Assessing transferability of ecological models: an underappreciated aspect of statistical validation

By S. Wenger, Julian D. Olden. 2012

583 citations recorded by the source corpus.

Subfields: Ecological Modeling, Ecology, Nature and Landscape Conservation

View Paper: https://www.semanticscholar.org/paper/a076597d205ad2519ff061529f67705b1a7d0e43

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecological Modeling

In Ecology

In Nature and Landscape Conservation

Citations in the corpus

Establishing the link between habitat selection and animal population dynamics (2015, methodology): “HSFs are supported by extensive statistical theory, widely available software, and graphical diagnostics, and they frequently outperform more opaque machine-learning methods such as neural nets (Wenger and Olden 2012).”
A comprehensive evaluation of predictive performance of 33 species distribution models at species and community levels (2019, background): “…carries the cost of increasing the number of degrees of freedom, which, in turn, increases the risk of statistical overfitting and thus modeling noise rather than signal (Ara ujo et al. 2005, Randin et al. 2006, Wenger and Olden 2012, Merow et al. 2014, Garc ıa-Callejas and Ara ujo 2016).”
The priority of prediction in ecological understanding (2017, background): “More recently, Bahn and McGill (2013) and Wenger and Olden (2012) explicitly address the issue of assessing the transferability of habitat occupancy models.”
A review of methods, data, and models to assess changes in the value of ecosystem services from land degradation and restoration (2016, background): “One of the major problems with using models is the trade-off between specificity and applicability, which the user must assess when determining the transferability between sites (Wenger and Olden, 2012).”
Dynamic population mapping using mobile phone data (2014, background): “sets are sampled from geographically distinct regions (56), allowed for a quantitative assessment of the extrapolation capacity of the model (57, 58).”
Performance evaluation and hyperparameter tuning of statistical and machine-learning models using spatial data (2018, methodology): “…spatial resampling methods for performance estimation has been emphasized repeatedly in recent years (Gei et al., 2017; Meyer et al., 2018; Wenger & Olden, 2012), such techniques have not been adopted in all cases (Bui et al., 2015; Pourghasemi & Rahmati, 2018; Smoli´nski & Radtke, 2016;…”
blockCV: an R package for generating spatially or environmentally separated folds for k-fold cross-validation of species distribution models (2018, methodology): “The package also allows division of the study area into vertical or horizontal bins of a given height or width (rectangle blocks, Figure 1b,c), as used by Wenger and Olden (2012) and Bahn and McGill (2012) respectively.”
Partial versus Full Species Distribution Models (2012, methodology): “For reasons of clarity; this is different from testing how well models fitted within artificial boundaries are capable of predicting a species’ full extent of occurrence, known as transferability studies (Wenger & Olden 2012; Zurell et al.”
Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models (2020, background): “We remark that blocked CV can be implemented in ‘biomod2’ through its in-built ‘BIOMOD_cv’ function, which can create spatially sliced blocks (Wenger and Olden 2012).”
Importance of spatial predictor variable selection in machine learning applications - Moving from data reproduction to spatial prediction (2019, background): “…spatial prediction models, and the performance estimates must account for that which is increasingly and consistently recommended in recent literature (e.g. Wenger & Olden, 2012; Juel et al., 2015; Roberts et al., 2017; Valavi et al., 2018; Poh-jankukka et al., 2017; C´anovas-Garc´ıa et al., 2017).”
Selecting predictors to maximize the transferability of species distribution models: lessons from cross‐continental plant invasions (2017, background): “…transferability can 67 result from many, possibly interrelated factors, such as violation of the niche conservatism 68 assumption (Broennimann et al., 2007; Early & Sax, 2014; Guisan et al., 2014) or methodological 69 limitations (Randin et al., 2006; Peterson et al., 2007; Wenger & Olden, 2012).”
Protected areas network is not adequate to protect a critically endangered East Africa Chelonian: Modelling distribution of pancake tortoise, Malacochersus tornieri under current and future climates (2020, methodology): “Disadvantages of using cross validation, are that only part of the data can be used for model fitting and that it is challenging to obtain test data that is statistically (spatially) independent of training data (but see Hijmans 2012, Wenger and Olden 2012).”