Long-term monitoring of river water nitrate: how much data do we need?

Abstract

Long records of river water quality are invaluable for helping to understand the biogeochemistry of hydrological systems. They allow relationships to be established between changes in water quality (including seasonal cycles, episodic responses and long-term trends) and potential drivers, such as climatic forcing or human activity; they can act as a stimulus for process-oriented experimental research; they can be used to help to make predictions about future temporal and spatial patterns; and they can help to guide management options to mitigate water pollution. In this paper we present the case in favour of maintaining some long records of river water nitrate concentration at “benchmark” sites, in terms of enhancing process understanding and identifying system lags. Many long-term time series of nitrate concentration data share similar features including a pronounced seasonality characterised by a clear winter maximum, an upward trend in the post-war period followed by a levelling off, or even a decline in the last 20 years, and unusually high concentrations following drought years. Concentrations in any one year are often dependent on conditions in previous years; relationships can be established between concentrations and hydrological drivers (such as rainfall) with different lag periods which can yield information about supply or transport limitations to nitrate transfers. The interpretation of any record is dependent on its length: short records have a high potential for misinterpretation. Often, the value of long records only becomes apparent when analysed in retrospect, perhaps yielding insight into processes and phenomena for which the data collection programme was not originally designed. We, therefore, urge monitoring agencies to devise a strategy for maintaining long records – at least for a few benchmark stations.

---

T. P. Burt

Tim Burt is Professor of Physical Geography and Master of Hatfield College at Durham University. He specialises in catchment hydrology, water pollution and long-term monitoring of the natural environment.

N. J. K. Howden

Nicholas Howden is a Lecturer in the School of Applied Sciences at Cranfield University. A catchment hydrologist and hydrogeologist, he specialises in improving conceptual understanding and modelling of long-term catchment-scale biogeochemical process.

F. Worrall

Fred Worrall is Reader in Environmental Chemistry at Durham University. He specialises in biogeochemistry of catchment systems, with a particular interest in the generation of dissolved organic carbon (DOC) in upland river environments.

M. J. Whelan

Mick Whelan is a Senior Lecturer in the School of Applied Science at Cranfield University. He is an environmental chemist specialising in improving understanding of the fate and transport of pesticides, nutrients and other chemicals through river and groundwater systems.

---

Environmental impact

Nitrate is one of the most problematic all water pollutants, causing nutrient enrichment of water bodies and contamination of drinking water. It is now widely acknowledged that agriculture is the main source of nitrate pollution in surface waters and groundwater in rural areas of Western Europe and USA. Rising nitrate concentrations have been of concern for several decades, but there is still relatively little information about long-term trends, especially in groundwater-dominated catchment systems. This paper argues the importance of maintaining some benchmark records (>30 years) of river water nitrate concentrations such that long-term trends may be identified, characterised and understood. Such records are necessary to provide a context for policy-driven measures designed to reduce nitrate losses from agriculture.

Nitrate in water

Estimates suggest that human activity has doubled the rate at which biologically available nitrogen enters the terrestrial biosphere compared with pre-industrial levels.¹ Increased inputs of nitrogen to land have resulted in: a deterioration of drinking water quality and may have contributed to eutrophication in some receiving water bodies with consequent increased occurrence of algal blooms, reduced dissolved oxygen levels and loss of habitat.^2,3 Nitrate is also allegedly toxic to humans, although the evidence for this is controversial.⁴ It is, therefore, important to understand and quantify nitrogen exchanges to and from the terrestrial biosphere. A number of empirical approaches have been developed for predicting the influence of natural and anthropogenic changes on nitrogen transfers from land to water (e.g.ref. 5). However, most of these approaches have been calibrated using data for a limited number of years (commonly from the mid-1990s) from catchments where nitrogen fluxes could be increasing in response to changing anthropogenic influences.

High nitrate concentrations in rivers and groundwater have been a matter of concern throughout the developed world for several decades.^6,7 Initially, the main concern related to the potential effects of high nitrate concentrations in drinking water on human health, reflected in the European Drinking Water Directive (80/778/EEC). More recently, legislation (Nitrates Directive: 91/676/EEC) has been driven by concerns about the role of nitrate in the nutrient enrichment, or eutrophication, of surface waters, including coastal waters where primary productivity is often nitrogen-limited (e.g.ref. 8,9). In England and Wales, 61% of the nitrogen which enters surface waters is estimated to originate from agricultural land and 32% from sewage effluent,¹⁰ although these proportions will vary in different catchments. In rural catchments, most stream and river water nitrate is thought to originate from farmland >90%.^11,12

Monitoring river water nitrate

“We need good monitoring and research as a basis for all the [UK] Government's environmental policies. Without that we cannot base our decisions on the best available scientific and technological advice and analysis”.¹³

Among its eight definitions of the word monitor, the Shorter Oxford English Dictionary includes ‘(to) maintain regular surveillance’ and ‘something that reminds or gives warning’. In the context of river water quality, most monitoring relates to compliance with statutory requirements. Since the latter part of the 19^th Century, water supply companies in the UK have been required to report on the quality of raw water being abstracted from surface and ground water as well as the treated water being supplied to customers. Nitrate was often included in the list of water quality variables determined. In the 20^th Century, water company data have been supplemented by measurements taken by government agencies (such as the Environment Agency (EA) in England and Wales) in order to evaluate water quality in relation to the ecological status of the aquatic environment or to fulfil EU or other international obligations. In some places, data may also be available from research projects; usually these are for short periods of no more than one or two years, but rarely long-term studies like the UK Environmental Change Network (ECN: http://www.ecn.ac.uk/)^14,15 or the US Long-Term Ecological Research network (LTER: http://www.lternet.edu/) in the USA can build up much longer records. For some water bodies, extensive data archives have now been constructed. Although most recent data (such as that collected by the EA) are in digital format, older data are usually to be found only in paper records and require transcription before statistical analysis can be carried out. It should be noted that there are a number of potential issues with long-term data sets which must be considered prior to their analysis. These include the possibility of changes in analytical methods which may introduce inconsistencies in the time series (often identified as discrete discontinuities in values), changes in the frequency of sampling and changes in the location at which samples were collected.

Why are long-term data sets important?

For much of the 20^th Century, there was an expectation that many environmental systems (climatic, hydrological, ecological, geomorphological) had some characteristic average long term state about which there would be small fluctuations but for which statistical stationarity could be assumed.^16,17 Concerns about the potential for significant and relatively rapid anthropogenic change began to emerge, with respect to the global climate system, in the 1970s. More recently, there has been an increasing awareness about the possibility for change in other systems, either as a consequence of changing climatic drivers (e.g.ref. 18) or independently – for example, as a result of changing land use or land management or due to changes in atmospheric deposition (e.g.ref. 19,20). A number of approaches exist within which change can be analysed; one example is the DPSIR framework, which involves consideration of Drivers, Pressures, States, Impacts and Responses (see, for example, review by Svarstad et al.).²¹ Long data sets allow us to identify changes in the state or condition of monitored environmental systems and to evaluate the drivers for these changes. For example, it may be possible to attribute environmental deterioration (e.g. water quality changes resulting from increases in the discharge of sewage effluent) or improvements (e.g. from enhanced sewage treatment provision) to human activities and to evaluate the socio-economic and political factors which influence the direction of change (e.g. prices, regulations, behavioural change).

The fact that change is now expected as a matter of course in many systems has implications for the interpretation of experimental science (whether in the laboratory or in the field). Manipulative experiments, which can be repeated at any place, any time, are vital to an understanding of process and response. However, their very independence in time and space can place limits on their value for helping to predict the behaviour of environmental systems, given the influence of contingent events.²² Likens²³ therefore contends that short-term experiments cannot completely substitute for long-term studies, given the probability that prior conditions will impinge on the current state of the system. Long data sets allow the effects of prior influence on present conditions to be identified; this is the case irrespective of whether the monitoring programme was deliberately designed to address the possibility of long-term change or whether the data set has accumulated as a result of short-term surveillance. Pickett²⁴ points out that a documented past is labelled “history”, but even when it is undocumented, the past will still have had an influence on present and future trajectories of a given system. The goal of long-term monitoring is, therefore, to document the changing pressures, states, impacts and responses which would otherwise be lost to the historical record.

Accepting that change is commonplace and that documentation of change is desirable, three phenomena are profitably addressed by analysis of long time series:^25,24,14

1. Slow processes

Many environmental processes operate over relatively long periods of time, typically much longer than the span of a research degree or grant. The results of short-term studies can be misleading and may not reflect long-term trends because of the impact of transient conditions. This is particularly the case where the signal is weak or subtle: although a clear pattern may eventually appear, high-frequency variation will prevent its identification by short-term study.

2. Rare events

By definition, rare events do not occur very often. They may be periodic, and therefore to some extent predictable, but most are unpredictable. For the rarest, very long time series may be needed to intercept such an event. Given that low-frequency events are often high-magnitude (and, thus, may have a disproportionate influence on system response), the need for long, unbroken records becomes even more important.

3. Complex phenomena

Complex processes are those that have multiple influences and effects. Pickett²⁴ has argued that causality can only be evaluated in complex systems (large systems or areas) by observing the system over sufficiently long periods of time to encompass periods when different causes dominate its structure or function. Some of these causal factors may not have been identified a priori and, therefore, may not have been investigated via manipulative experiments.

What can long nitrate time series reveal about river systems?

In answering this question, the first three points are drawn from Lawton;¹⁶ the fourth point comes from our own work.²⁶

1. Long-term nitrate data can give warning of undesirable changes taking place.

By definition, monitoring can give warning of unwelcome change. Where regular surveillance is employed, this may well be aimed at identifying sudden (acute) changes requiring immediate action. However, the accumulation of a long record also allows slow, subtle (chronic) deterioration to be detected too – an added benefit almost certainly not thought about when the original monitoring system was established.

It was analysis of long-term records for a few English rivers that first identified rising nitrate concentrations as a cause for concern. Very high nitrate concentrations were observed in the very wet period that followed the 1975–76 drought in Britain (e.g.ref. 27). This led to analysis of the available long-term nitrate records and in turn to a very influential report by the Royal Society.²⁸Fig. 1 reproduces Figure 18 from the Royal Society²⁸ report and shows clearly how “unwelcome” changes were evident in the long-term trends in relation to the maximum acceptable concentration for nitrate in drinking water under the European Drinking Water Directive (80/778/EEC). Upward trends from the 1960s meant that episodes like the high post-drought nitrate concentrations in 1976–77 started from a baseline which was already high, with the consequence that concentrations were more likely to exceed the legal limit.

Fig. 1 UK rivers where long-term monitoring data are available for average annual nitrate concentrations (reproduced after Royal Society),²⁸ for the period 1928 to 1978.

2. Long-term changes in river water nitrate concentrations need explaining

The identification of temporal concentration changes in long nitrate time series requires explanation and has prompted the generation of hypotheses. Such changes may include trends, cycles, rare events, and major shifts in system behaviour. Data series, such as those shown in Fig. 1, have catalysed a wide spectrum of research activity, including field and laboratory studies of nitrogen turnover and nitrate leaching in soils, additional catchment monitoring studies and mathematical modelling of nitrate leaching and river nitrate concentrations (see reviews in Burt et al.).⁷

Fig. 2 updates Fig. 1, extending the data sets with concentrations measured over the last thirty years. The new data put those presented in the original Royal Society²⁸ report into a longer term context. Concentrations in the rivers Lee, Thames, Essex Stour and Bedford Ouse appear to have levelled off whereas concentrations in the river Frome, which is fed by groundwater from the Chalk aquifer, are still rising. These and other long-term trends are the subject of ongoing research. For example, export coefficient modelling has been used to estimate nitrate export from catchment areas, comparing estimates with observed nitrate concentrations, and predicting future nitrate concentrations under various land-use change scenarios.^11,29 Results suggest that much of the rising trend in the post-war period was caused by increased inputs of inorganic nitrogen fertiliser, augmented by nitrogen mineralisation in soil resulting from ploughing land which had been under grassland for a long period.³⁰Fig. 3 shows average rates of nitrogen fertiliser use on arable and grassland in England and Wales between 1943 and 1989 collated by Mittikalli and Richards.³¹

Fig. 2 UK rivers where long-term monitoring data are available for average annual nitrate concentrations (reproduced after Royal Society),²⁸ updated for the full period from 1928 to present.

Fig. 3 Estimated changes in fertiliser N use on arable land and grassland over the period 1930–1989 based on Mittikalli and Richards.³¹ Data for missing years have been estimated using linear interpolation.

Process-based models of nitrate transport have been developed at a variety of temporal and spatial scales. In general, models developed or applied at the field or small catchment scale (e.g.ref. 32) tend to be more complex than those employed to describe the behaviour of larger catchment areas, where the emphasis may shift from nitrogen cycling to transport processes (e.g.ref. 33). For large surface water catchments, there is an increasing need to take in-channel nutrient cycling and transport into account and to integrate inputs from point and diffuse sources. This is ignored in small-catchment models such as those described by Whelan et al.³² but is central to the INCA model.³⁴

Of course, long data sets demand time series analysis, both to decompose series into their constituent components (such as periodic variations and trends) and to provide prediction of future trajectories. For example, Howden and Burt³⁵ fitted equations based on the shape of a solute breakthrough curve to describe temporal changes in nitrate concentrations in the Frome and Piddle catchments in Dorset (UK) as an alternative to linear regression models. Solute breakthrough curves were fitted to 34 nitrate time series dating from the 1970s. Examples from four stations are shown in Fig. 4, fitted with linear and breakthrough curve equations. The results showed that S-shaped curves were able to provide a superior description of temporal change over linear models in most cases. Contrasting responses between headwater and downstream sites reflected differences arising from the contrasting influence of Chalk groundwater between the Frome and Piddle catchments. These analyses also moved the interpretation of rising nitrate concentrations from one which was statistically based to one with some hydrological meaning. In the case of the groundwater-dominated catchments, the appearance of high nitrate concentrations in stream water following land use-induced changes in nitrate leaching from the soil is likely to be considerably delayed by long travel times in both the unsaturated and saturated zones of the Chalk aquifer. The transport of nitrate through such systems is, therefore, a solute breakthrough phenomenon and the S-curve represents an appropriate physical description of system changes. It should be noted, however, that in other catchments similar S-shaped curves may also be useful for describing temporal trends in stream water nitrate concentrations, although solute breakthrough per se may not be the main process responsible for the trend.

Fig. 4 Example of fitting linear trends and breakthrough curves to observed annual average nitrate concentrations in the Frome and Piddle catchments in Dorset, UK (after Howden and Burt).^35,36

The Slapton Wood catchment is a small mixed land-use catchment in Devon (UK), where previous analysis of nitrate concentrations since 1970 had indicated an upward trend³⁷ followed by a relatively stable period.³⁸ Although the observational record dates back only to 1970, Burt et al.¹¹ estimated nitrate concentrations as far back as 1930 using an export coefficient model. The combination of modelled nitrate concentrations from 1930 to 1970 with measured concentrations thereafter suggests that the temporal pattern of stream nitrate concentration at Slapton also takes the form of an S-shape, placing the previously identified “trend” into a longer term context (Fig. 5). The soil leaching response, likely to take the form of an S-shaped solute breakthrough curve, will be attenuated by the delays caused by flow through shallow groundwater systems. It is doubtful that the breakthrough curve at Slapton is purely the result of solute transport delays (as is the case in the Frome and Piddle) because of the relatively rapid response of nitrate concentrations in the Slapton Wood stream to hydrological events such as the 1976 drought and subsequent wet winter. However, it may be that the previous assumption that the hydrology and hydrochemistry of this catchment are controlled purely by shallow subsurface flow may need to be revised to give more room for deeper subsurface flow with longer residence times. This was something anticipated by Chappell and Franks³⁹ from an analysis of hydraulic conductivity but difficult to corroborate without independent evidence. The potential for long solute residence times in groundwater and throughflow pathways has been illustrated by the results of stable isotope studies in the Maimai catchment in New Zealand (e.g.ref. 40) and by tracer studies such as that described by Owens and Edwards⁴¹ who observed bromide concentrations above background levels in groundwater samples for approximately 10 years after a single application of bromide salt to a small catchment. Maximum concentrations were not observed until 3 years after application and concentrations remained above background for over ten years.

Fig. 5 Breakthrough Curve fitted to data from the long-term monitoring data from Slapton Wood in Devon. Here, the time series of observations from 1970 has been extended back to the 1920s using an export coefficient approach.

3. Long-term nitrate data are essential for testing hypotheses undreamt of at the time the monitoring was set up

Inevitably, new questions arise in science and, whether a monitoring programme was designed for some scientific purpose or not, long time series can sometimes allow new questions to be addressed. For example, Burt et al.³⁷ assessed trends and variability in annual mean nitrate concentration in the Slapton Wood catchment over the period 1971–1986 using multiple linear regression. They predicted observed annual mean nitrate concentration on the basis of year, annual rainfall (R) and annual rainfall lagged by one year (R_[-1]) and by two years (R_[-2]). Where statistically significant (p<0.05) positive correlations were observed between concentration and R_[-n], it was suggested the system was “transport limited”, i.e. the nitrate concentration in the river was controlled by how much nitrate could be flushed from soil stores by hydrological processes operating in the catchment. Where statistically significant negative correlations were observed between concentration and R_[-n], the system was inferred to be “supply limited”, i.e. the availability of nitrate in soil stores was assumed to limit nitrate transfer to the stream, so that as the amount of water available for flushing solute through the catchment increased, the observed nitrate concentration decreased. The analysis suggested that, after accounting for the trend, nitrate concentration was controlled by antecedent rather than concurrent conditions. In other words a ‘memory effect’ was evident, in which relatively dry years were followed by higher concentrations than expected in subsequent years and vice versa. Burt and Worrall³⁸ presented a re-analysis of data from Slapton Wood using a “moving window” approach.⁴² Trends (1971 to 2005) in the observed annual average nitrate concentrations were again considered using linear regression models. The results showed that the upward trend in nitrate concentrations reported by Burt et al.³⁷ had not continued and that the ‘memory effect’ had reversed in sign, suggesting a change in the balance of underlying processes and a switch from supply- to transport-limited controls on nitrate leaching (see Fig. 6). Howden⁴³ has argued that the change in ‘memory effect’ could be partly attributed to the “moving window” methodology and that, to identify a change, comparison between two “moving windows” should require the window pair to contain mutually exclusive groups of years (e.g. for a 35-year time series (1971–2005) analysed with a moving window, the window 1971–1985 could only be compared with windows starting in 1986 or later). Since the 15-year moving windows illustrated in Fig. 6 are mutually exclusive for windows at the beginning and end of the time series, the memory effect reversal could be confirmed. This approach has led to a more thorough analysis of the concepts underlying the moving window analysis so that future interpretation of long time series can be more robust.⁴³ Subsequently, Burt et al.²⁶ applied the moving window methodology to a 70-year time series of annual mean nitrate concentrations in the river Stour in Essex, using window widths of 5, 10, 15 and 30 years, demonstrating short-term but not long-term reversals in memory effect in that case.

Fig. 6 Partial correlations between annual average nitrate concentrations and year (y) and rainfall (r) with the latter lagged by one (r-1) or two (r-2) years. Variable held constant denoted in CAPS.

Further examples where hypotheses have been generated about very different water quality phenomena from those for which the original monitoring programme was set up include:

- identification of system instability following deforestation which was hypothesized to result from N-saturation.⁴²

- hypotheses relating the relative importance of climate and large-scale land use change in the Mississippi basin from an analysis of alkalinity records since 1904⁴⁴ and nitrate flux since the 1960s.⁴⁵

4. Long-term nitrate data allow recent observations to be placed into a longer context

Given the way in which short-term variability can mask long-term change, the ability to place a short observational record into its historical context is invaluable. Burt et al.²⁶ studied the patterns of short and long-term change in nitrate concentrations in the River Stour (Fig. 7). The record (from 1937 to present) is typical of many lowland rivers in England, showing a progressive increase from the 1940s to the beginning of the 1980s followed by an apparent very gradual decline. The upward trend largely reflects the intensification of agriculture during and after World War II. The subsequent stabilisation of concentrations and apparent decrease appears to have started in the early 1980s, before any nitrogen control measures were introduced (e.g. Nitrate Vulnerable Zones Designation to meet the requirements of the EU Nitrates Directive, 1991). Current levels of nitrate leaching appear to reflect a new steady state which could be a function of relatively intensive present-day land use and management methods. If this is the case, then a significant reversal of the post-war increase in concentrations may not be possible without a radical shift in cultivation practices with potential implications for food production, price and domestic self-sufficiency. Such a shift would undoubtedly be politically unpopular with many farmers. Climatic fluctuations induce high-frequency variations around the long-term trend. Use of moving window analysis showed that statistically significant trends could be identified for almost any time period, with short-term trends reflecting climatic drivers and long-term trends reflecting broader shifts in nitrogen supply. Burt et al.²⁶ argued that the long-term response times for nitrate in this catchment was of the order of 20–30 years (which corresponds approximately with the time required for re-equilibration of soil organic matter levels in ploughed-out pasture).³⁰ Management decisions made on the basis of shorter data sets (even extending up to 15 years) could, therefore, be misleading: interventions could be judged unsuccessful, even if they had, in fact, had significant positive impact in the long term, and vice versa. However, as Fig. 7 shows, generally, long-term trends begin to emerge after a decade. It is once again clear that the historical context is crucial: certain years are very influential in the time series (e.g. very high concentrations after the droughts in 1976 and 2003) and can have disproportionate influence over short-term trends. Any policy-related changes are bound to be contingent on the precise time at which they are introduced in relation to pre-existing trends or particularly influential hydrological events. Short-term fluctuations in the Stour nitrate concentrations since the 1980s are more likely to reflect climatic variation than the impact of any changes in farming practice therefore.

Fig. 7 Nitrate concentrations measured on the River Stour in Essex 1937 to date.

How much data do we need?

It is too easy to argue that, because a long record has been built up over the years, it must be maintained in perpetuity at all costs. Of course, where resources are scarce, tough decisions must be taken. There does, however, seem to be a lack of a collective strategy within which to consider whether to continue long-term monitoring at a particular site or not. Given that the true value of any record is often only apparent in retrospect, there seems great value in maintaining a few “benchmark” sites for which very long records already exist. Thus, guidelines are needed which will allow crucial data sets to be identified and protected as a national asset. Criteria should include: length of record, rarity of similar records within the region, the potential relevance of what is being monitored to broad questions of global change. This may be an area where an alliance of government agencies such as the Environment Agency, Meteorological Office and Natural England, perhaps under the auspices of a learned society, is needed to take a collective view, identify crucial data sets, and ensure ongoing funding. We do not advocate mindless continuation of records for their own sake, but simply an evaluation of long records against agreed criteria before any decision is taken to stop measurements.

We advocate the establishment of national and international data observatories in the public domain where long time series can be properly archived. This would have the added advantage of serving as a reminder to monitoring agencies that they must sustain a long-term perspective and ensure that benchmark sites continue to be sampled at high resolution, ideally one in every large river basin. Virtual observatories have several benefits: they permit new modelling developments by ensuring that relevant data linking long-term trends in water quality to socio-economic and climate change are readily available (cf. ref. 45). Such information can also be used to improve the public understanding of science, including its complexity, and to raise environmental awareness by demonstrating that potentially unwelcome changes are occurring. Such observatories would also inform government understanding of the value of long-term monitoring, particularly the point that there are no short cuts or quick fixes.^46,47 So, for a very few sites where valuable long records have been accumulated, it is truly necessary to advocate more of the same. In other cases, we may very well have to accept that resources cannot be found to maintain the site and that recording should cease.

In relation to current and future research agendas, there must be a focus on hypothesis-driven experimental science, of course, but this is not sufficient in itself and analysis of long data sets can provide important context, raise new questions to be explored and help us to understand the significance of experimental results for understanding the larger problem. Future research agendas must include consideration of long time series to the extent that these provide evidence for long-term change. For example, there appears to be a very strong link between changes in nitrate concentrations in ground and surface waters and changes in land use and land management but this is often associated with significant time lags. Without a long term context, there would be little evidence for these connections or for the benefits of management interventions designed to mitigate diffuse-source transfers. The current environmental agenda is dominated by the climate change question. From a catchment management perspective, it is important to improve our understanding of changes in flow regime, water quality and associated ecological status (cf. EU Water framework Directive) under expected climate change. This can be achieved by deriving relationships between past hydrochemical and hydroecological processes using existing long-term records, combined with data and understanding derived from process studies. This is crucial for developing cost-effective decision-support systems for sustainable water resource management which are appropriate for the present, and for informing mitigation, adaptation and restoration strategies which may be needed as our climate changes. Since such analysis depends on existing long records, by definition there needs to be some continuation of at least a few of these records to permit similar analyses in the future. We therefore need to be vigilant and ensure that monitoring networks are properly assessed and that no long-term record is discontinued without proper consideration of its ongoing merits. We do not advocate “more of the same” in every case, but protection of a few benchmark sites is essential.

References

1. J. N. Galloway, F. J. Dentener, D. G. Capone, E. W. Boyer, R. W. Howarth, S. P. Seitzinger, G. P. Asner, C. C. Cleveland, P. A. Green, E. A. Holland, D. M. Karl, A. E. Michaels, J. H. Porter, A. E. Townsend and C. J. Vorosmarty, Nitrogen cycles: past, present and future, Biogeochemistry, 2004, 70, 153–226.
2. R. F. Turner and N. N. Rabalais, Coastal eutrophication near the Mississippi River delta, Nature, 1994, 368, 619–621.
3. P. M. Vitousek, J. D. Aber, R. W. Howarth, G. F. Likens, P. A. Matson, D. W. Schindler, W. H. Schlesinger and D. G. Tilman, Human alteration of the global nitrogen cycle: sources and consequences, Ecological Applications, 1997, 7, 737–750.
4. T. M. Addiscott and N. Benjamin, Nitrate and Human Health, Soil Use and Management, 2004, 20(2), 98–104.
5. S. P. Seitzinger, J. A. Harrison, E. Dumont, A. H. W. Beusen and A. F. Bouwmen, Sources and delivery of carbon, nitrogen, and phosphorus to the coastal zone: an overview of Global Nutrient Export from Watersheds (NEWS) models and their application, Global Biogeochem. Cycles, 2005, 19, GB4SO1.
6. T. M. Addiscott, A. P. Whitmore and D. S. Powlson. ( 1991). Farming, fertilisers and the nitrate problem. C.A.B. International, Wallingford.
7. T. P. Burt, A. L. Heathwaite and S. T. Trudgill. ( 1993). Nitrate: Processes, Patterns and Management. Wiley: Chichester.
8. R. J. Sanders, T. Jickells, S. Malcolm, J. Brown, D. Kirkwood, A. Reeve, J. Taylor, T. Horrobin and C. Ashcroft, Nutrient fluxes through the Humber estuary, Journal of Sea Research, 1997, 37, 3–23.
9. C. J. Gobler, G. E. Boneillo, C. J. Debenham and D. A. Caron, Nutrient limitation, organic matter cycling, and plankton dynamics during an Aureococcus anophagefferens bloom, Aquatic Microbial Ecology, 2004, 35(1), 31–43.
10. D. T. E. Hunt, A. S. Dee, D. B. Oakes. ( 2004)Updating the estimates of the source apportionment of N to UK waters. Phase 2. DEFRA report by WRc plc. http://www.fwr.org/defrawqd/wqd0002.htm.
11. T. P. Burt, A. L. Heathwaite and P. J. Johnes, Stream water quality and nutrient export in the Slapton catchments, Field Studies, 1996, 8(4), 613–627.
12. H. Casey, R. T. Clarke and S. M. Smith, Increases in nitrate concentrations in the River Frome (Dorset) Catchment related to changes in land use, fertiliser applications and sewage input, Chem. Ecol., 1993, 8, 105–117.
13. HMSO ( 1990). This common inheritance. White paper on the environment. Her Majesty's Stationery Office: London.
14. T. P. Burt, Long-term study of the natural environment: perceptive science or mindless monitoring?, Progress in Physical Geography, 1994, 18, 475–496.
15. J. M. Sykes and A. M. J. Lane. ( 1996). The United Kingdom Environmental Change Network: Protocols for Standard Measurements at Terrestrial Sites. Her Majesty's Stationery Office, London.
16. J. H. Lawton. ( 1990). Problems and challenges of monitoring change. In Heal, O. W., Sykes, J. M and Howson, G., editors, Environmental Change Network: report on a workshop on the terrestrial network. Merlewood: Institute of Terrestrial Ecology, 9–14.
17. J. D. Phillips, Changes, perturbations, and responses in geomorphic systems, Progress in Physical Geography, 2009, 33, 17–30.
18. N. Gedney, P. M. Cox, R. A. Betts, O. Boucher, C. Huntingford and P. A. Stott, Detection of a direct carbon dioxide effect in continental river runoff records, Nature, 2006, 439, 835–838.
19. A. H. Johnson, A. Moyer, J. E. Bedison, S. L. Richter and S. A. Willig, Seven Decades of Calcium Depletion in Organic Horizons of Adirondack Forest Soils, Soil Sci. Soc. Am. J., 2008, 72, 1824–1830.
20. C. D. Evans, D. T. Monteith, B. Reynolds and J. M. Clark, Buffering of recovery from acidification by organic acids, Sci. Total Environ., 2008, 404, 316–325.
21. H. Svarstad, L. K. Petersen, D. Rothman, H. Siepel and F. Wätzold, Discursive biases of the environmental research framework DPSIR, Land Use Policy, 2008, 25(1), 116–125.
22. L. R. Taylor. ( 1989). Objective and experiment in long-term research. In Likens, G. E., editor, Long-term studies in ecology: approaches and alternatives. New York: Springer-Verlag, 20–70.
23. G. E. Likens. ( 1989). Long-term studies in ecology: approaches and alternatives. New York: Springer-Verlag.
24. S. T. A. Pickett Long-term studies: past experience and recommendations for the future. In: Risser P G, editor, Long-term Ecological Research, SCOPE 47, Wiley, 71–88, 1991.
25. J. F. Franklin. ( 1989). Importance and justification of long-term studies in ecology. In Likens, G. E., editor, Long-term studies in ecology: approaches and alternatives. New York: Springer-Verlag, 3–19.
26. T. P. Burt, N. J. K. Howden, F. Worrall and M. J. Whelan, Importance of long-term monitoring for detecting environmental change: lessons from a lowland river in south east England, Biogeosciences, 2008, 5, 1529–1535.
27. I. D. L. Foster and D. E. Walling, Effects of 1976 drought and autumn rainfall on stream solute levels, Earth Surface Processes and Landforms, 1978, 3, 393–406.
28. Royal Society ( 1983). The Nitrogen Cycle of the United Kingdom. Royal Society, London.
29. T. P. Burt and P. J. Johnes, Managing water quality in agricultural catchments, Transactions of the Institute of British Geographers, NS, 1997, 22(1), 61–68.
30. A. P. Whitmore, N. J. Bradbury and P. A. Johnson, Potential contribution of ploughed grassland to nitrate leaching, Agric. Ecosyst. Environ., 1992, 39, 221–233.
31. N. M. Mittikalli and K. S. Richards, Estimation of Surface Water Quality Changes in Response to Land Use Change: Application of The Export Coefficient Model Using Remote Sensing and Geographical Information System, J. Environ. Manage., 1996, 48, 263–282.
32. M. J. Whelan, M. J. Kirkby and T. P. Burt. ( 1995) “Predicting Nitrate Concentrations in Small Catchment Streams ” in Trudgill S. T. (ed.) Solute Modelling in Catchment SystemsJohn Wiley and Sons, Chichester, pp 165–192.
33. B. M. Jackson, H. S. Wheater, S. A. Mathias, N. McIntyre and A. P. Butler, A simple model of variable residence time flow and nutrient transport in the Chalk, J. Hydrol., 2006, 330, 221–234,  DOI:10.1016/j.jhydrol.2006.04.045.
34. P. G. Whitehead, E. J. Wilson and D. Butterfield, A semi-distributed Integrated Nitrogen model for multiple source assessment in Catchments (INCA): Part 1 – model structure and process equations, Sci. Total Environ., 1998, 210–211, 547–558.
35. N. J. K. Howden and T. P. Burt, Temporal and spatial analysis of nitrate concentrations from the Frome and Piddle catchments in Dorset (UK) for water years 1978 to 2007: Evidence for nitrate breakthrough?, Sci. Total Environ., 2008, 407(1), 507–526,  DOI:10.1016/j.scitotenv.2008.08.042.
36. N. J. K. Howden and T. P. Burt, Statistical analysis of nitrate concentrations from the Rivers Frome and piddle (Dorset, UK) for the period 1965–2007, Ecohydrology, 2009, 2, 55,  DOI:10.1002/eco.39.
37. T. P. Burt, B. P. Arkell, S. T. Trudgill and D. E. Walling, Stream nitrate levels in a small catchment in south west England over a period of 15 years (1970–1985), Hydrol. Processes, 1988, 2, 267–284.
38. T. P. Burt and F. Worrall, Non-stationarity in long time series: some curious reversals in the ‘memory’ effect, Hydrol. Processes, 2007, 21, 3529–3531.
39. N. A. Chappell and S. W. Franks, Property distributions and flow structure in the Slapton Wood catchment, Field Studies, 1996, 8, 559–575.
40. M. G. Sklash, M. K. Stewart and A. J. Pearce, Storm runoff generation in humid headwater catchments 2. A case study of hillslope and low-order stream response, Water Resour. Res., 1986, 22, 1273–1282.
41. L. B. Owens and W. M. Edwards, Long-term groundwater quality changes from a one-time surface bromide application, J. Environ. Qual., 1992, 21, 406–410.
42. F. Worrall, W. T. Swank and T. P. Burt, Changes in nitrate export due to ecological succession, land management and climate: developing a systems approach to integrated catchment response, Water Resour. Res., 2003, 39(7), 1177,  DOI:10.1029/2000WR000130.
43. N. J. K. Howden, Comment on “Burt T., Worrall F. 2007. Non-stationarity in long time series: some curious reversals in the memory effect. Hydrological Processes 21: 3529–3531”, Hydrol. Processes, 2008, 22, 2887–2889,  DOI:10.1002/hyp.7076.
44. Raymond et al., Anthropogenically enhanced fluxes of water and carbon from the Mississippi River, Nature, 2008, 451, 449–452,  DOI:10.1038/nature06505.
45. G. F. McIsaac, M. B. David, G. Z. Gertner and D. A. Goolsby, Nitrate flux in the Mississippi River, Nature, 2001, 414, 166–167.
46. D. Tilman. 1989: Ecological experimentation: strengths and conceptual problems. In Likens, G. E., editor, Long-term studies in ecology: approaches and alternatives. New York: Springer-Verlag, 136–57.
47. M. Holdgate. ( 1991). Foreword to Spellerberg, I.F., Monitoring ecological change.Cambridge: Cambridge University Press, xi-xii.

---

Footnote

† Part of a themed issue dealing with water and water related issues.

---