Adding the Trend column back to the rivers page
Daniel Reeve
Kevin Colburn
Daniel Reeve
Rob
USGS gauges are not 'real time'. Some gauges record data every 15 minutes, others record every 5 minutes, but all USGS that I have ever seen only post that data online ONCE PER HOUR, and that is usually between 20 and 30 minutes after the hour! (I.E., somewhere between 1:20 - 1:30 pm, they'll post the data for the hour from noon til 1:00pm.) SO ... by the time they post it, the data is 20-30 minutes old! This means any user who looks at the gauges (either USGS directly or the AW app or website) any time until 1:20 or 1:30 pm is only going to see the noon readings! All readings you see are AT LEAST 20 minutes old, and up to an hour and twenty minutes old.
(I do not program for AW, so I can't be sure how they receive their data, but my guess is they are not polling each and every gauge directly, getting up-to-the-minute readings. Rather (I suspect) they are polling the data after USGS has posted it to their online pages. PLEASE CORRECT ME IF THIS IS WRONG, but I highly doubt it is, because from my observations, what AW shows does not change until USGS has posted it's hourly data.)
In my area (and number of other places I'm aware of), some small fun runnable creeks (some with drainages under 20 square miles, and others up to 120 square miles) run only with significant rain. These can regularly go from unrunnably low, to runnable, to dangerously high, and back down to lame/tame or too low in 3 hours or less! By the time AW reports these are 'trending up', they are may already have peaked. By the time you can gather your gear, get there, and set shuttle, they're often already dropping back down. If you are relying on gauge data, you're likely to miss them nearly every time!
For the above reasons, trend actually is only meaningful for less flashy runs!
Rob
Rob
The best indicator of trend is simply the graph of the most recent values (whether its 2hr, 4hrs, 6, 8, 12, or 24).
DRC
Pro tip: Exclamation points and all caps
are the Internet equivalent of street corner preaching with a
megaphone. They're a good way to make people walk on by and
ignore everything you're saying.
No one is arguing that the trend tool is perfect. They're just saying that it's useful under certain circumstances. Give users the tool, give them an understanding of how it works, and let them decide when it's useful. It isn't our job to predict every possible way in which beta might be misused. (There are certainly ways in which existing AW beta can be misused, irrespective of the trend tool.)
To the best of my understanding, rivers
and creeks follow a logarithmic curve on the way down from a
runoff event (assuming there is no subsequent rain.) In my
previous home state, where I was streamkeeper for 8 years or so,
we observed that directly with multiple flashy creeks. You
could talk about them in terms of half-life: a creek with a
half-life of 1 hour would be flowing at half the current cfs an
hour from now, 1/4 the current cfs an hour after that, etc. We
had some 1-hour and 2-hour half-life creeks within 30 minutes of
where I lived, and both AW and USGS were mostly useless for
those. We had to look at rainfall gauges (which were
fortunately provided online by the local river authority),
predict when that rain would get down to the runnable sections,
and try to be there ready to go when it did. The USGS gauge was
useful mainly if the creek flashed a lot bigger than was
optimal. Knowing the half-life and gauge delay, we could use
that to figure out when it would be at the upper end of the
optimal range. Basically, getting any utility out of any online
source required specialized knowledge, but having those sources
was always better than not having them.
Expressing a linear trend (cfs/hour) is just expressing the instantaneous slope of the logarithmic curve. As you pointed out, the more flashy the run (the steeper the curve), the less that the instantaneous slope correlates to the actual long-term trend. That's fine, though. I think most paddlers understand that river flow doesn't change linearly. And really, the flashier the creek, the less that online gauges are useful at all. That's not just because of the delays you mention. It's also because the gauge is often many miles upstream or downstream of the run, so it doesn't reflect the actual flow on the run. Paddlers understand that too, and they tend to use other tools like Rivers.run or RiverBrain, which have better update rates than AW, for catching flashy runs. Mostly, though, the best tools for those are visual beta.
Where I found the trend tool particularly
useful was for rainfall-dependent rivers that rise and fall much
more slowly (like with a half-life of a day.) For flashier
stuff, it was still sometimes useful to see whether something
was on the way up or the way down, and potentially how quickly.
In other words, if you understand that the trend is only an
instantaneous slope of the curve and not really a long-term
trend, you can make some use out of that. Expressing the trend
as a half-life rather than a linear trend would be more useful
and correct for streams on the way down from runoff events, but
it would be just as incorrect as a linear slope in other
situations. It really just boils down to understanding the
limitations of the tool. The same is true for the gauges in
general.
DRC
--
You received this message because you are subscribed to the Google Groups "American Whitewater StreamTeam Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to american-whitewater-stre...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/american-whitewater-streamteam-forum/ed5cc144-7713-40de-9d40-5414e79ea4e2n%40googlegroups.com.
Rob
DRC
Being a professional tech writer is one of
many hats I wear, Rob. I wasn't saying "pro tip" ironically.
The slope of a curve is a useful metric.
It shows you whether the curve is ascending or descending, and
by how much at this particular moment. It's such a useful
metric that it has a special name in mathematics: the
derivative. If you understand that that's what the trend
feature expresses, then the feature is useful.
As you point out, showing another trend
metric such as half-life would be more fraught, because it would
imply that the curve will follow a certain shape. The slope of
the curve has no such implication. It merely tells you what the
curve is doing at this moment. It is not meant to be
predictive. Similarly, you can't take the slope of a car's
acceleration curve and predict the speed at which it will be
traveling an hour from now. However, if two cars are traveling
at 20 MPH, then there is utility in knowing that the slope of
one car's acceleration curve is -20 MPH/minute and the slope of
the other car's acceleration curve is 100 MPH/minute.
DRC
DRC
To put it differently, focus less on the
number, and think of the trend more as a qualitative metric.
What it really attempts to communicate is "rising quickly",
"rising slowly", "steady", "falling slowly", or "falling
quickly." However, just like the flow metric, translating a raw
trend number into a qualitative statement requires knowledge of
the river. 1000 cfs/hour could be a quickly or slowly rising
river, depending on the river. Thus, it's best for the web site
to simply present the numbers and let people apply their river
knowledge to them.
Paul Martzen
I think Rob put it best in his third message, "The best indicator of trend is simply the graph of the most recent values..."
A trend derivative is calculated from the graph, but only reflects a tiny portion of the graph. Any user can look at the graphs and get an overall sense of what the river or creek is doing. Looking at the graphs regularly seems much more educational to me, than looking at a momentary trend/derivative number.
As far as western rivers in California, we do have rain fed, flashy creeks, just not enough rain to make them flash as often as we would like. Most of the ones in my neighborhood don't have gauges, so I have to estimate from rivers that do have gauges. I always look at the graphs to have any chance of making a good guess.
Snowmelt rivers trend up and down every day, but the peaks and valleys trend up or down from day to day. Calculating a meaningful trend number could be very difficult and likely error prone. It is much easier for the user to look at the graph and then look at a weather report.
Let's not force our programmers to figure out which trend calculation to use for each different river, just to save the users from a single click to see the graph. The graph will tell them a lot more than any single trend number. In my opinion.
Paul
To view this discussion visit https://groups.google.com/d/msgid/american-whitewater-streamteam-forum/42d2936a-4045-4ad4-ac44-a19d2c814607%40gmail.com.
DRC
On Mar 2, 2026, at 11:23 PM, Paul Martzen <paul.m...@gmail.com> wrote:
To view this discussion visit https://groups.google.com/d/msgid/american-whitewater-streamteam-forum/a17579a7-86cc-420e-8677-15e06a696767%40gmail.com.
Rob
Thomas O'Keefe
Tom
Anthon Allred
My thought is that trend indicator based on past water level data would not be any more useful than the graph AW currently has on the "flow" page. But NOAA (the weather service) does do future water level predictions on its web site. Perhaps it might be useful to consider closer coordination with them.
Tony Allred Jr.
DRC
Jesus, Rob. Let it go, OK? I am not trying to convince you of anything. We're debating this for the benefit of people who actually make these decisions. I stepped in only because I didn't want attacks on what I deem a flawed-but-useful feature to go unanswered.
The harsh reality is that, on the modern
Internet, people my age and younger process arguments with all
caps, exclamation points, and scare quotes as inciteful rather
than insightful. (You can blame
certain politicians for that.) Thus, your initial comment came
across as scaremongering regarding something (the trend feature)
that isn't that big of a deal. I was not personally offended.
I was just pointing out that, in my opinion, your perfectly
valid argument was probably not landing, solely because of how
it was expressed. You could've just ignored that off-hand
remark rather than turning me into your personal white whale.
Believe me, I'm not worth the cost of the harpoon.
No one cares about our personal credentials. I mentioned that only because you said I was coming across as self-important by saying "pro tip." I could mention that another hat I wear is developing fluid flow monitoring software that performs similar statistical calculations to the ones we're discussing, but no one cares or should care. The only relevant argument is that the trend feature was deemed useful by some boaters, and that's sufficient to justify reinstating it.
Honestly, I'm sorry I said anything at
all.
I fully understand that for 'technical writing' you need to work within a constrictive set of grammatical rules which disallow unsanctioned use of bold, italics, capitalization, and exclamation points. I tend to write the way I feel and the way I talk (and the way most people talk), putting emphasis and inflection on specific words and phrases. I was not doing 'technical writing', I was doing informal writing. Sorry that you feel the need to restrict the freedom of others to write in a way which apparently offends your sensibilities.
And now you are now putting words in my mouth. I did not say that trend is "evil and useless". I said that (as it was naively being reported in the past, just subtracting the single value of the reading one hour ago from the single value of latest reading) it is far too prone to be entirely incorrect. I love having some metric to give an indication of rate of rise and fall.The case i have repeatedly made is that there needs to be some statistical 'data smoothing' to reduce or eliminate the impact of errant readings which regularly occur with USGS raw data.
Since apparently we are revealing our credentials to inform and reinforce the merit of our remarks ... I had a math major, computer science minor. I'll freely confess I have no formal emphasis in statistics, and my brain maxxed out when dealing with calculus, differential equations, the concepts of derivatives, and all that higher level mumbo jumbo. For exactly that reason, in my earlier post, that I wrote " I will admit that perhaps some more experienced statistician might be able to come up with a calculation which would be even more accurate more of the time." (Perhaps you could apply yourself more to that task than to taking others to task for their informal writing.)
On Monday, March 2, 2026 at 11:00:01 PM UTC-6 drcom...@gmail.com wrote:
Owen Coutts
Thanks for the discussion here, it's heartening to see so many people put so much thought into how this should work.
Some Background
- Go to the wave in Missoula and talk to folks waiting in line for their turn
- Go to the boat shop in Missoula
American Whitewater's codebase dates from 1996 and has the cruft to show for it. People have added many features that kind of work or used to work that feel broken and overwhelming today. It's often hard to know which are important enough to keep around and where we should build something simpler. As an example of something else removed, it used to be that our river system would allow a reach difficulty to get easier as the water goes up - I'm sure many of us have seen this happen but it led to situations where our data was just so consistently broken that people didn't trust the ranges.
This feature caused that feeling in the mountain west for beginners and intermediates. It also caused problems for local experts who could not explain what it was or what value it brought. The at the time owner of the boat shop who was doing a ton of instruction and interfacing through the community by running a boat library said "This feature has caused so much confusion, I wish it had been removed years ago".
The last piece of context to add is that the old system that showed the difference between the last two readings is gone, any system that does forecasting will be rebuilt from the ground up so we have a choice of when and how we build something in the space.
Values and How To Go Forward
- Adding things to a UI has a cost People don't "just ignore" things they don't understand or don't need and this has a cost for them. Things need to be generally useful to warrant putting them on the main interface pages.
- New + Infrequent Users are really important For the health of American whitewater, it's really important that we continue to build interfaces that are great for people who are new to our community or new to boating.
- Advanced Features can be Hidden Advanced folks who want a feature don't need to have those things on the main interface if provided another option
- Delta of Most Recent Readings (the one that we used to display):
- Pros:
- Fast - can quickly react to environmental changes
- Simple - of all the algorithms many people had an more or less correct understanding of how it worked
- Cons:
- Doesn't provide meaningful value for the majority of the rivers in our systems
- Provides incorrect and confusing data for the majority of rivers (it doesn't self opt out)
- Hard to interpret since our system doesn't yet have an understanding of how frequently a gauge updates and therefore doesn't display it which is key to understanding this metric when it's useful. We would likely need to learn which gauges update every e.g. 15 vs 60 min.
- Pros:
- NOAA Forecasts
- Pros:
- Uses professional humans to know the specifics of a given region / basin
- Has a wide geographical reach
- Does not provide incorrect forests for rivers where it doesn't know anything
- Cons:
- Slow to update
- Does not provide a forecast for commonly boated regions
- Pros:
- Linear Regression Model
- Pros:
- Likely more generally useful as a baseline
- Perhaps it could be tuned to provide some sort of directionality at the 24 hour level
- Could understand rivers with a diurnal pattern
- Cons:
- Would not be able to react to rain events because it's looking at historical data
- Does not have humans in the loop to know specifics
- Pros:
- A specific model / algorithm for a specific river
- Another agency (WSDOE, NOAA, Field Office, etc)
DRC
Thanks for your efforts, Owen. The only thing I would say is that, while talking to boaters, you should also talk to the streamkeepers who have been maintaining beta on the AW site for decades, as well as advanced boaters. There is a balance to be struck between making things accessible for new boaters and dumbing things down to the point that those with experience no longer take the site seriously. Out here in the Southeast, I struggle to think of any experienced boaters I know under the age of 60 who use AW as their go-to. That's a problem, because we need those experienced boaters to provide beta.
I feel like the technical gap between AW and other river apps has either closed or is closing fast, but there is still a big marketing/perception gap. The AW 2.0 rollout was frankly really buggy, and it made a lot of experienced boaters distrust the site/app. It's going to take some effort to bring them back. Attracting new boaters is a worthwhile effort, but we also need to retain experienced boaters in order to retain institutional memory.
I don't think the trend feature is make-or-break for any of the above. I think some people found it useful, self included. I wasn't married to it, but I also didn't want to see it burned at the stake. It could just as easily be moved into the graph section rather than displayed on the main page. (In fact, if the graph displayed a visual slope line, that would make it more clear that the trend is an instantaneous slope rather than a prediction.)
When I talk to experienced boaters, the #1 thing they mention about AW (and why they don't use it) is that it doesn't update quickly enough. The #2 thing they mention is that it's a moving target. Too often these days, you'll get used to how a feature works, then it'll either stop working or go away. When people spend their valuable time learning a system, they want that effort not to have been in vain. That doesn't mean keeping 30-year-old code, but it does mean that AW's change management could use some improvement.
tl;dr: Adding things to a UI has a cost,
but removing them or changing how they work has a cost as well.
DRC
Jim Mazzola III
Maybe it's just the pent-up angst everyone is feeling about not being able to get in the water just yet. :)
I will, however, weigh in on the trend line. I agree with the benefit of having it. We all understand it's not absolute, but it gives me a quick indication that I may need to investigate further.
I also agree with Darrell's comment that the 'community' we're serving with the information is important. The long-time boaters (regardless of age) simply want a quick overview of the usual level and trend. Less experienced boaters or boaters new to an area look for pictures, write-ups, directions, and warnings, possibly (although those are usually woefully out-of-date and never seem to get cleaned up. It IS two distinctly different audiences, and burdening one over the other with useless, cumbersome information OR not enough of the right stuff, will alienate someone!
I will say that the revision that was introduced 7-8 years ago was so 'buggy' that I lost interest and focused my efforts elsewhere.
The Carolina Canoe Club was fortunate to have a High School freshman write a river program that is pretty much unmatched in its instantaneous information without any fluff. It's called <rivers.run>, and I'm sure many of you know about this app.
This program does give trend indicators as a simple UP or DOWN arrow or a dash if it's stagnant. It also provides 'additional' nearby upstream gauges that serve as an 'indicator' of what may be coming down the river. When available, it also provides NWS, (National Weather Service) gauges that offer 'predictive' estimates of future flow, which also serve to help know just how much flashing of the river may be in store.
The app provides driving directions from any location with a one-button click to copy and send the link of river information to your friends.
And finally, the program also provides an 'overlay' of the current radar in the area so that additional knowledge can be gained to see if there's a storm brewing in and around the river destination, or live radar that needs to be consulted with.
Here's the French Broad as an example:
(Pick the river to expand it to see the gauges, also pick the 'Click load map' and the Weather toggle 'button' is in the lower left corner.)
https://rivers.run/#%7B%22tags%22:%7B%22query%22:%22cccwor%22%7D,%22normalSearch%22:%22French%209%22%7D
Here's what the one-click weather overlay looks like:
Granted, it's pretty granular, but it does give you an indicator of what's in the area.
Rest assured, software development in this arena is not standing still.
To view this discussion visit https://groups.google.com/d/msgid/american-whitewater-streamteam-forum/e9c61a61-949f-474b-8f06-ddfe755e04a3%40gmail.com.