Primary Datasets for Case Studies of River-Water Quality

Abstract

Level 6 (final-year BSc) students undertook case studies on between-site and temporal variation in river-water quality. They used professionally-collected datasets supplied by the Environment Agency. The exercise gave students the experience of working with large, real-world datasets and led to their understanding how the quality of river water is dependent upon both natural processes and human activities.

Keywords:

• Case studies
• Primary data
• Rivers
• Water quality

Introduction

It is well known that applied statistics is best learnt by the use of real rather than artificially generated datasets (Singer and Willett, 1990). Within the biosciences, experience of data processing is a fundamental component of learning and ideally students will process data that they have collected themselves. Such self-collected data although real are, however, liable to be limited in quality and quantity; an alternative is to use professionally-collected data — indeed it has been suggested that the Centre for Bioscience might host a data bank of professionally-collected biological datasets for use in learning and teaching (Anon., 2007).

In the context of human impacts on river-water quality, it is to an extent possible for students to collect their own data. For example, students in Biological Sciences, University of Hull, in the course of a Level 5 (second-year BSc) module, work in formal laboratory practical classes, in teams of three, to process water samples collected from four sites along 7 km of the Driffield Canal in East Yorkshire, northern England. The teams plan their own work, carry out multiple tasks in parallel and are able to show change along the canal in response to a sewage-treatment-works (STW) outfall. This is achieved by determination of phosphate concentration and algal growth potential, assessed by bioassay through culture in canal water of the unicellular alga Selenastrum capricornutum. The data collected, however, are insufficient for statistical analysis, give no clues to potential seasonal or long-term change in water quality, and only two variables are assayed.

Students have text books and the primary scientific literature as alternative sources of quantitative information about human impacts on rivers. Text books, however, inevitably tend to be limited by the necessity for generalization. For example, Mason (2002) has an account of oxygen sag curves (the decrease and subsequent recovery of dissolved oxygen concentration downstream of organic-rich discharges) in rivers. There are generalized curves to illustrate the effects of mild, intermediate and gross pollution and an example for a specific day in a named river during low-flow summer conditions. Nonetheless, limited information is conveyed about the extent of variation between rivers and the effects of season and weather (temperature, flow) conditions. Primary sources (e.g. Kang and Goulder, 1996; Ainsworth and Goulder, 2000; Oliveira and Goulder, 2006) tend to be more site specific, and may refer to rivers that students can visit; they also facilitate quantitative understanding of parallel change by water-quality and biotic variables in response to discharges. Their disadvantage, however, is that they essentially present the students with processed data — there is no opportunity to manipulate and develop a feel for how raw data might best be processed and analysed.

The present contribution describes the use of large, professionally-collected datasets as the basis of case studies undertaken during 2007–2008 by Level 6 (final-year BSc) students in Biological Sciences, University of Hull. The datasets describe water quality in rivers in Yorkshire, northern England and were made available by the Environment Agency, the statutory agency with responsibility for monitoring river-water quality. The case studies were part of a 20 credit module Threats and Remedies in Aquatic Environments. Half of the module comprised lectures and/or workshops with emphasis on anthropogenic threats and potential remedies in marine, estuarine and freshwater environments. This was assessed by a 2-hour unseen examination. The other half of the module was a case study, undertaken individually, requiring 100 hours of work and chosen from within the areas of marine, estuarine or fresh waters. Those students who chose fresh waters worked on the rivers datasets. The work was submitted as a 2500 word report plus figures and tables.

The datasets and tasks set

The first step was to acquire the datasets. An e-mail enquiry was sent to [email protected]; an address obtained from the Environment Agency website (www.environment-agency.gov.uk). From this came contact with the External Relations Team at York and receipt by e-mail in May 2007 of a list of 17 985 Environment Agency sampling points in Yorkshire and Northumbria; not all were currently in use nor regularly sampled. Each sampling point had an eight digit code, a site name, its grid reference, and the date last sampled. A request was made for datasets for sampling points on several Yorkshire watercourses, including Market Weighton Beck, the Driffield Canal, the River Wiske and Cod Beck. These were received by e-mail; the data for each sampling point were in a separate Microsoft Excel Worksheet. Also received were the Environment Agency’s conditions for use of the data. These are generous: the data are available for non-commercial use, and may be copied, given to others or published provided they are not altered and the Agency’s ownership is acknowledged.

The data received were for sampling points at which water is regularly collected, usually at approximately monthly intervals. The water samples were analysed for about 20 determinands (Table 1). There was minor variation in the number and selection of determinands between sampling sites and with time, but the principal water-quality variables that are customarily associated with monitoring the effects of STW effluents were always included (e.g. BOD, ammonia, orthophosphate, oxygen). Values for all the determinands were given for every sampling date. The data were posted on the University’s VLE (Blackboard Learning System) location for the module.

For their case study the students each chose one from a list of seven tasks (Table 2). The tasks were open ended; students were not given specific hypotheses to test. They were told which potentially relevant data were available (Table 2), and where the data were to be found, but they were not directed to use particular data. Much more data than needed were made available and students were required to select the sampling points, determinands, and sampling dates that best suited their personal approach to the task set. Students were given references relevant to the rivers from which the data were provided. The tasks set (Table 2) encouraged students to relate the outcome of their data processing to a wider environmental context than the rivers from which the data were obtained. Ongoing support was through one to-one tutorials which the students booked with the tutor by e-mail. Students were advised that they should come to these tutorials with provisional work plans ready for discussion; it was not acceptable for them simply to ask what they should do. Ten weeks were allowed for completion of the task.

Table 1 An example list of determinands

Determinands
pH
Temperature (°C)
Conductivity at 25°C (µS cm^−1)
BOD with inhibition of nitrification (mg l^−1)
Ammonia as N (mg l^−1)
Total oxidized N (mg l^−1)
Nitrate as N (mg l^−1)
Nitrite as N (mg l^−1)
Ammonia, unionized, as N (mg l^−1)
Suspended solids (mg l^−1)
Total hardness, as CaCO3 (mg l^−1)
Alkalinity, pH 4.5, as CaCO3 (mg l^−1)
Chloride ion (mg l^−1)
Orthophosphate as P (mg l^−1)
Magnesium (mg l^−1)
Calcium (mg l^−1)
Chlorophyll a (µg l^−1)
Copper (µg l^−1)
Zinc (µg l^−1)
Dissolved oxygen (% saturation)
Dissolved oxygen (mg l^−1)

The determinands listed above are for a sampling point on Cod Beck at Topcliffe (grid reference SE 411 761) in January 2000; the range of determinands for other sampling days and sampling points was broadly similar. Values were recorded with the units shown in brackets.

Table 2 The tasks set and the data provided

Tasks set	Data provided
Consider how the impact of STW effluent on recipient rivers is related to effluent quality. Illustrate by reference to Weighton Beck and Market Weighton STW.	Data from Weighton Beck downstream of the STW outfall from 1990 to 2006. The STW was modernized in 1998. Data for the effluent (key determinands only) were also provided.
Consider the downstream recovery of watercourses from pollution/enrichment by STW effluent. Illustrate by reference to the Driffield Canal.	Data from 3 sampling points on the Driffield Canal (0.6 km upstream of the STW outfall and from 2.0 km and 4.2 km downstream) from 2000 to 2006.
Consider how the water quality of streams and rivers varies with season. Illustrate by reference to Weighton Beck	Data from one sampling point on Weighton Beck from 1986 to 2007.
Consider the downstream recovery of watercourses from pollution/enrichment by STW effluent. Illustrate by reference to the River Wiske.	Data from sampling points on the River Wiske (2.5 km upstream of the STW outfall, and 1 km and 8 km downstream) from 2000 to 2006.
Consider how the water quality of streams and rivers varies with season. Illustrate by reference to the River Wiske.
Consider the downstream recovery of watercourses from pollution/enrichment by STW effluent. Illustrate by reference to Cod Beck.	Data from 4 sampling points on Cod Beck (4 km and 0.5 km upstream of the STW outfall and 2.5 km and 5.5 km downstream) from 2000 to 2006.
Consider how the water quality of streams and rivers varies with season. Illustrate by reference to Cod Beck.

How the students used the data

Thirteen students undertook the work and they used the data in very different ways; partly this appeared to be personal preference and partly it reflected the different tasks set. The number of determinands used by each student ranged from three to ten (Fig. 1a). The most popular were BOD (11 students), orthophosphate (9), pH (9), ammonia (8), nitrate (8) and oxygen as % saturation or mg l⁻¹ (7). The number of sampling points considered ranged from one to four (Fig. 1b). The length of data runs used varied between 3 years and 19 years; most students used runs of 6-10 years (Fig. 1c).

Figure 1 The distribution of (a) the number of determinands processed, (b) the number of sampling points considered, and (c) the length of data run used (one student did not specify how many years were used in comparison of summer and winter months)

Discussion

By using real data, collected professionally by the Environment Agency, the students were able to gain appreciation of how the quality of river water is dependent upon both natural features and human activities. They made decisions on what were key determinands, and selected data from specific sampling points and sampling dates. This led to understanding of how river water quality is related to time of year, to the location of STW discharges, and to change in the quality of STW effluents.

There were other positive features. (1) The students’ own choice of determinands and own selection of sampling points and dates led to great diversity in the content of the reports submitted — thus individuals were able to explore their preferred lines of investigation albeit within the general bounds of the task set and data provided. (2) There was appreciation that there are missing and anomalous values in real-world datasets, and that accommodation to these is necessary. (3) Experience was gained in working with numerical values that had different units (e.g. mg l⁻¹, µg l⁻¹, % saturation; Table 1). (4) The requirements of the Biosciences Benchmark (QAA, 2007) for independent work and development of critical and interpretive skills were met. (5) The need to base the work on site-specific unpublished data prevented plagiarism.

Extensive use was made of graphs and/or tables. Most students also employed statistical techniques; non-parametric tests were much used. A minority, however, made decisions on the validity of hypotheses by reference to gross differences shown by graphs or tables. The students had received no formal statistics teaching since their first year (Level 4) and with hindsight more guidance on statistical analysis might have been offered.

After completion of the module, and publication of provisional marks, feedback was sought from the students. A questionnaire was used that asked how well the case study had met five specific module learning outcomes relevant to: knowledge and understanding of biological processes in natural waters; the importance of physical and chemical features; human impacts; the ability to analyse data critically; recognition of general ecological concepts. The extent to which learning outcomes were achieved was indicated on a scale of 1–5 (from 1 = not at all to 5 = completely). Although only three students responded their views are of value. They clearly considered that the learning outcomes had been met; the mean score for the five outcomes was 4.1 (n = 15, range 3–5). These students were also asked how difficult and how interesting the task was, in comparison with assessments done for other modules; their view was that the task had been relatively difficult but interesting to do. The three respondents scored 3, 4 and 4 for difficulty (1 = very much easier to 5 = very much harder than other assessments). The scores for interest were 4, 4 and 5 (1 = very much less interesting to 5 = very much more interesting than other assessments).

The students also expressed the opinion that it had been useful to analyse actual data because this provided a better view of how the real scientific world works. In addition, they enjoyed the fact that the data were real, and that the task led to appreciation of what people working in the field have to cope with. They concluded that the task acted as a taster for the sort of job that they might eventually undertake.

Acknowledgements

I am grateful to the students; Peter Gaffney, Alexander Greenwood, Nichola Kirk, Joseph Lewis, Yan Liu, Daniel Lucas, Sean Manning, Andrea Robson, Adam Smith, Victoria Smith, Hollie Stephenson, William Troy and Sam Walton. I thank Sam Watson and the Environment Agency at York for the primary data, and Graham Scott for his helpful comments on the manuscript.