DISTRIBUTED MODELLING OF RUNOFF AND SEDIMENT YIELD BASED ON GEOSPATIAL TECHNIQUES

Abstract

Erosion of soil due to rainfall and runoff causes severe problems including reduction of reservoir capacity, deterioration in water quality and loss of soil cover, etc. Many engineering projects meet untimely demise because of increased sediment yield in the watershed due to excessive soil erosion. Hence measurement and modelling the processes of rainfall-runoff-soil erosion and sediment transport are very much required for soil and water conservation planning, control of reservoir sedimentation and for study of transport of pollutants carried by the sediments, etc. Over the past years researchers have developed and improved the techniques for both measurement and modelling the soil erosion and sediment transport processes. Though sophisticated measurement techniques are available, requirement of huge funds, difficulties in data collection, requirement of substantial land area, field personnel and automated equipment often make repeated field studies unfeasible especially in developing countries like India. Faced with these limitations, hydrologic modelling becomes a powerful alternative to quantify storm runoff, soil erosion and sediment yield for design and evaluation of alternative landuse and Best Management Practices (BMPs). Increasing development in technologies like remote sensing and Geographic Information System (GIS) and increasing knowledge on representation of the hydrological processes have collectively led to the development of process-oriented, Physically Based Distributed (PBD) models for a variety of hydrological applications. Though there are many more PBD models available, it is not always clear when and where to use which type of model. Comparisons of some of the models show that no one model works well in every situation of runoff and sediment yield generation in the watersheds. Many of the models are site specific and contain simplifications and assumptions that preclude their use universally. In most of the PBD models available in literature, the interception process is either ignored or not dealt extensively. However, in a real hydrologic system the process of interception may have considerable effect on volume of runoff and soil erosion due to raindrop impact in vegetated watersheds. As the proposed model is intended for vegetated watersheds, a simple two-parameter empirical based interception model has been adopted in the present study. vn In most cases of hydrological significance, use of kinematic wave (KW) approximation of the Saint Venant's equations for describing overland and channel flows is found to produce realistic results. Hence the kinematic approximation has been adopted for use in the present study. The channel flow is a major component of the runoff process in medium to large sized watersheds. However, channel flow component is not considered explicitly in several PBD models. Subsequently, this precludes successful applicability of such models in medium to large sized watersheds. To overcome this limitation, the channel flow process is modelled explicitly in the proposed model. Also there are PBD models having very complex structure and requires use of enormous data on landuse/land cover, management practices, etc. which are not feasible to obtain for most of the scenarios, even in developed countries due to requirement of huge man power and funds. In the present study, an attempt has been made to overcome this limitation to the extent possible by incorporating appropriate model components without compromising the physics of the rainfall-runoff-soil erosion and sediment transport processes. Several PBD models available in literature lack proper formulation for processes related to soil erosion due to raindrop impact, which is inappropriate given the current understanding of soil erosion process. To overcome this deficiency, a suitable formulation has been adopted in the present modelling. As the proposed model is a GIS based spatially distributed one, it necessitates the division of watershed into smaller and relatively homogeneous units for the present modelling. The watershed discretization involves breaking up of the heterogeneous and complex geometry into simpler and relatively homogeneous units while retaining the similitude with the natural watershed. Several schemes for watershed discretization are available at present. The choice of the discretization method is governed by many factors including nature and type of input data, size of the watershed, purpose of modelling, etc. Based on this, it is found that Critical Source Area (CSA) based discretization of watershed to form a cascade of overland planes and channels has been widely used and is quite adaptive method for representing the spatial heterogeneity of the watershed characteristics. Hence, the same discretization scheme has been adopted in the present study using TOpographic PArameteriZation (TOPAZ), a digital landscape analysis tool. Using TOPAZ, the hydrologic computational sequencing is also made for routing the generated surface runoff and eroded soil to the watershed outlet. vm The present study has been taken up with the above background to develop a relatively simple process-oriented PBD mathematical model capable of handling inputs from remote sensing, GIS and from conventional methods, for simulation of runoff and sediment yield during storm events in vegetated watersheds. It is also understood from the available literature that there is a great paucity of distributed data for verifying the distributed model outputs, thus limiting the validation of distributed outputs from the PBD models. However, during the course of this study, extra efforts have been made to collect such data at a manually operated internal gauging station in one of the study watersheds and used for model validations. The model developed in this study is capable of handling distributed information of landuse, soil type, topography, etc. of a watershed and generate runoff and sediment yield estimates in spatial and temporal domains. The forms of soil erosion by water such as rill erosion, interill erosion, gully erosion and channel scour and deposition are not modelled individually. No distinction has been made in the present work between bedload and suspended load transport. Therefore, spatial and temporal variation of sediment yield herein essentially means the spatial and temporal variation of total sediment load carried by the stream during a storm event. MODEL DEVELOPMENT In the present modelling approach, it is assumed that for any rainfall event in a watershed, a part of the rainfall is intercepted by vegetal cover which subsequently evaporates. Some part of the rainfall that reaches the ground surface after interception loss is met, infiltrates into the soil. The remaining part of the rainfall gets transformed as surface runoff. Surface runoff consists of overland flow and channel flow, and is assumed to be dominantly one-dimensional. In addition, the proposed model also simulates the process of soil erosion due to raindrop impact and surface runoff and the subsequent process of sediment transport. All the above processes are modelled in a distributed manner over the watershed elements derived based on the CSA concept. Computations are performed in all the discretized elements as per the hydrologic computational sequence derived using TOPAZ tool for routing the generated surface runoff and eroded soil to the watershed outlet. For computational purpose, the structure of the model is divided into two components, namely, hydrodynamics and soil erosion dynamics. IX Hydrodynamic Modelling The hydrodynamic component of the model is sub-divided into runoff generation and runoff routing phase. In runoff generation phase, interception loss is modelled empirically using the modified form of Merriam (1960). Some part of the rainfall that reaches the surface after interception loss, infiltrate into the soil, which is modelled using Smith and Parlange (1978) infiltration model for estimation of the excess rainfall. In runoff routing phase, the generated surface runoff is modelled using the one-dimensional depth averaged unsteady flow equations (Saint Venant's equations) under a KWapproximation for overland and channelized flow computations. The governing equations of the KW model developed for overland flow and channelized flow are then solved by employing weighted four-point implicit finite difference numerical technique. Soil Erosion Dynamic Modelling The subsequent processes of soil erosion and sediment yield which are dependent on the outputs from the hydrodynamic component are simulated using the basic governing one-dimensional equation for continuity of sediment mass. Auxiliary equations for detachment of soil by raindrop and flow are used in conjunction with sediment transport equations to solve the sediment continuity equation. Numerical solution for the sediment continuity equation also follows exactly the same procedures used to solve the flow continuity equation but in an explicit form. The routing of sediments through overland planes and channels to the watershed outlet produced the total soil sediment yield from a given watershed. MODEL APPLICATION AND RESULTS To test the predictive ability and performance of the proposed model, it has been applied for simulation of different storm events registered in the watersheds of Pathri Rao (India), Ganspoel (Belgium), Wl (USA) and W2 (USA) having varying climates. Data on temporal variation of runoff and sediment yield resulting from few storm events have been compiled for all the four study watersheds. In Pathri Rao watershed, in-situ field measurements have been made by the investigator of the present study for collection of precise data on temporal variation of runoff and sediment yield during different storm events in the year 2005. The rainfall has been measured using a self-recording rain gauge. Runoff have been determined by x manually measuring the flow velocity, flow depth and flow cross-sectional area during the storm events by establishing a gauging station at the watershed outlet and at an internal location as well. Sediment yield at the watershed outlet have been determined by collecting samples of sediment laden flow using bottle samplers during the corresponding storm events. Other data related to watershed characteristics such as area, slope, soil type and landuse, etc. have been derived using topographic maps, remote sensing techniques, field observations and GIS analysis. For the remaining watersheds, all the above datasets have been compiled from various scientific publications and processed in the GIS environment. Before selecting/estimating the values of calibration parameters, the order of sensitivity of the model outputs to some of the model input parameters has been ascertained through sensitivity analysis. In all, data for 19 storm events have been utilized for the four study watersheds and these have been used for calibration and validation of the proposed model. Results of the model application (calibration and validation) on the study watersheds indicate that the proposed PBD model well simulates the overall shape of runoff hydrographs and sedimentographs at the watershed outlets. The model results on runoff hydrograph have also been validated using data at the internal gauge station in Pathri Rao watershed and found to have good correlation with observed runoff hydrographs. Further, it is to be mentioned that this type of validation of outputs of a distributed hydrological model both at watershed outlet as well as at internal gauging station for such a relatively large sized watershed (Pathri Rao) makes this study a first of its kind especially in Indian watersheds. A study on the effect of interception process on the model outputs has also been carried out to demonstrate its significance in estimation of runoff and sediment yield. The results thus obtained justify the inclusion of interception loss component in rainfall-runoff-soil erosion and sediment yield modelling. Meaningful relationships have been identified, albeit qualitatively, between the calibrated optimum values of soil and landuse dependent model parameters and the physical characteristics of the computational elements in the study watersheds. With the use of GIS techniques, spatial and temporal distribution of runoff and spatial distribution of sediment yield for different storm events have been produced in the form of maps, which would be useful for effective implementation of soil and water conservation measures in the watersheds. xi