What is AUDPC and how trial teams use it

If you've run a fungicide efficacy trial, you've filled in a disease severity rating sheet more times than you'd like to count. AUDPC is what turns that stack of rating dates into one number you can compare across treatments.

AUDPC stands for Area Under the Disease Progress Curve. Plot disease severity (percent leaf area affected, incidence, whatever scale your protocol uses) on the y-axis and days after planting or after inoculation on the x-axis, and you get a curve. AUDPC is the area under that curve, calculated with the trapezoidal method:

AUDPC = Σ [(yi + yi+1) / 2] × (ti+1 − ti)

where yi is the severity rating at time i, and ti is the day that rating was taken. You sum that across every interval between rating dates for the full season. The result is a single value per plot that captures both how severe the disease got and how long it stayed severe across the season.

Why trial teams reach for it instead of a final rating

A single end-of-season rating tells you where disease landed, not how it got there. Two treatments can both finish at 60% severity and still perform completely differently: one held disease back until week 8 and then it blew up, the other let it creep in from week 3 and plateaued. Final-rating comparisons treat those as identical. AUDPC doesn't, because it integrates the whole progress curve. That's why it shows up in efficacy trial reports, variety screening data, and regulatory submission packages, it's the standard way to defend a claim that one treatment suppressed disease better over time, not just at one checkpoint.

It also normalizes trials that don't start disease-free at the exact same hour. If inoculation timing or natural infection pressure varies slightly across replicates, AUDPC still captures the trajectory rather than anchoring everything to day one.

Where the calculation quietly breaks

AUDPC is only as good as the rating dates feeding it. A few things that trip up field teams:

• Rating interval gaps. The trapezoidal method assumes severity changes roughly linearly between two ratings. If your scouts rate weekly but disease doubles in four days somewhere in that window, the curve smooths over the spike and your AUDPC undercounts it.
• Missed onset. If visual symptoms aren't scoreable until well after infection has started spreading, your first few data points are all zeros that should have had some value. That flattens the early part of the curve and shifts the whole number down, which matters a lot when you're comparing a slow-onset treatment against a fast one.
• Rater consistency. Severity scales are visual by nature. Swap raters mid-season, or have different people walking different blocks, and you've introduced noise that AUDPC can't tell apart from real treatment effect.
• Normalization. Comparing AUDPC across trials with different season lengths requires relativizing it (rAUDPC, dividing by the maximum possible area) or you're comparing apples to a much bigger apple.

None of this makes AUDPC less useful. It makes the rating schedule underneath it the thing worth getting right, because the calculation can't fix a gap in the input data after the fact.

That last point is where most of the actual risk sits. A weekly scouting pass across a multi-site trial program catches a lot, but it's still a snapshot taken on a schedule, and disease doesn't wait for the scout's route. The stretch between when an infection actually starts spreading and when it becomes visible enough to score is exactly what gets lost from an AUDPC curve, and it's usually the stretch that matters most for separating a good treatment from a mediocre one.

Flagging anomalous spectral signatures before visible symptoms show gives you an earlier anchor point for that curve, one closer to when infection began rather than when it crossed a visible threshold.

If your trial sites could use an earlier warning note to anchor the front end of that curve, that's what we're building.