forest fire; risk modeling; model optimization; South Korea