170 likes | 367 Views
open Modeller A framework for biological/environmental modelling. Inter-American Workshop on Environmental Data Access Campinas - SP, Brazil March 2004. Temperature. Precipitation. Species modelling.
E N D
openModeller A framework for biological/environmental modelling Inter-American Workshop on Environmental Data Access Campinas - SP, Brazil March 2004
Temperature Precipitation Species modelling Species model can be seem as a function telling the probability of the occurrence of some species for a given environmental condition. If we use xi to represent the i-th environment variable, then we have: Example with prob > 0,8: prob = F(x1, ..., xN)
Building a model Occurrence points are the geographical coordinates where the species was found (or observed). Pi = (Lati, Longi)
Temperature Precipitation Building a model For each occurrence point we find the values assumed for the environment variables. Doing that we transform de geographical occurrence points in nicheoccurrence points.
Temperature Temperature Precipitation Precipitation Building a model Based on the niche occurrence points we build a niche model, F(X), through the application of some algorithm (ex: GARP, GAM, Bioclim, Artificial Neural Networks, etc).
Temperature Precipitation Species distribution map The species distribution map is the result of the niche model application over some geographical region with known environment variables values. Thus, the species distribution map is a georeferenced map with species occurrence probabilities in its cells.
Warning! • Despite the terminology used here, strictly speaking, the distribution map shows the environmental similarities between distinct geographical regions according to the modelling algorithm metric. Using these similarities as probabilities of species occurrence must be done in a sensible way. • Some factors as natural barriers and historical influences are not caught by the distribution map. • The quality of the species occurrence data and the environment variable data are strictly related to the distribution map quality.
Distribution map for Terminalia argenteausing GARP algorithm. • Partnership: • Embrapa/UnB/IBAMA/RBGE • Internet downloaded: • Missouri Botanical Garden
Motivation • Read georeferenced environmental maps stored in different formats (GeoTiff, Arc/Info Grid, GXF, etc). • Deal with different coordinate systems and projections to combine the different maps and the species occurrence points. • Let the algorithm researchers concentrate in the algorithm development. • Permit the execution of different algorithms with exactly the same input, so they can be compared.
openModeller Modelling algorithms GARP Neural Networks Bioclim Precipitation openModeller Soil Temperature Environmental data Specimens
openModeller Modelling algorithms GARP Neural Networks Bioclim DiGIR portal Precipitation openModeller Soil Temperature Environmental data Select the environment variables Select the algorithm Send the species occurrence data Specimens Select the species’ name and the internet portals to be searched
Neural Networks Bioclim Soil GARP openModeller DiGIR portal ABCD portal Precipitation Temperature openModeller Modelling algorithms Environmental data Specimens
Library Desktop OR openModeller Web OR Soap OR ... openModeller client interfaces
openModeller algorithm interface Model Building Environmental values at species occurrence points. Ex: [20˚, 115 mm], [22˚, 100 mm] openModeller Modelling algorithm
openModeller algorithm interface Species distribution map generation For each resulting map cell, openModeller asks for the species occurrence probability. Ex: what is the probability for [30˚, 90 mm] openModeller Modelling algorithm Answer with the probability of occurrence Ex: prob = F( [30˚, 90 mm] ) = 0.8
The project • The core is been developed in C++ • Uses GDAL and proj4 open source libraries • Collaborative development • Distributed under GPL license Involved institutions: • CRIA – Centro de Referencia em Informação Ambiental • Poli USP - Escola Politécnica da Universidade de São Paulo • KU – Kansas University
Thank you http://openmodeller.cria.org.br mauro@cria.org.br Questions ?