reduceRegion() gives NULL for some small polygons

Question

I have a function that calculates the average rainfall over the first 6 months of the year over a 30 year period. The function receives a geometry, and returns the mean of pixels inside the geometry.

In some cases, with small polygons (about the size of a pixel) the result is NULL. it is not always the case, in tests I have made considerably smaller polygons that have given a result. Here is a code snippet that creates the null value.

var geometry = /* color: #0b4a8b */ee.Geometry.Polygon(
        [[[24.36479088838034, -33.36746983993103],
          [24.36024158347382, -33.42516561247864],
          [24.438305535991674, -33.4257574802331],
          [24.43181381987206, -33.37553285357873]]]);

var rain_temp = ee.ImageCollection("ECMWF/ERA5_LAND/MONTHLY_AGGR");

Map.addLayer(geometry);
Map.centerObject(geometry);
var avgRainfall6Coldest = function(polygon, startMonth){
  var startYear = 1990;
  var endYear = 2020;
  
  var rain_tempYears = rain_temp.filterDate(ee.Date.fromYMD(startYear,1,1),ee.Date.fromYMD(endYear,1,1));
  
  rain_tempYears = rain_tempYears.filterBounds(polygon);
  var monthlyAverages = function(month){
    return rain_tempYears.filter(ee.Filter.calendarRange(month,month,"month")).reduce(ee.Reducer.mean());
  };
  
  var yearlyAvg = ee.ImageCollection(ee.List.sequence(1,6).map(monthlyAverages)).reduce(ee.Reducer.sum());
  var imageVisParam = {"opacity":1,"bands":["total_precipitation_sum_mean_sum"],"min":0,"max":0.600,"gamma":1};
  Map.addLayer(yearlyAvg,imageVisParam);
  print(rain_tempYears.aggregate_count_distinct("year"));
  
  return yearlyAvg.select("total_precipitation_sum_mean_sum").reduceRegion(ee.Reducer.mean(),polygon);

};



print(avgRainfall6Coldest(geometry));

Link to the script in Earth Engine. https://code.earthengine.google.com/ec3cfbafb787c71658497e7b910d9acf

Any idea why this is occurring, and could there be a workaround?

Answer 1

Adding scale for the projection to work on inside .reduceRegion()should fix this here is a working version

Added line 11

var scale = rain_temp.first().projection().nominalScale().aside(print,'the scale of ERA5')

and line 31

.reduceRegion(ee.Reducer.mean(),polygon, scale)

Another thing to pay attention to when reducing over polygon is how pixel fractions are calculated/weighted see here and for more about scale here