Different results every time I calculate a weighted area average

[JackB]

Hi,

I'm trying to calculate the Niño 3.4 index from an netCDF file of ERA5 monthly SSTs from 1979 to 2019, but I'm getting quite a baffling problem. 

Part of the calculation requires a weighted area average to be found for the grid box [-5°N, 5°N, 170°W, 120°W]. However, every time I run this code, I get a different result for the weighted average with time. I'm really not sure what's going on as the grid area is always the same - the only bit that changes every time is the final weighted_avg cube. 

I've attached the code used below. The cube that is loaded in is just the ERA5 monthly SSTs from 1979 to 2019 (file size is 3.8 GB). I would be very grateful for any advice.

import iris

import iris.analysis

import iris.coord_categorisation

cube = iris.load_cube('.../ERA5/Monthly_SSTs.nc')

iris.coord_categorisation.add_month(cube, 'time', name='month')

iris.coord_categorisation.add_year(cube, 'time', name='year')

lat = cube.extract(iris.Constraint(latitude=lambda x: -5 <= x <= 5))

latlon = lat.intersection(longitude=(-170,-120))

    

latlon.coord('latitude').guess_bounds()

latlon.coord('longitude').guess_bounds()

grid_areas = iris.analysis.cartography.area_weights(latlon)

    

weighted_avg = latlon.collapsed(['latitude', 'longitude'], iris.analysis.MEAN, weights=grid_areas)

print (weighted_avg[0:10].data)  # This gives a different result every time I run the code