How do you model theta on expiration week, obviously there is not much interest left so does it become purely model based at some point? thanks hope I am clear.

Think about it this way you have 1min to expiry, whats theta zero? Nothing can happen right its either in the money or out. Until the underlying jumps 8% in the last 30s. It is precisely in this realm that Black Scholes fails completely the distribution of returns is extremely far from normal on these time scales. Remember volatility and time are the same thing as time gets shorter the volatility of volatility becomes very very important. Not so much a case of modeling theta but modeling the behavior of the underlying asset. If you watch a few close outs you will notice that the underlying tends to become very calm, may be an initial jump at the start of the day but then it will flat line with low volume until the very last moment.

I agree but.... I would only say this... why not add delta, vega and rho since all the risks are related.

rho is basically zero near expiration and vega is basically meaningless. Delta switches between 0 and 1.