Analyses
An analysis asks one question of a finished survey. Open the survey, choose Analyse, pick a question, optionally draw an area to restrict it to, and run it.
The work outlives the page. Close the browser; come back later.
Drawing an area
Every analysis can be run over the whole survey or over a shape you draw on it. Restricting it is faster and, for some questions, more accurate — a stockpile volume over the whole survey has to guess which heaps you meant.
If you draw nothing, the analysis covers the survey's footprint. That still counts against the area limit on your plan; "the whole survey" is bounded by the survey.
What it can answer
Stockpile volume
Measures the volume of a heap against the ground around it.
Draw the area a little wider than the pile. The ring of ground outside the heap is the evidence for where the ground underneath it is — there is no other way to know. Draw tight to the toe and the base is guessed from too little, and the number comes out low.
Reads the surface model, so it measures the top of whatever is there. A heap with a truck parked on it includes the truck.
Slope and aspect
How steep the ground is and which way it faces. Reports hectares by slope band and draws the steep country as its own layer. Reads the bare-earth model.
The usual uses: finding country too steep to work, planning access, and working out which faces get sun.
Water flow and ponding
Fills the ground to its brim to find where water sits after rain — how many hollows, how deep each one is, and how much it holds.
This is geometry, not hydrology. It tells you where water would collect on this shape of ground. It does not know about soil, infiltration, culverts, or how much rain fell.
Watercourse detection
Routes water over the ground to find where it runs — the channels and drainage lines, including the ones with no water in them today. A dry creek is still a watercourse to the terrain, and this finds it.
Erosion and gullies
Finds channels cut into the ground: incised, long, and carrying enough flow to keep cutting. Aimed at the erosion worth reaching before it widens.
Tree identifier
Counts canopy and measures each crown — area, spread, and height above the ground beneath it. Uses both elevation models: the difference between the top of the canopy and the bare earth under it is the tree's height.
Works well on separated trees. Closed canopy is harder, because the boundary between one crown and the next is genuinely ambiguous in the elevation model, and the count will read low.
Pipe finder
Finds long, straight disturbances in the bare-earth model — the trench a buried pipe was laid in, which outlasts the work by years. It reads three signals: the trench itself, lines that are a different temperature from the ground around them if the survey was flown with a thermal camera, and lines where what grows over them is a different colour from the paddock either side.
These are candidates to check, not confirmed pipe. A fence line reads exactly the same way. So does an old track, a buried cable, and a windrow. It narrows a paddock down to a few lines worth walking with a locator; it does not replace the locator.
Trained model
Finds whatever a model was taught to find, from examples somebody outlined on a survey. See Training a model.
Reading a result
Results draw over the map as outlines you can click. Clicking one shows what was measured about it — area, height, volume, whatever the analysis produces.
A result is a set of findings, and you can act on each one individually:
- Keep it and it stays in the count and the exports.
- Fold it away when two findings are the same thing counted twice.
- Throw it out when it is not the thing at all.
The totals follow. This matters: an automatic count that cannot be corrected is a number you either accept whole or discard whole, and neither is what you want when nine of ten are right.
How much to trust it
The terrain-derived analyses — slope, water, ponding, gullies, watercourses — are geometry over the elevation model. They are as good as the reconstruction is, and the reconstruction is good relatively and mediocre absolutely. Trust the shape; check the absolute elevations if they matter.
The detection analyses — trees, pipe, and anything from a trained model — are statistical guesses. They miss things that are there and find things that are not. Every one of them is a candidate to verify, and the fold/discard controls exist because verifying is part of the job.
Volumes are the ones people put in contracts. A properly overlapped flight over a well-defined heap with clean ground around it will typically land within a few percent. That is not survey-grade and should not be represented as such.