OpenQuake Engine 3.19.0

[Graeme Weatherill (@g-weatherill)]

• Contributed new GMPEs Weatherill2024ESHM20AvgSA,
  Weatherill2024ESHM20SlopeGeologyAvgSA, Weatherill2024ESHM20Homokedastic
  and GmpeIndirectAvgSA

[Paolo Tormene (@ptormene), Michele Simionato (@micheles)]

• Added geometry_file to discard ruptures

[Michele Simionato (@micheles)]

• Fixed exporting from calculations with a grand parent
• Added an uniqueness check to source IDs in the same branch
• Extended the command oq sample to multi fault sources
• Fixed the reading of ruptures with a multisurface with meshes
  of different lengths
• Fixed oq compare: the tolerance parameters were ignored
• Fixed a numerical precision error affecting the hazard curves at
  very high intensities
• Optimized multi fault sources by moving surface calculations in the
  preclassical phase
• Fixed disaggregation in case of mutex sources

[Paolo Tormene (@ptormene)]

• Added the possibility to display/hide announcements on top of all WebUI
  pages, setting them up via the Django admin interface

[Michele Simionato (@micheles)]

• Fixed ZalachorisRathje2019 (mag > Mb_ -> mag >= Mb_)
• Reduced memory consumption in multiFaultSources
• Implemented batch processing for AELO calculations

[Anirudh Rao (@raoanirudh)]

• Improved the performance of non-parametric liquefaction models

[Michele Simionato (@micheles)]

• Renamed 'fpeak' as 'f0' in the HassaniAtkinson2020 GMPE so that the
  Manea2021 GMPE can work
• Internal: extended the command oq mosaic sample_rups to multiple models
• After 3 years of deprecation, forbidden custom methods in GMPE classes
• Changed the universal installer to recreate the venv
• Reduced the number of tasks generated by disagg_by_src and added
  more checks on the naming conventions for the source IDs
• Fixed the gmf_data exporter with a filtered site collection
• Added missing validation on duplicate source IDs for source models
  in format nrml/0.5
• The calculation_mode=ebrisk was not honored when starting from GMFs
• Fixed an ordering bug in the IMTs causing two peaks in the UHS in
  some situations
• Fixed bug in the WongEtAl2022 GMPEs, reported by Thomas Bornstein
• Added a job.ini parameter smlt_branch to reduce the logic tree to
  a specific branch
• Internal: extended oq info to shapefiles

[Anirudh Rao (@raoanirudh)]

• Fixed the conditioned GMFs calculator in the case all the
  stations are filtered away

[Marco Pagani (@mmpagani)]

• Fixed the algorithm used to generate kite surfaces

[Michele Simionato (@micheles)]

• Changed the weighted quantile algorithm to not depend on numpy.argsort
  (since numpy 1.25 argsort produces a different sorting on machines with
  AVX-512 enabled processors, causing different quantiles when the weights
  are not all equal)
• Improved error message for missing fields in the exposure CSV files
• Internal: removed dbserver.user in openquake.cfg
• Fixed extract_from_zip when managing zip files coming from MACOSX

[Paolo Tormene (@ptormene)]

• Internal: extended the MosaicGetter to a GlobalModelGetter class with
  the ability to associate points to countries

[Michele Simionato (@micheles)]

• Removed the obsolete by_country hack when managing the global exposures
• Changed oq sample to use pandas for reading/writing CSVs, thus fixing
  the issue of quoted newlines
• Fixed oq check_input exposure.xml
• Added rupture_dict parameter to be used in scenarios

[Marco Pagani (@mmpagani)]

• Fixed a bug in the calculation of Rjb for gridded surfaces and added a
  dependency from alpha_shapes

[Chris di Caprio (@chrisdicaprio)]

• Contributed many new GMPEs for the New Zealand model

[Michele Simionato (@micheles)]

• Fixed wrong mean_rates_by_src in presence of the colon convention
• If there are no sources close to the (single) site, do not raise an error
• Fixed a bug with --hc taking the parent site collection instead of
  the child site collection
• Extended the exposure field mapping functionality to map the same
  input field to multiple columns in the asset collection
• Removed the experimental aftershock calculator; aftershocks can
  be computed by simply including a file delta_rates.csv

[Anirudh Rao (@raoanirudh)]

• Limited the cross-correlation coefficient of GodaAtkinson2009
  to have an upper bound of 1.0

[Paolo Tormene (@ptormene)]

• Added commands oq plot mean_hcurves_rtgm,governing_mce,disagg_by_src

[Michele Simionato (@micheles)]

• Raised the limit on the product num_assets * num_rlzs when using
  full enumeration by a factor of 10
• Fixed exporting-reimporting GMFs in hdf5 format
• Fixed a bug causing the insured losses to be larger than the
  ground losses in some cases
• Added command oq info executing
• Added extract/relevant_gmfs?threshold= functionality
• Added extract/ruptures?threshold= functionality
• Disabling the DbServer with dbserver.host = local
• Fixed disagg_source for same ID sources
• Disambiguated same ID sources by appending the branch ID after
  the exclamation mark and fixed disagg_by_src accordingly
• Fixed another bug in disaggregation for mutex sources
• Fixed a bug in disagg_source for mutex sources: src_mutex
  was not passed
• Internal: forced the colon convention on mutex sources
• Fixed a bug in the disaggregation with mutex sources incorrectly
  averaging over the rates and not the poes

OpenQuake Engine 3.16.7

[Antonio Ettorre (@vot4anto)]

• Upgraded Django version to version 3.2.21 with a security fix

[Michele Simionato (@micheles)]

• In case of a well known IndexError in disaggregation calculations, solved in
  newer versions of the engine, a helpful message is printed
• Backported fix on sanity check for avg_losses and aggrisk
• Backported fix to avg_gmf
• Backported a fix to oq reaggregate
• Backported fix to event based risk calculations with
  aggregate_by(parent) != aggregate_by(child)
• Backported fix to the risk_by_event exporter for some calculations
  starting from GMFs on a CSV file
• Backported fix to import_gmfs_csv, which was breaking sometimes with a
  filtered site collection

OpenQuake Engine 3.18.0

[Michele Simionato (@micheles)]

• Fixed the command oq extract ruptures producing wrong newlines on
  Windows

[Paolo Tormene (@ptormene), Michele Simionato (@micheles)]

• Fixed an ordering bug in the tagnames in extract_agg_curves breaking
  the QGIS plugin

[Michele Simionato (@micheles)]

• Fixed the function get_mean_stddevs_cy14 to recover the same results
  for the Canada SHM6 model as in engine 3.11
• Required at least Python 3.10.6 on mac and windows

[Claudio Schill (@claudio525)]

• Optimized the Kuehn sigma_mu_adjustment calculation by replacing the
  multi-step interpolation with a single 2D/3D space interpolation

[Michele Simionato (@micheles)]

• Fixed a bug in calculations with a filtered site collection using the
  HM2018CorrelationModel
• Internal: raised a clear error message when get_composite_source_model is
  called without passing a datastore in presence of multifault sources

[Christopher Brooks (@CB-quakemodel)]

• Added non-ergodic implementation of Zhao et al. (2016) GMM (ray-tracing
  of travel paths through anelastically attenuating volcanic regions and
  subsequent adjustment of the path term based on this distance to reduce
  the predicted ground-motion).

[Michele Simionato (@micheles)]

• Optimized (2x) the generation of ground motion fields
• Internal: using a single random number generator inside the GmfComputer
  and not one for each gsim, causing an artificial correlation between gsims
• Implemented maximum_distance_stations and fixed another site
  collection filtering bug in conditioned GMFs
• Extended conditioned GMFs to number_of_ground_motion_fields > 1
• Disabled SLOW MODE on macOS, since the amount of free memory returned by
  psutil cannot be trusted
• Speeding up count_ruptures for multifault sources, which gives a huge
  in some calculations (like event_based for Dominican Republic)

[Marco Pagani (@mmpagani)]

• Fixed a bug in the code that resamples a line, with a small effect
  on the hazard of models containing simple fault sources

[Michele Simionato (@micheles)]

• Optimized the calculation of quantile hazard maps (7x)
• Internal: stored _rates instead of _poes in classical calculations

[Lana Todorovic (@LanaTodorovic93), Anirudh Rao (@raoanirudh)]

• Added several regional liquefaction models to the secondary perils module,
  including:
  ZhuEtAl2017LiquefactionCoastal and ZhuEtAl2017LiquefactionGeneral,
  RashidianBaise2020Liquefaction, AllstadtEtAl2022Liquefaction,
  AkhlagiEtAl2021LiquefactionA and AkhlagiEtAl2021LiquefactionB
  (experimental),
  Bozzoni2021LiquefactionEurope, TodorovicSilva2022NonParametric
• Added new site parameters required by some of these new models to site.py
• Added LiqOccur as a valid IMT, referring to the occurrence or
  non-occurrence of liquefaction at a site
• Added basic documentation for the secondary perils module
  (originally written by @cossatot and extended by @LanaTodorovic93)
• Added rock-slope co-seismic failure computation of Grant et al. (2016)

[Michele Simionato (@micheles)]

• Fixed sanity check on avg_losses and aggrisk breaking for event_based_damage
  calculations following event_based_risk calculations
• Fixed a bug in avg_gmf causing the average to be computed incorrectly
• Removed the XML exporters for hazard curves, maps and uniform hazard spectra
  after nearly 5 years of deprecation

[Paolo Tormene (@ptormene)]

• In the WebUI users are not allowed to see other users outputs by default
  (the old default can be obtained setting ACL_ON = False)

[Manuela Villani (@mvillani), Michele Simionato (@micheles), Paolo Tormene (@ptormene)]

• Produced ASCE-7 and ASCE-41 outputs for the AELO project and hidden
  a few internal outputs

[Athanasios Papadopoulos (@thpap)]

• Adjusted the Swiss-specific implementations of the GMPEs used
  in the Swiss national seismic hazard (SUIhaz15) and risk (ERM-CH23) models.
  The new implementation returns all of total, intra- and inter-event sigmas,
  rather than just the total one

• Extended the ModifiableGMPE class by adding an
  apply_swiss_amplification_sa method that is used in ERM-CH23 to
  apply site specific adjustments for site effects and intra-event
  uncertainty

• Added ch_ampl03, ch_ampl06, ch_phis2s03, ch_phis2s06,
  ch_phisss03, ch_phis2s06 site parameters to the site.py file

[Paolo Tormene (@ptormene)]

• Upgraded requirements: Shapely to version 2.0.1 and Pandas to version 2.0.3

[Michele Simionato (@micheles), Paolo Tormene (@ptormene)]

• Implemented AEP, OEP curves

[Michele Simionato (@micheles)]

• Internal: changed events.year to be in the range 1..eff_time
• Enhanced oq engine --delete-calculation to remove calculation files
  even if the DbServer is on a remote machine
• Fixed another site collection filtering bug in conditioned GMFs
• Fixed a regression in oq reaggregate
• Better error message for logic trees with branchsets exceeding the limit
• Fixed a bug with mutex sources and disagg_by_src

[Manuela Villani (@mvillani), Kendra Johnson (@kejohnso), Michele Simionato (@micheles)]

• Implemented Risk Targeted Ground motion for the AELO project

[Paolo Tormene (@ptormene)]

• Fixed consequence calculations for 'homeless', 'fatality' and 'injury'
  with/without a specified 'time_event'
• Raise an early error if any station data has zero intensity value

[Michele Simionato (@micheles), Paolo Tormene (@ptormene)]

• Internal: checking encoding and newlines of all CSV files in the repository

[Christopher Brooks (@CB-quakemodel)]

• Implemented the Atkinson (2015) GMPE with the alt. eff. depth model

[Michele Simionato (@micheles)]

• Fixed a bug in event based risk calculations with
  aggregate_by(parent) != aggregate_by(child)
• Fixed a bug when exporting risk_by_event in absence of ruptures
• Fixed a bug by not exporting event_based_mfd in absence of ruptures
• Enabled the pointsource_distance approximation by default at 100 km
• Fixed an error with sources below the minimum_magnitude
• Introduced support for SLURM (experimental)
• Fixed bug in conditioned GMFs, due to the station sites being excluded
  from the complete site collection
• Reduced slow tasks and memory consumption
  in the event_based_risk calculator (3x for Chile)

[Paolo Tormene (@ptormene)]

• Raise an error if files specified in site_model_file do not have the same
  headers
• Give a better error message for risk calculations where the aggregate_by
  tag is specified but it doesn't exist in the exposure model

[Michele Simionato (@micheles)]

• Improved error message when quantiles are not supported in event based risk
• Reduced memory consumption in the master node in post_risk
• Reduced slow tasks in the ebrisk calculator and memory consumption
  in the master node

[Paolo Tormene (@ptormene)]

• Fixed bug #8907: unable to run ClassicalDamage demo on Windows

[Astha Poudel (@asthapoudel), Anirudh Rao (@raoanirudh), Paolo Tormene (@ptormene)]

• Updates to the infrastructure risk assessment and connectivity analysis
  module to incorporate additional performance metrics including Complete
  Connectivity Loss (CCL), Partial Connectivity Loss (PCL), Weighted
  Connectivity Loss (WCL), and Efficiency Loss (EL) based on
  Poudel et al. (2023).
• Implemented the possibility of including both weighted and unweighted
  graphs for infrastructure risk.
• Implemented the possibility of including simple undirected graphs, simple
  directed graphs, multigraphs and multi-directed graphs.
• Implemented the Bradley (2012) cross-correlation equations between peak
  ground velocity (PGV), peak ground acceleration (PGA), and 5% damped
  pseudo-spectral acceleration (SA)

[Michele Simionato (@micheles)]

• Moved the generation of the hazard curves from the GMFs in post_execute
• Fixed the error "You must supply a list of magnitudes as 2-digit strings"
  when starting from ruptures in a model with GMPETables (i.e. Canada)
• Replaced sequential tiling with parallel tiling in large classical
  calculations
• Huge reduction of the memory usage in the event_based calculator; also,
  reduced the slow tasks

OpenQuake Engine 3.17.1

[Michele Simionato (@micheles)]

• Fixed the associations event_id, rup_id, rlz_id

OpenQuake Engine 3.17.0

[Michele Simionato (@micheles)]

• Raise an early error if the user disable both realizations and statistics
• Internal: added dbserver.user in openquake.cfg
• Solved the hanging of classical calculations due to large zmq packets
  and reduced memory occupation

[Marco Pagani (@mmpagani), Michele Simionato (@micheles)]

• Added support for the "direct" method in MRD calculations

[Michele Simionato (@micheles)]

• Reimplemented the conditional_spectrum calculator as a post-processor

[Paolo Tormene (@ptormene), Michele Simionato (@micheles)]

• Raise an error when the same parameter is set in different sections
  of the job.ini file

[Michele Simionato (@micheles)]

• Fixed tau-phi inversion in LanzanoLuzi2019
• Fixed another bug with conditioned GMFs appearing as the error
  array[m, n, bad] = exp(mean_covs[0, g, m, n], im) TypeError: list indices must be integers or slices, not tuple

[Claudio Schill (@claudio525)]

• Fixed sorting bug in the sigma_mu adjustment factor in
  the Kuehn et al. (2020) GMM

[Michele Simionato (@micheles)]

• Fixed a subtle filtering bug in conditioned GMFs
• Fixed shakemap calculations to take the risk IMTs and not the hazard IMTs
• Extended shakemap2gmfs to amplify the GMFs based on user-provided vs30
  values
• Removed sorting of the sections in the multiFaultSource reader to keep
  the rup_idxs consistent with the order in the files
• GMFs too small to produce losses now give a warning rather than an error
• Fixed bug in get_ry0_distance breaking conditioned GMFs
• Made sure lon and lat are rounded to 5 digits in the site collection

[Graeme Wheaterill]

• Fixes to bchydro_2016_epistemic, eshm20_craton and lanzano_2019

[Michele Simionato (@micheles)]

• Extended sensitity analysis to file parameters
• Introduced new loss type residents

[Marco Pagani (@mmpagani)]

• Fixed bug in get_profiles_from_simple_fault_data affecting KiteSurfaces

[Michele Simionato (@micheles)]

• Fixed the rupture exporter in presence of GMPETables
• Optimized the postclassical phase
• Fixed a serious issue in the classical calculator with tiles
• Vectorized the association assets <-> hazard sites (up to 32x speedup)
• Fixed bug in disaggregation calculations associated to an error
  in the line [mag] = np.unique(np.round(ctx.mag, 6))
• Optimized the calculation of geohashes by using numba

[Paolo Tormene (@ptormene)]

• Extended total_losses to include business_interruption

[Michele Simionato (@micheles)]

• Recognized occupants_avg field in the exposure
• Optimized reading the exposure (both speed and memory)
• Optimized the rupture sampling for MultiFaultSources and improved
  the parallelization for MultiFaultSources and MultiPointSources
• Made the parameter truncation_level mandatory
• Fixed the damage state header in the aggrisk outputs
• Changed the order of the IMTs to be by period and not lexicographic
• Fixed the realizations extractor for scenario calculations reading
  ruptures with multiple different tectonic region types

[Kendra Johnson (@kejohnso)]

• Fixed a bug in disaggregation by Lon-Lat-TRT when there are multifault
  ruptures and multiple sites

[Michele Simionato (@micheles)]

• Deprecated the conversion in the exposure
• Added new loss types "area" and "number" for usage in risk calculations
• Made the contexts immutable and fixed mutability bugs potentially
  affecting the modules akkar_bommer_2010, bindi_2011_ipe, can_shm6_inslab,
  lanzano_2019, youd_etal_2002, zhang_zhao_2005, cauzzi_faccioli_2008_swiss,
  chiou_youngs_2008_swiss campbell_bozorgnia_2014
• Changed the disaggregation algorithm (see manual)
• Added a view risk_by_rup
• Internal: added command oq compare med_gmv <imt>
• Extended extract/ruptures to extract a single rupture given the rup_id
• Fixed a bug in classical_damage causing a ValueError: could not broadcast
  input array from shape (X,Y) into shape (X,)
• Added check for missing site parameters

[Paolo Tormene (@ptormene)]

• Internal: reduced the size of the gsim column in the view
  required_params_per_trt in the .rst report

[Michele Simionato (@micheles)]

• Optimized the generation of events
• Fixed the GMFs CSV importer so that it can import exported GMFs even in
  the presence of a custom_site_id
• Implemented sampling for mutex sources with grp_probability
• Introduced extreme_gmv parameter in the job.ini
• Introduced a new output "Annual Frequency of Events" and refined the command
  oq plot event_based_mfd?
• Moved the minimum magnitude filtering after the rupture sampling
• Changed the internals of rupture sampling and fixed the generated MFD
  in the case of a same source appearing in multiple logic tree branches
• Fixed a bug in scenario calculations causing a mysterious IndexError in the
  line rctx.clon = c.lons[rctx.sids]
• Changed the disaggregation exporter to add a column with the IML
  associated to the PoE for the mean disaggregation outputs
• Internal: added a parameter override_vs30 in the job.ini

[Marco Pagani (@mmpagani), Michele Simionato (@micheles)]

• Fixed a bug in the source writer when writing NonParametricSources

[Michele Simionato (@micheles)]

• Changed the conditional spectrum storage and fixed the exporter to export
  all PoEs

[Marco Pagani (@mmpagani)]

• Fixed conditional spectrum with investigation time different from 1 year

[Michele Simionato (@micheles)]

• Fixed documentation of the Extractor for hazard curves and maps
• Fixed bug in oq zip: when starting from a directory different from the
  job.ini directory, .hdf5 files for nonparametric ruptures were not included
• Honored minimum_magnitude for MultiFaultSources
• Overhauled the experimental postprocessing framework and fixed many
  bugs in the compute_mrd postprocessor

[Anirudh Rao (@raoanirudh)]

• Added a verification test suite for the Conditioned GMFs calculator

[Michele Simionato (@micheles)]

• Added output rates_by_src
• Fixed method '/v1/ini_defaults' returning NaNs for the site parameters
• Added output mean_disagg_bysrc for single site calculations with use_rates
• Added command oq reduce_smlt
• Fixed reinsurance calculations when there is a single loss type
• Changed the source seed generation to not depend on the source ID suffix
• Tracebacks were printed 3 times in the console: this is finally fixed
• Forbidden duplicate sources in atomic groups
• Forbidden duplicate mutually exclusive sources
• Fixed conditional spectrum calculator with multiple sites and implemented
  parallelization
• Fixed the source ID generation for CollapsedPointSources
• Fixed the error message for wrong weights in the gsim logic tree file
• Fixed KeyError in calculations using the EasternCan15Mid with PGA
• Replaced hard-coded limit to 25 levels in the MRD postprocessor
  with a simple warning

[Kendra Johnson (@kejohnso)]

• Added a job parameter 'reqv_ignore_sources', which indicates which
  sources - specified as a list of thier source ids - should not be collapsed
  using the equivalent distance

[Michele Simionato (@micheles)]

• Extended (re)insurance calculations to manage insurance deductibles at asset
  level
• Internal: added a parameter use_rates to compute statistical hazard curves
  in terms of rates, not probabilities
• Internal: changed disagg_by_src to work in terms of rates
• Added command oq runsite <lon>,<lat> for mosaic PSHA calculations
• Enhanced --reuse-input to reuse preclassical calculations if possible
• Normalized IMTs in the minimum_intensity dictionary (i.e. SA(1.)=>SA(1.0))
• Fixed event based calculations running out of memory due to hazard curve
  arrays being instantiated without need

[Julián Santiago Montejo Espitia (@SantiMon1417)]

• Contributed the GMPE Arteta et al. (2023) for crustal events on northern
  South America

[Michele Simionato (@micheles)]

• Setting both sites and site_model_file now works also for models
  with nonstandard site parameters, like EUR

[Paolo Tormene (@ptormene)]

• Added AELO mode for the engine webui, providing a web form to insert input
  values and launch a calculation. Input values are validated and the
  progress of the calculation can be monitored through the webui. On
  completion, the server sends an email to the job owner, providing feedback
  about failure or completion, with a link to retrieve a traceback or to
  browse the calculation outputs. Corresponding tests were added and the
  existing tests for the webui were improved, cleaning up calculations after
  they have been checked.

[Michele Simionato (@micheles)]

• Changed the default: now all realizations are consider in disaggregation
  calculations, not only the one closest to the mean hazard curve
• Added feature infer_occur_rates for multiFault sources
• Sampling the ruptures was raising a BlockingIOError sometimes

[Kendra Johnson (@kejohnso)]

• Made it possible to use extendModel for multiple branchSets in the
  source model logic tree

[Michele Simionato (@micheles)]

• Added an exporter for disagg_by_src
• Implemented mean disaggregation outputs, with the means computed in terms
  of rates, not probabilities

[Kendra Johnson (@kejohnso)]

• Corrected a bug when using the reqv feature: all sources were
  collapsed instead of the ones with the specified tectonic region types

[Michele Simionato (@micheles)]

• Changed the lon-lat binning algorithm to produce 4 times less bins
• Fixed install.py: it must not set the shared_dir
• Optimized disaggregation in performance, memory consumption and data
  transfer; also changed the task distribution algorithm

OpenQuake Engine 3.16.5

[Kendra Johnson (@kejohnso)]

• Added DEFINED_FOR_INTENSITY_MEASURE_COMPONENT = const.IMC.RotD50 to
  the NGAEast and NSHMP2014 GMPEs

[Michele Simionato (@micheles)]

• Fixed the hanging of classical calculations due to large zmq packets

OpenQuake Engine 3.16.4

[Antonio Ettorre (@vot4anto)]

• Upgraded Django version to version 3.2.19 with some security fixes

[Claudio Schill (@claudio525)]

• Fixed sorting bug in the sigma_mu adjustment factor in
  the Kuehn et al. (2020) GMM

[Graeme Weatherill (@g-weatherill)]

• Fixes to bchydro_2016_epistemic, eshm20_craton and lanzano_2019

[Marco Pagani (@mmpagani)]

• Fixed bug in get_profiles_from_simple_fault_data affecting KiteSurfaces

[Michele Simionato (@micheles)]

• Fixed tau-phi inversion in LanzanoLuzi2019
• Fixed shakemap calculations to take the risk IMTs and not the hazard IMTs
• Removed sorting of the sections in the multiFaultSource reader to keep
  the rup_idxs consistent with the order in the files
• GMFs too small to produce losses now give a warning rather than an error
• Fixed a subtle filtering bug breaking conditioned GMFs
• Fixed bug in get_ry0_distance breaking conditioned GMFs
• Fixed the rupture exporter in presence of GMPETables
• Fixed severe bug affecting the classical tiling calculator
• Fixed rounding bug in disaggregation (in the line
  [mag] = np.unique(np.round(ctx.mag, 6)))
• Fixed the error message for wrong weights in the gsim logic tree file

OpenQuake Engine 3.16.3

[Kendra Johnson (@kejohnso)]

• Fixed a bug in hazardlib.geo.surface.multi affecting the disaggregation

[Michele Simionato (@micheles)]

• Fixed the damage state header in the aggrisk outputs
• Fixed the realizations extractor for scenario calculations starting from
  ruptures with multiple different tectonic region types
• Fixed a subtle mutability bug affecting various GMPEs
• Fixed rupture sampling in the case of a same source appearing in multiple
  logic tree branches
• Fixed a bug in classical_damage causing a ValueError: could not broadcast
  input array from shape (X,Y) into shape (X,) in some situations
• Fixed the CSV importer so that it can import exported GMFs even in
  the presence of a custom_site_id

OpenQuake Engine 3.16.2

[Michele Simionato (@micheles)]

• Fixed a bug in extract_agg_curves affecting the QGIS plugin
• Fixed a bug in scenario calculations causing a mysterious IndexError in the
  line rctx.clon = c.lons[rctx.sids]

[Marco Pagani (@mmpagani), Michele Simionato (@micheles)]

• Fixed a bug in the source writer when writing NonParametricSources
• Removed the conditional spectrum calculator (it was experimental and
  not working properly in many cases)

[Michele Simionato (@micheles)]

• Now we log only when the DB schema is really upgraded
• Fixed the documentation about extracting hazard curves and maps
• Fixed KeyError in calculations using the EasternCan15Mid with PGA
• The minimum_magnitude parameter was ignored for multiFault sources
• Replaced hard-coded limit to 25 levels in the MRD postprocessor
  with a simple warning
• Fixed bug in oq zip: when starting from a directory different from the
  job.ini directory, .hdf5 files for nonparametric ruptures were not included
• Fixed a 32 bit / 64 bit bug in the hinge function of Kuehn et al. (2020)
• Forbidden source IDs with dots (".")

OpenQuake Engine 3.16.1

[Michele Simionato (@micheles)]

• Fixed event based calculations running out of memory due to hazard curve
  arrays being instantiated without need
• Sampling the ruptures was raising a BlockingIOError sometimes

[Kendra Johnson (@kejohnso)]

• Fixed bug in extendModel with multiple branchsets