Computational modeling techniques for species geographic distribution are critical to support the task of identifying areas with high risk of loss of Biodiversity. These tools can assist in the conservation of Biodiversity, in planning the use of non-inhabited regions, in estimating the risk of invasive species, in the proposed reintroduction programs for species and even in planning the protecting endangered species. Furthermore, such techniques can help to understand the effects of climate change and other changes in the geographical distribution of species. This chapter presents concepts related to the species distribution modeling and algorithms based on Neural Networks and Maximum Entropy as alternatives for modeling of species distribution. The algorithms were integrated into the open source tool called openModeller used by biologists and other researchers in this area. A case study of modeling the distribution of babaçu (Orbignya phalerata) in the Piauí State – Brazil is presented, evaluating the potential distribution of this species used to produce bioenergy. Fifty models were generated and merged the ten models with best accuracy for each algorithm. The results show that the models obtained by both are consistent. The models obtained with Maximum Entropy seem to reflect best the reality, considering the occurrence pattern of babaçu as a secondary species.
TopIntroduction
Brazil and other regions of the world face pressing environmental problems and Biodiversity loss associated with urbanization, industrialization, agriculture, use of resources and other human activities. These problems cut across national borders, scientific disciplines and observing systems. A key scientific and technical challenge, therefore, is to improve access to existing and emerging sources of environmental, biological, and socioeconomic data and to develop better ways to integrate these data to support a variety of disciplinary and interdisciplinary research and application efforts, and related policy-making initiatives (Canhos et al., 2005).
According to Myers et al. (2000), Brazil is believed to be the country with the richest flora in the planet, about one-sixth of the total. The Brazilian flora richness is distributed into one aquatic biome and six terrestrial ones: Amazon, Savannah, Atlantic Rainforest, Caatinga, Pampa and Pantanal. The Brazilian Biodiversity study can contribute economically and is very important for social and cultural life. Thus, the modeling of biological species can improve conservation and sustainable use of resources.
Advances in communication and information technology have provided an international environment for collaborative work, involving the sharing of knowledge. This fact, together with a clear demand from international conventions and treaties concerning environmental issues are contributing to a change in paradigm concerning free and open access to data, information, and software developments. These new demands are catalyzing the development of a new knowledge area, the Biodiversity Informatics (Canhos, 2003, Corrêa et al., 2006). Information is the basis for defining best practices and strategies for conservation and sustainable use of our natural resources. However, there are enormous gaps of data and knowledge about species occurrence and distribution in Brazil, and the use of species distribution modeling tools is a powerful means of overcoming them.
Computational modeling tools are essential to support the task of identifying regions that have high risk of Biodiversity loss. Such tools are important for directing the efforts to preserve the environment and the region’s most threatened species. According to Pereira and Peterson (2001), the use of tools for modeling the distribution of species can assist in the conservation of Biodiversity in planning the use of non-inhabited regions, prediction of invasion of species, the design of programs for the reintroduction of species and even in projects for protecting endangered species. Furthermore, such tools can help understand the effects of climate change and other changes in the distribution of species (Santana et al., 2008).
The modeling of species distribution is mainly based on the concept of a species ecological niche. The term ecological niche was defined by Hutchinson (1957) as an n-dimensional area where each dimension represents the range of environmental conditions or resources necessary for survival and reproduction of the species, such as temperature, humidity, food resources, light intensity, population density, among others.
Such models are built up grouping and classifying the collected environmental data in order to create a set of well-connected data points: the ecological niche. Neural networks are fairly conventional computational techniques appropriate for both types of problem in which the modeling of species distribution can be inserted: clustering and classification (Beale & Jackson, 1990).
The process of species modeling becomes complex, among other factors, due to the variety of the data nature and mainly because it requires the treatment of lack of data needed to build a reliable model. This kind of problem considers a restricted optimization problem. To solve this issue, the concept of entropy could be applied, because it defines a kind of measure in the space of probability distributions, so that distributions with higher entropy can be favored over others (Tavares, 2004).