Key takeaways
- A drainage decision is a decision about small elevation differences. The fall across a quarter section that decides where water goes can be a fraction of a metre, which puts the required accuracy well inside the range where ground control and checkpoints stop being optional.
- Photogrammetry returns the top of whatever light reaches. On a cropped, grassed or stubbled field, the surface produced is the top of the vegetation rather than the ground, and a drainage model built on it is a model of the canopy.
- Flight timing therefore decides the whole job. Bare ground after harvest or before emergence is the window. Snow removes the detail, standing water hides the bottom, and full crop makes the exercise meaningless unless LiDAR is used instead.
- Standing water extent on a dated capture is evidence in its own right. It shows where water actually sits, which is a different and sometimes more useful thing than where a model says it should sit.
- The useful products are tight contours, flow paths and catchment boundaries, a depression map with depth and stored volume for each low spot, and the position and grade of existing ditches, approaches and culvert inverts.
- A drone surface does not see under water, does not see tile or buried pipe, and says nothing about soil texture or infiltration, which is half of why water sits where it does.
- HEADS UP: drainage work that drains a wetland or alters the flow of a water body commonly needs authorisation, and water moved onto a neighbour is a legal matter rather than a survey one. The capture informs those questions and does not answer them.
Water sits in the same spot every spring. The low corner never dries in time to seed, the approach washes out, the ditch on the north side either is not doing anything or is doing something to the neighbour. Somebody suggests flying it, and the expectation is that a drone will produce a map with arrows on it showing where the water goes.
Aerial capture does produce that, and it produces it well under a specific set of conditions. Outside those conditions it produces something that looks exactly the same and is wrong. On drainage work more than on almost any other application, the value of the capture is decided before the aircraft leaves the ground.
Drainage is a small numbers problem
Most aerial measurement work is forgiving because the thing being measured is large. A stockpile is several metres tall. A pit face is tens of metres. An error of a few centimetres against a quantity like that is noise.
Surface drainage is the opposite. On flat Alberta farmland the total fall across the ground that matters can be a few tens of centimetres over several hundred metres. The question being asked is which way a surface slopes when the slope is a fraction of a percent, and whether a berm of a few centimetres is holding water back. At that scale the vertical uncertainty of the survey is not a footnote. It is comparable to the signal.
Which means the technical requirements are not negotiable:
- Real control, not an assumed datum. Either a properly initialised RTK or PPK solution or ground control points placed to cover the block, and in practice often both.
- A stated vertical datum and geoid model. A drainage surface that has to agree with a municipal culvert invert, a neighbouring survey or a later capture needs to be in the same vertical frame, and the wrong choice shows up as a shifted deliverable rather than an obvious error.
- Independent checkpoints. Points held out of the processing and measured against the finished surface, because an accuracy claim that has not been tested is a claim about the software settings.
A cheap capture flown without control will produce a smooth, plausible, confidently wrong surface, and on flat ground there is nothing in the output that makes the error visible.
What the surface actually is
Photogrammetry reconstructs the top of whatever reflected light back to the camera. On a bare, dry, tilled field that is the ground, and the result is excellent. On anything else it is not the ground:
- Standing crop. The surface is the top of the canopy. A canopy is not uniform in height, it is taller where the crop grew better, and crop frequently grows better where the ground is slightly higher and better drained. So a canopy surface can invert the very signal the capture was supposed to measure.
- Grass, hay and pasture. Better than standing crop and still not the ground, and the error varies with how thick the cover is.
- Heavy stubble and residue. A few centimetres of offset, which on a flat field is a meaningful fraction of the fall being measured.
- Snow. Smooths out exactly the small features that decide where water goes, and drifts preferentially into the depressions that matter most.
- Standing water. The surface returned is the water surface. The bottom of the slough is not in the data.
The timing conclusion follows directly. For a photogrammetric drainage survey the windows are after harvest before freeze up, or in spring after the ground is clear and before the crop is up. Those windows are short in Alberta and they are the same weeks everybody wants a drone for other reasons, which is an argument for booking the capture rather than deciding on it.
Where the ground cannot be cleared, the honest answer is a different sensor. LiDAR returns multiple reflections per pulse and can get a proportion of them to the ground through grass and light vegetation, and a bare earth surface is then produced by classifying the point cloud. It is a filtering decision rather than a direct measurement, and under a thick canopy it still struggles, but on grassed or lightly treed ground it is the difference between a drainage product and a canopy map.
What a drainage capture actually delivers
Assuming the timing and the control are right, the products that get used are these:
- Contours at a tight interval. On flat ground a quarter metre or finer, which is a request worth making explicitly because a default contour interval chosen for a construction site will show nothing on a field.
- Flow paths and catchment boundaries. Derived from the surface: where water collects, the route it takes, which part of the land drains to which outlet, and how much area reports to a given low spot or culvert.
- A depression map with numbers. Every closed low spot, its depth, its area and the volume it holds before it spills, and where it spills to when it does. This is the output that turns a conversation about the wet corner into a decision about whether it is worth addressing.
- Existing works, located and at grade. Ditches, field approaches, crossings and culvert ends, with elevations, so the question of whether an existing culvert is actually lower than the water it is meant to move becomes answerable.
- A dated record of where water was. If the capture is flown while water is standing, the extent of it is direct evidence rather than a model output, and it is the right thing to compare a later capture against.
The limits, stated plainly
A drainage plan is more than a surface, and the capture supplies the surface. It does not supply:
- Anything under water. The depth of a slough, the profile of a dugout, the invert of a submerged culvert. That needs the water gone or a method that measures through it.
- Anything buried. Tile, buried pipe, old drainage works somebody put in decades ago. Not visible, not inferable from the surface.
- Soil behaviour. Two low spots with identical geometry behave completely differently on clay and on sand. Infiltration, texture and compaction are a large part of why water sits where it sits, and no aerial sensor measures them.
- The design. Sizing a ditch or a culvert, setting grades, calculating a design flow and sealing the result is engineering work. The capture is the base the engineer starts from, and it does not replace the design or the professional sign off.
- Permission. Drainage work that affects a wetland or alters the flow of a water body commonly requires authorisation, and the determination of whether a given low spot is a wetland is a specialist assessment. Water moved onto a neighbour's land is a legal question with its own process. A capture makes these conversations better informed and does not settle any of them.
The pattern that works
The jobs that go well look alike. The ground is captured bare, with control, with the vertical datum written down. The depression map and the flow paths go to whoever is doing the design. If there is a dispute or a regulatory question, there is also a dated capture of water standing on the ground, which is harder to argue with than any model. And the same block is flown again after the work is done, in the same window with the same parameters, so the before and after are actually comparable rather than two separate surveys that happen to cover the same field.
That last point is the one most often skipped and the one that makes the file durable. The cost of the second capture is small next to the cost of re-establishing the baseline later. More on the general version of that problem in keeping repeat captures comparable, and the agricultural side of our work is on the agriculture page.
If there is a field or an acreage with a water problem on it, tell us what the decision is and when the ground will be bare. The window matters more than the schedule.
