Geoskill: Blue Carbon Assessment

Blue carbon ecosystem assessment - identify mangrove, salt marsh, and seagrass ecosystems, estimate carbon stocks, changes, and uncertainty using IPCC default factors or project-specific data.

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 生态系统分类栅格（GeoTIFF，class=1=红树林/2=盐沼/3=海草）。数据来源：GMW v3 / GLWD / 国家级海岸带调查。

👉 完整教程（含数据源 + 完整工作流）见仓库根目录 PREREQUISITES.md 1.1 节。

先准备 X 文件：先用 --synthetic 跑通，再换成真实数据。

快速试跑命令：

bash

python blue_carbon_assessment.py --synthetic --output-dir ./test

Blue Carbon Assessment

Identifies and maps blue carbon ecosystems (mangrove, salt marsh, seagrass), estimates carbon stocks by pool (above-ground biomass, below-ground biomass, soil), detects changes between time periods, and generates stratified sampling plans for field validation.

Trigger

Use when the user wants to:

• Estimate blue carbon stocks in coastal ecosystems
• Analyze carbon stock changes due to ecosystem loss or restoration
• Compare carbon densities across ecosystem types (mangrove, salt marsh, seagrass)
• Generate sampling plans for blue carbon field campaigns
• Produce screening-level carbon assessments for project planning

CLI Usage

bash

# Synthetic demo mode (no input files needed)
python scripts/blue_carbon_assessment.py --output-dir ./bca-output

# Assess only mangroves
python scripts/blue_carbon_assessment.py --ecosystem-type mangrove --output-dir ./bca-output

# With custom soil depth
python scripts/blue_carbon_assessment.py --soil-depth 0_200cm --output-dir ./bca-output

# Change detection with specific years
python scripts/blue_carbon_assessment.py --years "2015,2023" --output-dir ./bca-output

# With custom carbon factors
python scripts/blue_carbon_assessment.py --carbon-factors ./my_factors.json --output-dir ./bca-output

Parameters

Parameter	Default	Description
--ecosystem-raster	None	Path to ecosystem classification GeoTIFF (synthetic if omitted)
--ecosystem-type	all	Ecosystem type filter (mangrove/salt_marsh/seagrass/all)
--boundary	strict	Boundary type (strict/inclusive)
--years	2020,2023	Comma-separated years for change detection
--carbon-factors	None	Path to custom carbon factors JSON
--soil-depth	0_100cm	Soil depth for carbon accounting (0_30cm/0_100cm/0_200cm)
--uncertainty	0.95	Confidence level for uncertainty interval (0.5-0.99)
--output-dir	./bca-output	Output directory

Output

File	Description
report.html	Human-readable blue carbon assessment report
ecosystem_extent.npy	Ecosystem classification raster (synthetic mode)
blue_carbon_stock.npy	Carbon stock per pixel (tC)
change.geojson	Change analysis results
carbon_summary.csv	Per-ecosystem carbon stock summary
sampling_plan.geojson	Stratified random sampling plan
request.json	Analysis request metadata
dataset-manifest.json	Dataset inventory
output-manifest.json	Output file inventory
qa.json	Quality assurance checks

Key Algorithms

Carbon Stock Calculation

Total stock = Area x (AGB_density + BGB_density + Soil_density) where AGB = above-ground biomass, BGB = below-ground biomass. Soil carbon is depth-specific (0-30cm, 0-100cm, 0-200cm).

Uncertainty Propagation

Root sum of squares (RSS) of independent errors from area mapping (assumed 15% for raster-based), biomass factors, and soil factors. Confidence intervals use Z-scores (1.96 for 95%).

Change Detection

Compares ecosystem classification between two time periods. Loss = all carbon pools released (simplified). Gain = biomass accumulation only. Stable area = soil carbon accumulation at published rates.

Sampling Design

Stratified random sampling with Neyman-like allocation. Sample size per stratum: n = (Z^2 * CV^2) / E^2 where CV = coefficient of variation, E = margin of error.

Exit Codes

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

Important Limitations

• Default factors are screening-level: IPCC Tier 1 default carbon densities are global averages. Project-level assessments REQUIRE site-specific field sampling.
• Soil carbon dominates: Typically 70-90% of total blue carbon stock. Soil depth choice dramatically affects total estimates.
• Simplified change model: Loss assumes instant release of all pools. Gain assumes biomass only (no soil carbon recovery in first years).
• No tidal/seasonal adjustment: Ecosystem extent from remote sensing must account for tidal stage at time of image acquisition.
• Seagrass mapping uncertainty: Submerged vegetation is difficult to map from optical imagery; SAR or acoustic methods preferred.

References

• IPCC 2013 Wetlands Supplement (Tier 1 default factors)
• Fourqurean et al. 2012 (seagrass carbon)
• Alongi 2014 (mangrove carbon cycling)
• Ouyang et al. 2017 (salt marsh accumulation)

数据下载

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

bash

python blue_carbon_assessment.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 --image) 时，skill 自动下载数据。 当用户给 --image 时，走原文件路径 (向后兼容)。