REWT Rivers of England and Wales, Temporally

Background

The proof of concept

REWT follows a scoping exercise carried out in 2026 under the name Premodern Rivers, which produced a working reconstruction, two delivered datasets, and a map.

It began as a commission of three or four days. Justin Colson (Institute of Historical Research) wanted a GIS of rivers for a Domesday mapping project, and proposed the obvious construction: take Oksanen’s routes and splice them into the modern OS Open Rivers network.

What grew it was a cartographic decision, not a difficulty. A national river map is unreadable and enormous unless the smallest streams are dropped, and it is uninformative unless a river is drawn at a width that means something — so flow had to be computed, first to clip low-flow channels reliably and second to give every drawn line a width indicative of what it carried. That is a modest requirement with an immodest consequence: flow is modelled on the network, so the water has to be able to move through it. The modern survey, it turned out, does not let it — water cannot be followed from every stretch to the sea — and a flow model over a network the water cannot leave returns confident numbers about nothing.

So the three or four days became the finding that sets the order of everything since: repair the modern network first, because nothing downstream of it means anything until it is sound. The splice itself was dropped for a separate reason — where the synthesis’s courses are derived from parish boundaries they sit a median 236 m from the modern channel, so merging them would have reconciled two orders of accuracy silently at every junction. The project now declines to do the thing it was asked for, and the reasoning behind that refusal is most of what it has to offer.

That work is not public, and this page is the account of it. What it established, and what was and was not carried forward, is set out below; the regional source surveys it produced have been rewritten for this project and are the rebuilt landscapes.

What it established

That the thing is buildable. A national network can be assembled from open sources, dated, and modelled — and published as a map that a reader can interrogate.

That the modern survey is the problem. This is the finding that reorders everything. Defect after defect turned out to be a fault in the modern network’s topology rather than historical uncertainty: rivers ending inland, whole basins draining into nodes with no outflow, thousands of kilometres unreachable from the sea. Every one was invisible in the national totals and obvious on the map.

That a plausible number is the dangerous kind. A flow model will report a confident figure over a network in which the water cannot actually leave. Nothing in the summary statistics says so.

That corrections should arrive as records, not descriptions. A tool was built to let a reader click the line that is wrong and say what is wrong with it, so a correction carries the identifiers rather than a description of a place. It repeatedly found in minutes what analysis had missed for days.

What carries over, and what does not

To be carried over: the approach, the hand-made judgements and their reasoning, the tracing and correction tooling, and the source declarations. The judgements and the source declarations are here; the tooling is rebuilt rather than copied, and is not here yet.

Not carried over: the pipeline’s accumulated assumptions about the order of work. The scoping exercise grew twenty-three stages in which the sequence was load-bearing and largely undocumented, and several stages quietly overwrote each other. That is why REWT starts from a clean sheet rather than a fork, and why reproducibility is the first phase of work rather than an afterthought.

Naming

The earlier work is cited as Premodern Rivers; its releases keep their identifiers and remain valid. REWT is the name of the undertaking that follows it, and covers a wider span — from Domesday to the present-day channel it is built on — than premodern honestly describes.