Version 0.7.0 of the openModeller library released

Version 0.7.0 of the openModeller library is now available. This release contains the following main changes/features:

• New algorithm using Artificial Neural Networks;
• New option to generate distribution maps directly in ARC/Info ASCII grid format (byte or floating point);
• New keywords "Spatially unique" and "Environmentally unique" can be used to filter input points when using om_console;
• Change in AUC calculation: when no absences are provided, instead of using 1-specificity after generating pseudo-absences, it now uses the proportion of points predicted present for a large number of random background points;
• New method in the RocCurve class to calculate partial area ratio;
• New parameters to om_test: --calc-matrix, --calc-roc, --num-background and --max-omission.

Please note that the previous release (0.6.1) also includes these changes:

• Many adjustments in the command line tools;
• New method in the modelling service to perform external tests;
• Possibility for the modelling server to submit jobs to Cluster nodes via Condor.

More details about the latest release can be found here .

How to add new algorithms in openModeller

A new document is available to help those who are interested in adding new algorithms in openModeller. All algorithms must be written in C++. Apart from that, the framework is quite flexible to accommodate new algorithms since they are completely free to decide how distribution models will be represented. The new manual is available here.

openModeller being used by the GBIF data portal

The Global Biodiversity Information Facility (GBIF) recently released a new version of its data portal allowing users to generate niche models in real time. After searching for occurrence records, users have the option to create a niche model when visualizing results.

The data portal uses a Java client that can communicate with a remote modelling service through the openModeller Web Service protocol (OMWS). The Java client was developed by the University of Colorado, and the modelling service by CRIA. Initially, models are being generated only for present time using the Envelope Score algorithm.

GBIF is the world largest biodiversty data portal, currently serving more than 140 million specimen/observation records of which 80% are georeferrenced.

Version 0.6.0 of the openModeller library released

Version 0.6.0 of the openModeller library is now available. This release contains mainly new features, including:

• Drivers to directly read occurrences from GBIF or TAPIR/DarwinCore providers.
• A new command-line tool (om_points) that can be used to retrieve occurrence data using any of the available drivers.
• A new command-line tool (om_algorithm) to get information about the available algorithms.
• A new Maximum Entropy algorithm with two training methods: GIS (Generalized Iterative Scaling) and L-BFGS (Limited-Memory Variable Metric).

The GARP Best Subsets algorithm was also changed to accept the "max threads" parameter, which can be used to speed up the modelling process in multi-processor machines.

More details about this release can be found here:

https://sourceforge.net/project/showfiles.php?group_id=101808&release_id=611859

Released version 0.5.3 of the library

A new version of the openModeller library is available. This release (0.5.3) contains many bugfixes and a new algorithm called Aquamaps. The algorithms "minimum distance" and "distance to average" were removed since EnvironmentalDistance now provides the same functionality. The TerraLib drivers were updated for compatibility with TerraLib 3.2.0. There are two new classes for pre-analysis on input layers to perform jackknife and chi-square (only available as library classes - their functionality still need to be exposed in the command-line and/or Desktop interfaces).

More details about this release can be found here:

http://sourceforge.net/project/shownotes.php?group_id=101808&release_id=587262

openModeller Desktop version 1.0.6 Released

It is our great pleasure to announce the immediate availability of openModeller Desktop Version 1.0.6 openModelleris a generic framework for carrying out fundamental niche modelling experiments - typically used to predict species distribution given a set of environmental raster layers. The openModeller Desktop application builds on the openModeller library by providing a user friendly graphical user interface for carrying out niche modelling experiments.

This release introduces several new features including:

• Improved mapping capabilities - you can now specify occurrence symbol size and colour, and overlay a context layer (such as country boundaries) on your output map.
• Experiments run in their own thread.- this makes for a more responsive user interface and you can look at completed models whilst the experiment is still running.
• Web Service modelling improvements - improvement have been made to allow better job progress reporting when running as a web sevice.
• Absence modelling is now supported - previous versions of openModellerDesktop did not handle absence data properly so absence modelling was not available.
• GBIF Rest Data Fetcher (work sponsored by GBIF with our thanks!) - it is now possible to retrieve occurrence data using the new GIF Rest query service. The data fetcher has also had various cosmetic and useability improvements, and now functions as a wizard.
• New algorithms - openModeller Desktop now ships with two new algorithms (Support Vector Machines and Bioclim Score).
• Thresholding tool - you can now compute 'hotspot' models showing the predicted species count per cell by aggregating many model outputs into a single raster file. The threshold tool can also be used for 'concensus' modelling where many algorithms are used for a single species and all the resulting probability distribution maps are combined to provide a map indicating where most algorithms are in agreement that a species will ocurr at a given cell.

The release also includes numerous improvements, bug fixes and stability improvements. openModeller Desktop is available in source form, and as binary executables for Microsoft Windows, Mac OS X, and Ubuntu 7.10 GNU/Linux. All versions can be obtained from our download page. As an open source project, we provide support for using openModeller via our mailing lists and bug tracker:

• For general enquiries subscribe to our users mailing list.
• For developer related enquiries subscribe to our separate developers list.
• If you think you have found a bug, please report it using our bug tracker. When reporting bugs, please include some contact information in case we need help with replicating your issue.

Known Issues

The threshold tool can crash with large datasets, this will be addressed in next release.

Released version 0.4.2 of the library

A new version of the openModeller library is available. This release (0.4.2) contains some bugfixes and two new features:

• ROC curve was included in the model statistics (AUC is now automatically calculated by om_console, which generates pseudo-absence points if no real absence points are provided).
• Metric Chebyshev can be selected as a parameter in the Environmental Distance algorithm.

For developers, the Log object is now a singleton instead of a global variable.