After our conversations last week, we decided that it would be best if I focus my attention on mechanistically explaining the spatial and temporal patterns of DO in the anthropogenic run (i.e. the long hindcast). Analysis of WWTP loading will be a small part that we can investigate a little later.

With this new shift in scope, I have started to dig into hindcast results from 2014-2019.

This week, I re-created many of the figure that I had previously generated for 2013. I discovered that 2013 had much less hypoxia than the other six years (suggesting that 2013 had not fully spun up). Sadly, I also found that our 10 - 50 m depth signal is not present in the other years :(

More details below.

Note: 2016 is missing from several of the figures below. 2016 is a leap year, and I could not always think of a simple way to deal with the extra day. I will fix this issue in the future.

2D histograms

A few weeks ago, we discovered that for 2013, the hypoxic DO minima only appeared in the 10 - 50 m depth range. This was an exciting finding! However, this result may have been a symptom of the 2013 year being “not-fully-spun-up.” Figure 1 below shows 2D histograms of depth vs. concentration of the DO minima for every lat/lon cell in Puget Sound on every day of the year. Here we observe that hypoxic DO minima commonly extend as deep as ~180 m. Maybe 180 m depth is the model limit that describes the deepest depth that detritus can sink before it gets respired in the water column. In any case, 50 m is no longer our magic number.


Fig 1. 2D histograms of depth vs. concentration of the DO minima for every lat/lon grid cell on every day of the year. Data from Straits omitted. 100 x 100 bins.


Figure 2 shows a 2D histogram of the DO minima concentration vs. yearday for every lat/lon grid cell in Puget Sound. From this figure, we can see that there are many, many more grid cells that reach hypoxic concentrations in 2014 - 2019, compared to 2013. However, the hypoxic season mostly spans the same May 1 - Nov 30 dates that we observed previously.

Totally unscientific comment: I also greatly appreciate how the data in each panel continues into the subsequent panel– there’s a great wavy effect going on.


Fig 2. 2D histograms of DO minima concentration vs. yearday for every lat/lon grid cell. Data from Straits omitted. 365 x 100 bins.


Note that I also took a look at the s-level of the DO minima. For 2014 - 2019, the DO minima is most commonly in the bottom s-level (consistent with what we found before in our 2013 data). For the remainder of the blog post, the figures have been generated using data from the bottom layer only.

Average bottom DO during hypoxic season

Figure 3 shows the average bottom DO in Puget Sound during the hypoxic season (May 1 - Nov 30). The first panel shows the average of the 2013 - 2019 average concentrations (an average of averages). The remaining panels show the difference between a given year’s average concentration minus the average-average concentration (apologies for such poor words– we can talk during our meeting). Blue indicates that the year had a higher DO concentration than the average-average, while red indicates that the year had lower DO concentration than the average-average.

2013 has a substantially higher average bottom DO concentration in Hood Canal compared to the average-average. 2013 also has higher DO in Whidbey Basin. The missing hypoxic depths in Figure 1 likely come from these regions.


Fig 3. Average bottom DO in Puget Sound during hypoxic season.


To remove the bias that 2013 introduces, I have generated the same figure with 2013 omitted (Fig 4).


Fig 3. Average bottom DO in Puget Sound during hypoxic season, 2013 omitted.


I am intrigued by the DO variability between the years. 2015 appears to have larger DO deviations in Whidbey Basin, while the 2018 deviations looks nearly uniform throughout Puget Sound. In 2016 the Straits tended to have higher than average bottom DO, but Dabob Bay tended to have slightly lower than average bottom DO. In the coming week, I hope to develop some working hypotheses that might explain these observations.

It’s also good to note that the interannual variability in average bottom DO seems larger than the change in average bottom DO due to WWTP loading (from our old 2013 results). I’m bookmarking this thought for now, and will return later when the 2014 natural run is complete.

Re-creating figures from Khangaonkar et al., 2018

For the 2013 results, I re-created a few figures from Khangaonkar et al. (2018). I have repeated this process for the 2014 - 2019 results.

Figure 5 shows maps indicating the number of days with bottom DO < 2 mg/L. Again, the first panel shows the number of hypoxic days for the average year (calculated after finding average bottom DO concentration). The remaining panels show the difference between the number of hypoxic days in a given year compared to the average year.

Again, I observed that 2013 had significantly less hypoxia cmopared to the other years, so I have simply omitted it from the following figure so it will not bias the average.

Notably, 2014 seemed to have fewer hypoxic days in Hood Canal compared to the average, while 2019 tended to have more. The other years all tended to have a mix of more or less hypoxic days across Hood Canal.


Fig 5. Days with bottom DO < 2 mg/L.


Figure 6 shows a time series of bottom hypoxic area in Puget Sound (with the Straits omitted). Once again, 2013 appears to underestimate hypoxia (not fully spun up). The 2015 hypoxic season seems to start earlier than other years, while the 2014 season seems to start later. 2017 has some strong, short-lived, peaks in hypoxic area. What could be triggering these peaks?


Fig 6. Time series of bottom hypoxic area (DO < 2 mg/L).


Next steps

Now that I have a sense of what is going on in the 2014 - 2019 results, I can begin to develop hypotheses to methodically test (and hopefully explain mechanistically) why we observe the above spatial and temporal hypoxic patterns.

I will also fix issues with the leap day in 2016, and integrate this year into the rest of my analysis.

Course registration

Course registration for Autumn 2024 has begun. After the current quarter, I will need 6 more credits to complete my degree requirements.

During our last course-planning discussion, we came up with a list of three courses to take (each 3 credits):

  • OCEAN 587: Fundamentals of Climate Change (Autumn)
  • Engage Seminar: The Science Speaker Series (Winter)
  • CSE 412: Introduction to Data Visualization (Spring)

I haven’t signed up for OCEAN 587 yet for two reasons:

  1. There’s a chinese class I’m interested in taking which conflicts with OCEAN 587. Since CHIN 111 is the first in a series, I’m reluctant to miss it.
  2. Should I be thinking about quals for Autumn quarter?