The Eden of Apples
Welcome to Val di Non. Carved into the Italian Alps, this valley is an agricultural powerhouse, producing nearly 1 in 5 apples eaten in Italy under the premium Melinda consortium.
Over 20,000 fragmented parcels of land are painstakingly managed by 4,000 farming families. For centuries, the predictable cold winters and mild summers provided perfect alpine growing conditions.
centuries, the predictable cold winters and mild summers provided perfect alpine growing conditions.From the outside, the valley looks as healthy as it ever has. But look closer, and the climate is shifting the ground beneath their feet.
The Satellite's First Clue
In August 2017, the European Space Agency's Sentinel-2 satellite scanned the valley. It measured the Normalized Difference Vegetation Index (NDVI) — a raw measure of photosynthetic health and green leaf canopy.
The satellite confirmed what farmers could see: the trees were vibrantly healthy. In fact, NDVI was actually +1.5% higher than the historical average.
If the valley was greener than ever, the harvest should have been spectacular, right?
But the Trees Were Barren
What the satellite couldn't see was that the vibrant green leaves were hiding a catastrophe. The 2017 harvest was decimated.
Yield crashed by a staggering 60% across the valley. The trees had leaves, but no fruit. They were perfectly healthy, but economically barren.
This is the "Double Anomaly." How does an orchard flourish vegetation while its fruit production collapses entirely?
Cracking the Code with AI
To solve the mystery of 2017, we didn't just look at one satellite image. We extracted a 9-year, high-resolution feature matrix mapping 6 different multi-spectral indices (measuring canopy, chlorophyll, and water) across the entire growing season.
We trained an XGBoost prediction model to find what actually drives apple yield. The result shocked us.
The #1 predictor of harvest success isn't summer rain or how "green" the trees get. It is Day of Year of Spring Green-Up.
The exact week the trees wake from dormancy dictates their fate. Why?
The Silent Killer: Timing > Temperature
We needed irrefutable proof. We secured 32 years of historical ground-truth sensor data from the official MeteoTrentino Station Network. By auditing the precise daily minimum temperatures from stations deep within the valley, we found that 2018 had a much colder night (-6.7°C) than 2017 (-5.0°C). Yet 2018 was a record harvest.
2018 was a record harvest.The secret is phenology. In 2018, the deep freeze happened in March while the trees were asleep. But look closely at the timeline of April 2017:
In 2017, a freak false spring triggered early "green-up." The blossoms opened in mid-April. Then, on April 18th, an Arctic blast swept down the Alps, plunging temperatures to -5.0°C for four consecutive nights.
The freezing air murdered the delicate, fully-exposed blossoms. The tough vegetative leaves survived (hence the high NDVI), but the harvest was lost before summer even began.
The 2050 Reality & The "Blurry Valley"
The Prediction Challenge
To prepare for the future, we looked at global climate projections (CMCC CMIP6). However, standard climate models have a fatal flaw in mountainous terrain: the "Blurry Valley" problem.
This ignores the crucial physical phenomenon of Cold Air Pooling, where dense, freezing air sinks to the valley floor on perfectly clear spring nights. An "average" model predicting a safe 4°C will completely miss the devastating -4°C frost pocket down below.
The AI Downscaling Solution
We deployed our XGBoost ensemble to physically downscale the global climate models, mapping the exact topographic delta of every 1km grid cell. The results confirmed our worst fears: the southern half of the valley has plunged into a multi-year extreme water stress event (measrued via NDWI). By 2050, the "perfect alpine climate" will be forced 300 meters uphill.
The Threat Matrix for Melinda
By downscaling the CMCC data and combining the 9-year historical Spring Frost Frequency, Chronic Drought Stress, and Yield Volatility, we created a unified Climate Risk Score for every 1km grid cell in the valley.
This map is a surgical tool for adaptation. The cells highlighted in yellow, orange, and red are the most vulnerable terrain across multiple threat vectors.
The Precision Breakthrough: Orchard Masking
Initially, our satellite analysis included the entire 1km grid — rocky cliffs, the Noce River, and forests. This caused a massive "River Anomaly" where cell 14_19 was falsely flagged as a drought epicenter.
To fix this, we deployed Orchard Masking. We filtered the satellite pixels using the 42,000+ individual Rigenera apple plot boundaries. This surgical approach eliminates non-agricultural noise, leaving us with the pure spectral health of the trees themselves.
The True Epicenter: Cell 13_18
With the river noise gone, the true risk zone is revealed at cell 13_18 (highlighted in cyan). This area faces the highest cumulative threat from frost frequency and summer heatwaves when looking purely at the plots.
Precision data, applied to geographic reality, creates actionable intelligence. Melinda must prioritize these coordinates for the next generation of resilient orchards.
Interactive Exploration Unlocked
The map is now yours. Explore the Climate Risk Score across the valley and identify local pockets of vulnerability.