The Hydrological Impact of Low-Flow Alleviation Measures

Proceedings of the Institution of Civil Engineers Water Management 161 August 2008 Issue WM4 Pages 171–180 doi: 10.1680/wama.2008.161.4.171 Helen Clayton Susan Morris Neil McIntyre Marius Greaves Hydrologist, Jacobs, Reading, UK Senior Hydrologist, Jacobs, Reading, Senior Lecturer, Department of Civil Team Leader, Environment Agency, Paper 700051 UK and Environmental Engineering, Thames Region, Hatﬁeld, UK Received 02/10/2007 Imperial College London, UK Accepted 10/03/2008 Keywords: hydrology & water resource/ management/mathematical modelling

The hydrological impact of low-ﬂow alleviation measures

H. J. Clayton MSc, S. E. Morris MPhil, N. R. McIntyre PhD, CEng, MICE and M. Greaves MSc

The River Ver, a chalk stream in Hertfordshire, was In 1988, Halcrow6 concluded that low flows in the River Ver were identified as one of the ‘top 40’ low-flow rivers of attributable to the over-abstraction of groundwater for public England and Wales in 1992, and measures to improve water supply and proposed various mitigation options. flows were subsequently proposed under an alleviation of Consultation between the NRA, local action groups and Three low flows (ALF) scheme. Consequently, the largest Valleys Water PLC (a public water supply company) led to a groundwater abstraction within the catchment at Friars proposal to improve flows under the scope of the NRA’s Wash was closed, other than for emergencies. The alleviation of low flows (ALF) scheme. This was approved by impact of this closure on river flows as measured at a central Government and subsequently Friars Wash pumping gauge near the outlet of the Ver catchment is evaluated. station (PS) was closed on 26 May 1993. Despite the considerable A range of methods was employed: analyses of funding required to implement the scheme and the subsequent hydrographs, flow duration curves, flow percentiles, monitoring, little effort has been made to date to evaluate the baseflow index, correlations with data from an analogous success of the scheme in restoring low flows. The Environment catchment and rainfall–runoff modelling. Results suggest Agency (successor to the NRA), in its current role, has to fulfil an apparent increase in river flow in the 11 years since the commitments to local residents, manage the catalogue of closure at Friars Wash; however, this cannot be attributed environmental improvement work across the area and formulate a entirely to the ALF scheme. Further modelling decision as to how to proceed with river management under the investigations are recommended to support the Restoring Sustainable Abstractions programme. In support of assessment and to allow more localised effects on stream these needs, the present paper reports an analysis of data from the flow to be investigated. River Ver catchment, with the objective of identifying changes in the river’s flow regime following implementation of the ALF 1. INTRODUCTION scheme and whether any change identified can be attributed to the The four-year drought from 1988 to 1992 heightened concerns closure of Friars Wash PS. A selection of visual, statistical and over extreme low flows in rivers in England and Wales.1 The modelling analytical techniques is used. National Rivers Authority (NRA) subsequently identified 40 rivers suffering from severe low flows in Low Flows and Water 2. THE RIVER VER HYDROLOGY Resources—Facts on the Top 40 Low Flow Rivers in England and The topographical catchment of the River Ver extends over Wales.2 This publication brought to attention the need for approximately 134 km2, from the Dunstable Downs in the north, action to minimise the impact of the anthropogenic causes of southeast towards its confluence with the River Colne (Fig. 1). low flows. For example, 27 of the 40 rivers identified had low Land cover is mostly agricultural with the significant urban flows attributed to public water supply abstraction or to summer developments of St Albans and Hemel Hempstead in the lower spray irrigation under ‘licences of right’.3 catchment. The chalk aquifer underlying the catchment, formed predominantly from the Upper Chalk formation, provides the Owen4 proposed that over-abstraction of groundwater was an River Ver with a substantial baseflow component (base flow index important cause of unsatisfactorily low flows in groundwater- (BFI) ¼ 0.9). The groundwater catchment area fluctuates and has dominated rivers. In particular, his comparison of the River Ver been quoted as being between 89 and 121 km2, substantially and the River Chess, both chalk-stream tributaries of the smaller than the topographic catchment area. River Colne in Hertfordshire, UK, concluded that abstractions were having a major influence on the Ver’s flows. More The gauging station on the River Ver at Hansteads (located recently, increasing consideration has been given to additional approximately 1 km upstream of the confluence with the River causes of low flows besides abstraction. Acreman et al.3 put Colne) is a compound crump weir designed to allow accurate forward the notion that there may be multiple causes of measurement of flows in the range 0.085–2.83 m3/s. While the perceived degradation of a river’s flow regime; possible causes gauging station has been operational since 1956, data prior to the include land use change, field drainage and increasing public station’s relocation in 1968 are deemed unreliable. The mean daily awareness and expectation, as well as over-abstraction. In some flow at Hansteads for the period 1956 to 2001 is 0.42 m3/s, with a . 3 . 3 cases, low flows may be a natural occurrence, especially in corresponding Q95 and Q10 of 0 08 m /s, and 0 83 m /s chalk streams.5 respectively. A further gauging station was installed on the River

Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al. 171

Downloaded by [ University of Queensland - Central Library] on [23/12/15]. Copyright © ICE Publishing, all rights reserved. granted to Three Valleys Water cover the seven public water LUTON supply pumping stations. Together, these account for approximately 97% of the current licensed groundwater abstraction volume.

Kensworth Lynch PS Most of the water abstracted is ▲ transferred out of the catchment to supply Luton and Dunstable, or is supplied within Markyate STW the Ver, but transferred out of Friars the catchment to be treated and Wash PS ▲ subsequently discharged into HARPENDEN ▲ the Rivers Colne or Lea. Currently the largest licensed discharge to the River Ver is Redbourn from Markyate sewage treatment works (approximately 2 km upstream R. VER of Friars Wash). Available data indicate that the average Bow ▲ discharge from Markyate is Bridge PS approximately 0.015 m3/s (19%

▲ of the flow at Q95). ST ALBANS HEMEL HEMPSTEAD ▲ ▲ 3. LOW FLOWS IN THE VER Since 1950 increasing demands for water have dramatically increased abstraction from the underlying chalk aquifer and, Hansteads by the 1980s, it was estimated 0 1 2 4 " gauging station that 68–75% of recharge was 4,6 Km COLNE being abstracted. This R. appeared to coincide with the depletion of river flows and a change in the character of the Topographic catchment Portions of map based on information provided by the Environment Agency, river. For example, historic Approx. groundwater catchment Copyright © Environment Agency 2007 evidence suggests that the ▲ Groundwater pumping station Portions of map based on Ordnance Survey data, Crown Copyright © 2007 License no: 100026380 perennial source was in the region of Friars Wash,8 but by Fig. 1. A map of the River Ver topographic and groundwater catchments locating the Hansteads 1992 after an extended drought gauging station and Friars Wash pumping station period with continued abstraction, the perennial head is reported as having migrated Ver, approximately 19 km upstream at Redbourn in 1993 to 9 km downstream to just north of St Albans. In June 2005, some monitor low flows. Data from this gauge have not been used here 12 years after implementation of the ALF scheme, and following 7 owing to the lack of data prior to the implementation of the ALF months of below average rainfall, the perennial source was scheme. The mean annual rainfall (from 1956 to 2001) for the observed north east of Redbourn, 4 km downstream of the catchment is 719 mm.7 There is a lag of approximately 3 months historical Friars Wash source. Source migration is a natural between peak rainfall (October to January) and peak flow phenomenon in chalk streams and in 2000/2001 the source (February to March). A significant flashy response occurs, migrated to the highest recorded location, just north of Kensworth however, during storm events as a result of runoff from the urban Lynch, causing groundwater flooding in dry valleys.9 The historic areas, as illustrated for 1984 in Fig. 2. source quoted above may have been maintained by watercress beds, which were supported by groundwater augmentation but There are currently 18 licences granted for groundwater have since been abandoned. abstraction from 41 boreholes in or close to the groundwater catchment. The majority of the licences were granted under the In the mid-1980s Thames Water Authority (predecessor to the 1963 Water Act, and up until the 2003 Water Act, any reduction in regulatory bodies NRA and the Environment Agency, and the abstraction was at the licence holder’s discretion. Two licences water and sewerage service provider Thames Water Utilities)

172 Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al.

Downloaded by [ University of Queensland - Central Library] on [23/12/15]. Copyright © ICE Publishing, all rights reserved. Initially, a suite of visual and 1·2 0 statistical techniques was selected to increase the 1·0 likelihood of identifying any possible change. Rainfall– 25 /s

3 runoff modelling was selected 0·8 as a method of isolating and subsequently investigating the 0·6 50 impact of the contributing factors. Rainfall: mm 0·4 Owing to the natural inter-

Daily mean discharge: m 75 annual variability of stream 0·2 flow, three data periods each of Daily mean discharge 11 consecutive years were Daily mean rainfall defined for comparison: 1971 0·0 100 to 1981 (period 1), 1982 to 1992 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec (period 2) and 1994 to 2004 (period 3). The first two periods Fig. 2. Daily mean discharge (Hansteads) and daily mean areal rainfall (Chiltern East) for 1984 are prior to the reduction in abstraction at Friars Wash PS commissioned an investigation into potential methods of low- in 1993. For the first stage of analysis, the year of the ALF flow alleviation in the Ver. The studies by Halcrow,6,8 concluded implementation, 1993, has been excluded entirely to reduce the that over-abstraction, most notably at Friars Wash PS (located impact of any short-term aquifer response. approximately 20 m from the River Ver just north of Redbourn (Fig. 1)) and Bow Bridge PS (located 20 m from the River Ver, 5. VISUAL AND STATISTICAL ANALYSIS approximately 2 km north of St Albans) was the cause of low The ambiguous definition of ‘low flow’ means that techniques flows. Of the various proposals, the reduction of abstraction at used to report and analyse low flows vary widely.1,13,14 The most Friars Wash was preferred for environmental and technical relevant techniques are those presented by Gustard et al.14 Firstly, reasons. Three Valleys Water and the NRA signed an agreement to a visual comparison of the flow duration curves from the different reduce the maximum licensed annual abstraction from Friars time periods provides a qualitative indication of a change in flow Wash from 5 806 818 m3 (equivalent to an average flow of regime, focusing in this case on the low flows, of which an . 3 3 . 3 0 18 m /s) to 1 388 636 m (0 04 m /s) (the licensed maximum associated index is the 95 percentile flow (Q95). Secondly, the daily abstraction remained unchanged at 16 010 m3). More separation of the hydrograph into surface runoff and baseflow importantly, under the ALF scheme agreement, abstraction at components allows the creation of a 5-day baseflow time series14 Friars Wash was restricted to specified emergencies only. In order from which better comparisons of low flows can be made. Finally, to compensate for the loss of Friars Wash, Luton’s water double mass offset plots are used to identify any change in distribution system was revised to bring water in from Grafham relationship between two variables, in this instance areal rainfall Water in Cambridgeshire. Actual abstractions recorded for Friars and the flow at Hansteads. Offset plots are also reported using the Wash PS during the period June 1993 to December 2004, however, Hansteads flow and the observed flow in the nearby River indicate that abstractions took place in 116 months (out of a Mimram (at Panshanger Park gauge), which drains the same chalk possible 139), with an average monthly abstraction of 43 843 m3 outcrop and has a similar catchment area to Hansteads (0.017 m3/s). (133.9km2).

These alleviation measures, implemented from May 1993, resulted Flow duration curves (FDCs) plotted for each of the three periods in an abrupt reduction in the rate of groundwater abstraction (Fig. 3) indicate a more pronounced difference in the flow regime within the catchment and particularly in the upper catchment of before and after 1993 (between periods 2 and 3) than can be the Ver. Since the closure there have been suggestions of a identified between periods 1 and 2, and even after the bias of the corresponding increase in streamflow10,11 and some ecological extreme rainfall in 2000 and 2001 is removed. This difference is studies have recorded a recovery in the number of species present mostly displayed by the much greater maximum flows (>1.0m3/s) along the River Ver.10,12 There has, however, been no detailed in period 3 than in the previous two periods, themselves fairly objective assessment of any changes in stream flow regime and consistent, and more generally by the greater variation in the whether they can be attributed solely to the reduction in pumping range of flows. This difference between the three periods is at Friars Wash. It is the intention here to address this issue. mirrored in the baseflow duration curves, and suggests a change in flow regime in the periods before and after the ALF scheme. 4. METHOD They cannot, however, provide any explanation as to the cause of Rainfall, stream flow, potential evapotranspiration (PE) and the change. groundwater level data were supplied by the Environment Agency. Two stages of analysis were used By focusing the FDCs on the annual trends, analysis of the ranked

annual Q95 modelled flows (Table 1, discussed in more detail later) (a) visual and statistical analysis of low-flow data revealed that eight of the top 14 were from period 3. This suggests (b) rainfall–runoff modelling. that the lowest flows during period 3 were more frequently greater

Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al. 173

2·0 /s 3

1·5 Discharge: m 1·0

0·5

0·0 0 10 20 30 40 50 60 70 80 90 100 Percentage of time flow exceeded

Fig. 3. Flow duration curves for periods 1, 2 and 3 (with and without the years 2000/2001)

than during any other period. Rank Without ALF scheme (scenario 2): m3/s With ALF scheme (baseline): m3/s Some of this difference can, however, be directly attributed Year Q95 Q10 Year Q95 Q10 to extreme rainfall, and it is difficult to isolate this impact . . . . 1 2001 0 466 1 912 2001 0 677 2 110 on long-term stream flow. In . . . . 2 1979 0 361 0 925 1994 0 412 1 290 order to enable a more detailed 3 1971 0.255 0.569 2002 0.391 1.010 4 1983 0.254 0.723 1979 0.361 0.925 analysis of any change in flow 5 2002 0.251 0.858 2000 0.281 1.000 regime, annual FDCs were 6 1975 0.246 0.734 1971 0.255 0.569 analysed (curves not shown . . . . 7 1988 0 237 0 997 1999 0 254 0 790 here, but available in . . . . 8 1980 0 221 0 676 1983 0 254 0 723 Clayton15). 9 1994 0.210 0.937 1975 0.246 0.734 10 1995 0.201 1.059 1988 0.237 0.997 11 1982 0.195 0.628 1980 0.221 0.676 The three annual FDCs 12 1981 0.190 0.424 1995 0.218 1.150 signifying the greatest . . . . 13 1978 0 190 0 618 2003 0 213 1 210 discharge at low flows were all . . . . 14 1977 0 187 0 529 2004 0 207 0 597 from period 3: for the years 15 1972 0.187 0.586 1982 0.195 0.628 16 1986 0.144 0.398 1981 0.190 0.424 2001, 2002 and 1994. The 17 2000 0.140 0.851 1978 0.190 0.618 uncharacteristically high flow 18 1987 0.131 0.527 1977 0.187 0.529 regime observed in 2001 may . . . . 19 2003 0 126 1 125 1972 0 187 0 586 be attributed to extreme . . . . 20 1984 0 124 0 479 1986 0 144 0 398 rainfall during the winter of 21 2004 0.111 0.403 1987 0.131 0.527 22 1999 0.104 0.412 1998 0.125 0.491 2000/2001 (return period of 16 23 1990 0.091 0.479 1984 0.124 0.479 greater than 200 years ). In 24 1985 0.078 0.435 1990 0.091 0.479 1994, itself an unremarkable . . . . 25 1992 0 062 0 260 1985 0 078 0 435 year for rainfall, higher . . . . 26 1989 0 059 0 342 1996 0 071 0 437 observed flows can be partly 27 1991 0.055 0.197 1992 0.062 0.260 28 1974 0.049 0.431 1989 0.059 0.342 attributed to rainfall in 1992 29 1996 0.040 0.357 1991 0.055 0.197 and 1993, where total annual 30 1973 0.028 0.254 1974 0.049 0.431 rainfall is ranked eighth and . . . . 31 1976 0 013 0 200 1997 0 046 0 173 ninth respectively for the . . . . 32 1998 0 011 0 266 1973 0 028 0 254 period 1971 to 2004. A 33 1997 0.008 0.167 1976 0.013 0.200 comparison of total annual rainfall and annual FDC reveals Table 1. Ranked annual Q95 (and corresponding Q10) for the period 1971 to 2004, with impact of ALF scheme (modelled baseline flow) and without impact of ALF scheme (modelled flow from scenario 2) no consistent correlation, leading the authors to assume

174 Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al.

0·90

0·85

0·80 Annual baseflow index

0·75

0·70 1 2 3 4 5 6 7 8 9 10 11 Rank

Fig. 4. Plot of ranked annual baseﬂow for the three periods

that rainfall is not the only cause of the higher flows. In 1994, of baseflow or total flow against areal rainfall or areal effective however, flows were some of the highest in the 34-year study rainfall, all produced very similar relationships with change points period, an increase significant even at low flows. If the high flows occurring at the same time. The first notable change is in 1968, during 1994 are to be at all attributed to the reduction in pumping when the Hansteads gauging station was rebuilt. The second from Friars Wash PS, why is this trend of higher flows not major change point in 1992 signifies a general decline in flows continued into 1995, 1996 and 1998? In fact these three years relative to rainfall, which may be associated with the end of the 4-

exhibited some of the lowest Q95 values (Table 1). Actual daily year drought. In 1994 a smaller reversal can be identiﬁed which returns data from Friars Wash PS provide no evidence to suggest a may be linked to the ALF scheme. This increase in stream ﬂow

link between these lower Q95 years and increased rates of relative to rainfall is short-lived, however, and in 1999 the stream emergency abstraction from the PS. flow–rainfall ratio declines more drastically, a change that cannot obviously be linked to any single cause. In order to assist in If the years influenced by the unusually high rainfall (i.e. 2000 and isolating the impact of climate on the flow regime, a double mass 2001) are excluded, the range of the period 3 flow duration curves offset plot was produced for the River Ver at Hansteads plotted is significantly reduced, implying that extreme precipitation against the nearby River Mimram as measured at the Panshanger partially explains the variation between the three periods, in gauge. The plot, which depicts the consistency of the relationship particular at the low flows. Baseflow annual FDCs were similar in between the two flow series, replicates the timing of the latter shape to the total flow annual FDCs and thus they provide no change points. This suggests that these change points cannot be additional insight. While flow duration curves have proved useful attributed to a change in climate, as it must be assumed that the in the analysis of differences between years, and in highlighting Mimram catchment would respond to that type of change in a some features of the flow regime, they do not explain the similar manner as the Ver catchment. It is possible that the differences. changes may be linked to groundwater level thresholds. The underlying cause remains, however, unknown; there is no good The BFI was calculated for each year of the three periods (Fig. 4). A evidence of a link to ALF interventions. plot of annual baseflow ranked (on baseflow) within each period clearly shows that there is increased baseflow in period 3 relative 6. RAINFALL–RUNOFF MODELLING to periods 1 and 2. This increase in baseflow could be groundwater Rainfall–runoff modelling allows the effects of individual factors, from the environs of Friars Wash that was previously intercepted for example abstractions or climate variability, to be isolated. The by groundwater abstraction. Alternatively, it could be from rainfall–runoff model employed in this study is the conceptual increased rainfall and recharge causing a regional increase in model,17 Catchmod v4.02.18 Catchmod consists of three stores: a groundwater levels without a corresponding increase in surface soil moisture store, based on the Penman drying curve, which runoff. estimates the rate of loss owing to evaporation, and two conceptual stores in series (one linear and one non-linear) to Offset plots are obtained by taking the difference between the represent the catchment storage. Catchmod permits the inclusion cumulative double mass curve and a straight line connecting the of daily groundwater abstractions, which are taken from the first and last points displayed over time. Any change in gradient second linear store. Surface water discharges and abstractions are indicates a change in the relationship and is referred to as a subtracted/added directly from/to the output flow. The model may ‘change point’. The double mass offset curve of cumulative be employed in a semi-distributed manner, that is the division of a Hanstead stream flow plotted against effective rainfall is shown in catchment into sub-catchments with discrete inputs and outputs Fig. 5. Double mass offset curves from the various combinations by the creation of independently modelled hydrological response

Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al. 175

300

200

100 ALF scheme implemented 0 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

–100 Doublemass 'offset': mm

General increase in –200 flows relative to rainfall

–300 General decrease in flows relative to rainfall Hansteads gauging –400 station rebuilt

Fig. 5. Doublemass offset plot of the River Ver at Hansteads against Chilterns East effective rainfall

zones, the output flows from which are added together to obtain (E).19 This is used as a mathematical measure of how well a model the overall catchment response. In the Ver model, two zones were simulation fits the observed data.20 It is defined as employed: one to represent the permeable areas of the catchment and one to represent the urban areas. P n ðO M Þ2 1 E ¼ 1 Pi ¼ 1 i i n ðO OÞ2 The conceptual nature of the model necessitates the calibration of i ¼ 1 i Catchmod to observed flow data. Three criteria were used when manually calibrating Catchmod to the Ver data: minimisation of Values of E range between 1 (perfect model fit) and 1.In the difference between the simulated cumulative volume of flow squaring the differences between the observed (O) and modelled and the corresponding observed volume; visually matching the (M) values there is a tendency for E to place an emphasis on the simulated and the observed flow duration curve, focusing on the fitting of peak rather than low flows; for that reason the present low flows; and maximisation of the Nash–Sutcliffe efficiency study does not rely entirely on this measure.

3·00

Observed flow Modelled baseline 2·50

2·00 Period 1 Period 2 Period 3 (Validation) (Calibration) (Validation) /s 3 1·50 Flow: m

1·00

0·50

0·00

81 86 89 91 94 96 97 01/7101/72 01/7701/78 01/90 01/ 01/ 01/00 01/04 01/ 01/ 01/01/7301/01/7401/01/7501/01/7601/ 01/ 01/01/7901/01/8001/01/01/01/8201/01/8301/01/8401/01/8501/01/01/01/8701/01/8801/01/01/ 01/01/01/01/9201/01/9301/01/01/01/9501/ 01/ 01/01/9801/01/9901/ 01/01/0101/01/0201/01/0301/ Date

Fig. 6. Final modelled plotted against observed ﬂows (River Ver at Hansteads) for periods 1, 2 and 3

176 Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al.

Downloaded by [ University of Queensland - Central Library] on [23/12/15]. Copyright © ICE Publishing, all rights reserved. scenarios. Table 2 summarises the scenarios presented in detail Model inputs below.

Baseline Modelled river flow calibrated and validated using Following calibration, various scenarios of abstractions and actual rainfall, potential evaporation and climate were simulated and compared with the baseline abstraction records Scenario 1 Model as calibrated for the baseline simulation with simulation (i.e. the calibration and validation results). no abstractions Scenario 2 Model as calibrated for the baseline simulation with Scenario 1 used the model to simulate the Ver as a ‘natural’ period 3 abstractions as for period 2 catchment, and predicted stream flow without any Scenario 3 Model as calibrated using synthetic annual average groundwater abstractions, but otherwise using the same inputs (1961 to 1990) daily rainfall and potential evaporation, and actual abstraction records as used for the baseline simulation. The output hydrograph illustrates that flow simulated under scenario 1 is noticeably Table 2. Catchmod scenarios greater than the baseline flow in both periods 2 and 3 (see . 3 Fig. 7). For example, the increase in flow at Q95 is by 0 31 m /s in period 3 and 0.41 m3/s in period 2. Scenario 1 strongly Period 2 (1982 to 1992) was selected as the calibration period, indicates that if all abstractions were ceased within the while period 1 (1971 to 1981) and period 3 (1994 to 2004) acted as catchment, there would be a significant increase in flow validation periods. Sensitivity analysis conducted prior to volumes throughout the range. calibration identified the most sensitive parameters. For scenario 2 the abstraction rates in period 3 were taken as Calibration of the model was conducted in three parts. The the average of the actual pumping rates from period 2, an baseflow component of the Ver catchment originates from the approximation of the ALF scheme not being implemented. The chalk and this was modelled first. Parameters were adjusted so abstractions in period 2 itself were not altered. Results are that the baseflow response followed the general peaks and troughs shown in Fig. 8. From mid-1994 the scenario 2 flow is visibly of the observed hydrograph. A second response area representing less than the baseline flow. Analysis of the hydrograph reveals the quick flow runoff component of the observed flow was then the most significant difference is at the lowest flows. For replicated prior to the final calibration of the two response zones example in 1998/1999 minimum stream flow is 0.01 m3/s together, when calibration focused on the quantitative indicators compared with 0.06 m3/s baseline, and in 2000/2001 the of goodness-of-fit. minimum flow is often almost half the baseline, and the

modelled increase in period 3 Q95 flow associated with the ALF The final Catchmod model performed well over both the scheme is 0.06 m3/s. Significant differences in low flows are not calibration and verification periods (Fig. 6). For all periods seen until mid-1995, 2 years after the ALF implementation together, the ratio of simulated to observed flow volume is 1.06, suggesting an extended groundwater storage recovery period. . and Nash–Sutcliffe efficiency is 0 89. Although there are Annual Q95 flows are ranked to allow comparison between years questions over the predictive reliability of models, the for both the baseline and scenario 2 data (Table 1). The increase

performance of Catchmod over the validation periods, which in Q95 values between the two scenarios during period 3 includes the reduction in abstractions after 1993, allows some suggests that the ALF scheme may have prevented extreme low conﬁdence to be attached to the subsequent simulated ﬂows on a number of occasions since 1993.

3·0 0

20 2·5 40 Rainfall 60 2·0 Scenario 1 flow Period 2 Period 3 80

/s Modelled baseline flow 3 1·5 100 Flow: m

120 Rainfall: mm 1·0 140

160 0·5 180

0·0 200

82 87 88 91 92 94 00 01 03 01/ 01/ 01/90 01/96 01/97 01/ 01/04 01/ 01/01/8301/01/8401/01/8501/01/8601/ 01/01/ 01/01/8901/ 01/01/ 01/01/ 01/01/9301/01/ 01/01/9501/ 01/ 01/01/9801/01/9901/ 01/01/ 01/01/0201/01/ 01/ Date

Fig. 7. Scenario 1 and baseline model output for periods 2 and 3

Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al. 177

20 2·5 40

Rainfall 60 2·0 80 /s 3 Period 2 Period 3 1·5 100 Scenario 2 flow

Flow: m Modelled baseline flow

120 Rainfall: mm 1·0 140

160 0·5 180

0·0 200

82 87 88 91 92 94 00 01 03 01/ 01/89 01/96 01/97 01/ 01/04 01/ 01/01/8301/01/8401/01/8501/01/8601/01/ 01/01/ 01/ 01/01/9001/01/ 01/01/ 01/01/9301/01/ 01/01/9501/ 01/ 01/01/9801/01/9901/ 01/01/ 01/01/0201/01/ 01/ Date

Fig. 8. Scenario 2 and baseline model output for periods 2 and 3

In order to eliminate the impact of climatic variability between the groundwater abstraction within the catchment (Fig. 9). The periods 2 and 3, scenario 3 was purely theoretical. Annual average simulation also reveals that the catchment response to such a (1961 to 1990) daily PE was used together with the daily rainfall reduction in groundwater abstraction can be slow, taking up to 2 selected for an ‘average’ year (1970) and the actual groundwater years under average rainfall and PE conditions. abstractions. Without the natural variability in PE and rainfall this . 3 scenario predicted an increase in period 3 Q50 flow of 0 16 m /s 7. DISCUSSION (approx. 38% of daily mean flow) associated with the ALF scheme, Visual inspection of the hydrograph from Hansteads gauging . 3 compared with a potential maximum Q50 increase of 0 51 m /s station shows a significant increase in stream flow following the (approx. 110% of daily mean flow) for the cessation of all implementation of the ALF scheme in 1993. This is confirmed by

1·0 0

Rainfall 40

60 Scenario 3 flow 80 /s 3 0·5 100 Flow: m 120 Rainfall: mm

140

160

180

0·0 200 4 5 6 9 5 8 2 96 198 1991 /19 200 01/198 01/198 01/ 01/199 01/1999 01/01/198201/01/198301/ 01/ 01/01/ 01/01/198701/01/198801/01/19801/01/199001/ 01/01/199201/01/199301/01/199401/01/19901/01 01/01/199701/ 01/ 01/01/200001/01/200101/01/ 01/01/200301/01/2004 Date

Fig. 9. Scenario 3 for periods 2 and 3, illustrating the simulated impact of the change in groundwater abstraction on ﬂow within the River Ver. Rainfall is from an average year

178 Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al.

Downloaded by [ University of Queensland - Central Library] on [23/12/15]. Copyright © ICE Publishing, all rights reserved. analysis of annual flow duration curves and Q95 values, and (a) There is strong qualitative and statistical evidence that flows substantiated by a consistent increase in annual BFI. Inspection of in the River Ver at Hansteads gauging station have increased double mass offset plots (Ver rainfall against Ver flow, and Ver following implementation of the ALF scheme in May 1993. flow against Mimram flow) revealed inconclusive evidence of a There is, however, insufficient evidence to suggest that this change in flow regime following 1993. Much of the increased flow increase is due entirely to the ALF scheme rather than in post-1993 can be linked to the unusually high rainfall in 2000/ response to climatic variability. 2001 rather than solely to a reduction in groundwater abstractions (b) Through use of a conceptual rainfall–runoff model at Friars Wash. Hence a rainfall–runoff model, Catchmod, was (Catchmod), substantial increases in simulated low flows can applied, with the aim of isolating climatic variability effects from be seen to be associated with the ALF scheme. abstraction effects. The modelled scenarios imply that the (c) Although the model performed well in terms of simulating reduction in abstraction associated with the ALF scheme results in flows at Hansteads under relevant changes in abstraction a significant alleviation of low flows at Hansteads. In period 3 rates, as a lumped conceptual model Catchmod is not capable . 3 (1994 to 2004) simulated Q95 increased to 0 09 m /s when the of isolating the influence of individual abstractions, or abstraction reductions implemented in 1993 were included in the simulating effects on localised flows. Thus the feasibility of simulation, compared with 0.03 m3/s without reductions. Without more physically based, distributed, integrated surface and this modelling it would not be possible from observed flow data groundwater rainfall–runoff modelling should be alone to isolate the changes in flow regime arising as a result of investigated. the implementation of the ALF scheme from those arising from (d ) Owing to the complex nature of the groundwater–surface extremes of weather/climate. water systems and the variability of climatic inputs, the present study has illustrated that it is essential to employ a Arguably, this analysis has illustrated the insights achieved by range of visual, statistical and modelling techniques in the applying a simple rainfall–runoff model to the ALF assessment evaluation of all ALF schemes. problem. An alternative view is that the simplicity of the Catchmod model has led to an exaggeration of the simulated ACKNOWLEDGEMENTS ALF impacts on the Ver. The conceptual nature of the model The data used in this study were provided by the Environment makes the inclusion of changes, other than those for the input Agency for England and Wales and Three Valleys Water plc. The rainfall, PE and abstraction time series, impractical. This means, work was sponsored by the Environment Agency, Jacobs and for example, that the impacts of land use changes (e.g. decline Imperial College, London. of watercress beds) are entirely excluded. The errors observed during validation periods were, however, small relative to the REFERENCES modelled impacts of abstraction scenarios, suggesting that, in 1. AGNEW C. T., CLIFFORD N. J. and HAYLETT S. Identifying and practice, they are not significant. Another issue that may limit alleviating low flows in regulated rivers: The case of the the reliability of the Catchmod results is that all groundwater Rivers Bulbourne and Gade, Hertfordshire, UK. Regulated abstractions are treated as spatially lumped. Local differences in Rivers: Research and Management, 2000, 16, No. 3, 245–266. transmissivity and variations in aquifer–river connection, which 2. NATIONAL RIVERS AUTHORITY. Low Flows and Water Resources. may be significant to the importance of abstractions relative to Facts on the Top 40 Low Flow Rivers in England and one another, are ignored. The satisfactory performance of the Wales. NRA, Bristol, UK, 1993, pp. 1–30. model in period 3, during which the Friars Wash abstractions 3. ACREMAN M. C., ADAMS B., BIRCHALL P. and CONNORTON B. were smaller than in the calibration period, however, allows Does groundwater abstraction cause degradation of rivers some confidence that the effects of Friars Wash are sufficiently and wetlands? Journal of the Institution of Water and well represented. Another possible reservation about this Environmental Management, 2000, 14, No. 3, 200–206. application of Catchmod is that uncertainty in parameter values 4. OWEN M. Groundwater abstraction and river flows. Journal has been neglected, and reliability has simply been declared of the Institution of Water and Environmental Management, ‘adequate’ on the basis of validation period performance. 1991, 5, No. 6, 697–702. Further studies might apply a series of equally suitable 5. SEAR D. A., ARMITAGE P. D. and DAWSON F. H. Groundwater parameter sets to generate an ensemble of simulations, allowing dominated rivers. Rivers. Hydrological Processes, 1999, 13, assessment of the significance of model results in the context of No. 3, 255–275. uncertainty. 6. SIR WILLIAM HALCROW &PARTNERS. Study of Alleviation of Low Flows Resulting from Groundwater Abstraction, Final Some of these model limitations could be lessened with the use of Report, Volume 4—Ver Case Study. Thames Water a more physically based, distributed, integrated surface and Authority, Reading, 1988, pp. 1–71. 21 groundwater rainfall–runoff model (e.g. MIKE-SHE ). Such a 7. CENTRE FOR ECOLOGY and HYDROLOGY. Hydrological Data model would also provide a distributed output of groundwater United Kingdom, Hydrometric Register and Statistics, levels and stream flows, allowing more local effects (e.g. 1996–2000. Institute of Hydrology and British Geological movement of the perennial head ) to be explored, but would Survey, Wallingford, 2003, p. 98. require additional data collection. The feasibility of such a 8. SIR WILLIAM HALCROW &PARTNERS. Study of Alleviation of modelling study should be considered, in terms of model Low Flows Resulting from Groundwater Abstraction, suitability and distributed data requirements, for example on land Interim Report, Annex C: River Ver Case Study. Thames cover, hydrogeological properties and meteorological inputs. Water Authority, Reading, 1987, pp. 1–55. 9. ROBINSON V., SOLOMON J. and MORRIS S. E. Groundwater 8. CONCLUSIONS Flooding in Thames Region, Winter 2000/2001. The main conclusions from this study are listed below. Environment Agency, Reading, 2001, Internal Report.

Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al. 179

Downloaded by [ University of Queensland - Central Library] on [23/12/15]. Copyright © ICE Publishing, all rights reserved. 10. HOLMES N. T. H. Recovery of headwater stream flora 16. MARSH T. J. and DALE M. The UK floods of 2000–2001: a following the 1989–1992 groundwater drought. hydrometeorological appraisal. Journal of the Institution of Hydrological Processes, 1999, 13, No. 3, 341–354. Water and Environmental Management, 2002, 16, No. 3, 11. EDMONDS-BROWN V. and POWERS E. River Ver Report, June 180–188. 1999. Prepared for Three Valleys Water PLC and the Ver 17. GREENFIELD B. J. The Thames Catchment Model. Thames Society, 1999. Water Authority, Reading, 1984. 12. LEEMING D. Ecological Resilience and Recovery from Drought 18. ATKINS CONSULTANCY. Catchmod User Manual, a Guide to in the River Ver Following Implementation of an Alleviation Using the Catchmod Rainfall–Runoff Modelling Software of Low Flows (ALF) Scheme. Draft PhD thesis, University of Package. Environment Agency, UK, 2005, draft. Hertfordshire, 1999. 19. NASH J. E. and SUTCLIFFE J. V. River flow forecasting 13. SMAKHTIN V. U. Low flow hydrology: a review. Journal of through conceptual models: Part I—A discussion of Hydrology, 2001, 240, 147–186. principles. Journal of Hydrology, 1970, 10, No. 3, 14. GUSTARD A., BULLOCK A. and DIXON J. M. Low Flow 282–290. Estimation in the United Kingdom. Institute of Hydrology, 20. BEVEN J. K. Rainfall–Runoff Modelling—The Primer. Wiley, 1992, Report No. 108, p. 88. Chichester, 2000, p. 319. 15. CLAYTON H. J. An Evaluation of the Impact of Measures to 21. REFSGAARD J. C. and STORM B. MIKE-SHE. In Computer Restore Flows in the River Ver. MSc dissertation, Imperial Models of Watershed Hydrology (SINGH V. P. (ed.)). Water College London, 2005, pp. 1–117. Resources Publications, USA, 1995, pp. 809–846.

What do you think? To comment on this paper, please email up to 500 words to the editor at [email protected] Proceedings journals rely entirely on contributions sent in by civil engineers and related professionals, academics and students. Papers should be 2000–5000 words long, with adequate illustrations and references. Please visit www.thomastelford.com/journals for author guidelines and further details.

180 Water Management 161 Issue WM4 The hydrological impact of low-ﬂow alleviation measures Clayton et al.