The groundwater models

Following a review of its groundwater modelling programme in 1998–1999, the Environment Agency developed a Framework for Groundwater Resources—Conceptual and Numerical Modelling (Brown and Hulme Citation2001). Since that time, the Environment Agency’s nationally coordinated programme of groundwater resource assessment and modelling has resulted in the development of calibrated, transient, regional groundwater models for almost all the major aquifer units in England (Whiteman et al. Citation2012).

In the Environment Agency’s Midlands Region, groundwater models have been developed for West Midlands Worfe (WMW; Soley et al. Citation2001), East Shropshire (ESM; Streetly and Shepley Citation2005), East Midlands Yorkshire (EMY; Shepley and Soley Citation2009) and Bromsgrove (Streetly and Shepley Citation2010) Permo-Triassic sandstone aquifers, as shown in Fig. 2. The use of these models to derive “natural” summer flows in the associated streams is described in Shepley and Streetly (Citation2007).

Fig. 2 Groundwater models and NEP/RSA investigation areas.

Display full size

The groundwater models have been constructed to quality criteria (Hulme et al. Citation2002) that are applied by the Environment Agency to all groundwater models used for catchment management. This includes validation of the groundwater models against observed groundwater level time series and total surface water flow time series measured at gauging stations (streamflow gauges for catchments in the range 30–400 km² with data for the last 30–40 years). In some cases, refinement at a more local scale has been carried out using spot flow gauging data, particularly on reaches where abstraction stress was considered to be significant.

The groundwater models use the MODFLOW code (McDonald and Harbaugh Citation1988) and take inputs from separate recharge models that are not dynamically linked to the groundwater model. The recharge models calculate routed runoff, and total surface flows are then calculated from the combined outputs of the groundwater and recharge models. For the ESM, the recharge model calculates runoff at a sub-catchment scale so that total flows can only be calculated at sub-catchment outflow points. The other three models use the 4R code (Heathcote et al. Citation2004) in which the runoff is routed via a stream network using the same grid as the groundwater model, thus providing inflows to the MODFLOW stream cells. This means that total flow can be calculated for every stream cell (250–500 m) in these models.

The models can be considered to be a reliable tool for assessing different water resource options at the scale at which they have been calibrated. In general, for the sites considered in this study, they are reliable tools for assessing flows at the sub-catchment scale being considered, but their use at the reach scale (<1 km) should be considered as a scoping calculation to be supported by more detailed local data.

While the 4R recharge models calculate a water balance every day, because the Permo-Triassic sandstone responds very slowly to recharge and provides baseflow at a fairly steady rate, the groundwater models typically use a monthly stress period. This means that the total surface water flows calculated by the models are generally monthly average flows. The evaluation techniques presented in this article require assessment of various flow statistics (Q₇₅, Q₉₅, etc.) based on daily flow data. The difference between daily and monthly average flow statistics is discussed in the Appendix, which shows that, for these catchments, it is possible to produce a simple correction factor to produce approximations to the daily flow statistics from monthly average flows calculated by the groundwater models. However, the spatial distribution of groundwater abstraction effects on surface water flows is largely unaffected by this aspect of the models, and application of the correction factor did not significantly affect the reaches considered to be affected by abstraction stress.

The development of these groundwater models has allowed various “predictive scenario” model runs to be carried out in order to simulate potential future conditions. In general, these repeat the climate sequence used in the historic, calibrated run (typically 40 years), starting from the latest simulated heads and flows, but change the abstraction and discharges applied. The key model predictive runs used in this article for the assessment of current impacts on surface water bodies are the following:

• –  Naturalized model: this represents “natural” conditions in the aquifer, with no abstractions and most other anthropogenic influences removed (however, no change to land use is applied).

• –  Recent actual model: this represents the aquifer conditions under the current abstraction rates (generally based on actual abstractions averaged over the last 5 yearsFootnote²).

By comparing the results from these two model runs, we can assess the effects that groundwater abstraction has on groundwater levels and surface water flows.Footnote³ Further predictive model runs can then be carried out to assess the implications of different abstraction regimes (or potentially different climate sequences).

Most of the NEP sites under investigation for 2010–2015 were located in catchments for which the Environment Agency had already developed a regional groundwater model. This allows the spatial and temporal effects of groundwater abstraction on streamflows to be assessed more reliably. This is in contrast to the work presented in Bradley et al. (Citation2013), which is based on data from similar investigations by STWL in 2005–2010, on sites that did not have the benefit of a groundwater model for exploring the potential for different water resource options to result in improvements of in-stream ecology.

Integration of groundwater model results with hydro-ecological relationships

The hydro-ecological relationships established at other similar sites in the region by Bradley et al. (Citation2013) (Fig. 1) were then used to indicate the likely impact of abstraction on the ecological status of in-stream macro-invertebrate communities using the following categorization:

• Green Flow reduction <60% Q₇₅, indicating that impacts of low flows on macro-invertebrates were unlikely in the absence of other pressures.

• Yellow Flow reduction 60–80% Q₇₅, indicating that impacts of low flows on macro-invertebrates were possible in the absence of other pressures.

• Red Flow reduction >80% Q₇₅, indicating that impacts of low flows on macro-invertebrates were likely in the absence of other pressures.

In applying these categories for this purpose, it is important to note that it was developed using data from sites that are affected by groundwater abstraction. It has been shown that groundwater abstraction impacts on streams sourced from the Permo-Triassic Sandstone, such as these, are fairly constant (in terms of magnitude) across the flow spectrum (Shepley and Streetly Citation2007). In other words, the general effect of groundwater abstraction from the Permo-Triassic Sandstone is to shift the whole flow duration curve (FDC) downwards by a constant amount and to have a fairly steady impact from year to year.

In contrast, surface water abstractions or reductions in flow as a result of impoundment may affect flows more abruptly, and the resultant shape of the impacted FDC may differ substantially from “natural”. Thus, while the relationships demonstrated in Bradley et al. (Citation2013) at a single flow percentile may be appropriate for managing groundwater abstraction stress, a more complex consideration of impacts at different flow percentiles may be required for managing surface water abstractions.

The approach outlined above can be compared to the Environment Agency’s national screening approach for identifying abstraction-related stress on aquatic ecology. The latter uses a system of Environmental Flow Indicators (EFIs). For streams in the study area, EFIs have been set at flow equivalent between 80% and 90% of the estimated naturalized Q₉₅ (i.e. abstraction reductions need to be less than 10–20% of Q₉₅)Footnote⁴ (UKTAG 2008). Where the estimated flows under the recent actual abstraction scenario were less than 50% of the EFI (i.e. overall flow reductions of 55–60% of Q₉₅) and the reduction was mainly due to public water supply abstraction, sites were put forward by the Environment Agency for the NEP programme of low flow investigations.

Comparison of different flow scenarios

Having established the hydro-ecology model, it was then possible to interrogate the results from any predictive model scenario and present them directly in terms of the risk of causing negative impacts to macro-invertebrate communities in the absence of other pressures. To demonstrate this, the hydro-ecological model was used to predict two operating scenarios: (1) the effect of reducing abstraction at the investigation sites on the macro-invertebrates in the surrounding streams and (2) the effect of augmentation of river flow by discharges from water treatment works. The results of these scenarios are discussed below.

Identification of reaches potentially impacted by abstraction-related flow stress (regional)

Figure 3 shows the results of applying the hydro-ecological relationship in Bradley et al. (Citation2013) to the recent actual abstraction scenario from the WMW model. Comparison with the equivalent figure applying the relevant EFIs (Fig. 4) suggests that the locally derived hydro-ecological relationship implies that there are fewer reaches in which the macro-invertebrates were likely to be negatively impacted by abstraction than appeared to be the case using the national screening EFIs as criteria. In general, the reaches where macro-invertebrates were considered to be at risk of being negatively impacted by groundwater abstraction using the former approach were restricted to small headwater streams.

Fig. 3 Assessment of likely ecological status of streams in WMW model area using local hydro-ecological relationship at Q₇₅ (Bradley et al. Citation2013). Colour coding of green, yellow and red is the same as categories described in the text. Blue areas show reaches where the flow under the recent actual scenario is greater than the naturalized flow; these increased flows are predominantly due to sewage discharges.

Display full size

Fig. 4 Compliance of flows in streams in WMW model area with WFD EFIs for Q₉₅ flows.

Display full size

This has been confirmed by comparison of an independent set of LIFE data for these streams (collected by the Environment Agency) with the effect of abstraction on river flows calculated by the groundwater models (Fig. 5). This retrospective check with independent data from similar sites within the study area provides a powerful validation of the hydro-ecological relationships developed by Bradley et al. (Citation2013).

Fig. 5 Hydro-ecology validation plot for the WMW area. Data for AMP4 sites (Bradley et al. Citation2013) and historical data from Environment Agency sites (flow impacts derived from groundwater model). The mean LIFE O/E value at each site is plotted. The trend line, its 95% confidence and prediction intervals are shown together with 0.94 LIFE O/E.

Display full size

Looking at Fig. 5, there is clearly a fair amount of variation in the y-axes (LIFE Observed/Expected), which may be related to temporal climatic variation and other environmental factors. This may merit more detailed local investigation of other pressures, but does not alter the overall conclusions with respect to groundwater abstraction stress.

Comparison of different abstraction scenarios (local)

Having established a methodology that effectively calibrates the groundwater model results against ecological status, it is then possible to interrogate the results from any predictive model scenarios and assess the implications of any proposed change in abstraction directly in terms of the implied risk to the local macro-invertebrate communities. An example of this is shown in Figs 6 and 7, which present the inferred ecological status of the streams around Bromsgrove as a result of two different abstraction scenarios: In the scenario presented on Fig. 7, the Wildmoor groundwater abstractions have been switched off. The length of river over which ecological status would be expected to improve as a result of this change can be assessed directly from these plots.

Fig. 6 Hydro-ecological assessment of streams on Bromsgrove aquifer for the recent actual scenario.

Display full size

Fig. 7 Hydro-ecological assessment of streams on Bromsgrove aquifer for the recent actual scenario, except Wildmoor PWS is switched off.

Display full size

Quantification of proportion of wastewater discharge in streams at low flows

The groundwater models can also be used to quantify the amount of sewage treatment discharge in the streams under different flow scenarios as shown in Fig. 8. In this case, the discharges from the treatment works (which are included in the 4R recharge–runoff model) have been switched off and the change in flows downstream has been used as a measure of the contribution of the discharges to flow. It can be seen that some streams have a significant component of wastewater discharge, and it can be anticipated that, in these streams, water quality may be a confounding influence on LIFE scores.

Fig. 8 Proportion of sewage treatment works (STW) discharge in streams at Q₉₅.

Display full size

Applications and potential limitations of the methodology

The discipline of hydro-ecology (and ecohydrology) is barely a decade old (Wood et al. Citation2007), and yet it is increasingly called upon to inform major decisions about the allocation of the world’s most precious natural resource. As water resource management increasingly considers the environment, it would benefit from quantitative predictions of the ecological effects of river flow alterations. Few studies, however, provide sufficient information to support the development of regional environmental flow frameworks, such as ELOHA (Poff and Zimmerman Citation2010, Carlisle et al. Citation2012). This study is currently unique in that it uses quantitative relationships between ecology and flow alterations (Bradley et al. Citation2013) and applies them in a method that is appropriate to inform regional water management decisions.

There has been extensive development and application of environmental flow methods globally (Tharme Citation2003, Acreman and Dunbar Citation2004, Dyson et al. Citation2008). Most of these methods use relationships between biota and river flows to infer impacts from abstraction or flow regulation in the setting of environmental flow standards (e.g. Bovee et al. Citation1998, Extence et al. Citation1999, Dunbar et al. Citation2010). These methods fall short in that they do not incorporate quantitative measures of hydrological alteration from reference conditions and, ultimately, rely on expert opinion to set environmental flow standards, such as the Environment Agency (UK) EFIs (Acreman et al. Citation2008).

Consistent with the ELOHA framework, the method presented here uses quantitative measures of deviation from reference conditions for both hydrology and ecology (Poff et al. Citation2010). The scale at which this method is applied is designed to bridge the gap between existing methods that are either highly detailed and site-specific (such as Physical Habitat Simulation; Bovee et al. Citation1998) or national-scale “rules of thumb” (such as EFIs; Acreman et al. Citation2008) that have insufficient certainty to apply to investigation sites and are therefore conservative and used as screening tools to trigger further investigation using a method such as the one presented here.

A major advantage of the method presented here, and a feature that is consistent with ELOHA, is that it can be applied along river reaches between existing assessment points, at the scale of the groundwater model analysis nodes (Poff et al. Citation2010). This will help to optimize resources by focusing sampling to areas of uncertainty and provide an assessment of possible ecological impacts of abstraction throughout entire river water bodies. In this way, the method overcomes the problems of relying on sample-specific impact assessment and quantifies the length of river that will benefit from any mitigation measures.

There are several limitations of this method that need to be considered:

• –  First, it is a regional model based on data from streams originating from Permo-Triassic Sandstone aquifers in the midlands of England. While the methodological approach might have more wide application, the underlying models presented in this study do not.

• –  Second, the underlying hydro-ecological relationships (Bradley et al. Citation2013) were calibrated on unpolluted streams with similar physical character and on reaches with no major morphological alterations. The relationship between river flow variability and ecology is dependent on the geomorphic context of the river channel (Poff et al. Citation2006) and especially the degree of artificial modification (Dunbar et al. Citation2010). This method cannot be applied with confidence directly to river reaches that are artificially modified or have poor water quality. However, the model that is independent of major environment pressures other than groundwater abstraction has a useful application in water company investigations where the impact of abstraction in the absence of other pressures needs to be established. Further work is underway to quantify the relationships between LIFE and the hydrological effect of groundwater abstraction under different levels of morphological alteration.

• –  Third, LIFE scores reflect the physical quality of river reaches for macro-invertebrates; they do not describe the quantity of aquatic habitat for macro-invertebrates (SNIFFER Citation2012). The standard sampling method prescribes a fixed sampling time, regardless of the size of the wetted habitat. The implications of this are that river habitats that shrink due to abstraction and retain their physical character in miniature will not be reflected in LIFE scores. Quantitative sampling of separate and visually distinct in-stream habitats, combined with mapping of habitat areas (Armitage and Pardo Citation1995), should be considered in future assessments of the impact of flow alterations on macro-invertebrates to account for habitat quantity. Also, while macro-invertebrates are widely used biological indicators of environmental perturbations, this method cannot be directly applied to impact assessments of other organisms, such as fish.

• –  Fourth, the measure of ecological impact on the underlying model relies on the prediction of a reference condition for macro-invertebrates in the absence of flow alterations owing to groundwater abstraction. The River Invertebrate Prediction and Classification System (RIVPACS) model that provides these predictions was originally designed for assessing the effects of water pollution, so the variables that are measured on site to generate the predictions are not all independent of the effects of abstraction (Wright et al. Citation2000). Work is currently underway to develop pressure-independent RIVPACS models for more reliable application to water resource investigations (SNIFFER Citation2011). Also, the reference sites used in the model might not all be independent of flow alterations. Ongoing work to improve RIVPACS for water resource investigations has identified 42 reference sites that should be excluded from the model (SNIFFER Citation2008).

These limitations associated with the RIVPACS model need to be tested in every individual water resource application. Bradley et al. (Citation2013) examined the community response of macro-invertebrate taxa to the effect of abstraction in order to test the validity of the RIVPACS-based hydro-ecology model described in this study. The response of the individual macro-invertebrate taxa to the effect of abstraction was consistent with the RIVPACS-based model, giving confidence to the methods presented here (Bradley et al. Citation2013). Furthermore, the results presented in Bradley et al. (Citation2013) were consistent with other studies on the impacts of abstraction on macro-invertebrates that showed that impacts were most obvious with substantial dewatering (Armitage and Petts 1992, Bickerton et al. 1993, Castella et al. Citation1995, Chessman et al. Citation2011).

Implications for water resource management in the area

EFIs are used by the Environment Agency as national flow screening thresholds. Where flows are below the EFIs, this indicates reaches in which flow pressure may compromise WFD ecological status. The values set for EFIs make this a conservative approach, as would be expected for a national screening system.

Where flows are currently assessed as being below the relevant EFI, this triggers further investigations to confirm ecological impact. For sites flagged up by this screening exercise, it is possible that site-specific flow and ecology data may show that abstraction effects on streamflows and ecology are acceptable and do not contribute to the failure of Good Ecological Status under the WFD.

The flow reductions required to reduce abstraction stress to acceptable levels (as assessed by the adoption of locally based hydro-ecologically derived flow targets) are therefore, in this case, substantially lower than those implied by adoption of the EFIs. For the Worcestershire Middle Severn Sandstone groundwater body (effectively the area covered by the WMW and Bromsgrove models), the difference between applying the site-specific targets and the nationally applied EFIs is ~80 ML/d (0.93 m³/s); that is ~20 ML/d or 0.23 m³/s reduction is required using local hydro-ecological targets as opposed to ~100 ML/d (1.2 m³/s) using EFIs.

Given that re-locating 1 ML/d of abstraction is typically costed at £1–3 millionFootnote⁵, the cost implications of these findings are significant and provide a clear justification for carrying out such local investigations. In addition, the ability to use the groundwater model results to quantify the reaches that might benefit from different abstraction regimes provides useful information for the cost-benefit analysis that will be used to judge the cost effectiveness of the options considered to remediate any impacts of over-abstraction.

Assessing the significance of anthropogenic discharges

As discussed above, several of the streams in the area that are affected by groundwater abstraction have the impact on flow effectively mitigated by discharge of treated sewage effluent. While this will benefit flows, the overall effect on stream ecology will be determined by the amount of dilution of the discharge and the level of treatment applied. This gives an indication of reaches in which the attainment of Good status may be limited by water quality rather than flow.

In some of the sites being investigated, STWL is considering closing small, inefficient or outdated sewage treatment works. The resultant relocation of this discharge of water may have significant effects on local streamflows and quality. However, this process is largely managed and regulated through different teams in STWL and the Environment Agency, and so integrating this into the overall water resource management system is a challenge.

Similarly, while reduction in STWL’s mains leakage rates—down from 584 ML/d (6.8 m³/s) in 1996/97 to 474 ML/d (5.5 m³/s) in 2011/12—has reduced abstraction demand, it has also reduced the return of almost a quarter of this abstracted water to streams in urban areas (where most of the leakage occurs). In some of the areas being investigated (e.g. Bromsgrove and Coventry), main leakage appears to make a significant contribution to the local water balance, and reductions in urban streamflows in these areas will potentially be an unforeseen side effect of increased efficiencies in urban water use.

Ongoing work

The success of the analysis outlined above in quickly giving STWL and the Environment Agency confidence in the likely status of rivers across the area with respect to abstraction stress has helped to focus monitoring in the current investigations on sites at which there was the greatest level of uncertainty about the likely degree of abstraction stress; that is the screening analysis identified sites that were very likely to be stressed, (and thus required only validation sampling to confirm this) and also screened out some sites at which stress was unlikely and further investigation was clearly not required. The STWL is now carrying out a major programme of hydro-ecological field investigations (flow, groundwater levels and ecology surveys) at the sites where the level of abstraction stress was uncertain so as to quantify the level of abstraction stress more accurately and, where this is significant, to help identify the flows required to reduce stress to an acceptable level. These new data will allow the results of this modelling study to be validated. These investigations include investigation of fish as well as macro-invertebrates and also assessment of potentially confounding effects (e.g. water quality, sediment, etc.).

In parallel, the Environment Agency has licensed STWL to use the regional groundwater models that it has developed to support these investigations. In some cases STWL has carried out further localized refinement of these models so that they are able to provide a more robust assessment of the likely implications of abstraction in the catchments being studied. The results of these investigations will be fed into STWL’s 25 year Water Resources Management Plan, which is currently under preparation.