None	Path to custom risk scoring rules JSON
--output-dir	./tdr-output	Output directory

Output

File	Description
facility_changes.geojson	Water surface change polygons (expansion/contraction)
catchments.geojson	Upstream catchment boundaries
screening_zones.geojson	Runout/impact zones
downstream_exposure.csv	Downstream exposure statistics
risk_report.html	Risk screening report
request.json	Analysis request metadata
dataset-manifest.json	Dataset inventory and quality info
output-manifest.json	Output file inventory and results
qa.json	Quality assurance checks
run.log	Execution log

Risk Levels

Code	Name	Description
3	HIGH	Combined score >= 70
2	MEDIUM	Combined score >= 40
1	LOW	Combined score >= 20
0	UNKNOWN	Combined score < 20

Risk Components

Hazard (40%)

Based on dam height, capacity, water surface change, catchment size, and runout potential.

Exposure (35%)

Based on number of objects in impact zone and exposure area.

Evidence (25%)

Based on observed water surface changes and catchment characteristics.

Key Algorithms

Water Surface Change Detection

Compares binary water masks between time periods to detect expansion and contraction areas. Calculates change percentage and extracts change polygons.

Catchment Analysis

Uses D8 flow direction algorithm on DEM to compute flow accumulation and delineate upstream watershed contributing to the dam location.

Runout Zone Estimation

Two simplified methods for screening:

• Simplified: Uses reach angle concept (height / tan(alpha))
• Energy_line: Uses H/L ratio to estimate flow extent

Both methods are SCREENING-ONLY and do not replace engineering analysis.

Downstream Exposure

Analyzes objects within potential impact zone. Counts buildings, roads, and other infrastructure at risk.

Deformation Analysis

Processes InSAR deformation data to identify areas with significant movement exceeding threshold.

Rainfall Scenario

Estimates peak flow using rational method (Q = C * i * A) for different return period events.

Exit Codes

Code	Meaning
0	Success
2	Argument error
3	Dependency missing
6	Data validation failure
7	Processing failure

Limitations

• SCREENING ONLY: Results are auxiliary analysis only and do NOT replace dam safety engineering assessment
• Simplified runout models are not substitutes for detailed engineering analysis
• Facility data completeness affects result confidence
• Free satellite imagery may not detect fine cracks or small deformations
• All results must be reviewed by qualified engineers before making safety decisions

Data Requirements

• DEM: Any resolution (30m SRTM or 12m PALSAR recommended)
• Water masks: Binary (0/1) rasters from NDWI/MNDWI or classification
• Deformation data: InSAR-derived rate map (mm/yr, optional)
• Facilities: GeoJSON with geometry and optional dam_height, capacity

References

• US Dam Safety Guidelines (FEMA 1998)
• ICOLD Bulletin 153: Tailings Dams - Risk of Dangerous Occurrences
• Hungr et al. 1984: Debris flow runout estimation

数据下载

本 skill 可自动从 Microsoft Planetary Computer 下载数据 (无需 API key):

bash

python tailings_dam_risk.py --bbox 116,39,117,40 --date-range 2024-06-01,2024-06-30 --output-dir <tmp>

• --bbox W,S,E,N: WGS-84 边界框 (西, 南, 东, 北)
• --date-range START,END: 日期范围 (YYYY-MM-DD,YYYY-MM-DD)
• --aoi-file <path.geojson>: 替代 --bbox 的 GeoJSON 多边形
• --cache-dir <path>: 缓存目录 (默认 ~/.geoskill_cache)

当用户只给 --bbox + --date-range (没有 --dem) 时，skill 自动下载数据。 当用户给 --dem 时，走原文件路径 (向后兼容)。