Google Earth Engine: problem with areal calculations on MODIS data

Question

While trying to calculate forest coverage per country with MODIS/006/MOD44B and LSIB data, I get the following error:

FeatureCollection (Error)
Unable to transform edge (89589.000000, 86393.171875 to 89589.000000, 86393.261719) from SR-ORG:6974 PLANAR [231.65635826399998, 0.0, -2.0015109353988E7, 0.0, -231.65635826400006, 1.0007554676994E7] to EPSG:4326.

The main part of code is below:

var modisScale = 250 // m
var forestImg = modis_filtered.gte(forestThreshold).rename('forest_cover');
    var areaImage = forestImg.multiply(ee.Image.pixelArea());
    
    var stats = areaImage.reduceRegions({
      collection: countries,
      reducer: ee.Reducer.sum(),
      scale: modisScale,
    });

I guess GEE stumbles on reprojecting MODIS SR-ORG:6974 projection. However, if I add .reproject('EPSG:4326', null, modisScale); before reducing, everything works.

var forestImg = modis_filtered.gte(forestThreshold).rename('forest_cover');
var areaImage = forestImg.multiply(ee.Image.pixelArea()).reproject('EPSG:4326', null, modisScale); \\ REPROJECTING

var stats = areaImage.reduceRegions({
  collection: countries,
  reducer: ee.Reducer.sum(),
  scale: modisScale,
});

The question is whether this proir reprojection to WGS84 is the correct way to go.

Complete scrit is here

Answer 1

LSIB includes Antarctica, which spans the pole. Filter that out and you should be fine:

var countries = ee.FeatureCollection("USDOS/LSIB_SIMPLE/2017")
    .filter("country_na != 'Antarctica'")