In almost every conversation about remote sensing for mine rehabilitation, “seeing weeds” quickly becomes the focal point.
“Can I see the weeds?”
“Can I count the weeds?”
“Which species are there?”
These sound like reasonable questions. But rather than take them at face value, I see them as shorthand for a bunch of concerns about risk, compliance, and confidence: Will weed pressure undermine rehabilitation success? Will problems be detected early enough to intervene? Will there be defensible evidence if decisions are questioned?
Those are legitimate concerns. But when “seeing weeds” becomes the dominant criterion for evaluating monitoring approaches, it often nudges rehabilitation programs in the wrong direction.
Why “seeing weeds” feels like the right focus
Weeds are tangible. You can point to them in the field. They feature prominently in management plans and risk registers. And for anyone who has watched a rehabilitation area fail because of weed dominance, they are an obvious warning sign.
It also doesn’t help that much of the current technology narrative reinforces this focus. High‑spatial resolution imagery, AI‑driven classifiers, and automated species detection maps promise the ability to see more, in more detail, than ever before. It’s easy to assume that if we can identify weeds from the drones, we must be reducing risk.
But visibility and confidence are not the same thing.
Not to mention the huge amount of training that needs to be done to feed those hungry AI models. And who really wants to do that?
What rehabilitation monitoring needs to achieve
Mine rehabilitation is rarely about the complete absence of weeds. Regulators are not looking for pristine, weed‑free landscapes. What they are looking for is evidence that a site is:
on a credible trajectory toward the approved final land use;
structurally and functionally stable;
increasingly self‑sustaining; and
comparable, over time, to appropriate reference (analogue) sites.
These are outcomes assessed through patterns, trends, and trajectories, not through single observations.
A rehabilitated area can contain weeds and still be progressing well. Conversely, it can appear uniformly green in a snapshot and be fundamentally unstable. Perhaps it’s dominated by opportunistic species that respond quickly to rainfall but do little to build long‑term ecosystem function.
When monitoring programs prioritise what is easiest to see at one moment in time, they can miss what matters most.
The limits of snapshot thinking
Much of the emphasis on “seeing weeds” comes from a snapshot mindset: if a problem is present, it should be visible in an image – or through ground survey.
The trouble is that rehabilitation systems are dynamic. Seasonal growth and senescence, rainfall variability, establishment phases, and disturbance legacies all shape how vegetation looks at any given moment. A single image or site check, no matter how detailed, captures only a sliver of that story.
In practice, this can lead to two common pitfalls:
False reassurance: short‑lived greenness interpreted as success
False alarm: temporary weed presence interpreted as failure
Neither is particularly helpful for decision‑making.
A more useful focus? Behaviour over time
If the real concern behind “seeing weeds” is risk and defensibility, then a more useful focus is not visibility, but behaviour.
Over time, functioning rehabilitation areas tend to show:
seasonal responses that align with climate and rainfall;
recovery after stress events such as drought;
decreasing volatility as systems mature; and
convergence toward the behaviour of analogue sites.
Areas at risk often show the opposite: exaggerated responses, persistent divergence from reference conditions, or increasing instability through time.
These patterns are often detectable well before problems are obvious on the ground. And long before a once‑a‑year survey or flight would capture them.
An important caveat…
None of this is an argument against weed mapping in general.
In precision agriculture, biosecurity, or targeted eradication programs, identifying specific weed species is often exactly the right focus. In those contexts, spatial precision is critical. And where weeds are genuinely affecting mine site rehabilitation, knowing where they are and what they are is important to target interventions.
But mine rehabilitation operates at different spatial and temporal scales, with different regulatory objectives. Here, confidence comes less from knowing exactly what is present in one place at one time, and more from understanding how systems are behaving through time.
Can we ask better questions?
“Seeing weeds” persists as a focal point because it feels concrete. But it is ultimately a proxy for deeper concerns about risk, accountability, and outcomes.
When monitoring programs are designed around better questions, the resulting evidence is often more useful, more defensible, and more aligned with how rehabilitation success is actually assessed. How about:
Is this area stabilising? Is it responding to seasons as expected? Is it trending towards becoming more like its analogue, year by year?
This post is the first in a short series exploring how we design monitoring programs for mine site rehabilitation, and why the questions we ask of our data matter just as much as the data itself.
Read on to see what good ecosystem rehabilitation looks like over time.
Banner image by Leiada Krözjhen on Unsplash
