Differences... ok... the mosaic mean behaves differently, they get pretty close when you move away from the edges (, but they are 'exact' in their behaviour of calculating the mean. I also can't see where the 'float32' dtype would matter. Someone should try the mosaic data set in pro IF 'float64' is the default...
r_mean.data[:5, :5]
array([[ 10.94, 1.05, 1.2 , 1.14, 0.94],
[ 16.11, 2.31, 3.41, 2.19, 0.6 ],
[ 1.06, 4.59, 1.22, 1.99, 2.45],
[ 1.17, 5.82, 0.86, 0.56, 2.86],
[ 6.88, 0.93, 2.95, 0.42, 0.45]])
r1m.data[:5, :5]
array([[ 10.94, 1.05, 1.2 , 1.14, 0.94],
[ 16.11, 2.31, 3.41, 2.19, 0.6 ],
[ 1.06, 4.59, 1.22, 1.99, 2.45],
[ 1.17, 5.82, 0.86, 0.56, 2.86],
[ 6.88, 0.93, 2.95, 0.42, 0.45]], dtype=float32)
r0m.data[:5, :5]
array([[ 11.09, 1.05, 1.2 , 1.13, 0.93],
[ 16.26, 2.31, 3.4 , 2.17, 0.61],
[ 1.06, 4.57, 1.2 , 1.97, 2.44],
[ 1.16, 5.78, 0.87, 0.56, 2.85],
[ 6.84, 0.93, 2.95, 0.42, 0.44]], dtype=float32)