Role of Remote Sensing and GIS inputs in physically based hydrological modelling
Dr. S. M. Seth Director, National Institute of Hydrology, Roorkee247667 (U.P.) INDIA The scope of hydrological applications has broadened dramatically over the past four decades. Although the problems of flood protection and water resources management continue to be of importance and relevance for the security of communities and for human, social and economic development, many applied problems relating to the wider role of hydrology have come into focus. The management of waste and pollution in the environment requires the characterisation of hydrological flow and transport processes in freshwater, soils and groundwater systems. These depend on complex and poorly understood hydrobiological and hydrogeochemical interactions. Furthermore, changes must be characterised at a range of scales, including global, and this requires an integrated approach in which hydrological processes are central to the global ecosystem. Underlying applied problems is a set of scientific issues to which answers must be found in order to make progress. One requirement is to improve process understanding, key uncertainties include hydrological and associated biological and geochemical processes. Some limitations are due to problems of observability (the ability to estimate process parameters from observations). This is partly due to a need for new measurement techniques (for example for subsurface flows) and partly results from difficulties in describing spatial variability identified by measurements at appropriate spatial and temporal scales. The combination of these issues is a known restriction on the application of computer modelling to many important applied problems. Watersheds, catchments, river basins are subjected to many types of modifications by human and natural activities. Such changes can be distinguished as point changes and nonpoint changes and affect virtually all elements of hydrologic cycle. Structural changes such as dam construction, channel improvement, detention storage etc. are examples of point changes. Forestry, agriculture, mining, urbanization etc. are nonpoint land use changes. There has been a growing need to study, understand and quantify the impact of major landuse changes on hydrologic regime, both water quantity and quality. This is necessary to anticipate and minimize potential environmental detriment and to satisfy water resources requirements. Hydrological Modelling Hydrological modelling is a powerful technique of hydrologic system investigation for both the research hydrologists and the practising water resources engineers involved in the planning and development of integrated approach for management of water resources. Hydrologic models are symbolic or mathematical representation of known or assumed functions expressing the various components of a hydrologic cycle. However, the term hydrological model is often understood to be and is used more narrowly as a computer based mathematical model. With the current rapid developments within computer technology and hydrology the application of computer based hydrologic models can only continue to increase in the near future. Various techniques are available in the literature for modelling hydrologic system. Simulation is one of them where a system is represented as a model and its behaviour is studied. Digital simulation is needed in watershed research because it is a complex system to be analysed by exact mathematical techniques. In digital simulation, system model is developed by a number of mathematical expressions that represent the various processes of the system and simulation is done by using a computer. Hydrological models can be classified in different ways. Broadly many of the models presented in the literature can be divided into deterministic and stochastic categories. A deterministic model is one in which the processes are modelled based on definite physical laws and no uncertainties in prediction are admitted. It has no component with stochastic behaviour i.e. the variables are free from random variation and have no distribution in probability. Deterministic models can be further classified according to whether the model gives a spatially lumped or distributed description of the catchment area, and whether the description of the hydrological processes is empirical, conceptual or fully physically based. The familiar classification of model classification is to classify them in three categories: a) Black box models, b) Lumped models and, c) Physically based models. Black box models Black box models are based on transfer functions which relate inputs with outputs. These models, as the name suggests, generally do not have any physical basis. Conceptual model Lumped conceptual models occupy an intermediate position between the fully physically based approach and empirical black box analysis. Such models are formulated on the basis of a relatively small number of components, each of which is a simplified representation of one process element in the system being modelled. Physically based models: The physically based models are based on our understanding of the physics of the hydrological processes which control the catchment response and use physically based equations to describe these processes. Also, these models are spatially distributed since the equations from which they are formed generally involve one or more space coordinates. A discretization of spatial and temporal coordinates is made and the solution is obtained at the node points of this discretized representation. This implies that these models can be used for forecasting the spatial as well temporal pattern of more than one hydrological variable. Such models require much of computational time and also require advance computers as well as a broad data base. Physically based distributed models do not consider the transfer of water in a catchment to take place in a few defined storage as in case of lumped conceptual models. From their physical basis such models can simulate the complete runoff regime, providing multiple outputs (e.g. river discharge, phreatic surface level and evaporation loss) while black box models can offer only one output. In these models transfer of mass, momentum and energy are calculated directly from the governing partial differential equations which are solved using numerical methods, for example the St. Venant equations for surface flow, the Richards equation for unsaturated zone flow and the Boussinesq equation for ground water flow. As the input data and computational requirements are enormous, the use of these models for realtime forecasting has not reached the `production stage' so far, particularly for data availability situations prevalent in developing countries like India. Role of Physically Based Distributed Models Physicallybased distributed models can in principle be applied to almost any kind of hydrological problem. Obviously, there are many problems for which the necessary solutions can be obtained using cheaper and less sophisticated empirical, lumped conceptual or statistical models. However, for the more complicated problems there may be little alternative, but to use a physicallybased distributed model. Some examples of typical fields of application are: Catchment changes These include both natural and manmade changes in landuse, such as the effects of forest fires, urbanization and forest clearance for agricultural purposes. The parameters of a physicallybased, distributed model have a direct physical interpretation, which means that they can be evaluated for the new state of the catchment before the change actually occurs. This enables the effects of changes to be examined in advance of such changes. In addition, the characteristically localized nature of catchment changes can easily be accounted for within the spatially distributed model structure. Ungauged Catchments An application in a previously ungauged catchment requires the initiation of a programme of field work to provide data and parameters for calibration. Here, the physical significance of its model parameters enables e.g. the SHE to be applied on the basis of a much shorter, and therefore more cheaply obtained, hydrometeorlogical records than is necessary for more conventional models. Similarly the catchment parameters can be estimated from intensive shortterm field investigations. Spatial variability Spatial variability in catchments inputs and outputs. Distributed models can be used to examine the effects on flood flow of different directions of storm propagation across a catchment and also the effects of localized river and groundwater abstractions and recharge. This facility is beyond the capability of lumped catchment models which can deal only with quantities averaged across the catchment. Movement of Pollutants and Sediments Movements of pollutants and sediments. In order to model the movement of pollutants and sediments, it is first necessary to model the water flows which provide the basic dispersion mechanism. Most water quality and sediment problems are distributed in nature, so distributed models are the most suitable for supplying the basic information on water flows. Topmodel The Topmodel is a variable contributing area conceptual model in which the predominant factors determining the formation of runoff are represented by the topography of the basin and a negative exponential law linking the transmissivity of the soil with the vertical distance from the ground level. In this model the total flow is calculated as the sum of two terms: surface runoff and flow in the saturated zone. The surface runoff, in the most recent versions of the model, is in turn the sum of two components, the first generated by infiltration excess and the second, referring to a variable contributing area, by saturation excess. Though a conceptual model, i.e. one in which the physical reality is represented in a simplified manner, the Topmodel is frequently described as being `physically based', in the sense that its parameters can be measured directly in situ (Beven and Kirkby, 1979). This definition is somewhat optimistic, in veiw of the doubts and uncertainties encountered even in defining the parameters of the `physically based models', as already mentioned. Topmodel performs what is called an `upward search for conceptualisation' from the soil column level to the catchment scale. Basin parameters are related to point estimates. The spatial variability of both soil water content and lateral drainage is related to that of soil and topographic characteristics by means of simple but meaningful assumptions. The model is also attractive because of its structural simplicity and parsimonious parameterisation. The TOPMODEL is one of the few conceptual models that accounts explicitly for the saturation excess overland flow mechanism and integrates the variable contributing area concept, both of which are essential to model the catchment accurately. Topmodel represents catchment topography by means of a topographic index, ln(a/tanB), where `a' is the area draining through a grid square per unit length of contour and `tanB' is the average outflow gradient from the square. The index is calculated from a Digital Terrain Map (DTM) across a grid covering the catchment. The grid must be sufficiently fine to resolve important characteristics and slope formations. A high index value usually indicates a wet part of the catchment; this can arise either from a large contributing drainage area or from very flat slopes. Areas with low index values are usually drier, resulting from either steep slope or a small contributing drainage area. Grid squares with the same index values are assumed to behave in a hydrologically similar manner. As a result of this assumption, the catchment's topography may be summarised by the distribution of the index values. She Model The Systeme Hydrologique Europeen (SHE), the Institute of Hydrology Distributed Model (IHDM), and the USDAARS small watershed model are the familiar models from this group. Because of their inherent structure these models also make very little use of contour, soil and vegetation maps, or of the increasing body of information in such areas as soil physics and plant physiology. Similarly, much historical information frequently consulted during project planning, for example crop yields over specific periods, survival patterns of particular types of vegetation and characteristics events occurring during floods and droughts, is not used directly. A considerable improvement in project planning could therefore be derived from the integration of such information into the modelling process. These observations do not imply any criticism of conventional rainfallrunoff models in relation to the more traditional applications in which they have clearly been successful, for example realtime flow forecasting and the extension of short stream flow records using longer rainfall records. However, they serve to underline some of the potential which a new approach in hydrological modelling might be able to fulfil. In particular physicallybased, distributed models can in principle overcome many of the above deficiencies through their use of parameters which have a physical interpretation and through their representation of spatial variability in the parameter values. (Storm,1989). Structure of SHE : SHE has been developed as a fully modular system for mathematical description of the land phase of the hydrological cycle. The system comprises the following models for description of water flows:
In addition to these water flow components, addon modules have been/are
being developed for :
Point and Space Scales The scope of hydrology is best defined by the hydrologic cycle. Depending on the hydrologic problem under consideration, the hydrologic cycle or its components can be traced at different scales of time and space. The global scale is the largest spatial scale and the watershed or drainage basin, the smallest spatial scale. Time scales used in hydrologic studies range from a fraction of an hour to a year or perhaps many years. The physics of the process by which rainfall is separated into surface runoff and infiltration, and further into evaporation and ground water recharge  i.e. the basic processes of the hydrological cycle  is best understood on the point scale. The basis for operational hydrology is the catchment or river basin, or urban area. The average of the parameters observed on a point scale is not necessarily representative of the conditions of a catchment. Hydrologists are keenly aware that what they observe on a point scale can not be integrated directly into area averages useable for operational hydrology, and that the spatial redistribution of the water cycle components must be considered. Use of Remotely Sensed Data The remotely sensed data (aerial photography and satellite imagery) provide spatial information about the processes of the land phase of the hydrological cycle. The land cover maps derived by remote sensing are the basis of hydrologic response units for modelling units. For an understanding of the hydrology of areas with little available data, a better insight into the distribution of the physical characteristics of the catchments is provided by image processing techniques. Some of the new measurement methods (photographic systems, active radar systems etc.) could yield assessment of areal distribution or atleast to some extent reliable areal totals or averages of hydrologic variable such as precipitation, evapotranspiration and soil moisture. Some of the main hydrological application field of remote sensing are:
Geographic Information System (GIS) focuses on proper integration of user and machine for providing spatial information to support operations, management, analysis and decision making. Since, GIS does not directly land itself to time varying studies, its features are utilised in hydrological studies by coupling it with hydrological models. Two types of approaches are possible for this purpose. In the model driven approach, a model or set of models is defined and thus the required spatial (GIS) input for the preparation of the input data and output maps. The other approach is the data driven approach. It limits the input spatial data to parameters which can be obtained from generally available maps, such as topographic maps, soil maps etc. The possibility of rapidly combining data of different types in a GIS has led to significant increase in its use in hydrological applications. It also provides the opportunities to combine a data from different sources and different types. One of the typical applications is use of a digital terrain model (DTM) for extraction of hydrologic catchment properties such as elevation matrix, flow direction matrix, ranked elevation matrix, and flow accumulation matrix. It also provides the ability to analyse spatial and nonspatial data simultaneously. Application of Topmodel Calibration and validation of the Topmodel was carried out on Hemavati catchment situated in Western Ghats. Raster DEM input for the model is generated through ILWIS after digitization contour map from Survey of India toposheets. In all 5 years of data was available for simulation study. Available data series was broken in two parts and ifrst part i.e. June 1975December 1977 was used for model parameter calibration and remaining data series i.e. January 1978 to December 1980 was used for model validation. Simulation results are encouraging. Model efficienty (NashSutcliffe) was more than 0.84 both for model calibration and validation on independent data series.(Jain, 1996). The Topmodel has been applied to Malaprabha catchment in Karnataka to simulate the daily flows at Khanapur. River Malaprabha is a tributary of river Krishna. The catchment area of Malaprabha upto discharge measuring site Khanapur is 520 Sq. Km. The model uses topographic index for the formation of runoff. The topographic index for Malaprabha catchment was derived by developing a Digital Elevation Model (DEM) by interpolating the contours in the basin at 300 m grid size. The results indicates that the model can be used to simulate the flows in the catchment quite accurately (the efficiency of the model is 0.89 and 0.79 respectively in calibration and validation run). Also, model is able to simulate the timing and magnitude of the peak flows satisfactorily.(Venkatesh and Jain, 1997) Application of SHE model for subbasins of river Narmada The distributed and physically based nature of the SHE requires that in each application study a vast amount of data and parameters describing the physical characteristics of the catchments are available. This includes data for catchment geometry, land use and soil parameters, surface flow characteristics and input data of rainfall, evapotranspiration, stream flow etc. The data handling package of the SHE Model has therefore been organised on the lines of a typical GIS format. A brief processor package produces maps of spatially distributed data and enables an automatic setup of input data for the SHE at a desired grid square scale. Digitised data such as contour lines from toposheets are transformed to average elevation values compatible with the chosen grid square size. Also soil landuse maps are digitised and codes attached to each type are allocated to each grid square. The SHE model was applied to six subbasins of river Narmada namely  Kolar, Barna, Sher, Ganjal, Hiren and Narmada upto Manot. The SHE model has been found to be successful for modelling the rainfall, runoff process in these subbasins within data availability constraint. These studies have also provided useful guidelines for carrying out systematic field investigations for determination of various parameters for application of a physically based modelling approach. Typical flow chart of SHE Programme package is shown in Fig.2. Typical studies were also carried out for Kolar basin to consider the effect of changes in soil depth, soil properties, vegetation, surface roughness characteristics etc. on water yield from the catchment. Some simple studies for hypothetical irrigation systems were carried out for Barna basin. These clearly demonstrated capabilities of physically based distributed models using spatial information in typical GIS type framework. Recommendations GIS's are being effectively used in a variety of hydrologic applications. However, keeping in view the cost of implementing a GIS, especially when the cost of data collection and manipulation is considered, it is necessary to evolve such methodologies of hydrologic analysis and modelling where the GIS data base can be shared for several related purposes. There is need for a coordinated programme for this pupose particularly with a view to avoid unnecessary duplication. General Remarks Until very recently, GIS was a tool for managing and analysing spatial data and hydrologists were collecting their own data and sharing it in a format specific to the model they worked with. The integration of these two technologies began slowly as GIS was used to perform overlays of basin characteristics for further analysis by FORTRAN programme or a statistical package. However, now significant developments have already taken place for the integration of GIS and hydrological modelling. The basin characteristics for use in hydrological modelling is now being readily derived from digital elevation models. GIS is a valuable tool for use in parameterization for large scale physically based distributed models and significant developments in this area are taking place. These would result in corresponding increase in operational use of such models. There is however, need for considerable efforts to obtain representative hydrological data and information for use in such applications. References
 
