Google Earth Engine: Converting raster to vector in the same resolution as the original image

Question

I am using Google Earth Engine to reduce a raster region to polygon vectors. What I want to achieve is:

• Obtain a set of polygons each of which matches the grid of measurement in the raster dataset. If the raster dataset's measurements are 0.25 x 0.25 degree, I want to break up the region in the map to polygons that align with these 0.25 x 0.25 degree grids.

Here's how I am doing the raster to vector conversion:

// a "geometry" polygon is imported

var raster_features = ee.ImageCollection('ECMWF/ERA5/DAILY')
                     .select('maximum_2m_air_temperature')
                     .filterBounds(geometry)
                     .filterDate('2017-01-01', '2018-12-31');
                   
var mean_features = raster_features.mean().multiply(10000).toInt();
var scale_to_use = mean.projection().nominalScale();

var vector_features = mean_features.reduceToVectors({
  reducer : null,
  geometry : geometry,

  //scale: 28000,
  scale: scale_to_use,

  crs :mean.projection().getInfo().crs,
  //crsTransform :mean_features.projection().getInfo().transform,

  geometryType: 'polygon',
  labelProperty : 'grid',
  eightConnected: false
});
Map.addLayer(vector_features, {}, 'vector_features')

The resolution of ECMWF/ERA5/DAILY dataset is 0.25 arc degrees. I was expecting the length and breadth of the grids returned by reduceToVectors to be 0.25 deg, but they are 1 deg instead. I want to get polygons in the resolution that the dataset is in, which is 0.25 degree. How do I achieve that?

Answer 1

I found that replacing the definition of scale_to_use variable as given below solves this problem. It may be because the image collection itself has a default scale value of 1 degree, which the images inside the collection have the correct values, i.e., 0.25 degree.

var example_image = raster_features.first();
var scale_to_use = example_image.projection().nominalScale();

Answer 2

@vpk I tried your script and the output was mostly correct, meaning that most grid squares matched individual pixels but there were some rectangular polygons that were created which were made of 2 pixels. Did you encounter this as well?