How Do River Nitrate Concentrations Respond to Changes in Land-use? A Modelling Case Study of Headwaters in the River Derwent Catchment, North Yorkshire, UK
- 336 Downloads
A combined semi-distributed hydrological model (CASCADE/QUESTOR) is used to evaluate the steady-state that may be achieved after changes in land-use or management and to explore what additional factors need to be considered in representing catchment processes. Two rural headwater catchments of the River Derwent (North Yorkshire, UK) were studied where significant change in land-use occurred in the 1990s and the early 2000s. Much larger increases in mean nitrate concentration (55%) were observed in the catchment with significant groundwater influence (Pickering Beck) compared with the surface water-dominated catchment (13% increase). The increases in Pickering Beck were considerably greater than could be explained by the model in terms of land-use change. Consequently, the study serves to focus attention on the long-term increases in nitrate concentration reported in major UK aquifers and the ongoing and chronic impact this trend is likely to be having on surface water concentrations. For river environments, where groundwater is a source, such trends will mask the impact of measures proposed to reduce the risk of nitrate leaching from agricultural land. Model estimates of within-channel losses account for 15–40% of nitrate entering rivers.
KeywordsNitrate Water quality Model Catchment Land management Agriculture
Nitrate is a key nutrient within riverine environments and is of environmental concern in relation to issues of eutrophication and potable water supplies  and is profoundly important across Europe with regards to basin-scale management under the EU Water Framework Directive . There is considerable evidence of strong links between nutrient concentrations and loads in rivers and the proportion of agricultural land in the basins which they drain [37, 43]. Evidence is apparent in many basins , including the Humber, UK , a catchment of which, the Derwent , is the focus of a case study reported here. A wealth of experimental monitoring has revealed that the rate of applied fertiliser is likely to have a significant impact on nutrient losses in terms of exports from agricultural plots . Changes in agricultural practices have been linked to increased river nitrate concentrations in some Scottish catchments . The attribution of wider impacts in waterbodies, however, has often been reliant on modelling studies [50, 62] largely due to the paucity of clear evidence of uni-directional change in nutrient loadings.
A key question arises: in terms of nitrate concentrations, how quickly would effects of altered manure and fertiliser regimes or broader land-use modifications be seen in waterbodies? The answer is unclear and subject to many uncertainties such as prevailing weather conditions, crop yields, soil processes and hydrological pathways (reflected in catchment residence times and associated retention, retardation and release of nutrients). In particular, storage in groundwater is difficult to quantify. Whilst on a regional scale, it does not appear to exceed 3% of the rate of anthropogenically derived inputs, it may be more important in localised areas . In this respect, in-river processes are also important, determined by bioavailability, nutrient limitation and flow conditions and other physical factors. Even at the plot-scale, despite the direct evidence potentially available through the deciphering of data from controlled manipulation experiments, identification of long-term steady-state conditions following a change may not be possible. This is due to experimental plots undergoing continual management change in order to test a wide variety of new and ongoing scientific hypotheses.
Conceptual understanding of catchment-scale nitrogen dynamics is embodied in models with a detailed physical basis such as those described by Young et al. , Arnold et al.  and Wade et al. . In contrast, many models developed to address the consequences of strategic, policy-driven changes in land management to reduce nutrient loss are structurally simpler and largely based on empiricism. In this regard, the use of simple models is pragmatic and appropriate  given the difficulties (and introduction of associated uncertainties) in estimating a wide range of environmental parameters across geographically extensive areas. Such models typically work on the assumption that land-use-specific quantities of nutrient will be available for leaching and that this quantity is dependent on an assumed level of fertiliser and manure input (e.g. [3, 22, 44]). Despite having the important capability to provide insights of eventual steady-state conditions, estimating the dynamics of response is unlikely to be possible. Hence, when using such models of this type, it is important to make additional and explicit consideration of the likely timescales of change.
To quantify the likely change in NO3-N concentrations if the system were to have reached steady-state. Here, land-use changes are treated in isolation, the climatic effects being removed by using the same 1990–1993 meteorological datasets for both periods.
To indicate additional impacts of short-term climate on NO3-N concentrations, which has been shown to be important in a number of catchments in the north of UK using other models (e.g. ).
In the light of evidence of drivers and dynamics of long-term change summarised in Section 2, the predictions are evaluated and an explanation of changes in NO3-N concentration in the long-term offered. The research provides an example of the management issues for the Humber, a major basin within the UK and a key area for integrated environmental research and management (ChREAM project, ).
2 Review of Evidence of Long-Term Changes in Nitrate-N Leaching
Until the 1980s and the UK implementation of the EC Drinking Water Directive, there were many cases identified of long-term increases in river nitrate concentrations (nitrate as nitrogen: NO3-N) . These are summarised by Jose  and Johnes and Burt , largely indicating change in southern and eastern UK. The adverse river impacts were broadly coincident with increased loadings from point sources along with an intensification of agricultural activity, which is reflected and summarised by Sylvester-Bradley  and Davies and Sylvester-Bradley  who discuss trends in crop-specific N application and N surplus in the context of varietal genetic developments. A levelling off of leachable NO3-N since the mid-1980s is apparent. It should be noted that DON can also be important, representing around 20% of total nitrogen leached in some agricultural systems where rainfall input is high , and the fraction appears to become increasingly labile as systems become more saturated with nitrogen under increased loading .
However, it would be naive to assume that relationships between NO3-N leached and changes either in land-use (such as ploughing up of grassland) or fertiliser N applications are simple or manifested rapidly . A great wealth of data arising from plot experiments is of potential worth in this regard. However, these experiments had originally all been designed for a myriad of different purposes (e.g. identifying impacts of crop rotation and residual matter on N mineralisation and leaching) and were often ultimately focused on economic optimisation of fertiliser recommendation with knowledge of field history. In this way, the design of experiments to meet specific research objectives have often, for the purpose of identifying long-term change in leachable NO3-N, resulted in the introduction of confounding factors. Nevertheless, there is considerable evidence of complex responses at the field scale. For example, under cereals, tracer studies using 15N indicate the efficiency of crop utilisation, showing that less than 2% of applied inorganic nitrogen was present in soil in inorganic forms at harvest . Potato crops may be less efficient, yet Davies and Sylvester-Bradley  observed only 8% remaining at harvest. Even in long-term plots where fertilisation rates have been very high (e.g. 144 kg ha−1 per year) the contribution of N originally added as fertiliser (at any time not just in the given year) to the pool of leachable NO3-N is likely only to be around 50% . Davies  asserts that “the effect on NO3-N leaching of even large reductions in fertiliser N input would initially be comparatively small, only achieving substantial reductions after many years as organic N reserves and soil fertility decline”. As yields and crop residues have increased in tandem with development of new crop varieties in the past 100 years, there has been an attendant increase in soil organic matter accumulation which provides enhanced sources for slow and persistent generation (through mineralisation) of leachable NO3-N.
Some quantification of change in land-use from pasture to arable cropping and its effects has been made in small catchments; Worrall and Burt  pinpointed the importance of the unlocking of reserves of organic N present in grassland soils in regulating NO3-N concentrations. The link between renewed availability of organic N and an increased capacity for soil denitrification in response to increased supply was identified. Sequestration of N, apparently progressing at an exponentially decreasing rate, occurs when land reverts to pasture . Addiscott and Mirza  assert that the impact of reserves of organic material acting as sources and sinks of leachable NO3-N should be given due consideration in assessing and modelling catchment-scale river NO3-N responses.
At field scale alone, it may take up to 30 years to reach a steady-state in respect of leached NO3-N following a management change . Stabilisation changes in soil organic N may take much longer still. Notwithstanding the delaying and moderating effects of catchment-scale processes, including the impacts of groundwater residence time (which according to isotope studies of Bohlke and Denver  usually exceeds 20 years), these observations at field and small catchment-scale themselves have important implications for the feasibility of implementing current legislation drawn up to improve the chemical and ecological quality of waterbodies (the EU WFD: ).
Of relevance to the scale of large catchments, East European states have experienced 50–90% decrease in fertiliser use in recent years , with an increase in grassland at the expense of arable land. The changes have occurred since the end of the Communist regimes and provide a valuable case study of river nutrient impacts of broad-scale management shifts. From the limited data available, any evidence of steady-state having been reached is hard to find. In rivers draining to the Baltic, spatial surveying showed that largest NO3-N losses were actually observed in sites with largest N deficits. Mirroring many conclusions from field scale work, Vagstad et al.  identified that soil mineralisation rather than fertilisers are the main source of leached NO3-N. The delay in response to changing fertiliser regime is exacerbated by the dampening effect of groundwater retention allowing for deceleration of response (e.g. inefficient field drainage) and for denitrification to take place. Evidence from Latvia reported by Stalnacke et al.  suggests that stream denitrification is more efficient in the smaller streams and drainage ditches than in large channels. With a notable exception being the 157,000-km2 Tisza River in Hungary , riverine NO3-N trends in response to land-use change are more noticeable in smaller East European basins (< 1,000 km2). Examples include Slovakian studies  and the Estonian sub-catchments of Lake Peipsi . An overall lack of response in East European rivers to land-use change in the late 1980s suggests that reductions in NO3-N export from catchments are likely to be harder and slower to achieve than responses to increases in N loading. According to Grimvall et al. , this is due largely to sustained N release from humus layers and groundwater aquifers. Even the rapidity of increases is not always apparent, especially in large catchments, for example, as shown by Stow et al.  in North Carolina. Export coefficient modelling over long-term can illustrate the expected change in nutrient leaching and comparison with observed concentration trends. There is an example of such a study from the River Kennet, UK  where, arguably, river data show a 20-year lag in the marked increase attributed to land practice changes and population expansion in the 1950s and 1960s and that this has yet to reach a steady-state.
To summarise, the review reveals that change in NO3-N concentration will be slow and possibly incomplete, i.e. of lesser magnitude than the driving changes in N inputs. Delay and diminution of change is likely to be increasingly apparent on moving up from plot to catchment-scale due to the dampening effects of many catchment hydrological and biogeochemical processes. The model chosen for the Derwent study does not represent the dynamics of how such changes in input are manifested in terms of river NO3-N concentration. It can, however, be used to quantify an expected eventual steady-state condition.
3 Case Study Background: Catchment and Model Description
A hypothesis to guide the modelling case study is proposed: “over a 14 year period, changes in river NO3-N concentration can be satisfactorily explained by climate variability and changes in land-use within that time”.
3.1 Study Catchment
Catchment area (km2)
Base flow index
Rainfall (mm yr−1)
Land-use percentages in the Pickering Beck and River Dove
Pickering Beck 1990–1993
Pickering Beck 2000–2003
River Dove 1990–1993
River Dove 2000–2003
3.2 CASCADE/QUESTOR Application
3.2.1 Arable Land N Inputs
The first component to the N input (soil mineralisation) was calculated based on the concept that NO3-N available for leaching is supplied through the mineralisation of topsoil organic matter. Two pools of mineralisable N were considered: (1) readily decomposable material derived from plant residues (Ndpm) and (2) more recalcitrant (largely humic) material (Nrpm). Quantities were determined on a daily basis by first-order kinetics with different decay coefficients . The processes are dependent on soil temperature, and values appropriate to the Derwent were taken from Green and Harding .
N potentially available (kg ha−1) from crop residues (Ndpm) from main crop types found in Derwent catchment
Residual N, kg ha−1
Residual N, kg ha−1
Peas and beans
Values by soil HOST class of Nrpm (kg ha−1) specified on 1st September each year
In the period between crop harvest and rapid spring crop growth, the mineralised pools of NO3-N are available for leaching. However, to discriminate between spring and winter-sown crops, a temperature-dependent plant uptake factor was included . For winter-sown crops, this factor was applied to the combined pools of Nrpm and Ndpm in October and November of each year and serves to reduce the amount available for leaching in those months. When verified against data reported for various plot experiments , total net mineralisation rates (Nrpm + Ndpm) simulated by the model appear satisfactory.
In terms of managed additions to arable land (the second N input component), impacts of adding fertiliser are assumed to be manifested in the quantity of N in crop residue (as described above). If additions are made above the economic optimum, these quantities will start to increase significantly . National surveys of fertiliser use report annually on use of organic manures , revealing that cattle farmyard manure (FYM) is applied to 59% of fields and cattle slurry to 28% of fields. Data suggest that, though high in N, other manure types (e.g. pig and poultry manures and slurries) are only applied to a very small percentage of fields. For poultry manure, there is also high geographical uncertainty, in terms of where they are actually applied, as BSFP  data reveal that there is a significant level of importing of such material. For these reasons, organic manure types other than cattle are excluded from the calculation of N inputs. Under mean regional climate, the MANNER model  was used to determine the amount of NO3-N likely to be leached from a range of representative application quantities, timings and incorporation routines for both cattle FYM and cattle slurry. For each of the crop types considered, MANNER indicated that an additional 2 kg ha−1 per month of NO3-N available for leaching should be included for the months of November, December and January.
For the Derwent, atmospheric inputs (the third N input component) were assumed spatially invariant for the purposes of modelling. A value of 20 kg ha−1 per year is quoted by Smith et al. . For case study purposes the value was deemed unchanging on an annual time-step and was partitioned on a monthly basis using information on seasonal variability of atmospheric inputs . It was assumed that for nitrogen-fixing crops (peas and beans), farmers allow for the nutritional value of deposited N in their fertiliser calculations and monthly inputs from atmospheric N are not assumed to contribute to the NO3-N available for leaching between April and August inclusive.
3.2.2 Grassland N Inputs
For grasslands, HRU-resolution agricultural census data were used in conjunction with typical stocking density rates  to determine the proportions of grassland (temporary and permanent) used for beef, dairy and cutting systems, respectively. BSFP  data indicate that typical fertiliser application rates for grasslands are 155 kg N ha−1 and 85 kg N ha−1 for temporary (less than 5 years since last seeding/re-seeding age) and permanent (age greater than 5 years) pasture, respectively. These figures were used as input to the N-CYCLE model  which, for regional mean climate and estimated atmospheric N input, was run many times using a range of grassland ages and soil texture classes. In the absence of readily accessible historic records, census data were used to indicate the likelihood that land was previously arable rather than pasture prior to seeding. From the N-CYCLE applications, mean values of leachable NO3-N per annum for temporary and permanent grassland were determined for each of the three grassland systems. These annual values were uniformly distributed across the months at which soils are typically at hydrological field capacity (September–March inclusive). For the model applications described here, based on assessment of the spatial variability in abundance of the different grassland systems, uniform catchment-wide annual values of 21 and 10.5 kg ha−1 NO3-N available for leaching were deemed sufficient and used for temporary and permanent grassland, respectively.
3.2.3 Calibration of Hydrology
Whilst there were suitable data from the Dove and Pickering Beck in the two periods of interest, intermittent problems with flow data quality at one of the sites are apparent at other times. Therefore, for consistency, and to illustrate the level of robustness of a model-based method for identifying relations between land-use change and water quality, calibration was undertaken in two other catchments chosen to have similar characteristics to the Dove and Pickering Beck. This approach to calibration is in keeping with the primary purpose of the method as an exploratory tool in appraisal and management of large heterogeneous basins/regions.
For the River Dove, calibrated parameters were derived using 1987–1988 data from Hodge Beck, an adjacent catchment of similar rainfall and BFI (Table 1). In terms of hydrogeology, Pickering Beck is a challenge for rainfall-runoff modelling in that a limestone aquifer and impermeable grits both occur extensively. Calibrated parameters for Pickering Beck were derived in the drier catchment of the Derwent above Low Marishes (1991–1992 data), which has a somewhat higher BFI value (Table 1). In making this choice, a higher priority was assigned to the accurate simulation of low flows in Pickering Beck as opposed to storm flows. The dominance of sub-surface flow pathways is greatest at low flows and hence, total river nitrate loads are primarily determined by leaching from the soil matrix at such times. Nash–Sutcliffe model efficiency statistics however are more sensitive to errors at high flows.
A fuller account of the calibration process and the application of calibrated parameters across the entire Derwent catchment are given in Appendix 1. Model start dates were 1st Jan 1989 and 1st Jan 1999 for the 1990–1993 and 2000–2003 applications, respectively. This was to allow for equilibration of water and N in the soil stores, which started empty of water content and at roughly 50% of the annual total soil accumulated N (distributed 2:1 between the upper and lower layers). Warm up periods of 1 year were also used prior to the calibration exercises described above.
4.1 Summary Description
Climate inputs (rainfall, PET)
Pickering Model 1
Pickering Model 2
Dove Model 1
Dove Model 2
Pickering Model 3
Dove Model 3
- As a basis for modelling NO3-N, flow simulation is acceptable (in terms of flow duration curves: Fig. 3), although mean flows are underestimated.
- Land-use changes indicate an increase in NO3-N leaching. However, this is moderated by the diluting effect of wetter conditions in 2000–2003 (Table 6).Table 6
Summary statistics of modelled flow, NO3-N concentration and NO3-N load
Mean flow (m3 s−1)
Mean mg N L−1
kg N ha−1 yr−1
Pickering Model 1
5.90 (obs = 2.78)
17.1 (obs = 9.5)
Pickering Model 2
Pickering Model 3
6.25 (obs = 4.30)
22.5 (obs = 18.4)
Dove Model 1
3.72 (obs = 1.18)
14.7 (obs = 7.7)
Dove Model 2
Dove Model 3
3.63 (obs = 1.33)
16.8 (obs = 10.4)
Observed data show that mean values and the positive upward trend in NO3-N concentrations are much greater in Pickering Beck than in River Dove (Table 6).
Although simulation of flow percentile values is largely adequate, there is a universal tendency for the model to underestimate variability at the extremes, particularly in Pickering Beck. During 2000–2003, moderate flows were underestimated (more notably in the River Dove). The curves reflect substantially higher rainfall in the early 2000s compared with the early 1990s. A very low value for 1st percentile observed flow at Pickering Beck in 2000–2003 was due to temporary gauging problems giving spuriously low flow values which bias the statistics. The value is anomalous and should be disregarded. The failure to represent extreme conditions results in underestimation of high flows, and for both periods, mean flow is underestimated in both catchments (Table 6). Underestimation of high flows is generally more substantial in Pickering Beck, being an unsurprising consequence of the calibration procedure employed (outlined in Section 3.2.3). This is reflected in model efficiency statistics. Values of 0.15 and 0.30 for 1990–1993 and 2000–2003, respectively, are attained for Pickering Beck whereas higher values are attained for River Dove (0.53 in 1990–1993 and 0.63 in 2000–2003).
The N observations in the two rivers almost exclusively represent baseflow conditions (only six of 184 observations occurred at flows exceeding Q95). Both sub-catchments have in excess of 40% agricultural land and although Pickering Beck has a considerably higher proportion of tilled land (with higher levels of N in crop residues than other land-cover types) the differences in mean observed NO3-N concentrations are large. The observed increases in NO3-N concentration between 1990–1993 and 2000–2003 differ substantially in magnitude between the two catchments: in the Pickering Beck, an increase of 55% was seen whereas, in contrast, the River Dove showed only a 13% increase. In the Dove, it appears there were some higher observed concentrations in the intervening period (late 1990s) lasting only transiently, yet, at around the same time a step-change may have occurred in the Pickering Beck. The temporal increase in NO3-N load and concentration in Pickering Beck is captured by the model but not to the extent indicated by the observed data. Simulations indicate little change in the River Dove.
4.2 Impact of Propagated Errors on Model Output
When considering the Nash–Sutcliffe efficiency statistic, hydrological model performance for the two sub-catchments compares acceptably to other similar recent studies especially given that application was at sites different to where calibration was undertaken and furthermore, for different time periods. Other studies, comparable both in terms of the complexity of model structure and parameterisation and in terms of input data needs, cite performance under calibration or validation for the location at which calibration was carried out. For example, Bell et al.  achieved an efficiency value for the Derwent at Buttercrambe of 0.20 (compared with a value of 0.26 in our studies). Silgram et al.  report efficiency values of between 0.40 and 0.85 for diffuse pollution modelling applications in the nearby River Ouse catchment. More generally, for hydrological modelling in the UK using the water quality models NIRAMS, SWAT and INCA (by [21, 41, 77] respectively), a substantial number of efficiency values below 0.5 have been reported amongst the more prevalent higher values. Returning to this study, the CASCADE calibrations for Hodge Beck and Low Marishes (Derwent) yielded efficiency values of 0.66 and 0.77, respectively.
Modified simulated NO3-N concentrations and loads allowing for bias in flow modelling and in estimates of the fraction of soil nitrate leached (represented by the first and second adjustment multipliers respectively)
Original simulation (mg N L−1)
Original simulation (kg N ha−1)
First adjustment multiplier
Second adjustment multiplier
Modified simulation (mg N L−1)
Modified simulation (kg N ha−1)
Pickering Model 1
Pickering Model 3
Dove Model 1
Dove Model 3
Modelling the link between land-use and river N concentrations in small sub-catchments encompasses some uncertainties in catchment N residence time and also land-use mapping. Historic, as well as contemporary, land-use may influence river N content in the two periods of interest. For Pickering Beck the model includes two HRUs at the downstream end that only partially lie within the catchment delineated by the N sampling site. Splitting up HRUs is not justified (Section 3.2). These two HRUs, although representing less than 15% of the total modelled catchment area, are almost exclusively tilled land with much higher mean N inputs than the remainder of the catchment.
Achieving a close goodness-of-fit to observations is not the primary objective of this modelling study. Here, in line with Wade et al. , the main role of the model is to focus on improving understanding of the less well-defined components of the catchment system. More specifically, it is used to understand the role that actual land-use change has on impacts in terms of river NO3-N concentrations and how these impacts may differ between catchments. The model applications, as they stand, isolate impacts of land-use change and avoid confounding factors of in-river processes. Results permitted further isolation of confounding factors due to climate (e.g. diluting and flushing effects) to be made.
5.1 Differences Between Catchments
The differences in NO3-N concentration observed between the two catchments are larger than can be predicted by land-use factors alone.
- Seasonality in observed NO3-N concentration (1999–2004) differs markedly between the two catchments (Fig. 7a) and is only reproduced by the model in the River Dove, not Pickering Beck.
As annual rainfall patterns were similar in the catchments due to their close proximity and comparable relief, land-use factors are effectively summarised by the modelled annual mean loads. For example, if the loads simulated by Pickering Model 1 and Dove Model 1 are compared (Table 6), it is apparent that a difference of less than 20% might be expected. However, observed concentration data suggest that Pickering Beck has over twice the NO3-N load of River Dove. Clearly, there are additional sources of N contributing in the Pickering Beck but not in the River Dove.
Plots of mean NO3-N concentration for each month of the year (Fig. 7a) indicate summer maxima in the Pickering Beck but winter maxima in River Dove. The model structure, even when adapted to represent groundwater hydrological response, can only partially capture the seasonal NO3-N trends seen in Pickering Beck. Winter minima are simulated in periods when there is a preponderance of large winter dilution events in very high flow (as in 2000–2003). Otherwise, winter maxima are simulated (e.g. Pickering Model 1: Fig. 6). Summer maxima could only be fully reproduced in model output if an additional deep sub-surface source of N were to be included which conceptually would represent relics of historic N inputs percolating through the unsaturated zone. Winter maxima are typically seen in agricultural catchments [65, 80] with flow dominated by shallow sub-surface components which become enriched in NO3-N during the autumn and winter months. In this respect, seasonality is represented adequately by Dove Models 1 and 3, although the observed data indicate that winter peak N concentrations occur later than simulation suggests. This indicates a within-catchment retardation of N not captured by the model. A similar effect is seen in Pickering Beck (Fig. 6). Scatter-plots of flow at time of sampling against NO3-N concentration (Fig. 7b) clearly show an inverse relationship for the Pickering Beck and a positive relationship for the River Dove.
The large differences between the two catchments in terms of seasonal NO3-N dynamics reflect and support findings from other researchers (e.g. [43, 54]) suggesting the importance of groundwater sources in catchments which, like Pickering Beck, do not exhibit winter river NO3-N maxima. In Pickering Beck, the flow and seasonality-related relationships were much less strong pre-1998, which may reflect an increase in recent years in the groundwater NO3-N concentration (see Section 5.2).
5.2 Temporal Changes in Nitrate-N Concentration
are sufficient to adequately explain the small increase in NO3-N concentration in River Dove and
can only explain a minor part of the increase in NO3-N concentration in Pickering Beck.
Higher amounts of NO3-N available for leaching due to consideration of minor (high N use) crops and possible impacts of change in fertiliser usage on cereals and grasslands could account for an increase in NO3-N loads. To ease computational requirements, the CASCADE model does not consider all the possible combinations of soil type and land-use in a catchment. However, over 98% of land-use in the Pickering Beck sub-catchment was accounted for in the model and, in any case, potatoes and oilseed rape were both explicitly included. Currently available evidence on fertiliser use  does not suggest an increase in application rate for the main agricultural land-uses in the catchment.
An increase in numbers of livestock types not considered in the model application. The only appreciable change in the sub-catchment is in the lower reaches where an order of magnitude increase in poultry numbers has taken place. In conjunction with the impacts of land-use change between 1990–1993 and 2000–2003 (captured by the model application), such an increase could fully account for the observed increase in concentration in Pickering Beck. Estimates using the MANNER model  suggest poultry in the Pickering Beck catchment could contribute an approximate annual load to the river in excess of 3 kg ha−1 in the 2000–2003 period. Ordnance Survey maps at 1:25,000 scale show a large poultry farm situated very close to the river in the lower reaches of the sub-catchment. However, at the monitoring site, the median soluble reactive phosphorus concentration (that would in high concentration indicate the importance of such a point source input) is very low (40 μg L−1), and there is no indication of any temporal trend. The lack of P input to the river suggest that the poultry manure (typical N/P mass ratio of 2:1) has either not reached the channel in the time period under consideration or has been exported out of the catchment. The second explanation is most likely.
An increase in rates of nitrification and mineralisation processes in soils as a result of prolonged dry conditions experienced across England between 1995 and 1997. Evidence of elevated river NO3-N concentrations in the Windrush in 1998 has been attributed to this mechanism (particularly in woodland soils) by Moorcroft et al. . Extensive studies in the US  have also shown that nitrification is enhanced on drying and re-wetting of soils leading to marked year-on-year variability in N export and a clear decrease in N retention in wetter years. Despite this evidence, it would seem unlikely that such climate-related factors would result in a sustained longer-term trend.
Impact of higher groundwater NO3-N concentrations. There is much evidence to suggest that concentrations in major aquifers (of which for the UK the Jurassic Corallian group is included) have gradually been increasing in recent decades [11, 13]. Typically, present-day concentrations are up to 30% higher than in 1990 and are likely to violate EC Nitrates Directive levels. In many cases, sampling boreholes are impacted upon by water percolating from agricultural land through the unsaturated zone. High levels of NO3-N resulting from peaks in agricultural N surpluses in the 1970s are likely to eventually reach the water table and thenceforth the river network. The National Rivers Authority  reported an increase in groundwater NO3-N levels in samples from the Corallian limestone, indicating their motivation to prevent further degradation of the quality of the water resource. In the region, groundwater quality problems are of local concern to water companies . It seems highly probable that changes in groundwater NO3-N concentration are impacting upon current and recent NO3-N concentrations in Pickering Beck.
The fourth hypothesis, in conjunction with impacts of the change in land-use since the early 1990s, clearly provides the most plausible explanation for the large increase in NO3-N concentration in recent years in Pickering Beck. The lack of evidence of change in the River Dove, which is not impacted by Corallian groundwaters to any great extent, gives indirect support to the chosen hypothesis.
Application of the diffuse pollution model (CASCADE) to flow and NO3-N dynamics in two rural sub-catchments of the River Derwent has provided insights into the significance of current land-use on NO3-N impacts in rivers. This has been achieved by considering two land-use datasets approximately a decade apart. In both sub-catchments, the more recent land-use data should have resulted in higher potential leaching of NO3-N to waterbodies. Observed data reflect these differences, indicating evidence of increasing NO3-N concentrations. Use of the model allows the changes to be set in the context of current climate and hydrological controls and in-river processes. For example, in Pickering Beck, an increase of 29% (for both load and concentration) driven solely by land-use change is estimated by the model. When including climatic effects the load increase is higher (31%), yet, the concentration increase is just 6% due to the wetter conditions of more recent years. However, additional inclusion of in-stream processes (as represented in QUESTOR and illustrated for a location a short distance downstream), reduces the NO3-N load by approximately 30%. It seems likely that headwater streams and ditches in agricultural areas of the sub-catchments act as significant sinks for NO3-N. The process of denitrification can be exploited in the management of rivers of high NO3-N concentration. Stream restoration measures to increase flow path retention times , for example by allowing reconnection of channels and associated floodplains , lead to increased denitrification rates.
Returning to our case study, differing responses are seen in Pickering Beck and River Dove and, although in these cases the NO3-N concentrations themselves are not at levels to cause legislative concern, explaining the contrast has important implications for general understanding of long-term nutrient dynamics in catchments. Large increases in NO3-N concentration have occurred in Pickering Beck since the early 1990s. Modelling shows that these cannot be explained by changes in land-use observed in the intervening period. Furthermore, it would not have been expected that land-use change could explain the rapidity of increases in NO3-N concentration. This is because research evidence, summarised in Section 2, suggests sizeable moderating and delaying factors which are manifested as the integration of (1) soil profile biochemical N cycling processes and (2) catchment-scale transit times typical of groundwater-influenced systems. Effectively, a steady-state condition resulting specifically from the land-use change is unlikely to have been reached by 2004. The changes in the Pickering Beck are explicable, however, in the light of long-term increases in groundwater NO3-N concentrations. There is much concern in the UK over recent and ongoing increases in NO3-N in major aquifers, and it seems this is having a discernible effect on recent trends in the water quality of the Pickering Beck. Case study data and modelling indicate that there is not a significant groundwater component to flow in the River Dove. Nitrate-N concentrations show evidence of change, but to a lesser extent than that seen in Pickering Beck.
In terms of the hypothesis proposed at the start of Section 3 (“over a 14-year period, changes in river NO3-N concentration can be satisfactorily explained by climate variability and changes in land-use within that time”) we draw the following conclusions. For Pickering Beck, there is clear evidence of other factors and the hypothesis is refuted. The impact of groundwater flow contribution suggests that NO3-N leaching due to historic (pre-1990) land-use is still strongly influencing river concentrations. For the River Dove, despite the general literature evidence to the contrary the hypothesis cannot be refuted, although in reality there are probably additional drivers. However, there is a clear need for additional incorporation of within-channel sources and sinks of N, raising issues of model structure and equifinality of parameterisation.
The authors thank Andrew Johnson (CEH) for comments on the manuscript. EDINA at Edinburgh University Data Library and Defra are acknowledged as the sources of the Agricultural Census data. The work was underpinned in part by funds from CEH, NERC and the Environment Agency. The analysis undertaken in this paper is stimulated, in particular, by a new initiative, the Catchment Hydrology, Resources, Economics and Management (ChREAM) project, funded under the joint ESRC, BBSRC and NERC Rural Economy and Land Use (RELU) programme (award number RES-227-25-0024).
- 3.Anthony, S. G., Quinn, P., & Lord, E. I. (1996). Catchment scale modelling of nitrate leaching. Aspects of Applied Biology, 46, 23–32.Google Scholar
- 6.Bell, V. A., Kay, A. L., Jones, R. G., & Moore, R. J. (2007). Development of a high-resolution grid-based river flow model for use with regional climate model output. Hydrology and Earth Systems Science, 11, 532–549.Google Scholar
- 9.Boorman, D. B., Hollis, J. M., & Lilly, A. (1995). Hydrology of soil types: a hydrologically-based classification of the soils of the United Kingdom, Report no. 126. Wallingford: Institute of Hydrology.Google Scholar
- 10.British Survey of Fertiliser Practice. (2005). Fertiliser use on farm crops for crop year 2004. London: Defra.Google Scholar
- 11.Burkart, M. R., & Stoner, J. D. (2002). Nitrate in aquifers beneath agricultural systems. Water Science and Technology, 45(9), 19–28.Google Scholar
- 12.Burt, T. P., & Haycock, N. E. (1993). Controlling losses of nitrate by changing land-use. In T. P. Burt, A. L. Heathwaite, & S. T. Trudgill (Eds.), Nitrate: processes, patterns and management (pp. 341–368). Chichester: Wiley.Google Scholar
- 13.Burt, T. P., & Trudgill, S. T. (1993). Nitrate in groundwater. In T. P. Burt, A. L. Heathwaite, & S. T. Trudgill (Eds.), Nitrate: processes, patterns and management (pp. 213–238). Chichester: Wiley.Google Scholar
- 14.Centre for Ecology and Hydrology. (2003). Hydrometric register and statistics 1996–2000. Wallingford, UK: Centre for Ecology and Hydrology.Google Scholar
- 15.Chambers, B. J., Lord, E. I., Nicholson, F. A., & Smith, K. A. (1999). Predicting nitrogen availability and losses following application of organic manures to arable land: MANNER. Soil Use and Management, 15, 137–143.Google Scholar
- 18.Davies, D. B. (2000). The nitrate issue in England and Wales. Soil Use and Management, 16, 142–144.Google Scholar
- 24.European Union. (2000). Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for community action in the field of water policy. Official Journal of the European Communities, 327, 1–72.Google Scholar
- 26.Fuller, R. M., Groom, G. B., & Jones, A. R. (1994). The land cover map of Great Britain: an automated classification of Landsat Thematic Mapper data. Photogrammetric Engineering and Remote Sensing, 60, 553–562.Google Scholar
- 27.Fuller, R. M., Smith, G. M., Sanderson, J. M., Hill, R. A., & Thompson, A. G. (2002). The UK land cover map 2000: construction of a parcel-based vector map from satellite images. Cartographic Journal, 39, 15–25.Google Scholar
- 31.Green, F. H. W., & Harding, R. J. (1979). Altitudinal gradients of soil temperatures in Europe. Meteorological Magazine, 108, 81–91.Google Scholar
- 33.Hallett, S. H., Jones, R. J. A., & Keay, C. A. (1993). SEISMIC: a spatial environmental information system for modelling the impact of chemicals. In A. R. D. Stebbing, K. Travis & P. Matthieson (Eds.), Environmental modelling: The next 10 years. Report of a Symposium at The Society of Chemical Industry, London, 16 Dec 1992, pp. 40–49.Google Scholar
- 34.Heathwaite, A. L. (1993). Nitrogen cycling in surface waters and lakes. In T. P. Burt, A. L. Heathwaite, & S. T. Trudgill (Eds.), Nitrate: processes, patterns and management (pp. 99–140). Chichester: Wiley.Google Scholar
- 36.Heathwaite, A. L., Burt, T. P., & Trudgill, S. T. (1993). Overview—the nitrate issue. In T. P. Burt, A. L. Heathwaite, & S. T. Trudgill (Eds.), Nitrate: processes, patterns and management (pp. 3–21). Chichester: Wiley.Google Scholar
- 40.Hutchins, M. G., Deflandre, A., & Boorman, D. B. (2006). Performance benchmarking linked diffuse pollution and in-stream water quality models. Archiv fur Hydrobiologie Supplement-Large Rivers, 17(161), 133–154.Google Scholar
- 42.Institute of Hydrology. (1980). Low flow studies report. Wallingford, UK.Google Scholar
- 43.Johnes, P. J., & Burt, T. P. (1993). Nitrate in surface waters. In T. P. Burt, A. L. Heathwaite, & S. T. Trudgill (Eds.), Nitrate: processes, patterns and management (pp. 269–320). Chichester: Wiley.Google Scholar
- 44.Johnes, P. J., Moss, B., & Phillips, G. (1996). The determination of total nitrogen and phosphorus concentrations in freshwaters from land-use, stock headage and population data: testing of a model for use in conservation and water quality management. Freshwater Biology, 36, 451–473.CrossRefGoogle Scholar
- 52.Lord, E. I. (1992). Modelling of nitrate leaching: nitrate sensitive areas. Aspects of Applied Biology, 30, 19–28.Google Scholar
- 54.Martin, C., Aquilina, L., Gascuel-Odoux, C., Molenat, J., Faucheux, M., & Ruiz, L. (2004). Seasonal and interannual variations of nitrate and chloride in stream waters related to spatial and temporal patterns of groundwater concentrations in agricultural catchments. Hydrological Processes, 18, 1237–1254.CrossRefGoogle Scholar
- 56.Ministry of Agriculture Fisheries and Food. (2000). Fertiliser recommendations for agriculture and horticultural crops (RB209). Norwich: HMSO.Google Scholar
- 58.National Rivers Authority (1994). River Derwent catchment management plan consultation report (p. 97). Leeds: National Rivers Authority Northumbria & Yorkshire Region.Google Scholar
- 61.Nix, J. (2000). Farm management pocketbook. Ashford: Wye College Press.Google Scholar
- 64.Scholefield, D., Lockyer, D. R., Whitehead, D. C., & Tyson, K. C. (1991). A model to predict transformations and losses of nitrogen in UK pastures grazed by beef cattle. Plant and Soil, 132, 165–177.Google Scholar
- 66.Shaffer, M. J. (2002). Nitrogen modelling for soil management. Journal of Soil and Water Conservation, 57, 417–425.Google Scholar
- 68.Shields, C. A., Band, L. E., Law, N., Groffman, P. M., Kaushal, S. S., Sawas, K., et al. (2008). Water Resources Research, 44: article number W09416.Google Scholar
- 72.Stalnacke, P., Vandsemb, S. M., Vassiljev, A., Grimvall, A., & Jolanki, G. (2004). Changes in nutrient levels in some Eastern European rivers in response to large-scale changes in agriculture. Water Science and Technology, 49, 29–36.Google Scholar
- 80.Whelan, M. J., & Kirkby, M. J. (1995). Predicting nitrate concentrations in small catchment streams. In: S. Trudgill (eds), Solute modelling in catchment systems. Wiley, pp. 165–192.Google Scholar
- 85.Young, R. A., Onstad, C. A., Bosch, D. D., & Anderson, W. P. (1989). AGNPS: a nonpoint-source pollution model for evaluating agricultural watersheds. Journal of Soil and Water Conservation, 44, 168–173.Google Scholar