JEQ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 9 August 2006
Published in J Environ Qual 35:1764-1778 (2006)
DOI: 10.2134/jeq2005.0477
© 2006 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via ISI Web of Science (3)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Holden, J.
Right arrow Articles by Horton, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Holden, J.
Right arrow Articles by Horton, M.
Agricola
Right arrow Articles by Holden, J.
Right arrow Articles by Horton, M.
Related Collections
Right arrow Watershed and Landscape Processes
Right arrow Wetlands and Aquatic Processes
Right arrow Hillslope Analysis
Right arrow Preferential Flow

TECHNICAL REPORTS

Landscape and Watershed Processes

Impact of Land Drainage on Peatland Hydrology

J. Holdena,*, M. G. Evansb, T. P. Burtc and M. Hortona

a School of Geography, University of Leeds, Leeds, LS2 9JT, UK
b Upland Environments Research Unit, The School of Environment and Development, University of Manchester, Mansfield Cooper Building, Manchester, M13 9PL, UK
c Department of Geography, Durham University, Science Laboratories, South Road, Durham, DH1 3LE, UK

* Corresponding author (j.holden{at}leeds.ac.uk)

Received for publication June 2, 2006.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
There is a long history of drainage of blanket peat but few studies of the long-term hydrological impact of drainage. This paper aims to test differences in runoff production processes between intact and drained blanket peat catchments and determine whether there have been any long-term changes in stream flow since drainage occurred. Hillslope runoff processes and stream discharge were measured in four blanket peat catchments. Two catchments were drained with open-cut ditches in the 1950s. Ditching originally resulted in shorter lag times and flashier storm hydrographs but no change in the annual catchment runoff efficiency. In the period between 2002 and 2004, the hydrographs in the drained catchments, while still flashy, were less sensitive to rainfall than in the 1950s and the runoff efficiency had significantly increased. Drains resulted in a distinctive spatial pattern of runoff production across the slopes. Overland flow was significantly lower in the drained catchments where throughflow was more dominant. In the intact peatlands, matrix throughflow produced by peat layers below 10 cm was rare and produced <1% of the runoff. However, in drained peatlands, matrix throughflow in deeper peat layers was common and provided around 23% of the runoff from gauged plots. Macropore flow, the density of soil piping, and pipeflow were significantly greater in drained peatlands than in intact basins. Gradual changes to peat structure could explain the long-term changes in river flow, which are in addition to those occurring in the immediate aftermath of peatland drainage.


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
UPLAND PEATLANDS often tend to be source areas for flooding rather than sinks (Burt, 1995). Blanket peatlands typically have high water tables and the peat is able to store little additional water. Saturation-excess overland flow or near-surface throughflow can dominate the response of intact blanket peat during a rainfall event (Holden and Burt, 2002a, 2003a, 2003b). As a result, runoff generation from upland blanket peat is extremely flashy; peak flows are high and there is minimal baseflow contribution (e.g., Price, 1992; Burt et al., 1997; Evans et al., 1999; Holden and Burt, 2003a). On the other hand, stream discharge tends to decrease to very small amounts during periods of low precipitation, with important implications for ecology and water management (Holden and Burt, 2003b).

Many peatlands have been artificially drained as a response to increasing agricultural demand in marginal areas, the drive to create land for forestry, the demand for extraction of peat for horticultural and energy products, and the perception that peat drainage would alleviate flood risk. The peatlands of The Netherlands, Finland, Ireland, and the United Kingdom are among the most heavily drained in the world (Baldock, 1984). However, the effect of drainage on the hydrological response of these peats is poorly understood. There are a number of reports which suggest that drainage can result in either increased or decreased flood peaks (Conway and Millar, 1960; Burke, 1975a, 1975b; Moklyak et al., 1975, Robinson, 1980; Archer, 2003; Holden et al., 2004).

The impact of artificial drainage on the hydrological response of peatland catchments was first experimentally investigated by Conway and Millar (1960). The conclusions of their paper and deductions made from it by other authors have become standard reference points for many studies on the hydrology of land use changes, and in particular on the impact of land management change in peatlands (Robinson, 1985). The paper is still frequently cited even today. Conway and Millar (1960) worked on four small peatland catchments in the north Pennines, UK; two had natural drainage channels and two had artificial networks of surface drains (locally referred to as ‘grips’). They concluded that streamflow production in blanket peatlands was more rapid where artificial drainage had taken place. There was an increased sensitivity of runoff response to storm rainfall with peak flows both higher and earlier. In contrast, intact basins exhibited a smoother storm hydrograph with greater lag times. However, data presented in Conway and Millar (1960) showed that artificial drainage did not affect the yearly catchment runoff efficiency (proportion of annual precipitation produced as stream discharge) at the time of the study. This was also demonstrated by a re-analysis of a further 3.5 yr of data from the site (which were not included in the 1960 paper) by Robinson (1985).

Other drained peatland catchments have, on the other hand, exhibited greater discharge during ‘low flow’ periods with a less frequent occurrence of small discharges, but also with smaller flood peaks (e.g., Baden and Egglesmann, 1970; Mustomen and Seuna, 1971; Heikurainen et al., 1978; Robinson, 1980). These changes have sometimes been attributed to catchment ‘dewatering’ due to slow drainage of the peat pore waters after drains have been installed. The drained Glenamoy catchment in Ireland was estimated to lose 1000 mm of water per year (Burke, 1975a) through dewatering during the phase immediately following drainage. While lowering of the water table increases short-term (storm-event) water storage and makes the runoff response to rainfall less immediate, in the medium-term water is lost from the catchment. This, of course, was partially the intention of those who installed peatland drains, but regardless, in the long-term, as peatlands dewater, they are also liable to subside (e.g., Anderson et al., 1995) so that, in fact, the temporary increase in water storage capacity may be lost. The catchment response may then be more flashy with concomitant increases in flood risk. Most studies are not maintained over a sufficient length of time to establish whether these effects occur, but certainly relaxation times are an important element that have been ignored in most peat drainage studies. Where longer-term studies have been performed, other factors such as forest plantation and tree stand maturity have played a role in altering water budgets (Holden et al., 2004).

It is expected that the installation of open drains (ditches) in peat catchments will reduce the amount of overland flow (Holden et al., 2004). Traditionally, this is considered an effect mainly resulting from water table drawdown around each drain (Dunn and Mackay, 1996). However, ditches running parallel to slope contour lines may also reduce peat saturation lower down the slope by redirecting upslope flows into the ditch. Thus, the foot of a drained hillslope would no longer receive all of the water that would once have drained down the section of slope, resulting in a lower water table downslope and reduced spatial and temporal occurrence of overland flow. However, there have been no observations of such an effect. It is not known to what extent overland flow is reduced in drained catchments, and how drainage impacts the relative partitioning between overland flow and subsurface flow. This is important for helping to explain differences in river flow response, and also for understanding nutrient and carbon cycling processes in peatlands. This is particularly important since peatlands store over one third of the world's soil carbon (Holden, 2005b).

A lowering of water table in peat can result in peat decomposition. Once peat dries it often becomes hydrophobic and cannot regain its initial moisture content (Egglesmann et al., 1993). Subsidence and irreversible drying of peats has been noted as a problem following drainage in New Zealand (Bowler, 1980) and in the fens of south-east England (Gilman, 1994). Holden and Burt (2002c) found permanent structural changes to blanket peats in northern England subject to drought simulation in the laboratory. This led to changes in the hydrological routing of water through the peat with more throughflow and macropore flow, and less overland flow. It is not known whether such effects would occur in association with open-cut ditches. In addition to any impact on the water table, open-cut drains expose peat to summer desiccation and winter frost action on their walls and floor, which may further alter peat structure and promote more macropore and pipeflow. There is some evidence that such processes are affected by drainage. Holden (2005a), for example, showed that peat catchments with artificial drainage had significantly more soil piping than those catchments without drainage. Soil pipes are important hydrological agents in blanket peat catchments, and pipes in intact basins have been shown to transport more than 10% of the river flow (Holden and Burt, 2002b). It is not known whether they transport relatively more water in drained catchments.

This study aims to test whether there have been any long-term changes in stream flow within two drained blanket peat catchments compared to two control catchments. The two treatment catchments were drained in the 1950s, resulting in increased sensitivity to rainfall but no change in annual water budgets in the years immediately following drainage. In the paper, results from a re-gauging of the four Conway and Millar (1960) study catchments will be presented. The study aims to (i) examine whether there are long-term changes in catchment runoff efficiency and storm runoff response in drained peat catchments that were not observed in the immediate few years following drainage, (ii) measure and compare the partitioning of runoff in intact and drained peat basins to evaluate the impact of drainage on overland and subsurface flow production, and (iii) test the hypothesis that the open-cut drains result in structural change in peat and enhance bypass flow through macropores and pipes and that these changes could, in part, account for any long-term change in streamflow in peatlands.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
The four blanket peat catchments are located within the Moor House National Nature Reserve, north Pennines, UK, and are shown in Fig. 1 (54°41' N, 2°23' W) with characteristics described in Table 1. The climate is ‘sub-arctic oceanic’ with a mean precipitation of 1981 mm yr–1 (Holden and Adamson, 2001, 2002). The original lettering, shown in Table 1, used by Conway and Millar (1960) to identify the catchments will be used throughout this paper to avoid confusion. There are two intact catchments (L and G) and two artificially drained catchments (N and S). In the 1950s experiment, Catchment N was gauged at just one location. However, it can be seen from Fig. 1 that the catchment consists of two slopes: one containing the artificial ditch network (Nd) and the other with a complex eroded gully network (Ne). It was, therefore, decided to gauge these two areas separately. Amalgamating the record allowed comparison with the earlier dataset while the separate records from Catchment N allowed the two areas to be considered separately. A further complicating factor is that the vegetation of Catchment N was partly burned in 1950 before the study site became a nature reserve. Nevertheless, Catchment S provides a good unburned comparison.


Figure 1
View larger version (30K):
[in this window]
[in a new window]
  		Fig. 1. The study catchments on the Moor House Biosphere Reserve, northern England.

 

View this table:
[in this window]
[in a new window]
  		Table 1. Summary of catchment characteristics.

 
Precipitation
Precipitation was measured at the Moor House field station within 1.6 km of all study catchments using an autographic rain gauge installed in 1952. The gauge recorded precipitation on a time chart and these charts have been preserved for analysis. An automated tipping-bucket rain gauge recorded the time of each 0.2 mm of precipitation for the study from 1998 onward. A manual rain gauge of standard British Meteorological Office design was emptied each week as a check. None of these gauges were heated. Snowfall can occur during winter months and this can add considerable error to precipitation records, particularly when it drifts into gauges and results in overestimates of precipitation. Temperature and snow days have been recorded at the site since 1931 and 1952, respectively, so it is possible to exclude snow days from analysis of hydrographs as required. Since 1991 snowfall has automatically been monitored at the site using albedo as a surrogate as described in Evans et al. (1999). Precipitation was also gauged in the early and later periods of study by gauges within Catchments N and G. When checked against the main Moor House field station gauge, there was not more than 0.7% difference in totals.

Streamflow
Discharge from the 1954 to 1962 period was measured using 90° V-notch weirs equipped with chart recorders. Data were considered to be of good standard by NERC (1975) during their national study of flow records except for some problems of overtopping during very large floods and the lack of sensitivity to very low flows. Catchments Nd and Ne were gauged in April 1998 and Catchments S, L, and G were re-gauged in June 2002 using a pressure transducer recording at 5-min intervals with a precision of ±1 mm in stage, and double V-notch weirs with a 22.5° notch below a 90° notch. The area of each catchment was surveyed using a differential GPS and values were used to determine catchment runoff efficiency.

The annual runoff/rainfall ratio was calculated for each catchment based on water years (which conventionally run from 1 October to 30 September in the UK) for all data available at each gauging station. For each catchment, fifty single-peak storm hydrographs were analyzed during both study periods for which there are records at all four stations (1957 to 1962; 2002 to 2004). The same storms were analyzed at all stations. Rainfall and stage chart records from the earlier period were analyzed at 10-min intervals; data from the later period were analyzed using 5-min intervals. Some of the most extreme storms were not included because they were complex in form or related to rain-on-snow events. Storms where snow or freezing occurred were excluded from analysis. Storm percentage runoff was derived by subtracting pre-storm discharge from the storm hydrograph. This value was also used to determine recession time from peak flow back to pre-storm runoff. A measure of the shape of the storm hydrographs was calculated for each storm by dividing the peak storm discharge by the total storm discharge. Storm characteristics were available for the recent period for the Nd and Ne subcatchments as well as combined values for the N catchment.

Hillslope Runoff Production
Runoff from the peat during natural rainfall events was recorded between October 2002 and October 2004 on two open-ended hillslope plots per catchment (including Nd and Ne) using aluminum throughflow troughs channeled into tipping-bucket flow recorders (Khan and Org, 1997; Holden and Burt 2003a). Flow could enter each plot from upslope in an unrestricted fashion. The troughs were 1 m wide and carefully embedded into the vertical face of the peat at 1-cm, 5-cm, 10-cm, 50-cm, and 1-m depth and at the base of the peat layer (Holden and Burt, 2003a). This could be done with minimum disturbance because there was rarely obstruction by rocks or large roots, and hence a rigid aluminum trough could be slotted into the peat face. Data were recorded at 5-min intervals. Dividers, which were 5 mm thick, were made from Perspex and were inserted flush with the edge of each trough to prevent leakage from upper layers into lower troughs and to prevent lateral flow. The plots were chosen so that each had the same topographic index (slope length divided by slope angle) to allow a fair comparison.

On one slope plot (50 by 50 m) within each of the L, G, S, Nd, and Ne catchments, detailed water table and overland flow sampling was performed using an equally spaced systematic grid of 100 dipwells and 100 crest-stage tubes per slope (Holden and Burt, 2003a). Crest-stage tubes were placed so that they had water entry holes flush with the peat surface and were used to determine whether overland flow had been present or not (i.e., whether the tubes contained water or were empty). The crest-stage tubes and dipwells were checked every 2 wk for 2 yr. The crest-stage tubes were emptied using a syringe, thereby keeping disturbance to a minimum. Site trampling was kept at a minimum but inevitably may have had an impact on results, in that compaction may have lead to an overestimation of overland flow occurrence.

A tension infiltrometer similar to that designed by Ankeny et al. (1988) and described further by Holden et al. (2001) was used in each catchment. By setting up a small water tension, the device excludes flow from the larger pores within the peat. Changing the tension (e.g., from –12 cm to –1 cm) allows calculation of the amount of water that infiltrates into macropores compared to smaller pores (Reynolds and Elrick, 1991; Holden et al., 2001). It was placed on the surface within a 100 x 100 m plot in each catchment (with the same mid-point topographic index). Twenty tests were performed on the surface of each plot, with the position of each test chosen randomly. The proportion of flux through macropores was calculated using the techniques described in Holden et al. (2001) with macropores defined as those larger than 0.1 cm (i.e., with a 3-cm pressure head distinguishing between matrix and macropores).

Soil pipes were located using ground-penetrating radar (GPR). This geophysical technique was used successfully by Holden et al. (2002) and Holden (2005a) to identify peat pipes without disturbance. Transects across the main slopes of the catchments were assessed and the pipes were identified from the radargrams. Once the pipes had been mapped, pipe discharge was then manually measured during four storm events in all catchments where pipes discharged into streams (Holden and Burt, 2002b). It was not possible to measure the discharge from all of the soil pipes, particularly those that did not directly exit into the stream channel, but it was considered that the majority of larger pipes were sampled.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
Figure 2 plots the monthly efficiency (percent of precipitation produced as streamflow) of the catchments for the 2004 water year. There is a seasonal effect with lower efficiencies in the summer months when there is greater evapotranspiration (Gilman, 1994; Evans et al., 1999). Efficiencies during the winter months for all catchments are high with values in excess of 90% being recorded for Catchments S and Ne. The artificially drained S and gullied Ne catchments appear to have higher monthly runoff/rainfall ratios than the two control catchments and the drained Nd catchment.


Figure 2
View larger version (23K):
[in this window]
[in a new window]
  		Fig. 2. Monthly precipitation and proportion of rainfall produced as stream discharge in each of the study catchments for the 2004 water year. L and G are the control catchments.

 
Comparison of annual runoff efficiency values are given in Table 2. June, July, and August are the only months of the year when soil moisture deficits are likely at the site (Evans et al., 1999). Therefore, to remove yearly climate effects, the table excludes values from June, July, and August. There is no change in efficiency for the two control catchments when comparing the 2003 and 2004 water years with those from the earlier study period. However, there is a change in runoff from the two artificially drained catchments. Catchment S has a higher efficiency with values over 81.0% compared with values of 74.1 to 76.0% in the earlier period. For Catchment N, runoff values are lower than in the earlier period. The lack of change in the control catchment response suggests that this is not a climate-related effect. Precipitation totals and the seasonal distribution of precipitation during the recent period were also comparable to those during the earlier period. Catchment Ne has a much greater efficiency (73.8 to 75.8%) than Catchment Nd (57.6 to 65.2%) (Table 2).


View this table:
[in this window]
[in a new window]
  		Table 2. Runoff/rainfall ratio, %, for the study catchments excluding June, July and August.

 
Table 2 suggests that catchment behavior has changed over time in the drained catchments. However, Catchment S had increased runoff efficiency since the 1950s and Catchment N had decreased in efficiency. Indeed the drained Catchment Nd has the lowest efficiency of all five gauging stations, whereas drained Catchment S now has the highest.

Figure 3 shows runoff from the five stream gauges for a 3-mo period during 2004. All catchments have a flashy response to rainfall with minimal baseflow and narrow, pointed hydrographs. Higher peak flows are recorded in the drained and gullied catchments while Catchment G has the most subdued response. While all catchments were very responsive to rainfall compared to most other small catchments (Burt, 1996) the control catchments appeared to be much less sensitive to rainfall than the other catchments.


Figure 3
View larger version (16K):
[in this window]
[in a new window]
  		Fig. 3. Rainfall and streamflow hydrographs for a 3-mo period during 2004 for each gauging site: (a) precipitation; (b) Catchment Nd, (c) Catchment Ne, (d) Catchment S, (e) Catchment L and (f) Catchment G. L and G are the control catchments. Precipitation and streamflow data are presented in mm at 5-min intervals for comparison.

 
As Fig. 4 shows, Catchment G is characterized by a much less peaked storm hydrograph than the other catchments. There are longer periods with intermediate flows when compared to other catchments. Figure 4 presents data from a series of storm events during a 4-d period in November 2003. The nature of the storm hydrographs and comparative times to peak can be seen. The two artificially drained catchments, Catchments Nd and S, tend to have higher peaks and peak earlier than the other catchments. Table 3 provides full details of mean storm parameters for each of the catchments. The control catchments have less spiky hydrographs (lower hydrograph intensity) and significantly longer lag times and recession times (p < 0.01) than the other catchments.


Figure 4
View larger version (27K):
[in this window]
[in a new window]
  		Fig. 4. Hydrographs from each gauging site during a 4-d period in November 2003. Rainfall and runoff is shown in mm per 5-min interval for each gauge for comparison.

 

View this table:
[in this window]
[in a new window]
  		Table 3. Mean storm hydrograph characteristics for the 1957 to 1962 period (old) compared to the 2002 to 2004 period (new). N = 50 for each case and the standard error of the mean for each parameter is indicated.

 
There have, however, been some changes in storm hydrograph characteristics between the 1950s and 2004 in Catchments S and N (Table 3). There was no significant difference between the two time periods for the undrained catchments for any of the storm parameters investigated. While the timing of the runoff for each storm would necessarily be dependent on rainfall intensity and duration, the number and range of storms covered in each study period allow for fair comparison. For the drained catchments, Catchment S (in 2004) had a significantly lower mean time to peak (time from rainfall onset to hydrograph peak) than in the 1950s, by about 20 min on average, and a shorter recession time. The hydrograph shape was also significantly less spiky (mean hydrograph intensity had reduced), yet the storm efficiency had risen significantly. For Catchment N, the mean peak lag time and time from centroid rainfall to peak discharge had significantly increased, whereas the mean hydrograph intensity and mean storm efficiency had significantly reduced.

Based on 2 yr of data, Table 4 shows the proportion of runoff partitioned between the peat layers produced at the runoff trough sites. There is a marked difference in runoff response between the intact and drained catchments. The intact catchments (L and G) are dominated by overland flow with very little runoff below 10-cm depth. However, the drained catchments produce a much greater proportion of runoff as subsurface throughflow from deeper peat layers. The gullied catchment (Ne) consists of some gullies that have incised by over 1.5 m into the peat. In this catchment 43% of runoff is produced from peat greater than 50 cm below the surface.


View this table:
[in this window]
[in a new window]
  		Table 4. Percentage of runoff collected in automated throughflow troughs from peat layers October 2002 to October 2004. Values rounded to nearest whole percent.

 
Example hydrographs from different peat layers are shown in Fig. 5 . Here runoff production from the peat is compared, as an illustrative example, between intact Catchment L (Fig. 5a) and drained Catchment S (Fig. 5b). Much more runoff is collected from the intact plot than from troughs on the drained plot. Despite the plots occupying similar points on the hillslope in terms of slope length and slope angle, the intact plot is able to receive water from the full extent of the upslope contributing area. However, the plot in Catchment S can only receive water from about 15 m upslope because ditches are spaced at that interval parallel with the contours. The peat responds rapidly to rainfall in both cases. Overland flow occurs at both sites, but the overland flow in drained Catchment S is much shorter-lived, with a narrower, more peaked hydrograph than that from Catchment L. Runoff from deeper layers plays a more dominant role in the overall runoff response of the plot in the drained catchment. While there is some slow seepage from the 5- to 10-cm layer following the storm in Catchment L, the runoff only amounts to between five and six percent of total flow. The same layer in the drained plot plays a more important role with 25% of storm runoff flowing from this peat layer. Contributions from deeper layers in the drained plot are smaller, but the hydrograph shapes are similar to those from upper layers.


Figure 5
View larger version (20K):
[in this window]
[in a new window]
  		Fig. 5. Runoff from automated plots during a storm event 28 Feb. 2003 to 1 Mar. 2004, rainfall and runoff data collected in 5-min intervals: (a) intact Catchment L, (b) drained Catchment S. Runoff volumes at the intact catchment plot were greater than at the drained plot because the upslope contributing area was larger in the intact catchment since it was not intercepted by drains. Therefore, different y-axes scales are used for (a) and (b).

 
Figure 6 shows the mean water table depth for the five monitored plots. The locations of the ditches are shown on Plots C and D. Plot E (within Catchment Ne) has been subject to erosion and therefore the surface is less uniform than the other sites. Mean water table depths are closer to the surface (usually within the upper 10 cm of the peat) in the intact plots and the water table depth is also more uniform. There is a slightly greater variability of mean water table depth in Plot B than in Plot A, which has a slightly less steep gradient. In the two plots where ditches run across the slopes, there are some distinctive patterns. Immediately downslope of a ditch, the mean depth to water table is lowest. However, further downslope from ditches, the depth to water table gradually increases, and the mean water table depth reaches a maximum around 2 m upslope of the next ditch. These patterns are also reflected in Fig. 7 and 8, which show the range in water table depth (distance between maximum and minimum depth of the water table during the study period) and the occurrence of overland flow across each plot. The water table range is greatest downslope of ditches and on some of the steeper parts of the eroded plot. In the intact plots, the water table range is at a minimum. Overland flow is much more frequent across the intact sites, having typically occurred between bi-weekly sampling on 80 to 100% of occasions, compared to much lower frequencies across drained plots. Immediately downslope of ditches (Fig. 8c and 8d) overland flow was not detected, or was rare. On the drained slopes, overland flow is more likely to occur just upslope of a ditch where water has accumulated from the upslope area between ditches and in association with a higher mean water table (cf. Figure 6). The low-gradient parts of Plot E have a high mean water table (Fig. 6e), low range (Fig. 7e), and high incidence of overland flow (Fig. 8e), whereas some of the steeper parts have much lower overland flow occurrences. The darker shaded areas of Fig. 8e represent locations where overland flow is more likely and hence during storm events it is more likely that these locations will be connected by overland flow.


Figure 6
View larger version (43K):
[in this window]
[in a new window]
  		Fig. 6. Mean water table depth for bi-weekly monitored plots October 2002 to October 2004 on (a) Catchment L, (b) Catchment G, (c) Catchment S, (d) Catchment Nd, and (e) Catchment Ne.

 

Figure 7
View larger version (43K):
[in this window]
[in a new window]
  		Fig. 7. Mean range in water table depth for bi-weekly monitored plots October 2002 to 2004 on (a) Catchment L, (b) Catchment G, (c) Catchment S, (d) Catchment Nd, and (e) Catchment Ne.

 

Figure 8
View larger version (47K):
[in this window]
[in a new window]
  		Fig. 8. Percentage occurrence of overland flow for bi-weekly monitored plots October 2002 to 2004, as a proportion of all visits on (a) Catchment L, (b) Catchment G, (c) Catchment S, (d) Catchment Nd, and (e) Catchment Ne.

 
The proportion of macropore flow to overall water flux into the peat within the experimental tension-infiltrometer plots was found to vary between 11.1 and 72.6% with a mean of 35.5%. Table 5 shows the mean values of the proportion of macropore flow in each catchment. Mann–Whitney U-tests showed that there were no significant differences between the L and G catchments and so the datasets were combined for statistical analysis. Similarly there were no significant differences between Catchments S and Nd and these too were combined and treated as the same dataset. Using combined data, it was found that macropore contribution to flux was significantly greater in drained Catchments S and Nd than the intact Catchments L and G (p < 0.001). The eroded Catchment Ne had significantly greater macropore flux than the open-cut drain catchments (p = 0.011).


View this table:
[in this window]
[in a new window]
  		Table 5. Percent contribution of macropore flow at the peat surface based on 20 samples from each catchment.

 
Similarly, pipe density was found to be greater in the disturbed catchments than in the intact ones, with Catchment Ne having the greatest pipe density at 86.3 pipes km–1 of GPR transect (Table 6). The proportion of streamflow emitted by pipes in the catchments follows a similar pattern with 9.6% in Catchment L, 13.2% in Catchment G, and at least 26.5% in the disturbed catchments where around one third of all streamflow is estimated to move through the pipe networks. In Catchment Nd there were two large pipes which were found to completely bypass the gauging structure. During the four sampled storms, these pipes produced discharge that would have increased the discharge measured at the Catchment Nd weir by about 9%.


View this table:
[in this window]
[in a new window]
  		Table 6. Pipe density as measured by the ground-penetrating radar (GPR) and proportion of streamflow manually sampled from pipes during four storms.

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
SUMMARY AND CONCLUSIONS
 REFERENCES
 
Drainage Impacts on Water Table
Drainage had a major impact on the spatial production of runoff and on the spatial pattern of water tables in the peat catchments monitored. The range of water table depths in peat, as well as the mean water table depth are both important for plant ecology in peatlands (Ingram, 1983). Ditches resulted in a change in the spatial pattern of both mean water table depth and its range of variation. The patterns identified in Fig. 6, 7, and 8 highlight the errors that would occur if traditional models of water table drawdown around peatland ditches were employed. These assume that there is an equal drawdown of water table on either side of a peat ditch as a simple function of the hydraulic conductivity of the peat and the depth of the ditch (e.g., Boelter, 1972; Dunn and Mackay, 1996). This may well be the case on flat peat, but where there is any slope there will be a topographic effect. The main impact of contoured ditches on water table and runoff production was witnessed downslope. Some ditches might, therefore, be more important than others in changing hillslope saturation and the propensity for overland flow simply due to their topographic position. Given that peatland scientists in many parts of the world are looking to block (infill) ditches and restore peatlands (Holden et al., 2004), this is an important point. In some areas (e.g., parts of northern England) there are thousands of kilometers of ditches (Lane et al., 2004) and it would be very expensive to block all these ditches. By taking topographic location into account, it should be possible to identify ditches that are of greatest importance and thus provide for more efficient targeting of resources for ditch blocking (Lane et al., 2003; 2004). For example, a dense ditch network on relatively flat terrain may be much less important to peat saturation and decomposition within the catchment than a few ditches running across a steeper slope. Peatland topography and ditch location should be taken into account when determining the priority locations for ditch blocking.

Drainage Impacts on Runoff Partitioning
Overland Flow
In drained catchments, much more of the runoff was produced by deeper subsurface flow than in intact catchments. Where overland flow did occur in the drained slopes, it was at the maximum distance downslope from the nearest upslope ditch, and was associated with a higher mean water table depth. This confirms earlier observations that infiltration-excess overland flow is not an important process in blanket peat (Holden and Burt, 2002a; 2003a, 2003b) because overland flow was only coincident with locations subject to saturation.

Matrix Flow
In intact catchments, despite saturation, lower peat layers do not produce much matrix flow because of their low hydraulic conductivity (Holden and Burt, 2003c; Rycroft et al., 1975). Slow drainage from deeper peat layers was enhanced in the disturbed catchments, indicating that the hydraulic conductivity is likely to have increased. Further work is required to establish whether hydraulic conductivity is affected by drainage.

Macropore Flow
Macropore flow, pipe density, and pipe flow were found to be significantly greater in the drained catchments than the intact sites. This is further evidence to demonstrate that structural change in the peat has changed the balance of hydrological processes operating at the study site. It is unlikely that these changes would have been immediately observed in the aftermath of drainage. Instead, the development of more macropores and pipes would have taken many years. This would explain why there has been a long-term change in flow response to drainage in these catchments that was not found during the 1957 to 1962 period. Changes to peat structural properties caused by desiccation are often not reversible (Egglesmann et al., 1993). Once pipe and macropore flow networks have developed, then peatland restoration (e.g., through ditch blocking) may not necessarily restore the peatland system to its earlier pristine state.

Drainage Impacts on Hydrograph Form
While in both intact and disturbed catchments, the stream runoff response is rapid, and the response in the drained catchments has a sharper peak. The hydrograph shape from the throughflow plots in each catchment was similar to that of the streamflow produced by each catchment. For the drained catchments there is a sharper overland flow response, which is short-lived, and this is represented by a narrower, more peaked hydrograph from the catchment outlet. The ditches act to efficiently remove runoff from the catchment. In the intact basins, overland flow response is more prolonged as drainage across the entire length of the hillslopes routes to the stream channel. While the intact peat catchments do have a less flashy response than the disturbed catchments, their response is still very rapid, being dominated by saturation-excess overland flow.

For the 2002 to 2004 dataset the two artificially drained catchments (Nd and S) had higher peaks and peaked earlier than the control catchments. Therefore, the results are comparable to those from Conway and Millar (1960). Robinson (1985) showed that topographical differences between the sites could not explain the differences in the earlier dataset, and that it was drainage that resulted in the flashier response. When comparing results from the two time periods, no difference in runoff efficiency or storm hydrograph response could be identified for the two control catchments. However, the response of drained catchments N and S to storm rainfall has changed over time. In both cases the hydrographs are less peaked with some increase in lag times compared to that in the years immediately following drainage. This might be expected over a period of 40 to 50 yr when ditches can become less efficient at removing water quickly from a catchment. Many ditches in peatlands have been reported to fill with vegetation or sediment (Robertson et al., 1968; Stewart and Lance, 1991; Fisher et al., 1996). Some of the ditches on Catchments Nd and S have partially revegetated with Sphagnum. This is occurring particularly on ditches that run across slope rather than down slope. However, the reduction in hydrograph peakedness may also be related to the long-term increase in throughflow in drained peats, as opposed to flashy saturation-excess overland flow.

Drainage Impacts on Runoff Efficiency
Results from this study suggest that drainage of Catchment S has resulted in a long-term increase in runoff/rainfall ratios despite no difference in the first few years following drainage. This result has never been reported from a drained peat catchment before. In most cases it is assumed that ditches that are not periodically cleared either become less effective over time (as they revegetate or infill with sediment), or the peatland will initially dewater in response to drainage. This would theoretically result in an increased annual runoff/rainfall ratio, which would decrease over time or stabilize once dewatering has ceased. The additional pipe and macropore flow in drained catchments explains why the water yield has increased in Catchment S. In contrast, drained Catchment N had shown an apparent decrease in water yield. This was found to be an artifact caused by pipe flow which now bypasses the gauging structure. In Catchment Nd, the gauged outlet of the contoured ditches was located at the confluence of the ditch network. However, this is not equivalent to the natural topographic flowpath for water from the hillslope had there been no ditches to divert flow. Therefore, if pipeflow develops across the slope, and this pipeflow does not feed into the ditch network, then flow will leave the catchment without passing the gauging structure. It was found that flow amounting to about 9% of gauged flow bypassed the Catchment Nd weir in this way. If it is assumed that these pipes were not present between 1954 and 1962 (although no observations are available to support this other than the paired catchment comparison of pipe density and macropore flow) and we add the bypassing pipeflow to the gauged flow, then we can say that there has probably been a significant increase in water yield from Catchment N since the 1950s (see Table 2). Indeed, the increase would probably have been even greater had there not been substantial revegetation of the eroded Ne subcatchment since the 1960s (Higgitt et al., 2001). The problem of flow bypassing the weir did not occur in Catchment S because the gauging site was located at the natural topographic outlet for the catchment and most bypassing flow was still likely to flow via the gauge.

Limitations
Comparisons between early data and data from the more recent period are likely to be effected by differences in the quality of measurement technique. Conway and Millar (1960) were careful to point out potential errors in their gauging techniques. We have had access to all of their original data, field books, and notes from 1954 to 1962 and have tried to eliminate periods of error where possible. However, the quality of discharge-gauging equipment has improved over the past fifty years and it is, therefore, likely that some of the measured differences could be related to differences in measurement technique. Furthermore, snow results in rain gauge catch errors. While we have attempted to remove the role of snow from the analysis by discounting snowmelt-related discharge events and periods when weir stilling pools were frozen, there may still be some snow-related errors in the study. There may also be differences in runoff production when there is frozen ground. The study is also limited in that no continuous data set was available since the 1950s for the four study catchments.


   SUMMARY AND CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 SUMMARY AND CONCLUSIONS
REFERENCES
 
This paper revisited the sites and data from the classic study by Conway and Millar (1960), which looked at the impacts of peat drainage on river flow. It has been shown that there are long-term changes to the hydrology, which are not effectively captured by short-term studies. Blanket peat catchments responded to rainfall with flashy hydrographs and short lag times and drainage of blanket peat resulted in an even greater sensitivity to rainfall with shorter lag times and flashier hydrographs than control catchments. In the few years immediately following drainage there were no differences in annual runoff efficiency between the treatment and control catchments. A re-gauging of the sites over forty years later revealed no change in the efficiency and storm hydrographs in the control catchments, but significant change in the treated catchments. In the drained catchments storm hydrographs were less flashy than in the years immediately following drainage, and the runoff efficiency was found to have increased. Also, it has been shown that ditches in peat result in changes to the spatial pattern of runoff production across hillslopes. The water table depth is increased in drained peat catchments more by topographic controls than by drawdown around individual ditches. There was a reduction of overland flow and relative increase in the proportion of matrix throughflow in drained peat catchments. There was also increased macropore flow and an increase in the density of soil pipes in drained peat catchments.

The above factors combine to explain observed long-term increases in annual runoff efficiency from drained peat catchments. The process response to peatland drainage is lagged and this may help explain the confusing range of existing data relevant to this problem and why some peat catchments experience long-term changes in river flow that do not occur in the immediate aftermath of drainage.


   ACKNOWLEDGMENTS
 
This work was carried out while J.H. was in receipt of a UK Natural Environment Research Council Fellowship NER/I/S/2001/00712. We thank English Nature for allowing us to work at Moor House National Nature Reserve. We also thank John Adamson of Environmental Change Network, Centre for Ecology and Hydrology, Lancaster, for access to the Moor House field data archives. The comments of a number of reviewers on earlier drafts of this paper were greatly appreciated.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via ISI Web of Science (3)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Holden, J.
Right arrow Articles by Horton, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Holden, J.
Right arrow Articles by Horton, M.
Agricola
Right arrow Articles by Holden, J.
Right arrow Articles by Horton, M.
Related Collections
Right arrow Watershed and Landscape Processes
Right arrow Wetlands and Aquatic Processes
Right arrow Hillslope Analysis
Right arrow Preferential Flow

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Vadose Zone Journal Journal of Plant Registrations
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal