Evaluating methods for estimating whole house air infiltration rates in summer: implications for overheating and indoor air quality

Purpose – Accurate values for infiltration rate are important to reliably estimate heat losses from buildings. Infiltration rate is rarely measured directly, and instead is usually estimated using algorithms or data from fan pressurisation tests. However, there is growing evidence that the commonly used methods for estimating infiltration rate are inaccurate in UK dwellings. Furthermore, most prior research was conducted during the winter season or relies on single measurements in each dwelling. Infiltration rates also affect the likelihood and severity of summertime overheating. The purpose of this work is to measure infiltration rates in summer, to compare this to different infiltration estimation methods, and to quantify the differences. Design/methodology/approach – Fifteen whole house tracer gas tests were undertaken in the same test house during spring and summer to measure the whole building infiltration rate. Eleven infiltration estimation methods were used to predict infiltration rate, and these were compared to the measured values. Most, but not all, infiltration estimation methods relied on data from fan pressurisation (blower door) tests. A further four tracer gas tests were also done with trickle vents open to allow for comment on indoor air quality, but not compared to infiltration estimation methods. Findings – The eleven estimation methods predicted infiltration rates between 64 and 208% higher than measured. TheASHRAEEnhanced derived infiltration rate (0.41 ach) was closest to themeasured value of 0.25 ach, but still significantly different. The infiltration rate predicted by the “divide-by-20” rule of thumb, which is commonly used in the UK, was second furthest from the measured value at 0.73 ach. Indoor air quality is likely to be unsatisfactory in summer when windows are closed, even if trickle vents are open. Practical implications – The findings have implications for those using dynamic thermal modelling to predict summertime overheating who, in the absence of a directly measured value for infiltration rate (i.e. by tracer gas), currently commonly use infiltration estimation methods such as the “divide-by-20” rule. Therefore, infiltration may be overestimated resulting in overheating risk and indoor air quality being incorrectly predicted. Originality/value – Direct measurement of air infiltration rate is rare, especially multiple tests in a single home. Past measurements have invariably focused on the winter heating season. This work is original in that the tracer gas technique used to measure infiltration rate many times in a single dwelling during the summer. Estimating infiltration rates in summer © Ben M. Roberts, David Allinson and Kevin J. Lomas. Published by Emerald Publishing Limited. This article is published under the Creative Commons Attribution (CC BY 4.0) licence. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and noncommercial purposes), subject to full attribution to the original publication and authors. The full terms of this licence may be seen at http://creativecommons.org/licences/by/4.0/legalcode. The authors wish to thank technical staff in the School of Architecture, Building and Civil Engineering at Loughborough University: Mark Harrod for ensuring health and safety requirements were met during tracer gas tests; Mick Barker for the delivery of numerous canisters of CO2 to the test houses and ensuring the regulator was connected safely; and Dean Sanham who maintained the School weather station. The authors acknowledge Loughborough University’s continued provision of 24-h security and ongoing maintenance of the test houses. Funding: This work was funded by the London-Loughborough Centre for Doctoral Research in Energy Demand (grant EP/L01517X/1). The current issue and full text archive of this journal is available on Emerald Insight at: https://www.emerald.com/insight/2398-4708.htm Received 6 June 2021 Revised 26 July 2021 Accepted 1 September 2021 International Journal of Building Pathology and Adaptation Emerald Publishing Limited 2398-4708 DOI 10.1108/IJBPA-06-2021-0085 Thiswork is also original in that it quantifies both the infiltration rate and its variability, and compares these to values produced by eleven infiltration estimation methods.

The prior research tends to focus on wintertime infiltration due to legitimate concerns about wintertime heating demand (Keig et al., 2016;Johnston and Stafford, 2017;Johnston and Miles-Shenton, 2018;Kisilewicz et al., 2019;Vega Pasos et al. 2019Mun et al., 2021). Yet, infiltration rates in summer cannot be ignored. Summertime overheating is an increasing problem in the UK and elsewhere, partly due to climate change, but also because homes are becoming more airtight  and infiltration is a modifier of indoor temperatures Taylor et al., 2018;Roberts et al., 2019a).
Similarly, concerns about indoor air quality have historically been greatest in winter, as it is assumed windows are more likely to be closed in winter and open summer. It is accepted that indoor and outdoor temperatures are one of the main drivers of window opening in UK Reference Topic Method used Palin et al. (1996) Mechanical ventilation BREDEM reduction Kalamees (2007) Measurement of air tightness Divide-by-20 Hacker et al. (2008) Emissions modelling Minimum IAQ standard Banfill et al. (2011) Mechanical ventilation Divide-by-20 a Porritt et al. (2011) Overheating modelling BREDEM reference Banfill et al. (2012) Retrofit modelling Divide-by-20 Mavrogianni et al. (2012) Overheating modelling Dwelling age b Oikonomou et al. (2012) Overheating modelling Dwelling age b Porritt et al. (2012) Overheating modelling BREDEM reference Mavrogianni et al. (2014) Overheating modelling Dwelling age b Taylor et al. (2014b) Overheating modelling SAP reduction Taylor et al. (2014a) Overheating modelling Dwelling age b Beizaee et al. (2015) Heating controls Divide-by-20 Taylor et al. (2015) Overheating modelling SAP reduction Hong and Kim (2016) Infiltration rates in high rise buildings Divide-by-20 Jack et al. (2016) Ventilation heat loss Divide-by-20 Sinnott (2016) Ventilation heat loss ASHRAE basic Taylor et al. (2016) Overheating modelling SAP reduction Symonds et al. (2016) Overheating modelling SAP reduction Jack et al. (2018) Co-heating test validation Divide-by-20 c Simson et al. (2017) Overheating modelling Estonian building regulations Symonds et al. (2017) Overheating modelling SAP reduction Echarri-Iribarren et al. (2019) Passivhaus evaluation Sherman simplified Crawley et al. (2019) Airtightness SAP reduction Li et al. (2019b) Overheating modelling CIBSE Guide B d Li et al. (2019a) In-use HTC estimation SAP reduction MHCLG (2019) Overheating modelling CIBSE Guide A e Parker et al. (2019) Overheating modelling SAP reduction Roberts et al. (2019a) Overheating modelling Divide-by-20 f Rodrigues et al. (2020) Thermal comfort modelling Sherman simplified g Note(s): Excludes studies which use the methods for comparison/validation purposes a Acknowledges that "divide-by-20" is not perfect, but is advantageous due to its simplicity b Dwelling age was established using guidance from Stephen (2000), but it is not apparent how infiltration was derived c Notes that only three of the seven participants used this method. Other participants used tracer gas. Acknowledges that "divide-by-20" was developed in the USA and may not be appropriate for UK homes d CIBSE (2016) e CIBSE (2007) f All four of the study participants used this method g Derived from Pulse test data, not fan pressurisation Table 1. Studies which use air infiltration estimation methods homes (Fox, 2008;Sharpe et al., 2015;Jones et al., 2017;Mavrogianni et al., 2017). However, recent guidance from the UK Government to occupants on how to reduce summertime overheating states that window openings should be reduced during very hot weather to prevent ingress of warmer outdoor air (MHCLG, 2021). This action could have implications for indoor air quality if background ventilation is insufficient. It is known that the transmission of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus, which causes coronavirus disease 2019 , occurs more readily in poorly ventilated indoor environments (Cevik et al., 2020;Meyerowitz et al., 2020;Greenhalgh et al., 2021;Jones et al., 2021;Qian et al., 2021). So reliable estimates of the infiltration rates during a summer heatwave when occupants are encouraged to close their windows is especially important.
Infiltration is expected to be lower in summer than winter because the variable drivers of infiltration, i.e. the wind speed and indoor-outdoor temperature difference (Labat et al., 2013), will also be lower (Binamu and Lindberg, 2002;Hong and Kim, 2016;Kisilewicz et al., 2019). Infiltration may also be lower in summer due to seasonal variation in building envelope air permeability which can be 20-40% lower in summer due to building materials expanding and contracting with seasonal changes in humidity (Warren and Webb, 1980b;Persily, 1983;Kim and Shaw, 1986;Bassett, 1992;Bracke et al., 2016). However, the estimates of infiltration rate may be least accurate in the summer season, with one relevant study showing that the difference between measured and estimated infiltration rates (Sherman Simplified model) were greater in summer and smaller in the winter (Kisilewicz et al., 2019). Because of growing concerns about summertime overheating, and to redress the wintertime focus of virtually all infiltration studies, this work measured infiltration rates in the summer.
The aim of this study is to examine whether infiltration estimation methods can accurately calculate summertime infiltration rates by comparing such estimates to infiltration rate measured using tracer gas. The key question is: "Are summertime infiltration rates produced by commonly used estimation methods reliable?".

The test house
A single, unoccupied, semi-detached test house, constructed in the 1930s, was used for all tests (Plate 1). Semi-detached houses are the most common dwelling type in England (25.3%) and most were built between 1919 and 1944 (52%) (MHCLG, 2018). Across England, there is surprisingly little variation between such 1930's semi-detached houses (Allen and Pinney, 1990).
The test house is located in a suburban residential area of Loughborough, UK (52.7710718N, 1.2242648W). The front of the house faces south-southeast (1608) towards a front garden and a road and the rear of the property faces north to a large back garden (Plate 2). The house has an exposed west-facing façade and adjoins to another semi-detached dwelling on the eastern side. The adjoining, identical dwelling was unoccupied during all testing, with no source of CO 2 present to influence the tracer gas tests. There are neighbouring houses of similar roof height to the east and west (Plate 2).
The house has a floor area of 85.4 m 2 (including both floors), an internally-measured envelope surface area of 226.0 m 2 and a total volume of 209.2 m 3 . Floor to ceiling heights are 2.5 m on the ground floor and 2.4 m on the first floor ( Figure 1). Scale drawings of the house and local site plans are publicly available (see Roberts et al., 2019b).
The house is constructed of uninsulated masonry-cavity walls, wet plastered with gypsum-based plaster on the internal walls, and a clay-tiled pitched roof. The ground floors comprise suspended timbers throughout (except the kitchen), which are ventilated by subfloor air bricks. In the kitchen the floor is solid concrete and unventilated. Carpet tiles, approximately 7 mm thick cover all floors apart from in the kitchen, upstairs bathroom, and Since construction in the 1930s, the fireplaces in the living and dining room have been removed, bricked up, and plastered over. In 2016, the existing single-glazed, wooden-framed windows and external doors were replaced with double-glazed uPVC elements. At the same time, the roof tiles were replaced with like-for-like clay tiles and the bitumen roofing felt was replaced with a vapour-permeable membrane. The loft (attic) was insulated with 300 mm rockwool at the joists. The loft hatch was insulated, but not draught-proofed. A tile-hung exterior wall on the front bay between the living room and front double bedroom had new like-for-like clay tiles overlaid on a vapour-permeable membrane. Some further modifications were made to the test house by the experimenters, prior to commencing the tests. The windows on the western façade were blocked with 50 mm foilbacked PIR insulation, cut precisely to size and inserted into the entire window reveal [2]. An airtight fit was assured using aluminium tape adhered to the insulation and wall surrounding the window, and the seal was verified using smoke sticks (Roberts et al., 2018). Wall vents were sealed using aluminium tape on the interior wall, but the exterior opening was left unsealed (Roberts et al., 2018). Fireplace vents were also sealed on the interior wall only. All sub-floor ventilation air bricks were left unsealed, with the size and locations available in Roberts et al. (2019b).
During the tracer gas and blower door tests, the houses were unheated and without mechanical ventilation (i.e. free-running

Note(s):
The party wall is to the right hand side of the drawing, and adjoins to a house of identical geometry, with a mirrored floor plan (not shown) Estimating infiltration rates in summer Windows were always closed during testing, in order to examine the infiltration rate only. To measure additional background ventilation, in some tests, trickle vents were opened via the flap shutter. All trickle vents were open or closed simultaneously, although examination of different numbers of trickle vents has been explored in previous work (Roberts et al., 2017). In total the combined equivalent area of the 19 trickle vents was 25,000 mm 2 (see Roberts et al. (2017) for trickle vent locations).
During all tests, the indoor dry bulb temperature was measured at one-minute intervals using a U-type thermistor hung from a tripod in the centre of every room at a height of 1.1 m from the floor and wired to a data logger ( Table 2). All thermistors were calibrated prior to use in a water bath against a calibrated thermometer. A tubular foil-coated bubble wrap radiation shield was placed over every tripod to reduce incoming solar radiation contacting the thermistor. This radiation shield design allowed air to move freely around the thermistor, and care was taken to avoid the thermistor contacting the shield or the tripod.
Outdoor dry bulb temperature was measured at the location of the test house at oneminute intervals, using the same type of calibrated thermistor as used indoors. To protect from rain and solar radiation, the thermistor was encased in a naturally-aspirated radiation shield, which was placed in the garden on the north side of the house (Plate 1b). Wind speed and direction were measured at 10-min intervals at the University weather station 1 km from the test house (Table 2). There may be small differences between the weather conditions at the test houses and weather station due to the differing topography and sheltering or canyoning effects of surrounding buildings and trees.

Measuring infiltration and background ventilation rate with tracer gas
Fifteen whole house tracer gas tests were conducted to measure infiltration (trickle vents closed). Two tests in March 2017 (spring), and the remaining 13 in June-August 2018 (summer) ( Table 3). Four further tests to measure the combined infiltration and background ventilation (trickle vents open) were completed in August 2018 (Table 3).
The UK summer of 2018 was the joint-hottest summer since records began in 1884 (McCarthy et al., 2019). For the two tests conducted during the spring, there was no heating in the test houses for some weeks prior and the temperature difference between indoors and outdoors remained small (Tables 3 and 4).
The maximum indoor temperature recorded during the tracer gas tests was 27.8 8C and the maximum outdoor temperature during testing was 26.1 8C. These maximum temperatures did not occur during the same test. The predominant wind direction was south-west (218.48) during the infiltration tests (trickle vents closed) and west-southwest (

Estimating infiltration rates in summer
The decay method, in compliance with the American Society for Testing and Materials (ASTM) standard (ASTM, 2000), was used for all tracer gas tests. Carbon dioxide (CO 2 ) was selected as the tracer gas. The tracer gas was injected from a 22 kg bottle of CO 2 that was connected via PTFE tubing to a solenoid valve attached to a large floor-standing fan to immediately distribute the gas away from the dosing location. Pilot testing revealed that proper distribution of tracer gas throughout the house could be achieved with two injection points: one on the ground floor in the hall, and one on the first floor on the landing. Pilot testing using gas monitors at different heights showed that additional tracer gas mixing with room air was required, and so additional floor-standing fans were placed in every room and switched on during the tracer gas injection period to ensure proper mixing and to reduce stratification of the tracer gas. During this time, the operator monitored the CO 2 concentration in each room to check for homogeneity and made adjustments to the placement of the injection fans as necessary. CO 2 was injected in sufficient quantities to achieve a minimum concentration of 2,500 ppm evenly dispersed throughout the house. Once this tracer gas concentration was achieved, injection of gas was stopped and the fans were left running until homogeneous distribution of tracer gas was achieved in all rooms, the fans were then switched off. Data were not analysed during periods when fans were running.
Tracer gas concentration was measured using a multi-zone gas sampler [3] which took regular samples of air from the volumetric centre of six rooms, three on the ground floor and three on the first floor: the living room, dining room, hall, front double bedroom, rear double bedroom and landing. Air samples were pumped through nylon tubes, passing a dust filter, into the sampler and transferred via PTFE tubing (the tubing material selected to minimise absorption of air samples) to a gas analyser [4]. The gas analyser used photoacoustic infrared spectroscopy to analyse the air sample for CO 2 concentration and compensated for water vapour (Lumasense Technologies Ltd, 2016). Sampling intervals were continuous, meaning that after a sample had been taken and analysed, the next was taken immediately. In practice the sampling interval was every three minutes, as it took around 30 s for each of the six sampling points. This measurement frequency is greater than the minimum recommended sampling frequency of 15 min (ASTM, 2000).
The combination of multi-zone sampler and gas analyser were chosen because they can monitor six separate zones (rooms) simultaneously and had a high degree of accuracy (±3% [5]). The ASTM standard recommends ±5% accuracy (ASTM, 2000). The gas analyser was factory calibrated immediately prior to the experiments commencing and a nozzle calibration procedure was conducted by the operator prior to each individual test.
Data analysis began with a conformity check. To meet uniformity of concentration criterion stipulated in the ASTM standard (ASTM, 2000), the tracer gas concentration must differ between sampling points by less than 10% of the average concentration for the whole measurement zone. During the initial dosing and mixing phase, heterogeneity of concentration was expected, and analysis of the decay did not begin until homogeneity (within 10% of the mean) was established and the decay of tracer gas had begun. Tests were rejected from the analysis if the tracer gas concentration at one or more sampling points  (ASTM, 2000). Air change rate (1/h) was derived from the slope of the regression line that represents ln(C N ) against time (Equation 1) (Roulet and Foradini, 2002). Being a natural constituent of air, CO 2 decay will never reach zero. To account for this, ambient CO 2 was measured as 430 ppm [6], using the method outlined by Roulet and Foradini (2002).
where C N ppm 5 the normalised concentration of CO 2 at time t, C(t) ppm 5 indoor concentration of CO 2 at time t, C(0) ppm 5 indoor concentration of CO 2 at start of test, C o ppm 5 outdoor concentration of CO 2 (430 ppm).

Measuring air permeability with blower door tests
Blower door tests were used to measure the whole house air flow rate of the test house after the method described by ASTM (2007) and ATTMA (2016) on days between January and March 2017. The air flow was induced and measured by a Model 3 Minneapolis Blower Door via depressurisation. The blower door test method uses the relationship between flow through the envelope and the pressure difference across it to quantify air permeability (Sherman, 1987). The test pressure difference of 50 Pa is selected due to being achievable by standard blower door devices in most houses, but high enough to make the test reasonably independent of weather influences (Chan et al., 2013). The fan speed was controlled and data were logged using the Tectite Express 3.6 software (TEC, 2021). The Tectite Express software took at least 10 measurements of air flow rate at a range of building pressure differences between approximately 90 Pa and 20 Pa at intervals of between 5 and 10 Pa, which always included a measurement of air flow rate at a pressure difference above and below 50 Pa. The whole house air flow (m 3 /h) measured by the blower door was normalised by dividing by either dwelling surface area or volume to calculate q 50 or n 50 respectively. Two types of tests were conducted as per the ASTM standard (ASTM, 2007). With all trickle vents closed (n 5 34), and with all trickle vents open (n 5 8). The tests were conducted on 13 separate days in January, February, and March 2017 by the same operator. The blower door was placed into the same opening for all tests (dining room external door, Figure 1). This door was chosen after pilot testing alternative doors which were found to be inadequate due to shape (the front door was curved at the top and this made fitting a tight seal difficult, even after a section of wood was cut to fit) or because of obstructions from internal doors (kitchen external door) (Roberts et al., 2017).
The weather conditions during the blower door tests were recorded (Table 5). Most blower door tests were conducted under a south-westerly (220.58) wind direction, as is the prevailing wind direction in the UK Although six of the 34 blower door tests with trickle vents closed were conducted under wind speeds which exceeded the 6 m/s maximum recommended wind speed for testing (6.2-6.8 m/s) (ATTMA, 2016), they did not deviate from the mean value by more than 0.1 m 3 /h m 2 and were retained in the dataset.  6 and 7). The weather conditions during the tracer gas tests were previously noted in Section 2.2. The infiltration estimation methods required inputs from measured values as well as other assumptions (Tables 6 and 7). Where a method required data from a blower door test, this was taken as the mean value from the 34 tests with trickle vents closed.
Seasonal variation in building envelope air permeability could result in air leakage being lower in summer than winter (Warren and Webb, 1980b;Persily, 1983;Kim and Shaw, 1986;Bassett, 1992;Bracke et al., 2016). Thus, conducting the blower door tests at different times of the year to the tracer gas infiltration measurements could reduce the reliability of the infiltration estimation methods. To mitigate this, six of the blower door tests and two of the tracer gas tests were conducted in the same month (March 2017) (Table 3) with these results similar to the findings of the entire dataset. Furthermore, an additional fan pressurisation test was done in September 2017 (for quality assurance purposes and not included in the analyses), which yielded a result similar to that which had been recorded in the January-March 2017 tests (14.38 m 3 /h m 2 @ 50 Pa). Therefore, the influence of seasonal variation in The letter "M" denotes that the mean measured value for ΔT or wind speed was used (unique for all 15 tests) Table 6. Assumed values for each of the ASHRAE Basic, ASHRAE Enhanced, K-P UK, K-P US, LBL, and Modified K-P infiltration estimation methods IJBPA building envelope air permeability on the reliability of the infiltration estimation methods is deemed to be negligible in this study.
The equations for each method are listed in the referenced sources and repeated with full description of the specific inputs for this house in Roberts (2020).
(3) K-P (Kronvall-Persily) UK (Kronvall, 1978;Persily, 1983). Which uses a divisor of 20 to reduce the blower door data. The method is applied differently in the United States and the United Kingdom: in the K-P UK model, air permeability (q 50 ) is used (Poza-Casado et al., 2020).
(5) Lawrence Berkeley Laboratory (LBL) model (Sherman and Grimsrud, 1980;Sherman and Modera, 1986 The letter "M" denotes that the mean measured value for ΔT or wind speed was used (unique for all 15 tests). "R" denotes that a reference value for wind speed was used (subject to the month of the test) Table 7. Assumed values for the SAP Algorithm (measured wind), SAP Algorithm (reference wind), SAP q 50 /20 (measured wind), SAP q 50 /20 (reference wind), and Sherman Simplified infiltration estimation methods Estimating infiltration rates in summer (6) Modified divisor K-P model n 50 /30 (Dubrul, 1988;Liddament, 1996). As the K-P US model, but the leakage-infiltration ratio is 30, as suggested for low-rise buildings by Dubrul (1988) and Liddament (1996).
(8) SAP Algorithm (reference wind) (BRE, 2012). Reference wind values for wind speed for the corresponding month as supplied in SAP.
(11) Sherman Simplified model (Sherman, 1987). This method usually requires annual average wind speed and indoor-outdoor temperature difference, but in this analysis the average during each of the tracer gas test intervals was used.
Standard metrics were derived to enable the difference between the measured and estimated infiltration (i.e. the error E) to be quantified (Table 8). For each of the standard metrics, E t 5 P t À M t , where P t is the estimated infiltration rate for a particular test t, M t is the measured infiltration rate for the matching test t, and n is the total number of infiltration rates being compared.

Results
The tracer gas test results are provided to establish the measured infiltration rate; the blower door test results are given, as these inform many of the infiltration estimation methods; the measured and estimated infiltration rates are compared.

Infiltration rate and background ventilation rate measured by tracer gas
The mean air change rate values for infiltration (trickle vents closed) and background ventilation (trickle vents open) as measured by tracer gas were calculated (Table 9) from the tracer gas decays (Figure 2). The mean air change rate was lower when trickle vents were closed than open, and the difference in means was statistically significant (p < 0.05).

Air permeability measured by blower door tests
The mean air permeability (q 50 ), as measured via blower door tests (  (HM Government, 2010), though comparable to dwellings of similar age (Stephen, 2000). Opening all trickle vents resulted in a statistically significant increase in mean air permeability (p < 0.05). It was decided that a single mean value for q 50 and n 50 could be used in the infiltration estimation methods that required it. This was on the basis that the sample means, as calculated from 34 blower door tests with trickle vents closed had a very small standard error which indicated that the population mean would fall between ±0.09 m 3 /h m 2 @ 50 Pa (and 0.09 ach @ 50 Pa for n 50 ) in 95% of cases (Table 10). Thus, a single central mean value for q 50 and n 50 was used in all the infiltration estimation methods that required it.
Some infiltration estimation methods required parameters from the blower door tests instead of q 50 or n 50 , namely flow coefficient C, flow exponent n and effective leakage area (ELA). The value for each of these input parameters was taken as the mean from 34 blower door tests with trickle vents closed (Tables 6 and 7).

Estimating infiltration rates in summer
Qualitative analysis of air leakage paths using neutrally-buoyant smoke and infrared thermography showed the main routes for infiltration to be: at plumbing and electrical service envelope penetrations; around some plug sockets; at the floor/wall interface and around skirting boards; at the loft hatch; and under window ledges. Windows were generally well sealed, but there was some leakage around trickle vents, even in their closed position. This is reflected in earlier work in this house which found the air permeability to be 2% lower with trickle vents in their closed position and sealed with tape compared to being closed but unsealed (Roberts et al., 2017).

Comparing measured and estimated infiltration rates
All infiltration estimation methods predicted a higher mean infiltration rate than was measured by tracer gas (Table 11 and Figure 3). The mean infiltration rates estimated by each of the methods varied from 64 to 208% higher than the mean measured infiltration rate.
The ASHRAE Enhanced method estimated the mean infiltration rate that was closest to the mean measured value, but was still significantly (p < 0.001) higher than measured (Tables 11 and 12, and Figure 3). The K-P US model mean was furthest from the mean measured value [7]. Dividing n 50 by 30 (K-P modified divisor) was the second most accurate method, but a divisor of 58 would be required to perfectly predict the mean infiltration rate measured in this study (under summer weather conditions). Thus, none of the infiltration estimation methods could accurately predict the mean infiltration rate in this house, under the summer weather conditions experienced.
Whilst all the mean estimated infiltration rates were significantly different from the mean measured value, some of the individual data points were close to the corresponding measured value, and in some cases lower than the measured infiltration rate (Table 12 and Figure 4). The ASHRAE Basic, ASHRAE Enhanced, LBL model, and the Sherman Simplified method all have minimum errors of ≤0.10 ach (positive and negative, i.e. greater and less than measured). Equally, some single data points are much higher than the measured value (see maximum error in Table 12 and Figure 3). Therefore, the methods trialled cannot be relied upon to make consistently reliable estimations of the summertime infiltration rate.
The estimation methods which accounted for the weather conditions (and indoor temperature) were generally more reliable estimators of infiltration than those which did not (e.g. K-P US and K-P UK methods) (Figure 3). Using a locally measured value for wind speed generally improved the predictions compared to a reference value for wind speed (e.g. compare SAP measured wind to SAP reference wind).
However, the use of a variable input: locally-measured wind speed and the indoor-outdoor temperature difference, meant that the range of infiltration estimation was, unsurprisingly, higher than in the estimation methods which did not account for wind (or used a reference value) and, more interestingly, greater than the range in measured values (Figure 3 and Table 11). Yet, despite the greater range in estimated infiltration rates when locally measured wind speeds were used, the mean values were still generally closer to the measured value than if no wind speed data were used.
The variation in the 15 measured infiltration rates may be attributed to the varying weather conditions in each of the 15 tests (Roberts, 2020). The slope of the regression line for each estimation method is indicative of each method's response to those weather conditions (Figure 4). The ASHRAE Enhanced, LBL and Sherman Simplified methods all appear to respond correctly to the change in weather conditions, with the slope angle of the lines close to 1, i.e. the line of quality, albeit significantly above (Figure 4). The other estimation methods using measured wind speed are somewhat reliable predictors of the change in infiltration rate. The two SAP methods with reference wind do not display the correct rate of change, neither do the K-P UK, US, or n 50 /30 methods, due to having no weather data input (Figure 4).  Table 11. Descriptive statistics for the measured and estimated infiltration rate Estimating infiltration rates in summer 4. Discussion This paper is, to the authors' knowledge, the only time that eleven infiltration estimation methods have been compared to measured infiltration in a UK dwelling. Also notable is the summertime focus of this research, where most previous studies have examined wintertime infiltration. This research has shown that none of the infiltration estimation methods tested were reliable predictors of the mean infiltration rate under the summertime conditions in which they were examined. The mean infiltration rates estimated by the 11 methods were between 64 and 208% higher than the mean measured infiltration rate. The commonly used K-P UK and K-P US ("divide-by-20") methods should not be used to estimate summertime infiltration in typical UK homes. This finding is supported by previous studies (Keig et al., 2016;Johnston and Stafford, 2017;Cardoso et al., 2020;Vega Pasos et al., 2020) and, crucially, this study adds evidence with a summertime focus.
For the house tested in this work, and under these weather conditions, a divisor of 58 for q 50 and n 50 was required to perfectly estimate the mean measured infiltration rate. This is even higher than the divisor of between 37 and 39 suggested elsewhere (Vega Pasos et al. 2019. It is possible that seeking a single divisor for estimating infiltration from blower door tests is futile for anything other than very low resolution, low reliability estimates of annual average infiltration rate: there are complex geometrical considerations and infiltration is highly dynamic and weather-dependent (Jones et al., 2015).
Generally, the methods which were sensitive to the varying nature of infiltration rate were the more reliable estimators of the mean infiltration rate. The methods which accounted for the wind speed, indoor-outdoor temperature differences, and perhaps other information   These findings have implications for the reliable prediction of overheating risk and cooling demand. Overheating risk may be higher than models predict, as ventilative cooling is not as high as assumed by modellers. This could mean that appropriate and necessary passive overheating mitigation strategies are not implemented at the design stage. This could lead to overheating in new homes and the subsequent uptake of energy intensive airconditioning.
Looking to the future, cooling demand is likely to increase in UK homes (Gupta et al., 2015), and reliable estimations of infiltration will be needed to accurately predict cooling load. Windows are likely to be closed when mechanical cooling is operating, and so infiltration rather than ventilation through purpose-provided openings will become the dominant air exchange path. Reliable estimation of infiltration is therefore imperative to design cooling systems of appropriate size.
These findings also indicate that indoor air quality may be worse than expected, due to the air change rate being lower than current estimates. The test house used in this research was at risk of poor indoor air quality when the windows were closed in summer, and even when the trickle vents were open. A new UK dwelling of equivalent size to the test house requires a background ventilation rate of 25.62 l/s (0.44 ach), which is 22.5% higher than the 20.92 l/s (0.36 ach) that was measured (HM Government, 2010). In contrast, had the infiltration estimation methods been used to assess the indoor air quality, all but the ASHRAE Enhanced  Measured infiltration rate plotted against estimated infiltration rate IJBPA method would have (incorrectly) assumed indoor air quality to be satisfactory in the house, even without the trickle vents open. Therefore, households may be at risk of exposure to poor indoor air quality in summer when the windows are closed on hot days to prevent ingress of warmer outdoor air. This has implications for health and wellbeing. Ultimately, this work suggests that trickle vents cannot be relied on for the provision of satisfactory indoor air quality in summer. Further measurement in a wider range of homes is urgently needed to address this. The ASHRAE Enhanced method (ASHRAE, 2013), which is called the Flow Coefficient model in Energy Plus (DoE, 2020)) shows the greatest potential for adaptation and adjustment to summertime conditions, based on these results. The ASHRAE Enhanced method estimated a mean infiltration rate that was both closest to the measure infiltration rate and the gradient of the regression line (Figure 4) was similar to the gradient of the line of equality. Thus, it appears that this method was reliably sensitive to the changes in infiltration rate. Therefore, the ASHRAE Enhanced method shows the greatest promise and further investigation and refinement of this approach for estimation of infiltration rate in summer is recommended.

Conclusion
Due to the time, expense and intrusive nature of measuring infiltration directly using tracer gas in homes, infiltration rate is usually estimated, and often using information collected in a blower door test. This paper adds to the body of evidence that the commonly used methods of estimating infiltration rate are inaccurate in UK homes, and so infiltration is not reliably estimated. Importantly, where previous work has tended to focus on wintertime infiltration, this study demonstrates the magnitude of infiltration estimation error in the summer season. Infiltration in summer is important when predicting summertime overheating risk, calculating mechanical cooling loads, and assessing indoor air quality on hot days when occupants are encouraged to close windows to prevent this ingress of warmer outdoor air.
Infiltration rate (trickle vents closed) was measured by 15 whole house tracer gas tests in a typical semi-detached English dwelling built in the 1930s. Thirty-four blower door tests were carried out under the same conditions to measure the dwelling air permeability. Four tracer gas and eight blower door tests were also conducted with trickle vents open to measure background ventilation rate. The majority of tests took place during the joint-hottest summer on record in the UK. The measured infiltration rate was compared to that estimated by eleven infiltration estimation methods. The key conclusions are the following: (1) None of the eleven infiltration estimation methods trialled was a reliable estimator of the mean infiltration rate measured in the test house.
(2) If the commonly used K-P "divide-by-20" rule of thumb is used to estimate infiltration, the divisor for q 50 and n 50 should be replaced by 58. However, attempts to define a single value to reduce blower door data (q 50 or n 50 ) to infiltration rate is futile when considering the highly dynamic and weather-dependent nature of infiltration.
(3) Infiltration estimation methods which account for wind speed (especially when locally measured), indoor-outdoor temperature differences, and perhaps other information about the building and wider site are recommended to achieve more reliable estimates of infiltration, but still differences between the measured and estimated infiltration remained.
(4) The ASHRAE Enhanced infiltration estimation method was closest to the mean measured infiltration rate and demonstrated a similar rate of increase in estimated infiltration in line with the measured value. Thus, this method holds the greatest Estimating infiltration rates in summer potential for adjustment and adaptation to be more suitable for estimating infiltration in typical UK homes during summer.
(5) Incorrectly estimating infiltration could lead to incorrect assumptions regarding indoor air quality, overheating risk and cooling loads. In this test house, indoor air quality was measured as unsatisfactory even with trickle vents open, despite all but one infiltration estimation method (ASHRAE Enhanced) predicting a sufficient air change rate even with trickle vents closed.
Further research is required to adjust the existing infiltration estimation methods which show the greatest potential, i.e. the ASHRAE Enhanced method, for more reliable estimation of infiltration rate under summer weather conditions. This work has presented evidence from a typical UK home, but future work should also consider new build homes, which are likely to be built to higher standards of airtightness. This will allow for more reliable infiltration estimation methods to be developed and thus for proper ventilation strategies to be designed both to ameliorate summertime overheating and ensure provision of satisfactory indoor air quality.
Notes 1. Figures for the global energy demand and CO 2 emission due to heat production are limited. The quoted values include heat consumption for space heating and water heating in buildings, for cooking and for operating industrial processes.
2. This endeavour was primarily to reduce solar gains entering via glazing on the western façade but is noted here as it is likely to very slightly reduce the infiltration rate compared to a similar house of comparable age and construction. 6. ppm 5 parts per million.
7. The values for q 50 and n 50 were similar because the building surface area and volume are similar. Thus, the K-P US and K-P UK methods yielded similar, but not identical, results. The values would be different in a house where the surface area to volume ratio is not close to 1.
8. In dwellings of non-standard construction, caravans, a divisor for n50 of 40 has been suggested (Miles-Shenton et al., 2015).