mask = np.ma.masked_where(
modis_proj_array[1] == 0, modis_bin_array
) # remember to use the proj tif 2nd layer as the mask for binary tif creation
cloud_cover_pixel = (mask == -999).sum()
if cloud_cover_pixel == 0:
break
if cloud_cover_pixel > 0:
valid_date.drop(time)
print 'drop bin file at {}'.format(modis_bin_path)
else:
print modis_bin_path
print os.path.isfile(modis_bin_path)
os.remove(modis_bin_path)
new_bin_array = np.where(modis_proj_array[1] != 0,
modis_bin_array, -999)
array_to_raster(output_path=modis_bin_path,
source_path=modis_proj_path,
array_data=new_bin_array)
print 'cloud cover interpolation for {}'.format(
modis_bin_path)
else:
valid_date[bin_col_name].ix[time] = modis_bin_path
valid_date.to_csv(valid_date_path)
else:
print 'failed to create modis binary file'
print 'snow_cover_3: swe and modis binary file is done'
if calculate_total_sublimation:
var_array_sum_dict['total_sub'] = np.where(
mask,
var_array_sum_dict['ueb_Ec'] + var_array_sum_dict['ueb_Es'],
-99999.0)
var_array_sum_dict['water_loss'] = np.where(
mask, 100 * var_array_sum_dict['total_sub'] /
var_array_sum_dict['uebPrec'], -99999.0)
# export array sum as tif
for var in var_array_sum_dict.keys():
sum_tif_path = os.path.join(stats_folder,
'sum_{}_{}.tif'.format(var, year))
if not os.path.isfile(sum_tif_path):
array_to_raster(output_path=sum_tif_path,
source_path=tif_path,
array_data=var_array_sum_dict[var],
no_data=-99999.0)
# step2 get mean from array sum tif ##############################################
# make sure to add the total_sub and water_loss variable
if calculate_total_sublimation:
var_list.extend(['total_sub', 'water_loss'])
# get domain average for annual sum
domain_ave_df = pd.DataFrame(index=range(start_year, end_year),
columns=var_list)
for year in domain_ave_df.index:
for var in domain_ave_df.columns:
sum_tif_path = os.path.join(stats_folder,
'sum_{}_{}.tif'.format(var, year))
if os.path.isfile(sum_tif_path):
var_array_annual_mean_dict['total_sub'] = np.where(
mask,
(var_array_sum_dict['ueb_Ec'] + var_array_sum_dict['ueb_Es']) / yr,
-99999.0)
var_array_sum_dict['water_loss'] = np.where(
mask, 100 * var_array_sum_dict['total_sub'] /
var_array_sum_dict['uebPrec'], -99999.0)
var_array_annual_mean_dict['water_loss'] = np.where(
mask, 100 * var_array_annual_mean_dict['total_sub'] /
var_array_annual_mean_dict['uebPrec'], -99999.0)
# step 2 export array sum and annual mean as tif #############################################################
for var in var_array_sum_dict.keys():
array_to_raster(output_path=os.path.join(stats_folder,
'sum_{}.tif'.format(var)),
source_path=tif_path,
array_data=var_array_sum_dict[var],
no_data=-99999.0)
array_to_raster(output_path=os.path.join(
stats_folder, 'annual_mean_{}.tif'.format(var)),
source_path=tif_path,
array_data=var_array_annual_mean_dict[var],
no_data=-99999.0)
# step3 calculate the domain average #######################################################
if calculate_total_sublimation:
var_list.extend(['total_sub', 'water_loss'])
# calculate the
domain_ave_df = pd.DataFrame(index=var_list, columns=['sum', 'annual_mean'])
np.save(file_path, stack_array)
stack_sum = np.where(mask == False, np.nansum(stack_array, axis=0), -999)
np.save(file_path+'_sum', stack_sum)
stack_sum_dict[data_type] = stack_sum
ma = np.ma.masked_equal(stack_sum, -999, copy=False)
plt.imshow(ma, interpolation='nearest')
plt.colorbar()
plt.title('plot of sum of {} type pixels'.format(data_type))
plt.savefig(file_path + '.png')
plt.clf()
# export result as raster data
array_to_raster(output_path=file_path + '.tif',
source_path=swe_proj_path,
array_data=stack_sum,
no_data=-1)
except Exception as e:
continue
# calculate accurate data and evaluation #########################################################
print 'step3: calculate error stats '
snow = np.where(mask == False, (stack_sum_dict['B'] + stack_sum_dict['D']), -999.0)
dry = np.where(mask == False, (stack_sum_dict['A'] + stack_sum_dict['C']), -999.0)
snow_error = np.where(mask == False, stack_sum_dict['B']*1.0/snow, -999.0)
dry_error = np.where(mask == False, stack_sum_dict['C']*1.0/dry, -999.0)
for data_type, stack in zip(['snow', 'dry', 'snow error', 'dry error'], [snow, dry, snow_error, dry_error]):
try:
file_path = os.path.join(stats_folder, '{}_{}'.format(data_type, model))
np.save(file_path, stack)
days_of_snow = np.where(nan_mask == False,
np.nansum(all_snow_stack, axis=0), np.nan)
file_path = os.path.join(
stats_folder, 'days_of_snow_{}_{}'.format(name, swe_bin_col))
np.save(file_path, days_of_snow)
fig = plt.imshow(days_of_snow, interpolation='nearest')
plt.colorbar()
plt.title('plot of {}'.format('days of snow from {}'.format(name)))
plt.savefig(file_path + '.png')
plt.clf()
# export result as raster data
array_to_raster(output_path=file_path + '.tif',
source_path=swe_bin_path,
array_data=days_of_snow)
# stack layers and calculate oa values
nan_mask = np.isnan(model)
all_stack = np.stack(layer_stack)
missing_data = np.where(nan_mask == False,
np.isnan(all_stack).sum(axis=0), np.nan)
mismatch_data = np.where(nan_mask == False,
np.nansum(abs(all_stack), axis=0), np.nan)
oa_data = 1 - mismatch_data.astype('float32') / (all_stack.shape[0] -
missing_data)
result = {
'oa': oa_data,
'missing': missing_data,