[GIS] Implementing a classification on Google Earth Engine

classificationgoogle-earth-enginelandsat 8

I am attempting to implement a classification based from a decision tree (attached as image below) where there are 5 classes. These classes are based off the different water reflectance values in Landsat 8 bands (namely NIR, Red and Green). I have no idea how I should go about doing this. Someone suggested using image.expression however it has not worked for me. This is the code I've used thus far:
(landsat is landsat 8 surface reflection)

var image = ee.Image(landsat.filterDate ('2018-01-01', '2018-03-01').filterBounds(point).sort('CLOUD_COVER').first());

//print image 
print('a landsat scene', image);

//implement MNDWI

var green = image.select('B3');
var swi = image.select('B6');
var mndvi = green.subtract(swi).divide(green.add(swi)).rename('MNDWI');

var mndvipara = {min: 0, max: 0.7, palette: ['white', 'blue']};


Map.addLayer(mndvi, mndvipara, 'MNDVI image');

//classification attempt 
var lpb =image.expression (
'nir > 0.018', {
'nir': image.select ('B5')});
Map.addLayer (lpb, {min: -1, max: 1, palette: ['FF0000', '00FF00']});

[1]: https://i.stack.imgur.com/TKSbZ.png (if image doesn't work, classification can be found at https://www.sciencedirect.com/science/article/pii/S0303243415000641 – Figure A.1

Best Answer

Yes, you can use ee.Image.expression() to implement a classification decision tree. Although it is unclear which Landsat 8 image collection you are using, if it is the Landsat 8 Surface Reflectance collection you will need to include the scaling factor in the calculation: 

var lpb = image.expression(
  'nir > 0.018', 
  {
    'nir': image.select('B5').multiply(0.0001)
  }
);

While using ee.Image.expression() will work, it may be cleaner to implement this decision tree using a series of conditional or boolean operators.

Related Question