EVENT-BASED RAINFALL-RUNOFF MODELING USING HYBRID APPROACHES

Abstract

The correct estimation of the surface runoff from a catchment after the occurrence of rainfall is required for the planning, operation, design, development, and management of water resources. The behavior of surface runoff is affected by various hydrological and climatic factors such as soil characteristics, antecedent precipitation, topography, land use and land cover, drainage density, rainfall pattern, etc. The best runoff estimation comes from water level observation made at a gauging station which is converted to flow estimates by rating curve. This is quite tedious work and increases estimation costs. The estimation of direct runoff from a basin using observed data can be done by empirical and statistical techniques, and more commonly using rainfallrunoff models. The modeling approach depends on the tools and the skill availability, money and time available, and purpose of the modeling. If the number of parameters increases in a rainfall-runoff model, it will increase difficulty in parameter identification. In such a case, it becomes very difficult to find a better-calibrated value for unidentifiable parameters. So, it is necessary and better to formulate or develop models that are parametrically efficient. The estimation of runoff depends on the conceptual and physical basis of the model and the quality of the measured data. For individual event-based modeling, rainfall-runoff models are required to solve a wide variety of water resources problems. In event-based rainfall-runoff modeling which simulates runoff without acknowledging base flow, many models exist. Under the law of parsimony, the hydrologists prefer a rainfall-runoff model that requires the least input parameters with better runoff simulation. Among rainfall-runoff models, the soil conservation servicescurve number (SCS-CN) method (presently also known as the Natural Resources Conservation Service curve number (NRCS-CN) method) is the most commonly and widely used method across the globe and there exists ample documentation in its usage. This method is the most frequently used method for runoff estimation because of its simplicity and the reasonable accuracy. The model requires watershed and rainfall data which are easily available and can be applied to large watersheds with different land uses. Due to its modesty, the model can be used as a component by various hydrological computer-based models and has been the focus of much discussion among hydrological scientists for the past five decades. This method converts rainfall into runoff using a curve number which is derived from 5-day antecedent rainfall, soil, and watershed characteristics. For calculating event-based runoff, this model gains preference over other models because of many reasons. The method converts event rainfall to direct surface runoff using curve number, which is derived from five-days antecedent rainfall and watershed characteristics. Many researchers have diagnosed several structural inconsistencies in this model and developed SCS-CN based more efficient models. These formulated versions of the SCS-CN model become complex and require more parameters for estimating runoff. Despite many efforts, some of the basic issues still remain ignored in these models. On the basis of comprehensive study, many issues have been addressed in this research work that had not been discussed in the past and have led to various hybrid forms of SCS-CN based model. This study has been carried out on 114 US watersheds and considers only significant runoff producing events, for which the runoff coefficient (C) value is greater than 0.12 to minimize the biasing effect. The parameters of different versions of hybrid models have been computed using the iterative nonlinear least-squares fitting technique and minimize the sum of the squared difference between computed and observed runoff.