0

I wanted to calculate the NDVI time series of multiple points using Google Earth Engine. I'm using Collection 2 Level 2 Landsat-7 and Landsat-8 imagery to calculate the NDVI time series. However, when I try to harmonize the Landsat-7 ETM values to Landsat-8 OLI values using the tutorial, I get the same value of 0,74 for each NDVI value of the Landsat-7 images. I use .map to apply the harmınization function to all my images but can not solve the problem. How can I get the real/calculated NDVI values?

Here is my code:

 // Import the necessary libraries
var geometry = ee.FeatureCollection('users/maggie06/Kirsehir-5S');

//Image Collection Filter
var colFilter = ee.Filter.and(
    ee.Filter.bounds(geometry), ee.Filter.date('2014-09-22', '2017-07-02'),
    ee.Filter.lt('CLOUD_COVER', 50), ee.Filter.lt('GEOMETRIC_RMSE_MODEL', 10),
    ee.Filter.or(
        ee.Filter.eq('IMAGE_QUALITY', 9),
        ee.Filter.eq('IMAGE_QUALITY_OLI', 9)));

// Filter the images to only include images with less than 50% cloud coverage
var filtered_l7 = l7.filter(colFilter);
var filtered_l8 = l8.filter(colFilter);

// Applies scaling factors.
function applyScaleFactors(image) {
  var opticalBands = image.select('SR_B.').multiply(0.0000275).add(-0.2);
//  var thermalBand = image.select('ST_B6').multiply(0.00341802).add(149.0);
  return image.addBands(opticalBands, null, true);
//              .addBands(thermalBand, null, true);
}

var scaled_l7 = filtered_l7.map(applyScaleFactors);
var scaled_l8 = filtered_l8.map(applyScaleFactors);

// Rename the Landsat 7 bands in the collection
var renamed_l7 = scaled_l7.map(function(image) {
  return image.select(['SR_B1', 'SR_B2', 'SR_B3', 'SR_B4', 'SR_B5', 'SR_B7', 'QA_PIXEL'],
                       ['blue', 'green', 'red', 'NIR', 'swir1', 'swir2', 'pixel_qa']);
});

// Rename the Landsat 8 bands in the collection
var renamed_l8 = scaled_l8.map(function(image) {
  return image.select(['SR_B2', 'SR_B3', 'SR_B4', 'SR_B5', 'SR_B6', 'SR_B7', 'QA_PIXEL'],
                       ['blue', 'green', 'red', 'NIR', 'swir1', 'swir2', 'pixel_qa']);
});

// Mask clouds, cloud shadows, and snow
// Define the mask function using the pixel_qa band
var mask = function(img) {
  var qa = img.select(['pixel_qa']);
  // Clouds
  var cloud = qa.bitwiseAnd(1 << 3).neq(0);
  // Cloud shadows
  var shadow = qa.bitwiseAnd(1 << 4).neq(0);
  // Snow
  //var snow = qa.bitwiseAnd(1 << 5).neq(0);
  
  var mask = cloud.or(shadow);
  //.or(snow);
  return img.updateMask(mask.not());
};

var masked_l7 = renamed_l7.map(mask);
var masked_l8 = renamed_l8.map(mask);

//Harmonization Formula
var coefficients = {
  itcps: ee.Image.constant([0.0003, 0.0088, 0.0061, 0.0412, 0.0254, 0.0172]).multiply(10000),
  slopes: ee.Image.constant([0.8474, 0.8483, 0.9047, 0.8462, 0.8937, 0.9071])
};

// Define function to apply harmonization transformation.
var harmonised_l7 = masked_l7.map(function (image) {
  return image.select(['blue', 'green', 'red', 'NIR', 'swir1', 'swir2'])
    .multiply(coefficients.slopes)
    .add(coefficients.itcps)
    .round()
    .toShort()
    .addBands(image.select('pixel_qa')
  );
});

// Apply the Transform function to the L7 ETM collection
//var harmonised_l7 = masked_l7.map(etm2oli);

// Merge the two collections
var merged_collections = harmonised_l7.merge(masked_l8);

// Define the NDVI function
var calcNDVI = function(img) {
  var ndvi = img.normalizedDifference(['NIR','red']).rename('NDVI');
  return img.addBands(ndvi);
};

// Apply the NDVI function to the collection
var ndvi_collection = merged_collections.map(calcNDVI);

// Function to map over the FeatureCollection - Saving NDVI values to GoogleDrive:
var extractNDVI = function(feat) {
  // get feature geometry
  var geom = feat.geometry();
  // function to iterate over the yearly ImageCollection
  // the initial object for the iteration is the feature
  var addProp = function(img, f) {
    // cast Feature
    var newf = ee.Feature(f);
    // get date as string
    //var date = img.date().format();
    var date = ee.String(img.get('system:index'));
    // extract the value (first) of 'waterClass' in the feature
    var value = img.reduceRegion(ee.Reducer.first(), geom, 30).get('NDVI');
    // if the value is not null, set the values as a property of the feature. The name of the property will be the date
    return ee.Feature(ee.Algorithms.If(value,
                                       newf.set(date, ee.String(value)),
                                       newf.set(date, ee.String('No data'))));
  };
  var ndvi_values = ee.Feature(ndvi_collection.iterate(addProp, feat));
  return ndvi_values;
};

var ndvi_values = geometry.map(extractNDVI);

// Export the NDVI values and dates to an Excel file
Export.table.toDrive({
  collection: ndvi_values,
  description: 'Kirsehir_5s_NDVI_harmonized',
  fileFormat: 'CSV'
});

1 Answer 1

0

I don't think that is the intended use of ee.ImageCollection.iterate, but there is a simpler way of retrieving time series over multiple points for an image collection: using ee.Image.reduceRegions.

Here's a full minimal example consisting of: (1) defining a small feature collection (2 points), (2) retrieving a simple NDVI image collection (from Landsat 9), (3) retrieving the time series as a FeatureCollection, (4) making a chart showing the time series over the two points. You can easily integrate step 3 with your feature colleciton and your NDVI image collection.

var fc = ee.FeatureCollection([
  ee.Feature(ee.Geometry.Point(-123.75, 41.26), {name:"point1"}),
  ee.Feature(ee.Geometry.Point(-110.28, 38.44), {name:"point2"})]
)
Map.addLayer(fc)

var ndvi = ee.ImageCollection("LANDSAT/LC09/C02/T1_L2")
.filterDate("2023-01-01", "2023-05-30")
.filterBounds(fc)
.map(function(img){
  return img.addBands(
    img.normalizedDifference(["SR_B5", "SR_B4"]).rename("NDVI")
    )
}).select("NDVI")

var timeSeries = ndvi.map(function(img){
return img.select("NDVI")
  .reduceRegions({
    collection: fc,
    reducer:ee.Reducer.first(),
    scale: 100  // see https://developers.google.com/earth-engine/guides/scale#scale-of-analysis
    })   // Use reduceRegions in img, to get the first() value for each feature (point) in the collection.
  .filter(ee.Filter.notNull(["first"])) // Don't keep the features in which the value is null 
  // (because each image doesn't cover all points in the feature collection)
  .select(["first", "name"],["NDVI", "name"]) // rename the property "first" (from the ee.Reducer.first() to NDVI)
  .map(function(f){
    return f.set({date:img.date(), 
    landsatIndex:img.get("system:index")})
  }) // Assign the "date" property to each feature.. and optionally other metadata from each Landsat image. 
}).flatten() // Map the above function over all the images, and flatten the resulting FeatureCollection
.sort("geometry")// Sort the features by date

print(timeSeries)

var chart = ui.Chart.feature.groups(timeSeries, "date", "NDVI","name")

print(
  chart.setOptions({"title": "NDVI time series over two points"})
  )

https://code.earthengine.google.com/c8f72ee152c9d6f3657a7da8e762d3a5

3
  • I implemented your suggestion in my code. I can extract the time series values without the Landsat-7 harmonization part included. When I add the harmonization function, I got the 'Error in map(ID=1_LE07_176033_20140830): Image.date: Image '1_LE07_176033_20140830' does not have a 'system:time_start' property.' error. Here is my revised code: code.earthengine.google.com/b8470d07eb33014f778308c62b004a9d
    – Ayda Aktas
    May 31, 2023 at 11:52
  • That's a separate issue and one answer is here: gis.stackexchange.com/questions/422664/… -- you can use copyPoperties or other methods to ensure that the system:time_start is propagated through your operations. The way I used .addBands in the part where I defined ndvi in the short example above ensured that the system:time_start property wasn't lost, so .copyProperties wasn't needed. May 31, 2023 at 12:30
  • When I implemented your code, I also did the same while defining ndvi. Later, I also added .copyProperties to the harmonization function. Then I turn to the beginning, where I get the same value (0,74) for each point of Landsat-7 NDVI value for the whole time series.
    – Ayda Aktas
    May 31, 2023 at 13:41

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.