Research papersImprovement of the integration of Soil Moisture Accounting into the NRCS-CN model

- The relationship of rainfall variability, initial soil moisture and runoff is examined.
- A new approach based on the Soil Moisture Accounting procedure is proposed.
- The advantages of the proposed model and the influence of scale effects are analysed.

Rainfall-runoff quantification is one of the most important tasks in both engineering and watershed management as it allows the identification, forecast and explanation of the watershed response. This non-linear process depends on the watershed antecedent conditions, which are commonly related to the initial soil moisture content. Although several studies have highlighted the relevance of soil moisture measures to improve flood modelling, the discussion is still open in the literature about the approach to use in lumped model. The integration of these previous conditions in the widely used rainfall-runoff models NRCS-CN (e.g. National Resources Conservation Service - Curve Number model) could be handled in two ways: using the Antecedent Precipitation Index (API) concept to modify the model parameter; or alternatively, using a Soil Moisture Accounting (SMA) procedure into the NRCS-CN, being the soil moisture a state variable. For this second option, the state variable does not have a direct physical representation. This make difficult the estimation of the initial soil moisture store level. This paper presents a new formulation that overcomes such issue, the rainfall-runoff model called RSSa. Its suitability is evaluated by comparing the RSSa model with the original NRCS-CN model and alternatives SMA procedures in 12 watersheds located in six different countries, with different climatic conditions, from Mediterranean to Semi-arid regions. The analysis shows that the new model, RSSa, performs better when compared with previously proposed CN-based models. Finally, an assessment is made of the influence of the soil moisture parameter for each watershed and the relative weight of scale effects over model parameterization.