Inspired by our prior research on efficient wave modeling, Nate and I have added high-resolution wave predictions in the Salish Sea to The Windline! Below is an example of the results for Bellingham. Wave heights are shown in color. Wave periods are shown by the white contours labeled in seconds and wave directions are indicated by the grey arrows. Currently we are updating these predictions every 6-hours for 48-hour forecasts using the most accurate wind predictions from Env Canada HRDPS.

So how are we creating these? First we need to talk about the region and why this was not a simple as you'd might expect. In the Salish Sea model resolution is important (e.g. see our prior post). Complex shoals, narrow bays and inlets, and highly irregular islands and coastline of the region means we need super high resolution models with grid sizes around 50-100m (160-320ft). Typical wave models at these scales require significant amounts of computer power. For example, it could take several days to simulate just a single day's predictions on a powerful multi-core desktop. Clearly this won’t work, the forecasts will be completely outdated by the time they are ready! And since renting more powerful clusters from Amazon or Google is not in our budget, we needed an alternative approach.

As researchers Nate and I worked on highly efficient wave modeling approaches that are 1000’s of times faster than traditional approaches (SWAN) while sacrificing only minor amounts of precision (0-5%, typically). Using pre-computed look-up-tables, linear propagation of offshore swell, and multi-dimensional interpolation we quickly generate high resolution wave estimates. For those detail-obsessed individuals (nerds like us) we refer you to the published article, Efficient modeling of wave generation and propagation in a semi-enclosed estuary. For everyone else I’ll pull out some highlights about the approach below:

First, the Strait of Juan de Fuca. Because we know, sometimes Fort Ebey really goes off when the conditions are just right. Below you’ll see offshore waves simulated from different angles below. Waves coming from 278-282 appear to hit Fort Ebey. Keep that in mind when checking observations at Neah Bay Buoy.I'll add that It takes about 3-4 hours for waves at Neah Bay to make it to Fort Ebey.

Wave period is important too. Long wave periods (15 and above) are refracted (turned) towards the shore more strongly and don’ttend to make it down the Strait without running into the North or South boundaries. Your best bet is around 13-seconds, these waves tend to arrive with more of their energy at Whidbey Island.

While offshore swell is probably the most fascinating, the local winds here drive the majority of wave energy. We model these two, separately with precomputed tables of waves for the region at 50-meter (160ft) resolution. These are created for all wind speeds, directions, and tides possible in the region. If you want to dive into extensive detail here you can always meander to Nate’s Masters thesis.

With the nerdy details out of the way, we hope you find these predictions useful. We are currently creating predictions for Bellingham and the Juan de Fuca basin. Head over to our Waves Page and let us know what you think and what you'd like to see next!

Nate and I have spent a lot of time with various weather products across the Salish Sea. In our previous lives, as USGS researchers we sought after long historical weather climatologies we could use for modeling waves and water levels. Most searches ended with the conclusion that archived operational forecasts from Environment Canada were best at resolving local winds. Yes, better than the National Weather Service. As wind-sport enthusiasts we have shifted our focus more recently to forecasts and continue to make the similar conclusions.

Earlier this year we saw some powerful storms come through in early January. Looking back we see Environment Canada’s(blue) outperforming Windfinder’s superforecast (orange). Now this is windfinder's proprietary super high resolution model that they run in-house. While it does do a better job predicting wind speeds below 5-mph, we are not so excited about its performance when it matters.

The above observations are from Locust Beach (48.773°N, 122.532°W) where the NSKA maintains a wind station well exposed to south blows. Env Canada’s high-resolution model) does a great job estimating the timing and magnitude of most of the events in this window. In comparison, Windfinder’s super forecast missed the peak conditions by 5-10 mph. If you are looking to go winging, kiting, etc, this is problematic because we know there is a huge difference between 12 and 22 mph winds. Because power goes as the square, 22 mph winds nearly 3x more powerful than 12 mph winds!. That is certainly going to impact gear selection

So why the difference in forecasts? While we don’t have all the details regarding model configuration and boundary forcing we do know that Env Canada runs their predictions at higher resolution, 1-2.5km compared to 5km by Windfinder. So let’s compare winds from models at different resolutions. Below we look at peak storm conditions during a serious event on March 10th 2016 modeled at 3 different resolutions.

At 12km (about the width of Bellingham Bay) wind speeds are forecasted at just half the speed of higher resolution models! While, at least in hindsight, it might seem obvious that high resolution models would result in improved predictions in our geographically complex region, I hope we have convinced you that resolution matters.

At The Windline we want to make it easy for you to find the most accurate predictions. In addition to having Env Canada’s models in our data pipelines, we have recently added the NOAA HRRR model which makes forecasts at 3km resolution across the US. As we collect more data we hope to revisit this analysis, stay tuned!