User:Thejoyforever/sandbox

Environmental niche modelling(1)

Environmental niche modeling (ENM) is one of a series of frameworks which employ mathematical and computational tools to define, describe and quantify the relationship between a species and its geographic distribution and typically then apply this information to map the distribution of the species in changed environmental conditions or at different special or temporal scales (predictive models?). ENM relies on the existence of digital systems for mapping environmental and climatic layers. The method can be in principle applied to any taxonomic group, although should be considered on a case-specific level….

Over the past few decades, there has been tremendous interest in this field in the wake of growing concern over the impact of environmental changes, specifically habitat loss and climate change, on biodiversity. This, in conjunction with the rise of GIS tools, has lead to tremendous advancements, and the emergence of numerous methods.



Despite its great potential for …restoration and conservation planning, there is still debate as to the reliability of predictions obtained through these means, different models consistently returning widely varying results. However, a burgeoning field, many improvements, including ensemble forecasting.

Confusion about terminology


The spur of new developments in the field has created a degree of confusion concerning the terminology used, with a consensus yet to be established within the scientific community [3], despite the publication of several comprehensive reviews [2,3,12] and frameworks [0,1,10,12]. Terms including Environmental niche modeling, Ecological niche modeling, Species distribution modeling [3,10], Habitat suitability modeling[10], Habitat distribution modeling [3], Environmental niche theory and even Bioclimatic envelope models have been poorly distinguished in the literature, and used to refer to the same concept. Furthermore, the term species distribution model has been used to describe both the species niche and the suitability of a habitat to support a species [13]. However, within certain author groups, several clear distinctions have been drawn.

Bioclimatic Envelope Models[4] are referred to as a narrower section of Species Distribution Models, which rely on a strong correlation between current biogeographical distributions of species and large-scale climate variation, thus enabling projected distributions to be inferred based on climate predictions.

For Habitat Suitability Models (HSM), the environmental variables are related to species fitness, whereas for Ecological Niche Models, they are related to the probability of species occurrence [10].

Some authors have identified Ecological Niche Models as defining potential distributions, whereas Species Distribution Models involve actual distributions [13].

The ecological niche concept
A dominant concept of the 1970s, the ecological niche underpins the functioning of ENMs as a fundamental expression of the relationship between an organism and its environment [9], even though use of its initial form has declined massively, with even recent publications arguing its inappropriateness [7]. An essential distinction for this application is between the fundamental and realized  niche[2]
 * Fundamental vs realized niche

Correlative vs. Mechanistic
An essential classification of ENM is based on their underlying methods[3]. While correlative models are applied where strong, often indirect, connections exist between species distribution records and environmental variables, mechanistic methods, also known as ecophysiological or process-oriented models, aim to simulate the biophysiological mechanisms thought to dictate the organism's observed preference or tolerence to the the environmental variable[21], based on extensive knowledge of the species biology. It follows that correlative methods predict the realized niche, whereas mechanistic methods model the fundamental niche.[?] Correlative techniques ... Mechanistic methods show increased phenotypic plasticity and are robust enough to be applied at various spacial and temporal scales, there are few species for which a detailed quantitative response to environmental variables is known

Other classifications
Static models are limited to time-independent predictions, at equilibrium, whereas dynamic models enable the prediction of time-dependent dynamic responses to a environmental change [21], [2].
 * Static vs Dynamic

Profile techniques employ presence-only data, whereas Group discriminating techniques require presence-absence data. This means that profile techniques can be used in the case of cryptic species, where absence cannot be reliably confirmed. In some models, pseudo-absence data is infered from the distribution of the presence points and can be used to implement a group-discriminating paradigm. [cn]
 * Profile vs group discriminating

This classification is concerned with whether species are regarded individually or within community assemblages, respectively [2].
 * "Gleasonian" vs "clementian"

Modelling Methods
A recent review by Franklin et al. [13] proposes the following overview of currently used algorithms:

1. Statistical models

 * The linear model
 * Generalized linear models (GLM)
 * Generalized Additive Models (GAM)
 * Multivariate adaptive regression splines (MARS)
 * Multivariate statistics approaches
 * Bayesian approaches
 * Spatial autocorrelation (SAC)

2. Machine learning methods

 * Decision tree methods
 * Ensemble methods applied to Decision trees
 * Bagging trees
 * Boosted additive trees
 * Random forests
 * Artificial neural networks (ANN)
 * Genetic algorithms
 * Maximum Entropy (MaxEnt)
 * Support vector machines
 * Ensemble forecasting [1]

3. Classification, similarity and other methods for presence-only data

 * Envelope models and similarity measures
 * Species presence versus habitat availability
 * Resource selection functions
 * Ecological niche factor analysis (ENFA)
 * Genetic algorithms for rule production (GARP)
 * MaxEnt
 * Habitat suitability indices and other expert models

Choosing a model
it has been found that model choice is more reliably connected with data availability and purpose of the study than the modeling algorithm itself, suggesting that in choosing a model factors such as [cn] should be considered.

Several frameworks [] have been elaborated in order to guide researchers in choosing and applying Environmental Niche Models. However, results from different frameworks have been largely incoherent [], leading to calls for improved communication betweeen proponents of distinct paradigms[].
 * Frameworks

Data sampling

 * Biological data
 * Environmental data

Evaluation

 * The Kappa statistic (Cohen’s Kappa)

Predictive models
One of the most promising and most spectacularly developping[31] applications of Environmental Niche Modelling is that of projecting impacts of future environmental and climatic changes on species distribution.

Use of Global Climate Models
See also: Global_climate_model, economics of global warming

Challenges and Limitations
instance, serious shortcomings have been identified in the popular stepwise selection procedures in regressions (see Guisan et al. 2002) and new approaches have recently been proposed, such as multi-model inference, boosting and model averaging (Wintle et al. 2003), shrinkage methods[31]

Advances and Development
species migration, population dynamics, biotic interactions and community ecology into SDMs at multiple spatial scales. Addressing[31]