OpenQuake Engine 3.26.1

Release 3.26.1

[Matteo Nastasi (@nastasi-oq), Antonio Ettorre (@antonioettorre)]

• Fix and improve of post installation scripts

Release 3.26.0

[Amir Shahmohammadian (@Shahmohamadian)]

• Fixed a bug in the treatment of PGV period in the Jayaram and Baker
  (2009) spatial correlation model resulting in a wrong coefficient
  b = 22.0 instead of b = 25.7

[Michele Simionato (@micheles)]

• Optimized ebrisk calculations by using larger GMF arrays
• Saving memory when running several jobs together with a pool
• Fixed a bug with IMT-dependent weights being not normalized, breaking
  the CND model with use_rates=true

[Paolo Tormene (@ptormene)]

• Fixed a unit-scaling bug in GMF calculations; acceleration IMTs are
  converted to g as before, while velocity measures (PGV) preserve their
  native cm/s units.

[Paolo Tormene (@ptormene)]

• Added exporter for exposure_by_liquefaction_lse and
  exposure_by_landslide_lse (Exposure grouped by Region and by
  Liquefaction/Landslide LSE)

[Michele Simionato (@micheles)]

• Parallelized the exports in workflow calculations
• Fixed the exporters aggexp_tags and aggrisk_tags
• Reduced the number of tasks and the memory consumption in event_based_risk
• Optimized rupture sampling for sources with high multiplicity
• Internal: storing occ_by_trt_smr in the SES.hdf5 file
• Introduced strict flag to disable checks which are possibly too strict
• Internal: storing the real occurrence_rate in the ruptures dataset,
  not multiplied by smweight
• Fixed the slow tasks in AELO, due to accidental disabling of the filtering

[Savas Ceylan (@sceylan)]

• Added EdwardsFah2013Foreland3Bars, BindiEtAl2011Rhypo,
  ECOS2009VariableDepth and ECOS2009FixedDepth GSIMs with
  verification tests.

[Michele Simionato (@micheles)]

• ebrisk calculations with not enough GMFs now just log an error
  instead of failing
• Cached the multifault sections, with a 6x speedup when sampling the
  ruptures of the NZL model
• Avoided copying sources when building the CompositeSourceModel,
  thus reducing the memory occupation and time by 75% for NZL

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

• Implemented secondary perils for ShakeMap calculations

[Catarina Costa (@catarinaquintela)]

• Fixed landcover coefficient assignment on Nowicki Jessee landslide model

[Michele Simionato (@micheles)]

• Saving year and ses_id in the events dataset of SES.hdf5 files
• Splitting multifault and multipoint sources with the colon convention
• Fixed missing weight in AvgPoeGMPE breaking the realization exporter

[Christopher Brooks (@CB-quakemodel)]

• Fix a bug observed in the Aki and Richard's convention check in
  complex faults.

[Michele Simionato (@micheles)]

• Added --kfilter capability to oq run workflow.toml
• Optimized reading and sampling multiFaultSources
• Saved memory in classical calculations with large sources affecting many
  sites, by yieldings contexts
• Changed the rupture sampling algorithm so that the source groups are
  consistent between event based and classical calculations; removed
  useless copies of the sources and improved the performance and memory usage

[Fahrettin Kuran (@kurannn)]

• Added additional magnitude-scaling relationships to wc1994.py
• Added additional magnitude-scaling relationships to thingbaijam2017.py

[Michele Simionato (@micheles)]

• Added checks on MAX_NUM_RUPTURES and ps_grid_spacing
• Added command oq info db to extract the path to the sqlite file
• Implemented a workflow to run the 2026 Global Risk Model

[Thuany Costa de Lima (@thuanycostadelima)]

• Fix Allen (2022) depth truncation bug: hypocentral depth is now
  clamped to [10, 500] km before evaluating the cubic log-depth
  scaling term, as recommended in the original paper (Eq. 4, p. 1049).
• Updated test CSVs to reflect the corrected implementation.
• Added Bayless and Somerville (2024) GMM for Australia:
  BaylessSomerville2024Cratonic and BaylessSomerville2024NonCratonic.

[Michele Simionato (@micheles)]

• Changed postjobs.build_ses to store oqparam/{model}
• Saved memory in the rupture sampling algorithm for fault sources
  (simple, complex, characteristic)
• Extended use_rates to work also with IMT-dependent weights

[Christopher Brooks (@CB-quakemodel)]

• Add ability to specify a down-dip fraction for hypocentre placement
  on the rupture surface within a hypoDepthDist for area, point and
  multi-point sources with a QA test in classical/case_31.
• Fix exporting of ruptures with surfaces defined using a GriddedSurface

[Michele Simionato (@micheles)]

• Storing also aggexp_by outputs if avg_losses_by is enabled
• Extended the rupture exporter to also work with multi-model SES files
• Keeping the correlation matrices at 32 bit to save memory and to double
  the speed

[Christopher Brooks (@CB-quakemodel)]

• Modify MorikawaFujiwara2013 to compute the distance to volcanic
  front (xvf) on a ctx-by-ctx basis instead of requiring it as a
  site parameter for consistency with NIED's own implementation.
• Expand GridAdjustedGMPE to provide raytracing capabilities for
  distance-based corrections and update the QA test (classical/case_11)
  to test this feature.
• Add the 2025 NSHM variants of the anomalous intensity correction
  and the PH correction to the MorikawaFujiwara2013 and some QA
  tests to check the site location-specific component of these
  corrections (classical/case_30).
• Added ability to specify a distribution of hypocenters within a
  simple fault source. Tested in classical/case_28 and unit tests.
• Added the BCHydro adjusted versions of the original NGA GMMs (median
  scaling and single station sigma) and some unit tests.

[Marco Pagani (@mmpagani)]

• Fixing the calculation of ruptures with clusters and mutually exclusive
  sources

[Michele Simionato (@micheles)]

• Computing the GMFs with a single core in scenario calculations,
  changing a bit the numbers (akin to a change of seed), but saving a
  lot of memory, especially in conditioned computations
• Changed the distribution in damage calculations to split in blocks
  of assets, with a huge performance benefit for large scenarios

[Christopher Brooks (@CB-quakemodel)]

• Add ability to specify a magnitude-dependent aspect ratio which
  is evaluated using a predefined expression via a MagDepAratio
  class.
• Add the magnitude-scaling relationship used in the Northern
  Volcanic Arc (NVA) source zone of the BC Hydro SSC logic tree.
• Add the NIED 2025 sigma model to the MorikawaFujiwara2013 GSIM
  and some unit tests.
• Add ability to specify sentinel (-999) z_ref values when using the
  prepare_site_model command.
• Add a relative epistemic uncertainty for seismogenic depths in the
  source model logic tree. Some unit tests added and a QA test in
  logictree/case_33.
• Add ability to specify reference z1pt4 (like for z1pt0, z2pt5 etc.) in
  the job file (QA test in classical/case_13 with MorikawaFujiwara2013).

[Michele Simionato (@micheles)]

• Internal: added an utility hdf5.reduce_file to reduce HDF5 files
• Added parameter rupture_id to be used with rupture_model_file
  to run scenarios starting from a SES.hdf5 file

[Christopher Brooks (@CB-quakemodel)]

• Added new epistemic uncertainties in the source model logic tree
  for the rupture aspect ratio (setAspectRatioAbsolute and
  setAspectRatioRelative). Unit tests added and a QA test in
  logictree/case_32.

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

• Updated build_exposure_hdf5 to work with the 2026 Global Risk Model

[Amirhossein Mohammadi (@amirxdbx)]

• Add Karimzadeh 2023 Azores Islands GSIM implementation with
  verification tests.

[Michele Simionato (@micheles)]

• In event based calculations do not collect together sources following
  the colon convention to avoid duplicated rupture IDs

[Christopher Brooks (@CB-quakemodel)]

• Add GridAdjustedGMPE - a utility GSIM class allowing the
  adjustment of a GMPE using h3-gridded adjustments already
  defined within a hdf5 file. Examples of the hdf5 file structure
  are available in classical/case_11 and the associated unit tests.
• Introduce explicit handling of sentinel z_sed values (-999) within
  the Conterminous US 2023 NSHM's Coastal Plains site amplification
  adjustment model to turn off this correction for such sites.

[Michele Simionato (@micheles)]

• Logging and storing the engine version and not the DbServer version

[Matteo Nastasi (@nastasi-oq)]

• Integrate Christopher Brooks (@CB-quakemodel) PAPERS script to run calculation
  starting from hypothetical ruptures into engine as new API
  entry-point

[Michele Simionato (@micheles)]

• Added variable keep_trt to select a branchset in the gsim logic tree
• Filtering the sites around the rupture also in scenario_damage
• Extended oq shakemap2gmfs to work without a site model file

[Christopher Brooks (@CB-quakemodel)]

• Added the AlternativeCharacteristicMFD class (unit tests and also
  a QA test in classical/case_10)
• Fix some unit conversion bugs in the MorikawaFujiwara2013 and
  also a bug where the xvf cap of 75 km was being applied incorrectly.
• Modify openquake/hazardlib/tests/gsim/utils.py to support checking
  of JMA values too (supported IMT for MF13 GSIM).

[Paolo Tormene (@ptormene)]

• If any station is farther than the site association distance,
  the job keeps running instead of raising a SiteAssociationError

[Michele Simionato (@micheles)]

• Internal: added postproc.debug_rupture
• Saved avg_gmf always in impact mode
• Saved relevant_events always when using --hc

[Christopher Brooks (@CB-quakemodel)]

• Refactored check_gsims to permit unsupported IMTs when a weight
  of zero is specified for them in an IMT-dependent weighted GMC LT.
  A test case (classical/case_39) is added which checks this behaviour
  (permit zero weight for unsupported IMT, and raise error if non-zero
  for same unsupported IMT).

[Michele Simionato (@micheles)]

• Added maximum_depth dictionary (in km per tectonic region type)
  to the job.ini accepted parameters, used to discard deep ruptures
  in event based calculations

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

• If any worker dies before completion, the corresponding master and
  all its subprocesses are killed and the event causing the
  interruption is logged

[Christopher Brooks (@CB-quakemodel)]

• Added capability to specify an IMT-dependent vector of deltas
  to apply to total sigma, tau and phi within ModifiableGMPE.
• Fix a bug where ESHM20 subclasses of BCHydro without a FABA
  taper model incorrectly fellback to the original BCHydro backarc
  term, causing a missing attribute error for backarc. Updated
  classical/case_61 to now test this scenario.
• Correct the _apply_subduction_trench_correction term in
  Si and Midorikawa (1999)'s NorthEastCorrection subclasses
  to use rhypo instead of rrup.

[Anirudh Rao (@raoanirudh)]

• Support sampling from the truncated multivariate normal distribution
  using Paul Brunzema's implementation of the Botev (2017) algorithm
  (https://github.com/brunzema/truncated-mvn-sampler), permitting
  truncation in calculations employing a cross-IM correlation model
• Introduce two new job parameters: truncation_level_between and
  truncation_level_within allowing the user to specify different
  truncation levels for between-event residuals and within-event
  residuals in a scenario or event-based calculation

[Christopher Brooks (@CB-quakemodel)]

• Add ability to specify deltas for tau and phi within ModifiableGMPE.

[Michele Simionato (@micheles)]

• Extended the cross checking between IMTs and GSIMs to ConditionalGMPEs
• Extended export_job_zip to scenario calculations

[Anirudh Rao (@raoanirudh)]

• Added the Holmgren et al. (2024) GMM for Southeastern Africa
• Removed experimental liquefaction models and refactored
  the secondary perils test suite
• Supported embodied carbon as a new loss category

[Michele Simionato (@micheles)]

• Improved error message when all assets are filtered out
• Fixed a bug when prefiltering sources around the poles
• Made it possible to run risk calculations depending on AvgSA
• Internal: inferring the slow tasks information from starmap_info
• Removed numpy.random.seed in the engine code base
• The NegativeBinomialTOM can now be applied to all parametric sources