def _aggregate_state(self):
"""Aggregate data out of the state file and load it into `met_data`"""
# Precipitation record
trailing = self.state['prec']
begin_record = self.params['start'] - pd.Timedelta("90 days")
end_record = self.params['start'] - pd.Timedelta("1 days")
record_dates = date_range(begin_record, end_record,
calendar=self.params['calendar'])
trailing['time'] = record_dates
total_precip = xr.concat([trailing, self.met_data['prec']], dim='time')
total_precip = total_precip.rolling(time=90).mean().drop(record_dates,
dim='time')
self.met_data['seasonal_prec'] = total_precip
# Smoothed daily temperature range
trailing = self.state['t_max'] - self.state['t_min']
begin_record = self.params['start'] - pd.Timedelta("90 days")
end_record = self.params['start'] - pd.Timedelta("1 days")
record_dates = date_range(begin_record, end_record,
calendar=self.params['calendar'])
trailing['time'] = record_dates
dtr = self.met_data['t_max'] - self.met_data['t_min']
sm_dtr = xr.concat([trailing, dtr], dim='time')
sm_dtr = sm_dtr.rolling(time=30).mean().drop(record_dates, dim='time')
self.met_data['smoothed_dtr'] = sm_dtr
# Put in SWE data
self.met_data['swe'] = xr.Variable(('lat', 'lon'),
self.state['swe'].values)
def _aggregate_state(self):
"""Aggregate data out of the state file and load it into `met_data`"""
# Precipitation record
assert self.state.dims['time'] == 90, self.state['time']
record_dates = date_range(self.params['state_start'],
self.params['state_stop'],
calendar=self.params['calendar'])
trailing = self.state['prec']
trailing['time'] = record_dates
total_precip = xr.concat([trailing, self.met_data['prec']],
dim='time').load()
total_precip = (
cnst.DAYS_PER_YEAR * total_precip.rolling(time=90).mean().sel(
time=slice(self.params['start'], self.params['stop'])))
self.met_data['seasonal_prec'] = total_precip
# Smoothed daily temperature range
trailing = self.state['t_max'] - self.state['t_min']
trailing['time'] = record_dates
dtr = self.met_data['t_max'] - self.met_data['t_min']
if (dtr < 0).any():
raise ValueError("Daily maximum temperature lower"
" than daily minimum temperature!")
sm_dtr = xr.concat([trailing, dtr], dim='time').load()
sm_dtr = sm_dtr.rolling(time=30).mean().drop(record_dates, dim='time')
self.met_data['dtr'] = dtr
self.met_data['smoothed_dtr'] = sm_dtr
def setup_output(self, prototype: xr.Dataset=None):
if not prototype:
prototype = self.met_data
self.output = self.domain.copy(deep=True)
self.output.attrs = attrs['_global']
# Number of timesteps
if self.disagg:
delta = pd.Timedelta('1 days') - pd.Timedelta(
'{} minutes'.format(self.params['time_step']))
else:
delta = pd.Timedelta('0 days')
start = pd.Timestamp(prototype.time.values[0]).to_datetime()
stop = pd.Timestamp(prototype.time.values[-1]).to_datetime()
times = date_range(start, stop + delta,
freq="{}T".format(MetSim.params['time_step']),
calendar=self.params['calendar'])
n_ts = len(times)
shape = (n_ts, ) + self.domain_shape
dims = ('time', ) + self.domain_dims
coords = {'time': times, **self.domain.mask.coords}
for varname in MetSim.params['out_vars']:
self.output[varname] = xr.DataArray(
data=np.full(shape, np.nan),
coords=coords, dims=dims,
name=varname, attrs=attrs.get(varname, {}),
encoding={'dtype': 'f8', '_FillValue': cnst.FILL_VALUES['f8']})
def read_ascii(self, fpath: str) -> xr.Dataset:
"""Read in an ascii forcing file"""
dates = date_range(MetSim.params['start'], MetSim.params['stop'],
calendar=self.params['calendar'])
names = MetSim.params['in_vars'].keys()
ds = pd.read_table(fpath, header=None, delim_whitespace=True,
names=names).head(len(dates))
ds.index = dates
return ds
def read_ascii(data_handle, domain=None, is_worker=False,
start=None, stop=None, calendar='standard',
var_dict=None) -> xr.Dataset:
"""Read in an ascii forcing file"""
dates = date_range(start, stop, calendar=calendar)
names = var_dict.keys()
ds = pd.read_csv(data_handle, header=None, delim_whitespace=True,
names=names).head(len(dates))
ds.index = dates
return ds
def read_binary(data_handle,
domain=None,
iter_dims=['lat', 'lon'],
start=None,
stop=None,
calendar='standard',
var_dict=None) -> xr.Dataset:
"""Reads a binary forcing file (VIC 4 format)"""
dates = date_range(start, stop, calendar=calendar)
n_days = len(dates)
type_lookup = {'signed': 'h', 'unsigned': 'H'}
# Pack these for nicer syntax in the loop
var_names = var_dict.keys()
data_list = [[] for var in var_dict.keys()]
n_vars = len(var_names)
params = [p for p in var_dict.values()]
scales = [float(s.split()[0]) for s in params]
datatypes = [type_lookup[s.split()[-1]] for s in params]
with open(data_handle, 'rb') as f:
i = 0
points_read = 0
points_needed = 4 * n_days
while points_read != points_needed:
bytes = f.read(2)
if bytes:
# Get correct variable and data type with i,
# then unpack & scale
data_list[i].append(
struct.unpack(datatypes[i], bytes)[0] / scales[i])
i = (i + 1) % n_vars
points_read += 1
else:
break
# Binary forcing files have naming format $NAME_$LAT_$LON
param_list = os.path.basename(data_handle).split("_")[-3:]
params = {
"name": param_list[0],
"lat": float(param_list[1]),
"lon": float(param_list[2]),
"n_days": int(n_days)
}
# Assemble the dataset
data_dict = {c[0]: (['time'], c[1]) for c in zip(var_names, data_list)}
data_dict['day_of_year'] = (['time'], dates.dayofyear)
df = xr.Dataset(data_dict,
coords={
'lon': [params['lon']],
'lat': [params['lat']],
'time': dates
},
attrs={'n_days': params['n_days']})
return df
def setup_output(self, prototype: xr.Dataset = None):
if not prototype.variables:
prototype = self.met_data
self.disagg = int(self.params['time_step']) < cnst.MIN_PER_DAY
# Number of timesteps
if self.disagg:
delta = pd.Timedelta('1 days') - pd.Timedelta('{} minutes'.format(
self.params['time_step']))
else:
delta = pd.Timedelta('0 days')
start = pd.Timestamp(prototype['time'].values[0]).to_pydatetime()
stop = pd.Timestamp(prototype['time'].values[-1]).to_pydatetime()
times = date_range(start,
stop + delta,
freq="{}T".format(self.params['time_step']),
calendar=self.params['calendar'])
n_ts = len(times)
shape = (n_ts, ) + self.domain['mask'].shape
dims = ('time', ) + self.domain['mask'].dims
coords = {'time': times, **self.domain['mask'].coords}
self.output = xr.Dataset(coords=coords)
self.output['time'].encoding['calendar'] = self.params['calendar']
for p in ['elev', 'lat', 'lon']:
if p in self.params:
self.params.pop(p)
for k, v in self.params.items():
# Need to convert some parameters to strings
if k in ['start', 'stop', 'time_grouper', 'utc_offset']:
v = str(v)
elif k in ['state_start', 'state_stop']:
# skip
continue
# Don't include complex types
if isinstance(v, dict):
v = json.dumps(v)
elif not isinstance(v, str) and isinstance(v, Iterable):
v = ', '.join(v)
attrs['_global'][k] = v
self.output.attrs = attrs['_global']
dtype = self.params['out_precision']
for varname in self.params['out_vars']:
self.output[varname] = xr.DataArray(data=np.full(shape,
np.nan,
dtype=dtype),
coords=coords,
dims=dims,
name=varname,
attrs=attrs.get(varname, {}))
self.output['time'].attrs.update(attrs['time'])
def __init__(self, params: dict):
"""
Constructor
"""
# Record parameters
MetSim.params.update(params)
MetSim.params['dates'] = date_range(params['start'], params['stop'],
calendar=self.params['calendar'])
logger.setLevel(MetSim.params['verbose'])
ch.setLevel(MetSim.params['verbose'])
logger.addHandler(ch)
self.output = None
self.met_data = None
self.ready = False
def process_vic(params: dict, domain: xr.Dataset) -> xr.Dataset:
"""Process VIC-like data"""
read_funcs = {
"binary": read_binary,
"ascii": read_ascii,
}
if 'lon' not in params['iter_dims'] or 'lat' not in params['iter_dims']:
raise ValueError('Using VIC type input requires lat and lon to be'
' specified via `iter_dims` in configuration.')
# Creates the master dataset which will be used to parallelize
dates = date_range(params['start'],
params['stop'],
calendar=params['calendar'])
coords = {'time': dates, 'lon': domain['lon'], 'lat': domain['lat']}
shape = (len(dates), len(domain['lat']), len(domain['lon']))
dims = (
'time',
'lat',
'lon',
)
met_data = xr.Dataset(coords=coords, attrs={'n_days': len(dates)})
for var in params['forcing_vars']:
met_data[var] = xr.DataArray(data=np.full(shape, np.nan),
coords=coords,
dims=dims,
name=var)
# Fill in the data
for job in params['forcing']:
try:
_, lat, lon = os.path.basename(job).split("_")[-3:]
lat, lon = float(lat), float(lon)
if not domain['mask'].sel(lat=lat, lon=lon).values > 0:
continue
ds = read_funcs[params['forcing_fmt']](
job,
start=params['start'],
stop=params['stop'],
calendar=params['calendar'],
var_dict=params['forcing_vars'])
for var in params['forcing_vars'].keys():
met_data[var].loc[{'lat': lat, 'lon': lon}] = ds[var]
except (ValueError, KeyError):
continue
return met_data
def read_binary(self, fpath: str) -> xr.Dataset:
""" Reads a binary forcing file (VIC 4 format) """
dates = date_range(MetSim.params['start'], MetSim.params['stop'],
calendar=self.params['calendar'])
n_days = len(dates)
type_lookup = {'signed': 'h', 'unsigned': 'H'}
# Pack these for nicer syntax in the loop
var_names = MetSim.params['in_vars'].keys()
data_list = [[] for var in MetSim.params['in_vars'].keys()]
n_vars = len(var_names)
params = [p for p in MetSim.params['in_vars'].values()]
scales = [float(s.split()[0]) for s in params]
datatypes = [type_lookup[s.split()[-1]] for s in params]
with open(fpath, 'rb') as f:
i = 0
points_read = 0
points_needed = 4*n_days
while points_read != points_needed:
bytes = f.read(2)
if bytes:
# Get correct variable and data type with i,
# then unpack & scale
data_list[i].append(
struct.unpack(datatypes[i], bytes)[0] / scales[i])
i = (i + 1) % n_vars
points_read += 1
else:
break
# Binary forcing files have naming format $NAME_$LAT_$LON
param_list = os.path.basename(fpath).split("_")
params = {"name": param_list[0],
"lat": float(param_list[1]),
"lon": float(param_list[2]),
"n_days": int(n_days)}
MetSim.params.update(params)
params['elev'] = [[self.find_elevation(params['lat'], params['lon'])]]
# Assemble the dataset
data_dict = {c[0]: (['time'], c[1]) for c in zip(var_names, data_list)}
data_dict['elev'] = (['lon', 'lat'], params['elev'])
data_dict['day_of_year'] = (['time'], dates.dayofyear)
df = xr.Dataset(data_dict,
coords={'lon': [params['lon']],
'lat': [params['lat']],
'time': dates},
attrs={'n_days': params['n_days']})
return df
def _unpack_state(self, result: pd.DataFrame, locs: dict):
"""Put restart values in the state dataset"""
# We concatenate with the old state values in case we don't
# have 90 new days to use
tmin = np.concatenate((self.state['t_min'].isel(**locs).values[:],
result['t_min'].values))
tmax = np.concatenate((self.state['t_max'].isel(**locs).values[:],
result['t_max'].values))
prec = np.concatenate(
(self.state['prec'].isel(**locs).values[:], result['prec'].values))
self.state['t_min'].isel(**locs).values[:] = tmin[-90:]
self.state['t_max'].isel(**locs).values[:] = tmax[-90:]
self.state['prec'].isel(**locs).values[:] = prec[-90:]
state_start = result.index[-1] - pd.Timedelta('89 days')
self.state['time'].values = date_range(
state_start, result.index[-1], calendar=self.params['calendar'])
def _get_output_times(self, freq=None, period_ending=False):
"""
Generate chunked time vectors
Parameters
----------
freq:
Output frequency. Given as a Pandas timegrouper string.
If not given, the entire timeseries will be used.
period_ending:
Flag to specify if output timesteps should be period-
ending. Default is period-beginning
Returns
-------
times:
A list of timeseries which represent each of times that
output files will be created for.
"""
prototype = self.met_data
self.disagg = int(self.params['time_step']) < cnst.MIN_PER_DAY
if self.disagg:
delta = pd.Timedelta('1 days') - pd.Timedelta('{} minutes'.format(
self.params['time_step']))
else:
delta = pd.Timedelta('0 days')
if period_ending:
offset = pd.Timedelta('{} minutes'.format(
self.params['time_step']))
else:
offset = pd.Timedelta('0 minutes')
start = pd.Timestamp(prototype['time'].values[0]).to_pydatetime()
stop = pd.Timestamp(prototype['time'].values[-1]).to_pydatetime()
times = date_range(start + offset,
stop + offset + delta,
freq="{}T".format(self.params['time_step']),
calendar=self.params['calendar'])
if freq is None or freq == '':
times = [times]
else:
dummy = pd.Series(np.arange(len(times)), index=times)
grouper = pd.Grouper(freq=freq)
times = [t.index for k, t in dummy.groupby(grouper)]
return times
def disaggregate(df_daily: pd.DataFrame,
params: dict,
solar_geom: dict,
t_begin: list = None,
t_end: list = None) -> pd.DataFrame:
"""
Take a daily timeseries and scale it down to a finer
time scale.
Parameters
----------
df_daily:
Dataframe containing daily timeseries.
Should be the result of one of the methods
provided in the `methods` directory.
params:
A dictionary containing the class parameters
of the MetSim object.
solar_geom:
A dictionary of solar geometry variables
t_begin:
List of t_min and t_max for day previous to the
start of `df_daily`. None indicates no extension
of the record.
t_end:
List of t_min and t_max for day after the end
of `df_daily`. None indicates no extension of
the record.
Returns
-------
df_disagg:
A dataframe with sub-daily timeseries.
"""
stop = (df_daily.index[-1] + pd.Timedelta('1 days') -
pd.Timedelta("{} minutes".format(params['time_step'])))
dates_disagg = date_range(df_daily.index[0],
stop,
freq='{}T'.format(params['time_step']),
calendar=params['calendar'])
df_disagg = pd.DataFrame(index=dates_disagg)
n_days = len(df_daily)
n_disagg = len(df_disagg)
ts = int(params['time_step'])
df_disagg['shortwave'] = shortwave(df_daily['shortwave'].values,
df_daily['daylength'].values,
df_daily.index.dayofyear,
solar_geom['tiny_rad_fract'], params)
t_Tmin, t_Tmax = set_min_max_hour(solar_geom['tiny_rad_fract'],
df_daily.index.dayofyear - 1,
solar_geom['daylength'], n_days, ts,
params)
df_disagg['temp'] = temp(df_daily['t_min'].values,
df_daily['t_max'].values, n_disagg, t_Tmin,
t_Tmax, ts, t_begin, t_end)
df_disagg['vapor_pressure'] = vapor_pressure(
df_daily['vapor_pressure'].values, df_disagg['temp'].values, t_Tmin,
n_disagg, ts)
df_disagg['vapor_pressure'] = (df_disagg['vapor_pressure'].fillna(
method='ffill').fillna(method='bfill'))
df_disagg['rel_humid'] = relative_humidity(
df_disagg['vapor_pressure'].values, df_disagg['temp'].values)
df_disagg['air_pressure'] = pressure(df_disagg['temp'].values,
params['elev'], params['lapse_rate'])
df_disagg['spec_humid'] = specific_humidity(
df_disagg['vapor_pressure'].values, df_disagg['air_pressure'].values)
df_disagg['tskc'] = tskc(df_daily['tskc'].values, ts, params)
if 'longwave' in df_daily:
daily_lw = df_daily['longwave']
else:
daily_lw = None
df_disagg['longwave'] = longwave(df_disagg['temp'].values,
df_disagg['vapor_pressure'].values,
df_disagg['tskc'].values, params,
daily_lw)
df_disagg['prec'] = prec(df_daily['prec'], df_daily['t_min'], ts, params,
df_daily.get('t_pk'), df_daily.get('dur'))
if 'wind' in df_daily:
df_disagg['wind'] = wind(df_daily['wind'].values, ts, params)
if params['period_ending']:
df_disagg.index += pd.Timedelta('{} minutes'.format(
params['time_step']))
return df_disagg.fillna(method='ffill').fillna(method='bfill')
def wrap_run_cell(func: callable, params: dict, ds: xr.Dataset,
state: xr.Dataset, disagg: bool,
out_times: pd.DatetimeIndex):
"""
Iterate over a chunk of the domain. This is wrapped
so we can return a tuple of locs and df.
Parameters
----------
func: callable
The function to call to do the work
params: dict
Parameters from a MetSim object
ds: xr.Dataset
Input forcings and domain
state: xr.Dataset
State variables at the point of interest
disagg: bool
Whether or not we should run a disagg routine
out_times: pd.DatetimeIndex
Times to return (should be trimmed 1 day at
each end from the given index)
Returns
-------
df_complete: pd.DataFrame
A dataframe with the disaggregated data in it
df_base: pd.DataFrame
A dataframe with the state data in it
"""
lat = ds['lat'].values.flatten()[0]
lon = ds['lon'].values.flatten()[0]
elev = ds['elev'].values.flatten()[0]
params['elev'] = elev
params['lat'] = lat
params['lon'] = lon
df = ds.to_dataframe()
# solar_geom returns a tuple due to restrictions of numba
# for clarity we convert it to a dictionary here
sg = solar_geom(elev, lat, params['lapse_rate'])
sg = {
'tiny_rad_fract': sg[0],
'daylength': sg[1],
'potrad': sg[2],
'tt_max0': sg[3]
}
yday = df.index.dayofyear - 1
df['daylength'] = sg['daylength'][yday]
df['potrad'] = sg['potrad'][yday]
df['tt_max'] = sg['tt_max0'][yday]
# Generate the daily values - these are saved
# so that we can use a subset of them to write
# out the state file later
df_base = func(df, params)
if disagg:
# Get some values for padding the time list,
# so that when interpolating in the disaggregation
# functions we can match endpoints with adjoining
# chunks - if no data is available, just repeat some
# default values (this case is used at the very
# beginning and end of the record)
t_begin = [state['t_min'].values[-1], state['t_max'].values[-1]]
try:
nextday = out_times[-1] + pd.Timedelta('1 days')
t_end = [
ds['t_min'].sel(time=nextday), ds['t_max'].sel(time=nextday)
]
except (KeyError, ValueError):
# None so that we don't extend the record
t_end = None
# Disaggregate to subdaily values
df_complete = disaggregate(df, params, sg, t_begin, t_end)
# Calculate the times that we want to get out by chopping
# off the endpoints that were added on previously
start = out_times.values[0]
stop = (out_times.values[-1] + pd.Timedelta('1 days') -
pd.Timedelta("{} minutes".format(params['time_step'])))
if params['period_ending']:
start += pd.Timedelta('{} minutes'.format(params['time_step']))
stop += pd.Timedelta('{} minutes'.format(params['time_step']))
new_times = date_range(start,
stop,
freq='{}T'.format(params['time_step']),
calendar=params['calendar'])
else:
# convert srad to daily average flux from daytime flux
df_base['shortwave'] *= df_base['daylength'] / cnst.SEC_PER_DAY
# If we're outputting daily values, we dont' need to
# change the output dates - see inside of `if` condition
# above for more explanation
new_times = out_times
df_complete = df_base
# Cut the returned data down to the correct time index
df_complete = df_complete.loc[new_times[0]:new_times[-1]]
return df_complete, df_base
def wrap_run(func: callable, loc: dict, params: dict,
ds: xr.Dataset, state: xr.Dataset, disagg: bool,
out_times: pd.DatetimeIndex, year: str):
"""
Iterate over a chunk of the domain. This is wrapped
so we can return a tuple of locs and df.
Parameters
----------
func: callable
The function to call to do the work
loc: dict
Some subset of the domain to do work on
params: dict
Parameters from a MetSim object
ds: xr.Dataset
Input forcings and domain
state: xr.Dataset
State variables at the point of interest
disagg: bool
Whether or not we should run a disagg routine
out_times: pd.DatetimeIndex
Times to return (should be trimmed 1 day at
each end from the given index)
year: str
The year being run. This is used to add on
extra times to make output smooth at endpoints
if the run is chunked in time.
Returns
-------
results
A list of tuples arranged as
(location, hourly_output, daily_output)
"""
logger.info("Processing {}".format(loc))
lat = ds['lat'].values
elev = ds['elev'].values
swe = ds['swe'].values
df = ds.drop(['lat', 'lon', 'elev', 'swe']).to_dataframe()
# solar_geom returns a tuple due to restrictions of numba
# for clarity we convert it to a dictionary here
sg = solar_geom(elev, lat)
sg = {'tiny_rad_fract': sg[0], 'daylength': sg[1],
'potrad': sg[2], 'tt_max0': sg[3]}
# Generate the daily values - these are saved
# so that we can use a subset of them to write
# out the state file later
df_base = func(df, params, sg, elev=elev, swe=swe)
if disagg:
# Get some values for padding the time list,
# so that when interpolating in the disaggregation
# functions we can match endpoints with adjoining
# chunks - if no data is available, just repeat some
# default values (this case is used at the very
# beginning and end of the record)
try:
prevday = out_times[0] - pd.Timedelta('1 days')
t_begin = [ds['t_min'].sel(time=prevday),
ds['t_max'].sel(time=prevday)]
except (KeyError, ValueError):
t_begin = [state['t_min'].values[-1],
state['t_max'].values[-1]]
try:
nextday = pd.datetime(int(year)+1, 1, 1)
t_end = [ds['t_min'].sel(time=nextday),
ds['t_max'].sel(time=nextday)]
except (KeyError, ValueError):
# None so that we don't extend the record
t_end = None
# Disaggregate to subdaily values
df_complete = disaggregate(df, params, sg, t_begin, t_end)
# Calculate the times that we want to get out by chopping
# off the endpoints that were added on previously
start = out_times[0]
stop = out_times[-1] + pd.Timedelta('23 hours')
new_times = date_range(
start, stop, freq='{}T'.format(params['time_step']),
calendar=params['calendar'])
else:
# convert srad to daily average flux from daytime flux
df_base['swrad'] *= df_base['dayl'] / cnst.SEC_PER_DAY
# If we're outputting daily values, we dont' need to
# change the output dates - see inside of `if` condition
# above for more explanation
new_times = out_times
# Cut the returned data down to the correct time index
df_complete = df_complete.loc[new_times[0]:new_times[-1]]
df_base = df_base.loc[new_times[0]:new_times[-1]]
return (loc, df_complete, df_base)
def run(self):
"""
Kicks off the disaggregation and queues up data for IO
"""
time_dim = pd.to_datetime(self.met_data.time.values)
if self.params['annual']:
groups = time_dim.groupby(time_dim.year)
else:
groups = {'total': time_dim}
for label, times in groups.items():
logger.info("Beginning {}".format(label))
# Add in some end point data for continuity
times_ext = times.union([times[0] - pd.Timedelta("1 days"),
times[-1] + pd.Timedelta("1 days")]
).intersection(time_dim)
data = self.met_data.sel(time=times_ext)
self.setup_output(self.met_data.sel(time=times))
for i, j in self.locations:
locs = dict(lat=i, lon=j)
logger.info("Processing {}".format(locs))
ds = data.isel(**locs)
lat = ds['lat'].values
elev = ds['elev'].values
swe = ds['swe'].values
df = ds.drop(['lat', 'lon', 'elev', 'swe']).to_dataframe()
# solar_geom returns a tuple due to restrictions of numba
# for clarity we convert it to a dictionary here
sg = solar_geom(elev, lat)
sg = {'tiny_rad_fract': sg[0], 'daylength': sg[1],
'potrad': sg[2], 'tt_max0': sg[3]}
# Generate the daily values - these are saved
# so that we can use a subset of them to write
# out the state file later
df = self.method.run(df, self.params, sg,
elev=elev, swe=swe)
# Get some values for padding the time list,
# so that when interpolating in the disaggregation
# functions we can match endpoints with adjoining
# chunks - if no data is available, just repeat some
# default values (this case is used at the very
# beginning and end of the record)
if self.disagg:
try:
prevday = data.time[0] - pd.Timedelta('1 days')
t_begin = [self.met_data['t_min'].sel(
time=prevday).isel(lat=i, lon=j),
self.met_data['t_max'].sel(
time=prevday).isel(lat=i, lon=j)]
except (KeyError, ValueError):
t_begin = [self.state['t_min'].values[-1, i, j],
self.state['t_max'].values[-1, i, j]]
try:
nextday = pd.datetime(int(label)+1, 1, 1)
t_end = [self.met_data['t_min'].sel(
time=nextday).isel(lat=i, lon=j),
self.met_data['t_max'].sel(
time=nextday).isel(lat=i, lon=j)]
except (KeyError, ValueError):
# None so that we don't extend the record
t_end = None
self._unpack_state(df, locs)
df = disaggregate(df, self.params, sg, t_begin, t_end)
start = times[0]
stop = (times[-1] + pd.Timedelta('1 days')
- pd.Timedelta(self.params['time_step']))
new_times = date_range(
start, stop,
freq='{}T'.format(self.params['time_step']),
calendar=self.params['calendar'])
else:
# convert srad to daily average flux from daytime flux
self._unpack_state(df, locs)
df['swrad'] *= df['dayl'] / cnst.SEC_PER_DAY
# If we're outputting daily values, we dont' need to
# change the output dates - see inside of `if` condition
# above for more explanation
new_times = times
# Cut the returned data down to the correct time index
self._unpack_results((locs, df.loc[new_times[0]:new_times[-1]]))
self.write(label)