[Biomod-commits] Biomod-commits Digest, Vol 34, Issue 15

Mon Feb 20 14:47:43 CET 2012

Hello Damien and Andreas --

I just wanted to comment on Andreas' question #2, "Is the best model for my dataset also the best for a new dataset?"

This is not necessarily the case, and (in my opinion) is one of the main arguments for using an ensemble approach to forecasting species distributions onto new time period or locations.  A recent reference on this topic is:

Heikkinen RK, Marmion M, Luoto M (2011) Does the interpolation accuracy of species distribution models come at the expense of transferability? Ecography:1–13. DOI: 10.1111/j.1600-0587.2011.06999.x

Cheers,
Brenna

~~~~~~
Brenna Forester
PhD Student
Landscape Ecology Lab
Duke University
Durham, NC, US

________________________________________
From: biomod-commits-bounces at r-forge.wu-wien.ac.at [biomod-commits-bounces at r-forge.wu-wien.ac.at] on behalf of Damien Georges [damien.georges2 at gmail.com]
Sent: Monday, February 20, 2012 5:16 AM
To: biomod-commits at r-forge.wu-wien.ac.at
Subject: Re: [Biomod-commits] Biomod-commits Digest, Vol 34, Issue 15

Hi Andreas,

Here some comments and partial answer to your questions..

1) How can I obtain info about the predictive performance of a model on my already known localities? There are three functions, Models(), CurrentPred() and PredictionBestModel(), whose differences I find hard to understand.

The main function of BIOMOD is Models(), you have to run its function which all other functions needs to be computed.
Models run all selected models('GLM', 'GAM', 'RF'....) with data given to InitialState() function. It also does evaluation of each model predictive performance according to an accurancy metric (Roc, TSS or Kappa). This evaluations are stored in objects in your workspace called "Evaluation.results.Roc", "Evaluation.results.TSS" or "Evaluation.results.Kappa".

CurrentPred() returns the predictions of all selected models using InitialState() input data (considered as current data). It creates objects in the pred/ folder. Objects are a 4 dimensions arrays by species (may be several obects if you have decide to do binary or filtering transformation). All models, all PA selection, and all repetition are stored in this object (dim2, dim3 and dim4 respectively)

PredictionBestModel() does the same but keep only the best model for each Evaluation run and PA selection (so it's a 3D array). The best model is selected according to TSS, Roc or Kappa score. Noting is returne but objects are created in the pred/ folder.

2) Under what criterion can someone chose the best model for new datasets (e.g. other regions or climate change scenarios)? Is the best model for my dataset also the best for a new dataset?

That effectively the assumption we made.

3) What is the difference between Projections() and Ensemble.Forecasting()? Which of these two functions is suitable for predicting the species' presence in another region??

Projections give predictions of all models run on new environmental dataset.. (future or new zone for instance). It produce a 4D array like describe upper. object are stored in a new folder create when you compute this function.

Ensemble.Forcasting try to give you a consensus modeling projection, that means that every selected model will be contribute (more or less depending on arguments given) to a consensus and supposed better projection. So you get a 3D array containing the "consensus projections" for all evaluation run and all PA selections

4) I have information about my species' presence but not for their absence. Does this still mean that I have to create a presence-absence table and then use pseudo-absences?

Yes because models needs presences and absences to compute..

Hope that helps you,

Cheers.

Damien

On 19/02/2012 12:00, biomod-commits-request at r-forge.wu-wien.ac.at wrote:
> Send Biomod-commits mailing list submissions to
>       biomod-commits at lists.r-forge.r-project.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
>       https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/biomod-commits
>
> or, via email, send a message with subject or body 'help' to
>       biomod-commits-request at lists.r-forge.r-project.org
>
> You can reach the person managing the list at
>       biomod-commits-owner at lists.r-forge.r-project.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Biomod-commits digest..."
>
>
> Today's Topics:
>
>     1. New with BIOMOD;       how can I predict my species distributions
>        in another region? (Andreas Soteriades)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Sat, 18 Feb 2012 10:50:06 -0800 (PST)
> From: Andreas Soteriades<andreassot10 at yahoo.com>
> To: "biomod-commits at lists.r-forge.r-project.org"
>       <biomod-commits at lists.r-forge.r-project.org>
> Subject: [Biomod-commits] New with BIOMOD;    how can I predict my
>       species distributions in another region?
> Message-ID:
>       <1329591006.42296.YahooMailNeo at web121305.mail.ne1.yahoo.com>
> Content-Type: text/plain; charset="iso-8859-1"
>
> Hi,
>
> I am trying to learn BIOMOD by using the sample Sp.Data and by following the instructions in BIOMOD Tutorial, 2012.
>
> What interests me is to predict species distributions in other regions or under climate change scenarios. Unfortunately, I have not really understood what is the procedure to follow; I am afraid that the tutorial can be very confusing for non experts in SDM...?Here are my questions:
>
> 1) How can I obtain info about the predictive performance of a model on my already known localities? There are three functions, Models(), CurrentPred() and PredictionBestModel(), whose differences I find hard to understand.
>
> 2) Under what criterion can someone chose the best model for new datasets (e.g. other regions or climate change scenarios)? Is the best model for my dataset also the best for a new dataset?
>
> 3) What is the difference between Projections() and Ensemble.Forecasting()? Which of these two functions is suitable for predicting the species' presence in another region??
>
> 4) I have information about my species' presence but not for their absence. Does this still mean that I have to create a presence-absence table and then use pseudo-absences?
>
> Cheers,
>
> Andreas
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:<http://lists.r-forge.r-project.org/pipermail/biomod-commits/attachments/20120218/71c7bb87/attachment-0001.html>
>
> ------------------------------
>
> _______________________________________________
> Biomod-commits mailing list
> Biomod-commits at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/biomod-commits
>
> End of Biomod-commits Digest, Vol 34, Issue 15
> **********************************************

_______________________________________________
Biomod-commits mailing list
Biomod-commits at lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/biomod-commits