The demand layer

Who needs how much water, when, and what happens if they do not get it.

AREA MAPPING --> GEOLOGY --> WATER --> [ DEMAND ] --> SCHEME --> [OPTIMISATION]
                                          ▲
                                    you are here

Demand was previously a single line inside a scenario file — irrigation: annual_demand_m3: 3_000_000 — shaped by a fixed monthly curve and sourced from nothing. That is not good enough to size a scheme against, for a reason worth stating plainly:

A scheme sized against met demand is sized against the drought that already happened. What was abstracted in 2011 is not what was needed in 2011; it is what was available in 2011, which is a measure of the shortage rather than of the requirement. Sizing on it guarantees the scheme is too small, and does so invisibly.

So this layer has to answer a counterfactual: what would have been used, had it been there.

The three demands, and they are not the same kind of thing

1. Licensed abstraction — what is permitted

The EA abstraction licence register gives, per licence: point, purpose, annual and daily quantities, and period of validity. This is the ceiling and the easiest layer to build.

It is not demand. Licences are held unused, held at historic volumes that no longer reflect the cropping, and in some cases held defensively. Actual abstraction returns, where available, are the better measure of what was taken — but see the warning above about what “taken” means in a dry year.

2. Unconstrained agricultural demand — what would have been used

The one that matters, and the one that requires modelling rather than records. The shape of it:

    crop / land use  ->  potential evapotranspiration
                     ->  soil moisture deficit          <- the water layer
                                                           already computes
                                                           AET and effective
                                                           rainfall
                     ->  irrigation need to hold the
                        deficit below a target
                     ->  demand, daily, by location

The water layer already carries the expensive half of this: areal rainfall per catchment, a fitted AET of 470 mm/yr, and a soil-moisture store. What is missing is the land use — what is grown, where, and what deficit each crop will tolerate before yield is lost.

Two properties this must have that the placeholder does not:

• It must vary between years, and vary the right way. Demand is highest in exactly the years supply is lowest. A fixed annual volume removes the correlation that makes the problem hard, and therefore flatters every scheme tested against it.
• It must be a series, not a total. The scheme delivers water in July; a demand expressed annually cannot tell whether it arrived in time.

3. Environmental demand — layered, not single

Not one number and not one threshold. At least three distinct claims on the same water, with different legal standing and different timing:

The third of these is new to the project and is potentially the largest of them. The Somerset Levels are drained by rhynes whose levels are actively managed — held high in summer for stock watering, wet fencing and to stop the peat shrinking, drawn down in winter for flood storage. Holding a summer penning level across a flat moor is a real and recurring water requirement, and it is the most direct expression of the “return water to low-level watercourses” objective. See papers/water/rhynes.md for the network and the storage estimate.

The constraint layers for the designated sites are already fetched — core/constraints.py pulls SSSI, SAC, SPA and Ramsar. What is missing is the link from a designation to a water requirement, which is not in those datasets and generally has to be read from site condition assessments.

Chapters

Status

The finding that most changes the layer: the environmental demand splits by season and only half of it is a drought demand. The wildfowl and wader requirement is a winter–spring inundation, when water is abundant and the scheme has nothing to offer. The penning requirement is April–November, when water is scarce — and its own plans give three purposes, two of which (wet fences, livestock watering) are agricultural. So the summer environmental demand and the farm demand are not cleanly separable, and the priority question “farmers first or environment on a par” is partly a false choice on the Levels.

The denominator is no longer invented, but it is still a ceiling. “% of demand met” in the scheme runs should be read as % of a demand consistent with the licensed ceiling — not as % of what farming would have used had water been there. That counterfactual still needs building, and it is the one number that would change the scheme’s sizing rather than merely its labelling.

The interface with the scheme layer

Priority, as decided: farmers first, with environmental claims treated on a par where a specific requirement exists rather than a general preference. That ordering is a policy choice and is recorded here so it can be argued with rather than discovered in the code.

← All notes · More from Aquifer Storage