How to access image CRS transform from ee.Projection object in Earth Engine [duplicate]

Question

I used the following script to export an image from Earth Engine. I am trying to specify the crs_transform parameter argument in the Export.image.toDrive function using the "transform" property of the image's ee.Projection object. I can see that when I print the image's projection object that there is a field for "transform" containing a valid array element, but the export function throw's an error:

Invalid "argument, crs transform was not a string or array"

How can I retrieve the image's CRS transform?

var geometry = ee.Geometry.Rectangle([-114.3, 51.2, -113.8, 50.8]);
  
var S2_collection = ee.ImageCollection("COPERNICUS/S2_SR_HARMONIZED")
  .filterDate('2019-08-01', '2019-08-30')
  .filterBounds(geometry) 
  // ...map a cloud masking function
    
var S2_mosaic = S2_collection.median()
    
var projection = S2_mosaic.projection();
print(projection)
    
Export.image.toDrive({
  image: S2_mosaic,
  description: 'imageToDriveExample_transform',
  crs: projection.crs,
  crsTransform: projection.transform,  // <-- here's the problem
  region: geometry
});

Answer 1

You are calling projection.transform as if the projection variable is a JavaScript literal object, but it is an Earth Engine server-side ee.Projection object whose properties cannot be accessed by dot or bracket notation. You need to transfer the ee.Projection object client-side to work with it as a JavaScript literal object; you can use the evaluate function to do this:

projection.evaluate(function(proj_obj) {
  Export.image.toDrive({
    image: S2_mosaic,
    description: 'imageToDriveExample_transform',
    crs: proj_obj.crs,
    crsTransform: proj_obj.transform,
    region: geometry });
});

Here is some information on server vs client.

Also, the default CRS transform for image collection reductions (e.g. ee.Image.median) is 1-degree pixel scale in WGS84 EPSG:4326 projection. My guess is that it not what you want. If you'd like your export CRS to match the CRS of some other image in the collection, you should get it from an unaltered image like S2_collection.first().projection(). Here is more information on the default projection.