Dengue Risk Mapping with Climate Data and Remote Sensing

// Define study area: Lower Mekong Region (Vietnam / Cambodia / Thailand)
var roi = ee.Geometry.Polygon([
  [[102.0, 8.5], [109.5, 8.5], [109.5, 23.0], [102.0, 23.0]]
]);

// Temporal range: 2023 Monsoon Season (May - October)
var startDate = '2023-05-01';
var endDate = '2023-10-31';

// 1. ERA5-Land Temperature (Convert Kelvin to Celsius)
var era5 = ee.ImageCollection('ECMWF/ERA5_LAND/MONTHLY_AGGR')
  .filterBounds(roi)
  .filterDate(startDate, endDate);
var meanTempC = era5.select('temperature_2m')
  .mean()
  .subtract(273.15)
  .clip(roi)
  .rename('mean_temp_c');

// 2. CHIRPS Monthly Total Rainfall (mm)
var chirps = ee.ImageCollection('UCSB-CHG/CHIRPS/DAILY')
  .filterBounds(roi)
  .filterDate(startDate, endDate);
var totalPrecip = chirps.select('precipitation')
  .sum()
  .clip(roi)
  .rename('total_precip_mm');

// 3. MODIS NDVI (Scale factor 0.0001)
var modis = ee.ImageCollection('MODIS/061/MOD13A1')
  .filterBounds(roi)
  .filterDate(startDate, endDate);
var meanNdvi = modis.select('NDVI')
  .mean()
  .multiply(0.0001)
  .clip(roi)
  .rename('mean_ndvi');

// 4. WorldPop Population Density
var pop = ee.ImageCollection('WorldPop/GP/100m/pop')
  .filterBounds(roi)
  .mean()
  .clip(roi)
  .rename('population');

// Combine Predictors into a single multi-band Image
var predictorStack = meanTempC
  .addBands(totalPrecip)
  .addBands(meanNdvi)
  .addBands(pop);

// Map Center & Visualization
Map.centerObject(roi, 6);
Map.addLayer(meanTempC, {min: 20, max: 35, palette: ['blue', 'yellow', 'red']}, 'Mean Temp (C)');
Map.addLayer(totalPrecip, {min: 500, max: 2500, palette: ['white', 'cyan', 'navy']}, 'Total Precip (mm)');
Map.addLayer(meanNdvi, {min: 0.1, max: 0.8, palette: ['brown', 'yellow', 'green']}, 'NDVI');
Map.addLayer(pop, {min: 0, max: 500, palette: ['black', 'orange', 'white']}, 'Population', false);

// Define biological suitability thresholds for Aedes mosquito vector
// Optimal temperature: 24C to 32C; Rainfall: > 800mm cumulative; NDVI: > 0.45
var tempSuit = meanTempC.gte(24).and(meanTempC.lte(32));
var precipSuit = totalPrecip.gte(800);
var ndviSuit = meanNdvi.gte(0.45);
var popSuit = pop.gte(50);

// Rule-based heuristic risk score (0 to 4 scale)
var riskScore = tempSuit.add(precipSuit).add(ndviSuit).add(popSuit).rename('heuristic_risk');

// Sample spatial points for Machine Learning classification
// Risk score >= 3 -> High Risk Class (1), else Low Risk Class (0)
var binaryClass = riskScore.gte(3).rename('class');
var trainingData = predictorStack.addBands(binaryClass);

var samples = trainingData.stratifiedSample({
  numPoints: 300,
  classBand: 'class',
  region: roi,
  scale: 1000,
  geometries: true
});

// Train Random Forest Classifier (20 trees)
var rfClassifier = ee.Classifier.smileRandomForest(20)
  .train({
    features: samples,
    classProperty: 'class',
    inputProperties: ['mean_temp_c', 'total_precip_mm', 'mean_ndvi', 'population']
  });

// Classify environmental predictor stack
var rfRiskMap = predictorStack.classify(rfClassifier);

// Visualization
var riskVis = {min: 0, max: 1, palette: ['#1A9850', '#D73027']};
Map.addLayer(riskScore, {min: 0, max: 4, palette: ['white', 'yellow', 'orange', 'red', 'purple']}, 'Rule-Based Risk Score');
Map.addLayer(rfRiskMap, riskVis, 'Random Forest Dengue Risk Map');
Map.addLayer(samples, {color: 'black'}, 'Training Points', false);

// Print Feature Importance
var explain = rfClassifier.explain();
print('Random Forest Feature Importance:', explain.get('importance'));

// Export the classified Dengue Risk Map to Google Drive
Export.image.toDrive({
  image: rfRiskMap.toByte(),
  description: 'Southeast_Asia_Dengue_Risk_Map_2023',
  folder: 'GEE_Public_Health',
  fileNamePrefix: 'dengue_risk_rf_2023',
  region: roi,
  scale: 1000,
  maxPixels: 1e9,
  crs: 'EPSG:4326'
});