Tagged: data accuracy
2022-04-28 at 8:30 AM #16960
Our Mayfly 0.5b has been redeployed after sitting in storage for more than a year. Yesterday I downloaded a week’s worth of data and I’m wondering how to tell if it is accurate and that the CTD probes are working as they should.
From a quick scan of the data I’ve noticed a few things.
- the battery voltage is holding steady at above 4.0. Even if it dips a little (still above 4), it rebounds to the original voltage so it looks like the solar panel is providing sufficient power back to the battery – and that the battery is still good.
- the board temperature drops quite a bit at night – getting down to 5 degrees celsius. Is this normal? I figured that the insulation in the pelican case would provide a better level of insulation. What would be a safe operating temperature range for the board – both over the summer and winter?
- conductivity ranged b/t 260-285 during the week of deployment (4/20-4/27)
- water temperature ranged b/t 9-17 degrees – with no surprising jumps or other outliers
- depth ranged from 206mm to 160mm. The bottom of the probes are about 3 inches off the stream bed (rock and silt) so this is one set of measurements that I’m not sure about. How does the probe monitor the depth of the stream – establishing the depth of the water below the probe and above it for a total depth? I’m sorry if this is a basic question, but I didn’t attend a Mayfly training and was handed the equipment with instructions to “figure it out.”
- Turbidity low: there have been some big fluctuations with this one over very short periods of time… 2 to 27.7 to 4 over a 15 minute period. Maybe deer crossing the stream because the probe is located near a critter trail. The turbidity high went from 1 to 29 to 3 over the same period. It appears that any jump in the low sensor is nearly matched by a similar jump in the high sensor so I guess that indicates that they are both functioning – but how to determine if it is accurate?
Thanks for your time and help!
2022-04-28 at 11:02 AM #16961fisherbaParticipant
The CTD does not allow an internal calibration, but we do a calibration in the lab with conductivity standards and apply the calibrated data as a calculated variable.
This is maybe semantics, but the Mayfly wouldn’t be a source of quality (or the lack thereof). That’s the sensor’s job. CTDs are usually pretty durable.
2022-04-28 at 6:35 PM #16971neilh20Participant
One aspect of measurements the Mayfly does manage is the time. If the small battery gets used up, then it will reset to a default time. Another way of checking time is roughly accurate (+/ couple hours) is when the peak temperature occurs.
2022-05-03 at 11:04 AM #16994Shannon HicksModerator
DaveG, since you inherited a monitoring station from a previous owner, you didn’t get the full training we give station owners when we teach a workshop and/or install a station for them. I would suggest looking through all of our EnviroDIY Monitoring Station Manual since we thoroughly document all of the information about station assembly, installation, and maintenance. In particular, Section 8 – Monitoring Station Management addresses all of the common issues related to ongoing maintenance and support for a station, including links to quick guides for performing quality control and sensors calibration checks. There’s also 8 additional sections of appendices in the manual, with sample plots of data with explanations for the most common issues.
But for the quick answers to your questions above:
Battery voltage: the nominal voltage of the lipo battery is 3.7v. When fully charged, it will go up to around 4.2v, and can go as low as 3.55v. Below that, the logger code will prevent the sensors and cell modem from coming on in order to keep the battery from going below 3.5 v. Stations with a bad or dying battery will usually still operate on sunny days but will then stop at night, so that sort of data pattern would indicate that a new battery is required, but it sounds like your station’s battery is fine, and so is your solar panel. Occasionally the solar panels or their cables will degrade, causing the station to stop charging properly.
We usually program our loggers to transmit Mayfly board temperature (and now board humidity on v1.0 and above) so that it is some diagnostic info about the board conditions, along with battery voltage and the cellular signal strength. While the Pelican case looks like it’s an insulated box because of the foam, it definitely is not. Temperatures inside the case at night or in the shade are essentially the same as the surrounding air temperature, with a little attenuation because of the transfer of heat between the inside of the case and outside. And during the daytime if there’s direct sunlight on the case, board temperatures can get much hotter than the actual air temperature. We use the light colored tan or silver cases to prevent excess heating by the sun, and also because they visually blend in to the environment better. So in summary, you can’t really use the Mayfly board temperature as an accurate air temperature measurement, it’s only an indication of the temperature extremes the circuit board is seeing inside the case. The Mayfly circuit board can theoretically handle the standard industrial operation range of -40C to 85C (-40F to 185F), so it is unlikely your station will exceed those values. However, it is the lithium battery pack that suffers performance and lifespan issues at either of those temperature extremes so anyone planning to operate their station in the hot desert or the cold arctic might want to keep an eye on the lipo battery performance after a year or more. I’ve got stations here in Pennsylvania that have been deployed for 6 or 7 years with the same Lipo battery and logger board and have been operating continuously with no issues. The bigger killer of circuit boards and battery packs is humidity. Condensation will collect on the battery pack and logger board if there’s humid air in the Pelican case, and that condensation causes corrosion that can kill a Mayfly or battery pack. We recommend putting a desiccant pack inside the case of stations that have a tendency for humidity and condensation inside the case.
The conductivity range you mentioned seems about right for baseflow in your part of PA, however every stream and watershed is different, so you’ll want to take supplement measurements with a separate handheld conductivity meter to determine if the sensor is reading normally. There’s no way to recalibrate the sensor, it simply takes a reading and report it to the Mayfly board. The only time the reading is inaccurate is if the sensor is fouled or damaged. Fouling can occur on the electrodes of the CTD sensor that will affect the accuracy of the reading, but over-zealous cleaning of the CTD sensor could result in damage to the pressure measurement sensor that’s located right next to the electrodes, so extreme care should be taken when cleaning the CTD sensor. See section 8 of the Station Manual for more information about cleaning the sensor, and read the manufacturer’s manual (link available on that page, you have Hydros 21 Gen1 version) to learn more about how the sensor operates. The water depth is measured in millimeters of water above the pressure-sensing part of the sensor. That disk is about 15mm from the bottom of the sensor housing (on Gen 1 version). There’s a cutaway diagram in the Hydros 21 manual showing the location of the sensor components, and there’s also great information information in the quick guides (both the Maintenance Quick Guide and also in the QC Quick Guide) that can be found in section 8 of the EnviroDIY Manual.
The turbidity sensor on your station is an optical sensor (OBS3+) that can “see” about 16 inches into the water, so anything that is within that distance will affect the readings. Things like grass, sticks, fish, rocks, etc will cause the sensor to report a higher turbidity value than is actual, so you’ll need to investigate the area of influence around your sensor to see if there’s anything near the sensor in the direction that the sensor window is pointing. Sometimes mounting the sensors too close to the bottom of the stream will result in a constant reading that’s higher than normal, but if you’re seeing varying readings, my guess would be something loose nearby that’s waving in the current and causing the noise, or possibly fish activity. That particular sensor model was discontinued a few years ago by the manufacturer and has been replace by one with a much shorter measurement area, so shallow streams or cluttered channels will be much easier to monitor with the newer sensor (ClariVUE 10). Your turbidity sensor can’t be calibrated either, so the only way to know it’s working properly is to visually check that the sensor window is clean (refer to the Quick Guides for info about that) and check that you get a low reading when the water is clear. You can kick up some sediment around the sensor right at the time of measurement (every 5 minutes) to simulate high turbidity and then check that the sensor reported a higher value. You can also place your hand about one inch from the sensor window (while under the water) and you should get the max turbidity reading (around 250NTU on low scale and 1000NTU on high scale).
And as Neil mentioned, the CR1220 battery on the Mayfly board is there to keep the onboard clock synchronized. Your station is old enough that the backup battery might be getting low if the Mayfly was stored without the main Lipo battery connected for awhile. Is your station transmitting data to our website, or is it only reporting to the memory card? Does the timestamp on the data card appear accurate? For cellular-equipped stations, the logger contacts the internet time server upon initial startup to synchronize the clock, however your station is old enough that it might not have that feature, and of course it only works if you’ve got an active cellular data plan and a functioning cellular module. I think you’ve got an older 2G model which may not even have coverage in most places nowadays.
2022-05-03 at 11:17 AM #16995
Thank you once again for your thoughtful and thorough reply – as well as for your patience and understanding of my Mayfly ignorance for not having attended a workshop.
I greatly appreciate you sharing your time and knowledge. Your reply above is quite helpful.
The board already has corrosion issues on the main chip- likely humidity related as you mentioned – but it appears to be functioning properly at this point. I will suggest that we purchase a new board (and some desiccant) so we can replace it if/when it fails. There doesn’t seem to be any humidity impacts on the battery.
Our board is simply recording data to a micro sd card and we will be entering it on the network – but I wanted to make sure that it is functioning properly before I created the account and uploaded the data.
Thank you again for your kindness and support. It is greatly appreciated!
2023-06-27 at 8:15 AM #17935
After collecting nearly a year’s worth of data, it is clear that the depth sensor values are all over the place. I’m attaching a file that shows the variability between the staff gauge and the mayfly depth sensor. There doesn’t seem to be any pattern to the variation that I can determine. I tried graphing the data to see if something would become visible looking at the data visually – and the graph didn’t make make the results any easier to understand.
Could a bad circuit board (existing corrosion) account for this – or would that just be junk data (not numbers) – or is this indicative that the depth sensor itself isn’t functioning properly?
Thanks for your time and help.
2023-06-27 at 10:00 AM #17937Shannon HicksModerator
The CTD sensor you have is a digital output sensor, meaning all of the measurement happens onboard the sensor and it outputs a string of 3 numbers, all the Mayfly logger board is doing is capturing those numbers and storing them. If there was a problem with the Mayfly board, then you wouldn’t get the numbers at all, and there’s no calibration or conversion needed with those sensors, so the numbers being recorded on the Mayfly memory card are exactly the same numbers that are being sent directly from the sensor. If the data from the sensor doesn’t match up to your QA/QC measurements, then the sensor is either fouled or broken. There are a variety of ways those sensors can fail, sometimes it’s pretty obvious, and other times it’s more subtle and you need to look carefully at all the data and compare it to the real-world, hand-held measurements.
Do you have the continuous data file from the logger with all the data from the CTD sensor? Your excel chart with weekly reading indicates that something started affecting the depth readings in early January, but if you can graph all 3 CTD parameters with the continuous data, you can probably find a specific event that caused the failure. Being that time of year, it’s likely that either the sensor was subjected to freezing temperatures (which can cause catastrophic failure of one or more of the parameters) or the sensor cable jacket was damaged, which can affect the depth readings by interfering with the sensor’s atmospheric reference (essentially there’s a small air gap through the cable jacket that goes to the goretex vent on the cable end near the logger box that provides the atmospheric compensation). Or it could be that the sensor is just fouled with something on the pressure disk and needs a thorough cleaning. Have you removed it from the stream and examined it carefully and also examined the cable jack along the entire 30-foot length, looking for any nicks or cuts or holes? What depth and conductivity reading do you get on the logger when the sensor is out of the water?
2023-06-27 at 11:20 AM #17938
Thanks again for such a prompt and thorough reply, yet again.
We have continuous data starting in last July. The variation in depth over four months was consistently in the neighborhood of +265-275mm. There was a period of very cold temps around Christmas and the edge of the stream froze but I don’t think that the probes were encased in ice. However, after that freeze event, the variation in depth (b/t digital sensor and visual measurement on staff gauge) became much more variable, without any logical pattern.
I will do a thorough check of the cable, record readings when probes are out of the water, and see if I can clean the pressure disk. The turbidity measurements seem more accurate and responsive to weekly cleaning – both before and after the freeze event – so the issue is isolated to the depth sensor or the cable. Will post update later in the week when I get a chance to do more investigations and a more thorough cleaning.
2023-06-27 at 11:45 AM #17939
I just reviewed the data during the week of 12/19/22. There was a quite a spell with board temps down to -17 but I didn’t see any evidence of freezing in the data. The turbidity sensors appeared to be recording as normal (didn’t go to zero at all) and the depth didn’t show any big increase (as was illustrated in the troubleshooting section).
It is possible that the cold temps might have created a crack in the cable housing so I’ll have to look at that in more detail.
- You must be logged in to reply to this topic.