These things depend on timelike foliations of the spacetime.
Consider some timelike normal $n_{a}$ such that $n_{a}n^{a} = -1$ and $n_{a} \propto \nabla_{a}\tau$ for some function $\tau$ that gives the foliation.
You wish to ask whether some closed surface is a trapped surface. In that case, within the three slice, the intersection of the trapped surface and your leaf of constant $\tau$ will have a spacelike normal $s_{a}$ within the slice and associated 2-metric${}^{1}$ $q_{ab} = g_{ab} + n_{a}n_{b} - s_{a}s_{b}$, and the 2-surface will have the outgoing null normal $\ell^{a} = c\left(n^{a} + s^{a}\right)$ for any constant $c$. Then, any surface that satisfies:
$$\ell^{a}\nabla_{a}\left(q^{ab}\nabla_{a}\ell_{b}\right) = 0$$
will be a trapped surface, which captures the idea that the stack of surfaces on different leaves of $\tau$ will have $\ell$ as a Killing vector, which, in turn, captures the idea that that is the surface where "outgoing rays are frozen on the horizon"
Then, knowing all of this, it's just a matter of computing it yourself. But the definition does require that you either define a timelike foliation, or work out the somewhat dark art of null intrinsic geometry.
${}^{1}$technically, this is the pullback of the intrinsic 2-metric onto the 4-space