Classification of map

[Kamana Pathak]

Hello all, 

I am doing MaxEnt modeling of Sal, could you please help me with how can we classify the predicted map or how can we calculate the threshold for classification of the map as high, moderate, poor, and unsuitable ??

Sincerely

Kamana

[Amir Sohail Choudhury]

Hello, 

Please check the generated excel file there you find , Minimum training presence, 10percentile training presence, Equal training sensitivity and specificity. You can prepare binary model using any of these (above mentioned) thresholds, in this context you will find lots of literature. 

Further, you can use all these threshold values to produce classified HSM. 

Or Simply you can normalize your Suitable map to 1-100 percentage scale and then classify equally in any GIS software. 

--
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/461cadf3-175c-43e0-8c4e-115cedf81879n%40googlegroups.com.

[Belay Ejigu]

Dear Amir Sohail Choudhury,

Thanks a lot for your help to others!. Incredible!.

Please can say something on how maxent model calculates the probability values of covariates at each grid cell.

Thanks a lot in advance

Stay safe and blessed!

To view this discussion on the web visit https://groups.google.com/d/msgid/maxent/CALnNaESMAOCaQ05Odgq72nNMSxjHA%2B0TY9op9k78yBaAO84ePA%40mail.gmail.com.

[Kamana Pathak]

Amir Sohail Choudhury, 

At first, thank you for your response but could you please explain the process in detail about this statement "Simply you can normalize your Suitable map to 1-100percentage scale and then classify equally in any GIS software" as I am a new learner., also is it a scientific method to classify? And I want to categorize the suitability into 4 classes so the binary method would not work for me as far as I know. Thank you!!

[João Victor M. de Oliveira]

Hello, Kamana!

I have the same issue! I need to categorize my models into 4 classifications, but I don't have any references to do it. Did you get any answers after that?

Att.,
João.

[arjun nepal]

Hi Victor and Kamana

I hope this paper helps you. i referred to this a long time ago. Please check if it is relevant. 

https://www.cambridge.org/core/journals/journal-of-agricultural-science/article/prediction-of-bhutans-ecological-distribution-of-rice-oryza-sativa-l-under-the-impact-of-climate-change-through-maximum-entropy-modelling/EA4ADCA58B37EBF36EEA5A104B3DB987

Best wishes with your work 

To view this discussion on the web visit https://groups.google.com/d/msgid/maxent/4840373e-1532-4a01-99ac-7bdcc22b4a6fn%40googlegroups.com.

--

Arjun Nepal (Mr.)

Adjunct Lecturer/Research Student

Royal University of Bhutan 

College of Natural Resources, Punakha

Mobile no: +975_17553882

Alternative Email: nepaala...@yahoo.com

[Areej]

Hi Kamana

If you have ArcGIS you can use the Reclassification tool as it allows several classification techniques. The most common one is to classify equally and I saw some papers using the natural breaks classification. For the equal classification, many papers follow (0-0.2 not suitable, 0.2-0.4 least/marginal, 0.4-0.6 moderate, 0.6-1.0 highly suitable/optimal). If you will publish a paper out of your work, some journals will be ok with that while others might ask for an ecological explanation for the classification especially that you have specified the suitability degree, so you might put the range of numbers for each category/color instead of words and keep the legend title as something like relative habitat suitability or suitability ranks. 

For the natural breaks method (aka Jenks natural breaks), you can specify your desired number of classes for example 4. It works by creating "natural breaks that are inherent in the data itself where it reduces the variance within classes and maximizes the variance between classes" (quoted from the gibbon paper in the attachments). I personally used this method in some studies and I found the classification satisfying from visualization and ecological perspectives (where I explore any possible relation between the natural thresholds and the environmental variables that are strongly affecting my species distribution). Some studies made an extra effort and conducted a statistical method in R to help them determine the number of classes instead of determining it themselves -just to be more valid- So if you are good with R you can try that too. 

I hope my answer is useful