Will the US Geological Survey Go Metric?

I was asked the other day if I thought that the USGS would ever go metric.

I am unqualified to answer this question but I care about the implications of the issue. I started my field career working in Imperial units (also known as English) before the Water Survey of Canada converted to the International System of units (SI, commonly known as metric) in 1980 so I have a firsthand experience with the process and believe it would be much easier now with the advantage of modern electronic technology.

In order to coordinate data sources to tackle the ‘wicked’ 21st century problems of a stressed planet we will need to converge on international standards for data management and data publication. That a stakeholder like USGS remains a hold-out on one of the most basic of international standards speaks volumes to the challenges posed by jurisdictional policy barriers to international cooperation.  All of the costs of the change would be borne by the USGS but many of the benefits would be distributed worldwide. Furthermore USGS would still be obliged to provide reports in the units specified in legal documents for water licensing and regulation.

One major benefit to the world would be having the USGS Techniques and Methods papers published using SI units. The USGS commitment to updating these documents with new and emerging technologies and their commitment to free and open access are major advantages for any monitoring program needing a set of standard operating procedures.

These very rigorous, articulately written, documents would easily be the global industry standard if only they were written to International Standards.

The SI system of units is water-centric which should be argument enough to justify metrication. Dimensional transformations from flux to volume to mass to force can be calculated in the head to a first order of approximation. Smooth scaling over orders of magnitude further reduces the likelihood of undetected errors.

I challenge anyone to look at flow data in cfs and come up with a first order approximation of totals in acre feet without using a calculator.

It is important to understand the relation between sensor resolution and reporting resolution.  Sensor resolution is the smallest change a sensor can detect in the quantity it is measuring. In the analog days, this was about the thickness of the trace on a recorder chart, which for a standard chart scale was about 0.01 foot (0.003 m). This, conveniently, was about the same as the data accuracy for a staff gauge reading. Traditionally, it made a lot of sense to report the data at a resolution of 0.01 foot.

The achievable resolution of modern pressure transducers can be 0.01% or better even under difficult environmental conditions. For a 5 m range that would mean 0.0005 m change in stage is detectable. The accuracy of the data is still limited by reference gauge readings but detection of change within the time series is no longer limited by the recording device. The reporting resolution of 0.001 m used in SI systems provides a much better match to modern technology.

Consider the case where stage is rising at a rate of 0.006 m per hour.

A USGS gage would report the change in two steps of 0.01 foot whereas an SI system would report the change in 6 steps of 0.001 m. Therefore the SI system will represent the true dynamics of the underlying signal more accurately even though both systems have the same uncertainty with respect to gauge datum.

As I write this the discharge at the Stikine River at Wrangell (15024800) is 33,400 cfs and the discharge at the nearest upstream gauge is 219 m3s-1 at the Stikine River at Telegraph Creek (08CE001). There are a number of things that are interesting about this. First, it is fantastic that with a few mouse clicks I can access this data in near real time. Second, what a pain it is that the data are not in the same units and if I want to compare runoff then I am given the drainage area for Wrangell as 19,920 mi2 and the drainage area for Telegraph as 29,000 km2. Third, it is interesting that both agencies report data to 3 significant figures (actually the preliminary WSC data are not rounded for real-time publication but I want to make a point). The reporting resolution of the USGS data implies that the detectable change in discharge is 100 cfs (i.e. from 33,350 to 33,450) which is 2.83 m3s-1. The reporting resolution of the WSC data implies that the detectable change in discharge is 1 m3s-1 (i.e. from 218.5 to 219.5) which is 35.3 cfs. Both agencies use very similar technology, training and operating procedures so which one reports at a resolution best matching detectable change? The answer may be different than it used to be.

Part of me is inclined to root for under-dog legacy units like the miner’s inch and acre foot but it is fairly clear that these anachronisms are not going to prevail. The Omnibus Trade and Competitiveness Act of 1988 designated SI as the preferred system of weights and measures but stopped short of regulation, leaving US industry to adopt the metric system voluntarily. NASA adopted the SI system in 2007.

The USGS is one of the most trusted and respected agencies in the world. I don’t know if metrication is currently on their road map but when they do come to that fork in the road I am sure their choice will be well-informed by consideration of science, technology, societal benefits and stakeholder interests.

9 responses to “Will the US Geological Survey Go Metric?”

  1. Dave Gunderson May 15, 2012 at 11:16 pm

    I’m unqualified to answer this too. Equipment wise, the only physical changes would be resurveying every gaging site in the US to be in meters, changing gaging staffs to metric and swapping out Encoder wheels. Most electronic sensors can be re-programmed via SDI-12 commands or other serial interface. The big rub would be the documentation cut-over and re-working records to different standards. That is time consuming. Data Loggers would take the changes easy. Calibrations might be handled a little differently too.

    Your comment: “A USGS gage would report the change in two steps of 0.01 foot whereas an SI system would report the change in 6 steps of 0.001 m. Therefore the SI system will represent the true dynamics of the underlying signal more accurately even though both systems have the same uncertainty with respect to gauge datum.”

    True. I agree. Showing more resolution in a reading due to a smaller physical increments may seem like more accuracy but actually isn’t under real world conditions. Think of a sensor in actively moving water and the actual stability of the reading. Take several readings back-to-back and you see what I mean.

    I haven’t heard of a move within the USGS to go metric. If I had to ask someone, it would be Phil Turnipseed.

  2. Hi Dave,

    I’ll meet your point and raise it. Logging is typically done on a scheduled basis and so, in reality the data would be most likely be logged on a 15 minute basis. Typically, some electronic stilling is done on the sensor to average out any turbulence effects. For the case of a steady 0.006m rise over one hour the logged data using a ‘round half up’ rounding rule would look something like:

    Time Truth (m) Logged (m) Truth (ft) Logged (ft)
    0 0 0.000 0 0.00
    15 0.0015 0.002 0.004921 0.00
    30 0.003 0.003 0.009843 0.01
    45 0.0045 0.005 0.014764 0.01
    60 0.006 0.006 0.019685 0.02

    The SI data are a closer representation of what is really happening in the stream.

    Most of the time the difference between the two systems is inconsequential but the ‘step-wise’ appearance of the data imposed by logging to hundredths of a foot looks very odd particularly during episodes of low flow. Modern sensors can do much better than that.

    The scale of turbulence that you are talking about is more typical of kinematic effects of high stage where differences of a millimeter or so really don’t make a difference compared to the uncertainties in assumptions about what is really being measured (e.g. the assumption that water density is known and constant – which affects pressure transducers more so than float sensors).

    Changing the resolution of the data changes the precision. Uncertainty is a function of both bias and precision. If you would like to provide a mailing address I will send you an print of my recent paper on hydrometric uncertainty.

    There are many good reasons for inertia on the question of change. I do not mean to suggest what the USGS should or should not do. All I am saying is that if they haven’t looked at the question for a long time that some of the reasons for arguing against change may have since been over-taken by technological advancements. A counter-argument to this is, of course, that technological advancements should have also trivialized the issue of units conversion.


    • Dave Gunderson May 19, 2012 at 11:19 am

      I’ll call your bet.

      I see your point. Yes, more resolution for the record and the rounding issues we see in the data. A good hand…

      Now, let’s try to calibrate the site to those smaller increments and see if the real-time readings track what we see with the drop tape.

      I’ll raise the bet. What about site noise in the measurements due to water conditions? Will the lower resolution we have help or will it even matter?

      Sometimes, Stu. I think that we have to play the hand we’ve been dealt and try to finesse a few tricks along the way.

      What say you?

  3. Hi Dave,

    I want to preface this comment by pointing out that there is a linkage back to our other thread of conversation about training and how the challenges of stream gauging resist prescriptive recipe book solutions. Getting the best possible data requires an artful balance of techniques, technology and methods that are uniquely suited to local conditions. Formally structured training is inadequate to fully develop stream hydrographers for the task of getting the best possible data ‘under the circumstances’. In this case, you are examining the argument about a connection between a choice in system of units and data quality from the pragmatic perspective of net effect – in the context of the not-so-tidy ‘real world’. The ability to reconcile monitoring theory with the reality of field conditions is what distinguishes the experienced field hydrographer from everyone else.

    In theory, it makes sense to report data at a resolution that is close to the limit for detection of change in the sensor being used. It may be easier to explain in terms of temporal resolution. If you have a clock that can reliably detect change at a resolution of 1 second then rounding the time to the nearest minute will degrade the data. With rounding up to the nearest minute, two samples could be 59 seconds apart in time but share the exact same time-stamp whereas another two samples could be only 31 seconds apart and have different time-stamps. In other words, the choice of reporting resolution has an impact on precision, which has an impact on data uncertainty. The uncertainty due to imprecision is independent of the uncertainty due to bias. Bias error may be due to: uncertainty in the time zone; uncertainty in the reference standard used to set the time; and uncertainty due to clock calibration.

    These arguments about the effect of choices in temporal resolution have parallels in magnitude resolution. The choice between Imperial and SI units invokes a 3-fold difference in magnitude resolution. When WSC converted to metric in 1980 I could have argued (and probably did!) that the change in reporting resolution made no sense because it exceeded the native detection-of-change resolution of the technology. I remember being generally in favor of metrication but dis-satisfied with the change in resolution as it applied to measurement of discharge. Monitoring technology has since caught up and, arguably, surpassed the SI reporting resolution.

    You are correct that the uncertainties due to correction of the water level time-series to a reference datum – in the context of dynamic streamflow – are large with respect to the uncertainties due to reporting resolution. I don’t think that you are arguing that improvements in data precision are a bad idea but rather that the reward for doing so (in terms of data quality) will be lost in the context of all of the other sources of noise in the data.

    I would not argue this point for the 99% use-case. In absolute terms the error in calibration of the logger signal to an external measurement dominates the overall error. Furthermore, when the data are published at a resolution of daily mean values any gain in reducing precision error is lost in aggregation, whereas error with respect to the external reference is preserved. However, for any given error with respect to the reference gauge, the SI data will reveal the internal dynamics of the unit value signal more accurately. This can be very useful in the 1% case (e.g. during droughts and low flow episodes) where relative change in the data becomes important.

    We are by no means done with this topic. There are still many different directions we can take this conversation.


  4. Dave Gunderson May 24, 2012 at 1:35 am

    Done with the topic? No sir, we have only started…

    Directions to go? Setting up instruments to capture the site conditions? Measurement techniques? Logging methods? The pros/cons of certain instruments to use? Field automation? Telemetry? Sensors?

    Here to discuss, learn and toss in an occasional viewpoint from a field person.

  5. Hi Dave,

    We are getting a bit removed from the original question about metrication but you have planted seeds for about half a dozen new blog posts. It would be great if I can create new opportunities for you to share your insight with anyone who finds their way to the Hydrology Corner blog.


  6. Hey Stu, you need an old guy to do the acre-foot conversion. One cfs-day = two acre feet.

    Incidentally, when Canada switched to metric dam3 was selected as the means of displaying annual volumes since the number of digits ahead of the decimal place was roughly similar to that for acre-feet. Other options such a megalitres, cubic metre or cubic kilometres were rejected. This was at a time when the “blue books” were the gold standard for presenting hydrometric data – page layout was important.


  7. Hi Bob,

    I am very glad you provided this context. I do remember that working with English units was easier than working in metric. For one thing carrying two decimal places in mental arithmetic is easier than carrying three. I imagine that I knew this trick for conversion from cfs-day to acre feet at some point and then long since forgotten it. As with all unit systems there is an inherent logic with English units that makes it attractive for a variety of reasons.

    How convenient that the area that could be plowed in one day by an ox (43,560 ft2) http://en.wikipedia.org/wiki/Acre is just over ½ of the number of seconds in a day (86,400). Ultimately, this is a reminder that the purpose for monitoring, in many cases, is for agricultural water supply. I wonder though, in a day and age when farmers are surveying and leveling their fields to a very high precision using lasers and/or GPS technology, how the reference to oxen is still relevant.

    I do find the water-centric structure of the metric system very convenient.

    I would be delighted if you would consider contributing some articles to this blog. The history of hydrometric monitoring is an under-told story. The current generation of stream hydrographers would be well served if they had an opportunity to learn more about the origin and development of modern methods, techniques, technologies and publication protocols.

  8. Ronald T. Richards June 25, 2013 at 2:42 pm

    I retired from the USGS Caribbean Water Science Center in 2008. I now teach science in Puerto Rico where my job is made harder by the mixing of units. Distances on the highway are in kilometers and speed is in miles per hour. Gasoline is sold by the liter and milk is sold by the gallon. It costs the economy money when people have to be trained in two systems and when they make mistakes.

    The SI system was created by the French revolution and spread across the world by the Russian and Chinese revolutions. There are transitions costs but in the long run it would be better for both the US and Puerto Rico to switch units.

Join the conversation