Building resilience to overheating into UK 1960s hospital buildings within the constraint of the national carbon reduction target: adaptive strategies

The National Health Service (NHS) Estate in England includes 18.83 Mm 2 of acute hospital accommo- dation, distributed across 330 sites. Vulnerability to overheating is clear with 15,000 excess deaths occurring nationally during the July 2003 heatwave. The installation of mechanical cooling in existing hospitals appears to be the inevitable recommendation from NHS patient safety risk assessments but the carbon implications would undermine the NHS Carbon Reduction Strategy. NHS CO 2 emissions constitute 25% of all public sector emissions, equivalent to 3% of the UK total. In the post-2008 economic climate, the likelihood of wholesale replacement of the NHS Estate is signi ﬁ cantly diminished; refurbishment is now of increasing interest to the Trusts that together make up the NHS. The research project ‘ Design and Delivery of Robust Hospital Environments in a Changing Climate ’ seeks to understand the environmental performance of the current NHS Estate and, from this, to establish its resilience. To this end, hospital buildings operated by four NHS Trusts are being monitored and simulated using dynamic thermal models calibrated against measured data. Adaptive refurbishment options are proposed and their relative performance predicted against the existing internal conditions, energy demands and CO 2 emissions. This paper presents ﬁ ndings relating to one representative type building, a medium-rise ward block dating from the late 1960s. It shows that this particular type may have more resilience in the current climate than might have been expected, that it will remain resilient into the 2030s, and that relatively non-invasive measures would extend and increase its resilience whilst saving energy. (cid:1) 2012 Elsevier Ltd. All rights reserved.


Introduction
This paper investigates the resilience of various refurbishment schemes for a typical hospital tower building in the UK to climate change, focussing on summertime overheating. The National Health Service (NHS) Estate (England) comprises 28.38 Mm 2 of accommodation. In England, there are 330 acute hospital sites with a gross floor area of 18.83 Mm 2 ; 8.3 Mm 2 is occupied by patients [1]. The NHS is required by law to reduce its carbon emissions [2] and stringent targets for energy demand have been set. The NHS reports that it is currently responsible for 30% of UK total public sector carbon emissions, and 3% of all UK emissions [3]. Its annual carbon footprint, as of 2007, was 21 Mt pa (million tonnes of CO 2 ) of which 24% can be attributed to building energy [3]. Although energy is being used more efficiently, consumption has risen 40% since 1990 and increased by 2 Mt between 2008 and 2009 [3]. Attempts to reduce consumption in England by 0.15 MtC (million tonnes of carbon) between 2000 and 2010 appear to have failed [4]. According to the Department of Health's 'Health Technical Memorandum 07-02 0 , 44% of the energy used in a typical UK hospital is attributable to air and space heating [5]. 'Health Technical Memorandum 07-07' calls on NHS organisations to achieve targets for delivered energy of 35e55 GJ/100 m 3 for new buildings and major refurbishments, and 55e65 GJ/100 m 3 for less intensive refurbishments of existing facilities, 1 for all building uses including space heating, hot water, lights, appliances and catering [6]. Data shows that energy use in English hospitals is often far in excess of these levels [7].
The challenge of reducing CO 2 emissions and energy demand is amplified by the health implications of a changing climate. 2 The NHS is required to provide a safe and comfortable environment for patients and visitors (more than 1 million every 36 h) and staff (1.4 million employees, 5% of the UK workforce [8]). The 2003 heatwave led to 15,000 excess deaths in Northern France, the UK heatwave of June/July 2006 is thought to have led to an increase in deaths over baseline mortality of 4%, and there were approximately 300 excess summer deaths after the 2009 heatwave between 30 June and 2 July [9]. Individuals sensitive to high temperatures are likely to be present at all times in hospitals, including those with compromised thermoregulatory systems (the elderly, the chronically and severely ill, those on certain medications that impair perspiration) as well as those who cannot take action in the face of high temperatures (small children, the bed-bound, patients with mental illnesses) [10]. However, very few existing buildings on the NHS Estate were designed to be air-conditioned; indeed, many are poorly insulated and often over-glazed, leading to increased risk of summertime overheating, even in recently completed buildings [11].
It is understandable that NHS hospital trusts may retrofit air conditioning to ensure patient safety. In fact, some commentators suggest that such a strategy would be sensible (London School of Hygiene and Tropical Medicine contribution to the EPSRC Adaptation and Resilience to Climate Change Coordination Network 'Overheating Seminar to address Policy Questions', at the Department of Communities and Local Government, 1 December 2010, attended by the author). While natural ventilation is promoted by the NHS for non-critical spaces such as wards and offices [6], there are perceived barriers: concerns about infection control, security and safety when using operable windows, and a risk-averse procurement environment [12]. There are few examples of innovative natural ventilation/passive cooling strategies being used in hospital buildings, although recent research suggests that 70% of an acute hospital could be naturally ventilated while notional propositions for such strategies have been made [10,12]. Thus although the NHS Heatwave Plan [9] advocates a 'passive approach' to combating heatwaves, including in the medium term (10e30 years outlook) a 'focus on building design of hospitals . to aid passive cooling where possible', it also suggests that the NHS should 'target vulnerable areas (patients, medications, IT) with air conditioning.' However, the take-up of air conditioning at the significant scale of the NHS Estate would clearly disadvantageously affect the national carbon reduction programme.
The DeDeRHECC project adopts the premise that by reducing the inherent overheating risk and energy demand of existing buildings through refurbishment, it is possible to reduce CO 2 emissions, achieve the energy targets, and improve resilience to climate change [13,14]. 3 Energy demand reduction also provides NHS Trusts with added protection against fuel shortages and price rises. Within the context of mandatory carbon 'budgets', carbon savings through reduced energy use can be directed to patient care [15]. The focus on a particular building type, governed by specific criteria, can be justified by a.) the significant proportion of UK emissions generated by the NHS; b.) the scale of the NHS Estate and the recurrent types that comprise it; and c.) the role of the NHS in safely providing healthcare as a public service. It is also hypothesised that a building envelope that is well insulated, with measures to control solar gain and passively driven ventilation, will make buildings more resilient in the face of power loss and other natural and man-made catastrophes; i.e. a building with greater all round resilience. In this respect, many of the issues are applicable to other highly serviced building types.
The paper quantifies the impact of each proposed refurbishment option on the energy demand and CO 2 emissions of the building. The paper briefly reviews the current status of the NHS building stock and outlines the characteristics of the case study tower block. The current performance of selected spaces is reported. The paper presents the results from modelling the current and future performance of a key space, diagnoses the outcomes and proposes various refurbishment options to increase resilience. The performance evaluation criteria and model calibration are described. Lomas & Giridharan [16] explain the modelling exercise in greater detail in a related paper.

The case study building: the Addenbrooke's Ward tower
Addenbrooke's Hospital is located to the southeast of Cambridge (Fig. 1). The main ward tower, investigated here, was built between 1967 and 1972 comprising a ten-storey slab block. An initial survey of NHS hospital sites by the authors has identified 50 buildings of this basic type with an average of seven storeys, suggesting that an assessment of the inherent resilience of this type configuration for a hospital may have considerable potential impact. The research team's intent was to provide evidence to policymakers deciding whether to decommission these buildings, invest in relatively minor and less disruptive interventions, or contemplate more substantial interventions to realise long-term resilience. Fig. 2a and b shows two floors, levels 8 and 6 of the Addenbrookes Ward Tower, comprising general and trauma wards. The figures indicate the spaces in which data have been collected. The tower is 120 m long on its SW/NE axis and variously 14.1 and 18.3 m deep. All floors have the same overall geometry consisting of a long central corridor to which rooms of 5.7 m depth are connected. On the north side, single patient rooms are found at the ends (5.7 m deep) with offices and utility space to the middle. On the south side, multi-bed wards (10.2 m deep from corridor to window wall) occupy the wider end parts of the building and have a projecting bay area; there are also some further single bedrooms. As Fig. 3 shows, there is a half-height service floor between the 8th and 9th floors (the latter designed as an isolation unit for infectious patients) and an air handling plant and tank room above floor 9. The main plant room is below in a basement. The occupied levels have a structural floor to ceiling height of 3.66 m with, as designed, a 0.90 m void above the suspended ceiling. Fig. 4, cross section through a typical floor, shows the relative proportions of the single and multi-bed spaces and also the relative height of the wards.
The windows run as a continuous ribbon at all levels on both facades incorporating opaque panels. Precast concrete panels form the spandrels. Figs. 3 and 5 show the original fenestration comprising centre pivot teak-framed windows, c. 2 m in height (Uvalue estimated as 5.6 W/m 2 K). The continuous glazing here yields a façade glazing ratio of 57%.
The commissioning clients and their designers were aware of the likelihood of unwelcome solar gains and glare arising from the southeasterly exposure because solar gain had already proved to be an issue in previous buildings on the site ([17, minute of 1 March 1963]. In the case of the ward tower, direct solar gain was to be mitigated by 'external blinds made of tygan which could be operated from inside the rooms', to be fixed directly to the teak frames [17, minute of 23 August 1965]. 'Tygan' was the brand name of a type of polyvinylidene chloride plastic, in effect horticultural open woven shading cloth [18]. It is not known whether the clear variant was proposed (which could reduce light transmission by between 21% and 36%) or the green version (by 53e55%) [18]. However, in January 1969, it was reported that 'none of the blind manufacturers would guarantee an external fabric blind under the conditions prevailing on the elevations of the ward block and as a result of experience 3  gained in their use on other tall buildings, "Tygan" blinds were no longer manufactured' [19]. The proposed defensive measure was abandoned. Records show that there were complaints of excessive heat in the completed building [20].
Whereas contemporary guidance recommended natural ventilation for wards, at Addenbrooke's it was thought that a mechanical solution was required to achieve the then desired air change rate of 2.5 ach À1 [17, minute of 11 October 1963]. It was stated that air in winter would be heated to offset the air change heat loss only. Air was delivered to wards via ducts in the suspended ceiling of the corridor; bathrooms and utility rooms only were provided with extract ventilation. The defined temperature rise was 42 F (23.3 K), thus with a 25 F (À3.9 C) outdoor temperature, an indoor temperature of 67 F (19.4 C) would be maintained indefinitely. Patent 'Frenger' radiant ceiling heating panels were installed, comprising perforated metal trays onto the top of which is clipped metal piping to take low-pressure hot water.
The designers' principal concern was to achieve comfortable temperatures in winter, rather than mitigate the summer overheating risk. Ministry of Health guidance, then just published, called for wards to be heated to 65 F (18.3 C) when the external temperature was 30 F (À1 C) [21]. The mid-1960s design team Addenbrooke's Hospital, showing major construction phases. The Masterplan delivered an attenuated east-west spine, allowing the ward tower to enjoy southerly views over the countryside. Key Phase I, 1957e1962: pavilions orientated perpendicular to the spine, a conventional hospital arrangement with its roots in the nineteenth century Phase II, 1967e1972: notably, the main ward block is parallel with the spine in defiance of the traditional hospital arrangement and the recommendations of the Nuffield study 1973e1984 Post 1984 Spine on three levels: Level 1 services/supplies/waste; level 2 public/patients/staff; level 3 clinical/surgical. was aware that the heating of the earliest buildings at Addenbrooke's had been inadequate during the exceptionally cold winter of 1962e1963 [17]; the cold winter of 1947 may have also been on their mind.

Subsequent interventions and the building today
The hospital Estates staff report that there were no major changes to the external envelope until 2000 when the eastern Fig. 2. a. Plan of Level 8 in the ward tower at Addenbrooke's Hospital. The most noticeable change to the original design comprises the removal of the enclosed day spaces in bays at the ends of the wards. . Key: Spaces being monitored; Two multi-bed wards (8MB4, 8MB10); A nurses' station (8NS). b. Plan of Level 6 in the ward tower at Addenbrooke's Hospital. The most noticeable change to the original design comprises the removal of the enclosed day spaces in bays at the ends of the wards. Key: Spaces being monitored; a seven-bed ward (6MB7); a nurses' station (6NS). half of the building was extended on levels 1e3 only (Table 1). In 2004, the windows were replaced with aluminium-framed SAPA Glostal low-emissivity double glazed windows (U z 1.9 W/m 2 K) with significantly reduced opening area (Fig. 5). The lower c.90 cm panes are top-hung and operable to provide a maximum opening of about 100 mm, restricted for patient safety, whilst the upper panes (also c. 90 cm tall) are fixed. The window-to-floor area ratio on the SE side, which of course is exposed to solar Key: 1 Single bedroom with centrally controlled mechanical supply and room user operated natural ventilation; 2 Internal storeroom: extract only with grille in doorway to draw in air from wards; 3 Corridor: no extract, warm air from wards dwells; 4 Six bed ward: centrally controlled mechanical supply and room user operated natural ventilation; 5 Concrete structure: piers enable flat slabs with no downstands; 6 Mechanical air supply; 7 Radiant ceiling: perforated metal trays with attached low pressure hot water heating pipes; 8 Main air supply duct; 9 Main services spine; 10 As built window arrangement; 11 Current window arrangement. gains, is thus about 35% in the deeper rooms and as much as 60% in the shallower spaces, although solid lightweight panels have been used in some panes to reduce the area of glass. The opaque panels below the windows remain as built and so are uninsulated (U z 2.1 W/m 2 K). The concrete spandrel above the window is insulated (U z 0.5 W/m 2 K). Even ignoring the heat bridging at the floor to outer wall junctions this yields an area-averaged Uvalue of about 1.2 W/m 2 K.
The heating and ventilation system today is little changed from the original. The radiant ceiling panels are said to operate at 30 Ce35 C, switching off at ambient temperatures over 18 C. Though ceilings in many wards have been replaced with gypsum tiles in a grid, radiant panels have been retained; fan coils have been added. Some floors have part perimeter heating. A small number of specialist rooms have their own dedicated supply and extract while local cooling has been added in certain areas. For all other spaces the central roof-located air-handling unit, which has a single inlet and supply fan, delivers air to the wards on the SE side and to treatment rooms on the NW side. The fan is twin speed but was observed to run at maximum speed over 1.5 years of data collection, 2010e2012. Insofar as the Trust Estate office did not have any information on the fan speed, current flow rate is taken as the maximum speed, i.e. 0.22 m 3 /s. The air is heated to 18 C prior to delivery to the building. There is no return air route and so all delivered air escapes from the building either through the designed toilet (and other) exhausts, through windows and doors or through other gaps in the fabric. The nurses' stations are thus vented inadequately by the air flowing back along the central spine to the extracts in the toilets and kitchens. This effectively offsets any risk of draughts from the ingress of cold air, but is, of course, extremely energy inefficient. The heat for the radiant panels and ventilation air originates from the central-site, gas-fired steam plant. The steam is delivered to plate heat exchangers in the base of the Tower that provide hot water at 80 C in winter and 70 C in summer.

Measured performance of existing wards
The internal temperatures in five spaces are currently being recorded at hourly intervals using Hobo pendant loggers. 4 The spaces are ( Fig. 2a and b): B a seven bed ward on level 6 (6MB7) B a nurses' station on level 6 (6NS) B two multi-bed wards on level 8 (8MB4, 8MB10) B a nurses' station on level 8 (8NS) Initial results from this work have been reported for a 46-day (1104 h) period, 1 July to 15 August 2010, by Lomas and Giridharan [16]. The temperatures in one ward, the 10-bed ward on Level 8 (8MB10), are illustrated in Fig. 6, alongside the Cambridge air temperature for the period and locally measured solar radiation intensity. During this period, the Level 8 ward temperatures were between 21.4 C and 28.5 C. 45% of the hours during the measurement period had internal temperatures over 25 C, which for healthy people is seen as the value above which thermal dissatisfaction will occur, with 28 C being the upper limit of thermal comfort acceptance. There were 38 night time hours (taken as 21:00 and 06:00) above 26 C (i.e. 8% of the total).
Considering all the wards during the monitoring period, the temperatures ranged from 21.2 C to 28.5 C but with no more than 6 h (0.5% of the total) above 28 C, which suggests the wards will operate for the whole year within NHS HTM03-01 [22] and CIBSE Guide A [23] guidelines (Table 2). During the night however, all three wards were warm with, depending on the ward, 4%e9% of   night time hours over 26 C, above which temperature healthy people experience poor quality sleep [23]. The recent thermal comfort standard BSEN15251 takes account of human adaptation to recent ambient temperatures and thus occupants' preference for higher temperatures inside naturally ventilated buildings during warm summer spells [24]. Thus the envelope of acceptable temperatures has upper and lower limits that increase with the running mean of the ambient temperature and are applicable to different categories of persons; Cat I being applicable to very sensitive and fragile persons with special needs and thus seems appropriate to hospital wards (Fig. 7). Spaces of 'normal occupancy' (Cat II) might include administrative offices, consulting rooms, and perhaps nurse stations, whilst Cat III might be relevant to public arrival and waiting areas etc.
During the recording period, none of the monitored wards experienced more than 26 h when the measured temperature was above the BSEN15251 Cat I limit (see e.g. Fig. 7); 55 h, i.e. 5% of the total, might be seen as unacceptable (Table 2). 5 Although window opening frequency was not recorded, site visits revealed that windows were open to their maximum limit (100 mm) during warmer weather from early in the morning until late evening. Temperature records in the larger deep-plan wards indicated a gradient from the warmer core to the cooler perimeter.

Modelling the existing building e current climate
To predict the annual frequency of overheating of the wards and the energy demands and CO 2 emissions in the current climate, the dynamic thermal model IES was used [26]. This software was chosen as it is widely used by UK building engineering consultancy firms including partners in the 'DeDeRHECC' project. Its application to the refurbishment of hospitals would thus meet with general understanding and interest. Dynamic thermal models are routinely used by practitioners and academics to predict the future performance of existing buildings, or the current and future performance of refurbished (or new) buildings. Since measured temperatures were available, it was possible to calibrate the IES model prior to embarking on the assessment of the proposed refurbishment options.
Contemporary building energy models, with their powerful graphical interfaces, enable large models of whole buildings to be created relatively easily and it is tempting to capitalise on this capability. When comparing multiple refurbishment options for large, complex multi-cellular buildings, as here, it was clear that it would be prohibitively time consuming to model the entire Tower. In any case, because big models require users to make many more assumptions, it may be debatable whether the insights relating to overhearing risk that might be gained would be any more reliable, not least because calibration is only possible for the few spaces for which temperature data is likely to be available. In this respect, the work is intended to demonstrate a methodology that NHS Trusts could adopt for their own Estates.
The work concentrated on creating a well-calibrated model for a space with reliable measured temperatures. The focus is the southeast-facing seven-bed ward, 6MB7 (Fig. 2b), which is typical of the wards of similar orientation in the Tower building. The choice of a southeast-facing room, without shading from surrounding buildings, was driven by the central interest of this project, summertime overheating risk. The modelled ward had a floor area of 70 m 2 , a volume of 190 m 3 , and there were two double glazed window systems with four and six components (Fig. 5). The perforated metal radiant ceiling warmed the space which was mechanically ventilated. Details of the model and the sources of the data are in the Appendix and the environmental control strategy is also included in Table 3.
Although many of the parameters fed into the model were known from the plans and sections and through site visits, some were uncertain. Sensitivity analysis revealed that the summertime temperature predictions were especially sensitive to the uncertainties in the mechanical ventilation rate, the window opening strategy and the inter-zonal flow between the room and the void above the suspended ceiling. Temperature predictions were however less sensitive to uncertainty in the background infiltration rate and the window and wall U-values. As is normal when modelling, many thermal factors were ignored, such as heat bridging and inter-room heat exchange (surrounding spaces were assumed to be the same as the modelled space and doors closed). The adjustments to the three most sensitive input parameters is, therefore, also compensating for the uncertainty in many other features.
It proved rather difficult to devise values for the three parameters that simultaneously produced good predictions for exceedences of the HTM03 criterion and both the lower and upper . Measured temperature, ambient temperature and its running mean for a seven-bed room on Level 6 (6MB7). 5 BSEN15251 is applicable to temperatures monitored during occupancy and, strictly, these should be operative temperatures. BSEN15251 Cat I boundaries (see Fig. 7). Therefore, priority was given to predicting reasonable HTM03 exceedences. The final model predicted maximum, mean and minimum temperatures between 1st July and 15th August of 28.6, 23.9 and 20.9 C respectively, with 12 h above the HTM03 threshold of 28 C. The corresponding measured values were 28.4, 25.0 and 21.0 C and there were actually 6 measured hours above 28 C (see Appendix A). Part of the difference is due to the small difference in the weather between Cambridge and Bedford from where the data used in the simulations was obtained. 6 Hourly differences between measurements and predictions are largely due to the impossibility of replicating the pattern of the actual manual operation of windows. Spot checks were undertaken during hot spells.

Predicted performance of existing building e current climate
The model devised was used in conjunction with weather data for 2010 taken at Bedford [16] to predict the annual overheating risk and the annual energy demands for the existing ward. For the period MayeSeptember 2010, the model predicted that there would be 16 h for which the internal air temperature exceeded 28 C (Table 4) and 789 annual hours over the BSEN15251 Cat I upper temperature threshold, which, assuming occupancy for all hours in the year, exceeds the 5% acceptable threshold (i.e. 438 h). Although the ward space temperatures exceeded the BSEN15251 Cat I upper threshold, these results are better than might be expected for a building of this era. However, this robustness is an accidental by-product of the building's air leakiness and considerable mass, and comes with an energy penalty.
The predicted environmental energy demand for Cambridge in 2010 was 101GJ/100 m 3 , with the bulk of the energy being used to heat the fresh air (46%), to warm spaces via the ceiling (38%), and for driving the fan that delivers the air (14%). This substantially exceeds the NHS gross target for refurbishments of 55e65 GJ/ 100 m 3 ( Table 2). The uncontrolled loss of heat by leakage of air from the building leads to this high energy demand.
Concerning CO 2 emissions, CIBSE provides in Technical Memorandum TM46 benchmarks that are used for determining the operational rating of buildings [26]. 7 The whether adjusted benchmark for 'Hospitals; clinical and research' in the Cambridge (East Anglia) region is 142.4 kgCO 2 /m 2 ( Table 2). The predicted energy demand equates to 179 kgCO 2 /m 2 , which exceeds the TM46 benchmark by some 26%.
It is worth noting that the NHS and CIBSE benchmarks include all energy use, for space heating and ventilation, and also for hot water, medical equipment, small power, pumps, controls, lifts etc, which on average, across all NHS buildings, are responsible for approximately 44% of the emissions. Thus, in spaces like wards, which use energy for heat, ventilation, lighting and small power only, the energy and CO 2 emissions ought to be around half of the benchmark values if, overall, NHS Trusts are to meet their targets.
It is evident that, whilst the wards in the Addenbrooke's Tower provide satisfactory comfort levels, this is achieved at the cost of excessive energy demands. The researchers investigated whether refurbishment could curb the heat lost through uncontrolled leakage of ventilation air from the building, whilst also reducing There is a particular premium on rapidly executed, light-touch interventions in NHS buildings, the product of economic circumstances as well as a justifiable focus on improving the patient experience and infection control. 9 Table 1 records that the majority of interventions in the Addenbrooke's Tower have been largely cosmetic. NHS Trusts' decision-making with respect to their estates is driven by capacity control and so the long or medium term loss of a significant portion of the available space is disruptive (Jan Filochowski, West Hertfordshire Hospitals NHS Trust to authors, Project Sounding Panel meeting, March 2011). There are patient safety issues in the large-scale temporary dislocation of clinical activities (Patricia Young, National Patient Safety Agency, to authors, 2010) while project Partner NHS Trust Chief Executives have commented on the harmful effects on patients of dust arising from construction work, especially the difficulties in cocooning neighbouring spaces, denying them natural ventilation and potentially exacerbating summertime overheating risk (Filochowski to authors).
Five refurbishment options were considered by the authors, ranging from light-touch options to more invasive measures that implement innovative passive strategies. Each option exploited the inherent characteristics, mechanical installations and capacity in the existing building. The strategies sought also to improve the internal environment in the wards and other spaces, 10 and in this regard the features that are considered in the BREEAM Healthcare 2008 scheme [27] are pertinent, not least because Trusts are increasingly interested in the BREEAM rating that their buildings might achieve. The scheme gives the Energy section the highest weighting of all the nine BREEAM sections (19%) 11 but the second highest weighting is given for the 'Health and Wellbeing' section (15%). 12 This includes provision of: daylight, adequate view out; glare control; natural ventilation; thermal zoning; and thermal comfort.
Each of the options is described and illustrated (Figs. 8e12) with the key features and environmental control strategies (heat loss and solar control, ventilation, heating and cooling) being summarized in Table 3. For each option the predicted annual energy demands and CO 2 emissions of one ward, the refurbished multibed ward 6MB7, were predicted using IES and the Bedford 2010 weather [28] and compared with the NHS and TM46 benchmarks (Figs. 13 and 14). The internal temperatures predicted for the summer period MayeSeptember (153 days, 3672 h) were assessed using the HTM03-01 and CIBSE overheating criteria (virtually all annual overheating hours will occur during the MayeSeptember period). For the three refurbishment options incorporating operable windows, the BSEN15251 approach was also used to assess the internal temperatures; the two refurbishment options, in which the building is sealed, mechanically ventilated, heated and cooled, were not assessed using the BSEN15251 approach. By way of context there were just 4 h in Cambridge in 2010 when the ambient temperature exceeded 28 C [16].
It is important to note that in all the simulations, the ventilation, heating and cooling control strategies (Table 3) were devised with the provision of overall summertime thermal comfort (as defined by the guidance) as the priority. Less attention was given to the Fig. 7. Thermal model wireline of simulated ward room. 8 Lomas and Giridharan [16] examined the potential for patient operated internal ceiling fans to improve summertime temperature conditions, but this measure would not impact on the energy use and CO 2 emissions. 9 A parallel study of recent NHS refurbishment and construction projects forms part of the 'DeDeRHECC' project; the authors are working with researchers from the Engineering Design Centre at the University of Cambridge and the Design Group at the Open University to assemble detailed case histories. 10 For example [16], shows that the nurses' stations in the core of the Tower substantially overheat. 11 The lower the predicted emissions the greater the number of BREEAM credits that are awarded and additional credits can be earned by meeting the demand from renewable energy sources (BREEAM Issue Ene5). Further, a high credit score for 'Reduction of CO 2 Emissions' (Issue Ene 1) is mandatory if an Excellent or Outstanding rating is to be achieved. Innovation credits can be earned through exemplary levels of CO 2 reduction and daylight provision. 12 To elaborate the BREEAM issues: daylight -daylight factor, DF > 3% over 80% of floor space of wards (Issue Hea 1); provision of adequate view out e most easily fulfilled by wards less than 7 m deep (Hea 2); glare control e especially through occupant-controlled blinds (Hea 3); provision of natural ventilation e operable window area to 5% of floor area or, in deep-plan rooms, provision of cross ventilation (Hea 7); and Thermal Comfort e meeting the HTM03-01 overheating criterion (Hea 10); and Thermal Zoning e to enable occupant control of heating and cooling systems (Hea 11). surrounding five spaces all assumed to be at same temperature as modelled space except in Option 5 in which the space below, the void, was exposed to external conditions. In all options the internal doors closed. Vertical elevations have 100 mm insulation and roof 300 mm insulation k ¼ 0.025 W/m K; window area, 12.2 m 2 , i.e.17% window-to-floor area ratio. Volumes vary -with existing ceiling at 2.7 m height, 190 m 3 ; with exposed concrete ceiling at 3.9 m height, c270 m 3 ; and with part exposed ceiling height 3.3 m on average, Option 3, 227 m 3 . b All mechanically ventilated strategies have heat recovery at 60% efficiency, with summer by-pass function. c Option1 e existing windows with additional single pane protecting interstitial blinds in ventilated cavity, U ¼ impact of the chosen control strategies on the night time temperatures or the annual energy demand and CO 2 emissions. Thus there were times when the night time temperatures dropped below the BSEN15251 Cat I and Cat II envelope. By micro-adjustment of the heating set-point it was possible to avoid this, however, such refinements were not generally pursued in this work. 13 Similarly, changes to the control strategy might reduce the energy demands and CO 2 emissions slightly and in practice facilities managers manually adjust control settings in response to weather changes and other events, but in a thermal simulation model this is not easily done. Thus in interpreting the refurbishment options' performance, the night time temperatures are afforded little weight and the energy demand figures may not be optimally low. Nevertheless the results give a valid strategic impression of the relative performance of the existing building and refurbishment options. Option 1: SMVHC: Sealed building, Mechanical Ventilation with Heating and Cooling, Fig. 8.
The application of 'PassivHaus' principles to a substantial public building. The external envelope is overclad to improve substantially air-tightness and U-values. The relatively recent high-performing thermal break aluminium-framed double glazing is retained. The ventilation system operates at 6 ach À1 , as required by HTM03-01, windows are sealed and a third layer of glazing added internally. The glazing is shielded by new interstitial blinds. (This arrangement, which contrasts with the usual location of the interstitial blind in a vented cavity on the external side of the double glazing, was adopted in an attempt to retain the existing windows and thus to minimize disruption). The radiant ceilings are retained but are not active; perimeter heating is not employed in this first option. Resilience to overheating is provided by increased mechanical ventilation with some mechanical cooling for peak lopping when required.
The peak lopping strategy ensured that there were few summertime hours above each of the HTM03-01 and CIBSE thresholds ( Table 4). The annual predicted energy demands and emissions were 59 GJ/100 m 3 and 137 kgCO 2 /m 2 respectively (Figs. 13 and 14), which is within the NHS target for refurbished buildings (but of course excludes delivered energy, which would add to this figure). Very little energy was used for cooling but the ventilation energy demand to deliver the high airflows was significant. Overall the design achieved substantial improvement over the performance of the existing building.
The envelope is thermally upgraded as option 1. The 2004 double glazing is retained but solar shading is provided by fixed external shades tailored to shield all southerly glazing through the potential overheating season. The mechanical supply delivers the fresh air requirement only, i.e. below 6 ach À1 , and heat recovery is provided. The original radiant ceiling installation is refurbished and employed in winter to provide additional radiant heating (boosting the ducted warm air supply as required). In summer, cooler water (at 22 C) is pumped through the system. Almost all the insulation originally provided to the soffit above the radiant ceilings is presently missing; in this scenario it is essential to replace it.
The simulations indicated an insignificant number of summertime hours above each of the HTM03 and CIBSE thresholds ( Table 4). The annual predicted energy demands and emissions were 46 GJ/ 100 m 3 and 102 kgCO 2 /m 2 respectively (Figs. 13 and 14). This represents a further improvement: reduced energy demand, especially for the fans, which are delivering 4 ach À1 as in the existing building; reduced heating energy demand, as the heat is delivered directly to the space; and yet the space offers the potential for individual control of heating and cooling via the radiant ceiling.
Option 3: NVMVPH, a hybrid option, with Natural Ventilation and concurrent Mechanical Ventilation supply, heat recovery, opening windows and Perimeter Heating (Fig. 10).
Option 3 adopts similar thermal upgrading to the envelope, retains current double glazing units, and enables all of the glazing to be opened by occupants in peak summer periods. A high degree of occupant control contributes to the overheating defence strategy of this option. In addition, the suspended ceilings are cut back as far as the supply ductwork allows, in order to expose the concrete slabs inboard from the window walls. Perimeter heating is introduced. In some ways the services strategy is similar to that in the existing building, but with perimeter heating rather than a radiant ceiling, and much improved fabric, shading and also heat recovery.
There were no predicted summertime hours above the CIBSE, HTM03-01 or indeed the BSEN15215 thresholds ( Table 4). The annual predicted energy demands and emissions were 40 GJ/100 m 3 and 111 kgCO 2 /m 2 respectively (Figs. 13 and 14), with ventilation  (Fig.11). This option dispenses with the mechanical ventilation system. It upgrades the external envelope to the standard of Options 2 and 3 and provides perimeter heating with actuated trickle vents below fully operable occupant-controlled windows, adapting the existing installation. Furthermore, each floor can be cross-ventilated by threading crossover ducts in alternating directions across the width of the floor plate. The option envisages active use of night ventilation and removes all suspended ceilings to expose the full flat concrete soffit above all patients. The aim is that vigorous night ventilation will cool the soffits.
There were no predicted summertime hours above each of the thresholds (Table 4) and the annual predicted energy demands and emissions were just 20 GJ/100 m 3 and 44 kgCO 2 /m 2 respectively; the omission of fans being the key to such low energy demand.
The principle of natural ventilation explored in Option 4 is reinforced with stacks to develop greater pressure differences and hence flows as required. Stack ventilation of high rise buildings has been the subject of some speculative work [29] which introduced the idea of segmenting a building into bundles of floors that could be more easily served by shorter stacks. A potential difficulty is the effectiveness of stub stacks on the windward face in which the flows may reverse with a reversing flow regime set up on each floor. In a hospital where the avoidance of the risk of airborne infection spread is clearly important, reversing flow regimes are unacceptable. This option removes the envelope of a floor at mid-height to provide a free air environment in which the stacks to the lower four floors can terminate. The stacks are strictly dedicated to one space per cell, as the part elevation/section reveals. The occupation of the elevation by deep stacks reduces the glazed area beneficially and the depth also beneficially shades the glazing through the critical summer overheating period. All windows are rendered operable and the cross-ducts of Option 4 are introduced below a fully exposed flat concrete soffit. The intention would be to nightventilate vigorously within the comfort parameters of sleeping patients and patrolling medical staff.
For this final and more elaborate option, there were no predicted summertime hours above each of the comfort thresholds (Table 4) and the annual predicted energy demands and emissions were just 32 GJ/100 m 3 and 67 kgCO 2 /m 2 respectively (Figs. 13 and 14). This option was modelled with not only the façade exposed to external conditions but also the floor (below which is the void, i.e. the open area intended to provide greater exposure to the mid-height stack terminations). Thus the heat loss is greater than for Option 4, which has only the façade exposed to external conditions (Fig. 12). The additional energy, compared to Option 4, may be avoided by more refinement to the ventilation control strategy. The other floors would have a better performance.
In summary therefore, considering the performance of all the options, it would seem possible for all five of them to yield internal thermal comfort conditions that are an improvement on the existing building. Furthermore, with appropriate control of the mechanical systems, BMS controlled openings, manually operated windows, perimeter heating, or radiant ceilings, improved occupant control is possible. The energy demands and CO 2 emissions of all the options are lower than for the existing building but the ability to achieve this performance over the entire building implies reliable room-by-room temperature control, either occupant or BMS controlled, which is something that is entirely absent in the current building.
It is nonetheless important to note that these predictions do not include energy use for matters unconnected with space conditioning (small power, medical equipment, restaurants etc), which can be 44% of the total. Given this, it is likely that only Options 2e5 could plausibly meet the NHS energy target and CO 2 emissions benchmark of 55e65 GJ/100 m 3 . Importantly this excludes the option which has a mechanical supply of 6 ach À1 as stated in HTM03-01. We observe therefore that unless highly efficient heat recovery with a summer by-pass is used, the HTM03-01 mechanical ventilation requirement and NHS energy demand targets may be mutually incompatible; efficient fans are essential.

Modelling the refurbishment options e future climate
To assess the performance of the wards under the Cambridge climate of the future typical and extreme weather, as represented by Test Reference Years (TRYs) and Design Summer Years (DSYs) 14 were created for the 2030's the 2050's and the 2080's. Also, to provide a compatible benchmark against which to compare these predictions data was generated in a similar way for the current climate of the 2000's. The procedure for creating the weather data has been described in detail elsewhere [16] so is described in brief here.
The TRYs and DSYs for the current climate were created by the Prometheus project research team at Exeter University from the 25 years of hourly data that was available from the Bedford station (1980e2004) using the standard CIBSE method as described in Levermore and Parkinson [30]. The 2005TRY is built by chaining the most average January to the most average February etc and then smoothing the joins. The 2005DSY is simply the third hottest of the 20 years based on the average drybulb temperature form April and September; for the Bedford data set this was 1997. The 2005TRY has a broadly similar number of hours above 25 C and 28 C to the 2010 year; a similarity that extended across all hours from 20 C to 28 C. Thus the year of monitoring would seem, quite fortuitously, to be rather typical of the current climatic conditions in Cambridge. Fig. 10. Option 3: NVMVPH, natural and mechanical ventilation, perimeter heating. Key: 1 Insulation on roof; 2 Air handling unit; 3 Supply duct; air supplied to wards via ceiling, fresh air only; 4 100 mm insulation to external elevations; 5 External shading; 6 Opening windows; 7 Perimeter heating with low-level air supply via windows; 8 Ceiling exposed save for ventilation duct; 9 Mechanical extract via bathrooms; 10 100 mm external insulation; 11 Opening windows; 12 External shading.
The weather for the future years was also created by the Exeter team by the method described by Eames et al. [28] using the UKCP09 weather generator [31] assuming the global A1B emissions scenario as described in the IPCC Special report on Emissions Scenarios (i.e. a globally technologically advanced world in which energy production includes a broad portfolio of fossil-fuel and nonfossil-fuel sources) [32]. 15 Annual hourly TRY and DSY weather files were created for the 2030s, 2050's and 2080's for the 5 km grid square covering Cambridge.
The temperatures in ward 6MB7 were predicted using IES and compared using the CIBSE, HTM03 and BSEN15251 overheating criteria; the existing building and all five refurbishment options were modelled with all eight weather years ( Table 5). The BSEN15251 assessment method is particularly relevant in the context of climate change, as it takes account of occupants' adaption and thus preference for warmer conditions in naturally conditioned spaces, as the external environment warms. Indeed, one must question the Fig. 11. Option 4: CVPH, natural cross ventilation, perimeter heating. Key:1 Insulation on roof; 2 Air handling unit for bathroom extract only; 3 Suspended ceiling in corridor only; 4 100 mm external insulation; 5 High level air outlet route and grille to achieve cross ventilation (shown alternately on each floor); 6 Opening windows; 7 Low-level air supply via opening window behind perimeter heating; 8 Ceiling slab exposed; 9 Extract via bathrooms (not shown); 10 Opening windows; 11 100 mm external render; 12 Grilles for air outlet; 13 External shading. 15 This is the emissions trajectory the world is currently following. relevance of comfort standards intended for free-running buildings that are based on fixed temperature criteria when comfort in future warmer weather conditions is being assessed.

Future performance of the existing and refurbished building
The predicted future performance of the existing buildings and the refurbishment options is shown in Table 5. It has been noted [16] that the existing building would be deemed too hot by 2030s, as indicated by the HTM03-01 criterion, especially in an extreme year. However, using the BSEN15251 method the overheating in 2030s appears less serious in both typical and extreme years, especially in the spaces intended for normal, non-clinical, occupancy (as judged by the hours above BSEN15251 Cat II limits). 16 Beyond the 2030s the existing ward displays a marked increase in the frequency of overheating in extreme years.
The two refurbishment options in which the building is sealed and mechanically cooled (Options 1 and 2, SMVHC and SMVRHC) remained comfortable in both typical and extreme years to 2080s, in Fig. 12. Option 5: SVPH, natural stack ventilation, perimeter heating Option 5: 1 H-pot stack termination, robust to wind turbulence; 2 Insulated perimeter stacks external to existing elevation; 3 BMS controlled damper at entrance to stack, manual over-ride; 4 Level 6 envelope removed to create free space connecting north and south sides; 5 Lower bank of five floors connected to stacks; 6 All glazing openings under user control; 7 Perimeter heating connected to actuated vents; 8 Terminations connect to form full cross H-pot arrays, dividers guard against cross flows and risk of airborne cross infection; 9 Each stack comprises four discrete cells isolated from point of entry to cross pot; 10 Actuated points of entry into stacks; 11 Void through tower; 12 Lower ranges of stacks terminate in plane of void to enable 360 access to stack termination; 13 Glazing reduced in size and shaded within thickness of superimposed 'double façade'; 14 Highlighted floor modelled, slightly a typical as above void. U-value for floor 0.07 W/m 2 K. 16 The similar performance in the TMY and DSY years is because the BSEN15251 method accounts for occupants' adaptation to the warmer conditions of the hotter year.
that the internal temperatures remained below 28 C for all but a few hours. 17 Option 2, which has the external solar shading, rather than the internal inter-pane shading, and local cooling provision via the radiant ceiling, performs better than Option 1. The external shading improves resilience at little maintenance overhead by reducing the reliance on the mechanical cooling systems. Thus comfortable conditions are maintained through to 2080s but with a much lower energy penalty. The cooling strategy offered by incorporating large areas of radiant ceiling panels invites the use of low-carbon cooling systems, such as ground (or air) source heat pumps.
In typical years, the hybrid Option 3 (NVMVPH) and naturally ventilated Options 4 and 5 (CVPH and SVPH) remain comfortable until the 2080s. The improved façade insulation and solar shading, with a contribution from the internally exposed thermal mass, enable this performance. These results concur with those reported earlier for stack ventilated buildings outside the London environs [10]. All three refurbishment options produced spaces that have fewer hours with air temperature over 28 C than there are ambient hours over 28 C in the 2030s, 2050s and 2080s (see footnote to Table 5).
In extreme years, using the BSEN15251 method for appraisal, the hybrid Option 3 (NVMVPH), clinical spaces (Cat I) are comfortable until the 2050s, but non-clinical spaces remain comfortable even in the 2080s. Clinical spaces become unacceptably warm by the 2080s. The presence of a mechanical ventilation system with a return air route enables retrofitting as the climate warms, for example by the addition of a cooling coil in the air handling units (AHUs). In fact, by the 2080s, AHUs will have been through several refurbishment and replacement cycles. This option is only viable from an energy demand perspective if the airflow rates are as low as possible and very efficient heat recovery system and fans are used. The design outperforms the existing hybrid building because of the substantially improved fabric and external shading.
In extreme years, the natural ventilation options, either cross ventilation (CVPH) or stack ventilation (SVPH) performed well, as judged by the BSEN15251 method, in the 2030s. However, by the 2050s performance is unacceptable in Cat I spaces and by the 2080s also in Cat II spaces. In these extreme years the number of hours internal temperatures exceeded 28 C was greater than exceeded externally, suggesting that passive night time cooling is less effective in extreme years. The stack-vented option performed marginally better in environmental terms than the cross-vented option, though there is an energy penalty on the floor with the exposed slab (as is illustrated in the bar chart, Fig. 13). Table 6 sketches out the likely operational implications of the options for the hospital campuses, assembled in consultation with staff from Partner NHS Trust estates management teams. There is a clear premium on avoiding the wholesale closure of patientoccupied space to achieve these ends.
In summary it would seem that, even in extreme years, all five refurbishment proposals are resilient to climate change up until the 2030s. Thereafter, designs that incorporate mechanical cooling, probably in association with mechanical ventilation, remain comfortable in extreme years. Refurbishments offering the potential for the future addition of cooling would seem prudent. They would reduce current energy demands, CO 2 emissions and initial cost, whilst providing a degree of future-proofing.
However, caution is required in the interpretation of the predicted performance. The configuration of the wards may change, the lighting systems, heating arrangements, density of occupation and installed and portable equipment loads are likely to change through the period considered. Table 1 gives a detailed account of planned interventions since completion, none for 17 years, refurbishment of several wards at 28e30 years, largely cosmetic, and a sequence of more fundamental interventions in years 37e40. Addenbrooke's Estates staff advise that there is a planned maintenance/refurbishment cycle of 15 years for wards and so two cycles may be anticipated by 2050. The control strategies for Options 4 and 5 would ideally be 'remembered and retained' within subsequent work; the required 'as built' manuals may help preserve the principles but after the second cycle the memory can be expected to lapse with a subsequent effect on performance and patient discomfort, which could provoke intervention. Changes in lighting systems and medical technology would influence internal heat gains. The pressure is towards ever more energy efficient equipment but a significant increase in the amount of equipment, likely as hospitals become ever more specialised, may have an impact. Predicted internal temperatures are most heavily influenced by the climate, the shading, the assumed ventilation strategy and the heating and cooling technique and for each refurbishment strategy these have been fully defined ( Table 3). Predictions of future energy demands and CO 2 emissions are avoided because the primary thrust of this paper is overheating assessment, and because these will be heavily influenced by the efficiency of the building and hospital campus plant and changes in the carbon intensity of the fuels used; especially electricity.
In response to short term changes in the weather or the longer term drift of the climate, some adjustments to the control regimes would certainly be made by occupants and more importantly by the facilities managers. Rather than speculate about these possibilities, the control regimen for each refurbishment option was held constant for all the years studied. Thus the predictions show the relative    The HTM03 threshold is based on air temperature and rest are based on dry resultant temperature. The grey shows where limiting criteria are exceeded. The darker grey indicates where the exceedence is deemed important in that it could not be easily corrected by refining the control strategy. It is assumed that during the winter half of the year (October to April) the space will not overheat due to elevated ambient temperatures and solar gains. performance of the different options and give a very good indication of their inherent resilience untainted by unsynchronised tuning of the model.

Conclusions
This paper has reported on the current performance, refurbishment potential and modelled future performance of a representative 1960s UK hospital ward tower. The original designers of the Addenbrooke's building were not oblivious to environmental concerns and considered the use of shading on the glazed southfacing elevation, but this measure was omitted.
The existing wards in the Addenbrooke's Tower provide satisfactory comfort levels in the climate of Cambridge c. 2010 but this is achieved at the cost of excessive energy demands and CO 2 emissions, largely due to loss of heat through uncontrolled air leakage. The building remains resilient until the 2030s but becomes problematic thereafter. The current performance of the building is the result of its control strategy (which lacks exhaust) and the extent of the glazing. The results confirm that BSEN15251, which takes human adaptation to ambient temperature into account and provides comfort criteria appropriate to both clinical and non-clinical spaces gives a more favourable reflection of the resilience of spaces with natural ventilation and operable windows, than assessment methods such as HTM03-01 that use a fixed criterion. Fixed comfort limits, such as HTM03-01 look increasingly inappropriate in times of weather that is more variable and that will gradually warm over time.
Refurbishment, in particular to curb the heat lost through uncontrolled leakage of ventilation air from the building, could reduce energy demands and maintain or even improve the internal environment in the current climate. Space by space control and monitoring potential to enable optimised control is essential. The five refurbishment options also offer the opportunity to increase the resilience of the building in a warming climate whilst saving energy. Relatively modest interventions could achieve substantial savings whilst achieving future resilience in at least typical years. Beyond the 2030s, a degree of mechanical intervention will be required in extreme years and provision should be made for its eventual installation. Current work by the authors is exploring the 'whole life' financial costs of the options. Unless highly efficient heat recovery with a summer by-pass is used, with efficient fans, the HTM03-01 mechanical ventilation requirement of 6 ach À1 and NHS energy demand targets of 55e65 GJ/100 m 3 may be mutually incompatible, suggesting that further research in this area could be useful to the Department of Health.
Although focussed on a hospital building, the adaptive strategies explored in this paper could be applied to other non-domestic building types in temperate climates, for example office towers, in that they offer options for refurbishment incorporating the potential for upgrading in response to climate. The case study suggests that wholesale air conditioning is not necessary now, nor in the immediate future, and is indeed undesirable in financial as well as in energy use terms. model.
The ward was modelled using the IES software [25]. For the current building and the five refurbishment options ward 6MB7 was simulated. Its energy use and internal temperatures was taken as being indicative of the relative performance of the current building and the various refurbishment options. The model was calibrated as outlined in Section 4 and as described more fully by Lomas and Giridharan [16].
This Appendix describes the model of the current Addenbrooke's Tower building. The variations to simulate each refurbishment option are given in Table 2 and its footnotes. The model is illustrated in Fig. A1 and the key parameters listed in Table A1. The architectural elevations and plans, Figs. 2b, 3 and 5, give further insight.