Google Earth Engine – Troubleshooting ReduceRegions Not Returning Values in Google Earth Engine

google-earth-enginegoogle-earth-engine-python-apiland-coverreduceregion

I am trying to reduce the percentage of crops from the CGLS-100 Landcover dataset for a set of points uploaded as asset from a shapefile. This is an operation that I've done tens of times, but for some reason in this particular instance reduceRegions doesn't return any value from the image – only the fields from the original shapefile. I need someone to check my sanity:

# Load landcover
LC = ee.ImageCollection("COPERNICUS/Landcover/100m/Proba-V-C3/Global")\
    .filterDate('2015-01-01', '2015-12-31').first().select(['crops-coverfraction'])

# Load POIs
POI = ee.FeatureCollection('users/sebastienbiasse/barangays')

# Produce buffer around POIs
POI = POI.select(['NAME_1', 'NAME_2', 'NAME_3']).map(lambda feat: ee.Feature(feat).buffer(1000))

# Reduce
col = LC.reduceRegions(collection=POI, reducer=ee.Reducer.mean(), scale=100)

# Remove geometry
col = col.map(lambda feat: feat.setGeometry())

# Request GEE to perform its magics
ee.batch.Export.table.toDrive(collection=col, description=f'LC_crops8', folder='CALDERA', fileFormat="CSV").start()

No error is returned, but the output csv file contains only the fields system:index, NAME_1, NAME_2, NAME_3, and .geo. I expect a field called crops-coverfraction as requested. Other SO answers to similar problems suggest that one source of error could be the size of the featureCollection and mine happens to be rather large (84k points). I get the same problem when using this instead:

ee.batch.Export.table.toDrive(collection=col.limit(1000), description=f'LC_crops8', folder='CALDERA', fileFormat="CSV").start()

Can anyone spot an error?

The bizarre thing is that I reduce another image ~10 lines above this and it works perfectly.

Best Answer

I don't usually work with Python so I worked out a solution directly on the GEE code editor. From what I can gather, there are two things that are messing up your code:

First, I think you might have to limit the number of properties to something ≤5000, though this might not be the case when exporting the results to the Drive.

Second, for reasons unbeknownst to myself, reduceRegions needs a crs projection (even though it's supposedly optional) to work in this case.

Here's the GEE code I wrote to test your python code: https://code.earthengine.google.com/ee203842d4f9020ee718f1f6fb8451b7

Related Solutions

Google Earth Engine – How to Fix reduceRegions Is Not a Function Error

reduceRegions is a method of an Image not an ImageCollection and winter_day_C is an ImageCollection, not an Image.

If you want to reduce spatially to create a single mean value per city by image (date), then map reduceRegions to each image in the ImageCollection.

// Import day and night time land surface temperature (LST) by season from MODIS
var modis = ee.ImageCollection("MODIS/006/MOD11A1");

// WINTER.  
var winter_day = modis.filter(ee.Filter.date('2009-12-01', '2010-03-01')).select('LST_Day_1km');

// Convert to Celsius. 
var winter_day_C = winter_day.map(function(image) {
  return image
    .multiply(0.02)
    .subtract(273.15)
    .reduceRegions({  // <---- reduceRegions on the Image object
      collection: cities,
      reducer: ee.Reducer.mean(),
      scale: 100,
    })
    .copyProperties(image, ['system:time_start']);
});

To avoid the "Error: Unable to use a collection in an algorithm that requires a feature or image. This may happen when trying to use a collection of collections where a collection of features is expected;" do what the error tells you to and "use flatten, or map a function to convert inner collections to features."

// Export 
Export.table.toDrive({
  collection: winter_day_C.flatten(),  // <----- flatten()
  description: '2010_WinterDayLST_export',
  fileFormat: 'CSV'
});

Google Earth Engine – How to Resolve No Output Issue in GEE

The objects are working as intended, they are references to objects in the Google Earth Engine servers.

Read up on the difference between server and client side objects. The doc is javascript based, but applies to the python API as well. You can use getInfo() to convert to a client side object, but do so only if you really need to.

Warning: Use getInfo() with caution!

You shouldn't use getInfo() unless you absolutely need to. If you call getInfo() in your code, Earth Engine will open the container and tell you what's inside, but it will block the rest of your code until that's done. (You can optionally provide a callback function to avoid that, but the better option is to do everything on the server, if possible.)

import ee
ee.Authenticate()
ee.Initialize()

serverside_number = ee.Number(10.3)
clientside_number = serverside_number.getInfo()

print(type(serverside_number))
print(type(clientside_number))

<class 'ee.ee_number.Number'>
<class 'float'>

polygon = ee.Geometry.Polygon(
    [[[-122.092,37.424],
    [-122.086,37.418],
    [-122.079,37.425],
    [-122.085,37.423]]])

polygonArea = polygon.area(maxError=1).getInfo()
print(type(polygonArea))
print(polygonArea)

<class 'float'>
314214.1383491078

Best Answer

Related Solutions

Google Earth Engine – How to Fix reduceRegions Is Not a Function Error

Google Earth Engine – How to Resolve No Output Issue in GEE

Related Question