Note

Palaeochannel or ditch? Reading linear cropmarks on aerial photographs

Rolling farmland hills divided by hedgerows and a wooden field gate, seen from ground level.

You're looking at a long, pale or dark streak running across a scanned vertical or an old RAF sortie, and the desk-based assessment needs a call: is this a relict watercourse or a backfilled ditch. Both show up as linear cropmarks. Both can run for hundreds of metres. The interpretation you write down changes the mitigation strategy, so it's worth slowing down on the three things that actually separate them.

Planform does most of the talking

A cut feature, whether it's a boundary ditch, a trackway hollow, or a pipe trench, was dug by someone who wanted it to go somewhere in particular. It holds straight runs, turns at corners, meets other ditches at right angles or at a clean junction. Field boundaries especially tend toward rectilinear blocks, even on a slope where water would never have run that way.

A palaeochannel doesn't care about property lines. It meanders because moving water meanders: low-sinuosity in a confined valley, high-sinuosity with cut banks and point bars on a floodplain, sometimes splitting into a braided pattern with two or three channels running in parallel then rejoining. If the mark curves back on itself, narrows and widens along its length, or has a scroll-like sequence of parallel arcs beside it, you're looking at a channel, not a cut. Ditches don't leave scroll bars. Rivers don't dig corners.

Width is a second clue worth checking alongside sinuosity. A ditch tends to hold a fairly constant width along a given run, because someone dug it to a plan. A palaeochannel's width drifts, sometimes doubling where the channel widened at a bend and pinching down where it narrowed again. If you can trace the mark and the width is doing something irregular that tracks with the curvature, that's channel behaviour, not ditch behaviour.

What the crop is actually telling you

Both features can produce the same basic crop response for the same basic reason: deeper, moister soil over the fill lets roots run further and the crop grows taller, greener, later to ripen than the surrounding field. That's your differential growth cropmark, and on its own it doesn't distinguish watercourse from cut feature. Where it gets useful is in combination with context.

Palaeochannels usually sit at a consistent point in the local topography, following a shallow low that you'd expect a valley floor or old floodplain to produce, and they often show alluvial colour changes in the soil alongside the cropmark where the channel has been ploughed into. Ditches cut across contours as often as they follow them, because enclosure boundaries and droveways don't follow drainage logic. If the mark's course lines up with the valley's natural fall and the geology mapping shows river terrace or alluvial deposits in that field, weight the palaeochannel reading. If it ignores the contours and lines up with a known field system or a parish boundary on the tithe map, weight the ditch.

It's also worth checking whether the mark has a terminus or a junction. Ditches end at gateways, join other ditches at enclosure corners, or stop dead at a trackway. A palaeochannel either runs off the edge of your survey area or peters out gradually as it shallows toward its source. A clean right-angle junction is close to diagnostic on its own.

None of this replaces borehole or test-pit confirmation once you're on the ground, but it changes what you write in the assessment and what you ask the geoarchaeologist to look for first. A mark you've flagged as a probable palaeochannel points the watching brief toward waterlogged preservation and the need for environmental sampling. A probable ditch points it toward structural phasing and dating evidence from the fill.

The trouble with all of this is that single-date aerial photo coverage gives you one crop, one growth stage, one set of conditions, and sometimes the mark barely shows at all. Running a seasonal NIR cropmark pass over the whole survey block catches marks that one historic sortie missed, and gives you a georeferenced layer to check planform and width against rather than squinting at a scanned print on a lightbox. If the desk-based assessment has a linear mark you can't call with confidence, that's exactly the gap this kind of pass is built to close.

Get started

← Back to the blog