limiting factor map for replicated/averaged model

[Areej]

Hi all

I found that maxent produces the limiting factor map (LFM) if the model was not replicated (rep=1), in other words the code needs a single lambdas file. Since LFM is based on lambdas file and environmental variables and considering that we have multiple lambdas files for the replicated model, I'm thinking that  it would be possible to obtain the LFM if we calculated the average for all lambdas file and obtained a single file. 

I have actually tried that and was able to obtain the LFM for a replicated model. When I averaged lambdas files, I compared the resulted entropy and background data values with the one reported in the results.csv file and they were the same meaning, as I assume, that I obtained the correct lambdas file for my replicated/averaged model. 

I searched literature to support this practice but couldn't find a paper that did it clearly. Some papers actually reported the MoD map (most dissimilar variable) as the LFM which is wrong according to my understanding as they are different (MoD depends on MESS which in turn depends on env variables solely, while LFM uses the lambdas values for each feature used in the model). 

Does anyone have an opinion on the validity of this practice? and if possible, a supporting paper?

Best