OpenQuake Engine 3.16.0

[Matteo Nastasi (@nastasi-oq)]

• Bumped oq-libs to version 3.8.0

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

• Fixed the agg_curves extractor to also extract individual curves, thus
  fixing the QGIS plugin visualization

[Michele Simionato (@micheles)]

• Optimized the rupture sampling for MultiFaultSources
• Supported collect_rlzs together with disagg_by_src to avoid running
  out of memory in models with many realizations and many sources

[Marco Pagani (@mmpagani)]

• Implemented the correction of Lanzano et al. (2019) as described in
  Lanzano et al. (2022)

[Michele Simionato (@micheles)]

• Fixed disaggregation calculations when too few tasks were generated

[Anirudh Rao (@raoanirudh), Michele Simionato (@micheles)]

• Changed the algorithm used in classical_risk and classical_bcr
  calculations to take into account the risk_investigation_time

[Michele Simionato (@micheles)]

• Added a warning for pointlike sources with magnitudes >= 8
• Optimized the disaggregation by using our own truncnorm_sf function and
  not scipy

[Anirudh Rao (@raoanirudh)]

• Implemented Conditioned GMFs as defined in Engler et al. (2022)

[Matteo Nastasi (@nastasi-oq)]

• Removed python3-engine-master and python3-engine-worker ubuntu packages

[Michele Simionato (@micheles)]

• Internal: renamed rup_id -> seed in event based calculations to avoid
  confusion
• Parallelized the reinsurance calculation

[Marco Pagani (@mmpagani)]

• Completed implementation of grp_probability for SourceGroups with mutex
  ruptures

[Michele Simionato (@micheles)]

• Huge improvement of the task distribution when using zmq; also, used
  less memory on the master node
• Fixed context collapsing in the multi-site case
• Optimised postclassical, both in terms of memory in the master node
  and reading performance in the workers
• Added parallel.multispawn facility in oq engine --multi --run
• Improved oq engine --multi to run at most 5 jobs at the time
• Fixed bug introduced in engine 3.9 producing wrong numbers in the case of
  mutex ruptures (i.e. the New Madrid cluster in the USA model)
• Added --venv option to the universal installer to install in custom places
• Bug fix: the avg_losses-rlzs output in classical risk was stored
  incorrectly
• Speed up preclassical calculations in presence of complex fault sources and
  similar; for instance there is a 3x speedup for the SAM model

[Trevor Allen (@treviallen)]

• Added class Allen2012_SS14 to Allen (2012) GMM to incorporate Vs30 scaling
  based on Seyhan & Stewart (2014) amplification factors
• Added classes SomervilleEtAl2009NonCratonic_SS14 and
  SomervilleEtAl2009YilgarnCraton_SS14 to Somerville et al (2009) GMM to
  incorporate Vs30 scaling based on Seyhan & Stewart (2014) amplification
  factors
• Added Allen (2022) GMM for Banda Sea earthquakes observed in the North
  Australian Craton

[Michele Simionato (@micheles)]

• Moved the horiz_comp_to_geom_mean functionality from the gsim logic tree
  to the job.ini
• Optimized the parsing of XML files (35x for the Alaska model)
• Bug fix: ModifiableGMPEs with underlying tables were not receiving a
  single magnitude when calling the compute method
• Rewritten ModifiableGMPE to avoid a subtle memory leak

[Matteo Nastasi (@nastasi-oq)]

• Add WEBUI_PATHPREFIX variable to add a prefix to usual webui web path

[Michele Simionato (@micheles)]

• Fixed the exporter of avg_losses-stats when using collect_rlzs=true
• Mitigate the issue of ultra-long planar ruptures affecting many models
• Forced the usage of collect_rlzs for large exposures when computing
  avg_losses with many realizations
• Bug fix: min_mag and max_mag were not honored when using a
  magnitude-dependent maximum distance
• Reduced the memory consumption when using the ps_grid_spacing approximation

[Marco Pagani (@mmpagani)]

• Added a method to compute the mean rate density of two IMTs
• Added the possibility to define the edges of disaggregation bins
• Added a couple of utility methods to the weichert module

[Michele Simionato (@micheles)]

• Introduced automatic tiling for extra-large calculations
• Added command oq purge failed
• Fixed memory leak in the tiling calculator by resetting the ProcessPool
• Fixed an indexing error breaking the GMPE AtkinsonBoore2006 with
  stress drop adjustment

[C. Bruce Worden]

• Extended the AbrahamsonEtAl2014 GMPE to extrapolate the vs30

[Michele Simionato (@micheles)]

• Reduced the number of warnings when the horizontal component is not
  convertible into geometric mean in ModifiableGMPE

[Guillaume Daniel (@guyomd)]

• Added the highest magnitude bin to the Stepp (1972) completeness analysis
  and improved the legend in the plot

[Manuela Villani (@mvillani)]

• Added the Wong et al. (2022) ground motion model

[Marco Pagani (@mmpagani)]

• Added a constant_term attribute to the faultActivityData uncertainty
  model, with a default value of 9.1 for backward compatibility

[Michele Simionato (@micheles)]

• Removed support for Python < 3.8

[Prajakta Jadhav (@Prajakta-Jadhav-25), Dharma Wijewickreme (@Dharma-Wijewickreme)]

• Added GMPE for Zhang and Zhao(2005)

[Michele Simionato (@micheles)]

• Added conditional spectrum calculator in hazardlib
• Fixed the performance report that failed for NGAEast GMPEs
• Reduced the memory consumption in large classical calculations
  (in make_contexts)

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

• Fixed the calculation of conditional spectra

[Michele Simionato (@micheles)]

• Corner case bug: in the presence of parent site collection and a child
  site model the engine was associating the site model to the parent
• Strongly reduced the slow tasks and memory consumption in event_based_risk
  calculations and removed the need to keep (a lot of) the gmf_data table
  in memory
• Fixed a performance issue in event based risk calculations, due to the same
  aggregation IDs being computed for each task
• Increased sys.recursionlimit to solve a rare pickling issue and made
  the engine continue even if _csm is not stored
• Added a parameter max_aggregations with a default of 100,000
• Changed the risk calculators to reuse the hazard exposure (if any)

[Antonio Ettorre (@vot4anto)]

• Bumped h5py to version 3.7.0 and added fiona 1.8.21

[Marco Pagani (@mmpagani)]

• Added the get_dip and get_strike methods to the gridded surface

[Michele Simionato (@micheles)]

• Added calculation of reinsurance-risk_by_event and reinsurance-aggcurves
  given a reinsurance.xml file
• Supported absolute values for deductible and insurance_limit
  in insurance calculations (before we supported only fractions)
• Reduced precision in the header of the UHS csv files
• Managed the case with CoVs = 0 for the beta distribution