def run_scaling(args):
'''wrapper function for scaling tests'''
config = hosts[args.host]
vic_exe = VIC(args.vic_exe)
# write timing file header
header = string.Template(table_header)
header_kwargs = get_header_info(args.vic_exe, args.global_param)
header = header.safe_substitute(**header_kwargs)
with open(args.timing, 'w') as f:
f.write(header)
for i, kwargs in enumerate(config.profile):
if config.template:
# run on a cluster of some kind
# start by printing the template
print('-'.ljust(OUT_WIDTH, '-'))
print('{host} template'.format(host=args.host).center(OUT_WIDTH))
print('-'.ljust(OUT_WIDTH, '-'))
print(config.template)
print('-'.ljust(OUT_WIDTH, '-'))
template = string.Template(config.template)
run_string = template.safe_substitute(
vic_exe=args.vic_exe,
vic_global=args.global_param,
timing_table_file=args.timing,
i=i,
mpiexec=config.mpiexec,
**kwargs)
run_file = 'vic_{host}_{i}.sh'.format(host=args.host, i=i)
with open(run_file, 'w') as f:
f.write(run_string)
cmd = '{submit} {run_file}'.format(submit=config.submit,
run_file=run_file)
print(cmd)
if not args.test:
check_call(cmd, shell=True)
if args.clean:
os.remove(run_file)
else:
# run locally
n = kwargs['np']
print('Running {} with {} processors'.format(args.vic_exe, n))
if not args.test:
start = time.time()
vic_exe.run(args.global_param, mpi_proc=int(n))
end = time.time()
diff = end - start
with open(args.timing, 'a') as f:
f.write('%5s | %.2f\n' % (n, diff))
print('See %s for scaling table' % args.timing)
def run_scaling(args):
'''wrapper function for scaling tests'''
config = hosts[args.host]
vic_exe = VIC(args.vic_exe)
# write timing file header
header = string.Template(table_header)
header_kwargs = get_header_info(args.vic_exe, args.global_param)
header = header.safe_substitute(**header_kwargs)
with open(args.timing, 'w') as f:
f.write(header)
for i, kwargs in enumerate(config.profile):
if config.template:
# run on a cluster of some kind
# start by printing the template
print('-'.ljust(OUT_WIDTH, '-'))
print('{host} template'.format(
host=args.host).center(OUT_WIDTH))
print('-'.ljust(OUT_WIDTH, '-'))
print(config.template)
print('-'.ljust(OUT_WIDTH, '-'))
template = string.Template(config.template)
run_string = template.safe_substitute(
vic_exe=args.vic_exe, vic_global=args.global_param,
timing_table_file=args.timing, i=i,
mpiexec=config.mpiexec, **kwargs)
run_file = 'vic_{host}_{i}.sh'.format(host=args.host, i=i)
with open(run_file, 'w') as f:
f.write(run_string)
cmd = '{submit} {run_file}'.format(submit=config.submit,
run_file=run_file)
print(cmd)
if not args.test:
check_call(cmd, shell=True)
if args.clean:
os.remove(run_file)
else:
# run locally
n = kwargs['np']
print('Running {} with {} processors'.format(args.vic_exe, n))
if not args.test:
start = time.time()
vic_exe.run(args.global_param, mpi_proc=int(n))
end = time.time()
diff = end - start
with open(args.timing, 'a') as f:
f.write('%5s | %.2f\n' % (n, diff))
print('See %s for scaling table' % args.timing)
def main(cfg_file, nproc=1):
''' Main function
Parameters
----------
cfg_file: <str>
Input config file
nproc: <int>
Number of processors to use
'''
# ====================================================== #
# Load in config file
# ====================================================== #
cfg = read_configobj(cfg_file)
# ====================================================== #
# Process some cfg variables
# ====================================================== #
start_date = pd.to_datetime(cfg['FORCING']['start_date'])
end_date = pd.to_datetime(cfg['FORCING']['end_date'])
start_year = start_date.year
end_year = end_date.year
ens_list = range(cfg['FORCING']['ens_start'],
cfg['FORCING']['ens_end'] + 1)
# ====================================================== #
# Set up output directories
# ====================================================== #
dirs = setup_output_dirs(cfg['OUTPUT']['out_basedir'],
mkdirs=['forc_orig_nc', 'forc_orig_asc',
'forc_disagg_asc', 'forc_disagg_nc',
'config_files', 'logs_vic'])
# Subdirs for config files for ensemble
subdirs_config = setup_output_dirs(
dirs['config_files'],
mkdirs=['netcdf2vic', 'vic4', 'vic2nc'])
# ====================================================== #
# Load in domain file
# ====================================================== #
ds_domain = xr.open_dataset(cfg['DOMAIN']['domain_nc'])
da_domain = ds_domain[cfg['DOMAIN']['mask_name']]
lat_min = da_domain['lat'].min().values
lat_max = da_domain['lat'].max().values
lon_min = da_domain['lon'].min().values
lon_max = da_domain['lon'].max().values
# ====================================================== #
# Load in and process Newman ensemble forcings (for prec, Tmax and Tmin)
# and orig. Maurer forcing (for wind speed)
# ====================================================== #
# --- Load Maurer forcings --- #
print('Processing Maurer forcings...')
# Loop over each year
list_da_wind = []
for year in range(start_year, end_year+1):
print('Year {}'.format(year))
# --- Load in netCDF file for this year --- #
da_wind = xr.open_dataset(os.path.join(
cfg['FORCING']['maurer_dir'],
'nldas_met_update.obs.daily.wind.{}.nc'.format(year)))['wind']
# --- Mask out the target area --- #
da_wind = da_wind.sel(latitude=slice(lat_min, lat_max),
longitude=slice(lon_min, lon_max))
da_wind = da_wind.where(da_domain.values)
# --- Rename lat and lon --- #
da_wind = da_wind.rename({'latitude': 'lat', 'longitude': 'lon'})
# --- Put in list --- #
list_da_wind.append(da_wind)
# Concat all years together
da_wind_allyears = xr.concat(list_da_wind, dim='time')
# --- Load Newman forcings --- #
print('Processing Newman forcings...')
# If 1 processor, do a regular process
if nproc == 1:
# Loop over each ensemble member
for ens in ens_list:
load_and_process_Newman(ens, cfg, da_domain, lat_min, lat_max,
lon_min, lon_max, start_date, end_date,
dirs, da_wind_allyears)
# If multiple processors, use mp
elif nproc > 1:
# Set up multiprocessing
pool = mp.Pool(processes=nproc)
# Loop over each ensemble member
for ens in ens_list:
pool.apply_async(load_and_process_Newman,
(ens, cfg, da_domain, lat_min, lat_max, lon_min,
lon_max, start_date, end_date, dirs,
da_wind_allyears,))
# Finish multiprocessing
pool.close()
pool.join()
# ====================================================== #
# Convert orig. forcings to ascii format
# ====================================================== #
print('Converting orig. netCDF forcings to VIC ascii...')
# --- Setup subdirs for asc VIC orig. forcings for each ensemble member
# --- #
list_ens = []
for ens in ens_list:
list_ens.append('ens_{}'.format(ens))
subdirs_output = setup_output_dirs(
dirs['forc_orig_asc'],
mkdirs=list_ens)
# --- Prepare netcdf2vic config file --- #
dict_cfg_file = {}
for ens in ens_list:
cfg_file = os.path.join(subdirs_config['netcdf2vic'],
'ens_{}.cfg'.format(ens))
dict_cfg_file[ens] = cfg_file
with open(cfg_file, 'w') as f:
f.write('[options]\n')
f.write('files: forc_orig.{}.nc\n')
f.write('verbose: True\n')
f.write('output_format: ASCII\n')
f.write('out_prefix: forc_orig_\n')
f.write('coord_keys: lon,lat\n')
f.write('var_keys: pr,tasmax,tasmin,wind\n')
f.write('start_year: {}\n'.format(start_year))
f.write('end_year: {}\n'.format(end_year))
f.write('latlon_precision: {}\n'.format(
cfg['OUTPUT']['latlon_precision']))
f.write('\n[paths]\n')
f.write('in_path: {}\n'.format(os.path.join(
dirs['forc_orig_nc'],
'ens_{}'.format(ens))))
f.write('mask_path: {}\n'.format(cfg['DOMAIN']['domain_nc']))
f.write('mask_varname: {}\n'.format(cfg['DOMAIN']['mask_name']))
f.write('ASCIIoutPath: {}\n'.format(
subdirs_output['ens_{}'.format(ens)]))
# --- Run nc_to_vic --- #
# If 1 processor, do a regular process
if nproc == 1:
for ens in ens_list:
nc_to_vic(dict_cfg_file[ens])
# If multiple processors, use mp
elif nproc > 1:
# Set up multiprocessing
pool = mp.Pool(processes=nproc)
# Loop over each ensemble member
for ens in ens_list:
pool.apply_async(nc_to_vic, (dict_cfg_file[ens],))
# Finish multiprocessing
pool.close()
pool.join()
# ====================================================== #
# Run VIC forcing disaggregator
# ====================================================== #
print('Running VIC as a disaggregator...')
# --- Setup subdirs for asc VIC disagg. forcings and VIC log files for
# each ensemble member --- #
list_ens = []
for ens in ens_list:
list_ens.append('ens_{}'.format(ens))
subdirs_output = setup_output_dirs(
dirs['forc_disagg_asc'],
mkdirs=list_ens)
subdirs_logs = setup_output_dirs(
dirs['logs_vic'],
mkdirs=list_ens)
# --- Prepare VIC global file for the disaggregation run --- #
# Load in global file template
with open(cfg['VIC_DISAGG']['global_template'], 'r') as f:
global_param = f.read()
# Create string template
s = string.Template(global_param)
# Loop over each ensemble member
dict_global_file = {}
for ens in ens_list:
# Fill in variables in the template
global_param = s.safe_substitute(
time_step=cfg['VIC_DISAGG']['time_step'],
startyear=start_year,
startmonth=start_date.month,
startday=start_date.day,
endyear=end_year,
endmonth=end_date.month,
endday=end_date.day,
forcing1=os.path.join(dirs['forc_orig_asc'],
'ens_{}'.format(ens),
'forc_orig_'),
grid_decimal=cfg['OUTPUT']['latlon_precision'],
prec='PREC',
tmax='TMAX',
tmin='TMIN',
wind='WIND',
forceyear=start_year,
forcemonth=start_date.month,
forceday=start_date.day,
result_dir=subdirs_output['ens_{}'.format(ens)])
# Write global param file
global_file = os.path.join(subdirs_config['vic4'],
'vic.global.ens_{}.txt'.format(ens))
dict_global_file[ens] = global_file
with open(global_file, mode='w') as f:
for line in global_param:
f.write(line)
# --- Run VIC --- #
# Prepare VIC exe
vic_exe = VIC(cfg['VIC_DISAGG']['vic4_exe'])
# If 1 processor, do a regular process
if nproc == 1:
for ens in ens_list:
vic_exe.run(dict_global_file[ens],
logdir=subdirs_logs['ens_{}'.format(ens)])
# If multiple processors, use mp
elif nproc > 1:
# Set up multiprocessing
pool = mp.Pool(processes=nproc)
# Loop over each ensemble member
for ens in ens_list:
pool.apply_async(run_vic_for_multiprocess,
(vic_exe, dict_global_file[ens],
subdirs_logs['ens_{}'.format(ens)],))
# Finish multiprocessing
pool.close()
pool.join()
# ====================================================== #
# Convert disaggregated forcings to netCDF format
# ====================================================== #
# --- Prepare config file for vic2nc --- #
print('Converting disaggregated forcings to netCDF...')
# --- Setup subdirs for VIC disagg. netCDF forcings for each ensemble
# member --- #
list_ens = []
for ens in ens_list:
list_ens.append('ens_{}'.format(ens))
subdirs_output = setup_output_dirs(
dirs['forc_disagg_nc'],
mkdirs=list_ens)
# --- Prepare netcdf2vic config file --- #
# Extract disaggregated forcing variable names and order
with open(cfg['VIC_DISAGG']['global_template'], 'r') as f:
global_param = f.read()
outvar_list = find_outvar_global_param(global_param)
for i, var in enumerate(outvar_list):
outvar_list[i] = var.strip('OUT_')
# Extract end date and hour
end_date_with_hour = end_date + pd.DateOffset(days=1) -\
pd.DateOffset(hours=cfg['VIC_DISAGG']['time_step'])
# Loop over each ensemble member
dict_cfg_file = {}
for ens in ens_list:
cfg_file = os.path.join(subdirs_config['vic2nc'],
'ens_{}.cfg'.format(ens))
dict_cfg_file[ens] = cfg_file
with open(cfg_file, 'w') as f:
f.write('[OPTIONS]\n')
f.write('input_files: {}\n'.format(
os.path.join(dirs['forc_disagg_asc'],
'ens_{}'.format(ens),
'force_*')))
f.write('input_file_format: ascii\n')
f.write('bin_dt_sec: {}\n'.format(cfg['VIC_DISAGG']['time_step']*3600))
f.write('bin_start_date: {}\n'.format(start_date.strftime("%Y-%m-%d-%H")))
f.write('bin_end_date: {}\n'.format(end_date_with_hour.strftime("%Y-%m-%d-%H")))
f.write('regular_grid: False\n')
f.write('out_directory: {}\n'.format(subdirs_output['ens_{}'.format(ens)]))
f.write('memory_mode: big_memory\n')
f.write('chunksize: 100\n')
f.write('out_file_prefix: force\n')
f.write('out_file_format: NETCDF4\n')
f.write('precision: single\n')
f.write('start_date: {}\n'.format(start_date.strftime("%Y-%m-%d-%H")))
f.write('end_date: {}\n'.format(end_date_with_hour.strftime("%Y-%m-%d-%H")))
f.write('calendar: proleptic_gregorian\n')
f.write('time_segment: year\n')
f.write('snow_bands: False\n')
f.write('veg_tiles: False\n')
f.write('soil_layers: False\n')
f.write('\n[DOMAIN]\n')
f.write('filename: {}\n'.format(cfg['DOMAIN']['domain_nc']))
f.write('longitude_var: {}\n'.format(cfg['DOMAIN']['lon_name']))
f.write('latitude_var: {}\n'.format(cfg['DOMAIN']['lat_name']))
f.write('y_x_dims: {}, {}\n'.format(cfg['DOMAIN']['lat_name'],
cfg['DOMAIN']['lon_name']))
f.write('copy_vars: {}, {}, {}\n'.format(cfg['DOMAIN']['mask_name'],
cfg['DOMAIN']['lat_name'],
cfg['DOMAIN']['lon_name']))
f.write('\n[GLOBAL_ATTRIBUTES]\n')
f.write('title: VIC forcings\n')
f.write('version: VIC4.2\n')
f.write('grid: 1/8\n')
for i, var in enumerate(outvar_list):
if var == 'AIR_TEMP':
f.write('\n[AIR_TEMP]\n')
f.write('column: {}\n'.format(i))
f.write('units: C\n')
f.write('standard_name: air_temperature\n')
f.write('description: air temperature\n')
elif var == 'PREC':
f.write('\n[PREC]\n')
f.write('column: {}\n'.format(i))
f.write('units: mm/step\n')
f.write('standard_name: precipitation\n')
f.write('description: precipitation\n')
elif var == 'PRESSURE':
f.write('\n[PRESSURE]\n')
f.write('column: {}\n'.format(i))
f.write('units: kPa\n')
f.write('standard_name: surface_air_pressure\n')
f.write('description: near-surface atmospheric pressure\n')
elif var == 'SHORTWAVE':
f.write('\n[SHORTWAVE]\n')
f.write('column: {}\n'.format(i))
f.write('units: W m-2\n')
f.write('standard_name: incoming_shortwave_radiation\n')
f.write('description: incoming shortwave radiation\n')
elif var == 'LONGWAVE':
f.write('\n[LONGWAVE]\n')
f.write('column: {}\n'.format(i))
f.write('units: W m-2\n')
f.write('standard_name: incoming_longwave_radiation\n')
f.write('description: incoming longwave radiation\n')
elif var == 'VP':
f.write('\n[VP]\n')
f.write('column: {}\n'.format(i))
f.write('units: kPa\n')
f.write('standard_name: water_vapor_pressure\n')
f.write('description: near surface vapor pressure\n')
elif var == 'WIND':
f.write('\n[WIND]\n')
f.write('column: {}\n'.format(i))
f.write('units: m/s\n')
f.write('standard_name: surface_air_pressure\n')
f.write('description: near-surface wind speed\n')
# --- Run vic2nc --- #
# If 1 processor, do a regular process
if nproc == 1:
for ens in ens_list:
cfg_vic2nc = read_config(dict_cfg_file[ens])
options = cfg_vic2nc.pop('OPTIONS')
global_atts = cfg_vic2nc.pop('GLOBAL_ATTRIBUTES')
if not options['regular_grid']:
domain_dict = cfg_vic2nc.pop('DOMAIN')
else:
domain_dict = None
# Set aside fields dict
fields = cfg_vic2nc
# Run vic2nc
vic2nc(options, global_atts, domain_dict, fields)
# If multiple processors, use mp
elif nproc > 1:
# Set up multiprocessing
pool = mp.Pool(processes=nproc)
# Loop over each ensemble member
for ens in ens_list:
cfg_vic2nc = read_config(dict_cfg_file[ens])
options = cfg_vic2nc.pop('OPTIONS')
global_atts = cfg_vic2nc.pop('GLOBAL_ATTRIBUTES')
if not options['regular_grid']:
domain_dict = cfg_vic2nc.pop('DOMAIN')
else:
domain_dict = None
# set aside fields dict
fields = cfg_vic2nc
pool.apply_async(vic2nc,
(options, global_atts, domain_dict, fields,))
# Finish multiprocessing
pool.close()
pool.join()
# Load config file
# =========================================================== #
cfg = read_configobj(sys.argv[1])
# =========================================================== #
# Set random generation seed
# =========================================================== #
np.random.seed(cfg['CONTROL']['seed'])
# =========================================================== #
# Run VIC
# =========================================================== #
# Create class VIC
vic_exe = VIC(cfg['VIC']['exe'])
# Run VIC
vic_exe.run(cfg['VIC']['global'],
logdir=cfg['OUTPUT']['vic_log_dir'])
# =========================================================== #
# Extract VIC output soil moisture (layer 1) at the end of
# each day, and perturb
# =========================================================== #
# Load VIC output
ds = xr.open_dataset(cfg['OUTPUT']['vic_output_hist_path'])
# Resample surface sm to daily mean
da_sm1_true = ds['OUT_SOIL_MOIST'].sel(nlayer=0)
da_sm1_true_daily = da_sm1_true.resample(dim='time', freq='D', how='mean')
# Reset time index to noon on each day
da_sm1_true_daily['time'] = pd.date_range(
'{}-12'.format(cfg['TIME_INDEX']['start_date']),