User memory limit exceeded generating collection mean and time series chart in Earth Engine

Question

I'm new to GEE and I'm trying to create time series of NEX-GDDP Maximum data.

var roi = ee.FeatureCollection('users/prantoroy/Jessore_Dist');
var tempset = ee.ImageCollection("NASA/NEX-GDDP");

var startD = ee.Date('2000-01-01');
var Range = ee.DateRange(startD,startD.advance(20,'year'));

var intersect = tempset.filterDate(Range);

var selectBands = intersect.select('tasmax');

var temptoC = selectBands.map(function (img){
  return img
  .subtract(273.15)
  .copyProperties(img, ['system:time_start']);
});
    

var chart = ui.Chart.image.series({
  imageCollection:temptoC ,
  region:roi,
  reducer: ee.Reducer.mean(),
  scale:10000,
  xProperty:'system:time_start'
}).setOptions({
  title: 'Air Temperature 1990 to 2020 (Mean) Jessore District, Bangladesh',
  vAxis: {title: 'Min Temp'}
});
    
print(chart);
    
var MeanLST = temptoC.mean().clip(roi);

Map.centerObject(roi);
Map.addLayer(MeanLST,
            {min: -1, max:1, palette: ['blue','red']},
            'Max Temp 2000 to 2020');
Map.addLayer(roi);

But every time I get the same error.

Error generating chart: User memory limit exceeded.

I tried decreasing the years to 10, 5 and 2, but the problem still persists. I understand that my code is not efficient at all. Can someone please help me with a working version?

Answer 1

You are trying to calculate a temporal mean and regional mean time series chart for 260,666 observations. This is just a lot of data. Two issues I see here:

1. The NASA/NEX-GDDP dataset includes 21 models and multiple greenhouse gas concentration scenarios, each at daily cadence (each unique combination is an image in the collection). Try filtering to a single model and scenario:

var  nex = ee.ImageCollection("NASA/NEX-GDDP")
  .filterDate('2010-01-01', '2010-01-02');  // 1 day
print('Unfiltered observations:', nex);
print('Models:', ee.List(nex.aggregate_array('model')).distinct());
print('Scenarios:', ee.List(nex.aggregate_array('scenario')).distinct());

// Filter by model and scenario.
var nexFiltered = nex.filter(ee.Filter.and(
  ee.Filter.eq('model', 'BNU-ESM'), ee.Filter.eq('scenario', 'rcp85')));
print('Filtered observations:', nexFiltered);

2. Daily precision for calculating a 20-year mean seems unwarranted (i.e. a mean of monthly means will likely get you about the same answer.) My recommendation would be to use the NEX-DCP30 ensemble dataset, which is a monthly statistic. Note that it also includes a variety of greenhouse gas concentration scenarios, so you'll want to filter by 'scenario'.

Additionally, I see the you are interested in historical climate; note that NEX-GDDP data are modeled projections (there is a "historical" scenario pre-2006, but better historical datasets exist). You might check out ERA5 monthly and daily aggregates instead.