Behind every forecast is a ladder of data, physics, and AI, built to highlight the best and illuminate the worst in the snowpack.
Good forecasts can still lead to bad turns. Bluebird skies and a few fresh inches don't guarantee soft, consistent snow, sometimes you step out of the car into wind-scoured crust or dense, heavy powder. Standard weather apps tell you what falls from the sky, not what happens once it hits the ground.
We take a different approach. Instead of stopping at snowfall totals, we build a multi-layered picture of the snowpack, combining satellite data, physical snow models, and terrain-aware analysis. We don't just forecast the weather; we forecast the skiing. And that means looking beneath the surface too: tracking weak layers, temperature gradients, wind slabs, and avalanche stability as they evolve.
Think back to those days when the weather looked perfect but the skiing wasn't, or when conditions looked marginal, yet you found pockets of incredible snow. Those moments come down to what's happening inside the snowpack. Our system captures the nuances: how storms arrive, how they set up, how wind transports snow, how new layers bond with old ones, and how those structures strengthen or fail over time. WinterScience brings all of this into a single, clear view, what you ski on and what you ski over.
Garbage in, garbage out. To run the world's most advanced snow physics models, we ingest a massive spectrum of data points. We categorize our inputs by precision:
The gold standard. Real-time temperature, wind, and precipitation data from on-mountain sensors. This is as accurate as it gets.
High-resolution precipitation scanning. We calculate coverage percentage (0% to 100%) to understand exactly where snow is falling, even between stations. This fills critical gaps in mountainous terrain.
Broad-scale atmospheric data that fills in the gaps where radar and stations can't reach. Provides coarse but comprehensive precipitation metrics across all terrain.
The baseline simulation of the atmosphere. We combine multiple model ensembles to capture the full range of possible outcomes.
Why this matters: We use this historical data to build a "memory" for the snowpack. We know exactly how much rain fell on the snow three weeks ago, or how much wind hammered the ridge last night. This ensures our current conditions report is physically accurate, not just estimated.
All of these inputs are combined with numerical weather simulation data to produce the most accurate historical precipitation and weather record possible. This becomes the foundation for everything else, the more complete our inputs, the more accurate our models become.
Most apps use a single, generalized model for a massive region. We run custom AI models tailored to the complex terrain of each mountain.
Every ski area in our app gets a proprietary AI forecast that outperforms standard industry models. This forecast is built from the ground up for mountain terrain, understanding how elevation, aspect, and topography affect weather patterns.
All areas receive satellite precipitation coverage and benefit from our advanced model ensembles, ensuring state-of-the-art forecasting as good or better than the best available.
For mountains where we have dense input data, weather stations plus radar coverage, we run Hyper-Local AI models. These focus enormous computing power on a specific micro-climate, capturing terrain-specific patterns that generalized models miss.
The result: forecast accuracy that reflects the true personality of every slope, ridge, and valley. This is custom forecasting purpose-built for individual resorts.
You've likely noticed that mountain wind forecasts are rarely accurate. That's because standard models smooth out the peaks. Our Hyper-Local models understand how wind accelerates over ridgelines and funnels through couloirs, aligning with what you actually feel on the chairlift. You may have noticed that wind forecasts and wind sensors on ski mountains don't usually line up well, but in our case, they do.
Forecasting the weather is only half the battle. We also run physics-based simulations of the snowpack itself, tracking every layer, every grain type, and every structural change from the moment snow hits the ground.
We utilize the SLF SNOWPACK model, developed by the WSL Institute for Snow and Avalanche Research SLF in Davos, Switzerland. This is the same scientific standard used by avalanche professionals worldwide. It simulates the interaction of energy, mass, and momentum within the snow cover.
The model has been designed, validated, and refined over many years of rigorous field testing. Our snowpack models are updated hourly using both historical and forecasted data, giving you unprecedented detail about current and future snow conditions.
Our system tracks the evolution of snow grains, faceted crystals, rounded grains, surface hoar, and crusts deep within the pack. We model density, shear strength, and grain size at every layer, creating a digital snowpit for every mountain.
This level of detail allows us to understand not just how much snow is on the ground, but what kind of snow it is and how it will behave.
Based on a modified version of the SLF drift model, we calculate how wind moves snow from windward to leeward slopes. This forecasts depth variability across different aspects and elevations.
In the app, you'll see different snow depths expected across various aspects and elevations, showing you exactly where the wind has deposited or stripped away snow.
We implement Punstable, a machine learning model developed by the WSL Institute for Snow and Avalanche Research SLF. This model aggregates multiple snow stratigraphy features into a probability of instability for each layer within the snowpack.
Trained on hundreds of observed stability tests and validated with high accuracy (88% on independent data), Punstable correctly classified 69% of avalanche days and 75% of non-avalanche days across five winter seasons in the Swiss Alps. This provides critical, scientifically-validated safety information for understanding snow stability.
Why this matters: We don't just say "30 inches base." We know if that base is hollow, solid, or sitting on a layer of ice. We understand the structure of the snowpack and can tell you how new snow will bond with the old surface.
The more direct inputs these models have, such as weather station precipitation, the more accurate they become. But we go through extensive lengths to merge all available radar, satellite, and model data to get the most accurate picture possible.
This is where it all comes together. Our AI, SnowSense, reads the history, the forecast, and the physics. It "reads" the snowpack itself—density, crusts, grain types, and bonding.
Its purpose is simple: show you where the skiing is good right now.
By analyzing how new snow settles on the old surface, it generates aspect- and elevation-specific forecasts. Maybe north-facing trees are still holding powder. Maybe south-facing slopes have a breakable sun crust. It sees wind crusts, sun crusts, softening, and everything in between.
It's unlike anything else because it isn't just another AI using the same weather data. It's built on the full depth of available intelligence: physics-based snow models, storm structure, and terrain-aware analysis. SnowSense understands the snowpack.
The result: the clearest picture of not only how the mountain will ski, but where the best skiing will be.
All of this has been developed by scientists and snow professionals, forecasters, avalanche experts, and data scientists who have spent decades in the mountains. WinterScience is based on methods and data that have been successfully used operationally in professional avalanche forecasting and mountain safety.
It's built for those who demand more from a forecast, skiers, guides, patrol, and anyone who lives for the alpine. We bring decades of mountain experience into a single intelligent platform, merging expert insight with cutting-edge technology.
In general, all snowpack models, snow drift models, and instability analysis are available for all ski areas listed in our app. However, the inputs differ by location.
All areas have satellite precipitation coverage, but only some have weather station precipitation coverage and varying degrees of radar precipitation coverage. We also ingest weather station temperature and wind data where available.
Areas with a more complete set of inputs receive our Enhanced Hyper-Local AI weather forecast, but all ski areas get our proprietary AI forecast that is state-of-the-art and as good or better than the best of the rest.
Want to see what's available for your mountain? Check out our detailed coverage table.