Historical flood occurrence

Remote Sensing

GIS

Research

Extracting the historical average of flood occurrence by month from satellite imagery, to add probability information to WFP’s VAMPIRE flood alert system

Author

Benny Istanto

Published

April 9, 2020

Inspired from Dr. Ate Poortinga’s work in extracting water occurrence from satellite image, I tried to do similar analysis to get historical average of flood occurrence by month. The result from this analysis will be added as a probability information to enhance WFP’s VAMPIRE flood alert system (Extreme rainfall-triggering flood) that currently I am working on.

The steps below describe how to extract monthly water occurrence and seasonality water using Google Earth Engine code editor.

Step 1. Define geographic domain

var point = ee.Geometry.Point(111.9124, -6.9454);
var rectangle = ee.Geometry.Rectangle(57.0,-13.0,175,57.0);
var bbox = ee.Feature(rectangle).geometry();
Map.centerObject(point, 11);

Step 2. Import JRC Global Surface Water and extract permanent water from Seasonality layer

var jrcA = ee.Image('JRC/GSW1_1/GlobalSurfaceWater');
var seasonality = jrcA.select('seasonality').clip(bbox);
var jrcB = ee.ImageCollection("JRC/GSW1_1/MonthlyHistory");

Step 3. Define study period (example: extract January data from 1984 - 2018)

var startYear = 1984;
var endYear = 2018;
var startMonth = 1;
var endMonth = 12;

var startDate = ee.Date.fromYMD(startYear, startMonth,1); 
var endDate = ee.Date.fromYMD(endYear, endMonth,31); 

var months = endDate.difference(startDate, 'month').divide(12).toInt().add(1);
print('Number of month selected : ', months);

var myjrc = jrcB.filterBounds(bbox);
var myjrc = jrcB.filter(ee.Filter.calendarRange(startYear, endYear, 'year')).filter(ee.Filter.calendarRange(startMonth, endMonth, 'month'));

Step 4. Detect all observation data with water

// Detect observations
var myjrc = myjrc.map(function(img){
  // observation is img > 0
  var obs = img.gt(0);
  return img.addBands(obs.rename('obs').set('system:time_start', img.get('system:time_start')));
});

// Detect all observations with water
var myjrc = myjrc.map(function(img){
  // if water
  var water = img.select('water').eq(2);
  return img.addBands(water.rename('onlywater').set('system:time_start', img.get('system:time_start')));
});

Step 5. Calculate flood occurrence

// Calculate the total amount of observations
var totalObs = ee.Image(ee.ImageCollection(myjrc.select("obs")).sum().toFloat());
// Calculate all observations with water
var totalWater = ee.Image(ee.ImageCollection(myjrc.select("onlywater")).sum().toFloat());
// Calculate the percentage of observations with water
var floodfreq0 = totalWater.divide(totalObs).multiply(100);

// Mask areas that are not water (land)
var landMask = floodfreq0.eq(0).not();
var floodfreq1 = floodfreq0.updateMask(landMask); //Historical water occurrence (permanent water + flood)

// Extract permanent water, threshold >= 82% occurrence
// 82% came from https://www.frontiersin.org/articles/10.3389/fenvs.2019.00191/full
var gte82Mask = floodfreq1.gte(82);
var pw82 = floodfreq1.updateMask(gte82Mask);

// Subtract water (flood) occurrence with permanent water, threshold >= 82% occurrence
var gt82Mask = floodfreq1.gte(82).not();
var floodfreq2 = floodfreq1.updateMask(gt82Mask);

// Extract permanent water, threshold >= 10month/year occurrence
var gte10Mask0 = seasonality.gte(10);
var pw10 = seasonality.updateMask(gte10Mask0);

// Subtract water (flood) occurrence with permanent water, threshold >= 10month/year occurrence
var gte10Mask1 = seasonality.gte(10).mask(1);
gte10Mask1 = gte10Mask1.eq(1).not();
var floodfreq3 = floodfreq1.updateMask(gte10Mask1);

Step 6. Visualisation and add map to canvas

// Seasonal Water - Orange to Red is persistent water for at least 10 month/year
var seasonalityVis = {min: 0.0, max: 12.0, palette: ['f7fbff','deebf7','c6dbef','9ecae1','6baed6','4292c6','2171b5','08519c','08306b','fd8d3c','f03b20','bd0026']};

// Flood Occurrence
var floodViz = {min:0, max:100, palette:['c10000','d742f4','001556','b7d2f7']};

// Permanent Water >=82%
var pw82Viz = {palette:"ff9e00"};

// Permanent Water >=10month
var pw10Viz = {palette:"38f500"};

// Add map to canvas
Map.addLayer(floodfreq3.clip(bbox), floodViz,"FloodFreq without PW >=10month", true);
Map.addLayer(floodfreq2.clip(bbox), floodViz,"FloodFreq without PW >=82%", false);
Map.addLayer(floodfreq1.clip(bbox), floodViz,"FloodFreq", false);
Map.addLayer(pw10.clip(bbox), pw10Viz,"Permanent Water >=10 month/year", false);
Map.addLayer(pw82.clip(bbox), pw82Viz,"Permanent Water >=82% occurrence", false);
Map.addLayer(seasonality.clip(bbox), seasonalityVis, "Seasonal Water", false);

Step 7. Export the result to Google Drive

Export.image.toDrive({
  image:floodfreq3,
  description:'FF_January_1984_2018',
  folder:'GEE',
  scale:30,
  region: bbox_LKA,
  maxPixels:1e12
});

Step 8. Setting legend

// Set position of panel
var legend = ui.Panel({
  style: {
  position: 'bottom-right',
  padding: '8px 15px'
  }
});
 
// Create legend title
var legendTitle = ui.Label({
  value: 'Probability of Flood Ocurrence (%)',
  style: {
  fontWeight: 'bold',
  fontSize: '18px',
  margin: '0 0 4px 0',
  padding: '0'
  }
});
 
// Add the title to the panel
legend.add(legendTitle);
 
// Create the legend image
var lon = ee.Image.pixelLonLat().select('latitude');
var gradient = lon.multiply((floodViz.max-floodViz.min)/100.0).add(floodViz.min);
var legendImage = gradient.visualize(floodViz);
 
// Create text on top of legend
var panel = ui.Panel({
  widgets: [
    ui.Label(floodViz['max'])
    ],
  });
 
legend.add(panel);
 
// Create thumbnail from the image
var thumbnail = ui.Thumbnail({
  image: legendImage,
  params: {bbox:'0,0,10,100', dimensions:'10x200'},
  style: {padding: '1px', position: 'bottom-center'}
  });
 
// Add the thumbnail to the legend
legend.add(thumbnail);
 
// Area text on top of legend
var panel = ui.Panel({
  widgets: [
    ui.Label(floodViz['min'])
    ],
  });
 
legend.add(panel);
Map.add(legend);

Figure 1. Google Earth Engine user interface

Full script here