Helping Napa Valley winegrowers reduce smoke taint during the 2025 wildfire season using machine learning algorithms.
87%
less smoke-tainted grapes
$340K+
saved in crop losses
96%
accuracy
The problem
The 2025 Pickett Fire burned across Napa Valley in late August, which released smoke during harvesting season. Smoke particles settle on grape skins and absorb into the fruit, which creates bitter, ashy flavors that ruin premium wine. Valley terrain and wind trapped smoke in certain microclimates for up to 2 weeks while other vineyard blocks were clear, but winegrowers had no way to see these differences. As a result, growers faced choosing whether to harvest underripe grapes immediately to minimize smoke or wait for the optimal ripeness and risk losing entire crops to smoke.
The goal
Provide smoke exposure data so winegrowers can forecast where terrain and weather concentrated the heaviest smoke across their vineyard blocks. Using satellite imagery and machine learning models, Agsight helped winegrowers identify which rows experienced dangerous smoke levels and which grapes are ripe enough for immediate harvest, which allowed them to selectively pick high-risk fruit overnight while protecting blocks that could continue developing.
01
Decomposing the problem.
We worked with winegrowers in Calistoga, St. Helena, and Howell Mountain to understand their challenges during the Pickett Fire. Growers shared frustrations about inability to visually assess taint levels before harvest and pressure to make million-dollar decisions based on incomplete data. By interviewing 6 winegrowers from 4 representative vineyards to understand their decisions during the wildfire, we crafted our solution.
How are you deciding which vineyard blocks to harvest first?
How did you assess whether your grapes were ripe enough to harvest?
Which rows of grapes were most exposed to smoke vs. which remained relatively clear?
What tools or data sources did you wish you had access to when the fire started?
How are you deciding which vineyard blocks to harvest first?
How did you assess whether your grapes were ripe enough to harvest?
Which rows of grapes were most exposed to smoke vs. which remained relatively clear?
What tools or data sources did you wish you had access to when the fire started?
How did you decide which blocks to harvest?
How much of your crop do you estimate was affected by smoke taint?
How did you account for terrain or wind affecting smoke concentration across your property?
How did you determine if smoke was affecting different parts of your property?
How did you decide which blocks to harvest?
How much of your crop do you estimate was affected by smoke taint?
How did you account for terrain or wind concentrating smoke across your property?
How did you determine if smoke was affecting different parts of your property?
We created a salinity management system that combines all 3 insights in a way that optimizes irrigation and soil fertility to sustainably increase yield.
Our machine learning models combine smoke exposure maps with ripeness data to generate row-by-row harvest recommendations, which allows for overnight picking of high-risk ripe grapes.
Smoke taint data at your fingertips.
Satellite imagery tracks smoke plumes to identify which rows experience heavy exposure based on terrain features like valley floors that trap smoke versus ridge-top blocks with better air circulation.
Know which crops to harvest during wildfires.
Spectral analysis shows growers which high-exposure rows are ripe enough for immediate harvest versus underripe areas worth protecting.
04
Impacting real farmers on real issues.
To validate our solution, we conducted field trials across 4 vineyards to assess its ability to reduce smoke taint.
"Agsight's smoke maps showed my east-facing Cabernet rows got hit hardest because winds had pushed smoke into that section. I harvested just those 4 acres overnight, which saved about 60% of my premium fruit that would've been ruined if I'd waited or picked everything early."
Steve Rasmussen
Palisades Vineyard
"Agsight told me my upper vineyard had minimal smoke since the plume stayed low in the valley. I could see which rows were at optimal ripeness, so I focused on harvesting the lower blocks that got smoked and let the hillside grapes finish maturing. That kind of precision during a fire is impossible without real-time data."
Adrian Arroyo-Moye
Arroyo Winery
JOIN OUR NEWSLETTER
Sign up for exclusive promos, feature drops, farm tips, and more.
.png)
.png)
.png)
.png)
.png)
.png)
.png)
