Welsh Index of Multiple Deprivation (WIMD) 2025 technical report - Physical environment domain

The purpose of this domain is to measure factors in the local area that may impact on the wellbeing or quality of life of those living in an area. The domain has a relative weight of 5% in the overall index.

Air quality score

The air quality sub-domain comprises three indicators based on the population weighted average concentration values of the following key pollutants:

• Nitrogen dioxide (NO₂)
• Particulates less than 10 µm in diameter (PM₁₀)
• Particulates less than 2.5 µm in diameter (PM_2.5)

NO₂ is a gas that is mainly produced during the combustion of fossil fuels with a variety of negative health and environmental impacts.

Particulate matter (PM) is everything in the air that is not a gas and therefore consists of a huge variety of chemical compounds and materials. Some of these can be toxic. Depending on particulate size, they can enter the blood stream and become lodged in vital organs such as the heart or brain. 

More information on why NO₂ and PM are measured can be found in their respective reports in DEFRA’s air quality statistical releases (GOV.UK). 

The indicators in the air quality sub-domain are created using measurements of pollutants that could have negative effects on human health and/or the environment, based on the best medical and scientific understanding, and are proposed as a proxy measure of the quality of the surrounding environment. Poor air quality suggests proximity to certain activities such as traffic, domestic combustion and industrial sites: activities that could have a negative impact on quality of life, the local environment and health. 

Type of indicator

The three separate indicators that form the air quality sub-domain are:

• Population weighted average concentration value of NO₂
• Population weighted average concentration value of PM₁₀
• Population weighted average concentration value of PM_2.5

The average concentration values are measured in micrograms per cubic metre (µg/m³).

Numerator

N/A

Denominator

N/A

Source and time period

The sources used to form the air quality indicators are:

• pollutants data: Department for Environment, Food & Rural Affairs (DEFRA), 2023
• Small Area Population Estimates (SAPE): Office for National Statistics, 2022
• dwellings data: OS AddressBase, 2025 

Additional notes

Each year the UK Government’s Pollution Climate Mapping (PCM) model calculates average pollutant concentrations for each square kilometre of the UK. The model is calibrated against measurements taken from the UK’s national air quality monitoring network.

This data is combined with small area population estimates and dwelling data to provide population weighted data.

For each census output area (statistical geographic units comprising around 150 properties), the pollutant concentrations are weighted by the number of dwellings in each square kilometre to give an average NO₂, PM_2.5 and PM₁₀ concentration across the census output area.

For each Lower layer Super Output Area (LSOA), a population-weighted average over its constituent census output areas were calculated to give an average NO₂, PM_2.5 and PM₁₀ concentration.

Comparability with WIMD 2019

Broadly comparable. From the 2022 update of the air quality national well-being indicators onwards, a methodological improvement was implemented to the way in which the dwelling weights are calculated, as the original process used for estimating the air quality indicators (prior to 2022, including WIMD 2019 indicators) was not calculating the weights in the way intended. An assessment of the impact on the historic data has been undertaken and the impact is small. Given that the air pollution data is modelled and the population estimates are subject to rebasing following the Census, there is existing uncertainty associated with these estimates. Due to this uncertainty, the lack of detailed historic dwelling data and the small impact of the methodological change, the historic data has not been revised.

Flood risk score

Type of indicator

The flood risk indicator considers the proportion of households at risk of flooding from rivers, the sea and surface water flooding but it does not account for flood defences. A flood risk score between 1 and 100 is generated for each LSOA (see below for further information).

Numerator

N/A

Denominator 

N/A

Source and time period 

Flood Risk Assessment Wales (FRAW) data, Natural Resources Wales (NRW), 2025

Additional notes

The flood risk indicator considers the proportion of households at risk of flooding from rivers, the sea and surface water flooding but it does not account for flood defences. The risk is based on predicted frequency, rather than the level of damage caused by flooding.

The risk categories used are as follows:

• low risk - less than 1 in 100 (1%) chance in any given year
• medium risk - less than 1 in 30 (3.3%) but greater than or equal to 1 in 100 (1%) chance in any given year
• high risk - greater than or equal to 1 in 30 (3.3%) chance of flooding in any given year 

To ensure the areas at risk of more severe flooding rank as more deprived than areas at risk of less severe flooding, the following weighting was given: 

• the number of households in an area at high risk was multiplied by 24
• the number of households in an area at medium risk was multiplied by 4
• the number of households in an area at low risk was multiplied by 1

More information on the methodology for deriving the above weights is available at annex 9.1.

Note that, in cases where households were at different levels of risk from different types of flooding, the highest risk level was given priority. Each of these numbers is calculated for each LSOA and then added together to give total normalised number of households at a risk of flooding per LSOA. This number is then divided by the total number of households in the LSOA to give the proportion of households at risk of flooding. These values are then ranked and exponentially transformed to produce an overall flood risk score.

Comparability with WIMD 2019

Broadly comparable.

Proximity to accessible natural green space

Type of indicator

Proximity to accessible natural green space is the percentage of households in each LSOA that are within 300 metres (an approximate 5 minute walk) of an accessible natural green space.

Numerator

Counts of residential dwellings at an LSOA level within 300 metres of an accessible, natural green space.

Denominator 

Number of residential dwellings in LSOA.

Source and time period

The following sources were used to form this indicator:

• properties in scope: Ordnance Survey's National Geographic Database (NGD) Built Address, September 2025
• OS MasterMap Topography Layer®, September 2025
• OS Open Greenspace, September 2025
• National Trails: NRW, accessed October 2025 (publication date 18 June 2025)
• NRW Open Access, accessed September 2025
  • Open Country
  • Other Statutory Access Land
  • Registered Common Land
  • Other Dedicated Land
  • Dedicated Forests

Additional notes

This indicator measures the proportion of households in each LSOA that are within 300 metres (an approximate 5 minute walk) of an accessible natural green space. To calculate this indicator, Greenspace footprints were derived from OS MasterMap Topography Layer®, scope defined by OS Open Greenspace combined with NRW’s recognised natural greenspace typologies, National Trails and typologies containing Open Access from the Countryside Rights of Way Act 2000. The output highlights sites that could confidently be described as natural feeling places to which the public have right of access. Sites such as golf courses, allotments and cemeteries were excluded from the list. To approximate a 5 minute walk, polygons with a radius of 300 metres around the green space sites were created. All in scope residential dwellings (sourced from OS NGD Built Address) were then intersected with these polygons and flagged if they were within 300 metres of an accessible, natural green space. Counts of dwellings were then aggregated to an LSOA level to calculate the proportion of dwellings within 300 metres of an accessible, natural green space. Further detail on the derivation of this indicator can be found in annex 9.2.

Comparability with WIMD 2019

Somewhat comparable. This indicator is an updated version of the 2019 indicator and now includes the addition of National Trails.

Ambient green space score

Type of indicator

This indicator measures the ambient greenness within each LSOA. A Mean Normalised Difference Vegetation Index (NDVI) is generated for each LSOA.

Numerator

N/A

Denominator 

N/A

Source and time period

The indicator was generated by Population Data Science, Swansea University. To generate the indicator the following sources were used:

• NDVI values for this indicator, 3m surface reflectance satellite imagery from Planet, 2024
• residential dwellings were sourced from AddressBase® Plus, 2025

Additional notes

This indicator measures the ambient greenness within each LSOA. It is calculated as the NDVI within a 300 metre Euclidean buffer around each residential dwelling. Euclidean buffers are not bound to the boundary of an LSOA as it is recognised that an LSOA’s geography does not necessarily represent human behaviour. This removes any edging effect whereby a household may be located within a particular LSOA but their Euclidean buffer overlaps another. The NDVI calculates the normalised difference between the red and infrared bands for quantitative and standardised measurement of vegetation presence and health. Healthy vegetation has a high reflectance of Near-Infrared wavelengths and greater absorption of red wavelengths due to a greater chlorophyll composition. The Near-Infrared (NIR) and red (R) spectral channels of an image are used to calculate an index value, using the following equation:

NDVI=(NIR-R)

(NIR+R)

To calculate NDVI values for this indicator, we sourced 3m surface reflectance satellite imagery from Planet. Residential dwellings were sourced from AddressBase® Plus.

Comparability with WIMD 2019

Somewhat comparable. In 2019 NDVI scores were generated using aerial photography (with a 50 cm resolution), for WIMD 2025 satellite imagery (with a 3m resolution) was used. Therefore, we would expect some natural variation as a result of the change in image source.

Noise pollution

Type of indicator

Proportion of population exposed to a combined road and rail L_den greater than or equal to 55dB. L_den stands for Day-Evening-Night Level, and is used to describe average noise exposure over a 24-hour period, with penalties applied for evening and night-time noise.

Numerator

Population within LSOA exposed to a combined road and rail L_den greater than or equal to 55dB.

Denominator

Population of LSOA.

Source and time period 

Round 4 strategic noise maps, Noise Modelling System (NMS), Department for Environment, Food and Rural Affairs (DEFRA). Data relates to 2021.

Population: Census 2021, ONS.

Additional notes

Further detail on the derivation of this indicator can be found in annex 9.3.

Comparability with WIMD 2019

Not comparable as it is a new indicator.

There are 7 indicators split into 4 sub-domains that form the physical environment domain, weighted as follows. The sub-domain weights were agreed with the expert domain group, and 5% of each WIMD 2019 sub-domain weight is allocated to the new noise pollution sub-domain. 

• Air quality, 35%, formed of the following indicators that contribute equally to the sub-domain:
  • Nitrogen dioxide (NO₂)
  • Particulates less than 10 µm (PM₁₀)
  • Particulates less than 2.5 µm (PM_2.5)
• Flood risk, 35%
• Green space, 15%, formed of the following indicators that contribute equally to the sub-domain:
  • Proximity to accessible, natural green space
  • Ambient green space score
• Noise pollution, 15%

The domain has a relative weight of 5% in the overall index.

Air Quality sub-domain

To calculate the overall air quality sub-domain score, each indicator value was adjusted (via transformation) using a factor based on the objective, standard or risk factor for that specific pollutant and statistic. This method was developed to take into account air quality standards for each substance, which are based on the best medical and scientific understanding of their effects on health and/or the environment. The method also ensures that areas which have high prevalence of certain pollutants, but not others, are ranked as highly deprived; low levels of one pollutant will not cancel out the effect of a high level of another pollutant. 

The standards used to normalise the concentrations of these pollutants in WIMD 2025 are the World Health Organization (WHO) final interim targets for each pollutant:

• NO₂: 20µg/m³ annual average concentration
• PM₁₀: 20µg/m³ annual average concentration
• PM_2.5: 10µg/m³ annual average concentration

The transformed indicator values for each LSOA were averaged and ranked. These ranks were then exponentially transformed to produce a sub-domain score for each LSOA.

Green Space sub-domain

To calculate the overall green space sub-domain score, each set of indicator values was ranked and normalised. The normalised values were then combined using the following weighting:

• 50% proximity to accessible, natural green space
• 50% ambient green space score 

The combined values were then re-ranked and exponentially transformed to produce a sub-domain score for each LSOA.

There have been several methodological changes to the physical environment domain between WIMD 2019 and WIMD 2025. A full list of the changes is outlined below.

Noise pollution has been added as a new indicator; it is based on the proportion of the population exposed to noise pollution from road and rail sources greater than or equal to 55dB.

In the proximity to accessible natural green space indicator, national trails have now been included in this indicator, further detail of this is in the indicator section.

For the ambient green space indicator, the data source has changed for WIMD 2025 compared to WIMD 2019. In 2019 NDVI scores were generated using aerial photography (with a 50 cm resolution), for WIMD 2025 satellite imagery (with a 3m resolution) was used.

The weightings of the sub-domains have been updated for WIMD 2025 to account for the addition of the noise pollution indicator. In WIMD 2019 the sub-domains were weighted as follows:

• 40% air quality (3 equally weighted pollutant indicators)
• 40% flood risk
• 20% green space (of which half came from each of ambient green space and proximity to green space)

For WIMD 2025 the sub-domains are weighted as follows:

• 35% air quality (3 equally weighted pollutant indicators)
• 35% flood risk
• 15% green space (of which half came from each of ambient green space and proximity to green space)
• 15% noise pollution

In addition to domain ranks and indicator values, physical environment sub-domain ranks are also published on StatsWales.

The flood risk indicator considers the proportion of households at risk of flooding from rivers, the sea and surface water flooding. The risk is based on predicted frequency rather than the level of damage caused by flooding. The numbers of households at significant risk of flooding are given higher weighting than those at lower risk. Due to variability in the availability of data flood defences have not been taken into account when assigning flood risk categories.

The weighting methodology outlined in this annex was developed in partnership with Natural Resources Wales.

The weighting factors use the below risk categories, the same were used for WIMD 2019 as outlined in the 2019 technical report.

In WIMD 2025, the flood risk indicator is produced from Flooding Risk Assessment Wales (FRAW) data provided by Natural Resources Wales. The FRAW data uses the same risk categories as the NaFRA data (with the exception of the ‘very low’ risk category) and therefore we have used the same flood risk weightings that were derived for WIMD 2019.

Methodology for determining weights

As for the previous two indices, we use information on average flooding damages as produced by the Middlesex University Flood Hazard Research Centre in their Multi Colour Manual. 

The underlying assumption is that the impact on quality of life increases as potential flood damage rises. 

The Weighted Annual Average Damage (WAAD) represents the expected annualised economic damage from flooding, weighted by the probability of different flood events occurring. It is calculated as the area under the curve when plotting flood damage values against exceedance probability (the reciprocal of the return period in years). This approach accounts for both the severity and likelihood of flooding events.

• Damage values are derived from modelled events for specific return periods (e.g., 5, 10, 25, 50, 100 years).
• No differentiation is made between residential property types.
• WAAD calculations are based on properties within Flood Zone 2 (greater than 0.1% annual chance of flooding).
• For extreme events, more properties are affected, but average flood depths tend to remain below 1 metre.

The WAAD Estimation Tool, provided in the Multi-Coloured Manual, is designed to estimate potential benefits from reducing flood risk to properties. These benefits are expressed as economic damages avoided over a standard 50-year appraisal period.

WAAD ratios are then used to derive WIMD weights.

This ensures that areas with higher expected flood damages receive proportionately higher weights in the deprivation index.

Definition of accessible, natural green space

The definition for accessible, natural green space closely follows Natural Resource Wales’ (NRW) Greenspace toolkit.

The data set compiled to calculate the proximity to accessible, natural green space indicator in WIMD 2025 comprises, and builds on, a series of rules for accessibility and naturalness that can confidently be said to be natural-feeling places to which the public have a legal right of access.

NRW and Welsh Government recognise that this definition is not all encompassing and that citizens may perceive many more spaces (including urban parks) as natural. Further, there are many more accessible spaces than are defined by law. Therefore, to enhance this definition, urban and coastal coverage (polygons representing urban and coastal green and blue spaces) have been used to supplement data gaps. Note that these additional polygons are ‘likely’ to be accessible but do not strictly indicate legal right of access.

This definition serves to show all land and water in Wales that are not covered in man-made surfaces and could therefore potentially deliver health or well-being benefits. 

We are aware that this data set excludes some polygons which deliver ecosystem services (e.g. cycle paths, man-made play areas), but because the focus of this data set is on spaces which deliver health and well-being benefits by virtue of the natural nature of their surfaces we have deliberately excluded these.

NRW tested this rule-base by undertaking comparisons with aerial imagery as well as creating maps of familiar urban and rural areas to compare and verify the rule base delivered the intended result.

Data sources

Definition of accessible, natural greenspaces

• All polygons from Ordnance Survey MasterMap Topography Layer® that satisfy the ‘natural’, ‘multiple’ and ‘unknown’ make classification. For these polygons to pass through to the final inclusion stage they must geographically intersect with.
• All polygons detailing Open Access from the Countryside Rights of Way Act 2000. The Act gives a right of access on foot for the purpose of open-air recreation and, in Wales, the right given under the Act commenced in May 2005. Layers included are:
  • Dedicated Forests
  • Other Dedicated Land
  • Open Country
  • Other Statutory Access Land
  • Registered Common Land

Supplementary green spaces to account for urban and coastal accessibility follow the definition of:

• all polygons from Ordnance Survey MasterMap Topography Layer® whose descriptive group satisfies the ‘tidal water’ classification
• all ‘natural’, ‘multiple’ or ‘unknown’ polygons from Ordnance Survey MasterMap Topography Layer® that intersect within the extent of OS Open Greenspace polygons, that adhere to the following typologies:
  • public parks or garden
  • playing field
  • play space
  • other sports facilities

Certain national trails follow roads which are not one of the above descriptions to include. The approach has been taken to buffer the trail by 300 metres and find the properties that intersect this buffer.

Routing methodology

Both network analysis and Euclidean distances (an ‘as the crow flies’ approach) were considered for modelling the travel-time for a residential dwelling’s degree of proximity to accessible natural green spaces. However, whilst network analysis is highly representative of real-world behaviours and the favoured approach, Welsh Government recognise inconsistencies and significant data gaps in both greenspace polygon access point data and a nationally consistent path network. Therefore, applying network analysis would be nationally inconsistent and unfit for WIMD at this present time.

Although Euclidean distances do not take into account real-world obstacles, their production can be nationally consistent and provide a statistically robust method to calculate the areas of accessibility around greenspaces. Therefore, Euclidean distances were used to produce the travel-time polygons around the green space data set for the purposes of WIMD 2019. 

It is widely accepted that residential dwellings are classified as having sufficient access to accessible natural greenspaces if they can reach sites within a 5-minute walk. In a study undertaken by the University of Manchester for the formerly known Countryside Council of Wales, (now NRW) there is clear evidence to show that, generally, citizens are highly unlikely to walk beyond the 5-minute threshold to access local green infrastructure. 

On average, a 5-minute walk roughly covers approximately 400 metres in distance. To account for the shortcomings of a Euclidean distance methodology, time-travel polygons were produced with a 300-metre radius to account for real-world obstacles in green space access.

All in scope residential dwellings (as sourced from OS NGD Built Address) were then intersected with time-travel polygons to flag whether they were within a 5-minute walk to an accessible natural greenspace according to the definition. 

Dwellings were then aggregated to Lower Super Output Area (LSOA) and the percentage of total dwellings within proximity of an accessible natural greenspace calculated.

Scope and background

Noise Consultants Limited (NCL) were appointed by the Welsh Government Environmental Protection Division in 2021 to prepare the Round 4 strategic noise maps and associated noise exposure statistics in Wales. NCL partnered with Mott MacDonald, Oden Systems, Acustica and Stapelfeldt (the ‘Project Team’) to develop the model and provide the required outputs. The maps and models were delivered to Welsh Government in 2023.

The requirement to produce strategic noise maps is due to the Environmental Noise (Wales) Regulations 2006. The Regulations require the following to be mapped:

• major roads (those with over 3 million annual movements)
• major railways (those with over 30,000 annual train passages)
• major airports (those with over 50,000 annual movements)
• sources in agglomerations (industry, road and rail sources)

The Regulations require that noise mapping is undertaken every 5 years. To date, there have been four rounds of strategic noise mapping, with the most recent round (Round 4) based on the situation in 2021. 

The model data and associated results used for the derivation of LSOA noise exposure statistics are taken from the Round 4 strategic noise mapping model and results. The noise mapping results therefore relate to the year 2021.

Following completion of the mapping, Welsh Government contacted NCL seeking to understand the feasibility of providing noise exposure statistics based on the results of the road and rail noise mapping for Wales at LSOA level, to factor into WIMD 2025. 

NCL had already completed and delivered LSOA-level noise exposure statistics to Deprivation.org to factor into the English Indices of Deprivation 2025, therefore to ensure consistency, the approach adopted for the Welsh LSOA noise-exposure statistics was similar.

Derivation of noise exposure statistics

Statistical boundary dataset

To produce the noise exposure statistics at LSOA level, the LSOA boundary dataset was sourced from the Office for National Statistics Open Geography Portal.

Noise metrics

Noise exposure statistics were produced from the noise mapping results for the L_den noise metric. The L_den noise metric, also referred to as the ‘day-evening-night level’, represents the annual average long-term noise over 24 hours, and includes the application of a 5 dB(A) penalty for noise within the evening period (19:00-23:00) and 10 dB(A) penalty for noise within the night time period (23:00-07:00). 

The penalties are applied to account for the increased sensitivity to noise levels within these periods.

The L_den noise metric is widely used in health effect studies, linking long term noise exposure to the risk of ischaemic heart disease (IHD), hypertension, stroke and annoyance. This metric is therefore considered appropriate for the WIMD. More information can be found in the World Health Organization’s Environmental Noise Guidelines for the European Region.

Noise sources included

The LSOA noise exposure statistics uses the Round 4 noise mapping results from road traffic and railway noise sources, where every public road and railway in Wales has been modelled and mapped. Industry noise in Round 4 has been modelled at a relatively high-level in comparison to road and rail, and is only modelled within agglomerations. Industry noise was therefore considered inappropriate for inclusion within the LSOA noise exposure statistics, and combined exposure data is based on the Round 4 road and railway sources only.

Overview of processing steps

This section provides a summary of the processing steps undertaken to generate the LSOA noise exposure statistics. A more detailed overview of the process undertaken to produce the LSOA-level noise exposure statistics is presented in Appendix A1.

Create road and rail building level results

The calculation of noise exposure at building level for road and railway sources requires associating modelling results for building façade receivers with the number of dwellings within buildings, as well as the assignment of the number of people to each dwelling. 

The approach to assigning calculated levels at the façade receivers to dwellings and people in dwellings is set out in CNOSSOS-EU. Three methods are available, summarised below:

Method 1: the location of individual dwellings is known

Where the location of individual dwellings is known (such as with detached, semi-detached, terraced houses, or apartment buildings where the internal division of the buildings is known), the dwelling and number of people within the dwelling is assigned to the façade receiver point at the most exposed façade of the dwelling. 

Method 2: information is available showing that dwellings are arranged within an apartment such that they have a single façade exposed to noise 

Method 2 applies to apartment blocks that have all windows within each apartment only facing one direction. Under this method, the dwellings and people in dwellings are assigned to all façade receivers associated with the building, weighted by the façade length that each façade receiver represents, resulting in the dwellings and number people within the dwellings being assigned the lowest, median, and highest calculated noise levels around the building façade.

Method 3: information is available showing that dwellings are arranged within an apartment building such that they have more than one façade exposed to noise

Method 3 applies to apartment buildings that have dwellings with windows facing more than one direction. This approach is also to be considered the default in situations where the layout of dwellings within a building is unknown. Under this method, the dwellings and people within dwellings are assigned noise levels above the median of all building façade receivers.

It has not been possible to consider Method 2, given that the layout of dwellings within the buildings considered in the Round 4 maps is unknown. Therefore:

• method 1 has been applied to buildings with one dwelling; and
• method 3 has been applied to all other multi-dwelling residential buildings.

Noise levels were assigned to population and dwellings using the methods described above by merging attribute data from the building’s dataset, façade receiver dataset, and all required information within the model results files. The estimated exposure results were output at each building for each representative receptor and each 1 dB noise exposure band for all calculated noise metrics, including L_den.

Assign LSOA code to extracted road and rail results

Spatial analysis was undertaken to identify the LSOA boundary that each building is within, and assign the building with its corresponding LSOA code (‘LSOA21CD’) and name (‘LSOA21NM’).

Calculate the consolidated noise level at the buildings

A consolidated noise level combining the building-level road and rail noise levels was calculated by means of logarithmic summation, as follows:

Derivation of LSOA noise exposure statistics

The processing set out above produced a building dataset where each building was attributed with the following:

• LSOA code and name
• population
• consolidated noise exposure (road and rail combined, L_den)
• road traffic noise exposure (L_den)
• railway noise exposure (L_den)

The attributed data was used to calculate the total population and population exposed to road, rail, and consolidated L_den, and output as a CSV file that included the population exposed to road, rail, and consolidated noise per LSOA in 1 dB bandings from 40 dB up to >=75 dB and the total population as calculated from the building results layer per LSOA.