the dismo package for modeling Maxent

[lyza...@gmail.com]

 i want to ask whether the "dismo" package in R for modeling species distribution Maxent NEED the jar programe "maxent.jar". I saw this in the R tutorial. WHY？Why don't I use that software directly?

[Adam Smith]

Hi Lyza,

You can use the software directly; you don't need dismo.  However, the field of niche modeling has advanced quite a lot since the original maxent jar program was written, and it's very hard to do modeling well just using that program. For example, you would probably want to geographically divide your training and test sites from one another to gain a better estimate of how well your model predicts in novel situations. Similarly, you would in many cases want to attempt to correct for bias in collecting which you can do by thinning presences in geographic or environmental space or by using a bias raster or some other means of estimating bias. Finally, you most likely need to do a lot of pre- and post-processiing of the occurrence data, environmental data, and output.  None of this cane be done inn the maxent jar program (or at least very little), but most of it can be done in R.  So using dismo with the maxent jar file is a nice way to keep your workflow all in one place.  The big advantages of using dismo (and R in general) versus using a mix of ArcGIS, teh maxent jar file, and maybe other programs is that you can easily repeat your analysis and tweak it as you go along.

I have personally found doing everything in R is far better than using different programs.  Niche modeling is often a very iterative process... you try one thing, it doesn't work, you try again, it seems OK, but you think you can do better, etc.  Data cleaning and processing are also very iterative--I often have 10-20 sequential versions of my presence data as I get to know it better and find errors (e.g., someone georeferenced a specimen to the zoo it was living in). It's very laborious to recreate the data cleaning process by hand every time I need to do another clean of the data.  R really helps with that since once I write the code it's all automated.

Using R has also been *very* helpful in revising models based on feedback from reviewers.  Sometimes I can show then that I did indeed try something they are suggesting but it didn't work.  And sometiems I just need to adjust a few lines and I can address their concerns.  If I did it all by hand I'd need to do everything again.

This all said using R is hard at first.  It requires a lot of perseverance to learn.  If you want to get into using R with niche modeling I *highly* suggest using the wallace packae by Jamie Kass, Rob Anderson, and others. It's allmost like a mini-GIS program and *much* easier to use than the maxent jar program. When you're done it shows you the R code so you can use that to study.

Best,

Adam

[lyza...@gmail.com]

Thank you for your answer. I also see a R package named maxnet by Steven Phillips. The tutoriaI shows that its structure is the same as for the 'Maxent' Java package. Can i use this package?

在 2018年10月5日星期五 UTC+8上午11:17:06，lyza...@gmail.com写道：

[Adam Smith]

Yes, maxnet is the open-source version of the maxent jar file. It should give you nearly the same answers depending on the settings you use.

Best,
Adam