A theoretical approach for estimating a probability distribution of damage degree of tunnel concrete liners in Hokkaido area is developed. First, a probability density function theoretically derived from a Poisson model describing random deterioration processes of tunnel concrete liners is compared with a frequency distribution of actual damage degree data. Next, in order to improve the approximation in the range of large damage degree, a concept of double-mode model is newly introduced into the Poisson model based upon a result of analysis on physical structure of temporal growth of damage degree. Finally, it is clari.ed that the double-mode Poisson model can give quite accurate estimation for tail properties associated with the probability distribution of damage degree. Further, it is shown that the proposed double-mode model can well reproduce temporal variation of the damage degree distribution.