Claudia's Placencia mangrove

gee.js

/**
  MANGROVE CLASSIFICATION SCRIPT 
  For Placencia 2025 USING SENTINEL 2
  2025/06/09
  Claudia Baron-Aguilar
  
Previous steps.
  1. Selection of best image (Using Image selection code)
  2. selection of positive and negative goriud-truth points 
      (Creating List of coordinates code with anotehr code)
  3. Build ground-Truth matrix with positive and negative points 
      with data from step 2 (Presence-absence/ gt_data import above)

Steps.
  1. Load image
  2. Preprocess image
  3. Sample data
  4. Train model & Classify
  5. Accuracy assessment
*/

////////////////////////////////////////////////////////////////////////////
//                            1. LOAD IMAGE                               //
////////////////////////////////////////////////////////////////////////////
// Previously obtained with "seleccion_imagen" code

// Image ID
var id = '20250321T161011_20250321T161849_T16QCD';

//Load image
var image = ee.Image('COPERNICUS/S2_SR_HARMONIZED/' + id);

// see image
print (image);
      
// Apply scale factor to the Image      
var scaledImage = image.multiply(0.0001);

// Print Scaled Image to the console to inspect

print('scaledImage',scaledImage);

var image = scaledImage; //Image is now equal to scaled Image

//COLOR PALETTES (RGB)//
var visualization = {
  min: 0.0,
  max: 0.3,
  gamma:2,
  bands: ['B4', 'B3', 'B2'],
};

//Visualize Image
Map.addLayer(image, visualization,"image");

//Now we have an Image that is ready to be processed

//////////////////////////////////////////////////////////////////////
// //                          2. IMAGE PREPROCESSING               //
// ///////////////////////////////////////////////////////////////////


//Calculate the ndvi (Normalized Difference Vegetation Index)

var ndvi = image.normalizedDifference(['B8','B4']);
print ('ndvi', ndvi);
var ndviVis = {min: -1, max: 1, palette: ['blue', 'white', 'green']};
Map.addLayer(ndvi, ndviVis, 'ndvi');

//Create Mask where NDVI is greater than the threshold

var ndviMask = ndvi.gte(0.25);

//Apply the ndvi mask to the image

var maskedImage = image.updateMask(ndviMask);
print ('maskedImage', maskedImage);
Map.addLayer (maskedImage, visualization, 'maskedImage');


// // // ////////////////////////////////////////////////////////////////////////////
// // // //                            3. SAMPLE DATA                              //
// // // ////////////////////////////////////////////////////////////////////////////

// Ground truth data

var data = gt_data;

print('Class Codes', data.aggregate_array('class').distinct());
// class Codes are:
//   0: no mangrove
//   1: mangrove
  
print('Ground truth Points per class', data.aggregate_histogram('class'));
// Number of ground truth points per class
//  0:362
//  1: 40
  
//select bands to sample
var samplingBands = ['B2', 'B3', 'B4','B5','B6','B7','B8','B8A','B9','B11','B12'];

// Sample multi-espectral dara using all ground truth points
var sampleData = maskedImage.select(samplingBands).sampleRegions({
  collection:data,
  properties: ['class'],
  scale: 10});
  
// Generate random numbers from 0 to 1 for each sampled class.
var random = sampleData.randomColumn('random');
print ('random', random);

// Split ground truth data:70% for training and 30% for validation, 
// to use in the classification model training
var trainingPoints = random.filter(ee.Filter.lt('random', 0.7));
var validationPoints = random.filter(ee.Filter.gte('random', 0.3));
print('Training Samples (70%)', trainingPoints);


// // // ////////////////////////////////////////////////////////////////////////////
// // // //                      4. TRAIN MODEL & CLASSIFY                         //
// // // ////////////////////////////////////////////////////////////////////////////

//Define and train the classifier. In this case the SVM Classifier.

var SVM = ee.Classifier.libsvm ({
  kernelType:'RBF',
  // Low gamma measn less curvature. 
  // High gamma means more curvature. 
  // Range between 0.001 and 100
  gamma: 4, 
  // Low cost: smooth decision surface. 
  // Higher cost: you aim to classify more points correctly 
  //    (more missclassification possible). 
  // Range between 1 and 1000
  cost: 50 
  
});

var model = SVM.train({
  features:trainingPoints,
  classProperty: 'class',
  inputProperties: samplingBands
});

// Classify the image using the trained classifier
var classified = maskedImage.classify(model);

// Define a palette for the distint classes
var classPalette = {min:0, max:1, plaette:['#3090C7','#CD7F32']};

// Dis[play map with classification

Map.addLayer(classified, classPalette,"SVM classification",false);

// Display presence pixels ONLY (Useful to export layer of only mangrove to QGis)
var class1Mask = classified.eq(1);
var class1Image = classified.updateMask(class1Mask);

Map.addLayer(class1Image, {}, "class1");

//We crate a mask associated to a polygon to clean the image in the more inland areas where 
//we know there is not mangrove
var class1ImageMasked = class1Image.updateMask(
  ee.Image.constant(1).clip(inlandPolygon).mask().not()
  );
  
Map.addLayer(class1ImageMasked, {}, 'clean class1');  

// // / ////////////////////////////////////////////////////////////////////////////
// // // //                      5. ACCURACY ASSESSMENT                            //
// // // ////////////////////////////////////////////////////////////////////////////

// Calculate accuracy using validation data created in point 3
// Classify the image using the trained classifier
var validation = validationPoints.classify(model);

// Get a confusion matrix representing expected accuracy (Using validation points- 30%)
var errorMatrix = validation.errorMatrix('class', 'classification');

print('validation overall accuracy', errorMatrix.accuracy());

//Estimate user and producer accuracies
var producerAccuracy = errorMatrix.producersAccuracy();
var userAccuracy = errorMatrix.consumersAccuracy();

print('Producer Acuracy SVM', producerAccuracy);
print('User Accuracy SVM', userAccuracy);

// The Kappa Coefficient is generated from a statistical test to evaluate the accuracy 
// of a classification. Kappa evaluates how well the classification performed 
// as compared to just randomly assigning values, i.e. did the classification do better 
// than random. The Kappa Coefficient can range from -1 t0 1. A value of 0 indicated that 
// the classification is no better than a random classification. A negative number 
// indicates the classification is significantly worse than random. A value close to 1 
// indicates that the classification is significantly better than random.

var kappa = errorMatrix.kappa();

print ('Kappa statistic', kappa);

//////////////////////////////////////////////////////////////////////////////////////
// // //                           6. EXPORT SHAPE FILE TO DRIVE                 // //
//////////////////////////////////////////////////////////////////////////////////////


// Si le cambias el nombre a la variable "roi", tambien tienes que cambiarlo en la linea "var region = roi ;"       
// cambiar especificamente al area que quieres guardar

// The order of the coordinates goes as follows:
// [[[westLon, southLat],
// [eastLon, southLat],
// [eastLon, northLat],
// [westLon, northLat],
// [westLon, southLat]]]); // repetiddo como la primera
 
var roi = // south area -Belize-
ee.Geometry.Polygon(
  [[[ -88.13157854678775,16.46213607050848],
    [-88.13157854678775,16.84992519338283],
    [-88.46631111711997,16.84992519338283],
    [-88.46631111711997,16.46213607050848],
    [-88.13157854678775,16.46213607050848]]]);
    
    // Save the cleaned Sentinel Mosaic RGB at full resolution
  var description = 'export_Mnagrove_layer_geotiff_' ;  // Description for the export task
  var region = roi ;                                    // Region of interest described above
  var scale = 10 ;                                      // Spatial resolution in meters. Put here the resolution of you image/satellite.
  var crs = 'EPSG:4326' ;                               // Coordinate reference system (WGS 84)
  var maxPix = 1e13 ;                                   // Maximun number of pixels alloed for export
  var fileNamePre = 'Mangrove_Placencia_2025' ;         // Name the file to save
             

Export.image.toDrive({
  image: class1ImageMasked, // The image variable you want to export
  description: description,
  region: region,
  scale: scale,
  fileFormat: 'GeoTIFF',
  fileNamePrefix: fileNamePre,
  crs: crs,
  maxPixels: maxPix, 
  skipEmptyTiles: true
  }) ;