SCS-CN BASED LONG TERM HYDROLOGIC SIMULATION

Abstract

The hydrologic cycle is a conceptual model that describes the storage and movement of water between the biosphere, atmosphere, lithosphere, and hydrosphere. Continuous accounting of this movement of water involves consideration of precipitation, surface loss, infiltration, and surface transport processes as a part of surface flow process and evapotranspiration, soil moisture redistribution, and ground water transport as a part ofsub surface flow. The Soil Conservation Service Curve Number (SCS-CN) method which is based on proportionality and water balance hypotheses has been widely used in the past to model the surface flow component via direct surface runoff of hydrologic cycle. Though the SCS-CN method was initially developed for computation ofsurface runoff from isolated storm events, it has been successfully employed in several long term hydrologic simulation (LTHS) models by accounting for the soil moisture status at previous time steps. Of late, Michel et al. (2005) critically reviewed the soil moisture accounting (SMA) procedure lying behind the original SCS-CN method and proposed a procedure more consistent from SMA view point. They pointed out several structural inconsistencies in the existing SCS-CN methodology, and in treatment of the antecedent soil moisture condition (AMC). A rigorous scrutiny of the procedure proposed by Michel et al. reveals a need for refinement particularly in defining the initial moisture level (V0) and the proposed SMA procedure. Hence, in this study, this SMA procedure is modified in different ways by re-casting and re-conceptualization; by incorporating variation ofdaily CN based on antecedent moisture amount (AM) and moisture availability prior to rainfall to avoid unrealistic sudden jump in computation ofV0 and further quantum jump in computation of direct runoff to make the SMA procedure amenable to continuous hydrologic simulation. In the existing SMA based long-term simulation models, the SMA concept based on SCS-CN method is used for computation of surface flow only and the potential of its extension to sub-surface flow computation has yet to be explored. There are numerous models, and programmes existing in literature for modeling sub-surface flow. For examples, Water and Agrochemicals in the Soil, Crop and Vadose Environment (WAVE) (Vanclooster et al., 1996), Soil Water Assessment Programme (SWAP) (Van Dam et al., 1997), DRAINMOD (Skaggs, 1980), Yuan et al. (2001) model, Base flow separation techniques and programmes, etc. Among these, Yuan et al. (2001) model uses the SCS-CN method to model sub-surface flow by modifying it through analogy for estimation of sub-surface drainage flow from rainfall. Their conceptual frame-work is further modified in this study to simulate the sub-surface flow components. In addition, the stores which are common component of rainfall-runoff model are used to route surface and sub-surface flow (for example, Putty and Prasad's (2000) two-stores SAHYADRI model, Mishra and Singh (2004a) versatile twostores model, and Geetha et al. (2007) LCRR three-stores model, etc.). The single linear reservoir (SLR) is used to route surface flow and the exponential store as described by Putty and Prasad for sub-surface flow. In the present study, the above three concepts of Michel et al., Yuan et al., and Putty and Prasad are amalgamated, and four new/modified LTHS models proposed to carry out the long term hydrologic simulation. The first model (Model-I) is designated as LTHS MICHEL I and it uses the expressions proposed by Michel et al. (2005) for soil moisture store level prior to rainfall occurrence (V0) for various antecedent moisture conditions (AMCs), viaAMC I, II, and III based on the antecedent rainfall to compute direct runoff (RO) and sub-surface flow computation based on the conceptual behavior of soil moisture store (SMS) and ground water store (GWS) as given by Putty and Prasad (2000). The second model (Model-II), designated as LTHS MICHEL II, is formulated based on AM due to 5 days antecedent rainfall (prior to the storm) to avoid sudden jump in CN and further quantum jump in V0 and, in turn, the modification in the computation of RO. In the previous two LTHS models, V0 plays a vital role in the improvement of SMA procedure via improvement in surface flow components. Despite this improvement, these models, however, do not contain any expression for V0. Therefore, in the third model (Model-Ill), designated as LTHS ASMA I, where ASMA stands for Advance Soil Moisture Accounting, an expression for V0 is proposed based on the value of pre-antecedent moisture level before the onset of rainfall (V0o). This forms the advanced soil moisture accounting (ASMA) procedure. In both, LTHS MICHEL II and LTHS ASMA I models, the sub-surface flow components are modeled similar to LTHS MICHEL I model. The fourth model (Model-IV), designated as LTHS ASMA II, is similar to Model-Ill (LTHS ASMA I) for surface runoff computation but differs in computation of sub-surface drainage flow. In this model, apart from use of ASMA procedure, an expression for sub-surface drainage flow is developed by modifying the concept of Yuan et al. In all these models, the total stream flow from watershed is quantified by incorporating sub-modules for surface and sub-surface flow components such as surface runoff, evapotranspiration, sub-surface drainage, lateral flow, percolation, base flow, and deep percolation. These models were tested for their performance using daily hydro-meteorological annual and seasonal (monsoon season from June to November) data series of 17 watersheds of various sizes/shapes and physical characteristics, and located in various agro-climatic zones of India. The available data was split into two groups, one for calibration and the other for validation using non-linear Marquardt (1963) algorithm by minimizing the sum ofsquares of the errors between observed and model computed runoff. The performance of these models was evaluated using different statistical criteria such as Nash-Sutcliffe efficiency (NSE), standard error (SE), and relative error (RE). For performance evaluation, the study watersheds were grouped into three categories depending on the runoff coefficient (C) (Gan et al., 1997, Geetha et al., 2008)) as dry (C < 0.36), average (0.36 <C < 0.65), and wet (C > 0.65). It is found that all the proposed models perform very good on the data ofwet watersheds, good to satisfactory on normally dry watersheds, and poorly in most dry watersheds and are capable of capturing the variability of curve numbers representing hydrological characteristics of the complex watersheds. Among the proposed models, the LTHS ASMA II model produces better results than others, followed by LTHS ASMA I, LTHS MICHEL II, and LTHS MICHEL I models. The existing lumped continuous SCS-CN based long term simulation models such as Michel et al. (2005) and Geetha et al. (2008) models were also tested on annual data series of study watersheds. When compared, the best performing model LTHS ASMA II also worked betterthan the existing models. The performance of the proposed LTHS models is also analyzed on the basis of subjective assessment through visual comparison between the observed and simulated runoff. The results were plotted for calibration and validation for all the models for all study watersheds, showing a close match between simulated and observed stream flows for most of the watersheds except for some deviation in simulating peak flows. Since the models also help understand and identify various processes/components involved in runoff generating in mechanism, these components are quantified to compare their significance in various high/low runoff potential watersheds. For example, the base flow is more significant in high runoff producing coastal watersheds than low runoff producing watersheds in the study, while the evapotranspiration shows reverse trends. All other components except deep percolation show linearly increasing trends with runoff coefficient and are more significant/dominant in high runoff producing watersheds. Deep percolation is dormant in high runoff producing watersheds. The sensitivity analysis of the above better performing 15- parameter LTHS ASMA II model indicates that the coefficient (y), related with pre-antecedent soil moisture store level (Voo), most sensitive parameter among all other parameters, while the curve numbers at the starting day of simulation for surface flow (CN0) and sub-surface drainage flow (CNd0) are also highly sensitive, in addition to the parameters related with soil characteristics, wilting point (0W) and field capacity (y/f). An effort was also made to minimize the number of parameters by fixing the insensitive or less sensitive parameters based on statistical analysis and this, in turn, resulted in formulation of a nine-parameter simplified LTHS ASMA SIMP model. The LTHS ASMA SIMP model performed as well as did ASMA LTHS II model, but with little reduction in model efficiency. For pragmatic application, model parameters are also related with measurable physical characteristics of the watersheds using step-wise backward elimination procedure via p-value of F-statistic of multiple regression analysis. In mostcases, the parameters exhibited a good relationship. « Keywords: Antecedent moisture condition, curve number, initial soil moisture level, long term hydrologic simulation models, SCS-CN, soil moisture accounting procedure, stream flow, watershed.