Competition sheds light on approximation methods for large spatial datasets