Small Scale Variation

The stationary small-scale variation is modelled as a multivariate Normal noise process:

$$\vec{q}\sim MVN(0,\vec{\Sigma})$$ (3)

The covariance matrix $\vec{\Sigma}$ will be of size $NT \times NT$. This is a large covariance matrix (e.g. $NT=10000\times100=10^6$), and hence modelling it will require some simplification to make it computationally feasible.