Powder Pipeline

An end-to-end data pipeline for skiers frustrated by snow conditions.

NOTE: I'm currently reviewing code and syncing to Github as I go. Until this note is removed, the tool will not function.

The Problem This Package Solves

You're an avid skier. You intend on getting as many days on fresh snow as possible. The question that you pose year-after-year is: "Which pass should I buy? Where should I plan my trips?"

The problems faced here are:

• Each pass provider (Epic, Ikon, Indy, Mountain Collective) offer different price points, early bird pricing, blackout days, etc...
• Every year we face challenges with climate change, El Niño/La Niña cycles, different opening and closing days for resorts, you name it.
• Every season pass has dynamic pricing every year, and there's no readily available historic record of pass prices (as of the writing of this README).
• Each pass has diffent resorts associated with it.
• Just buying EVERY pass invariably ends up wasting money unless you have a season like I had two years ago and you can eek out ~100 days of skiing. I don't think every working professional can do that.

The Solution

The solution is to gather as much information as possible from open sources. We can then create an output table and perform some analytics to get us the final ROI: # Days/ $Dollar for each pass.

The Approach

Normally, I'd call this a methods session, but the intended audience would be an avid skier, so it's an approach.

Setup

Clone this repo with

git clone https://github.com/kwierman/PowderPipeline

Then, edit the example config in example_config.yml to suit your system. Install in a venv using your tool of choice, and you're off to the races!

1. Use the internet's Way Back Machine to fetch the pass prices

Pass	Operator	Since
Epic Pass	Vail Resorts	~2008
Ikon Pass	Alterra Mountain Co.	2018
Indy Pass	Powder Alliance (independent)	2019
Mountain Collective	Independent co-op	2012

This should be relatively straight-forward, but bot mitigation tends to be a problem. The good news is that I have some experience using Playwright to get past this, which brings us to a brief aside on ethics.

Legal & Ethical Notes

• This tool uses the Internet Archive (Wayback Machine) for historical data — a public resource explicitly designed for archival research.
• Live scraping respects a 1.5-second delay between requests (REQUEST_DELAY).
• Data is used for personal/research analysis only.
• Always check a website's robots.txt and Terms of Service before scraping.

This should output a CSV with the following schema:

Column	Type	Description
pass_brand	str	Ikon / Epic / Indy / Mountain Collective
pass_name	str	Full product name (e.g. "Epic Local Pass")
season	str	"2023-24" format
price_usd	float	Full retail price at launch
early_bird_price	float	Pre-sale / early-bird price if available
early_bird_deadline	str	Date by which early-bird price expires
num_resorts	int	Number of participating resorts
blackout_days	int	Number of blackout days (−1 = no blackouts)
unlimited_days	bool	True if unlimited access
limited_days	int	Number of days if access is limited
resorts_list	str	Comma-separated resort names (when available)
source_url	str	URL from which this record was scraped
snapshot_date	str	Wayback Machine capture timestamp
scraped_at	str	ISO timestamp of when the script ran
notes	str	Free-text notes / anomalies

This is accessible from the CLI with the following command:

powderpipeline scrape ski-passes

In the future, we can replace this with two commands, one for backfilling the data, and one for getting the current price, and run this pipeline once a year on automation.

Before running any commands, reference, the setup guide above.

2. Obtain a list of resorts and which of the big passes (if any) they are associated with

Once again, we're going to need to do some web scraping. We can also use this step to fill in information regarding the location of each resort. The Python Package, GeoPy can help us out there. Here, we can reference open information to obtain an output csv with the following format:

Column	Type	Description
pass_brand	str	Matches "pass type" above.
name	str	The name of the resort. We can treat this as a primary key.
latitude	float	Location latitude, for finding snow depth later
longitude	float	Location longitude, for finding snow depth later
base_elevation_ft	int	The elevation of the base of the resort, in feet
summit_elevation_ft	int	The elevation of the summit of the resort, in feet
state	str	The name of the state for each resort. This can be used for later analysis
scraped_at	str	ISO timestamp of when the script ran
notes	str	Free-text notes / anomalies

I'm choosing to put this in csv format for the reason that this should not be a very large dataset and therefore can be safely housed in memory.

I intend to put this step behind a CLI with the command:

powderpipeline scrape ski-resorts

3. Get Snow Depth for each resort and additional information.

For this, meteo provides a rate-limited API which we can use to house this information. I'm going to store this in DuckDB so that I can slowly fill this. This should output a table of the format:

Column	Type	Description
ski_resort	str	Matches "name" in the ski resort data.
date	datetime	The date at which this measurement was taken.
base_snow_depth_ft	int	The snow depth at the base, in feet
base_snowfall_in	int	The snowfall for this day, in inches. Should be aggregate over the 24 hour period.
summit_snow_depth_ft	int	The snow depth at the summit, in feet
summit_snowfall_in	int	The snowfall for this day, in inches. Should be aggregate over the 24 hour period.
scraped_at	str	ISO timestamp of when the script ran
notes	str	Free-text notes / anomalies

This should run behind two commands:

powderpipeline scrape snowfall-backfill

This should be run as a backfill to get historic data.

powderpipeline scrape snowfall

This should be run under automation daily.

4. Run analysis

All of these individual datasets can be analyzed individually, but it's far more useful to combine these into a single dataset for later analysis. This should be output into parquet for analysis with Dask using the command:

powderpipeline analyze

5. Run Visualization

Likewise, there is a Dashboard built with Plotly Dash, that you can bring up with

powderpipeline viz

References

Data Model	Source Name	Link
Snowpack	USDA	Link