35 Years of Sewage in a Single Year
Playing with Southern Water's published 2024 spill data and national rainfall records. An AI-assisted evening project, not an investigation.
About this piece. I watched a television programme about Southern Water and wanted to see what the numbers on their website actually showed. This page is me exploring that dataset with AI help: charts, copy, layout.
Every figure below comes from discharge data Southern Water publishes on its website. Met Office rainfall is used only to compare timing and volume. I am not alleging wrongdoing or offering regulatory analysis.
The scale of published discharge
Combined Sewer Overflows exist because the alternative is worse. When heavy rainfall overwhelms a network, untreated waste needs somewhere to go. CSOs are the pressure-release valve, and their existence is legal.
What is harder to read from published tables alone is how activation thresholds apply in practice, how completely the reported data reflects conditions on the ground, and how a non-specialist is supposed to make sense of it all.
What follows is my attempt to read Southern Water's published figures against the rainfall record for the same year. The patterns below are in the spreadsheets. I am not drawing conclusions about intent, compliance, or blame.
- 12,142
- overflow events in the published record
- 886
- unique overflow sites across seven counties
- 48
- sites with discharge logged in every month
Monthly discharge against rainfall
Bars show total spill count. Line shows rainfall in mm. June stands out: the driest month of 2024, yet 147 spills appear in the published data. I cannot say why from these fields alone.
Source: Southern Water EDM 2024 / Met Office National Rainfall. February's high discharge hours (110,622) are partly related to the January 1st timestamp pattern described in section 02.
Forty-eight sites show discharge in every month of 2024, including June, the driest month in the dataset. Whether that reflects weather, infrastructure, reporting conventions, or something else, I cannot tell from published tables alone. It is a pattern that raised questions for me as a reader.
Questions the spreadsheets raised
Five patterns I noticed in Southern Water's published figures, cross-referenced with rainfall. Starting points for curiosity, not findings of fault.
68 overflow events are logged as beginning at 00:00:00 on New Year's Day, which looks like a monitoring-system reset artefact. The true start dates are not shown in the published fields. Duration is counted from 1 January in the record. One example: Lavant No.1 SSO (West Sussex) runs continuously to 28 August — one declared event, 242 reported spills, 5,772 hours in the dataset. The Lavant is a winterbourne chalk stream. I cannot say from these tables alone when discharge actually began or how the site should be read against regulatory thresholds.
Timestamp artefact — start dates unclear in published dataJune 2024 was the driest month in the rainfall record I used, at 22% of its long-term average. The published discharge data still lists 147 spills for that month. Across Kent, East Sussex, West Sussex, Hampshire and the Isle of Wight, 125 events appear. Some lasted under a minute; one in Kent runs for nearly 24 hours in the record. June 15th alone shows 71 spills, with no matching exceptional rainfall in the national dataset for most affected areas. I do not know the operational context behind those entries.
Low-rainfall month — mismatch between rainfall and published spillsFebruary received 87.1mm of rainfall. September received 120.1mm, 40% more. Yet February shows 110,622 hours of discharge against September's 20,052. Rainfall volume alone does not explain that gap. Plausible readings from the published fields: February totals may be inflated by the January 1st timestamp pattern (many events ending in February–March), or winter infrastructure was already under strain at the start of the year. I cannot tell which from the spreadsheet.
Seasonal contrast — rainfall does not explain the gap on its ownThe dataset separates Discharge Duration (actual sewage flow time) from Discharge Period (monitoring window). The gap can be large. Court Road Lewes CEO lists 35 spills in 42 hours of actual discharge, inside a declared monitoring window of 813 hours. Cross Lane East Gravesend lists 9 spills in under 5 hours within a 188-hour period. Without raw sensor logs, a reader cannot tell whether that reflects sensor behaviour, grouped intermittent events, or how spills are counted within a window. The published fields alone leave that open.
Field definitions — duration vs monitoring window often divergeRainfall does not follow a weekly schedule, yet Monday shows nearly double Tuesday's spill count (6,994 vs 4,202) in the published data. That pattern does not line up with anything obvious in the 2024 rainfall record. Plausible explanations might include batch reporting, monitoring schedules, or operational rhythms — but those are guesses. I have no visibility into how events are recorded day to day.
Day-of-week pattern — cause not clear from published fieldsDischarge intensity per millimetre of rainfall
Dividing discharge hours by rainfall to compare months on a like-for-like basis. January and February sit high, partly because of the January 1st timestamp pattern.
January's figure (1,293 hrs/mm) partly reflects the 68 events with Jan 1 start timestamps. Even setting those aside, winter months still read differently from summer in this ratio.
County by county
Southern Water operates across seven counties. The distribution of discharge hours in the published data varies widely. Some of that likely reflects geography and network size; some of it I simply cannot interpret without operational context.
Bubble area = number of active overflow sites. East Sussex and West Sussex carry the heaviest total discharge burden in the dataset; Kent has the largest number of individual sites. Isle of Wight shown separately: as an island, it cannot route overflow to adjacent catchments.
Highest-hour sites by county
| County | Total hours | Highest-hour site | Spills | Hours | Scale |
|---|---|---|---|---|---|
| East Sussex | 81,309 | Ham Lane Lewes New SSO | 259 | 3,449 | |
| West Sussex | 77,301 | Lavant No.1 SSO | 287 | 6,830 | |
| Hampshire | 57,792 | Kings Somborne SSO | 252 | 5,482 | |
| Kent | 50,757 | Hawkhurst North SSO | 213 | 4,085 | |
| Isle of Wight | 28,413 | Sandown New No.1 SSO | 155 | 2,328 | |
| Surrey | 8,029 | Lingfield SSO | — | — | |
| Wiltshire | 5,196 | East Grimstead SSO | — | — |
Cowes area overflow sites
Fourteen active overflow sites across Cowes and East Cowes. Bar length represents total discharge hours in the published record. Egypt Point CEO discharges directly to the Solent.
Egypt Point Cowes CEO discharged on 27 separate occasions in 2024, including summer months. The Solent is classified as a sensitive receiving water under the Urban Wastewater Treatment Directive.
The discharge data is public on Southern Water's website. Reading it as a non-specialist is harder than I expected. Field definitions, timestamp quirks, and the gap between rainfall and recorded events all leave room for questions I cannot answer from the spreadsheet alone.
What the public dataset cannot tell you
There are limits to what can be established from the discharge dataset alone. It does not include volume, so litres entering each watercourse cannot be calculated from these fields. It does not include water quality results or raw sensor logs that would allow comparison with declared timestamps.
What it does show is that the published format has quirks a casual reader needs to account for. The 1 January timestamp pattern means any reading of 2024 figures may include events whose real start predates the year. The spill count field is defined per event record, which makes site-to-site comparison less straightforward than the headline numbers suggest.
Monitoring complex infrastructure across hundreds of sites is genuinely difficult. I am not claiming bad faith — only that a curious reader working from public tables will hit limits quickly, and that those limits shape what anyone can fairly infer.
If you live in the South East and want to look yourself, Southern Water publishes EDM data on its website and the Environment Agency hosts a searchable portal by site. I am not offering regulatory advice; this page is my own evening with the spreadsheets.
The data is public. How each of us reads it is up to us.
Data note. Discharge figures come from storm overflow and emergency overflow datasets Southern Water publishes on its website (2024 EDM). National rainfall data (Met Office Hadley Centre) is used only for month-by-month comparison. All discharge figures are as reported in those public files.
The January 1st timestamp pattern affects 68 entries totalling 36,657 hours of declared discharge, included in totals as published and flagged here as potentially unreliable for timing analysis.
The ‘spills’ count in Southern Water’s dataset is a declared count within each event record; individual overflow events may contain multiple reported spills.