The Impacts of Climate Change on Historic Buildings: Heating and Overheating in the Tower of London

Abstract

The built environment requires retrofitting on a massive scale to both mitigate the impacts of climate change and adapt to future conditions. Buildings of high thermal mass offer useful insights into the limits of passive measures in futureproofing against overheating. Historic buildings typically have a higher thermal mass and also offer a paragon case to study improvement options that do not compromise heritage or character. This paper focuses on the Tower of London as a case study. Data from pre- and post-COVID-19 allows insights into the building performance in the absence of end users. A thermal model is calibrated to accurately represent both the physics of the building and the impacts of its occupants in use. Future weather files then test the extent to which the building’s thermal mass can mitigate against overheating under a range of climate warming scenarios. The results suggest that prolonged heat waves pose a serious risk to passive mitigation strategies as the mass of the building stores heat it cannot shed overnight. These scenarios also reduce the heating demand in winter. The results suggest that the built environment faces subtle design challenges in understanding the limits for passive design techniques versus the need for cooling in a warming climate. For the Tower in particular, a significant increase in overheating is likely in the coming decades.

1. Introduction

The world has already experienced 1.1 °C of warming compared to preindustrial levels. In July 2022, the UK experienced temperatures exceeding 40 °C for the first time [1]. Climate impacts are only accelerating, and these types of heat waves will become greater in severity and longer in duration. All buildings must adapt to remain fit for purpose and fit for the future.

Historic buildings offer a useful subgroup of the built environment for several reasons. Firstly, their age limits the range of potential adaptation measures that can improve their performance without compromising their character. These restrictions drive innovation in technical solutions and retrofit measures. Secondly, they offer useful examples of how buildings have evolved through past transitions to maintain modern standards of comfort. For example, any building that had failed to incorporate plumbing and electricity systems would have become largely unusable.

Historic buildings are also important cultural symbols, and often act as a catalyst, driving public interest and support for major infrastructure programmes. For example, Buckingham Palace switching from towngas to natural gas during the conversion programme of 1965–1977 was used to boost confidence in the North Sea gas programme nationwide.

Historic buildings typically have a high thermal mass, which is useful for exploring the limits of passive heating and cooling measures. Studies of thermal mass typically struggle to reconcile measured with monitored data, due to the large number of variables involved, particularly in understanding the building’s operation and users. As climate impacts intensify, mechanical cooling will become an increasing necessity in many cases but incurs its own energy use and climate emissions. A more granular understanding of the performance of passive measures under different climate scenarios will help buildings sense when mechanical cooling is essential and minimise its use while maintaining comfort under heat wave conditions.

Given the age of much of the European building stock, and the UK stock in particular, historic buildings are not niche but, rather, hold important lessons about climate adaptation measures that are highly relevant to roughly one third of the built environment. However, their unique features can also limit the generalisability of the findings for individual cases. For these reasons, there is a growing interest in the field of research into climate adaptation methods in historic buildings that do not compromise their character and more fully characterising a range of case study examples that increase the generalisability of the findings for other building types.

This study contributes to this literature by considering a case study for the Tower of London. It compares data from pre-COVID-19 to data recorded during lockdown. This gives rare insights into the performance of one of the most visited buildings in the world while is unoccupied. It also allows a close calibration of the modelled and measured data in the absence of occupancy variables. Finally, it uses this calibrated model to give new insights into the performance of high-thermal-mass heritage buildings in a warming climate and discusses the technical and practical feasibility of mitigation techniques.

2. Literature Review

2.1. Climate Change in Historic Buildings

Nearly all UK buildings require upgrades to improve their performance and reduce their emissions. The available solutions typically include a combination of fabric measures to reduce heat loss in the winter and heat gains in the summer, as well as heating system upgrades to eliminate the use of fossil fuels on site. The need for mechanical cooling is increasingly common in retrofit.

Historic buildings present an important usage type since they considerably narrow the range of technical measures that can be deployed without compromising their character. They are also typically space constrained and highly used environments, which imposes practical challenges that further restrict available measures. The coming decades will force retrofit decisions on the basis of energy use, carbon, and extreme temperatures outside of the building’s original design.

Despite wide consensus on its importance, there has been limited research into the adaptation process of cultural heritage to climate change [2]. Some of this adaptation literature focuses on building contents [3], degradation and failure of structures [4], and long-term impacts [5]. Hao et al. [6] carried out a thorough review of the impact of climate change on the internal environment of historic buildings. Most studies have focused on energy use, overheating, and moisture dynamics in the envelopes.

There is interest in the appropriateness of different retrofit measures for building quality, particularly in regard to moisture impacts [6]. There is also increasing study into the frameworks underpinning these decisions, in order to better manage trade-offs between conservation and energy performance [7].

A better understanding of the building performance characteristics for historic buildings will be critical in deciding the suitable combination of passive fabric measures and heating system options and whether mechanical cooling will be required in addition to passive cooling strategies. Most UK examples focus on domestic impacts rather than public buildings, and they note the need for further study on the role of thermal mass in future scenarios.

2.2. Heating and Overheating in High-Thermal-Mass Structures

Despite the importance of thermal mass in building design, it remains an under-researched area by comparison with thermal resistance and conductivity [8]. Conductivity describes the rate of heat transfer through a material. The conductivity and shape of the material determine its thermal resistance. Thermal resistance is the dominant characteristic in reducing fabric heat loss. Significant savings potential for the use of building thermal mass has been demonstrated in commercial buildings using simulation, controlled laboratory testing, and field demonstrations. However, the savings are sensitive to many factors, including utility rates, the type of equipment, occupancy schedule, building construction, climate conditions, and control strategy [9].

The thermal mass of a building is a key determinant in how the structure responds to changing loads with time. Most of the literature focuses on the relationship between thermal mass and cooling load, not on its heating load (e.g., [10]). Passive solar design is used to maximise the benefit of thermal mass throughout the year by optimising the building’s form, fabric, orientation, and ventilation to maximise energy efficiency and comfort [11]. Studies have shown that the effective utilisation of thermal mass as a passive design principle can reduce peak heating loads 18–50% as cited by [10].

Most studies agree that high thermal mass can be used to reduce energy use; however, some have found that in colder climates, high thermal mass can cause an increase in energy use [8]. In the UK, the potential to reduce heating loads in winter ranges from 10% up to 30% if sophisticated passive solar techniques are utilised [11].

Taken as a whole, high thermal mass is consistently beneficial in reducing summertime overheating but of variable benefit in reducing winter heating consumption. For low levels of insulation, a higher thermal mass can increase energy consumption. For high levels of insulation, a higher thermal mass will likely decrease energy consumption. High thermal mass is more likely to reduce energy consumption in the warmer southern parts of the UK and when there is a high continuous occupancy [12,13].

Overall, heating loads are primarily driven by the heat loss parameters of the building, and within the fabric, the level of insulation and resulting U-values typically dominate over the thermal mass in affecting heat demand one way or the other [14]. Thermal mass can be used to reduce peak heating loads but has a lesser impact on overall energy consumption [15].

Buildings with a very high thermal mass offer an opportunity to explore the limits of thermal inertia to impact comfort conditions, but these are underexplored in the literature due to the lack of case study examples. Historic buildings offer a useful case study group to address this gap. However, they present their own challenges in performance monitoring and creating suitably calibrated thermal models.

2.3. Performance Monitoring and Modelling Future Weather

As the climate warms, there is increasing interest in how to mitigate overheating in buildings. The UK has historically had mild summers and, therefore, low drivers for mechanical cooling in buildings. Even passive solutions are widely under-implemented and so, with a lack of both passive and active solutions in place, UK buildings are uniquely vulnerable to overheating in a rapidly changing climate.

Retrofitting UK buildings will be critical in making them fit for the future and mitigating the impacts of extreme temperatures across seasons. Passive solutions including shading and natural ventilation are favoured since active cooling incurs its own energy and climate impacts. However, most passive solutions to overheating rely on shedding excess heat overnight. During prolonged heat waves, buildings may lose the capacity to do this, thereby limiting the effectiveness of passive cooling strategies. A better understanding of building performance is critical for more accurately characterising the effectiveness and limitations of passive cooling solutions under different climate warming scenarios.

There has been poor agreement between the literature and data measured in practice [8]. This is possibly due to the reliance on sinusoidal temperature profiles in the calculation procedures, which vary from true weather data. Dynamic simulations have worked to improve upon these assumptions [16].

Hao et al. [17] modelled overheating risks under future climate conditions for Alpine homes under a range of retrofit scenarios. They found that most retrofit measures that improve energy performance also increase overheating and require careful planning across seasons.

In terms of design and modelling, Kosny and Kossecka [18] showed that many simulation programs and codes provide inadequate results when modelling high-mass buildings, as they were developed for and tested with structures with much lower thermal storage capacity. Most such studies also suffer from limited generalisability due to the building-specific nature of the calculations and, thus, their conclusions [8].

This paper, therefore, addresses these key gaps. It seeks to understand the vulnerability of historic buildings to climate change in the UK by focusing on the Tower of London as a case study. It will consider the performance impacts of high thermal mass across seasons, studying both heating and overheating under a range of climate warming scenarios. Additionally, it will address the gap between modelled and measured studies of the performance of high-thermal-mass buildings by controlling for occupancy and user variables using a unique dataset gathered during the COVID-19 pandemic lockdown. This approach was also used by [19] in calibrating a model of Italy’s Duomo di Milano, with the focus on the artwork and contents rather the building performance. The generalisability of the results will be discussed and compared to other hard-to-treat building types.

2.4. The Tower of London Case Study

Use:

The Tower of London is among the most famous buildings in the world. It was founded in 1066, with the White Tower at its centre, then expanded through the 1300s with the establishment of the inner and outer wards, and a wall enclosing a total site area of 4.9 hectares. It has been continually retrofitted with more modern amenities throughout history, but the general layout has remained unchanged.

Historically, it served as a castle and prison. Today, it is a World Heritage Site and active tourist venue and the most visited paid-for attraction in the UK with over 2.8 million visitors per year. It hosts a wide range of historical artifacts, among the most famous of which are the Crown Jewels of England.

The name Tower of London refers to the castle and its grounds, which is made up of a collection of buildings within two concentric rings of perimeter wall. Today the various buildings serve a range of uses including tourist/museum areas, a chapel, offices, catering, events venues, and private residences. This study focuses on the White Tower building only and uses information obtained by the researchers through a series of site visits and engineering surveys.

Construction:

The White Tower (originally a keep) lies at the centre of the complex and was built in 1078. The materials and architecture reflect this period and function, with a footprint of 35 by 32 m and measuring 27 m tall. The walls are primarily made of Kentish rag-stone, with thicknesses varying from 1 to 6 m. Figure 1 shows a window reveal through a six-meter-thick perimeter wall. There is a subgrade basement, three upper floors, and no insulation anywhere in the building.

There are very few service apertures and very little air leakage through joints or bridges due to the material thicknesses; however, the windows are all single glazed and poorly sealed. Some windows and doors are left open continually even during winter due to the high volume of visitor traffic through the spaces. Intermediate floors and some partitions are timber, but most partitions are stone, typically one or more meters thick. The building, thus, has an extremely high thermal mass, but the fabric and ventilation heat loss characteristics are unknown and will need to be determined through the course of this work.

Services:

The building services strategy relies on a wet heating system and natural ventilation. There is no comfort cooling or mechanical ventilation. There are radiators around the perimeters of most spaces (see e.g., Figure 1). Some spaces have displacement ceiling fans to redistribute air internally, but all fresh air is via natural ventilation and air leakage. The heating source is a hot water (~80 °C) district heating system. The heat network is fed by a gas-fired boiler housed in a support building that serves the White Tower itself, as well as several other buildings within the Tower of London complex.

There is no sub-metering on the portion of the heating circuit entering the White Tower; therefore, heating loads can only be estimated as a fraction of the overall load on the heat network. The pumps that control the amount of heat given to the Tower are set to maintain comfortable temperatures but have limited controls and feedback from the space. That is, if the space becomes densely occupied and overheats, the heating system will not adjust in response. Rather, is it set at a baseline level that will maintain comfort when the space has expected levels of occupancy.

Internal Gains:

Due to its high thermal mass, the White Tower currently maintains largely comfortable conditions in the summer. The building absorbs solar heat gains during the day and reradiates this heat at night, when the environment is cooler. Currently, peak solar gains are sufficiently low, and heat waves are short enough in duration that this daily cycle keeps much of the heat on the outer layers of the stone and reradiates it back to the environment rather than allowing it to penetrate as an internal heat gain within the space. This work will calibrate the model to characterise this current operating mode and test the extent to which it remains true in a warming climate with prolonged heat waves.

There are very few equipment gains, but there is a significant lighting load, particularly display lighting for museum pieces and artefacts. By far the most significant internal gains are from the occupants themselves in the form of both sensible and latent heat loads. The Tower counts overall admissions to the complex but not specific foot traffic through the White Tower building itself. Furthermore, most of the space has transient occupancy, with visitors constantly moving through. The daily profile of the occupancy gains for the space must, therefore, be determined by this work.

This study will explore two challenges that the Tower faces in the coming decades. Firstly, it will determine if the current building performance will be fit for purpose in a warming climate, and if not, what passive and active solutions would work across seasons. Secondly, it will consider engineering options to decarbonise its energy use without compromising its historic character.

3. Methodology

The aim of this study is to determine the impact of the White Tower’s high thermal mass on its performance and energy use both now and in future climate conditions. This is approached through the following steps:

• Create a thermal model of the White Tower.
• Calibrate the thermal model to temperature data recorded when unoccupied during lockdown in 2020.
• Add Tower occupancy profiles to the model to match recorded data from 2019.
• Determine the risk of overheating in the Tower using current weather.
• Use the calibrated model to predict the risk of overheating under future weather conditions.
• Compare the heating consumption for the unoccupied, occupied, and future weather cases.

This paper combines data from several sources. The researchers created the Tower geometry for the thermal model using a series of detailed site surveys accompanied by Tower estates personnel. The estates team also provided details of the heating system design and setpoints throughout the test period. The Historic Royal Palaces Armouries researchers recorded temperature sensor data for 10 locations around the Tower from July 2018 to July 2021. The sensors have a resolution of 0.1 °C. The uncertainty on each device was not available, but the core function of the sensors is to monitor environmental conditions needed to preserve historical artefacts and is, therefore, deemed sufficient to use for this study. From 20 March 2020, the Tower was closed due to COVID-19 restrictions. This closure gives a rare opportunity to study the temperature changes of a high-thermal-mass building independent of thermal gains due to occupancy.

External temperature data were obtained from a weather station at Elephant and Castle, approximately 1 km from the Tower site.

The thermal model was created using IESve 2021 software, which uses dynamic hourly simulations of heat conduction and storage to characterise the thermal performance of buildings. IESve is a leading UK software widely used in building simulation research.

Future weather files were used from the Prometheus project, which created probabilistic reference weather years using UK Climate Projections 2009 data [20]. Data from this source were chosen to best match the weather station at Elephant and Castle. The overall workflow is shown in Figure 2, and a summary of the weather file source locations used in this paper is given in

Table 1. Summary of weather data for measured and modelled scenarios.

Weather Data	Approx Climate Change	Scenario Name in Paper	Tavg	Tmax	% > 25 °C	Source
Elephant and Castle 2019	~1 °C (in 2019)	Current Actual	11.6	36.5	1.2%	Metroblue
IslingtonDSY 1961–1990	~1 °C (in 2019)	Current Model	11.4	28.8	0.6%	Prometheus
IslingtonDSY A1B 10th Percentile Scenario 2050, 2080	~1.5 °C (in 2080)	Low Emissions	12.4	29.9	2.3%	Prometheus
IslingtonDSY A1Fi 33rd Percentile Scenario 2050, 2080	~3 °C (in 2080)	Med Emissions	14.5	33.8	5.8%	Prometheus
IslingtonDSY A1Fi 90th Percentile Scenario 2050, 2080	~>4 °C (in 2080)	High Emissions	18.6	42.3	19.2%	Prometheus

below.

The future weather files were created in 2010 under the Prometheus Project and under-sampled extreme temperature peaks, in particular, prolonged heat waves. This is evidenced by comparing the future weather files with the actual external temperatures recorded at Elephant and Castle in 2019. While the average temperature and the % of hours over 25 °C are both broadly consistent between the measured 2019 data and the IslingtonDSY weather file, the actual peak is 36.5 °C compared to the model weather file peak of 28.8 °C. The future weather files would not reach a peak temperature of above 36 °C until well beyond 2050. The trend of actual weather preceding our climate predictions has continued, with record heat waves across the UK in summer of 2022 that were not predicted to happen for several decades to come.

Ongoing research is seeking to better accommodate for both peak heat and peak cold weather events in response to more recent, and much accelerated climate observations over the past few years. The findings of this paper are, therefore, based on characterising the thermal model accurately for its current performance and weather data today, then describing the relative changes in performance from that baseline under theoretical future weather with the characteristics of Table 1. The column indicating the approximate amount of climate change this corresponds to is given for illustration purposes only and does not represent a direct comparison between a single weather file and a specific degree of climate change.

This study created a thermal model of the Tower building and calibrated it to the temperature data recorded during lockdown. This will establish a baseline in which the thermal model is based purely on the physics of the building independent of occupancy variables.

Figure 3 shows the locations of 10 temperature and relative humidity sensors that were used to calibrate the model. The 10 sensor locations were selected to obtain a range of internal and perimeter locations. The gross floor area of each level is approximately 1200 m² with the net area varying with wall thicknesses at each level.

• Create a thermal model of the Tower

The geometry for the Tower thermal model was obtained through a series of site surveys carried out with Tower staff and facilities management. Measurements were taken of key wall thicknesses and window sizes. However, material thicknesses varied significantly throughout and, so, approximations needed to be made for the model.

Air permeability was set through apertures only, using the MacroFlo function within IES. MacroFlo simulates airflow through openings based on their geometry and the wind and buoyancy forces acting on the surfaces. It accounts for the opening crack characteristics, the shape, and timing of the aperture openings. This assumed that air leakage was through external apertures and interzonal air movement only. No additional ventilation or mechanical ventilation was applied in order to replicate the actual services in the building.

The geometry of window openings was created in the model to match what was observed in the manual surveys. Most include sash windows with the opening fraction limited for security purposes.

2.

Calibrate the thermal model to temperature data recorded during lockdown

The IslingtonDSY weather file was selected due to its similarity with the measured external weather data obtained from 2019 and 2020 for the Elephant and Castle weather station. The calibration considered both the detailed changes in temperatures across individual days, as well as the overall trends in temperatures across the entire year.

The model was calibrated to individual days using one week in summer and one week in winter. For summer, the week of 14–17 July 2020 was selected, and for winter, 17–23 March 2021. These weeks were selected because the temperature profiles of the measured data from Elephant and Castle very closely matched the IslingtonDSY weather file. They were both taken during lockdown and, so, the recorded room temperatures would not have been influenced by occupancy. One occurred in summer with high solar gains, the other in winter, allowing to calibrate the heating system setpoints and operation schedule.

The thermal mass, U-values, lighting, and ventilation schedules for the White Tower were adjusted to match these weeks as closely as possible. The heating and services profiles were set to match the timing given by Tower facilities management, and the levels were adjusted to calibrate the temperatures to the observed data. It was found that supplying a base heating temperature of 16 °C and an average lighting density of 3 W/m² gave the closest match. These assumptions were triangulated against site observations. This combination of metrics represents the actual conditions of the space and was also sufficient to match the winter temperatures. Internal apertures for interzonal air movement were created according to the site observations as described above. The aperture opening followed the occupancy schedule of

Table 2. Summary of occupancy schedule for White Tower.

			Open	Close
1 January	28 February	Tues–Sun	10h00	16h30
1 March	5 September	Daily	9h00	17h30
8 September	23 December	Tues–Sun	10h00	16h30
24–25–26 December			closed
27 December	31 December	Daily	9h00	16h30

and the total air change rates in the space, therefore, varied with weather conditions.

The models were calibrated on the basis of the average annual temperature, the peak temperatures, and the daily variations in temperatures. Tables 3–5 show a close alignment in the annual figures for all rooms; however, the sample images in Figure 4 show that it is not possible to perfectly match each of these at the same time for each room across seasons. A balance was struck to give the best overall match in terms of annual averages and daily cycles. When a perfect agreement was not possible across seasons, the model was more closely calibrated to the summer conditions.

3.

Add occupancy profiles to the model in order to match recorded data from 2019

Once the building physics were determined for the 2020 temperatures recorded under lockdown, occupancy profiles were added according to the public schedule for the Tower of London visiting hours as summarised in Table 2.

Using this schedule, a uniform distribution of people at a rate of 10 m² per person was added throughout the zones, and the model was recalibrated to the 2019 temperatures recorded by each sensor.

A suitable match to the recorded temperatures was found by assuming an average of 3000 visitors per day, which is in line with typical Tower attendance. Though Tower attendance varies considerably in practice, for modelling purposes, this was assumed to be evenly distributed throughout the space and over time.

4. Results

4.1. Summer Overheating

Table 3, Table 4 and Table 5 below show the average annual air temperature, the peak zone temperature, and the % of occupied hours above 25 °C for the range of scenarios studied.

The 2019 Actual column is sourced from the temperature sensor data recorded by the Armouries. The external temperature for the 2019 Actual column is from a weather station at Elephant and Castle, nearby the Tower site.

The 2019 Model column uses the IslingtonDSY 1960–1990 weather file, which, as discussed in the methodology section, was selected for its similarity to the Elephant and Castle weather data over several years.

Each table is formatted with the minimum values in green, the maximum values in red, and a gradient in between to visualise the distribution of the temperatures present. The room labels correspond to the locations in Figure 3 above. The general trends observed are that the basement level is always the coolest, with temperatures gradually increasing towards the higher floors of the building. The first two columns of each table show close agreement between the current measured data and the baseline model, due to the model calibration procedure described above. The temperatures then gradually increase from left to right in the table as the future weather files consider higher degrees of climate impacts in 2050 and 2080. The labels ‘low, med, and high’ correspond to the 10th, 33rd, and 90th percentile of future weather scenarios detailed in Table 2.

Table 3. Average zone temperature (°C) for actual and modelled temperatures for 2019. Modelled for 2050 and 2080. Zone map given in Figure 3. (Colour gradient from green = min to red = max).

Average Zone Temperature
	2019		2050			2080
	Actual	Model	Low	Med	High	Low	Med	High
External	11.6	11.4	11.8	13.2	16.2	12.4	14.5	18.6
B_S9	19.9	18.8	19.3	20.2	22.4	19.6	21.1	24.3
G_S10	18.2	19.7	20.3	21.2	23.7	20.6	22.2	25.6
G_S11	19.0	19.4	20.0	20.9	23.4	20.3	21.9	25.4
G_S12	19.5	19.5	20.1	21.0	23.5	20.5	22.1	25.5
G_S28	19.1	19.5	20.1	21.0	23.5	20.4	22.0	25.5
1_S14	21.2	19.9	20.5	21.3	23.8	20.8	22.3	25.8
1_S17	21.4	20.1	20.7	21.6	24.1	21.0	22.6	26.1
2_S30	20.5	20.1	20.7	21.6	24.1	21.0	22.6	26.1
2_S31	20.9	19.9	20.5	21.4	23.9	20.8	22.4	26.0
2_S8	21.0	20.0	20.7	21.5	24.0	21.0	22.6	26.1

Table 3. Average zone temperature (°C) for actual and modelled temperatures for 2019. Modelled for 2050 and 2080. Zone map given in Figure 3. (Colour gradient from green = min to red = max).

Average Zone Temperature
	2019		2050			2080
	Actual	Model	Low	Med	High	Low	Med	High
External	11.6	11.4	11.8	13.2	16.2	12.4	14.5	18.6
B_S9	19.9	18.8	19.3	20.2	22.4	19.6	21.1	24.3
G_S10	18.2	19.7	20.3	21.2	23.7	20.6	22.2	25.6
G_S11	19.0	19.4	20.0	20.9	23.4	20.3	21.9	25.4
G_S12	19.5	19.5	20.1	21.0	23.5	20.5	22.1	25.5
G_S28	19.1	19.5	20.1	21.0	23.5	20.4	22.0	25.5
1_S14	21.2	19.9	20.5	21.3	23.8	20.8	22.3	25.8
1_S17	21.4	20.1	20.7	21.6	24.1	21.0	22.6	26.1
2_S30	20.5	20.1	20.7	21.6	24.1	21.0	22.6	26.1
2_S31	20.9	19.9	20.5	21.4	23.9	20.8	22.4	26.0
2_S8	21.0	20.0	20.7	21.5	24.0	21.0	22.6	26.1

Table 4. Peak zone temperature (°C) for actual and modelled temperatures for 2019. Modelled for 2050 and 2080. Zone map given in Figure 3. (Colour gradient from green = min to red = max).

Peak Zone Temperature
	2019		2050			2080
	Actual	Model	Low	Med	High	Low	Med	High
External	36.5	28.8	29.8	35.0	37.8	29.9	33.8	42.3
B_S9	25.3	23.4	24.6	25.9	29.2	25.1	27.2	32.1
G_S10	27.0	25.8	27.7	28.7	32.5	28.1	30.0	35.4
G_S11	30.3	26.4	28.8	29.3	33.9	29.3	30.7	37.6
G_S12	29.3	25.8	28.1	28.6	32.8	28.5	30.4	36.9
G_S28	36.3	25.7	27.9	28.5	32.6	28.3	30.3	36.5
1_S14	30.0	27.2	28.7	30.0	33.5	29.2	31.4	37.2
1_S17	30.0	27.5	29.4	30.5	34.1	29.9	31.8	37.7
2_S30	29.4	27.3	29.0	30.3	33.7	29.5	31.6	37.5
2_S31	30.0	27.2	28.9	30.2	33.6	29.4	31.5	37.5
2_S8	31.0	27.5	29.1	30.5	33.8	29.6	31.7	37.7

Table 4. Peak zone temperature (°C) for actual and modelled temperatures for 2019. Modelled for 2050 and 2080. Zone map given in Figure 3. (Colour gradient from green = min to red = max).

Peak Zone Temperature
	2019		2050			2080
	Actual	Model	Low	Med	High	Low	Med	High
External	36.5	28.8	29.8	35.0	37.8	29.9	33.8	42.3
B_S9	25.3	23.4	24.6	25.9	29.2	25.1	27.2	32.1
G_S10	27.0	25.8	27.7	28.7	32.5	28.1	30.0	35.4
G_S11	30.3	26.4	28.8	29.3	33.9	29.3	30.7	37.6
G_S12	29.3	25.8	28.1	28.6	32.8	28.5	30.4	36.9
G_S28	36.3	25.7	27.9	28.5	32.6	28.3	30.3	36.5
1_S14	30.0	27.2	28.7	30.0	33.5	29.2	31.4	37.2
1_S17	30.0	27.5	29.4	30.5	34.1	29.9	31.8	37.7
2_S30	29.4	27.3	29.0	30.3	33.7	29.5	31.6	37.5
2_S31	30.0	27.2	28.9	30.2	33.6	29.4	31.5	37.5
2_S8	31.0	27.5	29.1	30.5	33.8	29.6	31.7	37.7

Table 5. Overheating hours (% > 25 °C) for actual and modelled temperatures for 2019. Modelled for 2050 and 2080. Zone map given in Figure 3. (Colour gradient from green = min to red = max).

% Hours > 25 °C
	2019		2050			2080
	Actual	Model	Low	Med	High	Low	Med	High
External	1.2%	0.6%	2.0%	3.0%	11.7%	2.3%	5.8%	19.2%
B_S9	0.1%	0.0%	0.0%	4.3%	34.4%	0.2%	20.9%	46.9%
G_S10	0.4%	1.2%	11.0%	22.4%	43.1%	14.3%	33.7%	53.2%
G_S11	1.5%	0.9%	6.1%	16.0%	41.9%	9.0%	31.1%	52.3%
G_S12	1.7%	0.7%	6.7%	18.2%	42.6%	10.0%	33.0%	52.9%
G_S28	2.2%	0.7%	6.9%	18.6%	42.4%	10.3%	32.8%	52.7%
1_S14	6.8%	7.9%	20.7%	27.0%	42.5%	22.9%	34.9%	52.2%
1_S17	8.4%	9.0%	22.0%	28.3%	43.9%	24.4%	36.9%	53.6%
2_S30	9.2%	11.6%	23.9%	28.9%	44.3%	25.8%	37.9%	53.1%
2_S31	11.4%	9.9%	22.4%	28.0%	43.2%	24.6%	36.6%	52.4%
2_S8	6.2%	11.0%	23.3%	28.8%	44.0%	25.4%	37.5%	52.9%

Table 5. Overheating hours (% > 25 °C) for actual and modelled temperatures for 2019. Modelled for 2050 and 2080. Zone map given in Figure 3. (Colour gradient from green = min to red = max).

% Hours > 25 °C
	2019		2050			2080
	Actual	Model	Low	Med	High	Low	Med	High
External	1.2%	0.6%	2.0%	3.0%	11.7%	2.3%	5.8%	19.2%
B_S9	0.1%	0.0%	0.0%	4.3%	34.4%	0.2%	20.9%	46.9%
G_S10	0.4%	1.2%	11.0%	22.4%	43.1%	14.3%	33.7%	53.2%
G_S11	1.5%	0.9%	6.1%	16.0%	41.9%	9.0%	31.1%	52.3%
G_S12	1.7%	0.7%	6.7%	18.2%	42.6%	10.0%	33.0%	52.9%
G_S28	2.2%	0.7%	6.9%	18.6%	42.4%	10.3%	32.8%	52.7%
1_S14	6.8%	7.9%	20.7%	27.0%	42.5%	22.9%	34.9%	52.2%
1_S17	8.4%	9.0%	22.0%	28.3%	43.9%	24.4%	36.9%	53.6%
2_S30	9.2%	11.6%	23.9%	28.9%	44.3%	25.8%	37.9%	53.1%
2_S31	11.4%	9.9%	22.4%	28.0%	43.2%	24.6%	36.6%	52.4%
2_S8	6.2%	11.0%	23.3%	28.8%	44.0%	25.4%	37.5%	52.9%

The two sets of results in the 2019 column, thus, represent a comparison between the measured and modelled temperatures of the zones indicated, during times when the Tower was occupied as usual. The year 2020 was not included as a complete year for this summary due to the lockdown in March.

A comparison of the 2019 and 2020 measured data shows that there is nearly 2 °C difference in the average temperatures of each space. Due to the similarities in the external weather data from 2019 to 2020, the difference in internal temperatures can be almost entirely attributed to changes in occupancy.

The 2050 and 2080 columns show the breakdown of temperatures modelled under the low, medium, and high emissions scenarios described above. As noted, these scenarios cannot be directly translated to a particular temperature change, but the low emissions scenario broadly corresponds to limiting to 1.5 °C of climate change; medium emissions corresponds broadly to ~3 °C; and high corresponds to over 4 °C of climate change by 2080.

For all zones, there is close agreement between the 2019 Actual and 2019 Model results for both the average zone temperature and the % of hours above 25 °C. There is a slightly larger variation between the 2019 Actual and 2019 Model results for the peak temperatures. This highlights the important reality that real peak temperatures are increasing beyond what typical weather files are assuming. The fact that there is minimal difference between the average actual and modelled temperatures suggests that in this case peaks are short in duration and their impact is small when averaged across an entire year.

In Table 4, the difference between the actual (36.5 °C) and modelled (28.8 °C) peak external temperatures is 7.7 °C, whereas the difference between the actual and modelled peak internal temperatures varies from 1.2 °C to 3.9 °C (for all zones except G_S28, discussed below). This suggests that the peak of 36.5 °C, while considerable for most buildings, was largely absorbed by the high thermal mass of the Tower.

The only zone with an exceptionally high peak temperature in 2019 Actual was S28. The sensor for this zone is placed near a window and is the only sensor for which there is potential to be exposed to direct sunlight during certain hours of the day. The fact that the average measured and modelled temperatures for this zone are nearly identical shows that this effect is minimal over the course of the day. The difference between the actual and modelled peak temperature for G_S28 can, therefore, be disregarded as an artefact of the measurement rather than an actual feature in the building’s comfort.

The close agreement between the 2019 Actual and 2019 Model across all other zones indicates the thermal model has been suitably calibrated to the observed data. The temperatures observed in the future 2050 and 2080 future weather files will, therefore, be reasonably extrapolated from this baseline and the current conditions in the Tower.

The basement zone S9 is the lowest in temperature due to it being largely below grade and exposed to ground. However, even this zone becomes problematically hot in medium to high emission scenarios.

For all other zones, the average temperature in the space is not strongly affected in the low emission scenario. In considering the peak temperatures, the difference between the external peaks and the internal peaks experienced by each zone is somewhat muted in the 2019 Model case due to the high thermal mass of the building. However, this gap narrows as the emissions increase due to the higher frequency and more prolonged nature of these peaks.

This is better illustrated by comparing the average temperatures to the percentage of hours over 25 °C. The average temperature barely changes for any zone from the 2019 Model, the 2050 Low, and the 2080 Low cases. However, the percentage of hours over 25 °C effectively doubles. The low emission case effectively assumes we cap emissions in line with the UN’s 1.5 °C targets. This means that even though, on average throughout the year, the Tower will be experienced largely unchanged, even in this best case ~1.5 °C scenario, the amount of overheating in the Tower will approximately double due to the prolonged nature of warm temperature events. This finding will be particularly sensitive to the duration of the peak events, and for this reason, further research will follow up on this finding by exploring this model’s findings with a series of weather files that more specifically model a range of durations for heat wave events.

Under the high emissions scenario, more than 50% of hours will exceed 25 °C. This means the Tower would require considerable adjustments to its operation to continue to function as a tourist venue.

4.2. Decarbonisation of Heating

As noted above, the heating system for the White Tower is part of a district heat loop served by a central gas boiler. The loop serves a range of other buildings, most of which do not have the high occupancy of the White Tower and, thus, require higher levels of heating to maintain comfort in winter. As such, based on this thermal model, only a small fraction of the heat from this loop would be required to maintain the indoor temperatures recorded in 2019.

However, since this was not directly metered as part of this study, the actual heating cannot be compared to the modelled heat load. As such, this study will consider future heating loads on a relative basis compared to the 2019 Model baseline heating consumption. In

Table 6. Future White Tower heating loads normalised against current year.

	2019	2050			2080
	Model	Low	Med	High	Low	Med	High
January	62%	64%	38%	22%	57%	34%	10%
February	100%	92%	63%	24%	87%	53%	13%
March	69%	78%	47%	7%	62%	20%	5%
April	17%	18%	10%	3%	15%	4%	0%
May	1%	2%	0%	0%	1%	0%	0%
June	0%	0%	0%	0%	0%	0%	0%
July	0%	0%	0%	0%	0%	0%	0%
August	0%	0%	0%	0%	0%	0%	0%
September	0%	0%	0%	0%	0%	0%	0%
October	3%	3%	1%	0%	1%	0%	0%
November	17%	15%	10%	2%	13%	6%	1%
December	51%	43%	31%	15%	45%	22%	7%
Total Heating Load (kWh) in each case as % of model	100%	99%	62%	23%	88%	44%	11%
Peak Heating Load (kW) in each case as % of model	100%	101%	72%	45%	105%	56%	16%

, the total heating consumption for the 2019 Model base case is given as 100% in the first column labelled ‘Model’, and the reduction in heating demand due to climate change in 2050 and 2080 is shown relative to this figure.

The results show that under the low emissions case, there is only a modest reduction in the overall heating consumption of 88% of its current total. The peak consumption actually very slightly increases by 2080.

In the medium emissions scenario, heating consumption decreases to 44% of the current levels by 2080, with the peak at 60% of the current peak heating demand.

In the high emissions scenario, the need for heating is reduced to only 11% of the current levels and requires only 20% of the capacity.

The White Tower’s gas boiler will need to be replaced by a low-carbon heat source in line with the Mayor’s vision for a Net Zero London [21].

Given the relatively low level of heating currently needed by the White Tower, and the reasonable probability that the need for space heating will decrease in the coming years, the Tower should seek a low-carbon heat source that is cost effective and non-disruptive. The current heating is distributed via radiators and hot water pipework under the suspended wooden floors in the tourist areas. There is currently no plant room within the White Tower and any modifications to create distribution pipework to connect a new system would obviously have to be undertaken with great care so as not to disrupt the character of the building.

The first high-level question is whether to continue with the central solution of the heat network or opt for a local building level solution. To continue with the central solution would require replacing the gas boilers serving the heat network with a low carbon heat source. There is a range of options that could do this, including heat pumps; however, each option presents practical challenges for not only the installation but also the distribution of the heat through the existing pipework. This solution would incur a high cost and complexity outside of the Tower but would likely avoid the need for any changes to existing systems inside the tower.

The local solution would require creating space for a new heating plant to feed the existing radiator system or creating a new distribution system, such as local electric or fan heaters. This would essentially take the approach of only partially treating the space for occupied areas.

Future work will investigate and compare these options in more detail, including modelling the impact of a variable pricing tariff and using the mass of the White Tower as a thermal battery to avoid peaks.

5. Discussion

This study created a thermal model of the White Tower at the Tower of London. It used detailed room-by-room temperature sensors to calibrate the model to a weather file that closely matched actual local weather in 2020. It used temperature data recorded during lockdown to calibrate this model to the actual building performance independent of occupancy variables. This gives a detailed insight into the building physics of high thermal mass buildings.

It then tested the calibrated model against future weather files. It firstly found that the actual peak temperatures recorded near the Tower of London in 2020 exceed the peak temperatures predicted for 2050.

The model finds that most of the Tower spaces are currently comfortable throughout the summer months. The high thermal mass of the structure absorbs the peak external temperatures and maintains reasonable comfort conditions throughout the summer.

However, the biggest change in the future weather will not be the peak temperatures but, rather, the duration of these peak events. The model considered three levels of climate change, low, medium, and high. These broadly represent 1.5 °C, 3 °C, or +4 °C of climate change by 2080.

Under the low climate change scenario, the frequency of hours over 25 °C in the tower more than doubles, and many spaces will require mitigating solutions such as increased ventilation or reduced occupancy to cope.

Under the medium and high climate change scenarios, the temperatures throughout the Tower would be above 25 °C for a third to half of the year. This means that the building would be very difficult to occupy without major adjustments to its function.

The critical difference between the current operating mode for the Tower and likely future modes is that presently, the thermal mass of the Tower acts as a buffer to reduce the impacts of peak temperatures. With future climate change, these peaks will endure for longer periods of time and the thermal mass of the Tower will begin to store excess heat that it cannot shed overnight.

The data in Table 5, showing the percentage of hours exceeding comfort temperatures, allows us to explore this for the Tower case study in more detail. It shows that for the 2019 modelled and measured data, the rate of overheating increases higher up in the building. This is because there is more thermal mass protecting the lower floors. However, in the future weather scenarios, the difference between the lower (higher mass) floors and the upper floors is reduced. Under the ‘high’ climate change scenario, there is nearly no difference between the lower and upper floors, suggesting that the buffering impact of the thermal mass has been overwhelmed by the overheating events.

The study also considered the impact of future climate change on the building heating loads. The heating loads are not directly metered but inferred as a fraction of the district heating circuit and modelled on a relative basis to the 2020 values. The model shows that under the low climate change scenario, the peak heating loads will not significantly change, and the annual heating consumption will decrease by around 10% by 2050. Under the medium and high climate change scenarios, both the peak and annual heating loads decrease significantly.

6. Conclusions

This study contributes to the landscape of specific case studies characterising the future performance of heritage buildings under different climate scenarios. It is consistent with the literature in finding that there will likely be a need for mechanical cooling or changes in use to maintain comfort. However, the extent of the changes needed was very sensitive to the nature of the climate change experienced. In particular, the duration of heat waves played a critical role in creating tipping points beyond which the thermal mass of the building was no longer able to shed excess gains and instead acted as a thermal battery.

Future work should explore the mitigation options in more detail and determine what, if any, passive solutions could allow the Tower to continue to function as it does presently. The study could be validated by further investigations of similarly high-thermal-mass buildings. The Tower of London represents an extreme case for studying thermal mass, with walls several meters thick. However, most studies of this topic consider roughly one quarter of the building stock across Europe to be ‘high thermal mass’ compared to modern buildings. Studies of historic buildings are, therefore, not one-offs, but useful data points on a very significant fraction of the built environment.

In extreme warming scenarios, mechanical cooling will be inevitable for some buildings to maintain comfort and safety. However, good passive design must still minimise the need for mechanical cooling in use. A more granular understanding of the limits of passive measures under prolonged heat waves will be essential to strike the right balance between passive and active cooling solutions.

Finally, there was an interesting trade-off across seasons. Climate warming inevitably reduces heating degree days and retrofit decisions today must not only balance loads across seasons but also consider future use. Here, again, the results were extremely sensitive to different climate scenarios, with the Tower’s heating demand in 2050 dropping to anywhere from 23% to 88% of today’s values. Building services engineering has historically been risk averse in its approach to heating system sizing, leading to oversized systems. Building services as a sector, must adopt a more flexible approach to passive design and managing loads across seasons, or we risk creating the worst of both worlds, over-designing both heating and cooling systems that are not optimised for any climate scenario.