Hyperspectral vs multispectral imagery for disease detection in trial plots

If your trial protocol already calls for weekly or biweekly UAV passes over efficacy plots, you've probably had the multispectral-vs-hyperspectral argument at least once this season. Usually it comes up right after someone asks why the NDVI map looked fine on Tuesday and the plot had visible lesions by the following Monday.

Multispectral sensors give you a handful of broad bands. Hyperspectral sensors give you dozens to hundreds of narrow ones across the same range.

What the extra bands actually buy you

A typical multispectral UAV rig gives you red, green, blue, red edge, and near-infrared, maybe five to ten bands total. That's plenty to compute NDVI or NDRE and flag general vigor loss, biomass drop, or canopy stress across a trial block. What it can't do well is tell you why a plot is stressed. A rust infection, a nitrogen deficiency, and early drought stress can all produce a similar dip in NDVI. For a trials lead trying to separate treatment effect from disease pressure from plain old soil variability across replicates, that ambiguity is expensive.

Hyperspectral imagery samples reflectance across dozens to hundreds of narrow, contiguous bands, spanning visible, red edge, near-infrared, and into the shortwave infrared. That density captures the shape of the reflectance curve, not just a couple of index points on it. Pigment shifts tied to specific pathogen responses, changes in canopy water content, subtle chlorophyll fluorescence effects. These show up as curve deformations well before a human walking the rows would call it symptomatic. Anomalies in that curve, tracked against known pathogen spectral signatures, are what let you flag an outbreak while it's still asymptomatic to the eye.

Spectral bands for disease scouting, in practice

For disease scouting specifically, a few spectral regions tend to carry the signal:

• Red edge (roughly 680-750 nm): sensitive to chlorophyll content changes, often the first region to shift as a plant diverts resources to fight infection.
• Near-infrared plateau: reflects internal leaf structure; pathogen damage to mesophyll tissue shows up here before visible wilting.
• Shortwave infrared (SWIR), where the sensor supports it: tracks water content, useful for distinguishing disease-driven stress from drought stress, two things that look very alike in the visible and NIR alone.

A multispectral sensor gives you one or two sample points across each of these regions. A hyperspectral sensor gives you ten or more per region, enough to tell whether a dip is a straight drop or a dip with a shoulder on one side, and that shape is what maps back to a specific pigment or water-content change rather than just "something's off."

When multispectral is still the right call

A weekly multispectral pass still earns its keep on large trial footprints where the goal is spotting gross treatment differences, lodging, or an irrigation line that's clearly underperforming. The flights are quicker to plan, the files are smaller to process, and most field teams already have an NDVI workflow built around that output. Hyperspectral earns its keep somewhere narrower: dense, small-plot trial layouts where the job is separating pathogen signal from nutrient stress, drought stress, and ordinary plot-to-plot variability, and where catching that signal while the plot still looks clean on the ground is the whole point of flying.

The resolution side matters too. At under 10 cm GSD from a UAV, individual plots and even sub-plot variation stay distinguishable, which is the level most efficacy trials actually need. Crewed aircraft can cover more ground per flight, but UAVs give you the repeat cadence a weekly scouting protocol depends on.

Where this fits your protocol

If your current scouting routine is a human walking transects plus a multispectral vigor map, hyperspectral doesn't replace either of those. It adds a layer in between: a weekly anomaly signal that tells you which plots to prioritize for closer inspection before anyone sees a lesion. That's the gap weekly hyperspectral flagging across trial sites is built to close, tracking spectral anomalies against known pathogen patterns so your field team knows where to look first.

We're still building this with a small early-access group of trials teams rather than selling it as a finished product, so if catching outbreak signals a week or two earlier would change how you run efficacy trials, it's worth a conversation.