def calc_pml_fatalities(fatalities, nu, return_period=None):
"""Compute the probable maximum loss fatalities (PMLf) curve for a standard
probabilistic EQRM risk run.
Inputs:
fatalities Vector containing the fatality estimates in numbers for
each event with the corresponding nu. note that this can be
aggregated fatalities across all sites or it can simply be
the fatalities at a specific site.
nu 1D column array the event activity of each of the simulated
events.
return_period provide single return_period if usng this function
to compute exceedance fatalties for desired return period.
If None a uniform sample of return periods will be created.
Returns:
pmlf_data (nx3) array containing the PMLf curve. The first column
contains the probability of exceedance (in one year) values,
the second column contains the direct financial losses
for each of the probabilities of exceedance and the third
column contains the financial losses as a percentage of the
total building value.
"""
# get total fatalities
# num.sum(saved_fatalities, axis=0) # sum rows
# Define return periods of interest and return *rates*
if return_period is None:
return_periods = num.logspace(1, 6, 25)
else:
return_periods = return_period
rtrn_rte = 1. / return_periods
# only 1 element on left gets first of returned tuple only?
(trghzd_agg, _, _) = calc_annloss.acquire_riskval(fatalities, nu, rtrn_rte)
ProbExceedSmall = 1 - num.exp(-rtrn_rte)
return [ProbExceedSmall, trghzd_agg]
def calc_pml_fatalities(fatalities, nu, return_period=None):
"""Compute the probable maximum loss fatalities (PMLf) curve for a standard
probabilistic EQRM risk run.
Inputs:
fatalities Vector containing the fatality estimates in numbers for
each event with the corresponding nu. note that this can be
aggregated fatalities across all sites or it can simply be
the fatalities at a specific site.
nu 1D column array the event activity of each of the simulated
events.
return_period provide single return_period if usng this function
to compute exceedance fatalties for desired return period.
If None a uniform sample of return periods will be created.
Returns:
pmlf_data (nx3) array containing the PMLf curve. The first column
contains the probability of exceedance (in one year) values,
the second column contains the direct financial losses
for each of the probabilities of exceedance and the third
column contains the financial losses as a percentage of the
total building value.
"""
# get total fatalities
# num.sum(saved_fatalities, axis=0) # sum rows
# Define return periods of interest and return *rates*
if return_period is None:
return_periods = num.logspace(1, 6, 25)
else:
return_periods = return_period
rtrn_rte = 1./return_periods
# only 1 element on left gets first of returned tuple only?
(trghzd_agg, _, _) = calc_annloss.acquire_riskval(fatalities, nu, rtrn_rte)
ProbExceedSmall = 1-num.exp(-rtrn_rte)
return [ProbExceedSmall, trghzd_agg]
def calc_pml(saved_ecloss, saved_ecbval2, nu):
"""Compute the probable maximum loss (PML) curve for a standard
probabilistic EQRM risk run.
Inputs:
saved_ecloss 2D array containing the damage estimates in dollars for
each building, multiplied by the survey factor. Note that
the matrix has one row for each simulated event and one
column for each building.
saved_ecbval2 1D row array containing the value of each building,
multiplied by the survey factor.
nu 1D column array the event activity of each of the simulated
events.
Returns:
pml_data (nx3) array containing the PML curve. The first column
contains the probability of exceedance (in one year) values,
the second column contains the direct financial losses
for each of the probabilities of exceedance and the third
column contains the financial losses as a percentage of the
total building value.
"""
# get total building value
TotalBVal2 = num.sum(saved_ecbval2)
AggEcLoss = num.sum(saved_ecloss, axis=0) # sum rows
# Define return periods of interest and return *rates*
return_periods = num.logspace(1, 6, 25)
rtrn_rte = 1. / return_periods
# only 1 element on left gets first of returned tuple only?
(trghzd_agg, _, _) = calc_annloss.acquire_riskval(AggEcLoss, nu, rtrn_rte)
ProbExceedSmall = 1 - num.exp(-rtrn_rte)
return [ProbExceedSmall, trghzd_agg, trghzd_agg / TotalBVal2 * 100]
def calc_pml(saved_ecloss, saved_ecbval2, nu):
"""Compute the probable maximum loss (PML) curve for a standard
probabilistic EQRM risk run.
Inputs:
saved_ecloss 2D array containing the damage estimates in dollars for
each building, multiplied by the survey factor. Note that
the matrix has one row for each simulated event and one
column for each building.
saved_ecbval2 1D row array containing the value of each building,
multiplied by the survey factor.
nu 1D column array the event activity of each of the simulated
events.
Returns:
pml_data (nx3) array containing the PML curve. The first column
contains the probability of exceedance (in one year) values,
the second column contains the direct financial losses
for each of the probabilities of exceedance and the third
column contains the financial losses as a percentage of the
total building value.
"""
# get total building value
TotalBVal2 = num.sum(saved_ecbval2)
AggEcLoss = num.sum(saved_ecloss, axis=0) # sum rows
# Define return periods of interest and return *rates*
return_periods = num.logspace(1, 6, 25)
rtrn_rte = 1./return_periods
# only 1 element on left gets first of returned tuple only?
(trghzd_agg, _, _) = calc_annloss.acquire_riskval(AggEcLoss, nu, rtrn_rte)
ProbExceedSmall = 1-num.exp(-rtrn_rte)
return [ProbExceedSmall, trghzd_agg, trghzd_agg/TotalBVal2*100]
def calc_annfatalities_deagg_distmag(saved_fatalities, nu, saved_rjb,
aus_mag, momag_bin, R_bin, Zlim,
R_extend_flag=False):
"""Calculate the % of annualised fatalities disaggregated by mag and distance.
It is the annualised fatalities disaggregated by mag and distance/ total
annualised fatalities, as a percentage.
Inputs:
saved_fatalities fatalities, dimensions(event, structure)
nu event activity, dimensions(event)
saved_rjb rjb distances array, dimensions(event, structure)
aus_mag earthquake magnitudes, dimensions(event)
momag_bin 1xn array of bounds for moment magnitude bins e.g.
momag_bin = [4.5:0.5:6.5];
Note that the value 0.0000000000001 is added to the last
entry in momag_bin to ensure that values corresponding to
momag_bin(-1) are captured i.e.
momag_bin = [4.5, 5.0, 5.5, 6.0, 6.5]
becomes
momag_bin = [4.5, 5.0, 5.5, 6.0, 6.5000000000001]
R_bin 1xm list containing bounds for distance bins e.g.
R_bin = [0:5:100];
Note that if R_extend_flag==1 R_bin is extended by one
element as follows; R_bin(end+1) = 100000. This is done
to ensure that all values > R_bin(end) are included in the
final R_bin.
R_extend_flag Whether to extend the R_Bin last bin to catch overflow:
True => extend R_bin (see R_bin)
False => do not extend R_bin
Zlim 1x2 array of z-axis limits.
Returns the normalised deaggregated loss, dimensions(mag_bin, dist_bin)
"""
# convert all 'array' data to real arrays
saved_fatalities = num.array(saved_fatalities)
nu = num.array(nu)
saved_rjb = num.array(saved_rjb)
aus_mag = num.array(aus_mag)
momag_bin = num.array(momag_bin)
momag_bin = num.array(momag_bin) # make sure we don't have a tuple
# Verify the array shapes
events = aus_mag.shape[0]
assert nu.shape == (events,)
# get total building value
# get annualised loss in #
(ann_fatalities, _) = calc_annloss.calc_annfatalities(saved_fatalities, nu)
# Setup moment magnitude bins
momag_bin[-1] = momag_bin[-1] + 0.0000000000001
#momag_centroid = momag_bin[:-1] + num.diff(momag_bin)/2;
#if len(momag_centroid) >= 3:
# momag_centroid[-1] = momag_bin[-2] + (momag_bin[-2] - momag_bin[-3])/2
# Setup distance bins (doesn't necessarily need to be Joyner-Boore distance)
Rjb_bin = R_bin[:]
if R_extend_flag:
Rjb_bin.append(10000)
#Rjb_centroid = Rjb_bin[:-1] + num.diff(Rjb_bin)/2.0;
#Rjb_centroid[-1] = Rjb_bin[-2] + (Rjb_bin[-2] - Rjb_bin[-3])/2.0
mLength = len(momag_bin)
RLength = len(Rjb_bin)
# prepare result array for deaggregated annualised loss in $
DeAggFatalities = num.zeros((mLength-1, RLength-1))
for i in range(1, mLength):
## TODO: convert to log()
## print('Now aggregating loss for magnitude greater than '
## '%.2f and less than %.2f' % (momag_bin[i-1], momag_bin[i]))
aus_mag = num.array(aus_mag)
# finding magnitudes in mag bin
#print aus_mag.shape
mInd = num.nonzero((momag_bin[i-1] <= aus_mag) ==
(aus_mag < momag_bin[i]))[0]
subSaved_fatalities = num.take(saved_fatalities, mInd, axis=0)
#print saved_rjb.shape, mInd
sliced_rjb = num.take(saved_rjb, mInd, axis=0)
for j in range(1, RLength):
FatalitiesMatrix = num.zeros(num.shape(subSaved_fatalities))
ind = num.logical_and((Rjb_bin[j-1] <= sliced_rjb),
(sliced_rjb < Rjb_bin[j])).nonzero()
FatalitiesMatrix[ind] = subSaved_fatalities[ind]
TempAggFatalities = num.sum(FatalitiesMatrix, axis=1)
#print TempAggFatalities
[_, TempPercFatalities, Tempcumnu_fatalities] = \
calc_annloss.acquire_riskval(
TempAggFatalities, nu[mInd], 0)
# convert recurrence rates (cumsum(nu)) to
# prob. of exceed in 1 year
TempProbExceed = 1 - num.exp(-Tempcumnu_fatalities)
TempIntPercFatalities = calc_annloss.integrate_backwards(
TempPercFatalities, TempProbExceed)
#print DeAggFatalities.shape, TempIntPercFatalities.shape, TempPercFatalities
if len(TempPercFatalities)>0:
DeAggFatalities[i-1, j-1] = TempIntPercFatalities[0]
else:
DeAggFatalities[i-1, j-1] = 0
# Normalise the Deaggregated loss by the annualised loss in $
NormDeAggFatalities = 100.0 * DeAggFatalities / ann_fatalities;
return NormDeAggFatalities
def calc_annloss_deagg_distmag(bldg_value,
saved_ecloss,
nu,
saved_rjb,
aus_mag,
momag_bin,
R_bin,
Zlim,
R_extend_flag=False):
"""Calculate the % of annualised loss disaggregated by mag and distance.
It is the annualised loss disaggregated by mag and distance/ total
annualised loss, as a percentage.
Inputs:
bldg_value value of each building, dimensions(structure)
saved_ecloss building loss, dimensions(event, structure)
nu event activity, dimensions(event)
saved_rjb rjb distances array, dimensions(event, structure)
aus_mag earthquake magnitudes, dimensions(event)
momag_bin 1xn array of bounds for moment magnitude bins e.g.
momag_bin = [4.5:0.5:6.5];
Note that the value 0.0000000000001 is added to the last
entry in momag_bin to ensure that values corresponding to
momag_bin(-1) are captured i.e.
momag_bin = [4.5, 5.0, 5.5, 6.0, 6.5]
becomes
momag_bin = [4.5, 5.0, 5.5, 6.0, 6.5000000000001]
R_bin 1xm list containing bounds for distance bins e.g.
R_bin = [0:5:100];
Note that if R_extend_flag==1 R_bin is extended by one
element as follows; R_bin(end+1) = 100000. This is done
to ensure that all values > R_bin(end) are included in the
final R_bin.
R_extend_flag Whether to extend the R_Bin last bin to catch overflow:
True => extend R_bin (see R_bin)
False => do not extend R_bin
Zlim 1x2 array of z-axis limits.
Returns the normalised deaggregated loss, dimensions(mag_bin, dist_bin)
"""
# convert all 'array' data to real arrays
bldg_value = num.array(bldg_value)
saved_ecloss = num.array(saved_ecloss)
nu = num.array(nu)
saved_rjb = num.array(saved_rjb)
aus_mag = num.array(aus_mag)
momag_bin = num.array(momag_bin)
momag_bin = num.array(momag_bin) # make sure we don't have a tuple
# Verify the array shapes
events = aus_mag.shape[0]
structures = bldg_value.shape[0]
assert saved_ecloss.shape == (events, structures)
assert nu.shape == (events, )
assert saved_rjb.shape == (events, structures)
# get total building value
tot_bldg_value = num.sum(bldg_value)
# get annualised loss in $
((ann_loss, _), _) = calc_annloss.calc_annloss(saved_ecloss, bldg_value,
nu)
# Setup moment magnitude bins
momag_bin[-1] = momag_bin[-1] + 0.0000000000001
#momag_centroid = momag_bin[:-1] + num.diff(momag_bin)/2;
#if len(momag_centroid) >= 3:
# momag_centroid[-1] = momag_bin[-2] + (momag_bin[-2] - momag_bin[-3])/2
# Setup distance bins (doesn't necessarily need to be Joyner-Boore distance)
Rjb_bin = R_bin[:]
if R_extend_flag:
Rjb_bin.append(10000)
#Rjb_centroid = Rjb_bin[:-1] + num.diff(Rjb_bin)/2.0;
#Rjb_centroid[-1] = Rjb_bin[-2] + (Rjb_bin[-2] - Rjb_bin[-3])/2.0
mLength = len(momag_bin)
RLength = len(Rjb_bin)
# prepare result array for deaggregated annualised loss in $
DeAggLoss = num.zeros((mLength - 1, RLength - 1))
for i in range(1, mLength):
## TODO: convert to log()
## print('Now aggregating loss for magnitude greater than '
## '%.2f and less than %.2f' % (momag_bin[i-1], momag_bin[i]))
aus_mag = num.array(aus_mag)
# finding magnitudes in mag bin
mInd = num.nonzero(
(momag_bin[i - 1] <= aus_mag) == (aus_mag < momag_bin[i]))[0]
subSaved_ecloss = num.take(saved_ecloss, mInd, axis=0)
sliced_rjb = num.take(saved_rjb, mInd, axis=0)
for j in range(1, RLength):
LossMatrix = num.zeros(num.shape(subSaved_ecloss))
ind = num.logical_and((Rjb_bin[j - 1] <= sliced_rjb),
(sliced_rjb < Rjb_bin[j])).nonzero()
LossMatrix[ind] = subSaved_ecloss[ind]
TempAggLoss = num.sum(LossMatrix, axis=1)
[_, TempPercEcLoss, Tempcumnu_ecloss] = \
calc_annloss.acquire_riskval(
TempAggLoss, nu[mInd], 0)
# convert recurrence rates (cumsum(nu)) to
# prob. of exceed in 1 year
TempProbExceed = 1 - num.exp(-Tempcumnu_ecloss)
TempIntPercEcLoss = calc_annloss.integrate_backwards(
TempPercEcLoss, TempProbExceed)
DeAggLoss[i - 1, j - 1] = TempIntPercEcLoss[0]
# Normalise the Deaggregated loss by the annualised loss in $
NormDeAggLoss = 100.0 * DeAggLoss / ann_loss
return NormDeAggLoss
