2

I'm trying to perform a time-series analysis of Sentinel-2 satellite imagery over Water, and to do that I've been muddling through using functions to perform analysis on images over an Image Collection.

I previously got further with my analyses with individual images using .mosaic(), but can't use that for the time-series. Using this earlier question and the GEE Guides I've been able to run NDVI and NDWI over a test area, create individual masks based on these, and 'Add' (Combine/Overlay/Union?) them into a new layer that has all the areas covered by both masks.

The issue I've run into is that when I have the final combined Mask Layer, the unmasked imagery doesn't match either the combined Mask, or the individual Masks. I suspect the imagery with the mask applied is an entirely different image from the original also, as much of the clipped image includes clouds, which the original L1C image doesn't contain.

For instance:

The Original Image Source Imagery (Sentinel L1C) The Combined NDVI/NDWI Mask (i.e. Areas of Water matching '1', red on the Mask) The Combined Mask (L1C Mask The same Combined Mask applied to the imagery (L1C Applied, in White-Gray), compared to the L1C Mask above The Combined Mask applied to the Imagery, entirely different Mask area used

So, what is causing a different Mask area to appear instead? What section of my code is causing this error, and is there a fix or easier alternative to the way that I have to apply a mask to each individual image in the Image Collection?

I have included the code below, as well as a GEE Link. I've tried my best to organise and explain what each section does, but as I'm very new to GEE and .js, there's probably something evidently wrong here and there. I'm happy to provide more clarification on anything.

// ### IMPORTED GEOMETRY ###
var ShannonGeom = 
    /* color: #d6cfcf */
    /* shown: false */
    /* displayProperties: [
      {
        "type": "rectangle"
      }
    ] */
    ee.Geometry.Polygon(
        [[[-9.061457413652322, 52.9400316020557],
          [-9.061457413652322, 52.6131383784731],
          [-8.26623608491697, 52.6131383784731],
          [-8.26623608491697, 52.9400316020557]]], null, false);

// ### FILTER PARAMETERS ### 
//These apply the parameters to stop repetitive data entry and searching for the correct line to enter them.
// CLOUD % format '00' , numbers only without quotes
var MIN_CLOUD_PERCENT = 50;
// START DATE & END DATE format: 'YYYY-MM-DD', include the quotes
var START_DATE = '2022-01-01';
var END_DATE = '2022-04-01';
// BOUNDS can either be a drawn geometry shape within GEE, or an imported SHP file
// See: https://developers.google.com/earth-engine/guides/table_upload#upload-a-shapefile
var BOUNDS = ShannonGeom;
// ZOOM is based on GEE's 1-24 level system. Larger number = Larger Zoom
var ZOOM = 10;
// PARAMETERS END - You don't need to change anything below this line to make the script function

// ### MISC ###
//Prints out the Parameters set
print('Filtering available Sentinel 2 imagery between ' + START_DATE + ' & ' + END_DATE + ' with ' + MIN_CLOUD_PERCENT + '% cloud over the area of interest.');
// Centre based on the Geometry (Region of Interest, Zoom Level)
Map.centerObject(BOUNDS, ZOOM, print('Map centered on Region of Interest'));
// Sets the default Map to Terrain mode (Roadmap overlain with hillsahde) 
Map.setOptions("TERRAIN");

// ### EEPALETTES & LAYER VISUALISATION ###
/* Required for most layer visualisations
*  See https://github.com/gee-community/ee-palettes for more information
*
*  IF IT FAILS TO LOAD the PALETTES, LOAD THIS URL FIRST, THEN REFRESH THE PAGE: 
* (https://code.earthengine.google.com/?accept_repo=users/gena/packages)
*/
var palettes = require('users/gena/packages:palettes');
// NDWI palette 
var NDWIPalette = palettes.cmocean.Ice[7].reverse();
// NDVI palette
var NDVIPalette = palettes.colorbrewer.RdYlGn[10];

// Truecolour (R-G-B) Visualisation
var rgbVis = {
  min: 0,
  max: 0.35,
  bands: ['B4', 'B3', 'B2'],
};
// NDWI Visualisation
var NDWIVis = {
  min: -1,
  max: 1,
  bands: ['NDWI'],
  palette: NDWIPalette
};
// NDVI Visualisation
var NDVIVis = {
  min: -1,
  max: 1,
  bands: ['NDVI'],
  palette: NDVIPalette
};
// NDVI Mask Visualisation
var NDVIMaskVis = {
  min: 0,       // Land Areas
  max: 1,       // Other Areas
  bands: ['NDVI_Mask'],
  palette: ['cccccc','088300'],
  opacity: 0.65
};
// NDWI Mask Visualisation
var NDWIMaskVis = {
  min: 0,       // Land and Non-Water Areas
  max: 1,       // Water Areas
  bands: ['NDWI_Mask'],
  palette: ['cccccc','0000ff'],
  opacity: 0.65
};
// L1C Water/Veg Mask Visualisation
var L1CMaskVis = {
  min: 0,       // Land and Non-Water Areas
  max: 1,       // Water Areas
  bands: ['L1CMask'],
  palette: ['cccccc','f90000'],
  opacity: 0.65
};

// ### CLOUD MASKING ###
// Sentinel 2 Cloud Masking Function using the 60m Cloud Mask Band
/*  Function to mask clouds using the Sentinel-2 QA band
 *  @param {ee.Image} image Sentinel-2 image
 *  @return {ee.Image} cloud masked Sentinel-2 image
 */
function maskS2clouds(image) {
  var qa = image.select('QA60');
  // Bits 10 and 11 are clouds and cirrus, respectively.
  var cloudBitMask = 1 << 10;
  var cirrusBitMask = 1 << 11;
  // Both flags should be set to zero, indicating clear conditions.
  var mask = qa.bitwiseAnd(cloudBitMask).eq(0).and(qa.bitwiseAnd(cirrusBitMask).eq(0));
   return image.updateMask(mask).divide(10000);
}
print('Sentinel 2 Cloud Mask Function Complete');

// ### IMAGE COLLECTIONS ### 
//Load and Map L1C imagery with the filter parameters applied
                  /*
                  *  Load Sentinel-2 'Harmonized' Top Of Atomsphere (L1C) data
                  *  Dataset details: https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S2_HARMONIZED
                  *  HARMONIZED makes sure scenes after 25 January 2022 have the same DN ranges as older L1C scenes.
                  *  Harmonised L1C Data is available from Sentinel 2 Launch (2015-06-23) onwards.
                  */
var S2_L1C = ee.ImageCollection('COPERNICUS/S2_HARMONIZED')
                  // Filter by Date Period (YYYY-MM-DD)
                  .filterDate(START_DATE, END_DATE)
                  /* 
                  *  Pre-filter to get less cloudy granules
                  *  'Default' is aiming for 10% cloud
                  *  Dependent on availability of cloud-free imagery in the time period set
                  *  Longer periods will take longer to load
                  */
                  .filter(ee.Filter.lte('CLOUDY_PIXEL_PERCENTAGE', MIN_CLOUD_PERCENT))
                  // Select only image tiles that fall within the Geometry (Region of Interest) to reduce processing time
                  .filterBounds(BOUNDS)
                  // Applies the S2 Cloud Masking Function to each image in the IC
                  .map(maskS2clouds)
                  // Clips each image in the IC by the Bounds to reduce processing time further
                  .map(function(image) {
                    return image.clip(BOUNDS);
                  });

print('Time, Date, Bounding, and Cloud Tile Filtering parameters set for Imagery');

// Add the pre-clipped IC's to the map
Map.addLayer(S2_L1C,rgbVis,'L1C');

// ### FUNCTIONS ###
  // Add NDWI band to IC
  var addNDWI = function(image) {
    return image.addBands(image.normalizedDifference(['B3', 'B8']).rename('NDWI'));
  }; 
  // Add an NDVI band to IC
  var addNDVI = function(image) {
   return image.addBands(image.normalizedDifference(['B8', 'B4']).rename('NDVI'));
  };
  // ### L1C MASK FUNCTIONS ###
  // Function to mask out NDWI (L1C)
  var WaterMaskL1C = function(image) {
   var NDWI = image.select(['NDWI']);
    return image.addBands(NDWI.gte(0.1).rename('NDWI_Mask'));
  };
  // Function to mask out NDVI (L1C)
  var VegMaskL1C = function(image) {
   var NDVI = image.select(['NDVI']);
     return image.addBands(NDVI.lte(0).rename('NDVI_Mask'));
  };
  // Function to combine both L1C Masks
  var CombinedL1CMask = function(image) {
    var NDWI_Mask = image.select(['NDWI_Mask']);
    var NDVI_Mask = image.select(['NDVI_Mask']);
     // Adds the 2 masks together, and then clamps the output to be binary to keep it suitable for masking
     return image.addBands(NDWI_Mask.add(NDVI_Mask).clamp(0,1).rename('L1CMask')); 
  };
  // Function to update the clipping of imagery to just the water areas, as defined by the combined Mask
  //.rename(['B1_Msk','B2_Msk','B3_Msk','B4_Msk','B5_Msk','B6_Msk','B7_Msk','B8_Msk','B8A_Msk','B9_Msk','B10_Msk','B11_Msk','B12_Msk','QA10_Msk','QA20_Msk','QA60_Msk','NDWI_Msk','NDVI_Msk','NDVI_Mask2','NDWI_Mask2','L1CMask2'])
  var L1CMasking = function(image) {
    var Mask = image.select(['L1CMask']);
     return image.updateMask(Mask.eq(1));
  };

// Applies all the functions to respective image collections
var S2_L1C_Func = S2_L1C.map(addNDWI).map(addNDVI).map(VegMaskL1C).map(WaterMaskL1C).map(CombinedL1CMask);

// Add the individual new bands to the Map
// NDVI
Map.addLayer(S2_L1C_Func, NDVIVis, 'L1C NDVI');
// NDWI
Map.addLayer(S2_L1C_Func, NDWIVis, 'L1C NDWI');
// NDVI Mask
Map.addLayer(S2_L1C_Func, NDVIMaskVis, 'L1C NDVIMask');
// NDWI Mask
Map.addLayer(S2_L1C_Func, NDWIMaskVis, 'L1C NDWIMask');
// Combo Mask
Map.addLayer(S2_L1C_Func, L1CMaskVis, 'L1C Mask');

/* Creating a new var to run the mask apply function, because otherwise for some reason it retroactively 
*  breaks all the earlier separate ND VI/WI masks
*/
var S2_L1C_Masked = S2_L1C_Func.map(L1CMasking);
// Clipped hopefully?!
Map.addLayer(S2_L1C_Masked, rgbVis, 'L1C MaskApplied');

1 Answer 1

2

I can understand what confused you. In your analysis, you have an image collection as a result. When you add it to the map, you expect in both cases to see the image from the same date. However, there is no guarantee that this will be a case.

I've modified you script in two places:

  • I added a line to keep the date property of each image line 109 image.updateMask(mask).divide(10000).copyProperties(image, ["system:time_start"]);
  • When adding images to the map, I've filtered for the same date:
Map.addLayer(S2_L1C_Func.filterDate('2022-03-14', '2022-03-15'), L1CMaskVis, 'L1C Mask');
Map.addLayer(S2_L1C_Masked.filterDate('2022-03-14', '2022-03-15'), rgbVis, 'L1C MaskApplied');

The output images look the same in this case.

The link to the script: https://code.earthengine.google.com/7b127af4156094802d41a2b0bc815e68

2
  • Oh I see! Thank you so much, this explains what was going on. So the masks are working, but displaying a masked image for the wrong date! Out of interest, is there any particular reason why GEE does this? I can't understand why .map showed the same NDVI, NDWI, and individual Masks, but not the final combined Mask and the resulting masked image. Thank you again! Apr 28, 2022 at 12:24
  • 1
    Image collections and feature collections don't have a particular order of elements in it. So when you add to the map a collection with multiple images, what you actually see on the map is on of those images or some combination of it. By the way, to explore that, you can go to inspector tab on the right (next to 'Console' and 'Tasks') and click on the map to see all the values of the clicked pixel from different image dates (great tool for debugging). Happy to help! Apr 28, 2022 at 13:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.