Bark beetles cause billions of dollars in timber losses globally every year. In Australia, the native Ips grandicollis and introduced species like the European spruce bark beetle pose ongoing threats to plantation forests. Early detection—finding infested trees before beetle populations explode—is the key to effective management. Drones equipped with multispectral cameras are being marketed as the solution. The reality is more nuanced.

The technology relies on a straightforward principle. Trees under stress from bark beetle attack show changes in their foliage before visible symptoms appear to the human eye. Specifically, chlorophyll content decreases and water stress increases in the canopy of attacked trees. Multispectral cameras can detect these changes in the near-infrared and red-edge bands of the electromagnetic spectrum.

A healthy tree reflects strongly in the near-infrared (NIR) and absorbs in the red band. The ratio between these—expressed as vegetation indices like NDVI (Normalised Difference Vegetation Index)—provides a numerical measure of tree health. A tree being attacked by bark beetles will show declining NDVI values weeks before the canopy visibly changes colour.

Several pilot programs in Australian plantation forests have tested this approach over the past two years. The results are genuinely promising for certain scenarios and genuinely disappointing for others.

In Pinus radiata plantations in Victoria and South Australia, drone surveys using DJI Matrice 350 RTK platforms with MicaSense RedEdge-P multispectral cameras successfully identified stress signatures in trees confirmed to be under bark beetle attack. Detection accuracy in these trials ranged from 72% to 88%, depending on infestation severity, stand age, and survey timing.

The 72% figure needs context. That’s for early-stage infestations where trees have been attacked but haven’t yet shown visible canopy symptoms. A 72% detection rate at this early stage is actually impressive—it means catching nearly three-quarters of attacks before they’d be noticed by ground-based surveillance.

But the false positive rate is the problem. In those same trials, 15-30% of trees flagged as potentially infested turned out to be healthy trees experiencing stress from other causes: drought, nutrient deficiency, root disease, or mechanical damage. Every false positive requires ground-truthing—sending someone to inspect the tree in person. When you’re surveying thousands of hectares, false positives add up fast.

Some researchers working with AI consulting companies have been developing machine learning classifiers that can distinguish bark beetle stress signatures from other stress types. These models use the full multispectral signature plus spatial patterns—bark beetle attacks tend to cluster and spread outward from initial infestation points, while drought stress affects trees more uniformly. Early results show false positive rates dropping to 8-12% with trained classifiers, but these models need substantial local training data to work well.

Survey timing is critical and often overlooked in marketing materials. The spectral changes associated with bark beetle attack are most detectable during specific phenological windows—typically mid-spring through early summer when trees are actively growing and the contrast between healthy and stressed foliage is greatest. Surveys done outside this window produce much less reliable results.

Flight altitude affects resolution and coverage trade-offs. Flying lower gives better spectral resolution per tree but covers less area per flight. At 60 metres altitude, individual tree crowns are clearly resolved. At 120 metres, you get four times the coverage but individual tree assessment becomes less reliable. Most operational programs compromise at 80-90 metres.

Weather matters too. Cloud cover changes the light spectrum reaching the canopy and reflected back to the sensor. Consistent lighting conditions are essential for meaningful spectral data. Wind affects drone stability and image quality. These constraints mean you can’t just fly whenever you want—you need the right conditions, which limits operational flexibility.

The Forest & Wood Products Australia (FWPA) research program has funded several of these trials and published findings suggesting that drone-based detection is cost-effective for plantations larger than 500 hectares when replacing or supplementing ground-based surveillance programs. Below that threshold, the technology and training costs are harder to justify.

Integration with existing surveillance programs is where the real value lies. Drones aren’t replacing ground-based surveys—they’re directing them. A drone survey identifies areas of concern, ground crews investigate those specific areas, and management decisions follow. This targeted approach is more efficient than systematic ground surveys of entire plantations.

For Australian biosecurity purposes, drone detection could significantly improve response times when exotic bark beetle species are detected. The faster an incursion is identified and delimited, the better the chances of eradication. Current ground-based surveillance for exotic species relies on trap networks and reactive inspections, which inevitably have gaps.

The technology is real and useful, but it’s not the silver bullet some vendors claim. Budget for ground-truthing, accept that false positives are part of the process, fly during the right conditions, and invest in training classifiers with local data. That’s the honest path to making drone-based bark beetle detection operational.