Grassland Degradation Monitor

Identifies grassland degradation and recovery trends from multi-temporal remote sensing data. Separates climate-driven from management-driven vegetation change and outputs management zones with restoration priorities.

Trigger

Use when the user wants to:

• Identify degrading grassland areas from multi-temporal imagery
• Evaluate restoration effectiveness (e.g., grazing exclusion, reseeding)
• Separate climate-driven from management-driven vegetation change
• Generate management zone maps with restoration priorities
• Conduct BACI (Before-After-Control-Impact) analysis for management assessment
• Monitor grassland health trends over years to decades

CLI Usage

# Basic analysis with default parameters (10 years, Theil-Sen, climate normalized)
python scripts/grassland_degradation_monitor.py --output-dir ./gdm-output

# Custom years and trend method
python scripts/grassland_degradation_monitor.py \
  --years 15 \
  --trend-method ols \
  --output-dir ./gdm-output

# Without climate normalization (raw trend only)
python scripts/grassland_degradation_monitor.py \
  --years 10 \
  --no-climate-normalize \
  --output-dir ./gdm-output

# With custom degradation schema
python scripts/grassland_degradation_monitor.py \
  --years 10 \
  --degradation-schema references/degradation_schema.json \
  --output-dir ./gdm-output

Parameters

Parameter	Default	Description
--input-ndvi	None	Input NDVI time series GeoTIFF (multi-band, optional)
--years	10	Number of years for analysis
--trend-method	theil-sen	Trend estimation: theil-sen, ols, mann-kendall
--climate-normalize	True	Apply climate normalization (residual trend)
--no-climate-normalize	-	Disable climate normalization
--degradation-schema	built-in	Custom degradation schema JSON
--output-dir	./gdm-output	Output directory

Output

File	Description
degradation_status.tif	Raster map of degradation/recovery status codes
trend.tif	Raster map of trend slopes
priority_areas.geojson	Point features for priority restoration areas
management_summary.csv	Area statistics and recommendations per class
timeseries.csv	Mean NDVI, precipitation, temperature per year
request.json	Analysis request metadata
output-manifest.json	Output file inventory and area statistics
qa.json	Quality assurance checks

Degradation Status Codes

Code	Name	Description	Recommendation
3	severe_degradation	NDVI declining >0.02/yr, 3+ years	Immediate restoration
2	moderate_degradation	NDVI declining 0.01-0.02/yr, 3+ years	Priority restoration
1	light_degradation	NDVI declining 0.005-0.01/yr, 3+ years	Preventive management
0	stable	No significant trend	Sustainable use
-1	light_recovery	NDVI increasing 0.005-0.01/yr, 2+ years	Monitor and maintain
-2	moderate_recovery	NDVI increasing 0.01-0.02/yr, 2+ years	Continue current practices
-3	significant_recovery	NDVI increasing >0.02/yr, 2+ years	Success case, replicate

Key Algorithms

Climate Normalization (Residual Trend)

Regresses NDVI against precipitation and temperature anomalies, then computes the trend of residuals. This separates climate-driven variation from management-driven change.

Trend Estimation

• Theil-Sen: Robust non-parametric slope (median of pairwise slopes). Recommended for noisy remote sensing data.
• OLS: Ordinary Least Squares. Efficient for clean data.
• Mann-Kendall: Normalized Kendall tau. Non-parametric trend strength.

Degradation Classification

Combines three factors:

1. Trend slope: Direction and magnitude of change
2. Persistence: Minimum consecutive years meeting threshold
3. Absolute state: Current NDVI level (very low NDVI escalates severity)

BACI Analysis

Before-After-Control-Impact design for management effectiveness: BACI = (treat_after - treat_before) - (ctrl_after - ctrl_before)

Exit Codes

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

Limitations

• Grazing intensity data typically unavailable
• Grassland type differences complicate threshold transfer
• Remote sensing productivity proxies do not replace field biomass measurement
• Climate normalization assumes linear climate-vegetation relationship
• Short time series (< 5 years) reduces trend reliability

References

• Fensholt & Proud 2012, Remote Sensing of Environment (residual trend)
• Ivits et al. 2013, Remote Sensing (European grassland trends)
• BGC ChinaGrass dataset for grassland classification

数据下载

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

python grassland_degradation_monitor.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)。