Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Titre : Predicting freshwater fish distribution in the Murray-Darling Basin

Auteur : Srivastava, Sanjeev Kumar

Université de soutenance : Australian National University

Grade : Doctor of Philosophy (PhD) 2005

Résumé partiel
The natural history collections (NHC) of the world hold a huge repository of species occurrence information that can be used for predicting species’ distribution. With the setting up of the Global Biodiversity Information Facility (GBIF) by the Working Group on Biological Informatics of the Organisation for Economic Cooperation and Development, increasingly more NHC data have become digitised and available online. Another important objective of GBIF is to set up a web site with a home page for all the species known to science having information on their life history traits. This study was conducted to examine the prediction of species’ distribution using such occurrence and life history trait datasets. For predicting species distribution in multi-dimensional environmental space using NHC data, several niche-based statistical models are used. The NHC data are considered as presenceonly data because they are random collections over time. Prediction from presence-only data is often erroneous when subjected to conventional statistical modelling. The modelling of such data is best done by niche-based techniques that use presence-only data and enable modelling of species distributions in geo-space, based on the environmental characteristics of known occurrence sites. These techniques can provide a curate and informative synoptic maps for all species with sufficient occurrence records from NHC at regional and national scale. One of the main constraints to predict freshwater fish distribution is the availability of base maps on which predictions are to be made. Another constraint is the collection and derivation of different spatial layers that can be related to the occurrence and abundance of fish species. River systems, being linear habitat features on the landscape, are inherently difficult to model particularly when distribution is to be predicted using spatially distributed parameters. However, drainage networks drain catchments, and this provides a convenient geographic unit for monitoring and managing the impacts of land-use activities on fish habitat and populations. With the advances in geographical information science and the availability of drainage enforced Digital Elevation Models and flowdirection grid, it is possible to derive drainage networks and catchment boundaries. The catchment boundaries and the drainage network of the MurrayDarling Basin (MDB), the study area, were de1ived and the whole MDB was divided into 463 sub-catchments. This enabled collection and derivation of over eighty spatial variables using catchment as a unit.

Présentation

Version intégrale (179 Mb)