Try solving for t here, then plugging that into the other equation (the one for y). This way, you're solving directly for s, which is what the question asks for.
One thing though, when you solved for t, it's not the sqrt(g) you're taking, it's the square root of y - y0 - x * tan(a) + x0 * tan(a)) / g
which may or may not be negative in its entirety