Calculate Landsat 7 NDVI time series and get the maximum NDVI value and correspond image using Google Earth Engine?

Question

I want to to calculate NDVI time series from 2015 to 2018 using Landsat 7 and get the maximum NAVI value in this period and it's corresponding image index. That is, the horizontal and vertical axes value of this picture, I want to get the Image index in Google Earth Engine, not the time.

code

var landsat7_SR = ee.ImageCollection("LANDSAT/LE07/C01/T1_SR")
                    .filterBounds(geometry); 
var ndvi_list = landsat7_SR.filterDate("1999-01-01", "2018-04-30")
               .map(function(image) { 
                 var ndvi=image.expression('float(b("B4") - b("B3")) / (b("B4") + b("B3"))').rename('ndvi');
                 return image.addBands(ndvi); 
                });

Answer 1

To get the maximum ndvi you could use qualityMosaic function. To be able to know which image ID the pixel comes from is more complicated because as it is not a numerical property, you have to encode it some how. This is how I'd do it, but it's just one way, there could be other ways (may be easier)

l7 = ee.ImageCollection("LANDSAT/LE07/C01/T1_SR")
l7 = l7.filterDate("1999-01-01", "2018-04-30").filterBounds(geometry)
       .map(function(image) { 
          var ndvi=image.expression('float(b("B4") - b("B3")) / (b("B4") + b("B3"))').rename('ndvi');
          ndvi = ndvi.multiply(100)
          return image.addBands(ndvi); 
        })
        // add date band
        .map(function(img){
          var millis = img.date().millis()
          var days = millis.divide(1000).divide(60).divide(60).divide(24).toInt() // 86400000
          var dateband = ee.Image.constant(days).rename('date')
          return img.addBands(dateband)
        })
        // add PATH ROW band
        .map(function(img){
          var path = ee.Number(img.get('WRS_PATH')).toInt()
          var row = ee.Number(img.get('WRS_ROW')).toInt()
          var pathi = ee.Image.constant(path).rename('path')
          var rowi = ee.Image.constant(row).rename('row')
          return img.addBands(pathi).addBands(rowi)
        })
        // cast to Int
        .map(function(img){
          return img.toUint16()
        })

// function to get the image ID from date, path and row
var get_id = function(date, path, row) {
  var d = ee.Date(ee.Number(date).multiply(86400000))
  var i = l7.filterDate(d, d.advance(1, 'day'))
            .filterMetadata('WRS_PATH', 'equals', path)
            .filterMetadata('WRS_ROW', 'equals', row)

  return ee.Image(i.first()).id()
}

// apply qualityMosaic
var mosaic = l7.qualityMosaic('ndvi')
Map.addLayer(mosaic, {bands:['ndvi'], min:0, max:80})

// function to be able to print the image ID when clicking on the Map over the image
var click = function(coords) {
  var lon = coords.lon
  var lat = coords.lat
  var p = ee.Geometry.Point([lon, lat])
  var values = mosaic.reduceRegion(ee.Reducer.first(), p, 30)
  var date = get_id(values.get('date'), values.get('path'), values.get('row'))
  print('ID on point', p, date)
}
Map.onClick(click)

Answer 2

I have given my suggestion regarding the "index problem" at this other question!

The problem can be solved basically in 3 steps (after you applied a cloudmask to Landsat data and attributed the NDVI band):

1. Computing the Maximum NDVI value using .max() on your imageCollection
2. Computing an image containing at each pixel the index of the max-value:
   • by transforming the imageCollection to a list,
     • mapping through it and applying .indexOf() and set the index as a property,
   • then transforming it back to an imageCollection,
     • mapping through it and setting the pixel value equal to the index where the NDVI band values correspond to the maximal NDVI values you computed in step 1, using .where() and .eq()
3. You can then actually compute a mosaic showing the other bands from the original collection that are responsible for the maximal NDVI value of your series:
   • by mapping through a sequence of indexes calling the original image by the index created in step 2, but cutting each image using .updateMask() to the extent its index was responsible for the maximal NDVI value.

In the end you can use the inspector to check out at any point what band values caused which maximal NDVI at which index.

See the full code at link here: https://code.earthengine.google.com/e5537a444207f0e746b39910f62b23bd