The equation listed in the figure appears so uneasy to handle. That is true since it has cyclic-like behavior with three levels.
However, I've tried simple math, and came up with another, equivalent, yet much clearer formula.
aZ(t-1) = Z1,
bZ(t-2) = Z2,
1./((2Pi)^0.5*d) = A -- The amplitude of et.
2*d^2 = B
and let go a little further and write:
So, your equation will reduce to (after eliminating et from both equations):
with K being the main parameter. This, I think, can be handled easier than the original formula, and it is valid only if it DOES NOT alter the behavior of Z(t-1) and Z(t-2), and that I don't know much about.