def get_NSE(values):
its = values['LZSN'], values['INFILT']
print(('running simulation LZSN: {:.1f}, INFILT: {:.4f}\n'.format(*its)))
# open the baseline HSPF model
with open(model, 'rb') as f:
hspfmodel = pickle.load(f)
# change the filename to prevent two simultaneous runs with the same name
its = working, values['LZSN'], values['INFILT']
hspfmodel.filename = '{}/LZSN{:.1f}INFILT{:.4f}'.format(*its)
# change the pan coefficient
hspfmodel.evap_multiplier = 0.76
# change the parameter values in all the land segments
for p in hspfmodel.perlnds:
p.LZSN = values['LZSN']
p.INFILT = values['INFILT']
# build the input WDM file
hspfmodel.build_wdminfile()
# build the UCI file (only need to save the reach_outvolume, ROVOL)
hspfmodel.build_uci(['reach_outvolume'], start, end, hydrology=True)
# run the simulation
hspfmodel.run(verbose=True)
# make an instance of the postprocessor to get the efficiency
p = Postprocessor(hspfmodel, (start, end), comid=gagecomid)
# the regressions between the simulated and observed flows can be
# computed using the "regression" function that returns:
#
# daily r2, log daily r2, daily NSE, log daily NSE
# montly r2, log monthly r2, monthly NSE, log monthly NSE
dr2, dlr2, dNS, dlNS, mr2, mlr2, mNS, mlNS = p.get_regression(gagecomid)
# close the Postprocessor
p.close()
return dNS
def get_NSE(values):
its = values['LZSN'], values['INFILT']
print('running simulation LZSN: {:.0f}, INFILT: {:.2f}\n'.format(*its))
# open the baseline HSPF model
with open(filename, 'rb') as f:
hspfmodel = pickle.load(f)
# change the filename to prevent two simultaneous runs with the same name
i = len(hspfmodel.filename) - 1
while hspfmodel.filename[i] != '/':
i -= 1
path = hspfmodel.filename[:i]
its = path, values['LZSN'], values['INFILT']
hspfmodel.filename = '{}/LZSN{:.0f}INFILT{:.2f}'.format(*its)
# change the parameter values in all the land segments
for p in hspfmodel.perlnds:
p.LZSN = values['LZSN']
p.INFILT = values['INFILT']
# build the input WDM file
hspfmodel.build_wdminfile()
# build the UCI file (only need to save the reach_outvolume, ROVOL)
hspfmodel.build_uci(['reach_outvolume'],
start,
end,
atemp=atemp,
snow=snow,
hydrology=hydrology)
# run the simulation
hspfmodel.run(verbose=True)
# calibration period
dates = start + datetime.timedelta(days=warmup), end
# find the common identifier for the NWIS gage
d = {v: k for k, v in hspfmodel.subbasin_timeseries['flowgage'].items()}
comid = d[gageid]
# make an instance of the postprocessor to get the efficiency
p = Postprocessor(hspfmodel, dates, comid=comid)
# the regressions between the simulated and observed flows can be
# computed using the "regression" function that returns:
#
# daily r2, log daily r2, daily NSE, log daily NSE
# montly r2, log monthly r2, monthly NSE, log monthly NSE
dr2, dlr2, dNS, dlNS, mr2, mlr2, mNS, mlNS = p.get_regression(comid)
# close the Postprocessor
p.close()
return dNS
# open the postprocessor to get the calibration info p = Postprocessor(hspfmodel, (start, end), comid = gagecomid) # calculate and show the errors in the calibration parameters. the product # of the daily log-flow and daily flow Nash-Sutcliffe model efficiency are # one possible optimization parameter for a calibration. the log-flow # captures relative errors (low-flow conditions) while the flow captures # absolute error (high-flow conditions). p.calculate_errors() # close the open files p.close() # now let's change the value of some parameters, re-run the model, and see # the effect on the calibration statistics. we will change the default # values by perturbing some calibration multiplier parameter by a "factor." # since the initial run has low storm volumes, flow should be shifted from # interflow to surface runoff by decreasing INTFW LZETP_multiplier = 1. # lower zone evapotranspiration parameter LZSN_multiplier = 1. # lower soil zone storage capacity UZSN_multiplier = 1. # upper soil zone storage capacity INTFW_multiplier = 0.8 # interflow inflow rate INFILT_multiplier = 1. # infiltration rate IRC_multiplier = 1. # interflow recession rate evap_multiplier = 0.76 # pan evaporation relative to potential ET AGWRC = 0.95 # groundwater recession rate (site-wide)
postprocessor.get_hspexp_parameters(verbose = False)
postprocessor.plot_hydrograph(tstep = 'monthly', show = False,
output = '{}/hydrography'.format(preliminary))
postprocessor.plot_calibration(output = '{}/statistics'.format(preliminary),
show = False)
output = '{}/calibration_report.csv'.format(preliminary)
postprocessor.calibration_report(output = output)
postprocessor.plot_snow(output = '{}/snow'.format(preliminary),
show = False)
postprocessor.plot_dayofyear(output = '{}/dayofyear'.format(preliminary),
show = False)
# have to close the WDM files to continue
postprocessor.close()
# calibration using the full period of record
# file path to place the calibrated model and results
calibration = '{}/{}/calibration'.format(destination, HUC8)
# path where the calibrated model will be saved/located
calibrated = '{}/{}'.format(calibration, gageid)
# make the directory for the calibration simulations
if not os.path.isdir(calibration): os.mkdir(calibration)
postprocessor.plot_calibration(output = '{}/stats'.format(calibration),
show = False)
postprocessor.plot_runoff(tstep = 'daily', show = False,
output = '{}/runoff'.format(calibration))
output = '{}/calibration_report.csv'.format(calibration)
postprocessor.calibration_report(output = output)
postprocessor.plot_snow(output = '{}/snow'.format(calibration),
show = False)
postprocessor.plot_dayofyear(output ='{}/dayofyear'.format(calibration),
show = False)
postprocessor.plot_storms(season = 'all', show = False,
output = '{}/storms'.format(calibration))
# have to close the WDM files to continue
postprocessor.close()
# run a validation simulation
with open(calibrated, 'rb') as f: hspfmodel = pickle.load(f)
# output variables
targets = ['water_state',
'reach_outvolume',
'evaporation',
'runoff',
'groundwater',
]
# build the input WDM file
# open the postprocessor to get the calibration info p = Postprocessor(hspfmodel, (start, end), comid = gagecomid) # calculate and show the errors in the calibration parameters. the product # of the daily log-flow and daily flow Nash-Sutcliffe model efficiency are # one possible optimization parameter for a calibration. the log-flow # captures relative errors (low-flow conditions) while the flow captures # absolute error (high-flow conditions). p.calculate_errors() # close the open files p.close() # now let's change the value of some parameters, re-run the model, and see # the effect on the calibration statistics. we will change the default # values by perturbing some calibration multiplier parameter by a "factor." # since the initial run has low storm volumes, flow should be shifted from # interflow to surface runoff by decreasing INTFW LZETP_multiplier = 1. # lower zone evapotranspiration parameter LZSN_multiplier = 1. # lower soil zone storage capacity UZSN_multiplier = 1. # upper soil zone storage capacity INTFW_multiplier = 0.8 # interflow inflow rate INFILT_multiplier = 1. # infiltration rate IRC_multiplier = 1. # interflow recession rate evap_multiplier = 0.76 # pan evaporation relative to potential ET AGWRC = 0.95 # groundwater recession rate (site-wide)
def get_NSE(values):
its = values['LZSN'], values['INFILT']
print('running simulation LZSN: {:.0f}, INFILT: {:.2f}\n'.format(*its))
# open the baseline HSPF model
with open(filename, 'rb') as f: hspfmodel = pickle.load(f)
# change the filename to prevent two simultaneous runs with the same name
i = len(hspfmodel.filename) - 1
while hspfmodel.filename[i] != '/': i -= 1
path = hspfmodel.filename[:i]
its = path, values['LZSN'], values['INFILT']
hspfmodel.filename = '{}/LZSN{:.0f}INFILT{:.2f}'.format(*its)
# change the parameter values in all the land segments
for p in hspfmodel.perlnds:
p.LZSN = values['LZSN']
p.INFILT = values['INFILT']
# build the input WDM file
hspfmodel.build_wdminfile()
# build the UCI file (only need to save the reach_outvolume, ROVOL)
hspfmodel.build_uci(['reach_outvolume'], start, end, atemp = atemp,
snow = snow, hydrology = hydrology)
# run the simulation
hspfmodel.run(verbose = True)
# calibration period
dates = start + datetime.timedelta(days = warmup), end
# find the common identifier for the NWIS gage
d = {v:k for k, v in hspfmodel.subbasin_timeseries['flowgage'].items()}
comid = d[gageid]
# make an instance of the postprocessor to get the efficiency
p = Postprocessor(hspfmodel, dates, comid = comid)
# the regressions between the simulated and observed flows can be
# computed using the "regression" function that returns:
#
# daily r2, log daily r2, daily NSE, log daily NSE
# montly r2, log monthly r2, monthly NSE, log monthly NSE
dr2, dlr2, dNS, dlNS, mr2, mlr2, mNS, mlNS = p.get_regression(comid)
# close the Postprocessor
p.close()
return dNS
