I have a collection of 12 .grd files located in the same directory, all formatted as 2-dimensional arrays.
To calculate their average values, I initially used a simple loop method
ave_value = np.zeros_like(test_array) # creating an array with zeroes
for i in range(0,12,1):
file_path = %(i+1)+'.grd' #1.grd, 2.grd,...
# read_grd is a custom function
value, nodata_value = read_grd(file_path)
value[value == nodata_value] = np.nan
ave_value_+=value
ave_value = ave_value/12.0
However, I soon realized that this approach did not account for positions containing np.nan values.
For instance, while position [2,2] in one grid file was np.nan, it contained actual values in other datasets. In the averaging process mentioned above, the resulting average value in ave-value also turned out to be np.nan.
Is there a more effective solution to average different 2-dimensional arrays considering the presence of np.nan values? For example, if the specific position contains np.nan values in 3 files, we should only average the corresponding values from the remaining 9 files.
Any advice or suggestions you can provide would be greatly appreciated.