A Retired NASA Engineer Calls Amazon’s Noise Study “NOT CREDIBLE”
Dr. John Lyver, a Gainesville resident with a Ph.D. in Computational Data Science and a Certified Safety Engineer, analyzed Amazon’s revised January 9th noise study and found it to be both “NOT CREDIBLE and NOT Defensible.” Read his full letter to the Warrenton Town Council explaining Amazon’s willfully flawed study and why the Council should vote to deny the SUP in February.
Mayor and Councilmen of the Town of Warrenton,
I was pleased to be able to attend and speak to you during the Warrenton Town Council (TC) Public Hearing on the proposed Amazon Data Center SUP on Tuesday, January 10, 2023. My concern is that you have only heard a very tainted single view, that of Amazon, and have not heard the real physics of the situation. I am writing this note to help you understand just what this proposed data center will do to the Town.
First let me reintroduce myself. I am Dr. John Lyver, Ph.D. and I am a resident of Gainesville, VA, but love coming to Warrenton to relax and shop. I am a volunteer and do not have any “dog in the fight”, so I am giving you an unbiased independent analysis. I spent 25 years as a NASA Engineering Manager working on leading and performing analyses of the safety of nuclear missions heading to Mars and beyond. During that time, I earned a Ph.D. in Computational Data Science and earned a Certified Safety Professional license (basically, a Professional Engineer in Safety). So the work I have been doing on the noise and safety aspects of the data centers is something of which I have many years of experience.
Below is some information that I feel is vital to your understanding of the noise implications from approval of the SUP. Please accept my apology for its length, but this is important. At the Public Hearing, I offered to meet with each of you and/or your staff members to provide more details and explanation.
Amazon submitted Noise Study dated January 9, 2023:
First, let me say that I was the reviewer that found the September Amazon noise study to be NOT CREDIBLE. The January Noise study has not corrected any of the misinformation and bad physics that was in the September Study. I find the January 2023 noise study to be both NOT CREDIBLE and NOT Defensible. Let me explain:
The same contractor, same “SoundPLAN” model, and very similar assumptions were used in both studies.
The noise study did NOT incorporate the Town Zoning Administrator, Mr. Walton's, guidance letter to Mr. Foote on December 16, 2022. In fact, they again analyzed for the wrong conditions, locations and measurement requirements. Nor did they analyze against the -10 dB(Z) and -15 dB(Z) corrections to the Town Zoning Ordinance noise limits. (BTW: Please pass my appreciation on to Mr. Walton for an excellent letter. It is probably the best I have seen in explaining noise ordinances throughout the region.)
The study cited the August 2022 SUP provided materials and did not update for the October and December design changes.
The study recognizes that noise from diesel-electric generators will increase the overall noise levels. However, does not cite the noise that they will produce.
The study states: generators will only run for a “limited time” but do not cite what that time is. At other Amazon sites in Northern Virginia, generators run for time periods beyond the cited ANSI S12.9-defined short duration period. Plus, Amazon regularly runs generators in a contiguous series, making the noise generation well beyond the ANSI direction for measurement periods.
The study does not cite references, nor do they cite verifiable data. For example, the study cites noise from a generic Trane Corporation brand chiller but do not cite the source or the noise data itself. Trane produces a wide variety of chilling systems and sizes, so it is impossible to verify the assumptions made.
When the study does cite noise readings, the study does not cite the distance where the noise was taken. This makes it impossible to validate the calculations. Noise decreases by a ratio of [(1/r)^1.5], where r is the ratio of the cited distance to the observed distance.
The study does not combine the noise sources on the property into a single noise level. The study only cites individual levels. Mr. Walton’s letter clearly states that total noise from the site is what must be compared to the Ordinance. When the noise levels are combined, they clearly exceed the Ordinance limits in nearly all cases.
The study cites a drop of 5 dB(A) to 10 dB(A) noise generation decrease at night for a building operating at steady state 24/7. This is not correct. Plus the Ordinance uses dB(Z), making the claim difficult to explain.
The Study does not follow commonly used scientific principles in plots or how topography interacts with noise waves. They have topographic low areas stopping sound and hills increasing noise in some plots. Same as in Sept report.
The study’s plots show the center of the isopleth ‘circles,’ not at a noise source, but rather, in the road on the north side of the building. This factor alone discredits the entire analyses as poor modeling.
As for the SoundPLAN model, it does not consistently use similar modeling between frequency octaves. The sound attenuation modeling per distance varies. Same issues as in Sept.
The background reading plots show that the modeling was done at 15-25 dB(Z) lower than what was measured.
The Warrenton Zoning Ordinance Chapter 11, section 11-3.10.2, Item #8 REQUIRES “Noise impact and abatement studies to determine potential impact on adjoining properties and neighborhoods.” There is no discussion of impacts due to noise.
Finally, there is no mention of the study for the data center’s impact or compliance with the Fauquier County Noise Ordinance, even though over 1,000 residents of Fauquier County (living outside the Town boundary) will be affected by the noise.
For these reasons, the January 9, 2023 noise study is NOT CREDIBLE and NOT DEFENSIBLE and should be withdrawn as the September study was withdrawn. At a minimum, it should be rejected by the Town as a valid submission as required for SUP submissions and the SUP submission declared as incomplete.
Enforceability of Town Noise Ordinance:
The TC members had a large discussion during the meeting about how the Town can effectively shut down the data center if the data center exceeds the Town Ordinance. May I direct your attention to Virginia Code Title 15.2, Chapter 9, section 15.2-980 " Civil penalties for violations of noise ordinances” which states: "Civil fines will not exceed $250 for the first offense and $500 for each subsequent offense.” Nowhere in this Section does it permit shutting down the facility. Additionally, since the Code of Virginia does not include noise limits within their building code ordinances enforced through zoning, I contend that should the Town attempt to shut down the data center simply for noise violations, Amazon’s plethora of lawyers will take the Town to Court within minutes. This lawsuit will cost the Town large sums of money with no relief for the residents due to excessive noise. I would strongly recommend that the Town Attorney provide you with a thorough analysis of civil penalty limits due to noise. I am not a lawyer, but have been told this directly by the Acting Prince William County Executive, Mr. Elijah Johnson, during discussions of noise enforcement in PW County.
Reduction of Noise from an Amazon Data Center in Manassas:
Let me provide you with a short recap of the noise generation from the Amazon data center cite on Tanner Way in Manassas, Virginia.
The site consists of a four-building complex with approximately 220,000 square feet for 3 of the 4 buildings and the 4th building, slightly less. The first of these buildings became “operational” early in 2022 (February, we believe).
On May 16, 2022, a complaint was filed with the PW County Police for excessive noise. The PW Police responded that evening and performed noise readings. They found that the noise level was at the PW County Noise ordinance limit (55 dB(A)) at the resident’s homes, about 1,200’ from the closest operating building. The Police did not take readings at other homes which are closer to the data center buildings, which are as close as 600’ away from the buildings.
In development of my noise model, I discovered that the Tanner Way buildings closely matched the noise generation profile with other similarly designed buildings. My model predicted similar readings to the Police's readings.
In July, meetings began between Amazon, PW County officials, and the neighborhood HOA to discuss the complaint (Great Oak Subdivision). Amazon representatives were invited to the site and acknowledged the noise levels were loud.
As the meetings continued, Amazon tried to wrap the air exhaust vents with a sound-absorbing blanket as a short term aid. After about 100 blankets were applied, the noise levels did not decrease, in fact, they increased.
Amazon is now working on re-engineering of the exhaust fans by adding wind bands and new higher power fan motors. Amazon has committed to a 5 to 8 dBA source level reduction, which the community believes will still not be enough to bring noise levels under the outdated (1989) Prince William County noise ordinance (a low bar). During this time, noise has now increased to over 65 dB(A), which is over 10 dB(A) above the PW County Noise Ordinance limit. Further, certification of this solution has taken nearly six months, pushing their phased installation to start in Feb 2023, and plan to complete in Feb 2024. The community will have lived with unabated data center noise for two years at completion of this unproven solution.
Additionally, I have used the Tanner Way noise readings collected to validate and sharpen my model. What I am finding is that my model is conservative and actual noise readings are about 2 dB(A) higher than I predicted. Additionally, I was able to ‘work why model backwards’ to discover that on May 16, 2022, the complex was operating at a level of about 10% of total electrical power. These results were presented at one of the neighborhood meetings and were not denied or contradicted by an Amazon Vice President.
(Note: The heat generated that must be removed is a function of the I-squared-R losses in the electric circuits in the computers and is a function of the amount of current (the “I”) and the circuits involved (the resistance “R”). The result is that heat generated is a function of the power consumed in the complex.
(Note: Let me invite you to travel to the Great Oak neighborhood to listen to the Tanner Way noise generation for yourself. You will be shocked at the tenor and intensity of the noise along the western end of "Winged Elm Drive". Personally, I have heard the complex’s distinctive noise from about 1-1/2 miles away, above the background noise of Manassas.)
Proposed Amazon Noise Studies:
Amazon will offer to perform noise studies for the jurisdiction as they have done in Warrenton. One study I have reviewed is the Amazon provided noise study for the Tanner Way complex that was provided to PW County and Great Oak HOA. After review:
It was found that one of the 3 meters taking readings was placed in a depression and did not produce representative readings.
One of the meters taking readings stopped working during the analysis and was never restarted.
The modeling used did not correctly average the dB(A) readings. Mathematically, the results were incorrect/invalid.
The modeling discounted noise levels for some background noises (they claimed were frogs), but did not provide adequate discounting methods for these readings. This introduced more error in the results.
The modeling did not report the power level of the complex at the time of recording. It was believed to be about 30% or less.
When I reanalyzed the raw data, my results showed that Amazon exceeded the PW County noise limits even though they were only operating at a small fractional of their planned operational power consumption.
Additional facts to consider:
Amazon defines “Operational” as when they can process the first electron off the internet. The noise comes from the cooling systems to exchange the I—squared-R heat losses from the computers to the environment. At 2-4 months after becoming “operational,” noise levels will exceed the Town limits and will only increase as operations increase. Amazon’s set of ‘promised’ noise studies from October 28 would be completed before the real noise begins.
The typical Amazon Data Center design, like the Amazon Tanner Way centers in Manassas, will exceed the Town’s noise ordinance in OVER 90% of the cases.
When the diesel generators are operating, the noise will increase further.
The noise from data centers can be heard up to 2 miles. That covers over 99% of the Town footprint.
Residents up to 1/2 mile from the proposed site will experience noise levels above the ordinance limits, strictly from the data center noise generated 24/7.
Approximately 1,000 Fauquier County residents outside the Town limits will have noise levels that exceed Warrenton’s limits, and exceed the County’s noise ordinance.
Excessive noise causes stress-related health effects like diabetes, sleeplessness, high blood pressure, nervous conditions, lack of concentration and so on. These effects will be especially hard on seniors, stay-at-home adults, residents of Poets Walk, and children.
Amazon is counting on your lack of expertise. Let me help you become better informed on the implications of approval of this SUP. Please let me know if you would like to discuss my analyses. I will be happy to meet with you and provide you with even more analyses on the noise predictions of the proposed SUP.
THANK YOU for your consideration and attention in this matter.