Recent Articles

Early-Warning Covid-19 Alert for San Juan County and Other Seasonal Resort Communities

[Co-author of this article is Peter A. Morrison, Ph.D., RAND Corporation Senior Demographer (retired) and President, Peter A. Morrison and Associates, Inc. He is an applied demographer based on Nantucket Island, MA. He graduated from Dartmouth College and holds a Ph.D. from Brown University. Email:; website: Peter A. Morrison]


Local governing boards in San Juan County’s seasonal resort communities need immediate guidelines for balancing the need to protect their residents’ health and safety while sustaining local economic viability. As the County undertakes to restore economic viability, it must remain alert to any harbingers of an emergent COVID-19 contagious surge so that local communities can implement emergency responses. We offer a simple demographic model for signaling such alerts. These alerts can inform day-by-day decisions about emergency preparedness and triggering pre-authorized mitigation responses. This model can derive useful “yellow” and “red” early warnings from the overall statistical trend in daily reported counts of local “confirmed coronavirus cases” which most health departments now collect. These “early warnings” can equip any local community to detect and react to an impending risk, or to confirm the ongoing absence of risk as previously-imposed restrictions are gradually relaxed.


The San Juan Islands apparently are moored at a “pre-surge” stage of COVID-19 contagious spread. The telltale sign of an emergent surge (sudden explosive growth) is missing. On March 22nd, San Juan County had one confirmed COVID-19 case and by May 28th, it had only 15 cumulative total cases. Whatcom County, by contrast, reported a cumulative total of 48 cases on March 24th and 380 on May 28th.

With a continuing influx of summer visitors, the residents of the San Juan Islands face an increasing risk of a surge in contagious spread. By the time any emergent surge became fully apparent, the prospects for effective early suppression will have narrowed. Intervening just a week or two earlier could be very advantageous in restraining an embryonic surge.

Additional summer visitors increase the effective population of San Juan County by about 52 percent above its year-round resident population count (San Juan Department of Community Development, 2018 Report, Table 4, which can be accessed here.) The 52% summer surge is a gradual procession of different people coming and going. So many different faces from other places almost certainly would amplify the risk of an outbreak, especially on Lopez, Orcas, and San Juan islands which are the destinations of most summer visitors. Like their counterparts elsewhere (e.g., Nantucket, Martha’s Vineyard, and other island seasonal resort communities), the San Juan Islands must balance the need to protect the health and safety of all summer residents while sustaining local economic viability.

Our approach (outlined below) addresses an immediate urgent need: detecting early harbingers of an emergent contagious surge, then acting promptly and forcefully to suppress further outbreak. This approach integrates two elements:

(1) an early-warning alert, signaled by the statistical trend in daily reported counts of local “confirmed coronavirus cases,” presently compiled by San Juan County for the State Department of Health, forwarded on to the Washington State Department of Health and from there to the Johns Hopkins Coronavirus Resource Site; and

(2) a menu of pre-authorized responses, ready for immediate implementation upon any alert.

Akin to warnings that precede an approaching hurricane, this early warning plus response protocol is designed to detect and respond to an impending risk, using readily available data. The “early warning” alert element is calibrated and ready to use now. The “preauthorized response” element must be fleshed out as a spreadsheet of policy actions for re-tightening whatever local restrictions may have been loosened.


We have adapted a traditional demographic model to estimate the near-term relative risk of COVID-19 spread for island communities and other local geographies. This model is well-suited to the “adaptive response” approach to re-opening local economies, whereby policy-makers incrementally loosen and re-tighten restrictions to gradually restore a healthy local economy.

Our transparent projection model plays one key role: It signals a statistically defensible early-warning alert—and, by extension, window of opportunity for response. The model incorporates the “ARIMA” method (Auto Regressive Integrated Moving Average), a standard forecasting tool that is well-suited to San Juan County’s present quiescent stage. Our model (shown below) was deemed adequate by the NCSS Statistical system, the software we used to implement it.

The principal virtue of our model is its minimal appetite for data. A user need only assemble the ordered time-daily total cumulative COVID-19 case count. For this application, we acquired these data from the Johns Hopkins University Coronavirus Resource Center.

The principle limitation of our model is its narrow applicability to the “pre-surge” stage of COVID-19 contagious spread. The advent of any subsequent surge in coming weeks/months would invalidate its underlying assumptions, rendering the model inapplicable thereafter. In short, our model has a proper use and can serve an important purpose for quiescent island resort communities as newcomers arrive from elsewhere.

The headline figure above this article displays the actual 68-day time series of the cumulative case count data from March 22nd through May 28th, 2020 (daily red dots). The middle solid line extending on to the right (beyond the red dots) traces projected values from May 29th onward through September 5th. This line represents a plausible “best estimate” of what lies ahead absent any departure from the prevailing quiescence before a summer influx of outsiders. The dashed lines above and below the solid trend line represent the 95% lower and upper prediction limits.

The further ahead we look, of course, the less certain we are about the future. Consistent with that reality, the dashed lines trace a widening cone of uncertainty the further ahead one looks. (One virtue here: our probabilistic model cannot tell a lie!)


Another virtue of a probabilistic model: we can judge the limits of the quiescence observed thus far—the continuing absence of any contagious outbreak. The 95% lower and upper prediction limits accompanying the forecast afford an objective basis for triggering actions as observed counts are recorded, day after day (The day your temperature reaches 100.4 degrees, it’s a fever).

Suppose the cumulative case count trends upward over several days, veering toward that upper 95% bound. Any accelerated upward trend would warrant close observation and perhaps triggering a “yellow light” warning alert for preparatory actions. Any trend nearing the upper 95% bound would signal firm evidence of an emerging contagious surge and might trigger a “red light” alert to prompt further immediate actions. In short, each colored alert would serve to activate pre-authorized “tightening” actions (perhaps following earlier incremental “loosening” steps).


However these objective triggers may be calibrated, their applicability will be to actions that have been pre-authorized. The central aim here is to buy valuable time—even one week—to activate pre-approved “tightening” policies. Those policies must be formulated and pre-approved for implementation at once. By the time a surge becomes apparent, the prospects for effective early intervention will have narrowed. What this calls for is a menu of “Emergency Preparedness” actions (see illustration below).

The May 29th through September 5th projection (day 69 onward) is a baseline “forecast” of what to expect, given the trend in confirmed cumulative cases from March 22nd to May 28th. Starting May 29th, the actual reported daily cumulative number of cases should be tracked alongside the corresponding baseline forecasted number shown.

To illustrate, consider the period from July 16th to 23rd, where the cumulative forecasted count remains at 25 each day from the 16th to the 20th and then increases to 26 on the 21st and stays there through the 23rd. If a new case was confirmed on the 16th, the count would be 26. if two new cases are reported on the 17th, the count would be 28 and if on the 18th and 19th there was one more new case, respectively, the count would be 29 and 30, At this point, the “yellow light” is flashing. If by the 23rd, the reported case count hit or exceeded the forecasted upper limit of 33, then the red light starts flashing.

Ours is not a perfect early-warning system. Persons already infected may remain asymptomatic before becoming sick enough to warrant a test. Virtually all tests are imperfect because they issue both false negatives and false positives. Testing rates could change insofar as one test supersedes another (less accurate) test and because more test per capita are given. There also is a lag between infection and confirmation. Lastly, our system would be vulnerable to misreported or suppressed data, as has occurred in Florida and Georgia. Also, vulnerable to an outbreak are the county’s four assisted living centers or a cottage processing operation, neither of which are likely to be routinely exposed to a surge of summer visitors.

In conclusion, proper application of this model requires reliable data, local area knowledge, and an appreciation of the logic and inherent limitations of this early-warning approach. The model’s principal virtue (a minimal appetite for data) must be balanced by conscious recognition of its inherent limitations (the model is inapplicable once an actual surge commence). As with a traffic signal at a two-way intersection, turning yellow informs decision-making; nothing immediately thereafter should be relied upon.

Once a community has agreed upon the “braking” actions to be taken, it needs only a properly calibrated ARIMA model to forecast the trend in confirmed covid-19 cases relative to the 95% upper prediction limit to implement a suitable “yellow-then-red light” system to prompt immediate action.

Shown here is our actual forecast with 95% prediction limits, the basis for the graph at the top of this article.

Comments by Readers

Wynne Lee

Jun 04, 2020

Excellent info and ideas, given that areas of Whatcom County see large increases in transient populations during summer & holidays, including Birch Bay, Lake Whatcom and Lummi Island. I think we should ask our County Council, Exec and city leaders to develop plans to deal with potential outbreaks.


Michael Riordan

Jun 16, 2020

Thanks for doing this analysis, David and Peter. As an Orcas Island resident, I’ve been following the Covid-19 outbreaks in San Juan County, as well as in Skagit and Whatcom, and have written about them in these electronic pages, so I am very interested in your results.

One question: What is the assumption behind the slope of the linear rise shown in your graph? There must be something that drives this rise, based on our tourist influx this summer. And it looks like you assume that any such outbreaks will be kept under control, otherwise the solid line would curve rapidly upward, correct?

Finally, are you aware that SJC announced its 16th case yesterday, June 15? It’s due to a new employee hired at Roche Harbor, which claims it is doing testing of its employees.


David A. Swanson

Jun 16, 2020

The slight upward trend is driven by the historical data series  to which the ARIMA model is fitted. Keep in mind that the trend is meant to be a “forecast” only n the absence of any conditions disrupting the trend driven by the historical series. As such (per our discussion in the article), it serves as a baseline (to include, especially, the upper prediction limit) against which to evaluate the appearance of new cases and determine if any actions are needed.


Michael Riordan

Jun 16, 2020

That may not be a good assumption, as the SJC Health Department seems to have the original outbreak, which was due to county residents, well under control and probably had eradicated the virus in the county. The summer cases will likely be due to non-residents coming here from off island carrying the virus, as I believe the 16th case mentioned above was. So you can’t just use historical data to predict the future.


David A. Swanson

Jun 16, 2020

GIven that the 16th case is counted as a case to resident of SJC, the new empoyee of the county is now residing in the county (if not, the new employee would not be counted as a SJC “case,” but instead the case would have been counted elsewhere). As such, it is consistent with the historical data used to generate the ARIMA “baseline” found in the ARIMA results. So, these historical  data appear to be serving the probabilistic ARIMA “warning system” as intended.

The fact that the title of the article contains “Seasonal Resort Communities” in it should have been a clue that the warning system is aimed at outbreaks stemming from, quoting, in part, your comment, “..non-residents coming here from off island carrying the virus.” Was it not clear to you that this, in fact, is the entire point of the suggested warning system?

Until now I thought the only place in the Salish Sea area that existed under a dome was Edgemoor. Thanks for enlightening me about SJC.





Michael Riordan

Jun 16, 2020

This was not a “new employee of the county” from what I’ve heard (facts are scarce because of privacy concerns) but a new employee of Roche Harbor (a private entity) who had been hired from off island, hence more equivalent to a visitor or tourist than a current island resident.


Early-Warning Covid-19 Alert for San Juan County and Other Seasonal Resort Communities

By David A. SwansonOn Jun 02, 2020

[Co-author of this article is Peter A. Morrison, Ph.D., RAND Corporation Senior Demographer (retired) and President, Peter A. Morrison and Associates, Inc. He is an applied demographer based on [...]

6 comments, most recent 3 months ago

Bellingham Herald Reprints Seattle Times Article - But With Edit

By John ServaisOn May 29, 2020

​The graphic to right is an online screenshot of a Seattle Times article from Thursday, May 28. The Bellingham Herald reprinted the Times article today, Friday, May 29, in their morning print [...]

1 comment, most recent 3 months ago

Gridlock in the Making

By Dick ConoboyOn May 29, 2020

Yet another set of apartment buildings with a total of 118 studio units, called Samish View, is now proposed for the two parcels between the Elks Club and the Community Baptist [...]

4 comments, most recent 3 months ago

ACLU Files Suit for the Firing of Dr. Ming Lin

By John ServaisOn May 28, 2020

​This morning, the ACLU has filed a lawsuit in Whatcom County Superior Court against PeaceHealth, St Joseph Hospital, and TeamHealth for the wrongful termination of Dr. Ming Lin in March. [...]

2 comments, most recent 4 months ago

Would Adjudication Resolve Nooksack River Water Issues?

By Eric HirstOn May 19, 2020

How can we best resolve long-standing water-resource issues in the Nooksack River Basin? These problems–primarily too little water in the river and streams during the summer months–have been [...]

CityView Apartment Development - Update V

By Dick ConoboyOn May 18, 2020

An astounding 150+ comments have been received by the Planning Department regarding the CityView dormitory apartment complex. The initial comment period closed on May 8th, so the staff could begin preparing [...]

Corner of Elwood and Lincoln Soon to Be a Wee Tad Crowded

By Dick ConoboyOn May 16, 2020

As part of the pre-application process (Type II Process), Arbour North Architects hosted a public meeting for Samish residents, held on May 12th by way of a video conference call. [...]

2 comments, most recent 4 months ago

The Testing of Three Northwest Counties

By Michael RiordanOn May 14, 2020

Comparisons of the Covid-19 public-health responses and data from San Juan, Skagit, and Whatcom Counties reveal striking differences that may help these counties and Washington state chart a viable path [...]

Politicizing COVID-19

By Ho See-wingOn May 12, 2020

COVID-19 time is a time of contradictions. It’s a time of isolation, but also one of togetherness: we are isolated in necessary lockdowns, yet we are all being locked [...]

6 comments, most recent 4 months ago

The Pandemic And A Depression

By Guest WriterOn May 12, 2020

[This is a second article from our guest writer who does not wish to make his identity known. I refer to him as G, a retiree who moved to Bellingham [...]

3 comments, most recent 3 months ago

CityView Dormitory Complex - Please Don’t Approve This Proposal

By Alex McLeanOn May 10, 2020

Please don’t approve this proposal for the following reasons: 1) The Geology: The developed section under consideration, from reading the elevation demarcations on the site plan, has a drop of [...]

5 comments, most recent 4 months ago

MintTheCoin - Money for everyone

By Guest WriterOn May 08, 2020

[Our guest writer does not wish to make his identity known. I will refer to him as G, a retiree who moved to Bellingham over 15 years ago. He is a [...]

18 comments, most recent 4 months ago

The Lack of Containment Measures: Does It Constitute Senicide?

By David A. SwansonOn May 07, 2020

[This article authored by David A. Swanson with assistance from NWCitizen writer Dick Conoboy] Senicide, or geronticide, is the killing of the elderly, or their abandonment to death. Because the [...]

3 comments, most recent 3 months ago

Still a Dormitory, Still in Non-Compliance

By Guest WriterOn May 06, 2020

[Our Guest Writer, Linda Diebert, spent the first years of her career teaching young children here in Bellingham. She changed careers and became Transit Planner, Principal and Co-owner, TransitPlus, LLC, [...]

9 comments, most recent 4 months ago

WWU - Western - Dislikes Trees

By John ServaisOn May 04, 2020

​In a classic case of “cut down the trees to put up a parking lot,” Western Washington University (WWU) has for decades removed mature trees from its campus very quietly. [...]

9 comments, most recent 4 months ago

Monitoring the COVID-19 Pandemic in Benton County: The Good, The Bad, and The Ugly

By David A. SwansonOn May 04, 2020

The COVID-pandemic in Benton County has played out differently than in Whatcom County. The baseline forecast for its expected peak on April 29th was a total of 17,716 confirmed cases, which [...]

Whatcom County COVID-19: The Final Report

By David A. SwansonOn May 01, 2020

Modeling even simple aspects of the COVID-19 pandemic is more than challenging in the United States because of the sparsity of data.There is no comprehensive testing, and an active [...]

4 comments, most recent 4 months ago

CityView Apartment Development - Update IV

By Dick ConoboyOn Apr 26, 2020

Morgan Bartlett, the developer of CityView, the student rental dormitory project in the Puget Neighborhood, submitted his final application to the city and the city has deemed the application complete. [...]

5 comments, most recent 4 months ago

Fairhaven Flag To Remain At Half Staff

By Dick ConoboyOn Apr 25, 2020

The American flag located in front of Bellinham’s Fairhaven Library has been lowered to half staff to recognize those who have died as a result of the COVID-19 pandemic [...]

My mask protects you. Your mask protects me.

By Jane BrightOn Apr 24, 2020

“We’re all in this together,” means what exactly? The new PSA should be: “My mask protects you. Your mask protects me.” PSA announcements keep repeating the mantra, “We’re [...]

17 comments, most recent 4 months ago