Department for Environment, Food and Rural Affairs (DEFRA) – Supplementary written evidence (ALN0094)

Submitted by DEFRA following an evidence session with the Minister, Rebecca Pow MP, on 16 May 2023.

Noise

Metrics

The HoL Science and Technology Committee has asked a number of questions around the choice of metrics used to assess the impact of environmental noise on human health. A common understanding of terminology is helpful to achieve this.

The definition of noise is “unwanted sound”. It is important to remember that one person’s sound (e.g. their musical instrument) can be another person’s noise. The impact of that noise will be moderated by non-acoustic factors and the perceived reasonableness of the reason for that sound being generated (e.g. whether they like the sound of the instrument, whether they like the genre of music being played, time of day of practising, duration of practising, regularity of practising and so on). Hence, a noise is a complex response to a sound and a sound only becomes a noise, when there is someone (or some other species) who responds adversely to that sound.

Sound travels through a medium (usually air for human impact, but building structure can also be considered and for some other species, water might be the appropriate medium, etc) as pressure variations measured in Pascals (Pa). The threshold of human hearing is generally accepted to be around 20µPa (0.00002 Pa) and the threshold of pain is around 200Pa. This range of 10million different pressure levels are compressed for convenience using a logarithmic scale (the human ear responds logarithmically). The decibel is the ratio of a sound pressure referenced to the threshold of human hearing, which results in a range of 0-140dB.

Addition and subtraction of sound pressure need to take account of the logarithmic scale. Thus, two identical sources of sound will only add 3dB to the value of a single source. Conversely, to obtain a reduction in sound pressure level of only 3dB requires half of the energy in the source to be removed (e.g. switching off half of the number of a bank of identical generators will only result in a 3dB reduction in the sound pressure level). This contextualises how difficult it can be to engineer sound reduction solutions and therefore early consideration in the design phase of a project is generally more cost-effective than seeking solutions once operational.

Sounds will usually consist of different sound pressure levels at different frequencies, unless it is a pure tone. Frequency is measured in Hertz (Hz), and an example of a pure tone is a single note on a piano. The average healthy human ear can detect frequencies in the range of around 20Hz-20kHz. Frequencies are typically grouped into “octave bands” (c.f. to an octave on a piano) or 1/3-octave bands for analysis. Sometimes one or more tones are prominent in a source (called “tonal”) and this would be more noticeable than a broadband source of the same sound pressure level (amplitude), whose sound energy is distributed more evenly over the audible frequency range.

The human ear has a non-linear response to both amplitude and frequency. The average, healthy response is most sensitive in the mid-frequency range, which corresponds to the range in which speech typically occurs. However, from as early as 30-40yrs old age-related hearing loss can occur starting in the very high end of the frequency range. By age 50-60yrs, the mid-frequency range can be affected. This results in lower frequency sounds becoming more intrusive and discernment of speech becomes harder.

A filter is applied to measured or modelled sound pressure levels to reflect this non-linear response across the frequency range. It is known as “A-weighting”. Once applied, the frequency bands are summed (logarithmically) to obtain the overall A-weighted sound pressure level, denoted by dBA or dB(A). At this point, the sound pressure level is only applicable to the human response and biodiversity assessments are not applicable. However, if a value is reported in dBA, then it is known that the human response to the frequency of the sound has been accounted for. (Note that other weightings are available for very high sound pressure levels or those with a high proportion of low frequency noise, but A-weighting is applicable for the vast majority of the sources under discussion in the inquiry).

For high sound pressure levels, damage can occur to the ear, i.e. hearing loss. These sources are generally considered under the Control of Noise at Work Regulations 2005 and are not often encountered regularly in other environments. The Inquiry has generally discussed sources that whilst still associated with health impacts, are not sufficient to cause damage to the ear.

For all sources, amplitude is important. This can be assessed by looking at the maximum (LAmax) of an event or of a series of events and identifies the “highest amplitude”. “L” represents sound pressure level and “A” denotes that A-weighting has been applied across the frequency range.

The amplitude can be taken in account in planning decisions. It is referenced in documents including BS8233[1], WHO CNG[2], WHO NNG[3] and ProPG[4] (ProPG is referred to in the DLUHC Planning Practice Guidance for Noise) and generally indicates desirable internal levels for various living spaces.

The LAmax can be measured or modelled on a case-by-case basis. For example, HS2 estimates maximum noise levels at a distance from the track in its design documents. It is also regularly considered at individual airports. However, there is no internationally agreed method for strategic country or region wide assessments from road and rail sources using the LAmax metric.

Whilst it is known to contribute to additional awakenings or changes in REM-stage for sleep, long-term exposure is not yet robustly linked to health outcomes. This is an evidence gap and some research is occurring in the UK looking at airport noise, funded by DfT, however, studies of other sources are lacking.

The disadvantage of considering only the LAmax, as can be seen above, is that it doesn’t take account of rise time, duration, intermittency or the number of events in a given period.

The rise time is a measure how quickly the amplitude rises. It can be seen in the two graphs above that the sound source for the top chart has a faster rise time than that on the bottom chart. A faster rise time can generate a “startle” effect. Again, the long-term health impacts of the effect of the rise time are an evidence gap.

Similarly the duration of each individual event can affect the response, however, the evidence linking this response to long-term health outcomes is also insufficiently robust.

Intermittency can be considered as the gaps between events. Again, these are known to affect sleep depending how close together events occur, but the evidence linking any disruption to long-term health effects is insufficiently robust.

Therefore, these characteristics are considered on a case-by-case basis, but not in strategic country or region wide assessments. For example, BS4142[5] uses penalties for sources assessed as containing impulsive (high rise time) components or characteristics, and statutory nuisance can take account of how often an event occurs.

The LAeq,t is a metric which implicitly takes all of the above into account along with the number of events in a given period. In effect, it is the area under the graph (total energy) in the above figure. “L” represents sound pressure level and “A” denotes that A-weighting has been applied across the frequency range. The “eq” means “equivalent level. The official definition is that it is “the sound pressure level of a steady state sound that has, over a given period, the same energy as that of the fluctuating sound”. The “t” is the time period under consideration and can be anything from a few minutes to a whole day (0700-1900hrs), evening (1900-2300hrs) or night (2300-0700hrs) period. It can be thought of as averaging, but it must be remembered it is logarithmic averaging. Whilst it does result in a lower empirical value than the LAmax, for long-term health effects, the total exposure is of interest and not just one aspect of it. Most of the health effects assessments are not even based on a single 24hr sample, but on an annual average day to take account of seasonal variation in the source (e.g. for road traffic, weekends and school holidays can be quite different to other times in terms of traffic flow). Airports also consider the “summer average” to take account of increased flights compared to other times of the year. This is why it is important to be clear on the time period.

There are internationally agreed methods for modelling road, rail and aircraft LAeq,t (known as CNOSSOS-EU[6] and adapted for the UK). There are also national standards for modelling road (known as CRTN) and rail (known as CRN and adapted for high-speed rail) – these are currently being updated by a BSi committee. Therefore, the LAeq,t is used for strategic national and regional assessments as well as on a case-by-case basis.

Note these methods are not suitable for railways underground or in tunnels, but do take account of the noise as soon as a train emerges from a tunnel or underground. When considering human exposure to air-borne noise within the tunnel infrastructure, the Control of Noise at Work Regulations apply. For dwellings affected by underground railways, vibration needs to be considered, as the sound is not completely air-borne. Measurements can be made for specific stations on a case-by-case basis as each line, rolling stock and platform design differ.

Defra’s Noise Modelling System

Noise modelling has been carried out for a number of years on either a case-by-case basis or under the Environmental Noise Directive (transposed as the Environmental Noise (England) Regulations 2006), which considers large urban areas and major transport sources.

Our model differs in that it considers the whole country. Every public road and railway line (not just major sources or those in large urban areas) has been included. We believe it to be the first of its kind. In addition, it generates results to much lower noise levels than required by the Regulations, LDEN 40dB[7] and Lnight 35dB[8] at the façade of residential receptors (dwellings).

As well as LDEN and Lnight, the model output includes LAeq,16hr (0700-2300), Lday, and Leve. All metrics are A-weighted, taking into account the frequency response of the typical healthy human ear. They are all Leq-based metrics and take into account the number of events and the time of day that they occur.

Road Input Data

The road infrastructure is modelled in a 3-D environment, taking account of road gradient, cuttings, junction elevations, etc. Each section of road is attributed with road surface type, number of vehicles per time period, percentage mix of traffic type (heavy goods vehicles, light goods vehicles, cars), traffic speeds. Individual vehicles cannot be identified such as vehicles with modified exhausts, but instead average noise emission levels based on empirical data are assigned to give each road section an overall emission level.

Rail Input Data

The rail infrastructure is modelled in a 3-D environment, taking account of gradient, cuttings, elevations, etc. Each section of track is attributed with rail conditions, number of vehicles per time period, length of train (number of carriages/trucks to account for number of wheels), train speeds. Average noise emission levels based on empirical data are assigned to give each rail section an overall emission level.

Propagation Model

The terrain between source and receiver is modelled in a 3D environment, taking account of ground cover, buildings, barriers, bridges, etc. Meteorology is considered through long-term average wind direction, temperature and humidity data. Sound attenuation is calculated taking account of distance, screening, reflections and absorption.

Receptors

Every residential dwelling is included in the model. By calculating the noise exposure at all these locations, a comprehensive England-wide resource is available for research and using known epidemiological relationships, the burden of disease attributable to road and rail noise exposure can be calculated. National policy scenario assessments can be undertaken. The calculation results are expected to be published late summer 2023. All calculations are exposures at the building façade, no assessment of the internal noise environment is possible at this time. As the evidence base for long-term health impacts is also linked to external noise levels, this is not considered a disadvantage for strategic and national analysis. Note that for case-by-case assessments, e.g. in planning, internal noise levels and window behaviour (open/closed) can be accounted for.

Use by Local Authorities, Other Government Departments and Other Public Bodies

Data is stored in an open standard format allowing other public bodies to access the model for local assessments and mitigation planning. A portal is available giving them access to the modelling calculation software or data can be imported into other packages if preferred. The functionality to enable this workstream is still under development and we would hope to make it available from the end of the year (end 2023).

Local Authorities (LAs) have been engaged throughout the design and development process and different access routes have been identified for different purposes. At an entry level, LAs can view the output from the model and use it to inform planning decisions, e.g. in the development of a local plan, and also as a trusted source of data to verify values cited by third parties in planning applications. This will provide every LA with a level of noise data not previously available and requires little in the way of training or background knowledge.

For those with either in-house or contracted geospatial services, data can be downloaded and overlaid with other datasets, e.g. air quality, to identify areas of synergy. For those, with some acoustic modelling resource available to them, the option to use the system for scenario evaluation will be available. It is accepted that not all LAs will have this capability, however, most Local Highway Authorities will have access to some form of road modelling expertise, even if it doesn’t sit within the planning team at a LA. Finally, for those with access to their preferred modelling systems, the data can be imported into their software package of choice.

Defra has made the full functionality, research and development available to the Devolved Administrations. So far, the system has been chosen by the Welsh and Scottish Governments, enabling the potential for consistent GB-wide assessments to also be made.

Verification

The three calculation methods supported by the system were each developed through a process which included versification of the model results against measurement campaigns, both for the source emission levels, and the receiver levels including propagation.

The Control of Road Traffic Noise (CRTN) and Control of Railway Noise (CRN) methods have been the basis of noise impact assessments and environmental impact assessments within the planning and development process for over 30 years, and the results have been accepted as the basis of decisions related to local and national infrastructure developments. The CNOSSOS-EU methodology is more recent, but has been the focus of extensive international development and testing over the past 14 years by a wide range of users from academic, public body and commercial organisations.

CRTN and CRN include test cases, and the ISO17534 series sets out test cases for CNOSSOS-EU, and the calculation methodology coded in software is verified against these test cases. Source emission levels are based on verified data. There is no verification to date of the specific implementation of the model as developed by Defra (beyond using verified emission data and ISO17545 verified calculations), however, cost-effective options for this are currently being explored.

Delivery

The model is funded by Defra. Planning for the procurement and design of specification began in 2018, including a stakeholder workshop in 2019. The framework agreement to deliver the model was let via competitive tender to a team led by Noise Consultants Ltd. The supply chain includes academia and a number of SMEs, as well as a large multi-national consultancy.

Annoyance as a Health Impact

The WHO definition of health is “a state of complete mental, physical and social well-being”. In addition, annoyance in this field has a very specific meaning. It is very different from the everyday use of the word (e.g. I’m annoyed that I spilt my coffee). It takes account of the chronic aspect and the “highly” annoyed aspect.

When determining a health impact, the environmental epidemiology community do not rely on a single paper or a single study. This approach is not noise-specific, it is common, for example, for our understanding of how smoking is bad for health, or lead, or asbestos, air pollution, etc and relies on the same principles of piecing together a coherent picture from a large body of evidence, where each study can be considered as a single jigsaw piece.

Specifically for noise, it is difficult to point to a single paper that describes everything in detail, the knowledge is derived from many experimental and observational studies that helped to build that coherent picture.

However, some examples are:

Non-Acoustic Factors

The WHO recognise that in noise annoyance studies, non-acoustic factors may explain up to around one-third of the variance. The International Standards Organisation has recently commenced work on a technical specification to provide a definition, a conceptual framework and a categorisation framework for non-acoustic factors. There is a large participation from the UK on this working group reflecting the importance the UK acoustics community places on learning more about this important issue. This is the first towards standardisation and will improve comparability between studies, further increasing the robustness of research.

A proposed definition of non-acoustic factors is “all factors other than the objective, measured or modelled acoustic parameters which influence the process of perceiving, experiencing and/or understanding an acoustic environment in context, without being part of the causal chain of this process”. Non-acoustic factors can be grouped into three broad categories:

• Personal – strongly linked to an individual. Show stability over time and situation, vary between individuals. Examples include noise sensitivity, coping capacity, perceived control, perceived fear;
• Tangible – properties of the specific environment. Examples include access to green space, quiet façade, location of inhabited space(s), visual modifiers; and
• Psychosocial – shared between individuals of a community. Examples include perceived fairness, perceived community benefit/disbenefit, attitude towards noise authorities.

Non-acoustic factors are implicit in annoyance exposure-response functions and are therefore accounted for in those health impact assessments. However, it is less clear how relevant some of the above non-acoustic factors are in relation to physiological health effects, whereas factors like age and gender may have a larger role to play. As the evidence around non-acoustic factors grows, significant opportunities for both understanding the drivers of annoyance and opening up new possibilities for reducing the heath burden attributable to noise annoyance are envisaged. These opportunities may also be application to self-reported sleep disturbance.

Funding by Defra

Defra is investing around £6m in the development of the noise modelling system over 5 years.

In addition, in FY22/23, ~£200k was invested, including:

• 3 x Improving measurement and prediction techniques for planning and nuisance assessments;
• Improving modelling techniques for road and rail noise impact assessments;
• Benefits of green infrastructure in noise control
• 2 x Improving understanding of a diverse population

Two further projects have been commissioned for the IGCB(N):

• Meta-analysis for stroke and diabetes attributable to long-term environmental noise exposure;
• Developing the approach for valuing the impacts of hypertension.

Light

Existing research

Defra is not currently undertaking any research on the human health effects of artificial light. Clean Catch UK who are funded by Defra, are looking at possible solutions to bycatch reduction by examining existing evidence already available. Artificial deck lighting is thought to particularly attract seabirds at night, causing injury through collision with vessels, bycatch on hooked gear during night setting, and entanglement in lines and nets.

Research conducted in Scotland has shown that puffins and Manx shearwater fledglings are drawn to artificial lights on land, while scientists have found evidence of petrels off both Tristan da Cunha and South Georgia being attracted to fishing and other vessels due to their lights .

In the marine environment research is currently ongoing to measure the impact of artificial light and what the threshold of light exposure is to cause an impact (JPI Oceans joint action ‘Changing Marine Lightscapes’ has been approved for a knowledge hub and joint call, with UK academics, Cefas and Defra representation). The long term aim of this work is to determine what levels of light pollution cause an impact on the marine environment, to inform any consideration of future management approaches to address light pollution.

Published marine research has shown the effect on seabirds, with more collisions with buildings when lights show and higher numbers of fledglings grounding in areas of artificial light (Syposz et al 2018 & 2021).

Feasibility of light mapping

Evidence on artificial light is much less advanced than noise, and there is no current equivalent process for mapping artificial light. As the Committee has already heard, the CPRE conducts some light mapping but there are limitations to the technology and data currently available. For instance, some newer sources of light are less likely to be picked up by satellite imaging, providing an incomplete picture of the extent of artificial light. Further research would be needed to establish appropriate methodology, data sources, metrics and measures.

Other countries

Defra staff will shortly be meeting counterparts in France to learn more about the drivers for, and outcomes of, the artificial light legislation recently enacted there. We will also be looking into measures put in place by other countries across Europe and worldwide.

9 June 2023

10