OpenQuake Engine 1.3.0

[Matteo Nastasi (@nastasi-oq)]

• gunzip xml demos files after copied into /usr/openquake/engine directory

[Michele Simionato (@micheles)]

• Updated python-django dependency >= 1.6.1, (our repository already
  includes a backported version for Ubuntu 'precise' 12.04); this change
  makes unnecessary "standard_conforming_strings" postgresql configuration
  variable setting
• The event based risk calculator is able to disaggregate the event loss
  table per asset. To enable this feature, just list the assets you are
  interested in in the job.ini file: "specific_assets = a1 a2 a3"
• We have a new hazard calculator, which can be invoked by setting in the
  job.ini file: "calculation_mode = classical_tiling"
  This calculators is the same as the classical calculator (i.e. you will
  get the same numbers) but instead of considering all the hazard sites at
  once, it splits them in tiles and compute the hazard curves for each tile
  sequentially. The intended usage is for very large calculations that
  exceed the available memory. It is especially convenient when you have
  very large logic trees and you are interested only in the statistics (i.e.
  mean curves and quantile curves). In that case you should use it with the
  option individual_curves=false. Notice that this calculator is still in
  an experimental stage and at the moment is does not support UHS curves.
  Hazard maps and hazard curves are supported.
• We have a new risk calculator, which can be invoked by setting in the
  job.ini file: "calculation_mode = classical_damage"
  This calculator is able to compute the damage distribution for each asset
  starting from the hazard curves produced by the classical
  (or classical_tiling) calculator and a set of fragility functions. Also
  this calculator should be considered in experimental stage.
• A significant change has been made when the parameter
  number_of_logic_tree_samples is set to a non-zero value. Now, if a branch
  of the source model logic tree is sampled twice we will generate the
  ruptures twice; before the ruptures were generated once and counted twice.
  For the classical calculator there is no effect on the numbers (sampling
  the same branch twice will produce two copies of identical ruptures);
  however, for the event based calculator, sampling the same branch twice
  will produce different ruptures. For instance, in the case of a simple
  source model with a single tectonic region type, before we would have
  generated a single file with the stochastic event sets, now we generate
  number_of_logic_tree_samples files with different stochastic event sets.
  The previous behavior was an optimization-induced bug.
• Better validation of the input files (fragility models, job.ini)
• The ability to extract the sites from the site_model.xml file
• Several missing QA tests have been added
• The export mechanism has been enhanced and more outputs are being exported
  in CSV format
• New parameter complex_fault_mesh_spacing
• Some error messages have been improved
• A lot of functionality has been ported from the engine to oq-lite,
  i.e. a lite version of the engine that does not depend on
  PostgreSQL/PostGIS/Django nor from RabbitMQ/Celery. This version is
  much easier to install than the regular engine and it is meant for
  small/medium computation that do not require a cluster. The engine
  demos, have been moved to the oq-risklib repository, so that they can
  be run via the oq-lite command without installing the full engine.
• Currently the following calculators have been ported (all are to be
  intended as experimental): classical hazard, classical tiling, event
  based hazard, scenario hazard, classical risk, scenario damage,
  classical damage.