def calc_caloric_production_from_lulc(input_lulc_uri):
# Data from Johnson et al 2016.
calories_per_cell_uri = os.path.join(ag_dir, 'calories_per_cell.tif')
calories_per_cell = nd.ArrayFrame(calories_per_cell_uri)
ndv = calories_per_cell.no_data_value
# Project the full global to robinson (to match volta projeciton and also to allow np.clip_by_shape)
calories_per_cell_projected_uri = os.path.join(
project_dir, 'calories_per_cell_projected.tif')
if not os.path.exists(calories_per_cell_projected_uri):
calories_per_cell_projected = nd.reproject(
calories_per_cell,
calories_per_cell_projected_uri,
epsg_code=54030,
no_data_value=ndv)
else:
calories_per_cell_projected = nd.ArrayFrame(
calories_per_cell_projected_uri)
# Polygon of the Volta (also in Robinson projection)
aoi_uri = 'input/Baseline/PASOS_cuencas_robinson.shp'
# Clip the global data to the volta, but keep at 5 min for math summation reasons later
clipped_calories_per_5m_cell_uri = 'input/Baseline/clipped_calories_per_5m_cell.tif'
if not os.path.exists(clipped_calories_per_5m_cell_uri):
clipped_calories_per_5m_cell = calories_per_cell_projected.clip_by_shape(
aoi_uri,
output_uri=clipped_calories_per_5m_cell_uri,
no_data_value=ndv)
else:
clipped_calories_per_5m_cell = nd.ArrayFrame(
clipped_calories_per_5m_cell_uri)
# Load the baseline lulc for adjustment factor calculation and as a match_af
lulc_uri = 'input/Baseline/lulc.tif'
lulc = nd.ArrayFrame(lulc_uri)
lulc_ndv = lulc.no_data_value
# resample to LULC's resolution. Note that this will change the sum of calories.
calories_resampled_uri = 'input/Baseline/calories_resampled.tif'
if not os.path.exists(calories_resampled_uri):
calories_resampled = clipped_calories_per_5m_cell.resample(
lulc, output_uri=calories_resampled_uri, no_data_value=ndv)
else:
calories_resampled = nd.ArrayFrame(calories_resampled_uri)
# Base on teh assumption that full ag is twice as contianing of calroies as mosaic, allocate the
# caloric presence to these two ag locations. Note that these are still not scaled, but they are
# correct relative to each other.
# This simplification means we are doing the equivilent to the invest crop model beacause
# the cells to allocate are lower res than the target.
unscaled_calories_baseline = np.where(lulc.data == 12,
calories_resampled.data, 0)
unscaled_calories_baseline = np.where(lulc.data == 14,
0.5 * calories_resampled.data,
unscaled_calories_baseline)
# Multiply the unscaled calories by this adjustment factor, which is the ratio between the actual calories present
# calculated from the 5 min resolution data, and the unscaled.
n_calories_present = np.sum(clipped_calories_per_5m_cell)
n_unscaled_calories_in_baseline = np.sum(unscaled_calories_baseline)
adjustment_factor = n_calories_present / n_unscaled_calories_in_baseline
calc_baseline_calories = False
if calc_baseline_calories:
baseline_calories = unscaled_calories_baseline * adjustment_factor
baseline_calories_uri = 'input/Baseline/baseline_calories.tif'
# NOTE, this uses numdal in a weired way because it has THREE inputs (of which the last is jammed into kwargs).
baseline_calories_af = nd.ArrayFrame(baseline_calories,
lulc,
output_uri=baseline_calories_uri,
data_type=6,
no_data_value=ndv)
input_lulc = nd.ArrayFrame(input_lulc_uri)
unscaled_calories_input_lulc = np.where(input_lulc.data == 12,
calories_resampled.data, 0)
unscaled_calories_input_lulc = np.where(input_lulc.data == 14,
0.5 * calories_resampled.data,
unscaled_calories_input_lulc)
output_calories = unscaled_calories_input_lulc * adjustment_factor
output_calories_uri = os.path.join(
project_dir,
'calories_in_' + nd.explode_uri(input_lulc_uri)['file_root'] + '.tif')
output_calories_af = nd.ArrayFrame(output_calories,
lulc,
data_type=7,
no_data_value=ndv,
output_uri=output_calories_uri)
# output_calories_af.save(output_uri=output_calories_uri)
sum_calories = output_calories_af.sum()
print('Calories from ' + input_lulc_uri + ': ' + str(sum_calories))
return sum_calories
def calc_caloric_production_from_lulc_uri(input_lulc_uri, aoi_uri, output_uri):
# First check that the required files exist, creating them if not.
# Data from Johnson et al 2016.
working_dir = os.path.split(os.path.split(input_lulc_uri)[0])[0]
baseline_dir = os.path.join(working_dir, 'Baseline')
scenario_dir = os.path.split(input_lulc_uri)[0]
calories_resampled_uri = os.path.join(scenario_dir, 'calories_per_ha_2000.tif')
if not os.path.exists(calories_resampled_uri):
# Get global 5m calories map from base data
calories_per_cell_uri = os.path.join(ag_dir, 'calories_per_cell.tif')
calories_per_cell = nd.ArrayFrame(calories_per_cell_uri)
ndv = calories_per_cell.no_data_value
# Project the full global map to projection of lulc
calories_per_cell_projected_uri = os.path.join(baseline_dir, 'calories_per_cell_projected.tif')
output_wkt = nd.get_projection_from_uri(input_lulc_uri)
# print('projection', projection)
# output_wkt = projection.ExportToWkt()
calories_per_cell_projected = nd.reproject(calories_per_cell, calories_per_cell_projected_uri,
output_wkt=output_wkt, no_data_value=ndv)
# Clip the global data to the project aoi, but keep at 5 min for math summation reasons later
clipped_calories_per_5m_cell_uri = 'input/Baseline/clipped_calories_per_5m_cell.tif'
clipped_calories_per_5m_cell = calories_per_cell_projected.clip_by_shape(aoi_uri, output_uri=clipped_calories_per_5m_cell_uri,
no_data_value=ndv)
# Load the baseline lulc for adjustment factor calculation and as a match_af
baseline_lulc_uri = os.path.join(baseline_dir, 'lulc.tif')
baseline_lulc = nd.ArrayFrame(baseline_lulc_uri)
input_lulc = nd.ArrayFrame(input_lulc_uri)
# Resample baseline lulc to the intput_lulc (a slight size change happens with the scenario generator)
baseline_resampled_lulc = baseline_lulc.resample(input_lulc, discard_at_exit=True)
# Resample calories to lulc
calories_resampled = clipped_calories_per_5m_cell.resample(input_lulc, output_uri=calories_resampled_uri, no_data_value=ndv)
calories_resampled = None
input_lulc = nd.ArrayFrame(input_lulc_uri)
ndv = input_lulc.no_data_value
calories_resampled = nd.ArrayFrame(calories_resampled_uri)
baseline_lulc_uri = os.path.join(baseline_dir, 'lulc.tif')
baseline_lulc = nd.ArrayFrame(baseline_lulc_uri)
baseline_resampled_lulc = baseline_lulc.resample(input_lulc, discard_at_exit=True)
# baseline_lulc_uri = os.path.join(baseline_dir, 'lulc.tif')
# print('baseline_lulc_uri', baseline_lulc_uri)
# baseline_lulc = nd.ArrayFrame(baseline_lulc_uri)
clipped_calories_per_5m_cell_uri = 'input/Baseline/clipped_calories_per_5m_cell.tif'
clipped_calories_per_5m_cell = nd.ArrayFrame(clipped_calories_per_5m_cell_uri)
# Base on teh assumption that full ag is twice as contianing of calroies as mosaic, allocate the
# caloric presence to these two ag locations. Note that these are still not scaled, but they are
# correct relative to each other.
# This simplification means we are doing the equivilent to the invest crop model beacause
# the cells to allocate are lower res than the target.
unscaled_calories_baseline = np.where(baseline_resampled_lulc.data == 12, calories_resampled.data, 0)
unscaled_calories_baseline = np.where(baseline_resampled_lulc.data == 14, 0.5 * calories_resampled.data, unscaled_calories_baseline)
# Multiply the unscaled calories by this adjustment factor, which is the ratio between the actual calories present
# calculated from the 5 min resolution data, and the unscaled.
n_calories_present = np.sum(clipped_calories_per_5m_cell)
n_unscaled_calories_in_baseline = np.sum(unscaled_calories_baseline)
adjustment_factor = n_calories_present / n_unscaled_calories_in_baseline
unscaled_calories_input_lulc = np.where(input_lulc.data == 12, calories_resampled.data, 0)
unscaled_calories_input_lulc = np.where(input_lulc.data == 14, 0.5 * calories_resampled.data, unscaled_calories_input_lulc)
output_calories = unscaled_calories_input_lulc * adjustment_factor
output_calories_af = nd.ArrayFrame(output_calories, input_lulc, data_type=6, no_data_value=ndv, output_uri=output_uri)
print('Sum of ' + output_calories_af.uri + ': ' + str(output_calories_af.sum()))