What demand data actually exists

What was looked for, what was found, what was fetched, and what still needs asking for. First chapter of the demand layer.

The layer opened with a question it could not answer from anything on disk: does anyone record how much water a given place needs at a given time? For the environmental half the answer turns out to be yes, in numbers, and it had simply never been looked for. For the agricultural half it is partly.

1. Environmental — Water Level Management Plans

The answer to “does the EA record water required for certain sites at certain times” is yes. The Somerset Drainage Boards Consortium, with the Environment Agency and Natural England, publishes a Water Level Management Plan per protected moor. Each carries a table of target water levels in m ODN per control structure, summer and winter, with a grid reference for every structure.

Fetched and parsed by core/wlmp.py → projects/somerset/wlmp_levels.csv:

Of these, 51 carry both a summer and a winter level and 50 have a plausible seasonal difference:

The summer level is held a median 0.24 m above the winter level.

That single number replaces an assumption. The rhyne chapter had to assume a held depth; this is the seasonal uplift the system is actually required to maintain, and against the 5.20 km² of mapped rhyne water surface it is:

    5.20 km2 x 0.24 m = 1.25 Mm3 held through the summer
                        over and above the winter condition

Compared with the 3.0 Mm³/yr of irrigation the scheme runs assume, and the 0.64 Mm³/season of evaporation off the same surface, the environmental penning requirement is the same order of magnitude as the agricultural one. It is not a rounding error and it has been absent from every run so far.

The regime, from the plans themselves

• Seasonal switch dates are fixed and consulted on. Summer levels from 1 April, winter levels from 1 December, phased two weeks either side. Natural England must be consulted before moving them by more than two weeks — so the dates are a negotiated constraint, not an operator’s discretion.
• Summer penning serves three stated purposes: wet fences and livestock watering; maintaining the groundwater table through the growing season; and the conservation interest of the watercourses. The first two are agricultural. The summer requirement is not purely an environmental one and should not be counted as such.
• The bird requirement is winter and spring, and it is specified as inundation rather than level. Curry Moor’s plan asks for, in mid winter, “shallow pools up to 25 cm deep covering at least 50% of the majority of the fields” plus “deeper water roosts of at least 60 ha, with water 25 to 75 cm deep”; in early spring, surface water over up to 25% of most fields; tapering through May and June.

That last point matters for a drought scheme more than it first appears:

The two environmental demands sit in opposite seasons. The wildfowl and wader requirement is a winter-spring inundation, when water is abundant and the scheme has nothing to add. The penning requirement is April-November, when water is scarce. Only the summer half is a drought demand — and it is the half that is partly agricultural.

Curry Moor is fed from the Tone, which is the scheme’s own river

The plan states that water is supplied to the Curry Moor area from the River Tone upstream of New Bridge Sluice, through Knapp Inlet and New Bridge Inlet, and that Curry Moor in turn supplies North Moor, Salt Moor and Stan Moor in the summer.

The Tone is the river the scheme’s chosen intake sits on, and Bishops Hull is its largest gauge in the portfolio. So a scheme abstracting from the Tone in winter and returning water in summer is operating on the same channel that already supplies the moors — which is an opportunity for conveyance and a competition for resource, and the model currently represents neither.

The Consortium’s own description confirms the mechanism is already there: the Boards’ watercourses “may also be the primary routes by which a supply of water is brought to field ditches from the main rivers in summer”, and “the direction of the flow of water in the rhyne or watercourse system may be different in summer to winter”.

Caveats, and one of them is structural

• These came from the Internet Archive. The Consortium rebuilt its website in July 2026; /environment/wlmps/ and every WLMP PDF now return 404 and no WLMP page exists in the new navigation. The plans are still statutory; they are simply no longer served.
• They are the last published versions, not necessarily the operative ones — Curry Moor 2011, the rest 2009–10. Levels may have been revised. Confirm with the Parrett IDB or the EA Wessex area team before any of this reaches a design.
• Wet Moor did not parse and 43 of 97 rows lack a usable level pair. The parser reads a PDF text stream, not table geometry. Unparsed rows are counted, not silently dropped.
• The seasonal difference has a wide spread (IQR −0.30 to +0.30 m). Some structures genuinely pen lower in summer or sit open in winter; some of the spread is residual mis-pairing. Treat the median as indicative and read the CSV before quoting a specific structure.
• Each plan carries more than one level table — an existing/proposed comparison and a later one with gauge-board corrections applied — and the parser takes whichever window holds the most structures. Absolute levels can therefore differ by a few centimetres from the table a reader opens by hand. The seasonal difference, which is what this chapter uses, is the same in both.
• Discharge from the moors is tide-governed. The plans say so explicitly for Curry Moor Pumping Station (7.25 cumecs total capacity, “dependent on the intake conditions at the pumps and the water level in the River Tone”). The tidal boundary keeps reappearing.

2. Agriculture — land use

Crop Map of England (CROME) 2024 is fetched by core/landuse.py from the same Defra platform core/constraints.py already uses:

python download_data.py somerset --dataset landuse

320,760 hexagonal cells over 1,333 km² of the study box, in somerset.gpkg layer landuse_crome.

Grass dominates at roughly 720 km² of 1,333 km², which is what the Levels are, and it sets the shape of the demand: this is a grazing and hay landscape, not a high-value irrigated horticulture one. That has a direct bearing on the scheme’s value, because grassland irrigation is worth far less per m³ than potatoes or salad, and the scheme’s benefit case cannot assume the latter.

Three warnings on this dataset:

• Mean classifier confidence is 0.49. CROME is a random-forest classification of Sentinel-1 radar, not a return. Half the cells are at or below even odds. Cells below a threshold should be treated as unknown rather than as their modal class; the fetch keeps the prob field so that choice stays open.
• It is one year. CROME 2024 describes 2024, and cropping rotates. Earlier editions back to 2017 are on the same platform.
• It does not say whether a field is irrigated. Crop type is a proxy for irrigation need, never a record of irrigation practice.
• The full code lookup is Annex A of the CROME specification; only FA01, PG01 and NA01 were confirmed from it here. The AC** codes are arable classes whose individual crops are unresolved.

3. Abstraction licences — found, but read the licence terms

The dataset that answers “what is licensed, and what was actually taken” is the EA’s Water Rights Trading Data (England), which carries National Abstraction Licensing Database records: licence holder, abstraction point, water body, and six years of abstraction returns (2019–2024).

It is not fetched, and deliberately so. Its Open Government Licence carries a restriction to “a limited specified geographical area not exceeding 100 square kilometres and limited to internal use”. The study box is 1,333 km², thirteen times that. Pulling it wholesale would breach the terms.

Two lawful routes, and the choice is yours:

1. Clip to ≤100 km² around a chosen intake and keep it internal — enough to characterise demand near A3 without covering the region.
2. Request the wider extract from the EA, which is the documented route: enquiries@environment-agency.gov.uk for a single site, national.requests@environment-agency.gov.uk for multiple sites.

Licences under 100 m³/d are excluded from the dataset in any case, which will drop a long tail of small agricultural abstractions.

And the standing warning from the README applies most sharply here: the returns say what was taken, and in a drought what was taken is what was available. They are a measure of the shortage, not of the requirement. They bound the counterfactual from below and must never be used as the demand itself.

Status after this chapter

Sources: Water Level Management Plans, Parrett Internal Drainage Board with the Environment Agency and Natural England, via the Internet Archive. Crop Map of England 2024, Rural Payments Agency, OGL v3.

← All notes · More from Aquifer Storage