Why now?
Three acute information gaps
Anyone who buys olive oil, almonds, citrus fruit or bulk wine signs contracts months in advance – and the size of the harvest sets the price. Within just two years, the price of olive oil recently swung by more than a factor of two. And precisely now, key public estimates have disappeared or have been well off the mark.
Almonds · USA and world market
The most important volume anchor is gone.
In December 2025, after 60 years, the Almond Board of California stopped funding the USDA’s objective crop report (the July estimate).
Olive oil · Spain
Two weeks of uncertainty after the frost.
After the frost of 30 March 2026, technicians of the farmers’ associations needed more than two weeks for a reliable damage assessment. The 2025/26 season ended 5.6 % below the official autumn estimate (the MAPA aforo).
Wine · Spain and Italy
No early estimate any more.
In August 2026, the cooperatives in Castilla-La Mancha and the Italian associations dropped their traditional pre-harvest estimate.
OrchardCast closes these gaps: earlier, regionally resolved down to comarca level and with calibrated probability intervals.
More on the background
First estimates, for instance from trade associations, have often differed widely in the past and are not always free of vested interests. Official figures such as the Spanish aforo appear only shortly before the harvest. An independent, early and regionally resolved crop size estimate that states its uncertainty openly does not exist so far.
Project
Transparency and data accuracy for agricultural markets
We are currently preparing the launch of a new forecasting service designed to close the information gaps in the markets for Mediterranean permanent crops. Our goal is to provide buyers, traders and producers of olive oil, almonds, citrus fruit and wine with early, objective and verifiable crop size estimates.
To achieve this, we use modern methods: a purpose-built physiological growth model is continuously reconciled with the latest satellite data, weather models and official reference figures.
The result is a fully automated “harvest radar” that delivers well-founded, traceable assessments – from short-term damage assessments after frost events to long-term harvest forecasts for the following year.
Benefits
From a biological model to a basis for decisions
Four concrete advantages for bottlers, importers, almond shellers, large wineries, banks that finance harvests, and insurers.
0–3 days · Event assessment
Act within 72 hours of frost or heat.
Instead of waiting two weeks for the associations’ estimates, you receive an objective damage assessment for each comarca within 72 hours.
More
The assessment combines hourly weather data, local cold-air pools from a 30-metre terrain model and the phenological stage. Even in frost-free springs, the confirmed all-clear (“flowering passed without damage”) is of immediate value in your purchasing negotiations.
1–5 months · Season nowcast
Tell an apparent shortage from a real crop loss.
If the monthly production figures fall short of expectations during the olive harvest, the model shows whether the oil is merely reaching the market later or is really missing.
More
Rainy days often only keep machines out of the groves – the olives are still on the tree. Real losses come from frost on the fruit, from hail or from storms that knock the olives down. The model keeps the two strictly apart: the weather-driven progress of the harvest and the biological loss of volume.
3–18 months · Season and following year
Plan months before the official figures.
Long before the official Spanish aforo on 1 October, the model puts a figure on the yield potential of the coming season and of the one after.
More
It draws on canopy vitality, past, current and expected weather, the water held in the soil, the alternate-bearing phase (the switch between strong and weak harvest years) and winter chill.
Every forecast · Risk band
Calibrated risk bands instead of a single number.
Every forecast is an 80 % interval (10th to 90th percentile) from hundreds of scenarios computed in parallel.
More
After every season we transparently score our intervals against the official figures of AICA and MAPA. We analyse and publish a missed forecast – it is never quietly corrected.
Method
An automated data chain meets plant biology
Purely statistical models learn from average years. Frost during flowering, heatwaves or a break in the alternate-bearing cycle are often underestimated by them or not captured at all – yet that is exactly when the market needs the forecast most. OrchardCast therefore brings two worlds together.
01 · Data chain
All input data, processed fully automatically.
Plot-level areas (SIGPAC), trees and canopies from PNOA aerial images and LiDAR, cold-air pools from the 30-metre terrain model, Sentinel-2 and Landsat, ERA5-Land and ECMWF weather ensembles flow into a fully automated chain that we are currently building and that updates itself continuously.
02 · Process model
Biology instead of a black box.
At its core runs not a statistical model but knowledge of plant physiology: the soil water balance and weather-driven yield build-up, flower bud induction, alternate bearing, chill accumulation and stage-dependent frost damage determine the yield.
03 · Outlook
From the olive to further crops.
Around 80 % of the core biological processes – dormancy, phenology, frost, water stress – are shared across species. The system developed for the olive can therefore be transferred efficiently to almonds, citrus fruit, grapevines and pistachios in the main growing regions.
More on the method
A particle filter continuously reconciles the model with observations: satellite data on canopy vitality and, during the season, the monthly official production figures. To do so, the model computes hundreds of scenarios. They differ in their assumptions – for example in the course of the weather or in the model parameters. The spread of the scenarios gives the forecast interval.
Offer
From a free outlook to a data licence
Our model can cover different spatial scales – from the municipality through comarca, province and region up to nationwide forecasts. Because its basic unit is the individual plot (recinto), it can even be scaled down to single farms. Choose the level that suits you.
| Level | Content | Price |
|---|---|---|
| Newsletter | Free newsletter with the latest updates | free |
| Briefing / Season pass | Report on every forecast date with a regional table for each province; as a season pass also for the frost season or the olive harvest season | €300–500 per crop and year |
| Rapid event report | Damage assessment for each comarca within 72 hours of a major frost or heatwave | €500–1,000 per event |
| Pro | Data as CSV and API, resolution by comarca, event alerts within 72 hours, quarterly call | €2,500–3,500 per crop and year |
| Country bundle | all available crops at Pro level | €4,000–6,000 per year |
| Custom analysis | Your specific question, e.g. about particular sourcing regions or farms | on request |
| Institutional data licence | machine-readable raw data feed for banks, agricultural insurers, commodity traders and funds | from €12,000 per year |
Planned prices, as of October 2026. Founding members pay half in the first year (limited places). Questions about licences and custom analyses: info@orchardcast.com
Outlook
Coming soon
OrchardCast is currently under development, and this website will be expanded step by step. Here is what you can look forward to:
Live forecasts
Public summaries of our expectations for the upcoming seasons.
Case studies & track record
Open, time-stamped documentation of our forecasts, so that the quality of our models is documented transparently.
Model details
Data basis, physiological phases, forecast horizons.
Subscriptions
Detailed briefings, regional-level forecasts and data interfaces for professional clients.
Founder
Dr Oliver Tackenberg
OrchardCast is being developed by Dr Oliver Tackenberg. The reliability of its harvest forecasts rests on a foundation of more than twenty years of research in plant ecology and computational modelling. Predictive models, such as ecological niche models or dynamic growth models, use environmental data to calculate with great precision how plants respond to the weather or a changing climate, and where they will grow in the future.
His focus is on combining in-depth expertise in plant ecology with modern data-driven methods, in order to put biological know-how to practical use. For OrchardCast, he processes large volumes of data in a fully automated way – from geospatial data and satellite imagery to hourly weather records. By linking this information with physiological growth models and statistical particle filters, he translates the biological behaviour of the crops into robust harvest forecasts – including transparent information on forecast accuracy.
You can find more about his previous projects at i-flora.com and in the list of his more than 70 scientific publications on Google Scholar.
