We have a shiny new version of ENMTools ready to go! This version includes some minor bug fixes and adds a few new features. The new features are described in detail in the user manual, but here's a quick rundown:
*Handy tool for eliminating duplicate occurrence points from a .csv file using either exact location or an ASCII grid.
*New feature to make maps of the spatial distribution of residuals from a regression between two environmental variables (Warren and Moskwik, in prep).
*Tool to standardize raster files so that they sum to 1 over the geographic space.
*Tool for calculating range overlap from rasters, applying a user-selected presence/absence threshold.
*New rank-based overlap metric for rasters (RR) that estimates the probability that a pair of rasters agree in the relative ranking of any two patches of habitat (Warren and Seifert 2011).
*Addition of RR metric to hypothesis tests.
The manual has been expanded considerably, including some basic troubleshooting FAQs.
Subscribe to:
Post Comments (Atom)
Hi Dan, Excited to use this--I can't seem to get the program to run correctly though. It keeps giving me the error "no config file found" and the gui that launches only says "Welcome to ENMTools" with no buttons or anything. I saw in an earlier post that you asked if the person had the config file, and I don't...the only files that came with the download are 1.3_OSX.exe, 1.3.exe, _1.3.pl, the manual and a sample data file. This is on Mac OSX 10.6.8... Thanks for any help
ReplyDeleteSo you don't even see a menu bar up at the top?
ReplyDeleteoops, got it, nevermind. thanks!
ReplyDeleteGreat, glad to hear it!
ReplyDeletePossibly a question displaying my ignorance (easily done), but I am wanting to run ENMTools on a remote high performance cluster, the easiest method of doing which is via command line/scripts (ie not via the GUI). I can't find any reference to how do-able this is or how to configure ENMTools in this way (setting reps, ribbon widths, output locations etc). Would I be better off just being less lazy and learning to use a remote GUI viewer (which is an option on this system)??
ReplyDeleteCheers!
Bort
The easiest option is probably to use the method here:
ReplyDeletehttp://enmtools.blogspot.com/2009/10/how-to-run-enmtools-tests-on-cluster-or.html
Basically you use ENMTools to construct your replicates, and then submit those to Maxent on the cluster however you would normally do that. Then you can use ENMTools' scripting interface to get measurements of overlap for the hypothesis tests.
Thanks Dan, I hadn't conceptualised breaking it down like that, but makes sense, will try.
ReplyDeleteps, how easily could the range/ribbon breaking test be modified to include a sliding window option (ie where alternative breaks are along a lat/long axis, not randomised in orientation)? For me this is particularly of interest where there is a strong latitudinal gradient such that one does not expect a large proportion of the randomisations to be close to similar, the question being are breaks on a similar orientation any better or worse...
Another modification would be stepwise alteration of the width of the ribbon only, to see if the bounds of the region can be identified.
There would be problems maintaining sample sizes under these tests I imagine, but if that could be negated... thoughts?
Cheers!
Bort
Robert Anderson and I have been talking about doing something similar to the former idea, but I haven't really considered the latter before. You could do both fairly easily either manually or using the scripting interface. If you decide to give one or both of those a shot, I'd be really interested to see how it works out!
ReplyDeleteI have one particular data set that when I run niche identity, it cranks through to the end, puts out files at each step, but the final summary is empty. Other data sets/environ layers work. And, this combo works with maxent alone.
ReplyDeleteAny simple fixes? thanks Bob Zink, U MN zinkx003@umn.edu
Hi Bob! The quickest fix is to just measure the overlaps yourself using the scripting interface. There's a post about how to do this on the second page of the blog, here:
ReplyDeletehttp://enmtools.blogspot.com/2009/10/how-to-run-enmtools-tests-on-cluster-or.html
Let me know if you run into any problems. Cheers!
Hey! Congrats for the new release. I was trying to play around with your sample data and reading instructions (am a newbie myself).
ReplyDeleteIn order to compute, for instance, niche breath or overlap...do we need to run maxent separately beforehand. I know there is an option for the identity, etc. test in which it runs maxent internally, but the outputs do not give you the asc probabilities or binary maps to perform other tests....or I am not understanding anything?? Thank you in advance!
Pep
PS: a quick tutorial with your sample data would be very useful...
You do need to run Maxent separately in order to calculate breadth and overlap.
ReplyDeleteHi, Dan! Thanks for the tools; they're exactly what I needed and I'm slowly learning how to use them. Quick question about the correlation function, to identify environmental variables that can be eliminated. What is the downside of just leaving all of the variables in?
ReplyDeleteHi Diane! There are several reasons that people prefer models with fewer predictor variables, from simple parsimony to reducing spatial autocorrelation between predictors. In practical terms, some coauthors and I will demonstrate in a forthcoming manuscript that reducing the number of predictor variables tends to produce models that have a less smooth distribution of suitability scores across the landscape, i.e., the difference between suitable and unsuitable areas is greater. Whether that means the model itself is better is an open question, but the phenomenon is pretty unmistakable in our data.
ReplyDelete