Recent papers of interest

(Page under development....)

Niche limits, range limits and distributional modelling is a rapidly evolving field. I have tried to dedicate this page to extra-topical papers that were released since March 2014 with brief summaries/comments. I will try to continue adding to this page whenever possible.

Studies Directly Evaluating Correlative Species Distribution Models Through Experimental Field Translocations

Ebeling, S., Welk, E., Auge, H., & Bruelheide, H. (2008). Predicting the spread of an invasive plant: combining experiments and ecological niche modelEcography, 31(6), 709–719.doi:10.1111/j.1600-0587.2008.05470.x

Compared performance of an invasive shrub from common garden studies in its invaded range in Europe to distribution models developed from within the species native range. Their correlative distribution model used for the study showed strong agreement to the common garden field study. The authors were able to suggest niche conservatism between native & invaded range and then use the distribution model to predict areas of future invasion. 
Model: GARP, Variables: Climate, Scale: 1000’s km

Pattison, R., & Mack, R. (2008). Potential distribution of the invasive tree Triadica sebifera (Euphorbiaceae) in the United States: evaluating climex predictions with field trialsGlobal Change Biology, 14(4), 813–826. doi:10.1111/j.1365-2486.2007.01528.x

Using the CLIMEX program, Pattison & Mack generated a distribution model to predict the spread of an invasive tree. Then the correlated the results of their distribution model to seed and planting trials at different locations within and beyond the species range. There was a positive correlation between site germination rates & CLIMEX scores but not for growth rates of seedlings suggesting direct associations, if any, were life stage specific.
Model: CLIMEX, Variables: Climate, Scale: 1000’s km

McLane, S., & Aitken, S. (2012). Whitebark pine (Pinus albicaulis) assisted migration potential: testing establishment north of the species rangeEcological Applications, 22(1), 142–53. Retrieved from:

McLane & Aitken evaluated the relationship between the performance of field transplants of white bark pine to site suitability inferred from distribution models to better understand the feasibility of assisted migration. Transplant sites were established within and beyond the species range. For obvious reasons this study was only able to assess the relationship for small seedlings, but regardless they found strong agreement between site suitability and distribution models. Site climate conditions did vary considerably.
Model:, Variables: Climate, Scale: 100’s km

Sheppard, C., Burns, B., & Stanley, M. (2014). Predicting plant invasions under climate change: are species distribution models validated by field trials? Global Change Biology, Advanced online. doi:10.1111/gcb.12531

Ran MaxEnt models for three species invasive species in New Zealand using distribution records in their native range. The authors then transplanted experimental populations across New Zealand to a series of sites differing in their climatic conditions & therefore their inferred suitability. This study was perhaps the most direct evaluation of SDMs with field trials. In general there was moderate to strong agreement between transplant performance & site suitability from SDMs but not for all species. Also there were lots of outlier locations with sites inferred to be unsuitable, but had high transplant growth & visa-versa.
Model: MaxEnt, Variables: Climate, Scale: 100’s km

Isaac-Renton, M., Roberts, D., Hamann, A., Spiecker, H. (2014) Douglas-fir plantations in Europe: a retrospective test of assisted migration to address climate change. Global Change Biology. Accepted Manuscript (April 16, 2014).

This study arguably acted as the “biggest & greatest” meta-analysis for Douglas-fir assessing the relationship between transplant performance and inferred site suitability from distribution models. The authors were able to make use of years and years of field transplant data from field provenance trials in Europe. 2800 provenances X 120 sites in Europe. Distribution models were generated for fir from its native range in North America and projected to sites in Europe. The models generally were able to predict growth & survivorship along a large north – south latitudinal gradient, but failed to predict performance of coastal North American populations planted in Eastern Europe.
Model used: Random Forest, Variables: Climate, Scale: 1000’s km.

Swab, R.M., Regan, H.M., Matthies, D., Becker, U., Bruun, H.H. (2014) The role of demography, intra-species variation, and species distribution in species' projections under climate change. Ecography: 38, 221 - 230.

A lot was packed into this study. Researchers used data from a transplant study with a thistle across five sites in Europe spanning a massive climatic gradient. Population growth rates and vital rates were compared against climatic variables from sites as well as to SDM model output. There was no clear relationship between lambda and SDM scores of sites. It was suggested that the field performance of transplants was overwhelmed by site-level effects. This study also investigated metapopulation dynamics and local adaptation. 
Model: MaxEnt, Variables: Climate, Scale: 1000’s km.

Searcy, C.A. and Shaffer, B. Do Ecological Niche Models Accurately Identify Climate Determinants of Species Ranges? American Naturalist. 
Defining species’ niches is central to understanding their distributions and is thus fundamental to basic ecology and climate change projections. Ecological niche models (ENMs) are a key component of making accurate projections and include descriptions of the niche in terms of both response curves and rankings of variable importance. In this study, we evaluate Maxent’s ranking of environmental variables based on their importance in delimiting species’ range boundaries by asking whether these same variables also govern annual recruitment based on long-term demographic studies. We found that Maxent-based assessments of variable importance in setting range boundaries in the California tiger salamander (Ambystoma californiense; CTS) correlate very well with how important those variables are in governing ongoing recruitment of CTS at the population level. This strong correlation suggests that Maxent’s ranking of variable importance captures biologically realistic assessments of factors governing population persistence. However, this result holds only when Maxent models are built using best-practice procedures and variables are ranked based on permutation importance. Our study highlights the need for building high-quality niche models and provides encouraging evidence that when such models are built, they can reflect important aspects of a species’ ecology.


Antoine, M.E., & McCune, B. (2004). Contrasting fundamental and realized ecological niches with epiphytic lichen transplants in an old-growth Pseudotsuga forest. Bryologist 107, 163–173.Modelled abundance of lichen in relation to forest canopy height (A local correlative distribution model at the scale of m's). Then compared growth & biomass accumulation through an experimental field translocation of three species to the different forest canopy heights. Two of the species showed strong agreement between distribution models & subsequent biomass accumulation from the field transplant, but the anomalies for the third species was thought to be the result of competitive exclusion that was experimentally manipulated in the field study (Micro site differences, Not beyond range). 

Wright, J.W., Davis, K.F., Lau, J.A. et al. (2006). Experimental Verification of Ecological Niche Modelling in a Heterogeneous Environment. Ecology 87(10), 2433–2439.Scale: 500m/500m grid in old field meadow. Developed a SDM-GAM model for Collinsia spariflora (annual). Model parametized with occurence of the species ~ soil chemistry & soil depth variables. Planted seeds into grid of the species. Model of predicted suitability matched growth of seedlings, but not for all source populations. Model used: GAM, Variables: soil, Scale: 500m
Moore, K. A., Elmendorf, S.C. (2006). Propagule vs. niche limitation: untangling the mechanisms behind plant species' distributions. Ecology letters. 9: 797-804. 
Under the assumption that SDMs for species with high AUC scores (strong affinity for environment) will correspond with stronger fitness limitation rather than niche limitation, the authors transplant populations to suitable & unsuitable sites. These sites are defined as suitable or non-suitable. Enviornmetal variables are defined based on soil-type & plant community similarity/dissimilarity. Most of their study species showed strong evidence of niche limitation rather than dispersal limitation, but for two there was strong evidence of propagule limitation.  

Other studies with correlates of fitness and "environmental suitability" predictions from ecological niche models. 

VanDerWal, J.Shoo, L.P.Johnson, C.N. & Williams, S.E. (2009Abundance and the environmental niche: environmental suitability estimated from niche models predicts the upper limit of local abundanceThe American Naturalist174282291.
It is expected that there should be a positive association between high scores from an ENM and local abundance. But we know that ENMs are not perfectly modelling the physiological limits of a species so perhaps we will see that ENMs predict the upper limits of abundance. Positive associations were observed between ENM scores & abundace, but there was heavy noise for all species examined. 

ELMENDORF, S. C. and MOORE, K. A. (2008), Use of Community-Composition Data to Predict the Fecundity and Abundance of Species. Conservation Biology, 22: 1523–1532. 
The authors compare community-composition based model (similarity of communities) SDMs from basic plot-level environmental variables. To evaluate which model works best (the community model or the classic SDM) the authors planted seeds of four grass species across plots. The authors also compared the abundance of wild species to their predicted suitability scores. The community model outperformed the environmental SDM in terms of predicting abundance as well at the fitness of the transplants. 

Monnet, A., Hardouin, L.A., Robert, A., Hingrat, Y., Jiguet, F. (2014), Evidence of a link between demographic rates and species habitat suitability from post release movements in a reinforced bird population. Oikos, 000: 001-009. 
Birds were released from a release point and followed with the idea that the habitat suitability along their route of travel might be related to survival. If individuals moved to areas of lower habitat suitability their survival was also decreased, hence in the system Monnet et al worked with their SDM scores were related to bird survivorship. Environmental variables were derived at a 1km resolution and consisted of climatic & landcover variables. 

Tredennick, A.T., Adler, P.B. (2015) Do we need demogrpahic data to forecast population response to climate change? - PREPRINT BioRxiV
Individual level demographic models are more accurate than population-level density/occurrence models in terms of their ability to forecast effects of climate change. Population-level models developed based on occurrences tend to miss certain association between vital rates and environmental variables. But the main problem is that for individual-level demographic models, collecting enough data for is exceedingly laborious/difficult and probably not practical for most species. Unfortunately in this study predictions from both the individual demographic model and population-level occurrence/distribution model had a high degree of uncertainty. Neither approach could be deemed fully suitable in terms of applications of the predictions for management/conservation decisions. 

Other recent papers of interest

Blonder, B., Lamanna, C., Violle, C., and Enquist, B. (2014) The n-dimensional hypervolumne. Global Ecology and Biogeography. 23, 595 - 609.

A great overview of contemporary theoretical standpoints

Isaac-Renton, M., Roberts, D., Hamann, A., Spiecker, H. (2014) Douglas-fir plantations in Europe: a retrospective test of assisted migration to address climate change. Global Change Biology. Accepted Manuscript (April 16, 2014).

Diez, M., Giladi, I., Warren, R., Pulliam, H. (2014) Probabilistic and spatially variable niches inferred from demography. Ecology. Published online (January, 2014).

After six years of intensive demographic field studies following wild populations of Goodyera pubescens (an orchid in Eastern NA), these researchers incorporated fine scale environmental field measurements of soil moisture & solar isolation (at the plot level) into their demographic models in order to better understand how the growth rate of populations across the landscape can be explained by the species probabilist environmental niche. Despite predictions from basic niche theory, high lambda values did not correspond with optimal microsite conditions, suggesting strong meta-population dynamics were taking place + the authors also draw attention to the importance of hidden niche dimensions & importance experimental translocations.

Searcy, C.A., Shaffer, H.B. (2014) Field validation supports novel niche modelling strategies in a cryptic endangered amphibian. Ecography. 37: 001-010 (Early view).

An excellent example of the importance of validating a distribution model against a truly independent evaluation dataset rather than the more common cross validation approach. Also a nice comparison of local (regional) and global development.

Slavich, E., Warton, D.I., Ashcroft, M.B., Gollan, J.R., Ramp, D. (2014) Topoclimate versus macroclimate: how does climate mapping methodology affect species distribution models and climate change projections. Diversity and Distributions. (2014) 1 - 14 (Early View)

Although high resolution climate raster grids are tremendously useful for most niche modeling exercises, there are always challenges with climate resolution, accuracy and the discrepancy between weather station locations & the microsite of the organism of interest. In this study Slavich et al install over one hundred ibutton temp loggers in a small region & make their own climate grids that also incorporate micotopography & canopy cover. SDMs are developed for selected grasses and fern species and models are compared to SDMs developed from the more popular climate grids available online (~1km res). The customized climate grids developed for this project explained significantly more deviance than the more conventional sources & have import implications for modeled species occupying particular microsites such as cold air drainages.  

Merow, C., Latimer, A.M., Wilson, A. et al. (2014) On using integral projection models to generate demographically driven predictions of species' distribution: development and validation using sparse data. Ecography 37: 001-017.

Presence/absence data could be considered as the most simplified type of data set, with no insight into population specific growth rates, fecundity or survival (although we often try to speculate about these parameters from more conventional distribution models). This unique study develops a distribution model using demographic parameters that correlate with enviro variables to make spatial predictions. 

Warren, D.L., Cardillo, M., Rosauer, D.F., Bolnick, D.I. (2014) Mistaking geography for biology: inferring processes from species distributions.  Trends in Ecology and Evolution. 29 (10). 572 - 580.

Probably one of the most basic assumption of correlative niche modelling is that the distribution of a species over the landscape is indeed driven by the selected enviro variables more strongly than historical biogeograpahy. This is basic assumption is easily forgotten and we often fail to consider the implications with ecological niche models.

Merow, C., Smith, M.J., Edwards, T. C., Guisan, A., McMaon, S.M., Normans, S., Thuiller, W., Rafael, O., Zimmermann, N., Elith, J. (2014) What do we gain from simplicity versus complexity in species distribution models. Ecography 37: 001-015.

There is always such a push for the 'best model', even though all modellers are well aware that most controversial decision really just come down to different trade-offs. In this excellent review, we are reminded to focus on our study/research objectives, our dataset, and species/system rather than chasing the most recent trend or popular approach. Guidelines & recommendations are provided for attributes of models including: prediction, SAC, data res, sampling bias, sample size, hypothesis testing vs generation and the final application.

Jackson, H.B., Fahrig, L. (2014) Are ecologists conducting research at the optimal scale? Global Ecology & Biogeography (accepted article).

In landscape ecology, the effect and dependency of the spatial or temporal scale on the ecological question is usually given some consideration, but was the spatial scale chosen for study 'X' & question 'Q' really the most appropriate?  How many times do researchers choose the right scale for their study system? In this meta-analysis Jackson & Fahrig (2014) look at across a series of mulit-scale studies and concluded that a majority of the time, the largest ecological effect was found at what also happened to be the smallest or largest scale investigated, suggesting the true effect was outside of the range of the study.

Mellin, C., Mengersen, K., Bradshaw, C.J. (2014). Generalizing the use of geographical weights in biodiversity modelling. Global Ecology & Biogeography 23: 1314 - 1323.

The varying response of ecological processes & environmental relationships across space presents a major challenge for conventional SDMs. The variability in species-environmental response curves across different regions could be either from local adaptation or complex interactions with other variables. Regardless we refer to this phenomenon as non-stationarity. The most conventional approach to account for a dataset with heavy non-stationarity it compare the performance & predictions of global models to local or regional models, but usually it is hard to draw the line for certain regions or often outliers are clustered in strange patterns. Geographically weighted regression (GWR) is widespread across other research fields (e.g. municipal studies or health, crime, urban planning ect), but the incorporation of GWR into ecological SDMs has remained extremely rare, especially since most GWR packages have limited ability to work with binomial distributions.

Renwick, K.M., Rocca, M.E. (2014). Temporal context affects the observed rate of climate-driven range shifts in tree species. Global Ecology and Biogeography (accepted article).

When we consider range shifts and the delayed migration of species in response to climate change, we tend to occasionally over-simplify our idea of migration as being a smooth and gradual process. But in reality migrational lags are complex & usually driven by a variety processes such as reduced propagule pressure, major barriers over the landscape or mutualist interactions (e.g. pollinators, herbivorous, soil development at high altitudes ect). When we have historic survey records of a species range front from some time in the early ~ 1900's it is always tempting to think that the shift to present day conditions (2014) was continuious, but as Renwick & Rocca (2014) point out in this article, it is much more likely to have abrupt and sporadic range shifts and we don't really have any grounds to expect gradual transitions without a continuous survey.

Swab, R. M., Regan, H.M., Matthies, D., Becker, U., Bruun, H.H. (2015). The role of demography, intra-species variation and SDMs in species projections under climate change. Ecography 38: 221-230. 

No comments:

Post a Comment