Hydrological Modelling of the River Ythan using ArcInfo GRID

The River Ythan in Grampian Region, Scotland, is being investigated under an European Council Directive (91/676/EEC) for possible designation as a Nitrate Vulnerable Zone because of an increase in estuarine eutrophication in recent years. This paper explores the use of the GRID module of ArcInfo GIS for hydrological modelling of the River Ythan as part of the development of a spatial decision support system for the river catchment.

A hydrologically correct elevation grid was created for the catchment area using a DEM and digitised stream network. From this, the flow directions, flow accumulations at selected pour points and sub-catchment boundaries were created. These were compared with discharge values measured at the pour points and hand-delineated sub-catchment boundaries provided by the North East River Purification Board (NERPB). GRID was then used to measure the length of the stream channels and by assigning suitable flow velocities, an isochrone map was produced to show the spatial pattern of catchment hydrological response.

Currently this hydrological model is being extended to incorporate spatial variation in rainfall and evapotranspiration. The hydrological model is also being coupled with a hydrochemical model to predict the chemical composition of drainage water in the catchment. This will help policy makers decide on suitable methods for reducing nitrate concentrations in the River Ythan and provide a GIS framework for modelling hydrochemical responses to altered land use in the catchment.

The River Ythan in Grampian Region, Scotland is being investigated under a European Council Directive (91/676/EEC) for possible designation as a Nitrate Vulnerable Zone because of an increase in estuarine eutrophication in recent years.

Eutrophication is associated with increasing nitrate concentrations leading to increased growth of algae and weeds and has had an impact on seabird population in the Ythan estuary which is also a National Nature Reserve.[Raffaelli, 1989] It is estimated that 5,000 kg of nitrogen enters the river each day, overall only 2% of this nitrogen derives from sewage discharge.[Pugh, 1993] Most is derived from soil processes linked with land use practices in the catchment [MacDonald et al 1995]. Monitoring of the water quality has shown that nitrate concentrations have increased by 300% since the 1960s ( Figure 2).

The Ythan Catchment is about 685 km2 and contains gently undulating terrain. The river reaches the estuary at the town of Ellon. Elevations within the catchment range from sea level to 300m. The average annual rainfall varies from 700mm on the east coast to 900mm on the highest parts of the catchment. The increased loss of nitrogen to the estuary in recent years has been attributed to the high proportion of agricultural land in the catchment. Landuse in the catchment is dominated by agriculture (90% of total) which comprises of a mix of arable and livestock systems. The Scottish Office Agriculture Environment and Fisheries Department (SOAEFD) have launched a research project to investigate whether the river nitrate concentrations in the Ythan can be reduced by changing agricultural practices and improving the efficiency of nitrogen use.

Due to the inherent ability of Geographical Information Systems (GIS) to combine and analyse multiple data layers, ArcInfo has been used to store the spatial data sets available. The aspatial data concerning farming practices, fertiliser use and cropping for the area are held within a relational database which is linked to ArcInfo. The GRID module of ArcInfo is used to construct a spatially distributed model. This will be linked with a hydrochemical model which can be used to evaluate various land management scenarios and thus as part of an integrated management plan for the catchment, which aims to reduce nitrate concentrations in the River Ythan. This paper will concentrate on the hydrological modelling using GRID.

The flow chart in figure 3 shows the methodology followed during the project. The following data sets are being used:

DEM elevation data at 1:50,000 at 50m grid resolution

Soil coverage at 1:25,000

Land Coverage at 1:25,000

Catchment and subcatchment boundaries coverage (1:25,000)

Stream coverage (1:250,000)

HOST (Hydrology Of Soil Types) [Boorman et al.1995] coverage is also available at a resolution of 50m

The DEM, plus the boundary data of the Ythan catchment and the stream data, which had previously been edited to ensure no braided streams or lakes existed and all the rivers were flowing towards the estuary, were input into the GRID function TOPOGRID. TOPOGRID interpolates a surface from the height data, then modifies this surface to enforce the drainage representation contained in the stream coverage. This produces a hydrologically consistent surface. During the processing sinks were created, mostly by rounding errors in the data. The largest and deepest sinks were field investigated to check if they were actual topographic features; all were artefacts of the processing.

To begin analysis the flow direction from each cell down the grid had to established. This was achieved by creating a flow direction grid. The FLOWDIRECTION command in GRID computes the direction of flow for every cell to its steepest downslope neighbour. Following this, a flow accumulation grid was produced, using the FLOWACCUMULATION command, which accumulated all of the upstream water for each cell.

The objective, of this stage was to extract a drainage network from the hydrological surface created by GRID for all cells of accumulation values of 100 or over. However it is important to recognise that the drainage networks produced from GRID do not necessarily represent the natural system of river courses within a hydrological basin. Hydrological models work on the understanding that; 'Cells are defined as channels when they exceed a threshold contributing area resulting in a drainage system, that is actually suitable for hydrological interpretation' [La Barbera et al., 1993] Comparisons were made between the extracted stream network and the original digitised stream data. Results showed that in the worst case the stream network deviated by 2 cells equivalent to 100m.

Having created the flow accumulation grid, pour points could be found for each subcatchment by querying the values in the flowaccumulation grid directly from the screen. Where two or more streams joined the main channel near a subcatchment boundary and where there was a significant rise in the number of cells accumulated at that point, a pour point was assumed and the coordinates recorded in an ascii file. This was converted to a snap grid which held all the coordinates for the pour points. The WATERSHED function was used with the flowaccumulation grid and snap grid to find the subcatchments. However, not all the subcatchments in the Ythan have specific pour points as a number of diffuse reaches also exist.

Comparisons were made between the measured areas of the hand delineated subcatchments and those subcatchments produced by GRID. The method used is based on an approach for assessing the accuracy of area measurement that uses the Epsilon band concept (Aspinall, 1996). The grid coverage and hand delineated coverages were unioned and intersected and the sliver areas and perimeters for each subcatchment determined. The Epsilon band width was calculated as the width on a line of length P (mean perimeter) that contains 95% of the sliver area. Figure 4 shows the slivers between the GRID produced boundaries and hand delineated boundaries. As can be seen from Table 1, the Epsilon bands for Little Water and Cessnie subcatchments were substantially higher than the other subcatchments. When these were investigated in detail, the error was associated with the hand delineated subcatchment boundaries; these were subsequently altered. The largest slivers were field checked and found to be areas of flat land which is problematic for either method to define a specific boundary

Figure 4 Differences in Subcatchment Boundaries

Comparisons of discharge values

Only one continuous flow gauging station positioned at the Ellon estuary is currently used to measure the mean annual discharge. Using the GRID hydrological model, we can predict flow rates at the confluences of other tributaries which have not been monitored but have defined pour points.

The NERPB have recorded values of discharge at Ellon over time, and the average mean annual flow (Q) of the Ythan river at Ellon is 7.2 m3/s. This value was taken as a proportion of the total flow accumulation produced by GRID for the pour point at Ellon and proportionally distributed to the remaining pour points in the catchment. The results are presented in Table 2. (Note: the NERPB results for the subcatchments are derived from a model and are not actual measured values.) Because the Ythan catchment has a relatively uniform distribution of rainfall, runoff and land cover; the discharge value at the estuary can be distributed to the whole catchment to produce the recorded results. This would not be possible for more heterogeneous catchments. It can also be seen that the stream density is uniform throughout the subcatchments and there is a strong relationship between the area and predicted discharge highlighting the similarity of hydrological response over the entire catchment.

The hydrological model has so far been based on the whole catchment. To enable more detailed study of the catchment, two subcatchments have been selected for further modelling. The selection procedure was based on the choice of subcatchments which would best represent the catchment as a whole in terms of soil type, land cover, HOST (Hydrology Of Soil Types) [Boorman 1995], elevation and hydrology. The subcatchments also have to be of 'average size', with their tributaries discharging directly into the main stream of the river, not the estuary. The subcatchments were chosen using the above criteria and areal statistics produced for the soil type, land cover and HOST.

Once the stream network had been defined and the stream lengths from each cell to the outlet calculated, an average flow velocity was assigned to each cell. If both flow direction and velocity are known and the pathway from each cell to the outlet has been specified, then a grid can be created of the flow travel times. This uses time = distance / velocity, where the value in each cell is the time taken for the water from that cell to the watershed outlet.[Maidment, 1993] Using GRID it was possible to produce an isochrone map which gives an indication of the time to generate a discharge response at the catchment outflow. Data are currently being collected on velocities over a number of seasons and storm events and gauging equipment has been installed at the selected subcatchment outflows to measure actual discharge and validate the predicted values.

The present objective has been to model the Ythan catchment using the GRID hydrological model and compare the predicted data with actual data obtained from the NERPB. The results show that GRID can be used to detect errors in the hand delineation of the subcatchment boundaries. It also proved an effective means for assessing the discharge values modelled by the NERPB. Future work will concentrate on the development of subsurface water flow and spatial variation in rainfall and evapotranspiration at a catchment and subcatchment scale. Modelling at the subcatchment scale will use larger scale data including a more detailed stream network. One function of these developments will be to produce a grid of hydrological response times within the catchment.

The hydrochemical model is being studied at various scales, from the whole of the catchment to the farm to the field scale. Data are being collected on water quality within the subcatchments every fortnight and representative soil samples are also being taken regularly, from which an estimate of nitrate can be obtained. The approach to modelling at each scale will require different inputs and produce a range of information of differing levels of detail at each scale. The final objective of the project is to predict the effect of various land management scenarios on river nitrate concentrations for the whole catchment.

The hydrological and the hydrochemical models embedded in the GIS framework will provide policy makers with a decision support tool with which to assess the impact of land use scenarios in reducing the quantity of nitrogen leached to the estuary. This will help in the development of a management plan to reduce nitrate concentrations in the River Ythan and it's estuary, which will be required if the catchment is designated a Nitrate Vulnerable Zone.

This research is funded by the Scottish Office Agriculture, Environment and Fisheries Department.

Aspinall, R.J., (1996) 'Measurement of Area in GIS: A rapid method for assessing accuracy of field areas for IACS'. GISRUK Conference proceedings 1996

Boorman, D.B., Hollis, J. M., & Lilly, A. (1995) Hydrology of Soil Types (HOST) : a hydrologically-based classification of the soils of the United Kingdom. Report no. 126. Published by Institute of Hydrology

La Barbera, P., Lanza, L. & Siccardi, F. (1993) Hydrologically orientated GIS and application to rainfall - runoff distributed modelling: case study of Arno basin. In: application of Geographic Information Systems in Hydrology and Water Resources Management (ed. by K.Kovar & H. P. Nachtnebel), IAHS Publication No. 211 pp 171 - 179

MacDonald, A.M., Edwards, A.C., Pugh, K.B. & Balls, P.W. (1995) Soluble nitrogen and phosphorous in the river Ythan system, U.K.: annual and seasonal trends. Wat. Res., 29. 837-846

Maidment, R.D. (1993) Developing a spatially distributed unit hydrograph by using GIS. In: application of Geographic Information Systems in Hydrology and Water Resources Management (ed. by K.Kovar & H. P. Nachtnebel), IAHS Publication No. 211 pp 181 -192

Pugh, K.B., (1993) The nutrient status of the Ythan catchment and estuary. North East River Purification Board, Technical Report 93/1

Raffaelli, D., Hull., S. & Milne, H., (1989) Long-term changes in nutrients, weed mats and shorebirds in an Estuarine System. Cah. Biol. Mar., 30, 259-270