if it's possible it could be good to get and to study the noise samples. After that one can get estimations of the main statistical properties of the noise and build its model. At this stage it's could be usefull to define the goal of such modeling.
Best regards.
N.B. What do you mean under the type of noise? Additive, multiplicative, reverberation noises, broad- or narrowband or periodic noises or something else?
If it's additive, then measure it over a long enough time to cover the bandwidth of interest. Do it several times. Then study the noise. Otherwise, most cheat and like to addwhite noise to a paper because it drops out in the math. One of my pet peeves.
When modeling noise you must have in mind two features: probability density function and correlation function.
The most common realization of noise is Additive White Gaussian Noise (AWGN) which:
1. Adds to the signal [additive]
2. Have a constant power density functions (transform of the correlation function) [white]
3. Have a Gaussian or Normal probability density function [Gaussian]
Common alternatives for the Gaussian probability density function are the exponential, Rayleigh, Weibull and log-normal functions. It depends on your field of application.
Good Luck!
Article Modelación de las Distribuciones Rayleigh y Exponencial en M...