Skip to content

Commit abe910c

Browse files
fix: sync max-ndvi composite udp description with the markdown docs
1 parent f80f6ec commit abe910c

File tree

2 files changed

+4
-4
lines changed

2 files changed

+4
-4
lines changed

algorithm_catalog/max_ndvi_composite_files/max_ndvi_composite_description.md

+3-3
Original file line numberDiff line numberDiff line change
@@ -18,7 +18,7 @@ of other bands can be minimized to read only selected observations.
1818

1919
The image below shows a typical result over an agricultural area.
2020

21-
![max_ndvi_example.png](max_ndvi_composite_files/max_ndvi_example.png)
21+
![max_ndvi_example.png](./max_ndvi_example.png)
2222

2323
The examples below show typical resource usage figures. They illustrate that the cost varies as a function of the parameters,
2424
and most importantly that it is not possible to linearly extrapolate the cost from one example to another.
@@ -87,9 +87,9 @@ For individual time windows of up to 3 months, the method was efficient up to 10
8787

8888
Artifacts are expected over water and urban areas.
8989

90-
![max_ndvi_water_artifacts.png](max_ndvi_composite_files/max_ndvi_water_artifacts.png)
90+
![max_ndvi_water_artifacts.png](./max_ndvi_water_artifacts.png)
9191

9292
Residual cloud artifacts may be present in the composite, especially for smaller time windows or during cloudy seasons.
9393
The cloud artifacts are caused by the limited capabilities of the default Sentinel-2 cloud detection mechanism to correctly identify all clouds.
9494

95-
![max_ndvi_cloud_artifacts.png](max_ndvi_composite_files/max_ndvi_cloud_artifacts.png)
95+
![max_ndvi_cloud_artifacts.png](./max_ndvi_cloud_artifacts.png)

openeo_udp/examples/max_ndvi_composite/max_ndvi_composite.json

+1-1
Original file line numberDiff line numberDiff line change
@@ -379,7 +379,7 @@
379379
},
380380
"id": "max_ndvi_composite",
381381
"summary": "Max NDVI composite at 10m resolution.",
382-
"description": "# Description\n\nThis algorithm generates a Sentinel-2 based composite for a selected area and temporal extent. By default, the resolution of the output is 10 meters.\n\nThe used compositing method is the \"max-NDVI\" method, which selects the pixel with the highest NDVI value for each pixel location and within the time window.\nThe method falls under the 'rank composite' category, and ensures that selected spectral band values for any individual pixel all come from the same observation.\n\nThe method generates good results for timeseries analytics, but spatially neighbouring pixels may be selected from different observations,\nwhich leads to visual discontinuities in the result.\n\n# Performance characteristics\n\nThe method is computationally efficient, as it only requires the B04, B08 and SCL bands to determine the rank score. Loading \nof other bands can be minimized to read only selected observations.\n\n\n# Examples\n\nThe image below shows a typical result over an agricultural area.\n\n![max_ndvi_example.png](max_ndvi_composite_files/max_ndvi_example.png)\n\nA complete example including STAC metadata is shown here:\n\nhttps://radiantearth.github.io/stac-browser/#/external/s3.waw3-1.cloudferro.com/swift/v1/APEx-examples/max_ndvi_denmark/collection.json\n\nThe processing platform reported these usage statistics for the example:\n\n```\nCredits: 63 \nCPU usage: 47.743,722 cpu-seconds\nWall time: 1.948 seconds\nInput Pixel 10.997,635 mega-pixel\nMax Executor Memory: 3,239 gb\nMemory usage: 154.767.121,977 mb-seconds\nNetwork Received: 1.677.537.930.040 b\n```\n\nThe relative cost is 1 CDSE platform credits per km\u00b2 for a 3 month input window.\nThe cost per input pixel is 0.0057 credits per megapixel.\n\n# Literature references\n\nThe max-ndvi compositing method has been applied to multiple sensors, as described in literature:\n\nThis publication describes characteristics of the method when applied to AVHRR data:\nhttps://www.tandfonline.com/doi/abs/10.1080/01431168608948945\n\nThis publication applied it to Landsat data, for cropland estimation:\nhttps://www.nature.com/articles/s43016-021-00429-z\n\n# Known limitations\n\nThe method uses a vegetation index as scoring metric to determine the best pixel, making it only suitable for land applications. \nBare or urban areas may not be well represented in the composite. \n\nIt favours the observation which is least contaminated by atmospheric effects, but does not guarantee a fully uncontaminated composite.\n\nFor individual time windows of up to 3 months, the method was efficient up to 100x100km areas. For larger areas of interest, we recommend splitting the area into smaller tiles.\n\n\n# Known artifacts\n\nArtifacts are expected over water and urban areas.\n\n![max_ndvi_water_artifacts.png](max_ndvi_composite_files/max_ndvi_water_artifacts.png)\n\nResidual cloud artifacts may be present in the composite, especially for smaller time windows or during cloudy seasons.\nThe cloud artifacts are caused by the limited capabilities of the default Sentinel-2 cloud detection mechanism to correctly identify all clouds.\n\n![max_ndvi_cloud_artifacts.png](max_ndvi_composite_files/max_ndvi_cloud_artifacts.png)",
382+
"description": "# Description\n\nThis algorithm generates a Sentinel-2 based composite for a selected area and temporal extent. By default, the resolution of the output is 10 meters.\n\nThe used compositing method is the \"max-NDVI\" method, which selects the pixel with the highest NDVI value for each pixel location and within the time window.\nThe method falls under the 'rank composite' category, and ensures that selected spectral band values for any individual pixel all come from the same observation.\n\nThe method generates good results for timeseries analytics, but spatially neighbouring pixels may be selected from different observations,\nwhich leads to visual discontinuities in the result.\n\n# Performance characteristics\n\nThe method is computationally efficient, as it only requires the B04, B08 and SCL bands to determine the rank score. Loading\nof other bands can be minimized to read only selected observations.\n\n\n# Examples\n\nThe image below shows a typical result over an agricultural area.\n\n![max_ndvi_example.png](./max_ndvi_example.png)\n\nThe examples below show typical resource usage figures. They illustrate that the cost varies as a function of the parameters,\nand most importantly that it is not possible to linearly extrapolate the cost from one example to another.\n\n\n## 3-month composite over Denmark\n\nA complete example including STAC metadata is shown here:\n\nhttps://radiantearth.github.io/stac-browser/#/external/s3.waw3-1.cloudferro.com/swift/v1/APEx-examples/max_ndvi_denmark/collection.json\n\nThe processing platform reported these usage statistics for the example:\n\n```\nCredits: 63\nCPU usage: 47.743,722 cpu-seconds\nWall time: 1.948 seconds\nInput Pixel 10.997,635 mega-pixel\nMax Executor Memory: 3,239 gb\nMemory usage: 154.767.121,977 mb-seconds\nNetwork Received: 1.677.537.930.040 b\n```\n\nThe relative cost is 1 CDSE platform credits per km² for a 3 month input window.\nThe cost per input pixel is 0.0057 credits per megapixel.\n\n## 15-month composite over Denmark\n\nIn a second example, a longer compositing window was tested, generating a 3-band result. Here we see a lower cost per km², but a similar cost per input\npixel.\n\n```\nCredits: 189\nCPU usage: 77.621,979 cpu-seconds\nWall time: 5.499 seconds\nInput Pixel: 31.494,448 mega-pixel\nMax Executor Memory: 4,332 gb\nMemory usage: 564.094.942,143 mb-seconds\nNetwork Received: 872.636.866.126 b\n```\n\nThe relative cost is 0.03 CDSE platform credits per km² for a 15 month input window.\nThe cost per input pixel is 0.006 credits per megapixel.\n\n# Literature references\n\nThe max-ndvi compositing method has been applied to multiple sensors, as described in literature:\n\nThis publication describes characteristics of the method when applied to AVHRR data:\nhttps://www.tandfonline.com/doi/abs/10.1080/01431168608948945\n\nThis publication applied it to Landsat data, for cropland estimation:\nhttps://www.nature.com/articles/s43016-021-00429-z\n\n# Known limitations\n\nThe method uses a vegetation index as scoring metric to determine the best pixel, making it only suitable for land applications.\nBare or urban areas may not be well represented in the composite.\n\nIt favours the observation which is least contaminated by atmospheric effects, but does not guarantee a fully uncontaminated composite.\n\nFor individual time windows of up to 3 months, the method was efficient up to 100x100km areas. For larger areas of interest, we recommend splitting the area into smaller tiles.\n\n\n# Known artifacts\n\nArtifacts are expected over water and urban areas.\n\n![max_ndvi_water_artifacts.png](./max_ndvi_water_artifacts.png)\n\nResidual cloud artifacts may be present in the composite, especially for smaller time windows or during cloudy seasons.\nThe cloud artifacts are caused by the limited capabilities of the default Sentinel-2 cloud detection mechanism to correctly identify all clouds.\n\n![max_ndvi_cloud_artifacts.png](./max_ndvi_cloud_artifacts.png)",
383383
"parameters": [
384384
{
385385
"name": "spatial_extent",

0 commit comments

Comments
 (0)