STATEMENT OF COMMON GROUND - MODELLING

Transcription

1 WATER RESOURCES ACT 1991 (AS AMENDED) SOUTHERN WATER SERVICES OBJECTION TO ENVIRONMENT AGENCY S PROPOSALS FOR MODFICATION OF SOUTHERN WATER S LICENCES TO ABSTRACT FROM THE RIVER ITCHEN AT OTTERBOURNE; SOUTHERN WATER SERVICES OBJECTION TO ENVIRONMENT AGENCY S PROPOSALS FOR MODFICATION OF SOUTHERN WATER S LICENCE TO ABSTRACT FROM THE RIVER TEST AT TESTWOOD; SOUTHERN WATER SERVICES REPRESENTATION IN RESPECT OF THE ENVIRONMENT AGENCY S APPLICATION TO REVIEW ITS LICENCE TO ABSTRACT FROM THE CANDOVER BOREHOLES OBJECTION BY SOUTHERN WATER SERVICES LIMITED To the Environment Agency s proposals for modification of Southern Water Service Ltd s abstraction licence from the River Test at Testwood (licence no. 11/42/18.16/546) (APPLICATION NO. NPS/WR/024562) ( The Testwood application ) OBJECTION BY SOUTHERN WATER SERVICES LIMITED To the Environment Agency s proposals for modification of Southern Water Service Ltd s abstraction licences from the River Itchen at Otterbourne (licence nos. 11/42/22.6/92, 11/42/22.6/93, 11/42/22.7/94) (APPLICATION NO. NPS/WR/ ) ( The Itchen application ) REFERENCE TO SECRETARY OF STATE Of the Environment Agency s application to renew its licence for abstraction from the Candover boreholes (licence no. SO/042/0031/026) (APPLICATION NO. SO/042/0031/026) ( The Candover application ) STATEMENT OF COMMON GROUND - MODELLING Page 1 of 39

2 Introduction 1. This statement of common ground describes the industry standard modelling tools and approaches used by both the Environment Agency (EA) and Southern Water Services (SWS) to provide evidence for the Inquiry. The EA has used models to assess the impacts of abstractions and discharges on environmental receptors. SWS has used models to derive the possible frequency of Drought Order requirements, to make assessments of residual flows downstream of abstractions to inform associated environmental impacts, and to assess the potential of shortfalls to public water supplies. 2. The purpose of this document, as requested by the Planning Inspector, is to provide a description of the models and modelling processes which is agreed by the EA, SWS and Fish Legal, going into the inquiry. However, it is also important to note that, beyond this statement of common ground, some of the conflicting evidence brought to the inquiry will also relate to some of the detailed modelling assumptions made, and the interpretation of associated predictions. 3. Models play an important role in creating simulations of natural river flows for a range of water resource planning and environmental purposes. These include characterising the impacts of abstraction on the environment and in the development of water supply plans. Using a range of scenarios of different abstraction licensing and operational assumptions, models are used to assess the relative changes of different scenarios against a reference condition. Using models in scenario analysis can be helpful to understand how resilient the environment and water supply systems will be when faced with droughts different to and more severe than the historical events for which monitored evidence may be available. Please see issues in dispute (pages 37 39), point In planning for water resources in England and Wales, the source of river flow data for determining output of supplies and their impacts on river flows is usually gauged river flow records. For gauging stations incorporated in the National River Flow Archive the average gauged daily flow record length is around 40 years. A substantial proportion commence in the 1960s, a period of intense network growth Page 2 of 39

3 in much of the UK, and there are around 370 stations on the NRFA with record lengths over 50 years long Recent Government and Regulatory guidance requires water companies to consider severe and extreme drought events that require consideration beyond the information held in historic records. 6. The flow records for the Test go back to 1957 and 1958 for the Itchen. They don t include three of the major droughts of the 20 th Century in 1921/22, 1933/34 and 1943/44. Therefore, to assess performance of water supplies over a longer period to include these major droughts, river flows have to be generated from rainfall records, a few of which extend back to the early 20 th Century. 7. The situation for the Rivers Test and Itchen is further complicated by them being chalk streams, meaning that their catchments are largely underlain by highly permeable chalk rock, which absorbs most of the rainfall into groundwater aquifers. River flows are then largely driven by groundwater levels when groundwater tables intersect the valley sides at the spring line they generate the springs which are the main source of river flows. Conversely, heavy rain usually has little impact on river flows because the water is absorbed by the chalk, usually taking many weeks to sink down to the water table. The water table generally rises slowly through the winter, peaking in late spring and falling through the summer and autumn, when most rain is soaked up by vegetation. The process is illustrated in Figure 1: Figure 1 - Illustration of how river flows are generated in chalkstreams 1 National River Flow Archive website. Page 3 of 39

4 8. The natural process of converting rainfall into river flows is highly complex, depending particularly on the permeability and storage capacity of the chalk rock, which varies considerably throughout the valley. Therefore, the modelling of the generation of river flows from rainfall is also highly complex. The groundwater model of the Test and Itchen catchments uses state-of-the-art modelling software and has been undertaken by leading experts in the field. The modelled river flows are subject to considerable uncertainty, as will be shown later in this SOCG. 9. Appendix A is a table which lists all of the models and broader data analysis approaches which have been used to develop evidence to the public inquiry, including the interpretation of ecological responses to changes in river flows and temperature (in the second part of the table). The three water resources models in the first part of the table are of particular relevance to the inquiry and are the focus of this statement of common ground. They are described in the following order: the weather generator model; the Test and Itchen groundwater and river flow model and the Aquator water resource system model. 10. It is important to note for the purposes of this Inquiry that there are two versions of the Test and Itchen groundwater and river flow model that have been used. The first is a version produced by the EA in 2013 as part of a project involving SWS and Portsmouth Water. This version has been agreed as being fit for purpose in relation to the original modelling objectives set for its development. It provides outputs for the period and has been calibrated against measured data. Within the context of this Inquiry the EA s current (2013) version of the Test and Itchen groundwater and river flow model has been used by the EA to assess the impact on river flows and groundwater levels of its proposed use of the Candover Scheme. The EA has not used output from the Test & Itchen model in any of its technical assessment associated with the Testwood licence. Previous versions of the model were used to inform and develop the licence changes proposed in the River Itchen SAC Site Action Plan (2007). SWS have developed this model to produce a different version for use alongside the weather generator to produce estimates of river flow and levels for more extreme climatic events. This version was produced by SWS alone. Page 4 of 39

5 SWS have used this amended version of the model to generate river flows that are the natural resource simulations passed as an input to Aquator. The operation of the abstractions and discharges derived by Aquator are then passed back into the SWS version of the Test and Itchen model to simulate the associated patterns of flow and level impacts. Alongside the Aquator model, this model has also been used by SWS to assess how often Drought Orders would be needed in relation to the river flow conditions for both the Rivers Test and Itchen proposed by the EA. SWS have also used their version of the model to consider the potential impacts on river flow and groundwater levels of their proposed use of the Candover Scheme, alongside field evidence. Summary inputs, outputs and relationships between the models 11. Figure 2 shows how both the EA and SWS (and Portsmouth Water) have used the Test and Itchen groundwater and river flow model for the assessment of abstraction and discharge impacts on groundwater levels and river flows in the climate context of the recent January 1970 to March 2011 period. During this period more than 40 rain gauges and potential evaporation data from MOSES (Met Office Surface Exchange Scheme) are available to provide climate inputs to the model, and its historical simulation outputs can be compared with gauged river flows and observed groundwater levels at many locations as part of the process of model refinement. The extent and parameterisation of the model have been developing since As with all hydrological and groundwater models, its calibration is by no means perfect and there can be substantial differences between modelled flows and flows recorded by river flow gauges. Improvements to the groundwater and river flow model it are ongoing. As part of their ongoing modelling strategy the EA is currently reviewing, updating and calibrating this model. Impact predictions are derived by comparing outputs from runs with alternative abstraction, discharge and Candover Augmentation Scheme rules (e.g. the Natural scenario in which all these influences are switched off, and the Fully Licensed scenario which assumes maximum permitted abstraction is always taken, irrespective of need for it). Page 5 of 39

6 Rainfall from 40+ rain gauges and MOSES Potential Evaporation (PE), 1970 to 2011 Abstraction & discharge time series* Optional flow trigger rules for the Candover Augmentation Scheme* Test & Itchen GW & river flow model Conceptual understanding, boundaries and aquifer parameter assumptions River flows & GW levels, * *Model runs may use: historical abstraction and discharge time series for calibration comparisons with gauged flows and observed groundwater levels; or Scenario abstractions, discharges and Candover Scheme rules for predictive flows and groundwater levels (e.g. Natural, Recent Actual, Fully Licensed scenarios). Predicted impacts come from the differences between scenario runs (e.g. Fully Licensed minus Natural) Figure 2 - Use of the Test and Itchen groundwater and river flow model by the EA and SWS for environmental impact assessment in the context of the 1970 to 2011 climate. 12. Figure 3 summarises the relationships between the Test and Itchen groundwater and river flow model and the other modelling tools which have been used by SWS to consider public water supply provision and environmental impacts during drought periods different to and more severe than those that happen to have occurred in the recent monitored period. This work involves long prediction periods based on stochastically generated climate predicted from the recent historical record. Rainfall inputs are estimated at only 3 rain gauge locations and PENSE Potential Evaporation data inputs have been used for these prediction model runs. Each of these models is summarised below, illustrating how information is passed between them. Page 6 of 39

7 Long stochastic rainfall for 3 rain gauge locations & PENSE based PE sequences Weather generator model Test & Itchen GW & river flow model Operational river flows & GW levels during selected droughts* Natural river flows (80 historical years stochastic years) Operation assumptions for selected drought episodes* Public water supply demand Aquator water resources system model Abstraction, discharge & transfer time series* Demand saving & operating rules, other abstraction, discharge, licence constraint & Drought Order assumptions* River flows* Supply deficits* *Model runs use predictive scenario abstractions, discharges, licence conditions, demand constraints & drought order availability rules, assuming current public supply demands (e.g. with drought orders or without drought orders ). Predicted environmental impacts come from the differences between runs (e.g. with drought orders minus without drought orders ) Figure 3 - Summary model relationships and information transfers for SWS drought modelling 13. Firstly, the weather generator model has been used by SWS to project long stochastic time series of rainfall and potential evaporation based on an analysis of historical climate from the 20 th century. A long climate sequence has been compiled which combines rainfall records from three rain gauges at Otterbourne, Salisbury and Greywell, from 1918 to 1997 with 2000 year projected rainfall time series for these three locations from the weather generator model. These projected rainfall records have been used with associated projections of potential evaporation based on PENSE data (Penman Evaporation for South East England) to complete the climate inputs for modelling. Drought events of estimated return period severity have then been selected for further modelling and analysis. 14. Secondly, the Test and Itchen groundwater and river flow model has been used by SWS to simulate natural flows based on the long historical and stochastic climate inputs, without any abstractions or discharges. The river flow outputs from this model run provide the natural resource predictions passed as an input to Aquator. For selected drought periods the operation of the abstractions and discharges subsequently derived by Aquator can then be passed back into the Test and Itchen model to simulate operational scenario patterns of river flows and Page 7 of 39

8 groundwater levels during drought periods, for some aspects of further analysis. The Test and Itchen model is particularly important, alongside field testing evidence, for the assessment of the groundwater level, floodplain wetland and river flow impacts of the proposed Candover Augmentation Scheme operations. Southern Water's proposed drought order Scheme is to pump groundwater to augment low flows in the Chalk base flow dominated Candover Stream and middle River Itchen in order to support abstraction for public supplies and river flows to these abstractions. As stated in paragraph 10, the EA s current version of the Test and Itchen groundwater model has been used by the EA to assess the impact on river flows and groundwater levels of its proposed use of the Candover Scheme, which differs to the use by SWS set out above. Please see issues in dispute (pages 37 39), point Thirdly, the Aquator model has been used by SWS to inform its water resource planning, including its Water Resource Management Plan (WRMP) (Central and Eastern Areas only), and the Drought Plan (all Areas, including Western Area). Aquator includes a conceptual representation of SWS s water resource and supply system, the demands that need to be met, abstraction licence conditions and strategic operational rules. Together with time series of river inflows (taken from the Test and Itchen groundwater and river flow model), these are used to provide reference conditions against which the performance of the water supply system under different abstraction licence and operational scenarios can be compared. Aquator also produces simulations of river flows which have been influenced by the operation of abstractions and discharges, and by the assumed distribution of flow between river channels. These flow simulations are particularly important for the lower reaches of both the Rivers Test and Itchen over the less permeable deposits where there are multiple channels and surface water abstractions are governed by abstraction licence conditions including hands off flow constraints. Please see issues in dispute (pages 37 39), all points relevant. 16. SWS have chosen to use models such as the Weather Generator Model, the Test and Itchen Groundwater Model and the SWS Western Area Aquator model because they believe they represent good practice for water resources assessments. Page 8 of 39

9 This is because for their purposes, assessments need to consider supply resilience and river flow impacts under a fuller and wider range of events than can be found in the recent historical record. Please see issues in dispute (pages 37 39), all points relevant. 17. However, the use of these models for assessing performance and impact of supplies under the climatic conditions of the past 60 years can give less reliable results than using the gauged flow records that are available for both rivers since It should be noted that gauged records have to go through a process of naturalisation with respect to the abstraction and discharges occurring at the time a process which itself can be associated with some uncertainties, particularly in groundwater dominated areas where groundwater abstraction is a significant pressure. In the case of the River Test and River Itchen the influence of flow diversions, flow control structures and historical channel management practices can also be important. These factors may also vary in historical (and future) influence. 18. Each of these three models is described in more detail below, including the assumptions and approaches used to 1997 and long climate projection inputs from the Weather Generator model for drought resilience assessment. 19. In line with current regulatory guidance, SWS has to consider the resilience of public water supplies to drought events which are more severe (i.e. of longer return periods) than have occurred within the recent past, assessing the environmental impacts associated with alternative drought abstraction scenarios. 20. This has first involved looking back in time, to derive climate sequences from 1918 which cover the 1920 to 1922 drought event, considered an important severe event for assessment. Three rain gauges with good records for the early twentieth century Otterbourne, Salisbury and Greywell (Figure 4) were chosen to provide simplified rainfall inputs applied to the Test and Itchen model from 1918 to 1997 (with spatially distributed factoring). These rainfall records (and others) also fed into the stochastic Weather Generator modelling process. Page 9 of 39

10 Greywell Salisbury Otterbourne Figure 4 - Rain gauge locations from which historical gap-filled records back to 1918 and stochastically generated long time series are input into the Test and Itchen groundwater and river flow model (colours indicate the areas over which the rain gauge data are distributed) 21. The Weather Generator model SWS use was developed by Newcastle University (Serinaldi and Kilsby, ) and was originally used to support SWS s 2014 Water Resource Management Plan (WRMP14). The weather generator model has been subsequently updated, enhanced and refined for the draft 2019 WRMP which was discussed with the EA. Adaptations of the same model have also been used to support other Water Companies WRMPs, regional water resource modelling for Water Resource South East (WRSE) group and the National Water Resource Planning study Rainfall input data are based on historical records and vary in length by available gauge record. Data from the Otterbourne, Greywell and Salisbury rain gauges are used to generate input for the SWS Test and Itchen groundwater and river flow model. Data quality can vary substantially with time and records are often sparser 2 Serinaldi, F. and Kilsby, C., 2012, A modular class of multisite monthly rainfall generators for water resources management and impact studies. Journal of Hydrology, , pp Water Resources Long-term planning framework ( ), Water UK, 2016 Page 10 of 39

11 in the past. The Weather Generator can handle such data gaps, though it should be recognised that the bigger the gaps, the greater the uncertainty in model output. 23. To appropriately reflect UK climate variability, the weather generator also incorporates two regional scale climate indicators known to influence UK rainfall; specifically the North Atlantic Oscillation and Atlantic Sea Surface Temperature. The weather generator fits statistical distributions through these rainfall and climate indicator data and employs a stochastic procedure to generate rainfall time series at points where input rainfall data are provided (typically rain gauges). These sequences take account of spatial variability between gauges. Climate input data cover the 20 th century from 1908 to 1998 and are derived from publicly available data sets produced by the University of East Anglia Climate Research Unit and Met Office Hadley Centre. 24. The weather generator can produce extremely long time series of monthly rainfall, which by the nature of the stochastic process includes infrequent more severe drought events than have been experienced historically but which are historically plausible and consistent with the behaviour of the UK climate. Approximately ~1000 stochastic simulations of the historic climate sequence were generated which created a rainfall time series for each gauge of approximately 100,000 years in length. Additional analysis then classified each rainfall time series in terms of rainfall deficits (compared to long term average), and estimated drought indicators including calculation of Standard Precipitation Indices. Each period of rainfall deficit (or drought) was characterised in terms of overall duration of deficit, the intensity (i.e. the magnitude of the deficit) and its estimated probability. 25. It is not computationally practical to run the full 100,000 year stochastic weather sequence through either the Test and Itchen groundwater and river flow model, or the Aquator model, and hence a shorter 2000 year sequence was sub-sampled. The 2000 year sequence was chosen to be one that replicated the underlying average statistics of the parent data set in terms of rainfall metrics, i.e. a sequence was chosen that was not statistically significantly wetter or drier than the 100,000 year data set. Sensitivity testing of alternate 2000 year sequences was performed to validate the selection. Page 11 of 39

12 26. Corresponding Potential Evapotranspiration (PE) sequences for model input were generated using a nearest neighbour matching of historical PENSE-based PE data. This matches the synthetic rainfall to the closest observed month in the historical record and subsamples the associated PE data for that month. Some additional smoothing of the PE data is applied to account for observed persistence effects observed in long duration droughts. This procedure also disaggregates the monthly stochastic rainfall to a daily input sequence appropriate for use with the runoff and recharge component of the Test and Itchen groundwater and river flow model. 27. Design return periods of rainfall deficits and drought events were initially based on those required for the draft 2019 WRMP investment modelling from normal dry year (1 in 2), through moderate drought (1 in 20 to 1 in 100 year events), severe drought (1 in 200 year) and extreme drought (1 in 500 year). As a continuous time series of rainfall is simulated the modelling undertaken reflects a full continuum of drought events across and beyond this range and chosen events represent indicative samples of the data set for the stated drought probabilities. The Test and Itchen groundwater and river flow model The EA Test &Itchen groundwater model 28. The Test and Itchen model was built and is owned by the Environment Agency. It has been applied to a range of water resources investigations by both the EA and SWS over the last 15 years. The table below shows the development and versions of the model owned by the EA and developed and agreed between the EA SWS and Portsmouth Water: Page 12 of 39

13 Date Description Model Period Purpose Use 2003 River Itchen st time a groundwater River Itchen Groundwater (warm up period model of the entire Sustainability Study Modelling Study from ) Itchen catchment had 2005 Test & Itchen Groundwater Modelling Study (warm up period from ) been produced Extension of Itchen model into Test catchment Review of Consents Stage Test & Itchen model update 2013 Test & Itchen groundwater model refinement (warm up period from ) (warm up period from ) Model data sets updated from Jan 2003 to Dec 2005 Model extended and re-parameterised (SW had previously updated model in 2012 from ) Routine update for use of model in RSA investigations Routine update for use of model in RSA investigations and to assess Candover Scheme 2018 Test & Itchen groundwater model update and refinement Table 1 Likely to be Intention to update and recalibrate model and potentially move to a new version of the modelling code The Environment Agency Test and Itchen Groundwater Model Routine update for use of model in RSA investigations 29. It combines a daily simulation of rainfall to routed runoff and recharge carried out by the 4R code (Heathcote et al, ) with calculations of responding Chalk groundwater flow, level and flows to and from springs, winterbournes and perennial river channels carried out in the MODFLOW code (McDonald and Harbaugh, ) using half-monthly (~15 day) stress periods. Surface water abstraction and discharge influences are included in 4R, with groundwater abstractions in the MODFLOW simulation. The model simulation is carried out on a regular 250 m grid of cells. This 4R-MODFLOW code combination has been widely used for regional groundwater abstraction impact modelling across England by the EA and water companies and represents an industry standard approach. 30. The main purpose of such models is to understand how patterns of groundwater abstraction impacts vary both spatially and in time in the context of a naturally 4 Heathcote, J.A., Lewis, R.T. & Soley, R.W.N Rainfall routing to runoff and recharge for regional groundwater resource models. Quarterly Journal of Engineering Geology and Hydrogeology, 37, McDonald, M. G. & Harbaugh, A. W A Modular Three Dimensional Finite Difference Groundwater Flow Model. U.S. Geological Survey Techniques of Water Resources Investigations Report 06- A1. Page 13 of 39

14 fluctuating and regionally distributed water table, and combined with surface water flow influences. They are uniquely able to predict the changing patterns of river flow and groundwater level impacts during droughts and wetter periods. 31. Field-based test pumping, investigations and monitoring data also provide valuable information to characterise abstraction impacts, but only in the context of the observed hydrological conditions. Even with good field measurements, interpretive analysis is often required to derive impacts based on estimates of what would have happened without the abstraction or discharge being tested. Use of a groundwater and river flow model enables various scenarios to be tested to explore different ways of operating abstractions and/or the groundwater augmentation of river flows such as the Candover Scheme, under differing hydrological conditions. In this way, the likely impact of operating the abstractions and discharges at proposed licence or drought order rates, and under historical or design drought conditions, can be assessed alongside field evidence, and with due consideration of the uncertainties remaining in the model. 32. Field monitoring of levels and flows recorded during trial and operational pumping of the Candover Augmentation Scheme has been used to inform development of the EA's Test and Itchen groundwater and river flow model refining its parameterisation to better match observations. The model s boundary conditions were extended by the EA to cover the whole of the Chalk aquifer around the Candover Scheme and it was last updated in The model uses rainfall inputs from the many (more than 40) rain gauges located across the Test and Itchen catchments, and MOSES PE data in order to provide a simulation of the historical climate period 1970 to March When updated and extended in 2013, the model was accepted as fit for purpose, although its simulated flows, groundwater levels and impacts need to be considered alongside local field evidence. There is typically less confidence in absolute values of flow for the winterbourne reaches than at flow gauging stations and the model is not perfectly calibrated to match periods of winterbourne flow and drying out. This is set out in more detail in the Amec Foster Wheeler Technical Note 29388TN568 in the Inquiry core documents. Modelled impacts of abstractions may extend both spatially and temporally beyond what tends to be observed Page 14 of 39

15 during periods of test pumping. For those reasons, care must be used in analysing output and so the inquiry assessments consider observed as well as modelled data to assess potential impacts. 34. Concerns about the validity of the EA s Test and Itchen groundwater and river flow model had been raised previously by WWF/HIWWT whilst SWS were using it to assess the impacts of alternative Candover Augmentation Scheme operational proposals (letter to SWS of 17 th December 2014 during the work of the Augmentation Technical Working Group (TWG) of the Testwood 2015 RSA project). Exchange of commentaries between WWF, SWS and Amec Foster Wheeler has resolved that WWF and HIWWT accept that the Test and Itchen model should be used as the best available tool to inform assessment of the impacts of the scheme during drought, but with recognition of the caveats around its local accuracy and reliability, informed by field evidence. 35. Amec Foster Wheeler Technical Note 39715TN052i2 (in the inquiry core documents) summarises this exchange of commentaries and includes technical notes describing the latest rounds of Test and Itchen model refinement as well as a published paper with many examples of the use of such models for the assessment of groundwater abstraction impacts, written collaboratively by EA, SWS and consultancy staff. 36. The EA is conducting a routine review of the Test and Itchen groundwater and river flow model, in discussion with SWS, South East Water and Portsmouth Water, with the intention of updating the simulation period and further improving its calibration through parameter and boundary refinements. One of the aims of this EA review is to establish a new PE input data set which will be stable into the future. This is because the MOSES data on which its current fit for purpose calibration is based is being discontinued by the Meteorological Office. SWS version of the Test & Itchen groundwater and river flow model 37. SWS consider that a simpler input rainfall design - with fewer rain gauges - is necessary for the simulations for the inquiry, draft WRMP19 and Drought Plan which cover earlier historical droughts (such as ) and more severe and extreme droughts included within the longer period climate simulation derived stochastically by the weather generator model. This is based on only three rain Page 15 of 39

16 gauges, one potential evaporation time series and factoring according to long term spatial distributions. 38. It should be appreciated that the requirement to look at extreme droughts using stochastically generated rainfall and river flows is to meet a new Government requirement to look at resilience of supplies to droughts more extreme than those in the recent historic record. SWS applied the weather generation technique in preparing its 2014 WRMP. It was subsequently taken up in England and Wales within the National Water Resources Planning Framework study for Water UK in It has been used by the Water Resources South East Group s latest work and many water companies are using it for their latest WRMP preparations. The use and application of this technique remains in dispute by the EA and Fish Legal. Please see issues in dispute (pages 37 39), all points relevant. 39. For generation of river flow data used in the Aquator modelling for this Inquiry, SWS have used a different version of the Test & Itchen groundwater model, compared to the version used by the EA for environmental assessment and licence determination. The differences in quality of output and the significance of these differences is a subject of dispute in this Inquiry. 40. In order to extend the model back to before the drought it was also necessary to use PENSE-based PE input data for one climate station because MOSES data are not available for that period. The recent PENSE time series from 1970 was also used in association with the rainfall time series to derive related PE inputs for the long term stochastic climate simulations. 41. Two climate sequences were collated for an initial run of the Test and Itchen groundwater and river flow model in order to generate natural flows for the Aquator model: An 80-year historical period from 1918 to 1997; and A 2000-year stochastic sequence. 42. The flow calibration of the model based on the simpler 3 rain gauge and PENSE PE inputs, as used to generate river flows for the Aquator modelling, is not as close to gauged records during the 1970 to 1997 recent period as the 40+ rain gauge and MOSES PE version of the model which has been previously used by both the EA and SWS for environmental impact assessment modelling over the recent period. Page 16 of 39

17 The flows simulated from the 3 rain gauge and PENSE PE inputs are generally lower, particularly during the drought period when they fall well below the gauged record. Figure 5 shows that the 40+ rain gauge and MOSES PE 1970 to 2011 Test and Itchen model (red line), as used by the EA for environmental assessment and licence determination, which uses historical abstraction and discharge records produces a simulation of river flows at the Broadlands gauging station on the River Test which captures the low flow variability of the gauged record well. There are differences between the gauged and simulated flows in individual model stress periods (2 periods per month) but the flow duration curve statistical summary (Figure 6) is well matched below Q50 (i.e. the flow exceeded for 50% of the time). Both of the three rain gauge input models plotted in Figure 5 use recent abstraction demands rather than historical records which may explain some of the differences with the gauged data (although the differences in abstraction demands, perhaps about 20 Ml/d, are much less than the flow differences, which are often Ml/d). However, the application of PENSE potential evaporation data used for drought prediction runs carried out to date because it covers events in the early twentieth century results in much lower simulated flows (green line). This is particularly evident during the period but is also clear on the flow duration curves (FDC) shown on Figure 6. Page 17 of 39

18 Broadlands time series flows Figure 5 -Time series of river flows simulated by the Test and Itchen model at the Broadlands gauging station on the Test, compared with gauged records over the periods and inclusive. Page 18 of 39

19 Broadlands Flow Duration Curves Figure 6 - Flow duration curves for modelled and gauged flows at Broadlands on River Test 43. Although the SWS PENSE-based modelled flows follow the general pattern of the gauged river flows, there are at times substantial differences between modelled and gauged flows and between the three different model versions. These differences in flows could be significant when modelling the timing, frequency and duration of the cut-backs in abstraction from the Rivers Test and Itchen. These reductions are triggered when river flows drop below the proposed hands-off flows that are a primary focus of the Inquiry. The modelled flow differences will have an effect on the modelled frequency and duration of the demand restrictions, and it is the significance of this effect is a focus of the Inquiry. 44. The Aquator modelling has had to use the least well calibrated ( ) version based on the simpler three-rain gauge and PENSE PE inputs, because this is the only version able to cover the major historic droughts of 1921/22, 1933/34 and 1943/44 and the only version able to generate the stochastic droughts of various probabilities which are a key component of SWSs case. Page 19 of 39

20 45. The flow calibration of the SWS Test & Itchen model version used in the Aquator modelling, is not as close to gauged records during the 1970 to 1997 recent period as the 40+ rain gauge and MOSES PE version of the model which has been previously used by both the EA and SWS for environmental impact assessment modelling over the recent period. The flows simulated from the three-rain gauge and PENSE PE inputs are generally lower, particularly during the drought period when they fall typically Ml/d below the gauged record, as shown in Figure For example, simulated flows at Broadlands based on the PENSE PE input data are around 100 Ml/d (at Q98) TO 150 Ml/d lower than the gauged flows or if MOSES PE data are used. This applies throughout the range of flows normally encountered in spring, summer and autumn. Figure 7 and Figure 8 show there are also modelled flow discrepancies for the Itchen. The EA Test and Itchen modelled flows that do not fit as well as on the Test, particularly in , and the PENSE-based model flows are consistently lower. This means modelled river flows used in Aquator will have overestimated the frequency and duration of cut-backs in abstraction. The significance of this in respect of demand restrictions, drought orders, residual flows and environmental impact remain in dispute. Page 20 of 39

21 Allbrook & Highbridge time series flows Figure 7 - Time series of river flows simulated by the Test and Itchen model at the Allbrook and Highbridge gauging station on the Itchen, compared with gauged records over the periods and inclusive. (NB Run172 has no Candover Scheme which was actually operating during these drought periods) Page 21 of 39

22 Allbrook & Highbridge Flow Duration Curves Figure 8 - Flow duration curves for Test & Itchen modelled and gauged flows at Allbrook and Highbridge on River Itchen 47. When modelling the flows in the Great Test branch of the River Test and the triggering of the Great Test HOF, the model assumes that the River Test splits into the Great and Little Test branches in the ratio 2:1, according to the so-called Coleridge Award. The limited gauged flow records available show that there have often been substantial variations from this assumption. 48. Figure 9 shows that the PENSE PE long term average for the 1970 to 1997 recent historical period is only a little higher than the equivalent MOSES average (by 23 mm/a). Both show a warming trend through this period although this has not been carried forward into the projection of the recent PENSE data into the 2000 years of stochastic climate simulation. It is only when considering more detailed daily comparison of these alternative PE data sets through the year that the reasons for the differences in the water resources derived becomes apparent. There is more daily variability within the MOSES data compared (typically lower values on wetter days), compared with a smoother PENSE time series. MOSES PE also falls to zero and is often negative during winter periods whereas PENSE remains positive and higher at these times. Page 22 of 39

23 Historical 2000 year projection (no climate change) Figure 9 - Comparisons of MOSES and PENSE potential evaporation data inputs: long term averages ( ), annual totals and daily detail ( ) 49. This description of the difference which the input climate data sets can make needs to be borne in mind when considering the predictions from the Aquator modelling reported to date which have used the natural flows from the SWS Test and Itchen model based on the three rain gauges and PENSE PE inputs. SWS is undertaking a sensitivity analysis considering the use of MOSES as an alternative basis for potential evaporation projection to cover both the hindcast climate period before 1970 and the 2000 year stochastic simulation after The half-monthly natural river flow inputs derived from this initial SWS Test and Itchen model run are interpolated to daily time steps for use in the Aquator model. 51. The full 2080 year daily simulations carried out in Aquator determines the time series of abstraction and discharge operations required to meet public water supply demands, in so far as this is possible within the abstraction licence conditions (daily, monthly and annual volume limits and hands off flows), together with drought order assumptions. Various Aquator scenarios are run to test what difference the availability of drought orders (allowing abstractions below the hands off flows, or triggering the operation of river flow augmentation) makes to the flows in rivers, and the provision of the public water supply. The impacts of a drought order scenario are compared with a reference condition baseline the situation that would occur during drought but without the drought order. Page 23 of 39

24 52. The Test and Itchen groundwater and river flow model is the best available tool to assess the impacts of groundwater abstraction and augmentation from the Candover Scheme boreholes because it includes explicit, three-dimensional representation of groundwater and surface water processes. In order to assess the possible groundwater level and related impacts of operating the Scheme under different operating assumptions, SWS run the abstraction and discharge time series derived from selected scenario runs of the Aquator model through the SWS Test and Itchen groundwater and river flow model. Technical note 29388tn568, July 2015 (includes appendix C focused on the flow calibration of the EA model update, and appendix D on the groundwater level calibration, with a particular focus on Candover). 53. The Aquator time series include the periods and rates of augmentation pumping required from the Candover Scheme. These are split in the Test and Itchen model between a discharge at the existing outfall on the Candover Stream of up to 2 Ml/d, with most of the remainder, up to 25 Ml/d discharged further downstream to the River Itchen. The SWS Test and Itchen model has also been used to consider further alternative control rules for the 2 Ml/d environmental support component of the proposed Scheme which could continue to operate after the public supply support is switched off, or could be used more frequently, if deemed appropriate by the EA. 54. However, because the run times for the 2080-year climate sequence in the groundwater model are long, a selection of historical droughts and more severe, less frequent events selected from the 2000 year stochastic sequences were compiled, interspersed with three average rainfall recovery years, for a shorter simulation. The compiled sequence includes a period of run-in to each of the selected droughts, followed by a period of recovery comprising three years of average rainfall. This is possible because the Chalk aquifer responds rapidly to the climate. The shorter model runs provide a manageable approach to assessing the differences in impacts on flow and groundwater levels between the reference condition and drought order scenarios. 55. Table 2 is a summary of the SWS Test and Itchen groundwater and river flow model versions and run numbers which tracks the history of its development and includes the runs which have provided evidence for the Inquiry. Page 24 of 39

25 Note AMP6 = ; AMP5 = ; AMP4 = ; AMP3 = Date Description/Version Run Period modelled and weather inputs Use AMP3 AMP3 & AMP4 AMP4 AMP5 Itchen GW model: Phase 1 Improved Test & Itchen GW model: Phase 3 Improved Test & Itchen GW model: Phase 3 plus augmentation coding changes and extended period As above, plus stochastic hydrology e.g.49, 55, 58 (of Itchen Model) e.g. 53, 54, 57, 63 e.g. 65 e.g. 80 e.g. 108 AMP5 Post-refinement & extension e.g. 124 to 127 AMP5 Post-refinement & extension e.g. 126,127, 131, 132 AMP6 Post-refinement & extension e.g. 134, 135, 137 AMP6 Post-refinement & extension e.g. 140 to 154 AMP6 Pre-refinement & extension (for comparison with AMP4 and AMP5) e.g Standard Period: 1970 to rain gauges & MOSES PE Standard Period: 1970 to rain gauges & MOSES PE Extended Drought: 1913 to rain gauges & PENSE PE Extended Drought: 1913 to 1969; 1970 to year stochastic 3 rain gauges & PENSE PE & associated projections Standard Period 1970 to rain gauges & MOSES PE Standard Period 1970 to rain gauges & MOSES PE 45 year period selected stochastic droughts: 3 rain gauges & PENSE PE & associated projections Some runs use rain gauges & MOSES PE. Others use historical and stochastic climate splices based on 3 rain gauges & PENSE PE & associated projections 2000 year stochastic climate 3 rain gauges & PENSE PE & associated projections AMP6 Post-refinement & extension e.g year historical plus stochastic climate 3 rain gauges & PENSE PE & associated projections AMP6 Post-refinement & extension e.g AMP6 Post-refinement & extension e.g AMP6 Post-refinement & extension e.g rain gauges & MOSES PE rain gauges & MOSES PE 66 year period selected historical and stochastic droughts 3 rain gauges & PENSE PE & associated projections River Itchen Sustainability Study EA: HD Stage 3 RoC EA: CAMS EA: HD Stage 4 RoC SWS AMP4 WRI SWS AMP5 Water Resources Workpack EA Refinement SWS AMP5 Water Resources Workpack SWS: Testwood 2015 RSA Project: Augmentation Technical Working group SWS: Testwood 2015 RSA Project: Augmentation Technical Working group Assessment of potential impacts of SWS Candover Scheme for Technical Working Group meetings with EA, Natural England and Upper Itchen Initiative SWS: dwrmp19 DO SWS: to generate natural flows for Aquator for the Drought Plan & inquiry Inform Anton NEP Study on potential sphere of influence of Andover PWS Model updated to June 2017 and the different recharge forecasts to end 2020 to inform response to DEFRA on Water Resources situation SWS: with Aquator influences to assess SWS Candover Scheme impacts for Drought Plan & inquiry Table 2 SWS Test & Itchen groundwater and river flow model - key steps in development Appendix D sets out a table of the EA s runs of the Test & Itchen groundwater model for the Candover licence renewal. Page 25 of 39

26 The Aquator Model 56. Aquator is an industry-standard water resources modelling application used by water companies and others for a range of water resource assessment purposes. It is used to determine whether a water resource network can supply sufficient water from the sources represented in the model to meet demands under given hydrological and abstraction licence conditions. Aquator uses a conceptualisation of sources, water treatment works, demand centres and transfer links. Aquator performs daily water balance calculations on a daily time step to simulate how demands across the system would be met subject to the abstraction licence and operating conditions. 57. A set of reference conditions is specified for a model run to simulate abstractions, residual flows and any supply shortfalls against which the outcomes of alternative scenarios can be assessed. 58. The 2017 version of the SWS Western Area Aquator model was developed by SWS from an earlier 2015 version of the model. For the 2017 version, the previous Hampshire South Water Resource Zone (WRZ) has been subdivided into four separate WRZs: Hampshire Rural; Hampshire Winchester; Southampton East; and Southampton West. The key features of the 2017 version of the model are summarised in Technical Note DG16 which is included as Appendix B. The following sections of this Statement of Common Ground provide some complementary explanations. 59. All models are simplifications of real systems. Nonetheless, supported by the underlying models described above, the SWS Western Area Aquator represents the best means available for comparing the implications for abstraction, supply deficits and river flows of different assumptions regarding licence conditions and hands-off flow constraints, using the flow records generated for periods outside the availability of the 40-rain gauge data and the gauged river flow records. 60. However, the accuracy of the Aquator modelling is constrained by the quality of the river flow data used in the model. For the Aquator modelling used for this Inquiry, the river flow data has used the three-rain gauge, PENSE PE version, which does not match gauge flows as well as modelling using 40-raingauges and MOSES PE, as shown on Figures 5 and 6 above. Page 26 of 39

27 Key SWS assumptions for the SWS Western Area Aquator reference condition 61. For the purposes of this paragraph the reference condition is the run DP EA Section 52 licence proposals for River Test and River Itchen Level 1, 2 and 3 demand restrictions applied; Bulk Supply from Portsmouth Water included; no Drought Orders including no Candover Drought Order. The key assumptions for the reference condition (some of which are varied for different Scenario Runs) are all listed in this section:- Abstractions from the River Test are limited by the proposed EA Section 52 licence condition for Testwood, including a HoF for the River Test total outflow of 355 Ml/d; Abstractions from the River Itchen are limited by the Itchen sustainability reductions, in particular a HoF of 198 Ml/d at Allbrook and Highbridge and a HoF of 194 Ml/d at Riverside Park; Demand-side drought interventions are triggered when simulated residual flows at Allbrook & Highbridge breach specified Drought Trigger Levels (DTL)s; A constant 15 Ml/d abstraction at Gaters Mill supplies Portsmouth Water s own customers and an additional 15 Ml/d supply is available to SWS from Portsmouth Water when DTL Level 1 is breached; SWS existing infrastructure allows transfers from Southampton West to Southampton East water resource zones up to a maximum of 24 Ml/d. Transfers occur when abstraction from the SWS Lower Itchen sources is constrained either by monthly licence constraints and/or by the HoF of 198 Ml/d at Allbrook and Highbridge; and An indication of the reference conditions and the possible combination of options for different Scenario Runs is given in Table 3: Page 27 of 39