def get_wrf_fitness(param_ids, start_date='Jan 15 2011', end_date='Jan 16 2011',
bc_data='ERA', n_domains=1, correction_factor=0.0004218304553577255,
setup_yaml='dirpath.yml', disable_timeout=False, verbose=False):
"""
Using the input physics parameters, date, boundary condition, and domain data,
this function runs the WRF model and computes the error between WRF and ERA5.
:param param_ids: list of integers
corresponding to each WRF physics parameterization.
:param start_date: string
specifying a desired start date.
:param end_date: string
specifying a desired end date.
:param bc_data: string
specifying the boundary condition data to be used for the WRF forecast.
Currently, only ERA data (ds627.0) is supported.
:param n_domains: integer
specifing the number of WRF model domains. 1 - 3 domains are currently supported.
:param correction_factor: float
capuring the relationship between GHI and wind power density (WPD) errors averaged
across an entrie year. Calculated using opwrf/examples/Fitness_correction_factor.py.
:param setup_yaml: string
defining the path to the yaml file where input directory paths are specified.
:param disable_timeout: boolean (default = False)
telling runwrf if subprogram timeouts are allowed or not.
:param verbose: boolean (default = False)
instructing the program to print everything or just key information to the screen.
:return fitness: float
value denoting how well the WRF model run performed.
Fitness is a measure of accumlated error, so a lower value is better.
"""
if verbose:
print('- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -')
print('\nCalculating fitness for: {}'.format(param_ids))
# Create a WRFModel instance
wrf_sim = WRFModel(param_ids, start_date, end_date,
bc_data=bc_data, n_domains=n_domains, setup_yaml=setup_yaml, verbose=verbose)
# Check to see if WRFModel instance exists; if not, run the WRF model.
wrfout_file_path = wrf_sim.DIR_WRFOUT + 'wrfout_d01.nc'
orig_wrfout_file_path = wrf_sim.DIR_WRFOUT + 'wrfout_d01_' \
+ wrf_sim.forecast_start.strftime('%Y') + '-' \
+ wrf_sim.forecast_start.strftime('%m') + '-' \
+ wrf_sim.forecast_start.strftime('%d') + '_00:00:00'
if [os.path.exists(file) for file in [wrfout_file_path, orig_wrfout_file_path]].count(True) is 0:
# Next, get boundary condition data for the simulation
# ERA is the only supported data type right now.
vtable_sfx = wrf_sim.get_bc_data()
# Setup the working directory to run the simulation
success = wrf_sim.wrfdir_setup(vtable_sfx)
# Prepare the namelists
if success:
success = wrf_sim.prepare_namelists()
# Run WPS
if success:
success = wrf_sim.run_wps(disable_timeout)
if verbose:
print(f'WPS ran successfully? {success}')
# Run REAL
if success:
success = wrf_sim.run_real(disable_timeout)
if verbose:
print(f'Real ran successfully? {success}')
# RUN WRF
if success:
success, runtime = wrf_sim.run_wrf(disable_timeout)
if verbose:
print(f'WRF ran successfully? {success}')
else:
runtime = '00h 00m 00s'
else:
success = True
runtime = '00h 00m 00s'
# Postprocess wrfout file and ERA5 data
if success:
proc_wrfout_file_path = wrf_sim.DIR_WRFOUT + 'wrfout_processed_d01.nc'
if not os.path.exists(proc_wrfout_file_path):
if verbose:
print(f'Postprocessing wrfout file...')
success = wrf_sim.process_wrfout_data()
wrf_sim.process_era5_data()
# Compute the error between WRF run and ERA5 dataset and return fitness
if success:
mae = wrf_sim.wrf_era5_diff()
ghi_total_error = mae[1]
wpd_total_error = mae[2]
daylight_factor = hf.daylight_frac(start_date) # daylight fraction
fitness = daylight_factor * ghi_total_error + correction_factor * wpd_total_error
if verbose:
print(f'!!! Physics options set {param_ids} has fitness {fitness}')
else:
ghi_total_error = 6.022 * 10 ** 23
wpd_total_error = 6.022 * 10 ** 23
fitness = 6.022 * 10 ** 23
return fitness, ghi_total_error, wpd_total_error, runtime
def test_daylight_frac():
frac = daylight_frac('Jul 1, 2020')
print(f'The daylight fraction is: {frac}')
assert 0 <= frac <= 1
def wrf_errorandfitness_plot(wrfds,
paramstr,
save_fig=False,
wrf_dir='./',
era_dir='./',
fig_path='./',
verbose=False,
fitness_short_title='Model Fitness',
ghi_error_short_title='GHI Error (kWh m-2 day-1)',
wpd_error_short_title='WPD Error (kWh m-2 day-1)',
**kwargs):
"""
Plots the GHI error, WPD error, and fitness maps all within one 3-panel plot.
:param wrfds:
:param paramstr:
:param save_fig:
:param wrf_dir:
:param era_dir:
:param fig_path:
:param verbose:
:param fitness_short_title:
:param ghi_error_short_title:
:param wpd_error_short_title:
:param kwargs:
:return: None
"""
# Get the start_date and create the date string
start_date = str(wrfds.Time.dt.strftime('%b %d %Y')[0].values)
datestr = str(wrfds.Time.dt.strftime('%Y-%m-%d')[0].values)
# To start, we need to get the WRF map projection information (a Lambert Conformal grid),
# and find the domain boundaries in this projection.
# NOTE: this task MUST occurr before we regrid the WRF variables or the coordinates change and become incompatible.
wrf_cartopy_proj = get_wrf_proj(wrfds, 'dni')
proj_bounds = get_domain_boundary(wrfds, wrf_cartopy_proj)
if verbose:
print(f'WRF Projection:\n{wrf_cartopy_proj}')
print(f'\nDomain Boundaries:\n{proj_bounds}')
# Regrid the wrf GHI and WPD
# Due to some obnioxious xarry nuance, the easiest way to keep the regridding function from adding additional
# variables and coordinates to the xarray dataset in the outter scope, is just to open new datasets within
# wrf_era5_regrid_xesmf; so that's why we must again specify the wrf and era5 file names below.
wrf_file = f'wrfout_processed_d01_{datestr}_{paramstr}.nc'
era_file = f'ERA5_EastUS_WPD-GHI_{datestr.split("-")[0]}-{datestr.split("-")[1]}.nc'
wrfds, eradata = optwrf.regridding.wrf_era5_regrid_xesmf(wrfdir=wrf_dir,
wrffile=wrf_file,
eradir=era_dir,
erafile=era_file)
# Calculate the error in GHI and WPD
wrfds = optwrf.regridding.wrf_era5_error(wrfds, eradata)
# Calculate the fitness
correction_factor = 0.0004218304553577255
daylight_factor = helper_functions.daylight_frac(
start_date) # daylight fraction
wrfds[
'fitness'] = daylight_factor * wrfds.total_ghi_error + correction_factor * wrfds.total_wpd_error
# Create a figure
fig = plt.figure(figsize=(9.5, 3))
# Set the GeoAxes to the projection used by WRF
ax_fitness = fig.add_subplot(1, 3, 1, projection=wrf_cartopy_proj)
ax_ghierr = fig.add_subplot(1,
3,
2,
projection=wrf_cartopy_proj,
sharey=ax_fitness)
ax_wpderr = fig.add_subplot(1,
3,
3,
projection=wrf_cartopy_proj,
sharey=ax_ghierr)
# Create the filled contour levels
fitness_cn = ax_fitness.contourf(
wrfpy.to_np(wrfds.lon),
wrfpy.to_np(wrfds.lat),
wrfpy.to_np(wrfds['fitness']),
# np.linspace(0, np.amax(wrfds['fitness']), 22),
np.linspace(0, 7, 22),
transform=ccrs.PlateCarree(),
cmap=get_cmap("Greys"))
ghierr_cn = ax_ghierr.contourf(
wrfpy.to_np(wrfds.lon),
wrfpy.to_np(wrfds.lat),
wrfpy.to_np(wrfds['total_ghi_error']),
# np.linspace(0, np.amax(wrfds['total_ghi_error']), 22),
np.linspace(0, 2.5, 22),
transform=ccrs.PlateCarree(),
cmap=get_cmap("hot_r"))
wpderr_cn = ax_wpderr.contourf(
wrfpy.to_np(wrfds.lon),
wrfpy.to_np(wrfds.lat),
wrfpy.to_np(wrfds['total_wpd_error']),
# np.linspace(0, np.amax(wrfds['total_wpd_error']), 22),
# np.linspace(0, 5000, 22),
locator=ticker.LogLocator(subs='all'),
norm=LogNorm(vmax=5000),
transform=ccrs.PlateCarree(),
cmap=get_cmap("Greens"))
# Format the axes
time_string_f = wrfds.Time[0].dt.strftime('%b %d, %Y')
format_cnplot_axis(
ax_fitness,
fitness_cn,
proj_bounds,
title_str=f'{fitness_short_title}\n{time_string_f.values}',
cbar_ticks=[0, 1, 2, 3, 4, 5, 6, 7],
cbar_tick_labels=['0', '1', '2', '3', '4', '5', '6', '7'])
format_cnplot_axis(
ax_ghierr,
ghierr_cn,
proj_bounds,
title_str=f'{ghi_error_short_title}\n{time_string_f.values}',
cbar_ticks=[0, 0.5, 1.0, 1.5, 2.0, 2.5],
cbar_tick_labels=['0', '0.5', '1.0', '1.5', '2.0', '2.5'])
format_cnplot_axis(
ax_wpderr,
wpderr_cn,
proj_bounds,
title_str=f'{wpd_error_short_title}\n{time_string_f.values}',
cbar_ticks=ticker.LogLocator())
if save_fig:
file_type = kwargs.get('file_type', '.pdf')
plt.savefig(fig_path + file_type,
dpi=300,
transparent=True,
bbox_inches='tight')
plt.show()
def wrf_era5_plot(var,
wrfds,
erads,
paramstr,
src='wrf',
hourly=False,
save_fig=False,
wrf_dir='./',
era_dir='./',
short_title_str='Title',
fig_path='./',
verbose=False,
**kwargs):
"""
Creates a single WRF or ERA5 plot, using the WRF bounds, producing either a plot every hour
or a single plot for the day.
:param var:
:param wrfds:
:param erads:
:param paramstr:
:param src:
:param hourly:
:param save_fig:
:param wrf_dir:
:param era_dir:
:param short_title_str:
:param fig_path:
:param verbose:
:param kwargs:
:return: None
"""
# Format the var input
if type(var) is not str:
print(f'The var input, {var}, must be a string.')
raise TypeError
if var in ['GHI', 'ghi']:
var = 'ghi'
wrf_var = 'ghi'
era_var = 'GHI'
elif var in ['WPD', 'wpd']:
var = 'wpd'
wrf_var = 'wpd'
era_var = 'WPD'
elif var in ['ghi_error', 'GHI_ERROR']:
var = 'ghi_error'
wrf_var = var
elif var in ['wpd_error', 'WPD_ERROR']:
var = 'wpd_error'
wrf_var = var
elif var in ['fitness', 'Fitness', 'FITNESS']:
var = 'fitness'
wrf_var = var
else:
print(f'Variable {var} is not supported.')
raise KeyError
# Get the start_date and create the date string
datestr = str(wrfds.Time.dt.strftime('%Y-%m-%d')[0].values)
# To start, we need to get the WRF map projection information (a Lambert Conformal grid),
# and find the domain boundaries in this projection.
# NOTE: this task MUST occurr before we regrid the WRF variables or the coordinates change and become incompatible.
wrf_cartopy_proj = get_wrf_proj(wrfds, 'dni')
proj_bounds = get_domain_boundary(wrfds, wrf_cartopy_proj)
# We can use a basic Plate Carree projection for ERA5
era5_cartopy_proj = ccrs.PlateCarree()
# Now, get the desired variables
if var in ['ghi_error', 'wpd_error', 'fitness']:
# Due to some obnioxious xarry nuance, the easiest way to keep the regridding function from adding additional
# variables and coordinates to the xarray dataset in the outter scope, is just to open new datasets within
# wrf_era5_regrid_xesmf; so that's why we must again specify the wrf and era5 file names below.
wrf_file = f'wrfout_processed_d01_{datestr}_{paramstr}.nc'
era_file = f'ERA5_EastUS_WPD-GHI_{datestr.split("-")[0]}-{datestr.split("-")[1]}.nc'
wrfdat_proc, eradat_proc = optwrf.regridding.wrf_era5_regrid_xesmf(
wrfdir=wrf_dir, wrffile=wrf_file, eradir=era_dir, erafile=era_file)
# Calculate the error in GHI and WPD
wrfdat_proc = optwrf.regridding.wrf_era5_error(wrfdat_proc,
eradat_proc)
if verbose:
print(wrfdat_proc)
# Calculate the fitness
correction_factor = 0.0004218304553577255
daylight_factor = helper_functions.daylight_frac(
datestr) # daylight fraction
wrfdat_proc['fitness'] = daylight_factor * wrfdat_proc.ghi_error \
+ correction_factor * wrfdat_proc.wpd_error
# Define the time indicies from the times variable
time_indicies = range(0, len(wrfds.Time))
# Format the times for title slides
times_strings_f = wrfds.Time.dt.strftime('%b %d, %Y %H:%M')
# Get the desired variable(s)
for tidx in time_indicies:
timestr = wrfds.Time[tidx].values
timestr_f = times_strings_f[tidx].values
if hourly:
title_str = f'{short_title_str}\n{timestr_f} (UTC)'
else:
time_string_f = wrfds.Time[0].dt.strftime('%b %d, %Y')
title_str = f'{short_title_str}\n{time_string_f.values}'
# WRF Variable (divide by 1000 to convert from W to kW or Wh to kWh)
if not hourly and tidx != 0:
if src == 'wrf':
if var in ['ghi', 'wpd']:
plot_var = plot_var + (wrfds[wrf_var].sel(
Time=np.datetime_as_string(timestr)) / 1000)
else:
plot_var = plot_var + wrfdat_proc[wrf_var].sel(
Time=np.datetime_as_string(timestr))
elif src == 'era5':
if var in ['ghi', 'wpd']:
plot_var = plot_var + (erads[era_var].sel(
Time=np.datetime_as_string(timestr)) / 1000)
else:
print(f'Variable {var} is not provided by {src}')
raise ValueError
else:
if src == 'wrf':
if var in ['ghi', 'wpd']:
plot_var = wrfds[wrf_var].sel(
Time=np.datetime_as_string(timestr)) / 1000
else:
plot_var = wrfdat_proc[wrf_var].sel(
Time=np.datetime_as_string(timestr))
elif src == 'era5':
if var in ['ghi', 'wpd']:
plot_var = erads[era_var].sel(
Time=np.datetime_as_string(timestr)) / 1000
else:
print(f'Variable {var} is not provided by {src}')
raise ValueError
if hourly:
# Create a figure
fig = plt.figure(figsize=(4, 4))
# Set the GeoAxes to the projection used by WRF
ax = fig.add_subplot(1, 1, 1, projection=wrf_cartopy_proj)
# Make the countour lines for filled contours for the GHI
contour_levels = specify_contour_levels(var, hourly=True, **kwargs)
# Add the filled contour levels
color_map = specify_clormap(var)
if src == 'wrf' and var in ['ghi', 'wpd']:
cn = ax.contourf(wrfpy.to_np(wrfds.lon),
wrfpy.to_np(wrfds.lat),
wrfpy.to_np(plot_var),
contour_levels,
transform=ccrs.PlateCarree(),
cmap=color_map)
else:
cn = ax.contourf(wrfpy.to_np(erads.longitude),
wrfpy.to_np(erads.latitude),
wrfpy.to_np(plot_var),
contour_levels,
transform=ccrs.PlateCarree(),
cmap=color_map)
# Format the plot
format_cnplot_axis(ax, cn, proj_bounds, title_str=title_str)
# Save the figure(s)
if save_fig:
fig_path_temp = fig_path + str(tidx).zfill(2)
plt.savefig(fig_path_temp + '.png',
dpi=300,
transparent=True,
bbox_inches='tight')
plt.show()
if not hourly:
# Create a figure
fig = plt.figure(figsize=(4, 4))
# Set the GeoAxes to the projection used by WRF
ax = fig.add_subplot(1, 1, 1, projection=wrf_cartopy_proj)
# Make the countour lines for filled contours
contour_levels = specify_contour_levels(var, hourly=False, **kwargs)
# Add the filled contour levels
color_map = specify_clormap(var)
if src == 'wrf' and var in ['ghi', 'wpd']:
cn = ax.contourf(wrfpy.to_np(wrfds.lon),
wrfpy.to_np(wrfds.lat),
wrfpy.to_np(plot_var),
contour_levels,
transform=ccrs.PlateCarree(),
cmap=color_map)
else:
cn = ax.contourf(wrfpy.to_np(erads.longitude),
wrfpy.to_np(erads.latitude),
wrfpy.to_np(plot_var),
contour_levels,
transform=ccrs.PlateCarree(),
cmap=color_map)
# Format the plot
format_cnplot_axis(ax, cn, proj_bounds, title_str=title_str)
# Save the figure(s)
if save_fig:
file_type = kwargs.get('file_type', '.pdf')
plt.savefig(fig_path + file_type,
dpi=300,
transparent=True,
bbox_inches='tight')
plt.show()
def wrf_errorandfitness_plot(wrfdata, save_fig=False, wrf_dir='./', era_dir='./',
fig_path='./', verbose=False, fitness_short_title='Model Fitness',
ghi_error_short_title='GHI Error (kWh m-2 day-1)',
wpd_error_short_title='WPD Error (kWh m-2 day-1)'):
"""
:return:
"""
# Get the start_date and create the date string
start_date = str(wrfdata.Time.dt.strftime('%b %d %Y')[0].values)
# To start, we need to get the WRF map projection information (a Lambert Conformal grid),
# and find the domain boundaries in this projection.
# NOTE: this task MUST occurr before we regrid the WRF variables or the coordinates change and become incompatible.
wrf_cartopy_proj = get_wrf_proj(wrfdata, 'dni')
proj_bounds = get_domain_boundary(wrfdata, 'ghi', wrf_cartopy_proj)
if verbose:
print(f'WRF Projection:\n{wrf_cartopy_proj}')
print(f'\nDomain Boundaries:\n{proj_bounds}')
# Regrid the wrf GHI and WPD
input_year = helper_functions.format_date(start_date).strftime('%Y')
input_month = helper_functions.format_date(start_date).strftime('%m')
wrfdata, eradata = optwrf.regridding.wrf_era5_regrid_xesmf(input_year, input_month,
wrfdir=wrf_dir, eradir=era_dir)
# Calculate the error in GHI and WPD
wrfdata = optwrf.regridding.wrf_era5_error(wrfdata, eradata)
# Calculate the fitness
correction_factor = 0.0004218304553577255
daylight_factor = helper_functions.daylight_frac(start_date) # daylight fraction
wrfdata['fitness'] = daylight_factor * wrfdata.total_ghi_error + correction_factor * wrfdata.total_wpd_error
# Create a figure
fig = plt.figure(figsize=(9.5, 3))
# Set the GeoAxes to the projection used by WRF
ax_fitness = fig.add_subplot(1, 3, 1, projection=wrf_cartopy_proj)
ax_ghierr = fig.add_subplot(1, 3, 2, projection=wrf_cartopy_proj, sharey=ax_fitness)
ax_wpderr = fig.add_subplot(1, 3, 3, projection=wrf_cartopy_proj, sharey=ax_ghierr)
# Create the filled contour levels
fitness_cn = ax_fitness.contourf(wrfpy.to_np(wrfdata.lon), wrfpy.to_np(wrfdata.lat),
wrfpy.to_np(wrfdata['fitness']),
np.linspace(0, np.amax(wrfdata['fitness']), 22),
transform=ccrs.PlateCarree(), cmap=get_cmap("Greys"))
ghierr_cn = ax_ghierr.contourf(wrfpy.to_np(wrfdata.lon), wrfpy.to_np(wrfdata.lat),
wrfpy.to_np(wrfdata['total_ghi_error']),
np.linspace(0, np.amax(wrfdata['total_ghi_error']), 22),
transform=ccrs.PlateCarree(), cmap=get_cmap("hot_r"))
wpderr_cn = ax_wpderr.contourf(wrfpy.to_np(wrfdata.lon), wrfpy.to_np(wrfdata.lat),
wrfpy.to_np(wrfdata['total_wpd_error']),
np.linspace(0, np.amax(wrfdata['total_wpd_error']), 22),
transform=ccrs.PlateCarree(), cmap=get_cmap("Greens"))
# Format the axes
time_string_f = wrfdata.Time[0].dt.strftime('%b %d, %Y')
format_cnplot_axis(ax_fitness, fitness_cn, proj_bounds,
title_str=f'{fitness_short_title}\n{time_string_f.values}')
format_cnplot_axis(ax_ghierr, ghierr_cn, proj_bounds,
title_str=f'{ghi_error_short_title}\n{time_string_f.values}')
format_cnplot_axis(ax_wpderr, wpderr_cn, proj_bounds,
title_str=f'{wpd_error_short_title}\n{time_string_f.values}')
if save_fig:
print('Saving figure...')
plt.savefig(fig_path + '.pdf', transparent=True, bbox_inches='tight')
plt.show()
def wrf_era5_plot(var, wrfdat, eradat, datestr, src='wrf', hourly=False, save_fig=False,
wrf_dir='./', era_dir='./', short_title_str='Title', fig_path='./'):
"""
Creates a single WRF or ERA5 plot, using the WRF bounds, producing either a plot every hour
or a single plot for the day.
"""
# Format the var input
if type(var) is not str:
print(f'The var input, {var}, must be a string.')
raise TypeError
if var in ['GHI', 'ghi']:
var = 'ghi'
wrf_var = 'ghi'
era_var = 'GHI'
elif var in ['WPD', 'wpd']:
var = 'wpd'
wrf_var = 'wpd'
era_var = 'WPD'
elif var in ['ghi_error', 'GHI_ERROR']:
var = 'ghi_error'
wrf_var = var
elif var in ['wpd_error', 'WPD_ERROR']:
var = 'wpd_error'
wrf_var = var
elif var in ['fitness', 'Fitness', 'FITNESS']:
var = 'fitness'
wrf_var = var
else:
print(f'Variable {var} is not supported.')
raise KeyError
# To start, we need to get the WRF map projection information (a Lambert Conformal grid),
# and find the domain boundaries in this projection.
# NOTE: this task MUST occurr before we regrid the WRF variables or the coordinates change and become incompatible.
wrf_cartopy_proj = get_wrf_proj(wrfdat, 'dni')
proj_bounds = get_domain_boundary(wrfdat, 'ghi', wrf_cartopy_proj)
# # Rename the lat-lon corrdinates to get wrf-python to recognize them
# variables = {'lat': 'XLAT',
# 'lon': 'XLONG'}
# try:
# wrfdat = xr.Dataset.rename(wrfdat, variables)
# except ValueError:
# print(f'Variables {variables} cannot be renamed, '
# f'those on the left are not in this dataset.')
#
# # This makes it easy to get the latitude and longitude coordinates with the wrf-python function below
# lats, lons = wrfpy.latlon_coords(wrfdat['dni'])
#
# # I have to do this tedious string parsing below to get the projection from the processed wrfout file.
# try:
# wrf_proj_params = wrfdat.dni.attrs['projection']
# except AttributeError:
# raise ValueError('Variable does not contain projection information')
# wrf_proj_params = wrf_proj_params.replace('(', ', ')
# wrf_proj_params = wrf_proj_params.replace(')', '')
# wrf_proj_params = wrf_proj_params.split(',')
# wrf_proj = wrf_proj_params[0]
# stand_lon = float(wrf_proj_params[1].split('=')[1])
# moad_cen_lat = float(wrf_proj_params[2].split('=')[1])
# truelat1 = float(wrf_proj_params[3].split('=')[1])
# truelat2 = float(wrf_proj_params[4].split('=')[1])
# pole_lat = float(wrf_proj_params[5].split('=')[1])
# pole_lon = float(wrf_proj_params[6].split('=')[1])
#
# # Fortunately, it still apppears to work.
# if wrf_proj == 'LambertConformal':
# wrf_cartopy_proj = ccrs.LambertConformal(central_longitude=stand_lon,
# central_latitude=moad_cen_lat,
# standard_parallels=[truelat1, truelat2])
# else:
# print('Your WRF projection is not the expected Lambert Conformal.')
# raise ValueError
#
# # I need to manually convert the boundaries of the WRF domain into Plate Carree to set the limits.
# # Get the raw map bounds using a wrf-python utility
# raw_bounds = wrfpy.util.geo_bounds(wrfdat['dni'])
# # Get the projected bounds telling cartopy that the input coordinates are lat/lon (Plate Carree)
# proj_bounds = wrf_cartopy_proj.transform_points(ccrs.PlateCarree(),
# np.array([raw_bounds.bottom_left.lon, raw_bounds.top_right.lon]),
# np.array([raw_bounds.bottom_left.lat, raw_bounds.top_right.lat]))
# We can use a basic Plate Carree projection for ERA5
era5_cartopy_proj = ccrs.PlateCarree()
# Now, get the desired variables
if var in ['ghi_error', 'wpd_error', 'fitness']:
input_year = helper_functions.format_date(datestr).strftime('%Y')
input_month = helper_functions.format_date(datestr).strftime('%m')
wrfdat_proc, eradat_proc = optwrf.regridding.wrf_era5_regrid_xesmf(input_year,
input_month,
wrfdir=wrf_dir,
eradir=era_dir)
# Calculate the error in GHI and WPD
wrfdat_proc = optwrf.regridding.wrf_era5_error(wrfdat_proc, eradat_proc)
print(wrfdat_proc)
# Calculate the fitness
correction_factor = 0.0004218304553577255
daylight_factor = helper_functions.daylight_frac(datestr) # daylight fraction
wrfdat_proc['fitness'] = daylight_factor * wrfdat_proc.ghi_error \
+ correction_factor * wrfdat_proc.wpd_error
# Define the time indicies from the times variable
time_indicies = range(0, len(wrfdat.Time))
# Format the times for title slides
times_strings_f = wrfdat.Time.dt.strftime('%b %d, %Y %H:%M')
# Get the desired variable(s)
for tidx in time_indicies:
timestr = wrfdat.Time[tidx].values
timestr_f = times_strings_f[tidx].values
if hourly:
title_str = f'{short_title_str}\n{timestr_f} (UTC)'
else:
time_string_f = wrfdat.Time[0].dt.strftime('%b %d, %Y')
title_str = f'{short_title_str}\n{time_string_f.values}'
# WRF Variable (divide by 1000 to convert from W to kW or Wh to kWh)
if not hourly and tidx != 0:
if src == 'wrf':
if var in ['ghi', 'wpd']:
plot_var = plot_var + (wrfdat[wrf_var].sel(Time=np.datetime_as_string(timestr)) / 1000)
else:
plot_var = plot_var + wrfdat_proc[wrf_var].sel(Time=np.datetime_as_string(timestr))
elif src == 'era5':
if var in ['ghi', 'wpd']:
plot_var = plot_var + (eradat[era_var].sel(Time=np.datetime_as_string(timestr)) / 1000)
else:
print(f'Variable {var} is not provided by {src}')
raise ValueError
else:
if src == 'wrf':
if var in ['ghi', 'wpd']:
plot_var = wrfdat[wrf_var].sel(Time=np.datetime_as_string(timestr)) / 1000
else:
plot_var = wrfdat_proc[wrf_var].sel(Time=np.datetime_as_string(timestr))
elif src == 'era5':
if var in ['ghi', 'wpd']:
plot_var = eradat[era_var].sel(Time=np.datetime_as_string(timestr)) / 1000
else:
print(f'Variable {var} is not provided by {src}')
raise ValueError
if hourly:
# Create a figure
fig = plt.figure(figsize=(4, 4))
# Set the GeoAxes to the projection used by WRF
ax = fig.add_subplot(1, 1, 1, projection=wrf_cartopy_proj)
# # Get, format, and set the map bounds
#
# # Format the projected bounds so they can be used in the xlim and ylim attributes
# proj_xbounds = [proj_bounds[0, 0], proj_bounds[1, 0]]
# proj_ybounds = [proj_bounds[0, 1], proj_bounds[1, 1]]
# # Finally, set the x and y limits
# ax.set_xlim(proj_xbounds)
# ax.set_ylim(proj_ybounds)
#
# # Download and add the states, coastlines, and lakes
# states = cfeature.NaturalEarthFeature(category="cultural", scale="50m",
# facecolor="none",
# name="admin_1_states_provinces_shp")
#
# # Add features to the maps
# ax.add_feature(states, linewidth=.5, edgecolor="black")
# ax.add_feature(cfeature.LAKES.with_scale('50m'), alpha=0.9)
# ax.add_feature(cfeature.OCEAN.with_scale('50m'))
# Make the countour lines for filled contours for the GHI
if hourly:
if var in ['ghi', 'ghi_error']:
contour_levels = np.linspace(0, 0.75, 22)
elif var in ['wpd', 'wpd_error']:
contour_levels = np.linspace(0, 2500, 22)
elif var == 'fitness':
contour_levels = np.linspace(0, 1.5, 22)
else:
if var in ['ghi', 'ghi_error']:
contour_levels = np.linspace(0, 5, 22)
elif var in ['wpd', 'wpd_error']:
contour_levels = np.linspace(0, np.amax(wrfpy.to_np(plot_var)), 22)
elif var == 'fitness':
contour_levels = np.linspace(0, 10, 22)
# Add the filled contour levels
if var in ['ghi', 'ghi_error']:
color_map = get_cmap("hot_r")
elif var in ['wpd', 'wpd_error']:
color_map = get_cmap("Greens")
elif var == 'fitness':
color_map = get_cmap("Greys")
if src == 'wrf' and var in ['ghi', 'wpd']:
cn = ax.contourf(wrfpy.to_np(wrfdat.lat), wrfpy.to_np(wrfdat.lon), wrfpy.to_np(plot_var),
contour_levels, transform=ccrs.PlateCarree(), cmap=color_map)
else:
cn = ax.contourf(wrfpy.to_np(eradat.longitude), wrfpy.to_np(eradat.latitude), wrfpy.to_np(plot_var),
contour_levels, transform=ccrs.PlateCarree(), cmap=color_map)
# Format the plot
format_cnplot_axis(ax, cn, proj_bounds, title_str=title_str)
# # Add a color bar
# plt.colorbar(cn, ax=ax, shrink=.98)
# # Add the axis title
# ax.set_title(title_str)
# Save the figure(s)
if save_fig:
fig_path_temp = fig_path + str(tidx).zfill(2)
plt.savefig(fig_path_temp + '.png', dpi=300, transparent=True, bbox_inches='tight')
plt.show()
# Create a figure
fig = plt.figure(figsize=(4, 4))
# Set the GeoAxes to the projection used by WRF
ax = fig.add_subplot(1, 1, 1, projection=wrf_cartopy_proj)
# # Get, format, and set the map bounds
#
# # Format the projected bounds so they can be used in the xlim and ylim attributes
# proj_xbounds = [proj_bounds[0, 0], proj_bounds[1, 0]]
# proj_ybounds = [proj_bounds[0, 1], proj_bounds[1, 1]]
# # Finally, set the x and y limits
# ax.set_xlim(proj_xbounds)
# ax.set_ylim(proj_ybounds)
#
# # Download and add the states, coastlines, and lakes
# states = cfeature.NaturalEarthFeature(category="cultural", scale="50m",
# facecolor="none",
# name="admin_1_states_provinces_shp")
#
# # Add features to the maps
# ax.add_feature(states, linewidth=.5, edgecolor="black")
# ax.add_feature(cfeature.LAKES.with_scale('50m'), alpha=0.9)
# ax.add_feature(cfeature.OCEAN.with_scale('50m'))
# Make the countour lines for filled contours for the GHI
if hourly:
if var in ['ghi', 'ghi_error']:
contour_levels = np.linspace(0, 0.75, 22)
elif var in ['wpd', 'wpd_error']:
contour_levels = np.linspace(0, 25000, 22)
elif var == 'fitness':
contour_levels = np.linspace(0, 1.5, 22)
else:
if var in ['ghi', 'ghi_error']:
contour_levels = np.linspace(0, 5, 22)
elif var in ['wpd', 'wpd_error']:
contour_levels = np.linspace(0, np.amax(wrfpy.to_np(plot_var)), 22)
elif var == 'fitness':
contour_levels = np.linspace(0, 10, 22)
# Add the filled contour levels
if var in ['ghi', 'ghi_error']:
color_map = get_cmap("hot_r")
elif var in ['wpd', 'wpd_error']:
color_map = get_cmap("Greens")
elif var == 'fitness':
color_map = get_cmap("Greys")
if src == 'wrf' and var in ['ghi', 'wpd']:
cn = ax.contourf(wrfpy.to_np(wrfdat.lat), wrfpy.to_np(wrfdat.lat), wrfpy.to_np(plot_var),
contour_levels, transform=ccrs.PlateCarree(), cmap=color_map)
else:
cn = ax.contourf(wrfpy.to_np(eradat.longitude), wrfpy.to_np(eradat.latitude), wrfpy.to_np(plot_var),
contour_levels, transform=ccrs.PlateCarree(), cmap=color_map)
# Format the plot
format_cnplot_axis(ax, cn, proj_bounds, title_str=title_str)
# # Add a color bar
# plt.colorbar(cn, ax=ax, shrink=.98)
#
# # Add the axis title
# ax.set_title(title_str)
# Save the figure(s)
if save_fig:
plt.savefig(fig_path + '.pdf', transparent=True, bbox_inches='tight')
plt.show()
total_monthly_wpd_error = []
print('Calculating difference between WRF and ERA5...')
for start_date, end_date in zip(start_dates, end_dates):
wrf_sim = WRFModel(param_ids, start_date, end_date)
wrf_sim.DIR_WRFOUT = wrf_sim.DIR_MET4ENE + 'wrfout/ARW/default_all_2011/%s_' % \
(wrf_sim.forecast_start.strftime('%Y-%m-%d')) + wrf_sim.paramstr + '/'
error = wrf_sim.wrf_era5_diff()
total_monthly_ghi_error.append(error[1])
total_monthly_wpd_error.append(error[2])
sys.stdout.flush()
print('!Done!')
# Calculate the annual daily mean error for both GHI and WPD
print('Calulating the annual means...')
total_annual_ghi_error = sum(total_monthly_ghi_error)
daily_mean_ghi_error = total_annual_ghi_error / 365
total_annual_wpd_error = sum(total_monthly_wpd_error)
daily_mean_wpd_error = total_annual_wpd_error / 365
print(f'The annual daily mean GHI error is: {daily_mean_ghi_error} kW m-2')
print(f'The annual daily mean WPD error is: {daily_mean_wpd_error} kW m-2')
# Calculate the annual mean daylight fraction
daylight_factors = []
for jday in range(1, 366):
daylight_factors.append(hf.daylight_frac(jday))
mean_daylight_factor = sum(daylight_factors) / len(daylight_factors)
# Calculate the correction factor
alpha = (daily_mean_ghi_error / mean_daylight_factor) / daily_mean_wpd_error
print(f'The fitness function correction factor is {alpha}.')
