I assume that "K-2 parameter model" is Kimura's 2 parameter model (K80):

http://en.wikipedia.org/wiki/Models_of_DNA_evolution#K80_model_.28Kimura.2C_1980.29.5B2.5D

Models of evolution, like K2, can be used in two ways in phylogenetics:

1. most often we use them in model based methods, e.g. building maximum likelihood trees.

2. You can use K2 model to calculate pairwise distances between nucleotide sequences. Thus, you can transform a DNA multiple sequence alignment to a distance matrix, which can be used for distance based tree building, like NJ.

Both ways are used, but my personal opinion is that in most cases model based methods over-perform distance based methods, but there are scenarios when the usage of distance based methods is justifiable.

I suggest you to take a look at this article. It explains the reason:

Srivathsan, A., & Meier, R. (2012). On the inappropriate use of Kimura‐2‐parameter (K2P) divergences in the DNA‐barcoding literature. Cladistics, 28(2), 190-194.

Edit: I just realized that the question was from last April, even if it was suggested by ResearchGate. I apologize for the necropost.

http://onlinelibrary.wiley.com/doi/10.1111/j.1096-0031.2011.00370.x/full

Because It incorporates hidden changes/mutations among long diverged sequences.

The Kimura two parameter model will be calculate the number of differences accumulated between two sequences since their last common ancestor. Furthermore estimate the genetic distances among the sequences of alignment MEGA will build a matrix with the pairwise distances. For phylogenetic analyses the selection of the best model can be done using another programs.