I am working on LandSat 8 data to build a classifier for built up and non built up areas in a particular district. I was using the following code for classification and then to export my classified image into my drive:
var cloud_masks = require('users/fitoprincipe/geetools:cloud_masks');
var maskClouds = cloud_masks.landsatTOA();
//Loading India image, the extracting data for Haryana (a state in India) and then subsequently Ambala (a district in Haryana)
var bands = ['B1','B2', 'B3', 'B4', 'B5', 'B6', 'B7', 'B8', 'B10', 'B11'];
var india = ee.FeatureCollection('ft:1UDdgOCf8DoRJ9bVm-UVbR6CqxtkJToLQjTFd0r0Z','geometry')
.filter(ee.Filter.eq('Name','India'))
.geometry();
var india_image = ee.ImageCollection('LANDSAT/LC08/C01/T1_TOA')
.filterBounds(india)
.filterDate('2014-02-01','2014-09-01')
.sort('CLOUD_COVER')
.map(maskClouds)
.median();
var district = ee.FeatureCollection('ft:1PA2zwArj8EsplrX9eMxJ2H_TICyyx855KPnbJhC1','geometry')
.filter(ee.Filter.eq('name','Gurgaon'));
var haryana_image = ee.ImageCollection('LANDSAT/LC08/C01/T1_TOA')
.filterBounds(district)
.filterDate('2014-02-01','2014-04-01')
.sort('CLOUD_COVER')
.map(maskClouds)
.median();
var input = haryana_image;
input = addBands(input.select(bands));
india_image = addBands(india_image);
//Loading the points from the fusion table and training the classifier
var ft = ee.FeatureCollection('ft:1fWY4IyYiV-BA5HsAKi2V9LdoQgsbFtKK2BoQiHb0');
var ft_builtup = ft.filter(ee.Filter.eq('class',1)).limit(1200);
var ft_nonbuiltup = ft.filter(ee.Filter.eq('class',2)).limit(1800);
ft = ft_builtup.merge(ft_nonbuiltup);
var new_bands = ['B1','B2', 'B3', 'B4', 'B5', 'B6', 'B7', 'B8', 'B10', 'B11','NDBI','NDVI'];
function addBands(image){
var ndvi = image.normalizedDifference(['B4', 'B3']).rename('NDVI');
var ndbi = image.normalizedDifference(['B5', 'B4']).rename('NDBI');
var ndwi = image.normalizedDifference(['B6', 'B6']).rename('NDWI');
return image.addBands(ndvi).addBands(ndbi).addBands(ndwi);
}
// Load a Landsat 8 image to be used for prediction.
var training = india_image.sampleRegions(ft,['class'],30);
var trained = ee.Classifier.cart().train(training, 'class', new_bands);
input = input.clip(district);
input = input.classify(trained);
input = input.expression('LC==1?1:0',{'LC':input.select('classification')});
Export.image.toDrive({
image: input.clip(district),
description: 'Test',
scale: 30
});
I used this code for download in September and it gave me a .tif image of size 175.5 KB. Now I am running the same script again now in November but it gives me an error saying that the number of pixels limit is exceeded. So I modified it as:
Export.image.toDrive({
image: input.clip(district),
description: 'Test',
maxPixels: 499295920080,
scale: 30
});
This code now gives me 16 split .tif files, with their sizes summing up to a total of 298 MB.
My code remains untouched and I even checked my images of 3 months back, they are not corrupted or incomplete. I cannot figure out what happened in this span of 3 months that led to such differences. Could there be some changes in certain libraries I am using or the LandSat 8 images I am training and classifying on that are leading to such discrepancies?
