The type and number of variables comprising the dimensions of an environmental niche vary from one species to another and the relative importance of particular environmental variables for a species may vary according to the geographic and biotic contexts.
The choice of cell size (resolution) of environmental layers used in modelling is one important factor that may affect predictions.
I don't think there is a "right" grid size for niche model. 1 km grid is the most used resolution simply because the grid of bioclimatic variables is 1 km. The selection of grid size is really dependent on the available gridded environmental variables and the domain of the study.
I don't think there is a "right" grid size for niche model. 1 km grid is the most used resolution simply because the grid of bioclimatic variables is 1 km. The selection of grid size is really dependent on the available gridded environmental variables and the domain of the study.
Currently approx. 1 km resolution environmental grid files are available and it would be more justifiable to work with them. Working with coarse resolution files may not help us to clearly delineate range overlaps of two allied/dependent species (realised niche studies). From 1km gridfiles we can interprete local aspect of landscape which may add to understanding of the niche of the species concerned. Therefore I would suggest using grid files of resolution of 1 km or less.
Luis, cell size will depend of the spatial scale (cartographic scale) of your work or desired product. First you have to define this in order to know the minimum cartographic unit (a.k.a. minimun mappable area). Only when you have defined your working scale, you will be able to define the cell size.
Evidently, the selection of cell sizes of an analysis depends in practice on the actual spatial resolution and spatial accuracy of the available species and relevant environmental data layers. WorldClim for example suggest a much finer resolution than the underlying point data support (attached). Categorical data layers (e.g. analogue cover or soil maps) can be all too easily gridded into finer resolutions than the underlying sampling survey scheme supports.
My empirical experience on the landscape level is that range limits of explanatory variables are robust for resolutions from 30m to 1000 m for both plants and large mobile animals. Finer resolution is not necessarily better in plant SDMs as demonstrated in the second attachment. For large mobile animals fine spatial resolution would seem a priori unsuitable.
In testing spatial resolutions, I occasionally found that a environmental variables at one resolution was replaced by another (related) one at an alternative resolution. Related variables at the landscape level in the attached are climate, elevation and vegetation or vegetation and substrate.
Article Where the bears roam in Majella National Park, Italy
Article Fine resolution distribution modelling of endemics in Majell...
The selection of resolution indeed has a lot to do with the spatial extent of the work plus the size of the organism you are studying (but that would also define the extent).
Additionally, it is important to keep in mind that some modeling approaches are more sensitive to changes in resolution than others. In my experience maxent, for example, seem to be strongly affected by it.
Consult downloadable materials on https://www.researchgate.net/profile/Laszlo_Orloci including:
1. On sample size optimality in ecosystem survey. L. Orlóci, V. D. Pillar
2. The complexities and scenarios of ecosystem analysis Laszlo Orloci in Multivariate Environmental Statistics, Edited by G.P. Patil and C.R. Rao, 01/1993: chapter 20: pages 423-432; Elsevier Science Publishers.
3. Statistical Ecology ... Laszlo Orloci Chapter 5 Section 5.8 in Stat Ecology
4. Flexible gradient analysis: a note on ideas and applications. Otto Wildi, Laszlo Orlóci in Computer assisted vegetation analysis, Edited by Enrico feoli and Laszlo Orloci, 01/1991: pages 249-254; Kluwer Academic Publishers.
Etc.
Hi
You will find an example here
Pelosi, C., S. Bonthoux, F. Castellarini, M. Goulard, S. Ladet and G. Balent (2014). "Is there an optimum scale for predicting bird species' distribution in agricultural landscapes?" Journal of Environmental Management 136: 54–61.
The problem is that most geospatial/climatic data will be of insufficient scale for this problem. The appropriate scale would be one that matches the spatial extent of the environment used/integrated by the organism. So, for instance, I study lizards in the context of climate. Heat transfer (which governs body temperature, physiology, and behavior) occurs at the cm scale, while home range dynamics occur over tens of meters. Unless you collect your own data, there are few resources (outside of LIDAR, etc) that are available. Further, if one takes a statistical approach to the problem, you risk ignoring all of the biology that occurs at these scars. It's a tough problem with no clear answer.
For insects, you can use a 5x5m grid: we have used this grid size for mosquitoes. See also the papers by Burel and Baudry on Carabid beetles.
Best,
Guy
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology