It was a bright afternoon as I made my way across the prairies to Lethbridge. The sky loomed large and foreboding with the weight of the impending Q competition.  The sharpest hydrometric minds in the country were converging to test their mettle against one another at the 2017 CWRA Flo-tilla, Q-Competition, and BBQ.  While the intention was advertised as information sharing, skill enhancement, and scientific enquiry, we all knew that it was a high stakes battle of wit and nerve with only 1 winner.

The Players

I knew the field would be rife with talent:
Stu “El Conquistador” Hamilton: This globe-trotting data guru wasn’t afraid to try a few illegal moves when the ref wasn’t looking, such as the “last-minute transect” and the old “fictional biography”, both moves banned by the federation when Stu invented them.
Tyler “The Marathon Man” Veness. Despite Tyler’s hulking physique, he’s actually quite nimble in the ring.  AKA the “Stickman”, Tyler’s inventiveness in the field saw him making several measurements with different gear, not all “his” per se.
Willemijn “Fire and Ice” Appels uses “unconventional” means to measure flows.  Despite her generally reserved demeanor, her methods tested the extremes of normalcy.
Derek “The Bruiser” Brzoza works diligently behind the scenes to substantiate his uncanny ability to “see” the flow a priori.  It’s assumed any deviation from Derek’s guesstimate is error in the measurement.
Gabe “The Duckman” Sentlinger with his Quackbox attempts a salt dilution measurement in Extremely calm conditions, i.e. not suitable for salt dilution.
Jeff “The Wizard” Woodward seen here in his custom artwork van, circa 2015, employs black magic and substantial experience to pick the correct number out of thin air.

The Field

The Alberta Irrigation Technology Center (AITC) is about 5 minutes drive east of Lethbridge.  It represents the hydrographer’s dream/nightmare where the “True” flow is known to within 1%.  This is accomplished by a closed loop irrigation canal with pumps and in-pipe magmeters.  It’s capable of moving and measuring 1 to 60 lps.  In this theatre, an error of 1 lps can be ±100% error.

Stakes were high.  Our facilitators, Gerald Ontkean and Lawrence Schinkel played it very cool while competitors boasted of their prowess and brazenly called out wildly wrong flow estimates.  Gerald started the day with a red herring about a leaking canal, a leaking pipe, a changing flow, but it was a gambit to unsettle the competitors.

The Morning

After Gerald’s playful ruse with flow rates, the Q appeared to stabilize and competitors setup their equipment.  Gerald kept his secret well, but at the end of the day revealed the flow was 1.45 lps.  Methods and equipment used were:

• Sontek’s newest ADV, the Flow-Tracker 2, was on hand for current-metering at various locations,
• Fathom Scientific’s QiQuac Salt Dilution measurement system was plugged into the bottom of the diffuser pipe for both Salt Dilution and the novel Thermal Energy Conservation methods.
• The LSPIV method was implemented in the Imomo Discharge app,
• A Thermal InfraRed (TIR) Camera and temperature probes were deployed.
• The AITC’s Rubicon Flumegate and Broad Crested Weir were utilized.
• Several Scientific Wild-Ass Guesses (SWAG)s were hazarded.

ADV Current Metering (using Flow Tracker 2 by Sontek)

Sontek’s latest creation, on loan from Hoskin Scientific, for current metering is the Flow Tracker 2.  From the website:
“The new FlowTracker2 (FT2) handheld Acoustic Doppler Velocimeter (ADV®) has all the technology you have grown to know and trust with the original FlowTracker, but now comes with functional, modernized features   (Bluetooth, GPS and large color screen, to name only a few)…. Each step of the way FT2 guides you along the measurement process with visual prompts and SmartQC audio alerts just in case something important needs your attention.”
The FT2 automatically measures the depth of water as well, so that measuring the flow within a cell is very simple, and calculating the total Q, with QA/QC and uncertainty, occurs automatically and immediately after completing the transect.
According to Table 1, both Stu Hamilton and Jeff Woodward used the FT2 to make flow measurements.  However, Stu’s measurements varied wildly depending on the transect, ranging from 0.3 lps ±300% to 4.2 lps ±27.5%, to 0.8 lps  ±75.  Jeff employed his ample experience in a single velocity measurement to achieve a Q of 1.41 lps±25%.

Salt Dilution (using QiQuac by Fathom Scientific Ltd.)

Salt dilution is a well-established technique for measuring flows in turbulent watercourses (Østrem 1964, Church and Kellerhals 1970, Day 1976, Hudson & Fraser 2002, Moore 2005).  This method relies on “complete mixing”, which implies that water on one side of the channel moves to the other side between the point of salt injection and conductivity measurement.
As seen in Photo 1 above, this site is not well suited for salt dilution, although the constrictions at the flume gates do provide some mixing.  Because the morning flow was so very low, the channel was essentially a series of large pools with a very small flow passing through the gates.   This implied that the transit time would be very long, possibly hours to days, for all the injected salt to move through the system.  Therefore a modified approach was taken, shown in Photo 2.

Screwdrivers and other tools were used to encourage mixing in the pipe.  Originally, one probe was at the furthest point upstream while still fully submerged and the second probe was as far downstream as possible without being affected by stagnant water.  However, because the bottom of the pipe was sealed off, we found the 2^nd probe always displayed an attenuated pulse caused by stagnant water recirculating at the bottom of the pipe.  We therefore put both probes further upstream, as in Photo 2.  We then found agreement between the two.  However, to ensure complete mixing, we employed a power drill with an improvised paddle bit just downstream of the injection inlet.  We then measured a series of several reliable measurements, with an average of 1.4 lps ±7%.

Thermal Energy Conservation Method (The Kettle Method)

This is an experimental method and one of its first trials.  The theory is that energy is conserved when hot water is poured into a watercourse.  The first law of thermodynamics states that the internal energy of a closed system remains constant.  The SI unit of thermal energy is the Joule.  The specific heat (c) is the amount of energy required to heat 1 kg of a mass by 1°C.  For water, this is 4.186 kJ/kg/ °C, or 1 calorie/g/°C.  The heat energy gained by one mass in a closed, two mass system, must equal the heat lost by another.
(eq1)
The thermal energy contained in the Kettle (E_k) over background is given by
(eq2)
Where c_w is the specific heat of water, ΔT is the difference in temperature between the background (starting) temperature and the final. We can measure the change in thermal energy per litre of water that passes a point by integrating under the water-thermal energy breakthrough curve, which is given simply by:
(eq3)
 
Where ΔT is the change in temperature over the background, Δt is the time step, and V is the volume of water. To calculate the flow, the delta energy injected is divided by the area under the curve, given by:
(eq4)
 
As a check on the logic, we can compare units.
(eq5)
 
The Energy in the kettle is in Joules, and we’ve divided the volume out of the breakthrough curve area to get Joules/litre * seconds.  The Joules cancel out to give l/s.
We only had time for one test using this method at the diffuser pipe site.  We used the QiQuac probes to measure the boiling water temp with the misguided assumption that the same probe should be used to measure both.  This was misguided for two reasons 1) we only measured temperatures 4°C above background during the test and 2) we made the mistake of not letting the probes completely come to equilibrium with the streamflow, shown in Figure 1.
To remove the probe equalization curve, we use Newton’s Law of Cooling, and the equation:
(eq6)
 
Where T_o is the starting temperature, k is the conduction efficiency, and t is the time.  We found a k value which matched the decay of the temperature shown in Figure 1.  We then subtracted this background signal to achieve our residual temperature breakthrough curve shown in Figure 2.  Summing under the curve, and using this as the quotient in equation 4, we arrived at a Q of 1.8 lps.  This overestimation is probably due to conduction from the water to the pipe, or some other loss not accounted for.  This quick test is used as a proof of concept for further studies.

Thermal InfraRed (TIR) Imaging Methods

Willemijn Appels setup a TIR camera over the channel, shown in Photo 1.  The theory was that two methods could be employed: 1) measure the velocity of ice-cubes floating by and 2) measure the total surface temperature as a plume of kettle water passed by.
A characteristic image from the first test is shown in Figure 3.  The ice cubes are clearly visible in the TIR image, however they were clustered near the center for the channel making it difficult to achieve a channel wide estimate of velocity.
A characteristic image from the second test is shown in Figure 4.  The plume arrived on the left bank and appeared to change the temperature of the entire channel width.  Using a modified version of equation 4, it should be possible to estimate the total Q, however this has not yet been undertaken.  We anxiously await the results of this innovative test.

The Afternoon

We were told the flow was changed by the AITC overlords in the afternoon.  This was evident when several screwdrivers were shot out of the inlet diffuser holes by the increased flow.  After the channel had filled and reached equilibrium, competitors setup for a much larger (still less than 100 lps) flow.  Before any measurements, keen competitors entered their SWAG into the books.  Clearly deprived of similar field conditions for too long, Stu Hamilton estimated a measly 6 lps for what was later revealed to be 61.5 lps; out by a factor of 10.  Gabe Sentlinger, hoping to hitch a ride on the vast experience of one of WSC’s most revered alumni, but seeing clearly that the flow was greater than 6 lps, entered a conservative 7 lps into the books.  Derek Brzoza, not easily swayed by his peer’s poor judgement, entered an eerily accurate 57 lps into the ledger, which was Grade A in accuracy, but perhaps thrown off by his competitors wildly wrong initial guesses, hazarded a 20% uncertainty.  Tyler Veness, seeing the upward trend of estimates, boasted a Q so large nobody knew where to look, although somehow that estimate didn’t make it into the official record.  And with a SWAG range of roughly 10,000% (or 3 orders of magnitude), the clearly shaken competitors setup their instruments hopeful that science would prevail where clearly (most) judgement had failed.

LSPIV (Using the Imomo Discharge App by Photrack AG)

The following write up was contributed by Derek “The Bruiser” Brzoza.
Large Scale Particle Imagery Velocimetry (LSPIV or PIV for short) has been a technology used for many decades but with today’s suite of new tools readily available on smart phones and tablets, measurements are now a literal snap of the camera shutter. Discharge.ch has incorporated these technologies and provides a platform that’s ready to use once the app is downloaded (from https://discharge.ch/) [1] The discharge app was  deployed on the Samsung Tab-A.
Discharge.ch reads “The Discharge app is a non-intrusive, optical flow measurement tool, suited for natural water streams, irrigation furrows and water channels. The app is fully integrated in the web platform ‘discharge.ch’. At dedicated measurement sites the app can accurately determine water level and discharge. The discharge is calculated either via rating curve, or via surface velocity that is measured by the app. All calculations are performed directly on the smartphone, such that the app can operate in offline mode. An operational site requires 4 reference markers and a known cross sectional profile.”
PIV not only allows for measurements to be conducted on a real-time basis through the field installation of a remote camera it also allows for discharge measurements during times of river ice break-up when other methods fail. One potential draw back of PIV is that it likely won’t work in the rain as raindrops will alter the water surface references of the stream (to be field test).
The app itself is quite user friendly as I didn’t have the user manual to reference (first time user) and was able to navigate the site and measurement screens to capture a discharge measurement of 1.9 l/s versus the known 1.45l/s. This measurement was conducted within the confines of the Rubicon flumegate that provided a reliable water depth for the cross sectional area for the final calculation. I was quite impressed that the app could measure flows this low and within an error of 31% to the magmeter as flows this low are quite challenging to measure.
In the afternoon when flows were increased, two comparison measurements were conducted using the app. This time around, the measurements were hindered by numerous crashes of the software. I eventually found a work around by creating a new site for the measurements and was able to capture two successive measurements of 50l/s and 48.6l/s showing data reproducibility. This out of the box solution for discharge measurements is quite novel and will require further investigative trials and tests to determine its associated error. The software does allow for further measurement analysis via the web site where one can enter the actual cross sectional profile of the stream vs the default shapes provided.
The LSPIV app created by Ryota Tsubaki was also deployed during the flo-tilla. This software works on Apple IPhone and required either wifi or data as your measurements are emailed to you. The app works to capture the necessary field data and post processing is required to determine discharge. I am still working to post process the data collected by the app with the developer who has recently launched the proprietary software. As with all field data it is usually easy to collect but harder to process. Both software have great potential due to their low costs, versatility, and availability. Another interested side note is the ability of post processing video captured by the general public of flood events.
As an endnote to this study, we at Fathom have undertaken several measurements using the LSPIV method as implemented in the Imomo Discharge app at other natural stream sites.  Results are promising, showing repeatability with a Co-efficient of Variation (CoV) between 5%-36%.  The CoV appears inversely proportional to the surface velocity, or more likely the quality of the surface features.  Seeding with popcorn or other detritus like leaves and pine needles also helps with repeatability so long as they don’t sink or become stuck in eddies.  We believe LSPIV can be a very useful technique for interpolation or extrapolation after calibration against other conventional methods such as current metering and/or salt dilution.  The Imomo Discharge app is a very well-thought out and intuitive implementation with the benefit of GIS based web-database to track measurements over time.  Further tests are ongoing.  With respect to the app stability, the developer is very responsive to queries and has written “We are working heavily on increasing the app stability.”  We expect this app to be an integral part of our suite of hydrographic tools in the near future.

Salt Dilution (using QiQuac by Fathom Scientific Ltd)

Again, the QiQuac was setup, this time in the main channel.  There were two weirs forming constrictions.   The 1^st trial was just downstream of the first weir, but agreement between the left bank and right bank probe was elusive.  The next trial was downstream of the 2^nd weir and took longer for the pulse to come through.  There was closer agreement between probe 1 and 2, but still a significant difference was measured.  For the 3^rd trial, taking lessons learned from the morning measurement, GS used a 1×4 “paddle” and standing midchannel mixed vigorously as the salt wave passed by.  Agreement between LB and RB probes was achieved.

Results and Discussion

Table 1 shows the compiled results of the Flo-tilla with uncertainty stats calculated and colour coded. This table shows the party, the method, the estimated Q and associated uncertainty, and the “True” Q and error compared to the estimated Q.
It’s important to understand the difference between uncertainty and error.  Uncertainty is how well you know something, or the confidence you have in a measurement.  It should represent the 95% confidence interval, meaning you are 95% confident the true measurement is within this uncertainty bound.  Error is the difference between the estimate of Q and the “True” Q, here represented by both “Magmeter: %Diff from” and “Magmeter: Agrees with?”.  In this case, we’re taking the Magmeter results as the “True” Q, even though the Q varied between 1.40 lps and 1.50 lps, or ±6% in the morning, and 1% in the afternoon.  Accuracy is used as the complement of Error in this article:  a measurement with low error has high accuracy.
In this table, 7% has been used as the threshold between a Grade A (Green) measurement and a Grade B (Yellow) measurement, and 15% between Grade B and Grade C (Red).  Ideally, a measurement is both accurate and has a low uncertainty.  An accurate measurement with a high uncertainty is arguably no better than an inaccurate measurement with a low uncertainty; i.e. you cannot be confident in your measurement, or it’s wrong but you think it’s right. An inaccurate measurement with high uncertainty at least indicates that you cannot trust the measurement and know it.
Based on this discussion, a good measurement is green across the four coloured columns in Table 1.  In the morning only the SD measurement made with the QiQuac meets this criteria.  Jeff “The Wizard” Woodward made a Grade A measurement at the Rubicon Flumegate, unfortunately it was 10% different than the Magmeter flow, and therefore failed the Grade A accuracy test.  He later redeemed himself by estimating the flow accurately with a single velocity measurement and a SWAG, but with a high uncertainty of 25%.  Derek “The Bruiser” Brzoza made a Grade B volumetric measurement at the Rubicon Flumegate and passed the accuracy test coming within 10% of the magmeter estimate.
Notice that many measurements agree with the magmeter (there is no significant difference between the measurement and the magmeter flow when both uncertainties are considered), but have very large uncertainties and cannot therefore be relied upon.
In the afternoon, no measurement met the criteria of green across all columns.  Again Jeff “The Wizard” Woodward managed to come within 7% of the Magmeter flow using the Rubicon Flumegate, but lacked confidence in his abilities, assigning a 25% uncertainty to his estimate.  Tyler “The Marathon Man” Veness was on the opposite end of scale being highly confident in his wrong answer of 45 lps using the Flow Tracker 2.  The QiQuac SD measurement averaged 55 lps with a 10% uncertainty, which was just shy of containing the Magmeter flow for 61.5 lps.  This resulted in a FALSE in the “Agrees with Magmeter” column, although does still qualify as a Grade B measurement.  Derek Brzoza defied all convention going against several wildly wrong SWAGs to come within 7% of the Magmeter Q with his SWAG, although with a low confidence.
These concepts are illustrated in our target plot of Error vs. Uncertainty in Figure 5.  In this figure, the uncertainty is plotted as a function of absolute error for all measurements.  In the morning, measurements generally had a high uncertainty and low error, while it was the opposite case in the afternoon.  Only one measurement hit the bullseye of low uncertainty and error less than 7%, the SD measurement in the morning.  Two measurements hit the Grade B target in the morning and one in the afternoon.

Conclusions

Measuring low flows in seemingly ideal conditions for current metering (parallel banks, flat bottom with very little turbulence) is more difficult than expected.  In the morning, the very low velocities of the AITC irrigation canal proved problematic for the Sontec Flow Tracker 2, resulting in errors between 45% to 190%.  Both Froud and Volumetric methods at the Rubicon Flumegate failed to achieve a Grade A measurement in both uncertainty and accuracy.  Only the Salt Dilution measurement using the QiQuac achieved Grade A in both uncertainty and accuracy, although this method required significant time to fine-tune and adapt for the site, which was not well-suited to SD methods.  Forced mixing for SD is a new approach to this method born from the fertile soil of the flo-regatta.  Thermal Energy Conservation methods using temperature probes and TIR is a promising field of research but did not produce, reliable results on this day.
In the afternoon, experience guided several of the measurements towards a Grade B measurement in accuracy.  No single method achieved a Grade A in both uncertainty and accuracy.  The LSPIV method showed promise, but requires calibration against other methods.
It was surprising that conventional methods i.e. multi-panel current metering, produced results that were no better than ±45% in error in the morning  and ±27% in the afternoon.  Weir measurements were able to achieve results within 10% accuracy, but were tempered by good judgement.
In summary, the 2017 CWRA Flo-tilla, Q-competition, and BBQ in Lethbridge was a success.  Experience, tips, and tricks were exchanged and all attendees benefited from the experience.  New methods, both commercial and experimental, were tried and valuable information on how to improve on them was garnered.  It was good chance to meet colleagues from across the country, and objectively assess both the uncertainty and accuracy of their methods for measuring low flows.

The Winner(s)

Even though we were ALL winners for coming out to the 2017 CWRA Flo-tilla, Q-Competition, and BBQ, the mob that had assembled in the afternoon, fueled by free beer and meat, demanded champions be named (and the prize donors also needed to get their merchandise out).

The Morning

• Jeff Woodward went home with the closest Q estimate for his SWAG/ADV measurement of 1.41 lps ±25%
• Gabe Sentlinger rightfully claimed the award for closest measured Q within uncertainty for 1.40 lps ±7%

The Afternoon

• Derek Brzoza received the “Stu Hamilton Confidence in Truth Fabrication” award for bravery in the face of peer pressure for his initial SWAG of 57 lps ±20%.
• Willemijn Apels was named “Innovator of the Contest” for her TIR camera tests, rumoured to have been thought up while scanning her office for ideas.

YOU can be a winner too by coming out to the 2018 CWRA Flo-tilla, Q-competition, and BBQ in Victoria B.C. where we’ll be showcasing new technologies for flow measurements  The social aspect of these events cannot be trivialized;  the benefit to hydrometry and the country is tangible.  We look forward to seeing you there.

References

Church, M. and R. Kellerhals. 1970. Stream gauging techniques for remote areas using portable  equipment. Department of Energy, Mines and Resources Inland Waters Branch, Ottawa, Canada.Technical Bulletin No. 25.
Day, T.J. 1976. On the precision of salt dilution gauging. Journal of Hydrology 31:293–306.
Hudson, R., & J. Fraser. 2002. Alternative methods of flow rating in small coastal streams. B.C. Forest Service, Vancouver Forest Region, Nanaimo, B.C. Forest Research Extension Note EN-014.
Moore, R.D. 2005. Introduction to salt dilution gauging for streamflow measurement. Part 3: Slug injection using salt in solution. Streamline Watershed Management Bulletin 8(2):1–6
Østrem, G. 1964. A method of measuring water discharge in turbulent streams. Geographical Bulletin 21:21–43.
[1] Fudaa is another open source LSPIV app built windows based (https://forge.irstea.fr/projects/fudaa-lspiv/files).  Fudaa was not used for this experiment.