00058_modis_landcover_downscaling

/*
Tutorial Code by Amirhossein Ahrari
YouTube: https://www.youtube.com/@amirhosseinahrarigee

This code is part of a tutorial series on Earth Engine programming techniques
presented on the Amirhossein Ahrari YouTube channel. You are free to use and modify
this code for academic and non-academic purposes. Don't forget to subscribe to
the Amirhossein Ahrari channel and follow the videos to support the instructor!
*/ 
 

var cor = [45.68929902130395,37.11033465203267]

var geometry = ee.Geometry.Point(cor)
 
var table = ee.FeatureCollection("WWF/HydroSHEDS/v1/Basins/hybas_5")
 
var roi = table.filterBounds(geometry).map(function(feature){
  return feature.simplify(3000)
  })
 
Map.centerObject(roi)
Map.addLayer(roi)

var time_start = '2020', time_end = '2021'

var modis = ee.ImageCollection("MODIS/061/MCD12Q1")
.select('LC_Type1')
.filterDate(time_start, time_end).first();

var vis = {
  min: 1.0,
  max: 17.0,
  palette: [
    '05450a', '086a10', '54a708', '78d203', '009900', 'c6b044', 'dcd159',
    'dade48', 'fbff13', 'b6ff05', '27ff87', 'c24f44', 'a5a5a5', 'ff6d4c',
    '69fff8', 'f9ffa4', '1c0dff'
  ],
};

Map.addLayer(modis.clip(roi),vis, 'modis_lc_500m',false)


// predictor 

var landsat  = ee.ImageCollection("LANDSAT/LC08/C02/T1_L2")
.select('SR_B.*')
.filterDate(time_start, time_end)
.filterBounds(roi)
.filter(ee.Filter.lt('CLOUD_COVER', 10))
.map(function(img){
  var gain = ee.Number(img.get('REFLECTANCE_MULT_BAND_1'))
  var offset = ee.Number(img.get('REFLECTANCE_ADD_BAND_1'))
  var sr = img.multiply(gain).add(offset)
  var ndvi = sr.normalizedDifference(['SR_B5','SR_B4']).rename('ndvi')
  var ndwi = sr.normalizedDifference(['SR_B3','SR_B5']).rename('ndwi')
  var swir = sr.select('SR_B[6-7]')
  return ee.Image.cat([ndvi, ndwi, swir])
  });

var landsat_stat = landsat.reduce(ee.Reducer.percentile([25,50,75])) 

var image_2020 = ee.Image('JRC/GHSL/P2023A/GHS_BUILT_S/2020')
var built_2020 = image_2020.select('built_surface');
var visParams = {min: 0.0, max: 8000.0, palette: ['000000', 'FFFFFF']};

Map.addLayer(built_2020.clip(roi), visParams, 'urban',false);


var sar = ee.ImageCollection("COPERNICUS/S1_GRD")
.filterDate(time_start, time_end)
.filterBounds(roi)
.filter(ee.Filter.listContains('transmitterReceiverPolarisation','VV'))
.filter(ee.Filter.eq('instrumentMode','IW'))
.filter(ee.Filter.eq('orbitProperties_pass','DESCENDING')).select('VV')
.mean();

var dataset = landsat_stat.addBands(built_2020).addBands(modis).addBands(sar)

var training_data = dataset.stratifiedSample({
  numPoints: 100, classBand: 'LC_Type1', region: roi, scale: 100
  });
  
var model = ee.Classifier.libsvm().train({
  features: training_data, classProperty: 'LC_Type1', inputProperties: dataset.bandNames()
  });
  

var predictors = dataset.bandNames().remove('LC_Type1')

var modis30 = dataset.select(predictors).classify(model);


print(modis30)
  
Map.addLayer(modis30.clip(roi),vis, 'modis_lc_50m',false)