How to Mosaic an ImageCollection before downloading? - OpenET Dataset

[jamayor2]

Hello all,

I need to download the eeMETRIC Dataset from OpenET (monthly ET) for a specific watershed and I was following the Exporting Data from Google Earth Engine Notebook. I noticed that for each month, I got two scenes in my area of interest (i.e, 12s_20160101_20160131 and 13s_20160101_20160131). Before moving to mosaic in GIS, I am wondering if there is any option to mosaic them before downloading using geemap. I tried, but since is an Image Collection I was not able to figure it out.

Thank you for your help, the code is below:

import ee
import geemap
import os

def clipcol(image):
    clipimage = image.clip(geometry)
    return clipimage

geometry = shp.geometry()

eeMETRIC = ee.ImageCollection('OpenET/EEMETRIC/CONUS/GRIDMET/MONTHLY/v2_0') \
            .filterBounds(geometry) \
            .select('et') \
            .map(clipcol) 

count = int(eeMETRIC.size().getInfo())
print("Total number of images: {}".format(count))

for i in range(84, count):
    image = ee.Image(eeMETRIC.toList(count).get(i))
    region = image.geometry() # roi
    scale = image.projection().nominalScale() # specify the image resolution

    task_config = {
        'folder': 'eeMETRIC_SanJuan_Clipped', # output Google Drive folder
        'region': region,     # specify roi
        'scale': scale,       # specify image resolution
        'crs': 'EPSG:4269',
        'maxPixels': 1.0E13,
        'fileFormat': 'GeoTIFF'
    }
    
    description = image.get('system:index').getInfo()
    print('Exporting {}/{}: {}'.format(i+1, count, description))
    task = ee.batch.Export.image.toDrive(image, description, **task_config)
    task.start()

[giswqs]

Something like this:

https://book.geemap.org/chapters/09_timelapse.html#creating-time-series

collection = ee.ImageCollection('OpenET/EEMETRIC/CONUS/GRIDMET/MONTHLY/v2_0') \
            .filterBounds(geometry) \
            .select('et') \
images = geemap.create_timeseries(collection, start_date, end_date, region=geometry, frequency='day', reducer='mean')
geemap.download_ee_image_collection(images, out_dir, scale=30)

[jamayor2]

Thank you, Dr. Qiusheng, for your prompt response! I got a 'User memory limit exceeded' error when attempting to export monthly files for the entire year, given that the basin area is approximately 60,000 km². However, when I specified the start and end dates for only six months, the export process was successful. Do you have any recommendations for addressing these memory issues without losing spatial resolution? Thanks again for your assistance!

[jdbcode]

#1781 might help. It uses the parallelScale parameter of ee.ImageCollection.reduce which can divide your task up into smaller pieces, which can reduce the memory needed for each worker. It does a couple of other things that can help with memory, depending on what defines the region and the size of the collections being composited. It's quite possible that you will not be able to call download_ee_image_collection on composite and will need to export it as a batch task to asset, drive, or cloud. The batch stack has characteristics that often make it more robust for large tasks than the interactive stack. You also might need to break up the region into smaller parts and export them into an image collection and later calls ee.ImageCollection.mosaic on them; ee.Geometry.coveringGrid can be helpful for breaking up large regions into multiple exports.