Black is closing in on White’s king. A key feature of the position is the kernel of a discovery on the diagonal leading to g1; keep it in mind as you go through your thought experiments. The natural moves to consider first are any checks. Black has two: Rh2, which loses the rook without compensation; and Bg2, which is a lot more interesting because the bishop enjoys protection from the rook on f2, forcing White’s king to move. Where would it go? It would have two options: h2 and g1. Consider them both:
(a) If White plays Kh2, notice that the king would have made itself the target of another discovered check—using the kernel Black created when he moved his bishop in front of his rook. Black would be able to unmask a check by moving his bishop from g2, and so can think about targets the bishop might reach in two moves. The natural candidate is White’s queen. Hence Bf3+; and after White moves his king, Black has BxQ.
(b) Now suppose that in reply to Bg2 White plays not Kh2 but Kg1. Again the king would have walked onto the target square of a discovered check, this time the one whose kernel was in place from the beginning. It would be tempting now to play Rf5+, unmasking check and then winning White’s queen in exchange for a bishop after White plays KxBg2. But don’t jump to conclusions; there may be a less costly way to achieve the result. It would be preferable to first force White's king to h2 and then use the discovered check described in the previous paragraph. How? With another check: after White plays Kg1, Black replies Rxe2+. The bishop on e3 now gives check and the bishop on g2 still has protection. The king’s only legal escape is to h2—and then play proceeds as described a moment ago, with Bf3+ discovering check and then winning White’s queen.