I am creating a ML classifier to detect floods, but I need to find imagery to train it with. To do this, I am using a Fusion Table with thousands of flood events, their began/ended dates, and polygons of their extent.
I want to quickly find what imagery is available for these floods, then see if it is suitable for training. In GEE, I created a function to map over the fusion table, find Landsat image collections for the date range and bounds of a flood event, then put them together. The "putting together" part is causing me trouble.
I want to be able to set a flood ID property ('dfoID') for the images/image collections. That way I can see which flood they belong to later. Ideally, all the images from each flood event would be grouped into one image collection. However, because I map a Feature Collection, the output is a Feature Collection. I flatten this and then get one image collection but none of the dfoIDs remain on the images.
Here's the link to my code: https://code.earthengine.google.com/3c31496721f4a01c4d9c77933e408266
Any ideas for 1. Yielding an image collection of image collections; Or 2. Making the dfoIDs stick to either the individual images, or to an image collection of images for each flood event?