Interpreting maps from Maxent to assess relationship between species distribution and climate change

[U.R.]

Hello everyone,

I'm working with Maxent to assess the possible relationship between distribution of a species and climate change. I obtained three asc maps from Maxent for three different scenarios in which I have used several variables but I have only changed temperature and precipitation in each case. I was expecting to see a trend across the time, counting the number of pixels that are present in specific ranks (for example between 0.5 and 0.75, and 0.75 -1). I wanted to assess if the habitat suitability is changing with climate change.

The problem that i found is that every time that I run the analysis, Maxent gives me different values for the ranks that I have established. For example from 0.75 to 1, the number of pixels can vary from 400 to 700. I have tried to increase the number of replicates but I see that I continue having the same problem.

How can i know with which map should I interpret the results? Is there another way to assess if suitability is increasing or decreasing through time? I have made the maps for the different scenarios separately. Is there a way to make the analysis as a whole?.

Thank you very much!

[Jessica Beckham]

I have also had this problem. What is your max number of replicates? At 100 replicates, though my current distributions are relatively similar, I still have differing numbers of pixels for different classes. Thanks for posting and I look forward to hearing what others have done!

[U.R.]

Hi Jessica, it's good to see that I'm not the only one with this problem. I have tried 10, 30 and 100 replicates and anyway, results are slightly different each run and it's impossible to see a trend.

I think that in my case maybe it is because my species is fairly widespread and as consecuence my models are not too good (they range between 0.55 and 0.8 AUC). I had to discard too many variables that were highly correlated or had a negative Jackknife test gain, so in at least two of my models I'm working only with two variables and only one of them is climatic :(.

Maybe I have been too hard by removing variables but when I include some of those variables the AUC value goes down and the Jackknife test gain is fairly negative, so I should remove them, right?

Greetings

[Sanjo Jose]

Hi,

I had done some work with species distribution and climate change. I think its better to pick the map of the maxent run which have lower SD and high AUC value. Also check for the omission and ROC curves. High replication could produce more reliable model.

Regards.

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxent+unsubscribe@googlegroups.com.
To post to this group, send email to max...@googlegroups.com.
Visit this group at https://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

--

Sanjo Jose V,

Academy of Climate Change Education and Research,

Kerala Agricultural University,

Thrissur, Kerala,

India-680656.

[U.R.]

Thanks Sanjo. 

Do you think that is correct to use only one climatic variable (i.e. temperature and not precipitation)? If I add another temperature variables or precipitacion, the model gets negative Jakknife values and AUC goes down too. But I don't know if it is usual to work with just one climatic variable. Should I include another one to the model despite being worse?.

El lunes, 14 de noviembre de 2016, 16:55:27 (UTC+1), Sanjo Jose escribió:

Hi,

I had done some work with species distribution and climate change. I think its better to pick the map of the maxent run which have lower SD and high AUC value. Also check for the omission and ROC curves. High replication could produce more reliable model.

Regards.

On Mon, Nov 14, 2016 at 6:55 PM, U.R. <blog...@gmail.com> wrote:

Hi Jessica, it's good to see that I'm not the only one with this problem. I have tried 10, 30 and 100 replicates and anyway, results are slightly different each run and it's impossible to see a trend.

I think that in my case maybe it is because my species is fairly widespread and as consecuence my models are not too good (they range between 0.55 and 0.8 AUC). I had to discard too many variables that were highly correlated or had a negative Jackknife test gain, so in at least two of my models I'm working only with two variables and only one of them is climatic :(.

Maybe I have been too hard by removing variables but when I include some of those variables the AUC value goes down and the Jackknife test gain is fairly negative, so I should remove them, right?

Greetings

El lunes, 14 de noviembre de 2016, 13:22:31 (UTC+1), Jessica Beckham escribió:

I have also had this problem. What is your max number of replicates? At 100 replicates, though my current distributions are relatively similar, I still have differing numbers of pixels for different classes. Thanks for posting and I look forward to hearing what others have done!

On Thursday, November 10, 2016 at 8:29:22 AM UTC-6, U.R. wrote:

Hello everyone,

I'm working with Maxent to assess the possible relationship between distribution of a species and climate change. I obtained three asc maps from Maxent for three different scenarios in which I have used several variables but I have only changed temperature and precipitation in each case. I was expecting to see a trend across the time, counting the number of pixels that are present in specific ranks (for example between 0.5 and 0.75, and 0.75 -1). I wanted to assess if the habitat suitability is changing with climate change.

The problem that i found is that every time that I run the analysis, Maxent gives me different values for the ranks that I have established. For example from 0.75 to 1, the number of pixels can vary from 400 to 700. I have tried to increase the number of replicates but I see that I continue having the same problem.

How can i know with which map should I interpret the results? Is there another way to assess if suitability is increasing or decreasing through time? I have made the maps for the different scenarios separately. Is there a way to make the analysis as a whole?.

Thank you very much!

--
You received this message because you are subscribed to the Google Groups "Maxent" group.

To unsubscribe from this group and stop receiving emails from it, send an email to maxent+un...@googlegroups.com.

To post to this group, send email to max...@googlegroups.com.
Visit this group at https://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

[U.R.]

In fact I have removed all the variables with negative values in Jacknife test gain except one which had the negative value closest to 0. Do you think is that correct?