Pixel values in the masked area change from 'masked' to 0 during image export using geemap.ee_export_image_to_drive()

[DKchenliyuan]

Hello everyone!

I'm dealing with the Global Friction Surface and encountering an issue where the pixel values in the masked area 'automatically' change from 'masked' to 0 during image export using geemap.ee_export_image_to_drive(). This seems to happen when I convert the data type to Int.

Here are the relevant codes and images:

import ee
import geemap
ee.Initialize()

# Data preparation
roi = ee.Geometry.Rectangle([0,60,180,85], 'EPSG:4326', False)
friction_map = ee.Image("Oxford/MAP/friction_surface_2019")
ERA_Land = ee.ImageCollection("ECMWF/ERA5_LAND/MONTHLY_AGGR")
export_projection = ERA_Land.first().projection()
land_area = ERA_Land.first().select('skin_temperature').mask()

# build a fine-resolution mask
friction_motorized_mask = friction_map.select('friction').neq(friction_map.select('friction_walking_only'))

# convert it to the coarse-resolution mask
friction_motorized_mask_reproj = friction_motorized_mask.reduceResolution(**{'reducer': ee.Reducer.mean(), 'bestEffort': True}) \
            .reproject(**{'crs': export_projection, 'scale': export_projection.nominalScale()}) \
            .gte(0.1)

# calculate the mean value in the fine scale, reduce it to coarse resolution, and mask it by coarse-resolution mask
# convert the data type to Int
friction_motorized_reproj = friction_map.select('friction').updateMask(friction_motorized_mask) \
    .reduceResolution(**{'reducer': ee.Reducer.mean(), 'bestEffort': True}) \
    .reproject(**{'crs': export_projection, 'scale': export_projection.nominalScale()}) \
    .multiply(1e7) \
    .toInt() \
    .updateMask(friction_motorized_mask_reproj) \
    .updateMask(land_area) \
    .clip(roi)

Map = geemap.Map()
Map

visParams = {
  'min': 0.0022*1e7,
  'max': 0.04*1e7,
  'palette': [
    '313695', '4575b4', '74add1', 'abd9e9', 'e0f3f8', 'ffffbf', 'fee090',
    'fdae61', 'f46d43', 'd73027', 'a50026'
  ]
}
Map.addLayer(friction_motorized_reproj, visParams, 'friction_motorized_reproj')

geemap.ee_export_image_to_drive(friction_motorized_reproj, \
                                description = 'test_int', \
                                region = roi, \
                                crs = export_projection, scale = export_projection.nominalScale(), \
                                maxPixels = 1e13)

In the Map display, pixel values in the masked area are 'None':

However, the exported image, when viewed in QGIS, displays 0 in the masked area:

Then, I remove two lines related to data type conversion:

 #   .multiply(1e7)
 #   .toInt()

Results in the pixel values of the masked area in the exported image being 'None', as expected:

I'm puzzled about the cause of this issue. Any comments on this greatly appreciated.

[jdbcode]

You should be able to use the formatOptions.noData option to set the nodata value in GeoTIFF exports.

Based on your example, you'd do something like this:

# unmask the masked pixels - set to -9999 in this example
nodata = -9999
test_int = test_int.unmask(nodata)

geemap.ee_export_image_to_drive(friction_motorized_reproj, \
                                description = 'test_int', \
                                region = roi, \
                                crs = export_projection, scale = export_projection.nominalScale(), \
                                maxPixels = 1e13, \
                                formatOptions = {"noData": nodata})  # <---- add this to set noData!

Here is a toy example you can play with from the JS Code Editor (sorry, I don't have a py/geemap version): https://code.earthengine.google.com/443a155ef777970c727314318988e26e