Error in Google Earth Engine: Computed value is too large - classification

Question

I'm trying to achieve classification for image collection over 30 years as the code shows below:
var roi = ee.FeatureCollection("users/471033961/YellowRiver");
var empty = ee.Image().toByte();
var outline = empty.paint({
featureCollection:roi,
color:0,
width:3
});
Map.addLayer(outline, {palette: "ff0000"}, "outline");
//////////////////////////////landsat 5,7
function l57class(year){
var startDate = ee.Date.fromYMD(year, 1, 1);
var endDate = ee.Date.fromYMD(year+1, 1, 1);
var LC57_BANDS = ['B1', 'B2', 'B3', 'B4', 'B5', 'B7']; //Landsat57
var STD_NAMES = ['B2', 'B3', 'B4', 'B5', 'B6', 'B7'];
var cloudMaskL457 = function(image) {
var qa = image.select('pixel_qa');
// If the cloud bit (5) is set and the cloud confidence (7) is high
// or the cloud shadow bit is set (3), then it's a bad pixel.
var cloud = qa.bitwiseAnd(1 << 5)
.and(qa.bitwiseAnd(1 << 7))
.or(qa.bitwiseAnd(1 << 3));
// Remove edge pixels that don't occur in all bands
var mask2 = image.mask().reduce(ee.Reducer.min());
return image.updateMask(cloud.not()).updateMask(mask2);
}
var collection1 = ee.ImageCollection("LANDSAT/LE07/C01/T1_SR")
.filterDate(startDate, endDate)
.filterBounds(roi)
.filter(ee.Filter.lte('CLOUD_COVER',10))
.map(cloudMaskL457)
var collection2 = ee.ImageCollection("LANDSAT/LT05/C01/T1_SR")
.filterDate(startDate, endDate)
.filterBounds(roi)
.filter(ee.Filter.lte('CLOUD_COVER',10))
.map(cloudMaskL457)
var collection = collection1.merge(collection2).select(LC57_BANDS, STD_NAMES)
var image=collection.mosaic().clip(roi)
function NDWI(img){
var nir = img.select('B4');
var green = img.select('B2');
var ndwi = img.expression(
'(B3-B5)/(B3+B5)',
{
'B5':nir,
'B3':green
});
return ndwi;
}
function EWI(img){
var swir1 = img.select('B6')
var nir = img.select('B5');
var green = img.select('B3');
var ewi = img.expression(
'(B3-B5-B6)/(B3+B5+B6)',
{
'B6':swir1,
'B5':nir,
'B3':green
});
return ewi;
}
var MNDWI = image.normalizedDifference(['B3', 'B6']).rename('MNDWI');
var ndbi = image.normalizedDifference(['B6', 'B5']).rename('NDBI');
var ndvi = image.normalizedDifference(['B5', 'B4']).rename('NDVI');
var ewi = EWI(image).rename('EWI');
var ndwi = NDWI(image).rename('NDWI');
var lswi = image.normalizedDifference(['B5','B6']).rename('LSWI')
var nbr2 = image.normalizedDifference(["B6", "B7"]).rename("NBR2")
var awei = image.expression(
'4*(green-SWIR1)-(0.25*NIR+2.75*SWIR2)',{
green:image.select('B3'),
NIR:image.select('B5'),
SWIR1:image.select('B6'),
SWIR2:image.select('B7'),
}).float().rename('AWEI')
image=image
.addBands(ndvi)
.addBands(ndbi)
.addBands(MNDWI)
.addBands(ndwi)
.addBands(ewi)
.addBands(lswi)
.addBands(awei)
.addBands(nbr2)
var classNames = city.merge(water).merge(tree).merge(crop).merge(bare);
var bands = ['B2', 'B3', 'B4', 'B5', 'B6', 'B7','MNDWI','NDBI','NDVI','EWI','NDWI','AWEI','LSWI',"NBR2"];
var training = image.select(bands).sampleRegions({
collection: classNames,
properties: ['landcover'],
scale: 30
});
var withRandom = training.randomColumn('random');
var split = 0.8;
var trainingPartition = withRandom.filter(ee.Filter.lt('random', split));
var testingPartition = withRandom.filter(ee.Filter.gte('random', split));
var classProperty = 'landcover';
var classifier = ee.Classifier.smileRandomForest(100).train({
features: trainingPartition,
classProperty: 'landcover',
inputProperties: bands
});
var classified = image.select(bands).classify(classifier);
var test = testingPartition.classify(classifier);
var confusionMatrix = test.errorMatrix('landcover', 'classification');
print(year)
print('confusionMatrix',confusionMatrix);
print('overall accuracy', confusionMatrix.accuracy());
print('kappa accuracy', confusionMatrix.kappa());
Map.addLayer(classified,{}, "classified"+year);
Export.image.toDrive({
image: classified,
description: 'YR_cart'+year,
folder: 'YR_cart'+year,
scale: 30,
region: roi,
maxPixels:34e10
});
}
////////////////////////////////////////////////////////////////////////////////
function l8class(year){
var startDate = ee.Date.fromYMD(year, 1, 1);
var endDate = ee.Date.fromYMD(year+1, 1, 1);
var collection = ee.ImageCollection("LANDSAT/LC08/C01/T1_RT")
.filterDate(startDate, endDate)
.filterBounds(roi)
.filter(ee.Filter.lte('CLOUD_COVER',20))
var remove_cloud=function(collection)
{
var fun_calCloudScore = function(image) {
return ee.Algorithms.Landsat.simpleCloudScore(ee.Image(image));
};
var calCloudScore = ee.ImageCollection(collection)
.map(fun_calCloudScore)
;
var fun_maskCloud = function(image) {
var mask = image.select(['cloud']).lte(10);
return image.updateMask(mask);
};
var vsParam={band:['B2', 'B3', 'B4', 'B5', 'B6', 'B7'],min:0,max:0.3};
var maskCloud = ee.ImageCollection(calCloudScore)
.map(fun_maskCloud) ;
var maskCloud2=maskCloud.mean()
return maskCloud;
}
var image=remove_cloud(collection).mosaic().clip(roi);
function NDWI(img){
var nir = img.select('B4');
var green = img.select('B2');
var ndwi = img.expression(
'(B3-B5)/(B3+B5)',
{
'B5':nir,
'B3':green
});
return ndwi;
}
function EWI(img){
var swir1 = img.select('B6')
var nir = img.select('B5');
var green = img.select('B3');
var ewi = img.expression(
'(B3-B5-B6)/(B3+B5+B6)',
{
'B6':swir1,
'B5':nir,
'B3':green
});
return ewi;
}
var MNDWI = image.normalizedDifference(['B3', 'B6']).rename('MNDWI');
var ndbi = image.normalizedDifference(['B6', 'B5']).rename('NDBI');
var ndvi = image.normalizedDifference(['B5', 'B4']).rename('NDVI');
var ewi = EWI(image).rename('EWI');
var ndwi = NDWI(image).rename('NDWI');
var lswi = image.normalizedDifference(['B5','B6']).rename('LSWI')
var nbr2 = image.normalizedDifference(["B6", "B7"]).rename("NBR2")
var awei = image.expression(
'4*(green-SWIR1)-(0.25*NIR+2.75*SWIR2)',{
green:image.select('B3'),
NIR:image.select('B5'),
SWIR1:image.select('B6'),
SWIR2:image.select('B7'),
}).float().rename('AWEI')
image=image
.addBands(ndvi)
.addBands(ndbi)
.addBands(MNDWI)
.addBands(ndwi)
.addBands(ewi)
.addBands(lswi)
.addBands(awei)
.addBands(nbr2)
var classNames = city.merge(water).merge(tree).merge(crop).merge(bare);
var bands = ['B2', 'B3', 'B4', 'B5', 'B6', 'B7','MNDWI','NDBI','NDVI','EWI','NDWI','AWEI','LSWI',"NBR2"];
var training = image.select(bands).sampleRegions({
collection: classNames,
properties: ['landcover'],
scale: 30
});
var withRandom = training.randomColumn('random');
var split = 0.9;
var trainingPartition = withRandom.filter(ee.Filter.lt('random', split));
var testingPartition = withRandom.filter(ee.Filter.gte('random', split));
var classProperty = 'landcover';
var classifier = ee.Classifier.libsvm().train({
features: trainingPartition,
classProperty: 'landcover',
inputProperties: bands
});
var classified = image.select(bands).classify(classifier);
var test = testingPartition.classify(classifier);
var confusionMatrix = test.errorMatrix('landcover', 'classification');
print(year)
print('confusionMatrix',confusionMatrix);
print('overall accuracy', confusionMatrix.accuracy());
print('kappa accuracy', confusionMatrix.kappa());
Map.addLayer(classified,{}, "classified"+year);
Export.image.toDrive({
image: classified,
description: 'YR_cart'+year,
folder: 'YR_cart'+year,
scale: 30,
region: roi,
maxPixels:34e10
});
}
function main() {
var startYear = 2022;
var endYear = 2022;
for (var year=startYear; year<=endYear; year++) {
if (year >2012){l8class(year)}else{l57class(year)}
}
}
main();
I knew the study area was too big which lead to an error: "computed value is too large". Does anyone know how to solve this problem? better to revise based on this code if so.
Really appreciate it in advance!
To be honest I checked too many ideas but one of the effective ways I thought is to "break your study area into a bunch of tiles to calculate first, then combine all the results to get the final. However, It's just initial thinking that I didn't test.
Here is the link(you can ignore the comment in Chinese): https://code.earthengine.google.com/5d362229b959d96b9e4017571ba11a75