Explanation

Structures are safe against undermining when the actual exit gradient is less than the critical exit gradient. The critical value is the safety threshold.