Install
openclaw skills install @ruiduobao/geoskill-wetland-mappingopenclaw skills install @ruiduobao/geoskill-wetland-mappingMulti-source fusion wetland type mapping. Combining four co-registered raster layers (band order: NDWI/MNDWI, NDVI, DEM, SAR backscatter σ⁰), a physical-rule decision tree classifies each pixel into open water, swamp, mudflat, or non_wetland, with priority water > swamp > mudflat.
Example criteria: open water requires high NDWI and very low SAR (specular reflection at the water surface); swamp requires high NDVI (wetland vegetation), low-lying terrain, and relatively low SAR; mudflat requires low NDVI, wet conditions (relatively high NDWI), low-lying terrain, and relatively low SAR. The DEM is automatically normalized to low-lying degree (0 = lowest).
Typical applications: wetland resource baseline surveys, coastal/lakeshore wetland mapping, and preliminary surveys of mangrove and swamp distribution. The synthetic mode generates multi-source data containing all four wetland types, with overall classification accuracy >0.95 against the injected ground truth.
pip install 'numpy' 'rasterio' 'scipy'
python geoskill-wetland-mapping.py --bbox 116.0 39.0 117.0 40.0
python geoskill-wetland-mapping.py --bbox 116.0 39.0 117.0 40.0 --synthetic --output-dir ./out
python geoskill-wetland-mapping.py --input fused_4band.tif --output-dir ./out
Input band order: [NDWI, NDVI, DEM, SAR(dB)]; the DEM is automatically normalized to [0,1].
python geoskill-wetland-mapping.py --bbox 121.0 31.0 122.0 32.0 --synthetic --quiet --output-dir ./out
| File | Format | Description |
|---|---|---|
wetland_class.tif | GeoTIFF (int32) | 0=non_wetland, 1=water, 2=swamp, 3=mudflat |
area_stats.json | JSON | Per-class area/percentage + total wetland area |
output-manifest.json | JSON | Run manifest (includes accuracy QA in synthetic mode) |
--synthetic offline synthesis of multi-source scenarios (no network, no account required).--synthetic mode requires no network at all.MIT
多源融合的湿地类型制图。组合四个共配准栅格层(波段顺序:NDWI/MNDWI、 NDVI、DEM、SAR 后向散射 σ⁰),按物理规则决策树把每个像元分为 开放水域(water)、沼泽(swamp)、滩涂(mudflat)与非湿地(non_wetland), 优先级 water > swamp > mudflat。
判据示例:开放水域要求 NDWI 高且 SAR 极低(水面镜面反射);沼泽要求 NDVI 高(湿生植被)、地形低洼且 SAR 偏低;滩涂要求 NDVI 低、湿润 (NDWI 偏高)、低洼且 SAR 偏低。DEM 自动归一化为低洼度(0=最低)。
典型应用:湿地资源本底调查、滨海/湖滨湿地制图、红树林与沼泽分布摸底。 合成模式生成含四类湿地的多源数据,分类总体精度与注入真值 >0.95。
pip install 'numpy' 'rasterio' 'scipy'
python geoskill-wetland-mapping.py --bbox 116.0 39.0 117.0 40.0
python geoskill-wetland-mapping.py --bbox 116.0 39.0 117.0 40.0 --synthetic --output-dir ./out
python geoskill-wetland-mapping.py --input fused_4band.tif --output-dir ./out
输入波段顺序:[NDWI, NDVI, DEM, SAR(dB)];DEM 会被自动归一化到 [0,1]。
python geoskill-wetland-mapping.py --bbox 121.0 31.0 122.0 32.0 --synthetic --quiet --output-dir ./out
| 文件 | 格式 | 说明 |
|---|---|---|
wetland_class.tif | GeoTIFF (int32) | 0=非湿地, 1=水域, 2=沼泽, 3=滩涂 |
area_stats.json | JSON | 逐类面积/占比 + 湿地总量 |
output-manifest.json | JSON | 运行清单(合成模式含精度 QA) |
--synthetic 离线合成多源场景(无需网络、无需账号)。--synthetic 模式完全无网络。MIT