Can AI forecast timber harvest volumes from weather data?

Yes, but weather probably matters most through a route many forecasts don't model. A 2024 study of 25,387 harvester records, in plantations with no frozen-ground season, found rainfall, temperature and solar radiation added so little to predicting a machine's productivity that they were dropped. Weather lifted the best group of models' mean fit only from 0.63 to 0.65, though the authors still judge weather plus harvest data "an accurate approach". In our reading, weather's larger effect is on whether the ground and roads can be worked at all. A useful forecast predicts operable days from the operation's own history, then applies productivity to them.

What the usual answer says

The leading pages on the subject describe AI blending data sources. A January 2026 article on forest.fi says AI "improves forecasts of timber volumes, harvesting opportunities and assortments". The claim is about updating forest resource data rather than weather. The article adds that future assistants will "combine forest data, weather forecasts, cost estimates and legal requirements to generate a management recommendation". It gives no figures.

Combining the data is right. The question is which effect of weather the combination is meant to capture, and the evidence points away from the machine.

What weather does to the machine

Rodrigo Almeida, Richardson da Silva and Danilo Simões published a model comparison in Forests in 2024. They trained 24 algorithms on 25,387 records to predict harvester productivity in eucalyptus plantations. Each ran once on harvesting attributes alone and once with weather added. When they ranked the inputs, three weather features dropped out: "mean air temperature, mean global radiation, and mean rainfall were removed. These results indicate that these features have low relevance in the dataset." The inputs that mattered most were working hours and average tree volume, then stand age and operator experience.

On the test data, the best group of models averaged a fit of 0.63 without weather and 0.65 with it. The best single model went from 0.67 to 0.70. Wind, humidity and pressure did help a little, which the authors link to visibility and conditions in and around the cab. Their conclusion is that "The use of weather data combined with timber harvesting attributes in the model is an accurate approach for predicting harvester productivity".

Two limits matter here. The study covers eucalyptus plantations in a climate with no thaw, so its weather range never includes frozen or thawing ground. And its strongest input is working hours. A shift cut short by rain shows up as fewer hours, so the model may be absorbing part of weather's effect through hours rather than through the weather features.

What weather does to access

Seasonal access is where we think weather moves volume most. From Quarri's own work with a forestry operation: spring thaw cut harvest volumes by 80%+ from the winter peak.

That figure doesn't say why. Soft ground, road load restrictions, mill demand and planned shutdowns can all fall in the same weeks, and only some of those are weather. Our reading is that much of the drop is access, the days the ground and roads can be worked, but a forecast should record each cause separately so the model can tell them apart.

That effect is changing. Michael Kilgore, Charles Blinn and Stephanie Snyder ran focus groups with agency field foresters for a 2025 paper in the Journal of Forestry. In the paper's summary, foresters said "warmer winter temperatures and precipitation changes have generally led to less conducive conditions for winter harvesting". They also named factors that have nothing to do with weather: "poor timber markets and larger, heavier, and more productive logging equipment." It reports what foresters observed rather than a measurement, and it covers one region's frozen-ground season.

How to build the forecast

Treat volume as operable days multiplied by output per day, and forecast each part from its own data. Operable days depend on ground and road conditions, road restrictions and planned shutdowns. The first two follow weather with a lag: frost depth, thaw, days of rain. The operation's own daily history of volume, set against the weather that preceded it, shows where its thresholds sit. Output per operable day depends on the stand, the machine and the crew, and the one productivity study we found says weather adds little there.

In our view several years of daily records are the minimum. A single season contains one thaw, and a model trained on it learns that year's dates rather than the conditions that caused them. The forecast should be checked each week against what was actually hauled, because a forecast of access is only as good as its most recent season.

When it doesn't apply

Operations on well-drained ground or all-weather roads may rarely lose days to weather, and for them the productivity model is most of the forecast. Where daily volume records do not exist, only monthly totals, the thresholds cannot be learned and the seasonal pattern is all a model can offer. And beyond the range of a weather forecast, any volume forecast is a forecast of the season rather than the weather.

Quarri for forest management is built around how a forest operation runs, from the cruise to the settled account.

Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.