Strange prediction

[Champ]

I modeled an aquatic invasive plant using global occurrence data and bioclimatic (+elevation) variables. I used the mean of two GCMs for future prediction.  The prediction I received in the current climate is realistic; however, the projected map (RCP 4.5 for 2050) is strange (image attached) and there is no prediction for my area of interest (0 values). Can somebody tell me why I get this type of weird prediction? I used subsampling with 10 replicates, auto features. 

Many thanks 

[Bede-Fazekas Ákos]

Hello,
are you sure that all the predictors are in similar scale in the future dataset to that of the reference climate dataset? I.e. you did not rescaled (centered/standardized/studentized) the reference data. The unit of measurement does not differ (e.g. K vs. °C), etc.
And what about the CRS/projection of the two datasets? Are these the same?
Best wishes,
Ákos Bede-Fazekas
Hungarian Academy of Sciences

2020.07.20. 3:16 keltezéssel, Champ írta:

I modeled an aquatic invasive plant using global occurrence data and bioclimatic (+elevation) variables. I used the mean of two GCMs for future prediction.  The prediction I received in the current climate is realistic; however, the projected map (RCP 4.5 for 2050) is strange (image attached) and there is no prediction for my area of interest (0 values). Can somebody tell me why I get this type of weird prediction? I used subsampling with 10 replicates, auto features. 

Many thanks 

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/maxent/b2f712ce-0952-4c4f-9101-46c27d8c215bo%40googlegroups.com.

[Milan D]

Maybe more replicates??

This happened to me, but after a bit of faster manipulation, found the target location.

On Sun, Jul 19, 2020, 6:16 PM Champ <gefsecs...@gmail.com> wrote:

I modeled an aquatic invasive plant using global occurrence data and bioclimatic (+elevation) variables. I used the mean of two GCMs for future prediction.  The prediction I received in the current climate is realistic; however, the projected map (RCP 4.5 for 2050) is strange (image attached) and there is no prediction for my area of interest (0 values). Can somebody tell me why I get this type of weird prediction? I used subsampling with 10 replicates, auto features. 

Many thanks 

[Juliana Mota]

Hello, I'm having the same problem with the future projections of some species of insects, I still haven't discovered the error.

[champika kariyawasam]

Hi
Yes, all layers were in the same projection, spatial resolution and extent.

To view this discussion on the web visit https://groups.google.com/d/msgid/maxent/85d70dec-d309-8347-8150-2024ee87398c%40gmail.com.

[Juliana Mota]

Hello, forgive my English but have you tried to see if the pixel type is the same in the current and in the future? I had this difficulty and it was the pixel flot point and pixel of the integer type

[Champ]

Yes , the cell size is similar. 

I ran the models again and which resolved the problem.

thanks

[Rainer]

Hey

You should consider to limit dispersal for global predictions.

Further, you probably have to include at the very least one variable related to surface water (e.g. present or not, proportion of cell covered by, ... etc.)

Have fun!

Cheers
Rainer

https://www.researchgate.net/profile/Rainer_Wunderlich

[pode...@gmail.com]

hi! I have the same problem, what did u do to resolved them??

[Nor Zafirah Ab Lah]

Hi, i get the same problem for my maxent prediction using bioclim variables. I didn't get future prediction value in my AOI. I did tried to convert the value from the integer to float and make sure all the variables in same extent and cell sizes. But still I get the strange prediction. the value output is very2 small. dunno what is wrong.

[Rainer]

Dear Nor Zafirah and others,

if you ONLY use variables such as climate when other variables (in the listed cases possibly presence or area, flow velocity, nutrients, and temperature of WATER) in reality may more effectively constrain the distribution, then predicting to future climate scenarios will often result in drastic reductions of habitat suitability.

Mind that maxent itself has absolutely no clue about either ecology in general or the species you are modeling. Rule of thumb: if something that was around for ten-thousands of years somewhere is predicted to vanish in a few decades due to climate change alone then your model is likely to be inappropriate.

I would like to suggest to try again but include WATER related variables in your models if you model aquatic species. I would be happy to hear if you could solve the problem!

BTW. you can find water variables on earthenv.org!

R