Claim Assessment of a Rainfall Runoff Model with Bootstrap

Abstract

Since the inception in 1954, researchers started to scrutinise the United States Department of Agriculture (USDA) Soil Conservation Services (SCS) rainfall runoff model with different field data after the model produced inconsistent runoff prediction results throughout the world. This paper re-assessed two key hypotheses used by SCS where Ia = 0.2S and λ = 0.2 as a constant. The 112 original SCS data points were used to re-determine the correlation between Ia and S with Bootstrapping, BCa procedure. Both key hypotheses of SCS were proven to be statistical in-significant. Inferential statistics deduced that Ia ≠ 0.2S while λ is neither equal to 0.2 nor a constant at alpha = 0.01 level. Both hypotheses are not even applicable to the original dataset used by then SCS to formulate the rainfall runoff model. Ia = 0.112S fitted SCS original data points better at alpha = 0.01 level. The 1954 SCS proposal of Ia = 0.2S and λ = 0.2 committed type II error as pertain to its own dataset. Therefore, SCS rainfall runoff model cannot be blindly adopted. Practitioners of this model are encouraged to validate and derive regional specific relationship between Ia and S.