I have a binary data (1,0) to calculate spatial distribution of soil bacteria. I have to use indicator kriging (presence/absence) and fitting model of semivariogram. Then, which is the model semivariogram for binary data?
Use whichever semivariogram works with your data. The problem with indicator kriging can be that because your data is 0s and 1s that the differences that are plotted on the Y axis of a variogram cloud are all 0 or 0.5 - but once you've got enough of those and binned them you should get (for a large number of points) a smooth variogram.