Save-Best Joins

To save only the best match for each element in a collection, use an ee.Join.saveBest(). The saveBest() join functions in an equivalent way to the saveAll() join, except for each element in the primary collection, it saves the element from the secondary collection with the best match. Unmatched elements in the primary collection are dropped. Suppose the intention is to find a meteorological image closest in time to each Landsat image in the primary collection. To perform this join, the ee.Filter must be redefined for a single join condition (combined filters will not work with saveBest() since it is ambiguous how to combine ranks from multiple sub-Filters):

// Load a primary collection: Landsat imagery.
var primary = ee.ImageCollection('LC8_L1T_TOA')
    .filterDate('2014-04-01', '2014-06-01')
    .filterBounds(ee.Geometry.Point(-122.092, 37.42));

// Load a secondary collection: GRIDMET meteorological data
var gridmet = ee.ImageCollection('IDAHO_EPSCOR/GRIDMET');

// Define a max difference filter to compare timestamps.
var maxDiffFilter = ee.Filter.maxDifference({
  difference: 2 * 24 * 60 * 60 * 1000,
  leftField: 'system:time_start',
  rightField: 'system:time_start'

// Define the join.
var saveBestJoin = ee.Join.saveBest({
  matchKey: 'bestImage',
  measureKey: 'timeDiff'

// Apply the join.
var landsatMet = saveBestJoin.apply(primary, gridmet, maxDiffFilter);

// Print the result.

Note that a saveBest() join defines the name of the property with which to store the best match (‘bestImage’) and the name of the property with which to store the goodness of the match metric (‘timeDiff’). Inspection of the results indicates that a matching DAYMET image has been added to the property bestImage for each Landsat scene in the primary collection. Each of these DAYMET images has the property timeDiff indicating the time difference in milliseconds between the DAYMET image and the Landsat image, which will be minimum among the DAYMET images passing the condition in the filter.

Enviar comentarios sobre…

Google Earth Engine API