How Satellite Imagery and AI Predict Coffee Harvests Months Ahead

Coffee prices often move before most buyers understand why. A drought in Brazil, irregular rainfall in Viet Nam, or a disease outbreak in a key producing region can tighten supply months before the harvest reaches port. Satellite imagery and artificial intelligence are changing that by giving traders, roasters, and cooperatives earlier signals about crop health and likely yield.

What satellites can see

Satellites do not photograph coffee cherries directly from space. Instead, they measure how plants reflect light. Healthy coffee vegetation reflects light differently from stressed plants, and indices such as NDVI (Normalized Difference Vegetation Index), LAI (Leaf Area Index), and FAPAR (fraction of absorbed photosynthetically active radiation) help estimate canopy vigor and growth.[1][2]

By tracking these signals over time, analysts can identify patterns linked to flowering, vegetative growth, stress, and maturity. A field that shows declining vegetation health during a critical growth stage may indicate drought, nutrient stress, pest pressure, or reduced yield potential.

Where AI adds value

Satellite data alone produces a large stream of images and measurements. AI turns that stream into usable forecasts. Machine-learning models can combine vegetation indices with rainfall, temperature, soil data, farm management practices, historical yields, and even altitude to estimate production.

Research in Viet Nam’s Central Highlands found that models using satellite-derived NDVI, LAI, and FAPAR from the first six months of the year could forecast regional coffee yields with adjusted R² values of 64–69% and mean absolute percentage errors of 3.9–4.7%. In Brazil, studies using high-resolution imagery and machine learning have predicted yields up to nine months before harvest.[2][3]

These are not perfect predictions. But in an industry where a few percentage points of supply can affect prices, even a directional forecast has commercial value.

Why early signals matter

For an importer or roaster, an early harvest estimate affects purchasing strategy. If satellite data suggests a smaller Robusta crop in Viet Nam, a buyer may secure volumes earlier, review contracts, or diversify origins before prices rise. If the outlook is strong, the business may avoid overpaying in a panic-driven market.

For producer organizations, the value is different. Early forecasts can support harvest labor planning, drying and warehouse capacity, financing decisions, and communication with buyers. A cooperative that knows a region may face lower yield can plan logistics instead of discovering the shortfall during collection.

The limits of prediction

Satellite-based forecasting is powerful, but it is not fortune-telling. Cloud cover can interfere with optical imagery; smallholder farms may be too fragmented for field-level precision; and models trained in one region may not transfer cleanly to another. Coffee also has biennial bearing cycles, where trees naturally alternate between heavier and lighter production years, making historical yield data essential.

The most credible forecasts combine several inputs: satellite vegetation data, weather records, soil and agronomic information, farmer reports, and historical harvest outcomes. A study in Viet Nam found that combining agroclimate variables, satellite-derived evapotranspiration, and farm-management information produced reasonable yield estimates across major producing provinces.[4]

What it means for coffee businesses

The practical shift is from reacting to price movements toward anticipating them. Satellite and AI tools will not remove volatility, but they can shorten the time between a field-level problem and a business decision.

For a coffee professional, the useful questions are no longer only “What is the current price?” or “What did last season’s crop look like?” They are: What is the vegetation trend in my key origins? How does it compare with previous seasons? What weather risks are emerging? And how should purchasing, pricing, and inventory plans adjust?

Coffee will always involve uncertainty. But satellite imagery and AI make that uncertainty more measurable and in a volatile market, measurable uncertainty is a competitive advantage.

Citations:

[1] Early Prediction of Coffee Yield in the Central Highlands of Vietnam Using a Statistical Approach and Satellite Remote Sensing Vegetation Biophysical Variables https://orbi.uliege.be/bitstream/2268/292543/1/remotesensing-14-02975.pdf

[2] Early Prediction of Coffee Yield in the Central Highlands of Vietnam ... https://agris.fao.org/search/en/providers/122535/records/65df5d206eef00c2cea27acc

[3] Coffee yield prediction using high-resolution satellite imagery and crop nutritional status in Southeast Brazil https://www.sciencedirect.com/science/article/abs/pii/S235293852300174X

[4] Probabilistic yield forecasting of robusta coffee at the farm scale ... https://agris.fao.org/search/en/providers/122535/records/65df6896b766d82b18018e19

[5] An agronomically informed framework for temporally ... https://link.springer.com/article/10.1007/s11119-026-10454-2?error=cookies_not_supported&code=59b84f8b-a2aa-4aa5-a132-5d67b791fe95

[6] MAPPING THE COFFEE FIELD IN RIO BANANAL, ES, USING ... https://agris.fao.org/search/en/providers/124122/records/69e20e24920f500389b20772

[7] [PDF] Estimating Arabica Coffee Productivity Using ... - alice Embrapa https://www.alice.cnptia.embrapa.br/alice/bitstream/doc/1186268/1/AP-Estimating-Arabica-coffee-2026.pdf

[8] Coffee-Yield Estimation Using High-Resolution Time-Series Satellite Images and Machine Learning https://repositorio.usp.br/directbitstream/eb03ed96-4d80-4393-bb4b-a17701928041/3105395-Coffee-Yield%20Estimation%20Using%20High-Resolution%20Time-Series%20Satellite%20Images%20and%20Machine%20Learning.pdf

[9] mapping coffee management zones using satellite-derived ... https://www.alice.cnptia.embrapa.br/alice/bitstream/doc/1181979/1/AA-Mapping-coffee-SIAGRO-2025.pdf

[10] Smart Coffee: Machine Learning Techniques for Estimating Arabica Coffee Yield https://repositorio.usp.br/directbitstream/9e681c38-12bd-481f-9ba1-cd0b9e8ffc26/3228219-Smart_Coffee_Machine_Learning_Techniques....pdf

[11] Deep learning for large-scale mapping of coffee cultivation in Minas Gerais, Brazil https://www.tandfonline.com/doi/full/10.1080/10106049.2026.2743328

[12] The use of machine learning in digital processing of satellite images applied to coffee crop | CABI Reviews https://www.cabidigitallibrary.org/doi/abs/10.1079/PAVSNNR202015045

[13] Estimating Coffee Plant Yield Based on Multispectral Images ... https://agris.fao.org/search/en/providers/122436/records/67598b3fc7a957febdfba2cd

[14] Improving Coffee Yield Interpolation in the Presence of Outliers ... https://agris.fao.org/search/en/providers/122413/records/687a8e33464818917f33227b

[15] Dense Time Series of Harmonized Landsat Sentinel-2 and Ensemble Machine Learning to Map Coffee Produ

ction Stages https://www.mdpi.com/2072-4292/17/18/3168