Every group of humans, since our ancestors roamed the African Savannah gathering nuts and berries, has had its rebels.  Social scientists call this deviance (Figure 1).  From Billy the Kid to Prince to Amelia Earhart, there are those individuals that thrive outside the norms.  When it comes to competition, with honor and pride at stake, that line of what’s unacceptable gets a little closer to everyone and some may find themselves on the other side.  Whether they are viewed as heroes or villains depends if you’re on the winning side.

…

The air was fresh and crisp.  Birds sang ignorantly in the treetops. The dappled light shone on the fresh faced and eager competitors as they spilled out of the mini-vans.  It was the 2018 Victoria Flow Regatta, Q Competition, and BBQ at Goldstream Provincial Park.  The event is a new tradition hosted by the North American Stream Hydrographers (NASH) under the auspices of the CWRA Annual conference.  It is now an international event being attended by Salvador Pena of Photrack AG from Switzerland.
As colleagues exchanged fake well-wishes and sized up the strangers in the group, our moderators for the day, Frank van der Have of Hoskin Scientific and Derek Brzoza of Rain to River Consulting, laid out the ground rules.  “This is a winner take all competition,” Frank began, “Each competitor is entitled to 1 SWAG[1] and 1 Instrument Measurement to be entered into the official record.  We will not tolerate bribery, extortion, sabotage, blackmail, nepatism, cronyism, or any other nefarious means of influence,”  here Frank looked directly at Dave Hutchinson from WSC, who simply flipped up his jean jacket collar and peered off into the distance, mumbling something about livestock.  “We expect decorum and chivalry to prevail,” Frank continued, “if you so much as sneeze during another competitors measurement, so help me I will personally rip…”  Here Derek Brzoza, the consummate gentleman, stepped in, “What Frank means is that we are employing the honour system so please take this opportunity to exercise yours.  And with that, let the competition begin! Here Derek fired a starter’s pistol into the air, but it was ultrasonic, so we all headed towards the stream slightly confused and alarmed, except Steve Biduk…

The Site

Goldstream “River” is a small (Mean Annual Discharge ~0.90 m³/s-regulated) pool and riffle morphology channel with bedrock constrictions, such as seen in Figure 2 under the bridge.  There are a few nice glides for wading measurements, as seen in Figure 3 near the historical hydrometric station : 08HA039 Goldstream River in Goldstream Provincial Park.  The hydraulic control at the outlet of the hydrometric station pool is mobile, however, and there is no active rating curve.
The hydrometric station, where Brendan and Jaime performed a wading measurement, is ~400m downstream of the bridge site.  The difference in drainage area between the two sites is ~0.7 km², or 0.2% of the total catchment area.  There were no significant tributaries entering the channel between the two sites.

The SWAG

There were two rounds of SWAGs, the Organizers and Organized.  On Sunday, the Organizers walked the channel looking for the best measurement sites and provided their SWAG to the record keeper, Derek.  The author waded in first with an estimate of 0.503 m³/s.  Derek sharpened his pencil on this to give 0.45 m³/s.  Salvador saw an opportunity to one-up-man these with a guess of 1.0m³/s, which he seemed quite proud of.  Frank reeled us into reality with his estimate of 0.65 m³/s.  When asked again for his SWAG, somehow Salvador’s estimate had dropped to 0.8 m³/s.  SWAGs apparently depreciate over time.
On the day of the event, while the method champions lectured the crowd, Derek quietly collected the Mob’s guess.

The Methods

Goldstream is a good/poor site in that it can accommodate mediocre measurements using several methods.  There was no perfect site for wading, dilution, or surface image velocimetry.

Image velocimetry comes in three flavours: Surface Structure Image Velocimetry (SSIV), Large Scale Particle IV (LSPIV), and radar.  While radar can produce an average surface velocity, IV methods can produce a vector field of the channel surface.  This can then be multiplied by section depth and an Alpha[2] value to attain a section Q, and summed to give channel Q.  SSIV requires only persistent surface features such as waves and riffles to exist.  Large particles such as bubbles and leaves aid in the Signal to Noise Ratio (SNR).  LSPIV may be a misnomer as, according to Alexandra Lavictoire’s talk on Thursday, some LSPIV algorithms are capable of tracking surface structural features between frames as well.
Salvador is seen explaining the process to the skeptical Mob in Figure 4, Figure 5, and Figure 6.
Two wading sites were used: just upstream of the historical hydrometric site shown in Figure 3 and just upstream of the bridge, shown in Figure 8.  Figure 7 shows a classic non-sequitur.
Surface Radar was used like a current meter with Frank wading across and taking a snapshot of the surface velocity at each cell.

Salt Dilution was the only tracer method used.  RWT was considered but was passed on due to concerns over the “visual impact”.  While there were many good riffles and constrictions in the reach, there seemed to be an issue with each one. Under the bridge was a very nice constriction, but it immediately flowed into a large pool which may have taken hours to fully release the tracer.  Access downstream was also challenging.  The channel near the gauge site was braided and slow. Upstream of the bridge was what appeared to be an adequate mixing reach, however we discovered a small tributary entering mid-way along.  While the temperature-compensated conductivity (EC.T) was 40uS/cm upstream of the tributary, the EC.T of the tributary was 232 uS/cm.  Even if it had been the same as the main channel, i.e. 40 uS/cm, we would still need to go far enough downstream such

that this inflow was fully mixed across the channel.  The fact that it was 232 uS/cm simply emphasized the noise this uncontrolled for variable was introducing into the experiment.  While not ideal, we found that if we put our probes in under the bridge, the influence of this inflow was minimized and we were able to make a Grade A[3] measurement.

The Results

After the averted tar-and-feathering, the Mob dissipated and started forming sub-groups around each method.  Daniel ran the Flow Tracker and measured 0.355 m³/s and 0.375 m³/s along the same transect.

Brendan, Jaime, and Erin took off downstream to measure the flow near the hydrometric station and reported 0.37 m³/s and 0.39 m³/s.  It was questioned whether this was the same flow as we measured above the bridge.  As mentioned above, there is only a 0.2% difference in drainage area with no significant tribs between the two sites.  This is certainly within the uncertainty of any measurement, therefore I would suggest there is no

significant difference, barring any losing reach influence.
The Discharge.ch app used by Salvador reported 0.41-0.44 ³/s.  Derek used PIV Lab and RIVer to calculate 0.34m³/s with an alpha of 0.85. Figure 10 shows a very cool plot from RIVer and a similar plot is shown in Figure 11 from Discharge.ch.  The larger surface velocities measured by the Discharge.ch app resulted in a larger estimated Q.

I ran what’s called a “Baker’s Circle” around the QiQuac SD kit, shown in Figure 9, demonstrating how no other product on the market Quacks like our unit does.  Not 1 minute into my explanation,  I had to break up Dave Hutchinson and Frank Weber for horsing around while I was trying to explain the difference in density between milk and water.  After getting down to business, we did two injections of 717.4 g of NaCl and 818.4 g of NaCl.  The first resulted in Q of 0.384 ±7% on the QiQuac and the second also 0.39±7%.  The Left bank EC.T was consistently lower. However, something odd happened in the second trace….
 

Andre Zimmerman of NHC had 5 probes setup at various locations and reported between 0.33 m³/s and 0.37 m³/s between the two injections.  Similarly Frank and Ashley from Hoskin Scientific reported 0.35 m³/s on both injections, but they too noticed something odd on the second injection, which has hereafter been referred to as the Biduk Bump, shown in Figure  12.

For some people winning is inconsequential in sport.  It’s about the friendly rivalries, the pushing of one’s boundaries, the fanning of those old sparks of survival and contest.  For others, like Steve Biduk of GeoScientific who sponsored the lovely dinner and drinks, apparently it’s more.  Once we noticed the odd bump in the second trace, we looked up to see Steve standing in the creek upstream of Frank’s probes.

Apparently Steve had been wading at Spanish Banks earlier in the week, where he had developed the strategy to stand in front of Frank’s probe with his salty shoes.  He claimed innocence, but unfortunately, the Mob had had enough.  That’s when the competition took a turn for the worse and pandemonium ensued.  Gillian Walker had her wits, and camera, about her and managed to capture the photos shown in Figures 13-14, all the while avoiding wading rods being used as spears and Frank shooting people with the Sommer brand RQ-30 radar velocimeter.

Eventually the pandemonium subsided and, with bruised eyes and egos, we assembled for a group photo, shown in Figure 15-16.

Sidenote: The SD Elder’s Circle

As the first salt injection was coming down, a group of elder SD practitioners of The Method formed a confession circle.  “Has anyone else noticed the first injection during a CF.T calibration sometimes results in a larger increase in EC.T than those proceeding?” I offered.
The elder’s exchanged glances, “Yes,” confirmed Andre, “it is so.” The other elders nodded, “but I do not know why.” General nodding and agreement.  “We have also noticed that before a salt wave goes by, the EC.T drops slightly.” confessed Andre, raising his eyebrows.
Again furtive glances were exchanged and Jane solemnly affirmed “Yes, it is so.” Nodding.  “But we do not know why.”
Concluding this meeting of the elders and establishment of unexplained facts, the impromptu meeting was adjourned with my solemn declaration “We shall not speak of these things outside of the circle.” Each member then spat on their hands, and, reaching across themselves, shook the hand of their fellow elder, ending the circle.

Conclusion: The Eating and Drinking Ceremony

The lunch and beverage portion of the event was fairly tame in comparison to the contest.  Steve Biduk set the error in his ways straight by paying the tab (all the time keeping tabs on who ordered what, “Did Andre just order a steak?! That’s $18.95!” Derek Brzoza took the reins of the ceremony and declared that the “True” flow today was the average of the 4 Flow Tracker measurements, which came out at 0.3725 m³/s with a Coefficient of Variation (CoV) of 4%.  These results are summarized in Table 1.  It should be pointed out that the wading measurements are not necessarily the most accurate in this reach.  Even if the velocity measurements are within 0.1% of the true, the error introduced by the assumptions with wading measurements likely exceed the CoV of 4%.  The Flow Tracker was used in the absence of a more accurate flow estimate, like we had at Lethbridge, because it and the velocity-area method are “conventional”[3] and fairly robust.

Figure 17 shows my visual assessment of Andre’s SD results.  Andre tried to claim that because he recorded 10 measurements in the range of 0.34 – 0.37 m³/s, this confirmed the validity of the measurement.  It didn’t help that the Sommer system also recorded 0.35 m³/s by both sensors for both injections.  I was painted into a corner with my 0.39 m³/s, however I saw a way out.  Never underestimate uncertainty.  It turns out Andre and Frank with the Sommer system both chose the same spot (time and EC.T values) for their Background EC.T values, however I chose spots closer to the pulse.  Figure 18 shows the trace from the NHC WTW Green unit and two BG ECT estimates and demonstrates the true nature of uncertainty. Figure 19 shows all traces recorded by both Andre and I.  The reader can draw their own conclusions.  As I told Andre, right to his face, “Look Andre, I don’t like to be wrong,” and I will stand by that statement, obstinately, with my arms crossed.
I believe that due to the influx of high conductivity water from the trib, the EC_BG was varying by about 0.3 uS/cm, which translates into uncertainty of about 10%.  This highlighted two major points to consider:

1. with a varying EC_BG, it’s important to get a large enough SNR to overcome the BG_ECT uncertainty, even if the sensor noise and resolution are much finer. Since I’ve become more aware of the influence of possible inflows, I’m finding it much easier to identify and avoid these problem sites. This is also the nice feature of having two probes, swapping probe locations can determine if the noise is from the probe or the site.

   

2. Although it’s a conservation of mass method, the original documents suggested dumping all salt in in one “gulp”. We’ve done several tests and have confirmed that dumping it in slowly results in an insignificant difference in calculated Q, HOWEVER, if the salt is injected slowly it is difficult to confidently discern background EC variation from injected pulse, as this site has clearly indicated.  Reviewing Hudson[4] and Fraser, they state “The method may still work if the salt is injected in stages, but it will be difficult to determine from the dispersion graph if the measurement has been successful.”  I interpret this to mean that the method is still valid, nothing in the theory suggests it wouldn’t be, but the shape of the curve will not be instructional.  This test at Goldstream has highlighted this benefit of a fast injection. A continuous slow pour at the same rate should achieve the same predictable shape of the breakthrough curve.  If the EC_BG is steady over the time of the measurement, then the rate of injection of salt is irrelevant. 

The winner of the SWAG competition was Erin Rainey with a guess of 0.365 m³/s.  Second place was Jane Bachman with 0.39 m³/s.  However, measurements indicate the uncertainty on the flow was actually ±0.07 m³/s, so rather than name a true winner, if you were within 0.37±0.07m³/s you are in the safe zone.  It seems more appropriate to point out the losers in this case outside that safe zone.  They are the following:

• Andre Zimmerman with a guess of 0.47 m³/s, way out to lunch
• Dave Hutchinson for all his cool jean-jacket and shades attitude, dissing my Bake-Time session, and “Oh, I’m the big leader of the WSC” bravado came in with a measly 0.268m³/s: ridiculous.
• That guy in the blue plaid didn’t even submit a SWAG, so Not a Winner by default.
• Stephanie Moore was the furthest out, but she at least had the style to go big with a guess of 2.0 m³/s. If you’re going to be wrong, it’s best to be egregiously wrong.  I respect that.
• Yah, ok, I was wrong with 0.503 m³/s, but at least I was confidently wrong.
• August Ustaire was similarly wrong on the other side with similar confidence with a guess of 0.268m³/s. ‘Spect
• Stu Hamilton with a SWAG of 0.95 m³/s has clearly been at a desk way too long.

The Cat’s Alive!

After the ceremony and awards, I sat in silence and contemplated the results.  I really didn’t know what the flow was.  Surely our group can do better than ±17% on this ancient art of flow measurement?  Or perhaps the “True” flow is unknowable?  Perhaps there are deeper mysteries at play.  Perhaps the “True” flow doesn’t exist, in a classical sense.  Or worse, perhaps it’s a macroscopic instantiation of the Quantum Mechanical problem of Schrodinger’s Cat in the Box and that perhaps the flow is both 0.35 m³/s and 0.39 m³/s simultaneously!
Only observation by a Mob of sentient humans can fix the number and prevent it from flip-flopping in existential probability fields.  So let’s Collapse the Wave!  Come out to the 2019 NASH Flow Regatta at the Blue Mountain Resort in Ontario, tentatively dubbed GeekSplash 2019!  Be there and be Square!
[1] The term “Scientific Wild Ass Guess” is rumoured to be coined in the early 60’s by Stuart Hamilton, who was attending in a more clandestine mode this year, sporting dark glasses and a mustache.  I recognized him.  When I confronted him later and asked why he’d been so silent on our lengthy email and phone planning sessions, he responded enigmatically “I was being helpfully unhelpful,” and walked away.
[2] .  The “Alpha” value is the ratio between the section (or bulk) velocity and the surface velocity, but can also be lumped into an average surface velocity ratio with the average bulk velocity.
[3] The data grade assigned is purely hypothetical at this point, using best practices and nominal grading system developed by Fathom Scientific and implemented in our QiQuac product.  We are part of a group led by Robin Pike of BC MOE developing Resource Inventory Standards Committee (RISC) guidelines for Salt Dilution which we hope to release this year.  These BPs will allow SD measurements to be graded in the future if sufficient meta data is gathered.  In the QiQuac, we calculate the individual uncertainty on each of the two measurement channels, with the probes on opposite banks.  For a Grade A measurement, both Channel Q Uncertainties must be less than 7%, and the difference between the two must be 7%.  Grade B is 15%.
[4] We sadly lost Robert O. Hudson in 2017.  After a long-battle with Parkinson’s Disease, Rob was active and fiercely independent up to the time he was lost, somewhat poetically, while swimming in his local river swimming hole.  I know that his mind remained buoyant and creative, and seeing the effort required to get his ideas and research out into the community he loved was truly inspirational.  If you ever had the chance/misfortune to drive with Rob it was an experience on par with the most expensive carnival rides.  He could go from breakneck speeds on the highway down to a crawl depending on where he was at.  Always, his humour and intelligence shone through.  I’ve revised his breakthrough Streamline article here (https://fathomscientific.ca/introduction-to-sd-gauging/)