[GIS] Masking Water Bodies to build surface temperature raster slice of them in ENVI

envilandsat 8normalized-difference-water-index

I am looking at a region in the Alaskan tundra, there are plenty of small water bodies in my scene.

I need to mask these bodies, so that I can perform a surface temperature/emissivity normalization scene on only these water bodies. The ultimate product is to look at the original scene, and overlay the temperature masks unto the bodies of water.

How can I do this?

Edit: I am relatively new to ENVI, but I have tried doing a NDWI with band math.. It created a scene in which it identified possible water bodies in white and the rest in black. I could not figure out how to create a shapefile out of the white bodies it identified..

I have also considered classifying the water bodies but I am not sure how to go about this and ultimately create a mask.

I am working with Landsat 8 Data.

Best Answer

You are the right way. Follow this process

Calculate MNDWI with Band using (b1 - b2)/(b1 + b2) where b1 is green band and b2 is SWIR1. Read this paper for more information about MNDWI.
Also with Band math, create the water mask. Use (b1 le 0)*0 + (b1 gt 0)*1 using MNDWI result. Where le is lower/equal than b1 and gt greater than b1.
Go to Vector / Raster to vector and convert it. This will result in a .evf file, to save as shapefile use Vector / Classic EVF to Shapefile.

Related Solutions

[GIS] how i calculate land surface temperature using landsat 8 in Envi 5.1

Step 1: Convert from digital numbers (DN) to radiance
This is done by applying the multiplier and addition numbers as found in the metadata (.MTL) file. For the thermal bands (B10 and B11), the values are usually, but you should check the file:
Add: 0.1
Multiply by: 0.0003342 (3.3420E-04)

In ENVI you can apply this correction using 'band math':
float(b10)*0.0003342+0.1
This gives you the radiance value.

Step 2: Convert from radiance to kelvin
The formula needed here is
K2 / ln(K1/TOA_r + 1)
Again, the important values can be found in the metadata file. Usually, the K1 and K2 values are as follows:
K1_CONSTANT_BAND_10 = 774.89
K1_CONSTANT_BAND_11 = 480.89
K2_CONSTANT_BAND_10 = 1321.08
K2_CONSTANT_BAND_11 = 1201.14
In ENVI band math the formula becomes:
1321.08 / alog(774.89/B1+1)
Where alog is the ENVI band math version of the natural log.

This could be combined into one step - for example, band 11 becomes:
1201.14 / alog(480.89/(float(b11)*0.0003342+0.1)+1)

Counting Water Pixels in Landsat Collection using Google Earth Engine

It's going to be something like this:

var landsat8= ee.ImageCollection('LANDSAT/LC8_L1T_TOA').filterBounds(geometry)
var waterThreshold = 0; 

// water function:
var waterfunction = function(image){
  //add the NDWI band to the image
  var ndwi = image.normalizedDifference(['B3', 'B6']).rename('NDWI');
  //get pixels above the threshold
  var water01 = ndwi.gt(waterThreshold);
  //mask those pixels from the image
  image = image.updateMask(water01).addBands(ndwi);

  var area = ee.Image.pixelArea();
  var waterArea = water01.multiply(area).rename('waterArea');

  image = image.addBands(waterArea);

  var stats = waterArea.reduceRegion({
    reducer: ee.Reducer.sum(), 
    geometry: geometry, 
    scale: 30,
  });

  return image.set(stats);
};
var collection = landsat8.map(waterfunction);
print(collection);

var chart = ui.Chart.image.series({
  imageCollection: collection.select('waterArea'), 
  region: geometry, 
  reducer: ee.Reducer.sum(), 
  scale: 30,
});
print(chart);

I took the liberty of removing some redundant code from your function. Watch out for a reduction inside a map() for the reasons described here. Also, note that you may get no pixels in the result (waterArea=0) if the region overlaps the image footprint, but no valid pixels.