THE GREAT UK RAW SEWAGE SCANDAL

1. A dataset containing over 16,000 geocoded storm overflow site locations in England and Wales provided by The Rivers Trust was collected.
   1. The dataset contained metrics for overflow sites with EDM (Event Duration Monitoring) data, including the number of counted spills (12-24h count method), total spill hours, and the local authority within which the site was located.
2. The Met Office HadUK-Grid dataset was used to acquire daily gridded precipitation observations on a 1km grid over the UK.
   1. A spatial buffer of 5km around each overflow site was used to find the number of heavy rain occurrences in proximity to the overflow site.
      1. A heavy rain occurrence was defined as any day with 25mm or more of precipitation in any given area.
3. The EDM storm overflows data acquired from the river trust and precipitation observations from the Met Office were subsequently aggregated by local authority to calculate the total number of overflow sites, total counted spills (12-24hr count method), total duration of spills, average number of heavy rain days and average number of spills per heavy rain day.
   1. The 12-24hr spill count method is defined by gov.uk as follows:
      1. "Any discharge (or discharges) in the first 12-hour block are counted as one spill. Any discharge (or discharges) in the next and subsequent 24-hour blocks are each counted as one additional spill per block. Continue counting until there is a 24-hour block with no discharge"
   2. The average number of heavy rain days is calculated as follows:
      1. For each day, capture only rain within 5km of overflow site
      2. Filter out any instances of rain that are less than 25 mm
      3. Count the number of unique days that have had heavy rain for each overflow site. E.g., if for overflow site A, Jan 1st had two instances of 25mm of rain, and Jan 2nd had one instance of 25mm of rain, and the rest of the year had no rain, the number of unique days is two (as there are two days that had any heavy rain).
      4. Finally, average the number of unique rain days by local authority
   3. The average number of spills per heavy rain day is averaged for each individual overflow location, which is then aggregated by local authority (average) and will not be the same as (total spills/ heavy rain days).
4. Bathing water quality samples between 2017 - 2023 were collected from the Environment Agency and subsequently used to identify the average annual 95th percentile of both Intestinal Enterococci (IE) and Escherichia Coli (E. coli) traces (cfu/100 mL) in all bathing waters within 5km of an overflow site. Both IE and E. Coli are bacteria found in intestinal tracts of humans and are likely remnants of faecal matter.
   1. Average count of each of these factors are calculated as follows:
      1. For each overflow site, the 95th percentile of samples in that year/year group was obtained
      2. Then, the average of these samples was calculated for each local authority/year. Therefore, this is the average of the 95th percentiles.
   2. It must be noted that multiple factors contribute towards bacterial counts in bathing waters, including animal discharges. While filtering bathing waters to those within 5km of an overflow site helps narrow the likely cause of changes in bacterial counts, it may not be the sole contributing factor.
5. The 3-year annual 95th percentile average of EC and IE counts calculated in the previous step were then averaged by local authority to calculate the percentage change in both EC and IE counts between the periods 2018-2020 and 2021-2023. A normalised degradation score is subsequently calculated considering the percentage changes in both EC and IE counts. An additional bathing water quality score is calculated using the average annual 95th percentile of bacterial counts in the period 2020-2022.
   1. A yearly average from 2018-2023 was also calculated to show year on year how the samples change. The average sample size for each year/area is 87 samples, with the lowest average sample for an area is equal to 12 samples taken.
   2. For each year, the % change from the past year was calculated. These year-on-year % changes were then combined as a weighted average using the following weights:
      1. 2018/19 - 0.15
      2. 2019/20 - 0.15
      3. 2020/21 - 0.2
      4. 2021/22 - 0.2
      5. 2022/23 - 0.3
   3. This final weighted average % change shows how the bathing water quality has changed over the years, with a bias towards the past 3 years (as those are more important to today's water quality).
6. By using percentile thresholds from environment.data.gov.uk, the average percentile by local authority/district were evaluated.
   1. Most of the bathing waters in the UK are coastal, so only non-coastal districts (i.e. local authorities that do not have a border with the ocean) were evaluated with the inland bathing water thresholds, these are highlighted red in the data.
   2. The classifications were then converted to a numerical score in order to rank each local authority/district:
      1. Poor - 1
      2. Sufficient - 2
      3. Good - 3
      4. Excellent - 4
   3. These scores were then averaged to provide an overall look at the bathing water quality of each area.
7. EDM data detailing long-term trends was acquired from data.gov. Statistics provided in the dataset were used to provide breakdowns by water company comparing the number of storm overflows, average spills per storm overflow, average duration (hrs) per monitored spill and total duration of spill events monitored.
8. The long-term EDM trends dataset was also supplemented with bathing water quality samples to provide, for each year between 2017-2023, the average annual 95th percentile of EC and IE counts in bathing waters together with average spill duration (hrs) and the average number of spill events per overflow for the years (2017- 2022). The percentage changes in the aforementioned metrics were calculated as the average between 2018-2020 compared to the average of the remaining periods.

Bathing waters data (no date) Gov.uk. Available at:
https://environment.data.gov.uk/bwq/profiles/data.html
(Accessed: March 4, 2024).

CEDA Archive web browser (no date) Ceda.ac.uk. Available at:
https://data.ceda.ac.uk/badc/ukmo-hadobs/data/insitu/MOHC/HadOBS/HadUK-Grid/v1.2.0.ceda/1km/rainfall/day/v20230328
(Accessed: March 4, 2024).

Environment Agency (2024) “Event Duration Monitoring - storm overflows - annual returns.”

Event duration monitoring - storm overflows - 2021 (England and Wales) (no date) Catchmentbasedapproach.org. Available at:
https://data.catchmentbasedapproach.org/datasets/
2cf731d1782f4a899c317be83f221226_11/explore?location=>52.716396%2C-3.144202%2C7.00
(Accessed: March 4, 2024).

Event duration monitoring - storm overflows - 2022 (England and Wales) (no date) Catchmentbasedapproach.org. Available at:
https://data.catchmentbasedapproach.org/
datasets/cd83fb0f4b084557a0537fbbd7e20153_0/explore
(Accessed: March 4, 2024).

Georgiou, S. (2005) “Bathing water quality,” pp. 75–101. doi: 10.1007/3-540-27150-3_4.

Met Office et al. (2023) “HadUK-Grid Gridded Climate Observations on a 1km grid over the UK, v1.2.0.ceda (1836-2022).” NERC EDS Centre for Environmental Data Analysis.

• Only EDM sites containing local authority labels provided by the river trust were analysed for local authority breakdowns.
• Kensington and Chelsea has been omitted from our bathing water quality analysis as notes provided by the Environment Agency state animal faecal release as the likely cause of high bacterial counts.