I would like to classify glacial area by categories of snow (if there is) and ice, but what's most important: between old ice and fresh ice. They have different properties that are possible to recognize in field, but can you do this with satellite data? (preferably Landsat becasue of 30/15m spatial resolution)
-
1What characteristics do old and new ice have in the field? – Aaron♦ Mar 15 '16 at 22:43
-
11) Fresh snow is much much less compacted than old glacier ice (it becomes ice by compaction). Therefore this can be somehow connected with IR reflectance, which is absorbed by water. 2) Also fresh snow has albedo even up to almost 100%, but old snow can get as low as ~40% (no strict classification of course). I would like to make use of IR, since True Color composition isn't as useful as I'd like it to be. – adamczi Mar 17 '16 at 17:38
-
1This sounds like a straight forward image classification problem. You need to start with training data, which can be gathered in the field or by expertly selecting pixels from imagery. – Aaron♦ Mar 17 '16 at 21:59
-
3I think the way to go here is a supervised classification algorithm such as Maximum Likelihood, Random Forests, etc. that uses all of the available spectral bands. Are you familiar with these methods? I'm not sure what you mean by "IR composition". Are you referring to creating composite imagery such as false color composite (i.e. NIR, R, G)? If so, you are very limited in the applications of such products. – Aaron♦ Mar 19 '16 at 1:23
-
1@adamczi try working on google-earth-engine. supervised classification algorithms will be available as well as SAR data (either your upload or google's cloud). – csheth Jun 6 '17 at 14:11
You will have to use microwave data for this. Optical data just won't cut it. If you want to still go through with optical, do tell me what methodology you followed. Also a lot depends on the topography, LULC of your area. Microwave data classification is itself not straightforward, you'll have to consult a lot of literature and choose a methodology which suits you best. Please see the methodology I followed in my M.Tech Thesis: http://www.iirs.gov.in/iirs/sites/default/files/StudentThesis/Sanjay_MTech_2013-15.pdf
Please do ask if you have any question after going through the literature.
Here's an example that can help you get started on Google-Earth-Engine using Sentinel-1's C-Band:
var pt = ee.Geometry.Point(96.7868, 29.31409);
// Filter collection around point. Also read up on Sentinel-1's
// polarization
var collection = ee.ImageCollection('COPERNICUS/S1_GRD').filterBounds(pt)
.filter(ee.Filter.listContains('transmitterReceiverPolarisation', 'VV'))
.select('VV');
// select an appropriate date
var beforesnow = collection.filterDate('2016-11-01', '2016-12-01').mosaic();
var aftersnow = collection.filterDate('2017-02-01', '2017-03-01').mosaic();
// bands for Sentinel-2
var bands = ['B2', 'B3', 'B4'];
// Some Sentinel-2 images for reference
var S2 = ee.ImageCollection('COPERNICUS/S2').filterBounds(pt)
.select(bands);
var S2before = S2.filterDate('2016-10-01', '2016-11-30').mosaic();
var S2after = S2.filterDate('2017-01-01', '2017-02-01').mosaic();
Map.addLayer(S2before, {bands: ['B4', 'B3', 'B2'], min: 300,max: 5000}, 'S2 Before');
Map.addLayer(S2after, {bands: ['B4', 'B3', 'B2'], min:873,max: 12522}, 'S2 After');
Map.centerObject(pt, 13);
// you may change the min, max later when tinkering with the layers tab in // the map
Map.addLayer(beforesnow, {min:-30,max:0}, 'Before snow');
Map.addLayer(aftersnow, {min:-30,max:0}, 'After snow');
//Some information on the Sentinel-1 collection
print('Collection: ', collection);
You will have to classify the Image using supervised classification algorithms mentioned here: https://developers.google.com/earth-engine/classification
More on using Sentinel-1 https://developers.google.com/earth-engine/sentinel1
On Google Earth Engine and Glaciers: http://www.geo.uzh.ch/~mzemp/share/scratch/msc/MSc.Thesis_NoahZeltner_UsingGoogleEarthEngineForGlobalGlacierChangeAssessment.pdf
On SAR and glacier zones: http://www.sciencedirect.com/science/article/pii/S0034425713001703
