GEE-Land-Cover-Classification

//Using GEE to estimate agricultural land in Brazil //centre the map view over ROI Map.centerObject(ROI, 7); //DEM scene var demelevation = SRTMDEM.select('elevation');//elevation data from SRTM Digital elevation model (DEM) //print(demelevation) var slope = ee.Terrain.slope(demelevation);//slope of the landscape from SRTM Digital elevation model (DEM) //print(slope)

//Sentinel 2 Data // Now select your image type! var collection = ee.ImageCollection('COPERNICUS/S2_SR') // searches all sentinel 2 imagery pixels... .filter(ee.Filter.lt("CLOUDY_PIXEL_PERCENTAGE", 5)) // ...filters on the metadata for pixels less than 5% cloud .filterDate('2018-09-01' ,'2019-06-30') //... chooses only pixels between the dates you define here .filterBounds(ROI); // ... that are within your aoi print(collection); // this generates a JSON list of the images (and their metadata) which the filters found in the right-hand window. var medianpixels = collection.median() // This finds the median value of all the pixels which meet the criteria. var medianpixelsclipped = medianpixels.clip(ROI) // this cuts up the result so that it fits neatly into your aoi
var medianpixelsclippednew = medianpixels.clip(ROI).divide(10000)// and divides so that values between 0 and 1

// visualise the mosaic using various band combinations. //Map.addLayer(medianpixelsclippednew, {bands: ['B12', 'B11', 'B4'], min: 0, max: 1, gamma: 1.5}, 'Sentinel_2 mosaic',false) //Map.addLayer(medianpixelsclippednew, {bands: ['B4', 'B3', 'B2'], min: 0, max: 1, gamma: 1.5}, 'Sentinel_2 truecolour',false) //Map.addLayer(medianpixelsclippednew, {bands: ['B8', 'B4', 'B3'], min: 0, max: 1, gamma: 1.5}, 'Sentinel_2 falsecolour',false)

// Calculate NDVI var image_ndvi = medianpixelsclipped.normalizedDifference(['B8','B4']); print(image_ndvi); // this generates a JSON with metadata on the NDVI image

// Compute Vegetation Indices; first extract the various bands and name them var aerosol = medianpixelsclipped.select('B1'); var green = medianpixelsclipped.select('B3'); var red = medianpixelsclipped.select('B4'); var vre1 = medianpixelsclipped.select('B5'); var vre2 = medianpixelsclipped.select('B6'); var vre3 = medianpixelsclipped.select('B7'); var nir = medianpixelsclipped.select('B8'); var vre4 = medianpixelsclipped.select('B8A'); var swir1 = medianpixelsclipped.select('B11');

// Compute indices based on their formulae var brazilndvi = nir.subtract(red).divide(nir.add(red)).rename('NDVI'); var brazilndii = nir.subtract(swir1).divide(nir.add(swir1)).rename('NDII'); var brazilndwi = green.subtract(nir).divide(green.add(nir)).rename('NDWI'); var brazilsipi = nir.subtract(aerosol).divide(nir.subtract(red)).rename('SIPI'); //Calculate SIPI1 var brazilpssr = nir.divide(red).rename('PSSR'); //Calculate PSSRb1 var brazilsavi = (nir.subtract(red).divide(nir.add(red).add(0.428))).multiply(1+0.428).rename('SAVI');//Calculate SAVI

// Filter the collection for the VV product from the descending track var collectionVV = ee.ImageCollection('COPERNICUS/S1_GRD') .filter(ee.Filter.eq('instrumentMode', 'IW')) .filter(ee.Filter.listContains('transmitterReceiverPolarisation', 'VV')) .filter(ee.Filter.eq('orbitProperties_pass', 'DESCENDING')) .filterDate('2018-09-01' ,'2019-06-30') .filterBounds(ROI) .select(['VV']); //print(collectionVV); // Filter the collection for the VH product from the descending track var collectionVH = ee.ImageCollection('COPERNICUS/S1_GRD') .filter(ee.Filter.eq('instrumentMode', 'IW')) .filter(ee.Filter.listContains('transmitterReceiverPolarisation', 'VH')) .filter(ee.Filter.eq('orbitProperties_pass', 'DESCENDING')) .filterDate('2018-09-01','2019-06-30') .filterBounds(ROI) .select(['VH']); //print(collectionVH);

var VV = collectionVV.median().clip(initial); //median of selected band var VH = collectionVH.median().clip(initial);//median of selected band var ratio = VV.divide(VH).rename('ratio'); //band ratio

//print(VV); //print(VH); //print(ratio);

// Adding the VV layer to the map Map.addLayer(VV, {min: -14, max: -7}, 'VV',false); Map.addLayer(VH, {min: -20, max: -7}, 'VH',false);

// Compute texture measures of individual bands

var brazilb5sd = vre1.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilb6sd = vre2.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilb7sd = vre3.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilb8asd = vre4.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilndvitexture = brazilndvi.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilndvimean = brazilndvi.reduceNeighborhood({ reducer: ee.Reducer.mean(), kernel: ee.Kernel.circle(5), });

var brazilndvivariance = brazilndvi.reduceNeighborhood({ reducer: ee.Reducer.variance(), kernel: ee.Kernel.circle(5), });

var brazilndiimean = brazilndii.reduceNeighborhood({ reducer: ee.Reducer.mean(), kernel: ee.Kernel.circle(5), });

var brazilndwimean = brazilndwi.reduceNeighborhood({ reducer: ee.Reducer.mean(), kernel: ee.Kernel.circle(5), });

var brazilsipimean = brazilsipi.reduceNeighborhood({ reducer: ee.Reducer.mean(), kernel: ee.Kernel.circle(5), });

var brazilsavimean = brazilsavi.reduceNeighborhood({ reducer: ee.Reducer.mean(), kernel: ee.Kernel.circle(5), });

var brazilndiisd = brazilndii.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilndwisd = brazilndwi.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilsipisd = brazilsipi.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilsavisd = brazilsavi.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilndiivar = brazilndii.reduceNeighborhood({ reducer: ee.Reducer.variance(), kernel: ee.Kernel.circle(5), });

var brazilndwivar = brazilndwi.reduceNeighborhood({ reducer: ee.Reducer.variance(), kernel: ee.Kernel.circle(5), });

var brazilsipivar = brazilsipi.reduceNeighborhood({ reducer: ee.Reducer.variance(), kernel: ee.Kernel.circle(5), });

var brazilsavivar = brazilsavi.reduceNeighborhood({ reducer: ee.Reducer.variance(), kernel: ee.Kernel.circle(5), });

var brazilVVtexture = VV.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilVVmean = VV.reduceNeighborhood({ reducer: ee.Reducer.mean(), kernel: ee.Kernel.circle(5), });

var brazilVVvariance = VV.reduceNeighborhood({ reducer: ee.Reducer.variance(), kernel: ee.Kernel.circle(5), });

var brazilVHtexture = VH.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilVHmean = VH.reduceNeighborhood({ reducer: ee.Reducer.mean(), kernel: ee.Kernel.circle(5), });

var brazilVHvariance = VH.reduceNeighborhood({ reducer: ee.Reducer.variance(), kernel: ee.Kernel.circle(5), });

var brazilratiomean = ratio.reduceNeighborhood({ reducer: ee.Reducer.mean(), kernel: ee.Kernel.circle(5), });

var brazilratiosd = ratio.reduceNeighborhood({ reducer: ee.Reducer.stdDev(), kernel: ee.Kernel.circle(5), });

var brazilratiovariance = ratio.reduceNeighborhood({ reducer: ee.Reducer.variance(), kernel: ee.Kernel.circle(5), });

//Layer Stack, add all bands and texture measures together var mergedCollection = medianpixelsclipped.addBands(VV).addBands(VH).addBands(ratio).addBands(brazilndvi).addBands(brazilndii).addBands(brazilndwi).addBands(brazilsipi).addBands(brazilpssr).addBands(brazilsavi).addBands(demelevation).addBands(slope).addBands(brazilndvimean).addBands(brazilndwimean).addBands(brazilndiimean).addBands(brazilndvivariance).addBands(brazilndvitexture).addBands(brazilndiimean).addBands(brazilndwimean).addBands(brazilsavimean).addBands(brazilsipimean).addBands(brazilndiisd).addBands(brazilndwisd).addBands(brazilsavisd).addBands(brazilsipisd).addBands(brazilndiivar).addBands(brazilndwivar).addBands(brazilsavivar).addBands(brazilsipivar).addBands(brazilVVtexture).addBands(brazilVVmean).addBands(brazilVVvariance).addBands(brazilVHtexture).addBands(brazilVHmean).addBands(brazilVHvariance).addBands(brazilratiomean).addBands(brazilratiosd).addBands(brazilratiovariance).addBands(brazilb5sd).addBands(brazilb6sd).addBands(brazilb7sd).addBands(brazilb8asd); print('mergedCollection: ', mergedCollection); var band= ['slope','B2','B3','B4','B5','B6','B7','B8','B8A','VV','VH','ratio','NDVI','SIPI','NDII','NDWI','SAVI','NDVI_stdDev','NDII_stdDev','NDWI_stdDev','SAVI_stdDev','SIPI_stdDev','VV_stdDev','VH_stdDev','ratio_stdDev','B5_stdDev','B6_stdDev','B7_stdDev','B8A_stdDev']; //select bands to use for training data

//training data //create training GCPs var classNames=trnewpolycoffee.merge(newcoffeepolytr).merge(coffeepolystr).merge(coffeenewtrain).merge(newcoffeetrnew).merge(trnewpolyforest).merge(newforesttrnew).merge(newbeloforest).merge(geeforest).merge(trpolyurban).merge(newbelourban).merge(watertrain).merge(trnewpolywateragain).merge(newbelowater).merge(trnewpolynoncoffee).merge(newnoncoffeetrneww).merge(newbelononcoffee).merge(trpolyopenground).merge(foresttrainagain).merge(coffeetrainagain).merge(opengroundtrained).merge(morenoncoffeetrain).merge(shrubtr).merge(newopengroundtr).merge(newnoncoffeetr).merge(roadtr).merge(coffeepttr).merge(forestpttr).merge(geenoncoffee).merge(noncoffeepttr).merge(polyroadtr).merge(shrubpolytr).merge(waterpttr).merge(newpolyroadtr).merge(newbeloroad).merge(newbeloopenground); //create testing GCPs var testnames = urbantested.merge(newwatertest).merge(foresttested).merge(coffeetested).merge(opengroundtested).merge(noncoffeetested);

//create variable to read the different land cover classes (0-5: 0-coffee, 1-forest, 2-urban, 3-water, 4-noncoffee, 5-openground) var label = 'descriptio'

//Create training data var training = mergedCollection.select(band).sampleRegions({ collection: classNames, properties: [label], scale: 15 });

//train classifier var classifier = ee.Classifier.randomForest({ numberOfTrees: 30, variablesPerSplit: 3 }).train(training,label,band);

//Run the classification var classified = mergedCollection.select(band).classify(classifier);

//Display classification result Map.addLayer(classified.clip(ROI),{min: 0, max: 5, palette: ['purple', 'green', 'red','blue','yellow','black']},'classification');

//computes the accuracy of the classified image from the training dataset var confMatrix = classifier.confusionMatrix() print('Resubstitution error matrix: ', confMatrix); print('Training overall accuracy: ', confMatrix.accuracy());

//Create testing data var validation = classified.sampleRegions({ collection: testnames, properties: ['descriptio'], scale: 15 });

//Compare the landcover of the validation data against the classification result var testAccuracy = validation.errorMatrix('descriptio', 'classification');

//Print the error matrix to the console print('Validation error matrix: ', testAccuracy);

//Print the overall accuracy to the console print('Validation overall accuracy: ', testAccuracy.accuracy());

// Export the image, specifying scale and region. Export.image.toDrive({ image: classified, description: 'RFLandCoverClassificationBrazilNW', scale: 20, maxPixels:1e13, region: northwest, });

// Export the image, specifying scale and region. Export.image.toDrive({ image: classified, description: 'RFLandCoverClassificationBrazilNE', scale: 20, maxPixels:1e13, region: northeast, });

// Export the image, specifying scale and region. Export.image.toDrive({ image: classified, description: 'RFLandCoverClassificationBrazilSW', scale: 20, maxPixels:1e13, region: southwest, });

// Export the image, specifying scale and region. Export.image.toDrive({ image: classified, description: 'RFLandCoverClassificationBrazilSE', scale: 20, maxPixels:1e13, region: southeast, });

// Export the image, specifying scale and region.

Export.image.toDrive({ image: classified, description: 'RFLandCoverClassificationBrazil%accuracy', scale: 20, maxPixels:1e13, region: roibrazil, });