Environmental Mapping and Modelling of a Catchment using GIS

As farm production has intensified in recent years, the non-point source pollution of water has become a significant problem. At the Macaulay Land Use Research Institute (MLURI), we are applying GIS technology to help with the management of the River Ythan catchment, Aberdeenshire, Scotland. There has been a 300% increase in the amount of nitrogen entering the estuary since the 1960s which is thought to be related to changes in agricultural practice [MacDonald at al 1995]. The estuarine eutrophication has stimulated greater algae growth,which has reduced the amount of food available for wildlife in the National Nature Reserve (the Ythan estuary) [Raffaelli, 1989]. This paper discusses how, using spatial analysis techniques and a hydrological model in a GIS framework, we can assess the possible impact of land use scenarios for reducing nitrogen leached to the estuary. Using satellite imagery, we have gained an insight into the distribution of crops and organic nitrogen inputs throughout the catchment in an attempt to target areas of high risk N loss. With the development of a partially distributed hydrological model in GRID we can calculate the weekly mean flow at any point on the river network and the daily accumulated flow at the outlet allowing estimation on nitrate loads entering the estuary. As potentially higher quantities of N are leached from land under certain crop types, we have also investigated the cropping patterns in close proximity to the streams which may be of importance in deciding future abatement measures. GIS techniques have been particularly useful in enabling us to analyse the effects of agricultural non-point sources by the following means: the GIS is used to store, co-ordinate and manipulate the spatial and satellite data as well as acting as a link to databases holding socio-economic data in tabular form; the network analysis provides a means of interpreting pathways of nutrient loss and connects agricultural land use throughout the catchment with the estuary, allowing for the estimation of nitrate loads entering the estuary; the output from the GIS is used to communicate the results of analyses in a flexible and visual manner that is immediately understandable.

Introduction

As farm production has intensified in recent years, pollution from diffuse sources has become a significant problem. This has been suggested to be the case in the River Ythan, Aberdeenshire, Scotland. A three fold increase in nitrogen since 1960 has been observed, with 5,000kg of nitrogen currently entering the estuary daily [MacDonald at al 1995]. The Ythan estuary is a National Nature Reserve and recognised as an important breeding area for shorebirds. However, the increased eutrophication (nutrient enrichment of natural waters) has stimulated algal growth interfering directly with the availability of food for wildlife [Raffaelli, 1989]. In the long term, this may have dramatic consequences for the whole ecology of the estuary. The Macaulay Land Use Research Institute (MLURI), is currently undertaking a three year research programme to investigate whether the increased river nitrate concentrations in the Ythan have been influenced by changing agricultural practices. ArcInfo has played a key role in the project, co-ordinating and performing spatial analysis on the data as well as providing a framework for a hydrological model which could be used as a tool in the management of the River Ythan catchment.

Profile of the Catchment

The Ythan catchment (685 km²) is largely homogeneous in terms of soil types, rainfall and elevation. The land use is 90% agriculture comprising a mixture of arable and livestock farming. The analysis of river samples collected over the past 30 years has revealed a steady increase in nitrate concentrations. The geographic pattern of the increase has been shown to have a relatively constant profile along the river system, suggesting that the source is distributed throughout the catchment as a whole. It has been suggested that the recent changes in agricultural activity and land management that have occurred in the catchment could be a major contributing factor to the increased nitrate concentrations in the river and hence to nutrient input to the estuary.

A large variety of both spatial and attribute data were used in the project (see figure 1). The land, soils and hydrological spatial layers used are a combination of the Scottish Office, MLURI and Ordnance Survey data. The annual agricultural census data provided information on types and areas of crops grown and livestock numbers, which was combined with information on the application of organic manures in order to calculate N input. Specific information on actual practices was obtained from a farm nutrient questionnaire.

Figure 1. Data used in the project

Monitoring River Flow

An important component of any future managemnet strategies to reduce the river nitrate concentrations requires the prediction of the quantity of nitrate actually reaching the esturay. Therefore it is necessary to understand the dynamics of nitrate transport within the catchment. Using the GRID module of ArcInfo; the DEM at 50m resolution and the digitised streams were input into TOPOGRID to produce a hydrologically consistent surface. From this, the stream directions, stream accumulations and sub-catchment boundaries were calculated. The sub-catchment boundaries were subsequently compared with hand delineated boundaries, field checked and modified [McAlister et al. 1996]. Analysis of the hydrograph showed that 80% of flow in the Ythan catchment is subsurface with very little ground water contribution. Therefore a hill-slope model, written in FORTRAN, was developed and loosely coupled with ArcInfo to predict the amount of water flowing in the subsurface, from any point on the catchment, into the stream network on a daily basis [Dunn 1996]. The model used daily rainfall values as one of many physical parameters, while GRID was used to calculate the topographic parameters. These included:

• Distance from each cell in the catchment to its nearest stream along the flow path
• Slope between each cell and the nearest stream

To find the distance along a flowpath to the closest stream, the streams were coded as sinks and merged with the flow direction grid. Distances along the flowpaths were found using the FLOWLENGTH command. Figure 2 shows a categorised output of the flow length grid.

Figure 2. Distances along flowpaths

We based the calculation of the average slope angle on the difference in height between any cell in the catchment and the closest stream cell, divided by the distance between the two cells. The stream network was recoded to contain the elevation for each individual stream cell using the DEM. From this, sub-catchments for each cell on the stream network were calculated and coded using the elevation values of the origin pour points. To find the average height difference between any cell in the catchment and the stream network the sub-catchment and stream network grids were subtracted.

AML programs were written to obtain:

• the weekly mean flow at any point on the stream channel by accumulating the average weekly flow rates down the flow paths
• the daily accumulated flow at the outlet of the catchment by performing calculations on the VAT tables.

Calculating the sub-catchments for each stream cell enabled us to visualise which areas are contributing most flow to the river system and to consider these when targeting areas for N abatement schemes. [see figure 3]

Figure 3. Classified sub-catchment areas for each cell on the river network

Cropping Patterns

The range in farm types and management practices may potentially produce a spatial component to the nitrate loss. This may have a major influence when considering the relative location of crops with respect to the stream channels. ArcInfo was used to investigate the planting patterns of the crops and their proximity to the streams. Satellite imagery was used to identify the crop types in the catchment. A good quality, cloud-free, Landsat Thematic Mapper Image taken during the growing season in 1994 was used for the classification. The image, having been geometrically corrected and classified, was imported into the GIS and reclassified into spring, winter, grass and root crops.

From studying each class in turn, it appeared that the root crops were located towards the edge of the catchment boundary, the winter crops had a tendency to be planted on the lower elevations closer to the streams, and the spring crops were more prominent in the higher parts of the catchment. A more comprehensive analysis to determine the proportion of each crop type within a certain distance of the nearest stream was undertaken. Using buffers the percentage of each crop type planted at different distances from the streams was found. Different cell resolutions, acting as buffers, resampled at increments of 100 m, were used to mask out areas over the streams for each crop in turn. All the buffers, for each crop in turn, were combined and the percentage of crop in each distance class calculated. [see figure 4].

Figure 4. Percentage of crops in each distance class from the streams.

A compositional analysis was performed to test whether there was a spatial pattern in the distribution of crops in relation to the river network within the catchment. This compared the observed area of crops at different distances from the river with the area that would be expected if the crops are distributed at random in relation to the river network. Figure 5 shows the results of the analysis using a 95% significance level. At a distance of 400m from the streams grass is over represented, while spring crops are over-represented at a distance of 600m. At 900m and 1000m from the streams all the crops are grossly under represented implying that farmers tend to avoid planting in these areas (which have slightly higher altitudes). We can only conclude from the results that there is a random distribution of crops in 1994 in relation to the distance from the streams. However, the annual rotation of crops in the catchment means these findings may change from year to year.

Figure 5. t-test results of compositional analysis comparing expected areas and observed areas of crops at different distances from streams.

[The values in bold italics imply over representation of particular crops, values in bold imply under representation]

Organic Inputs

In the Ythan catchment it is necessary to identify and target areas of highest N input. From information provided by the farmers, we calculated N manure (kg/ha) values for each farm based on numbers of livestock and the farm area. Thiessen polygons were produced based on N manure (kg/ha) values and the farm dwellings to visualise the geographic pattern in the amount of Nitrogen manure (kg/ha) applied to the farms [see figure 6].

Figure 6. Thiessen polygons showing the total N(manure) applied to farms in the catchment

As many of the most intensive farms are also the smallest, their impact on the N input into the catchment may be overlooked on the map. The thiessen polygon map, as conventional maps, can be seen as land area cartograms because they are drawn in proportion to their land areas. However, using a cartogram drawn in proportion to N kg/ha, allows a clearer interpretation of the data, highlighting the extreme N values. As can be seen from figure 7, the cartogram suggests a more visual pattern of higher intensive manure input towards the edge of the catchment.

Figure 7. Cartogram showing the total N(manure) applied to farms in the catchment

The N (manure) thiessen polygon coverage was combined with each crop in turn to help identify relationships between manure application rates and specific crop types. As can be seen from figure 8, no pattern has emerged from this analysis.

Figure 8. The percentage of application rates in each manure N(kg/ha) class.

The domination of a farm type high in nitrogen production, in a particular sub-catchment, may help decision makers when planning abatement schemes for the catchment. Therefore the farms that replied to the questionnaire were classified by farm type and plotted as pie charts in their respective sub-catchments [see figure 9]. From studying the statistics no pattern could be found, with cereals and general cropping farms dominating in the majority of sub-catchments and pigs and poultry farms evenly distributed throughout the catchment.

Figure 9. Farm Types in Sub-catchments

Managing the Catchment

GIS techniques have had a key impact through each stage of this project in enabling us to analyse the effects of agricultural non-point sources of nitrate. The GIS has provided storage, co-ordination and manipulation of the spatial, satellite and questionnaire data. It has given us an insight into the distribution of crop planting patterns, organic N input, and agricultural practice throughout the catchment. The GRID module has also facilitated network analysis as a means of interpreting pathways of nutrient loss while connecting agricultural land use throughout the catchment with the estuary and thereby allowing estimation of nitrate loads entering the estuary. The output from ArcInfo has been very effective as a means of communicating the results of the analyses in a flexible and visual manner that is immediately understandable to non GIS users.

The final stage of the project, currently under completion, involves testing different land use scenarios, ultimately leading to the development of an economics policy of abatement measures to balance the relationship between land use activity and the amount of nitrate leached for the whole catchment. GIS has provided the framework for the project, allowing spatial analysis of the current problems while providing a predictive tool for future decision making regarding the environmental pollution in the catchment. It lets us link analysis of processes in the physical environment with economics and agricultural practices to solve an environmental problem.

Acknowledgements:

This project is funded by the Scottish Office Agriculture Environment and Fisheries Department (SOAEFD).

References:

Domburg, N., McAlister, E. and Edwards, A.C. (1996) Spatial Modelling of the nitrogen cycle in the River Ythan catchment. Conference on Information and Communication Technology Applications in Agriculture : state of the art and future perspective. Computers and Electronics in Agriculture.

Dunn, S.M., McAlister, E. and Ferrier, R.C. (In press) Development and application of a distributed catchment scale hydrological model for the River Ythan, North East Scotland. Journal of Hydrological Processes

Dorling, D (1995) Visualizing the 1991 Census. In S Openshaw (ed) A Census User's Handbook. London : Methuen

McAlister, E., Domburg, N., Edwards, T., Ferrier, B (1996) Hydrological Modelling of the River Ythan using ArcInfo GIS

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

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