“I just returned from Nepal!” I offered loudly to my colleague, Derek Brzoza, aprapos of nothing. “Boy, I’m really jet-lagged, but what a trip!” I complain-bragged.  Derek (shown in Figure 1) organized the Kelowna EFN Flow Regatta held at Mission Creek on Oct 19, 2018.  Derek sighed.  I could hear him thinking if he should take the bait; wisely he didn’t.  If you are looking to cut back on the number of colleagues that will speak to you, and lower opinions of you in general, these are great openers.  Since my encounter at the Boston Pizza in Kelowna, I’ve been trying not to indulge in this type of behaviour, instead waiting expectantly for people to ask me where I’ve been, then launching into a 20-25 minute narrative of my trip, complete with photos and insightful anecdotes on the failings of western civilization.

Thankfully I have a blog post and at least 3 avid readers. I’ve cut back the number of selfies from about 20 to just 1 (shown in Figure 2) . I’ll try to make this seem professional and hydrologically related.

Nepal (I just got back form Nepal. Boy! Was I jet-lagged).

Soo, the reason I went to Nepal was two-fold:

1. Some recent QiQuac customers from the US are developing a hydropower project on a glacier lake. This is brilliant for a few reasons:
   • It lowers the risk of a GLOF (no explanation necessary. It just sounds like something that should be avoided, like a large Muppet monster. Beware the GLOF)
   • It generates clean-green power.
   • It promotes economic development in a developing country.
2. My Little Buddy Nick Page said to me, he said “You have a choice: 1. You can go to Nepal or 2. I will kill you, assume your identity and I’ll go to Nepal.” Pretty clear choice.

I offered my assistance setting up a quality hydrometric program in exchange for having my expenses paid. So while it may not have been a smart financial move, the cost of living in Nepal is so low by western standards, I’m sure I spent less there than if I were working in North Van, thereby making money.  So it was kind of an investment?  Like I said, I’m no financial wiz.
So after several 12+hr flights, a 12 hour bus-jeep ride along the Annapurna “Highway”, and a 16 hour hike from 1,870mASL to 4,000mASL, we arrived at site.  Although it was many hours, the actual cost was less than an hour of helicopter time in BC. Go figure.
Setting up the hydrometric program consisted of three components (this is the hydrological relevant bit).

1. Establish a reliable and accurate Salt Dilution Instream Q (SDIQ) section for use with the QiQuac,
2. Establish reliable Benchmarks and Pressure Transducer installations,
3. Conduct a bathymetric survey of the sub-glacial Lake.

   

This is a tricky site, shown in Figure 3.  Water exits the lower sub-glacial lake and enters a very turbulent cascading falls section, but it does not constrict the flow.  We followed one of the many small side-channels (shown in blue) in the lower alluvial/colluvial fan upstream and it entered the falls below the lake, so this is a losing reach.
There are several side channels shown in Figure 3, but we’ve chosen to only focus on those labelled Side Channel A (SCA) and Side Channel B (SCB) in this study.  SCB appears to be a combination of slope runoff from the small Drainage Area (DA) downstream of Dona Lake, along with flow leaving the mainstem at the Falls and flowing through loose alluvium/colluvium to rejoin the mainstem at SCB-Confluence.  SCA is a more constrained flow bypassing the mainstem from the Bridge site to SCA-Confluence.
We first examined potential measurement sites.  We decided that the Falls was too braided and featured too few constrictions and too many points of departure to be considered for SDIQ.  The Gulley, downstream of SCB-Confluence was considered an excellent mixing reach where all flow is consolidated and there are several constrictions.  The drawback of this site is the difficulty (danger) of access, the contribution of flow from the DA d/s of the lake, and it also seemed to be windier and colder than upstream of the SCB-Confluence, which can complicate a measurement and lead to errors.
We began tests on Oct 7, 2018 upon arrival at site.  Measurement in the Gulley was with both probes on the Right Bank (RB) and separated by about 10m, which was the safest option that afforded complete mixing.  We injected u/s of SCB-confluence. This test resulted in a Grade A[1].  We then measured the flow in SCB to be less than 1% of the flow in the mainstem.  Given an inherent 2s (95% confidence) uncertainty of ~6% in the mainstem SDIQ, this was not a significant contribution.  We performed a measurement injecting d/s of SCA-Confluence and measuring u/s of SCB-Confluence and achieved a Grade B measurement due to incomplete mixing over this shorter reach.
The next day, we injected at the Bridge and measured u/s of the SCA-Confluence.  This resulted in a  Grade C measurement, again due to incomplete mixing.  We then moved the probes to u/s of SCB-Confluence, on LB and RB and achieved a Grade A measurement.  To confirm an accurate measurement, we moved to our proven Gulley SDIQ site and measured another Grade A measurement.  This confirmed that injecting at the bridge and measuring u/s of SCB-Confluence (indicated by the green line in Figure 3) was an adequate mixing length, for this flow.  It may not be a sufficient mixing length for larger flows and probes placed on opposite banks will confirm complete mixing.

Kelowna EFN Flow Regatta

“Just returned from Nepal, sure need this coffee!” I offered loudly to the woman behind me in the Starbucks line.  A subtle frown flashed across her forehead.  “Oh, you must be tired,” she replied, hesitantly.  “Well, they’re 12:45 ahead of our time zone, plus they’re a day ahead, so it’s like I’m still in the future, but it’s nighttime!” I said cleverly.  The woman averted her eyes to her phone, as did everyone else in the line, like synchronised swimmers diving underwater. “What can I get you, sir?” the Starbucks partner asked, “Well, it’s such a relief to finally get some good coffee, there’s no Starbucks in Kathmandu…”
I managed to make it to Kelowna, only swerving off the road a few times (those road divots are brilliant).  Nothing like waking up from that buzzing sound to get a much needed hit of adrenaline.  I made the mistake of not bringing snacks so I had to search around behind my seat for something to eat to stay awake.  I actually found some Dollar-store Glossette raisin knockoffs that had a nice zing to them.  I arrived at my lovely Pure Earth Cabin from AirBnB hosted by Brante.  Nestled in an apple farm in the Westbank, it was the perfect autumn setting with crates full of fresh crisp apples, leaves on apples trees starting to turn orange, farm and kids equipment scattered about.  Charm was up around 108%.

After settling in, I headed out for some dinner.  Found a Boston Pizza still open and, after some small talk about travelling and coming round to Nepal with my waitress, I settled into some serious work. Sidenote: RollerGate RC.
At about 11, when the local bars closed, a drunken young man staggered in, singing a Bob Seeger song, and ordered some Gin, and by Boston Pizza policy, some food.  I finished my meal and went to pay. “It’s so inexpensive in Nepal.  This pizza would have cost about 5$.  You can live on $10/day if you really tried.  And the craftsmanship, spectacular.  In all my travels, I was most impressed with the Nepali craftsmanship.” I observed astutely and loudly.  “Now, when I was in New Zealand, “ I started, but was abruptly grabbed by my collar and flung to the ground.  Startled, I saw the young drunken man swaying above me.  He regarded me through his hazy consciousness, and pointed a sausagey finger at me, “Nobody cares that you went to Nepal.  Nobody… Stop being a (belching) Cad! … Like a rock! Standing arrow straight…” He paid his bill and left, continuing to sing Bob Seeger, out of tune I might add.
I straightened up, the waitress feigned to help me but I could see her suppressing a smile.  I paid my bill sheepishly and left, fell asleep dreaming of apples, autumn leaves, and sausages.

By 6am the next morning I was up and Aut’umn.  I found a Starbucks (I’m trying to attract some corporate sponsorship here) but was quiet on the subject of travel.  I found Mission Creek and made the usual rounds of hellos to old colleagues and new faces.  The flow regattas are a great way to meet people face to face who you’ve only know by an email address.  I even found an old Football teammate who I might have had a son with! The weather and setting were spectacular, captured by Ashley in Figure 5.  Frank Van der Have, who sponsored the event, and paid for the beers and lunch (after one sponsor[2] bailed), started hauling gear out of his truck. We had several new and exciting devices to test.
The SVR (Stream Velocity Rod) is a hand-crafted analog device that measures the velocity head on a transparent measuring stick.  The technology has been around for some time[3], but Robin Pike with Ministry of Environment has perfected it and made it easily accessible to the professional and stream hydrologist hobbyist.  For further information, refer to this handy Stream Velocity Rod User’s Guide.
The Sontek Flow Tracker is the workhorse of the WSC fleet, harnessed and tamed by Tobin from WSC.  It measures point velocities using acoustic doppler technology, summing the Q in each panel to provide a total Q with uncertainty based on both ISO Standard 748 and a USGS Statistical Method.
The SonTek RiverSurveyor M9 is an ADCP device mounted in a raft capable of measuring depths from 0.2 to 80m.  Unfortunately for this site, the depth was shallow and much of the measurement was blanked out.  If we’d had a deeper, wider river we could have sampled the full capabilities of this device, which is equipped with a remote control diesel engine[4] and tiller.
The Sommer Tracer Q system is a Salt Dilution (and RWT with additional probes) flow measurement system.  The probes are BlueTooth connected to a laptop or handheld device, which records the EC.T over time and calculates Q.  Unfortunately the Sommer TQ doesn’t quack or make any noise at all, unless you drop it or bang it with a rock, which I didn’t!

The Patent Pending[5] QiQuac! System from Fathom Scientific has 100% guaranteed Quack[6]!  Like the TracerQ (made in Austria) it has two probes, one tethered by a long-range radio (>30m) and calculates Q in realtime.  However, that’s where the comparison ends!  The QiQuac is a standalone system not requiring expensive laptops, BlueTooth range extenders, or expensive European hairdoes, shown in Figure 6!  It’s made lovingly[7] right here in Beautiful BC by skilled craftspeople contributing to our shared economy[8]! It provides Industry Leading[9] uncertainty estimates, RISC compliant grading schemes[10], and a colourful[11] Liquid Crystal Display!  We’ve sold systems all over the world[12], including Japan and New Zealand, and even Andre Zimmerman has bought 2!
August Ustaire brought the Deeper Sounder, which is a new device born by the powerful crowd-funding movement.  It pairs with your Android or iPhone and tracks GPS location and depth up to 80 meters with an accuracy of ~±5%.  This is the same device I was introduced to in Nepal (I went to Nepal).  It uploads data to the cloud (Lakebook) where it can be easily managed, stored, and displayed.

LSPIV[13] (a.k.a. SSIV) is a new and exciting technology that uses your phone’s camera to record a video of the water surface, which is then orthorectified using measured reference markers, and processed to generate a velocity vector field.  Paired with a measured, or estimated, depth profile and assumed Vs/Vm[14] ratio (usually 0.85 is a reasonable estimate), the total Q can be estimated.  The Discharge.ch app was used, shown in Figure 7,  and video was also post-processed by Derek using RIVeR, shown in Figure 8.

The Mission Creek Coliseum

Despite its austere beauty and tranquility, Mission Creek was the site of a hydrometric battle of epic proportions.  Thankfully the carnage was limited to a few dead batteries (Frank’s laptop), compared to the mayhem of the Victoria Flow Regatta earlier in the year.

After it was established whose Salt Dilution system had the most Quack, we set up probes on either bank, downstream of the bridge.  I successfully distracted Ashley by continually pestering her and splashing her so that she messed up her CF.T factor and was forced to use default values.  Salt was injected about 400m upstream, shown in Figure 11.  This, again (sigh) is not a good site for salt dilution.  There are a few riffles upstream of the bridge, but no constrictions of note.  In a lapse of judgement (I was so jet-lagged!) we setup our probes downstream of large glides which attenuated the conductivity peaks, but more importantly were downstream of a culvert under the bridge on the left bank, spilling about 30 l/s of higher conductivity inflow.  Despite both Ashley and Derek’s warnings, I belligerently claimed “the higher conductivity won’t affect the measurement, the pulse will simply ride on top of it!” I guess I was thinking if the inflow was upstream of the injection point?  But we were getting a difference in EC.T of about 11 uS/cm, or about 5% higher of 210 uS/cm measured by both systems’ RB probes, shown in Figure 9 from the Sommer unit.  The noise in the LB signal (Probe A) is also indicative of local inflow.  The exact same thing happened in Victoria and affected our results.

Our first injection resulted in a Q of 1.3cms on the RB and 1.9 cms on the LB, on both the QQ and Sommer systems.  Assuming this was due to incomplete mixing (which is partially true) I injected the 2^nd dose about 750m upstream.  Unfortunately (for Frank) this is where the batteries died in the laptop used by the Sommer system to log and calculate the measurement (don’t look at me). The QiQuac merrily logged on, however and measured 1.5cms on the RB and 1.55cms on the LB.  Much closer, but, in hindsight, the LB measurement should be ignored, or at least taken with a grain of… salt.  For this reason I entered 1.5 cms into the regatta log book.  The uncertainty on the RB probe was +/-7%.
Tobin’s Flow Tracker produced a Q of 1.38cms ±2.7% at .  This is very close to the rating curve Q of 1.39cms at 10:00am and 1.43cms at 11:50am.  The results are shown in Figure 12.

The River Surveyor M9 was not well suited to this site given its blanking distance (depth below the water surface it cannot measure) of about 20cm.  It was only able to measure 61% of the area and estimated a Q of 1.07 cms.
August fired up the SVR and calculated 1.24 cms, shown in Figure 13, not bad.  Remember 1.38 cms from the FT is not necessarily the correct number. If we assign a 12% uncertainty to this Q, it’s not significantly different than the FT.
Well, according to Derek Brzoza, processing of the video using RIVeR resulted in a Q of 1.34 cms and with the estimated culvert flow of 30l/s is a total flow of 1.37l/s.  Similarly Salvador processed the video and survey data sent to him from August and arrived at a Q of 1.37cms.  Adding 30 l/s is 1.40cms.  But hey! It’s very promising technology.  I think the thing to remember is that 1.38±6% is the same as both of these numbers, and 1.24±10% and 1.5±7%.

So the PIV methods are very encouraging, if not a little too accurate.  We’d need more of these, and Flow Tracker measurements up and downstream to get a real sense of the true uncertainty.  More discussion is below.  The Salt Dilution methods didn’t achieve a Grade A (<7%) measurement due to local inflow and perhaps poor mixing.  If we’d had more time, we’d have moved the probes upstream of the culvert and injected even further upstream.

Anti-Climax

Over beers at the pub, somebody[15] hauled out the Flow Tracker manual and opened up the section on uncertainty calculations.  This always happens when a group of hydrologists start drinking.  Here’s the equation from the manual:

It’s shown here not as a mathematical reference, but just as a pretty decoration, like paisley wall paper.  I’m sure it means something, but even USGS doesn’t trust it and offer the Statistical method in the Flow Tracker as well.  The statistical method appears to try to take into account the shortcomings of wading measurement based on heuristic[16] methods.  They offer “Perhaps the primary advantage of the Statistical technique is that the estimated uncertainty takes into account variability in depth/velocity across the stream, and as such, includes measurement uncertainty, natural stream conditions (e.g., different bottom types), and the assumption that depth and velocity change linearly between stations.”  It seems to suggest they are questioning the latter assumption, or at least trying to assign an uncertainty to it.  From my brief review of the documentation, it seems the statistical method is usually larger.  But the thing not mentioned in the Flow Tracker documentation is that this is the 1 standard deviation result with a confidence interval of only 68%!  For non-statistical people, i.e. most reasonable people, perhaps this isn’t as shocking as I suggest.  But you should know that this confidence interval will not capture an adequate range of error[17].  The standard in science is to report 2s, or the 95% confidence interval.  The QiQuac always reports 2s.  So we are not comparing apples to apples if we compare a QQ measurement with ±7% 2s uncertainty with a Flow Tracker ±2.7% 1s uncertainty.  The single flow tracker measurement summary, provided in Figure 12, shows an ISO uncertainty of 2.7% and a USGS statistical uncertainty of 3%.  But multiplied by 2 we get ±5% and ±6% respectively, which is more reasonable, but I still think underestimates the error due to the assumptions in the method using only a single point flow measurement at 0.6*Depth.  The same document suggests the uncertainty associated with a single measurement at 0.6*Depth is ±7.5%.  This is the Upi value.  This value gets squared, summed, square-rooted, and otherwise abused, but I fail to see how the total error can be only 2.5% when one of the terms is 7.5%.  I might be confused though.  It’s happened before.
That’s my rant against the Flow Tracker Uncertainty Estimate.

Two kinds of people in this world:

Winners and Losers.  You know the Lindsey Buckingham song.  Well, I’d say we were 50-50 at this regatta.  If the Flow Tracker is the number, then the two PIV methods came the closest.  But the SVR (lower) and QiQuac (higher) are not statistically significantly different.  ADCP: way out to lunch.

In the always controversial and fiercely competitive SWAG contest, we again had a wild and hilarious range of guesses that we all (at the pub) had a good laugh at, especially if you didn’t attend.  Table 2 is a list of winners, from closest (to the Flowtracker) to the DFL (aka Go Big or Go Home award).  Now the GBOGH award I get, I admire that kind of bold confidence in the face of startling evidence to the contrary.  People like that get things built in the world much to the peril and demise of many other people, like the Othello tunnels or the Panama Canal (A Man A Plan A Canal Panama) .  But what I was baffled by, was how Chris won the SWAG competition with a 6^th place guess?  Well, I wasn’t in charge of the awards[18].  I went to complain loudly to the awards committee when I heard someone singing that familiar melody, “Like a Rock… Standing Arrow Straight” and I felt weak in the knees and decided to hold my complaint until next time.  Come out with your own complaints to the 2019 NASH Flow Regatta at the Blue Mountain Resort in Ontario![19]
 
[1] We have defined our Grading system based on BC Resource Inventory Standards Committee thresholds.  Grade A is assigned if Uncertainty on CH0, Uncertainty on CH1, and the %Difference in CH0 and CH1 Q are ALL less than 7%.  Grade B is if they are all less than 15%, and Grade C is assigned otherwise.  The 7% and 15% thresholds are user-defined settings in the QiQuac.  Note that the channel uncertainties are based on the 95% confidence interval, which is 2 standard deviations, not 1 standard deviation (68%) as the deceptive Flow Tracker reports, discussed below.
[2] Who will remain unnamed, but whose name rhymes with Leave The Duck.
[3] Fonstad, M.A., J.P. Reichling, and J.W. Van de Grift. 2005. The Transparent Velocity-Head Rod for Inexpensive and Accurate Measurement of Stream Velocities. Journal of Geoscience Education 53(1): 44-52.
[4] Correction, electric motor.
[5] We haven’t applied yet, that’s pending.
[6] On units that have an operational Quack system.
[7] This is a figure of speech.  There is no guarantee that love went into each unit, although some units may have been loved.
[8] This is actually true.
[9] I’m not sure what this means.
[10] This is likely going to be true.
[11] Green
[12] Technically on both sides of the International Dateline and both Hemispheres.
[13] Large Scale Particle Image Velocimetry (LSPIV) and Surface Structure Image Velocimetry (SSIV)
[14] Surface Velocity to Mean Column Velocity
[15] Whose name rhymes with Rubric Mendoza
[16] I had to look this up after I wrote it.  I know it’s a word, I just don’t know what it means: “A heuristic technique, often called simply a heuristic, is any approach to problem solving, learning, or discovery that employs a practical method, not guaranteed to be optimal, perfect, logical, or rational, but instead sufficient for reaching an immediate goal.” Wikipedia
[17] If you did 100 measurements in the reach at the same flow, 32 of them would be outside of this bounding box.  Shocking!
[18] His name rhymes with Dirk DeSouza
[19] I can’t believe I missed this https://youtu.be/wd3Mt8JBBBg in my first draft.
Thankfully Scott Babakaiff just reminded me. Hopefully we caught it in time and my E&O will cover me if someone files suit.