Hola Ricardo.

Working with geographic data supposes the generation of cartographic products to derive, so it is necessary to define the applied geographical scale. Depending of the geographical scale choosen for your study, you will select the spatial resolution of raster to be used in the analysis.

You need to have into account the minimum cartographic unit (or minimun mappable area), this depend of the cartographic scale of your analysis. All this will give you the level of detail in a visual interpretation.

Minimum cartographic unit can be defined as 0.5 mm of the scale of the map (spatial scale), althoug threre are other methods to define it (e.g. 10 times de size of the pixel).  If the rasters you used to generate the models were 1 km of spatial scale, the cartographic scale of your models is 1:2.000.000.

How this affect your analysis depends on what kind of analysis you are doing, With the information on spatial scale, you can avoid problems associated to spatial autorrelation.

Saludos

Hi

Good points Adrian, and this is one of those "SDMs are as much feel as science" things that we all have to acknowledge. So yes Ricardo, maxent (and many other SDMs) can be a real pain if you want nice simple quantifiable rules that you can run your analysis by without being held to account by a reviewer. My take has always been to look at the practicalities of answering my research question by asking:

Usually I start looking at all of the above while I'm doing preliminary variable evaluations, usually in a GIS environment, and I find the question of resolution quickly defines itself. However, although I like maxent, I never trust any one SDM. I make it a rule to run multiple variable combinations through multiple, and ensemble, model packages and make comparisons between them. The idea being to achieve a level of rigor through the congruence of a diverse suite of models and approaches. Biomod2 is excellent for doing this.

Cheers

Shaun

Maxent SDMs are remarkably resolution-robust. In the attached you can study an empirical example. Beyond the published 30 versus 90m resolution we have tested the 800m resolution and found our result robust at an order of magnitude range of resolutions. You may it interesting to have a look were the small variation in outcome of 30 versus 90m originated: the spatial configuration of one of the categorical environmental predictors (SDM with only contiguous predictor variables are more robust). And the finding that a finer resolution is not necessarily better may apply elsewhere.   

There are many considerations in selection of a resolution. Shaun mentioned a few. For maxent you should aim with your resolution at a number of species presences between 15 and 100 (see attached bear article).

Article Fine resolution distribution modelling of endemics in Majell...

Article Where the bears roam in Majella National Park, Italy