River capture

Why can the divide between two river basins creep across the land, handing one river's headwaters to its neighbour?

A drainage divide looks like the most permanent line on a map. Rain falling a metre to one side of it ends in one sea, a metre to the other side in another, and the line itself is drawn along ridge crests that seem about as fixed as anything in a landscape can be.

Yet there are places where the evidence flatly contradicts that. A dry notch sits in a ridge, far too broad and deep for the trickle that occupies it — or for nothing at all. Below it, a small stream wanders through a valley plainly cut by a river many times its size. And nearby, a river makes a sharp, almost right-angled turn, as though it had changed its mind. Something moved the line. What could move a watershed?

Begin by asking what fixes the divide's position in the first place. Two streams work back into the same high ground from opposite sides. The divide sits wherever their headward erosion has got to, so it is stable only while both are eroding at the same rate. That is not a law; it is a coincidence. So the useful question becomes: what makes one side erode faster?

Three things, mostly. A stream that reaches its base level — the sea, or a lower trunk river — over a shorter horizontal distance is steeper, and steepness raises the power a given flow exerts on its bed. A stream fed by more rainfall carries more water and does more work. And a stream cutting weaker rock removes more per unit of effort. Any of the three, acting for long enough, makes one head advance faster than the other, and the divide is simply the place where the advance has reached. It moves toward the slower side.

Now follow the migration to its conclusion. The aggressive stream's head creeps up and through the high ground until it intersects the channel of its neighbour at a point upstream of where the divide used to be. At that moment the neighbour's headwaters have somewhere lower to go, and they go there. The flow switches basins in an instant, though the approach to it took a very long time.

Then ask the interesting question — what happens next — because this is where the process stops being gradual. The aggressor has just acquired the discharge of another river's headwaters. More water means more stream power, so it incises faster, which lowers it further relative to its neighbour, which steepens its remaining headwater slopes, which speeds the divide's migration again. The victim has lost its supply and cannot cut, so it cannot push back. The advantage is self-reinforcing on one side and self-cancelling on the other, which is why captures tend to come in sequences rather than singly.

That leaves a set of fingerprints, and they are what turn the story into a claim you can check. At the junction, an elbow of capture: the diverted water arrives at a sharp angle rather than the smooth downstream-pointing junction rivers otherwise make. In the ridge, a wind gap: a valley with no river in it, because its river now flows elsewhere. Beyond the gap, a misfit stream — a small watercourse in a valley cut by something far larger.

Three refinements keep this honest.

The first is that "the divide creeps" is a description of a result, not of a thing that travels. Nothing moves along the ridge. Two erosion fronts advance at different speeds, and the divide is wherever they currently meet — a bookkeeping line, not an object.

The second is that capture is a threshold event riding on a continuous process. The migration is slow and steady; the diversion is abrupt. Both halves matter, and treating capture as sudden misses the long, invisible approach that made it inevitable.

The third is a caution against reading every elbow as a piracy. Sharp bends also occur where a river follows a fault or a joint set in the rock, and dry gaps through ridges are also cut by ice-marginal drainage or by a channel diverted by a glacier and never reoccupied. The classic account of divide migration by differential erosion is well supported, but which particular gap records which particular event is often argued, and the rates are now estimated by methods — cosmogenic nuclide exposure ages, and comparisons of channel geometry across a divide — that were not available when the idea was first framed.

So the falsification test should not rest on landform shape alone. If a capture happened, the beheaded valley below the wind gap must contain river gravels whose rock types come from the headwaters that were diverted away — material the present misfit stream has no access to and could not have delivered. What would refute it: gravels in that valley containing only local lithologies, with nothing sourced from the supposed captured catchment, and a wind gap whose floor sits at an elevation inconsistent with any through-going channel graded to the old outlet. That would leave the elbow and the gap needing a structural or glacial explanation instead.

• How comparing channel steepness on either side of a present-day divide predicts which way it is currently moving.
• Why a captured river's new, steeper profile sends a wave of incision upstream for a very long time afterwards.

Every term the collection defines is gathered in the glossary.