Update, February 5, 2026. With each validation month post, we often optimize precursors, round out missing tests and exercise other best practices. Here we’ve moved to the projected specific snow water content and deployed a different precursor set. And we’ve skipped ahead to simply demonstrate the raw error. In this new update, the color scheme has also been slightly adjusted to bring out the greatest contrasts. For these, the white contour line represents ‘zero’ error.
These are still a work in progress. Very much so, and in these preliminary stages, much experimentation with graphic options is tried. Over time we’ll reach a consistent template for skill documentation that works most intuitively with our products and their users.
More about this forecast/hindcast skill test: Specific Snow Water Content is a macroscopic snow parameter we selected as a baseline of study. In the future we will also project a related parameter that can be directly converted to inches and/or cm of snow. In the meantime, we start from this baseline and will use to evaluate the accuracy of each monthly snow projection.
December 24 2025: The ECMWF’s ERA5 monthly data is out for past November 2025. I’ve updated our method to apply 2024 data only towards projection of November 2025. for my part, I’m anchored to the Colorado River Basin and the connected states. Especially Arizona in this case, which saw some record rains in places last month. Several actual years of observations are mapped below culminating in the November 2025 specific rainfall at lower left, and the projection exercise for that same month, (again using only 2024 data) at lower right.
Another way to estimate the skill of the projection is through a residual map such as the raw difference between the two data sets.
A percent error map will be added, and these will be produced as standard output for all closed projections. All draft and all rights reserved.