stefanbrand / masterdatacube Goto Github PK

Grids covering the whole UTM stripe from equator to almost poles and from left projection bounds rounded to the nearest eastling of 10[m] resolution to match Sentinel-2 pixel grid. (UTM North projection definition projected Bounds: 166021.4431, 0.0000, 833978.5569, 9329005.1825)

• 10[m] resolution at zoom9
• buffer of 50[km] meters below equator (the Sentinel-2 Scenes in UTM North go lower than equator so let's give it some more space)

GridDefinition(
    grid=None,
    shape=[8, 1],
    bounds=[166020.0, -50000, 1476740, 10435760],
    srs={"epsg": 32632},
    is_global=False
)

Zoom0 tiles over the whole grid

Sentinel-2 Scene in UTM-32 with zoom9 tile edge fitting the 10[m] pixel grid:

50[km] buffer below equator

Contrary to the synchronous Sentinel Hub, requesting userdata.json doesn't work for Sentinel Hub Batch API. After removing

{
    "identifier": "userdata",
    "format": {
        "type": "application/json"
    }
}

from the requested responses, the process doesn't return FAILED anymore.

The API reference also states that requesting userdata is permitted: https://docs.sentinel-hub.com/api/latest/reference/#operation/createNewBatchProcessingRequest

In order to make results comparable, the employed hardware must be made transparent.

I cancelled this request after almost 4 days:

'id': '4293b3d0-afba-4df3-b7c9-6b0ed9725ea4',
'processRequest': {'input': {'bounds': {'bbox': [16.446445736463346,
    47.680841561177864,
    16.49776618971013,
    47.72587417451863],
   'geometry': None,
   'properties': {'crs': 'http://www.opengis.net/def/crs/OGC/1.3/CRS84'}},
  'data': [{'location': 'AWS:eu-central-1',
    'dataFilter': {'timeRange': {'from': '2017-09-01T00:00:00Z',
      'to': '2018-11-30T00:00:00Z'},
     'mosaickingOrder': 'mostRecent',
     'maxCloudCoverage': 100.0},
    'processing': None,
    'tileList': None,
    'id': None,
    'type': 'S2L2A'}]},
 'output': {'width': None,
  'height': None,
  'resx': None,
  'resy': None,
  'responses': [{'identifier': 'B02', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B03', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B04', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B05', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B06', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B07', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B08', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B8A', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B11', 'format': {'type': 'image/tiff'}},
   {'identifier': 'B12', 'format': {'type': 'image/tiff'}},
   {'identifier': 'NDVI', 'format': {'type': 'image/tiff'}},
   {'identifier': 'GNDVI', 'format': {'type': 'image/tiff'}},
   {'identifier': 'BNDVI', 'format': {'type': 'image/tiff'}},
   {'identifier': 'CVI', 'format': {'type': 'image/tiff'}},
   {'identifier': 'NDSI', 'format': {'type': 'image/tiff'}},
   {'identifier': 'NDWI', 'format': {'type': 'image/tiff'}}]},
 'dataProduct': None,
[...]
 'tilingGridId': 0,
[...]
 'status': 'PROCESSING',
 'userAction': 'CANCEL',
[...]
 'resolution': 10.0,
[...]
 'created': '2020-04-24T08:33:09.632Z',
 'valueEstimate': 8519.558595300987,
 'tileCount': 1,
 'tileWidthPx': 2004,
 'tileHeightPx': 2004}

As you can see in the screenshot below, the FLOAT32 variable CVI still only contains integers. In Sentinel Hub Process API everything works as expected.

The crs parameter doesn't seem to influence the output CRS. The result of the Batch API is always in UTM.

"processRequest": {
  "input": {
    "bounds": {
      "bbox": [],
      "geometry": {},
      "properties": {
        "crs": "http://www.opengis.net/def/crs/OGC/1.3/CRS84"
      }
    },
   "data": []
  },
  "output": {},
  "evalscript": "string"
},

Also the documentation specifically mentions UTM, but doesn't talk about other CRS.

Note that the processing results might be in different UTM zones (and thus different CRSs) if your area of interest intersects with more than one UTM zone.

The tilingGridId parameter also doesn't affect output CRS.

Polling in an interactive Python notebook is probably not the best idea. During polling, which can take hours, the notebook is busy and interactivity is gone. There must be better ways to do this.

I want to retrieve the total execution time of my request. Ideally the REST response schema for a Request ID would contain two elements:

• dateCreated and
• dateFinished

This way execution time could be retrieved at any point in the future (as long as the request is stored server-side).

Besides, the oauth2 token expires after 1 hour and has to be refetched periodically. Also, I'm dependent on my computer not going to stand-by. The Jupyter Lab tab has to be open at all times.

In order to compare performance with existing tools, the exact definition of the tiling grids used in Sentinel Hub Batch API is relevant information. Below is a table of available tiling grids:

Source

If I specify only one input band, the result is empty (all pixels contain value 0). If I run the same request with two input bands, output is as expected (pixels populated with real values).

According to the API docs there is a query parameter search when GETting a list of all batch requests.

I don't understand how I can use this parameter.