Why does Sobel use a sum of squares instead of just a sum?

I'd assume it's because we want to compute the magnitude of gradients.