Dissolved Oxygen Monitoring: The Reading Nobody Is Awake For
TL;DR: The UK’s statutory river standard for oxygen is not an average. It is a 10 percentile, which means the classification is decided by the worst tenth of the readings. In a river those readings happen shortly before dawn, when nothing has photosynthesised for eight hours and everything has carried on breathing. A grab sample taken during the working day is drawn from the top of that distribution, which is why dissolved oxygen measured by visit systematically flatters the water.
Last updated: 11 September 2026
On this page
Key takeaways
- Dissolved oxygen in a river follows a daily cycle: production only in daylight, consumption around the clock. The minimum arrives just before dawn.
- The statutory standard is expressed as a 10 percentile of percent saturation, not a mean. It is a rule about the bad tenth of the year.
- For type 1, 2, 4 and 6 rivers and salmonid rivers, good status needs 75% saturation at that 10 percentile. High needs 80%.
- For type 3, 5 and 7 rivers the same thresholds are 70% for high and 60% for good.
- Grab sampling in working hours draws from the upper half of the daily distribution, so it cannot estimate the lower tail the standard is written against.
- Continuous dissolved oxygen measurement is not about more data. It is about sampling the part of the day the rule is actually about.
Why does dissolved oxygen collapse overnight?
Because production stops and consumption does not. Plants and algae put dissolved oxygen into the water while there is light to drive photosynthesis. Every living thing in the river, plants included, takes dissolved oxygen out continuously. As soon as the light goes, one side of that balance switches off and the other keeps running until sunrise.
The University of Florida’s extension service puts the shape plainly in its water quality notes on the subject: graph a week of readings and you see seven peaks in the late afternoon and seven low points around dawn. That is not a fault or an anomaly. It is the normal daily respiration of a living watercourse, and the healthier and more productive the river, the larger the swing.
Three processes set the curve. Photosynthesis adds dissolved oxygen during daylight. Respiration removes it continuously. Reaeration, the physical exchange across the water surface, pulls the concentration back towards saturation from whichever side it has strayed. A shallow, turbulent, shaded stream has strong reaeration and a small daily swing. A slow, sunlit, nutrient-rich reach has weak reaeration and a large one.
Two things then make the pre-dawn minimum worse. Warm water holds less dissolved oxygen at saturation than cold water, so a hot still night lowers the ceiling at the same time as respiration is lowering the floor. And any organic load discharged into the reach, from a storm overflow, a farm, or a works, is consumed by bacteria that take their dissolved oxygen from the same water column.
What the UK standard actually asks for
Not an average, and this is the part that changes how you should monitor. The Water Framework Directive (Standards and Classification) Directions (England and Wales) 2015 set the river dissolved oxygen standard as a percentage of saturation at the 10 percentile. Ten percent of the record is allowed to sit below the threshold. Ninety percent has to sit above it.
The thresholds themselves are set out in Table 1 of the Directions, and they depend on which type the river has been placed in, which in turn depends on its altitude and its alkalinity:
- Types 1, 2, 4 and 6, and salmonid rivers: high 80%, good 75%, moderate 64%, poor 50%.
- Types 3, 5 and 7: high 70%, good 60%, moderate 54%, poor 45%.
Read that as an instruction to the monitoring, not just to the river. A 10 percentile is a statement about the shape of the lower tail of a distribution. To estimate it you need a sample that represents the whole distribution, including the hours nobody is on site. A dissolved oxygen record designed around a mean will pass a reach that the statutory test would fail.
Why a daytime sample cannot estimate a 10 percentile
Because it is not a random sample of the day. If dissolved oxygen follows a strong daily cycle and every visit happens between roughly nine and five, then every reading is drawn from the rising and peak part of that cycle. The result is not noisy, which would be survivable. It is biased, which is not.
Noise averages out with more samples. Bias does not. Take a hundred grab samples in working hours and you get a very precise estimate of the daytime distribution and no information at all about the pre-dawn one. The 10 percentile of your sample is not the 10 percentile of the river, and the gap between them grows with exactly the thing you care about most: the size of the daily swing.
The perverse consequence is that the reaches where visit-based dissolved oxygen sampling is most wrong are the productive, nutrient-enriched, slow-moving ones, because those have the biggest amplitude. The reaches where it is nearly right are the cold, turbulent, well-mixed ones that were never at risk. Your monitoring is most accurate where it matters least.
This is the same structural problem we describe in our piece on pressure transients, where the instrument is simply somewhere else in time when the interesting thing happens. Different parameter, identical failure: the sampling regime and the phenomenon are out of phase.
What a continuous record shows that visits cannot
Four things, none of which can be reconstructed from spot samples. Each of them changes a decision rather than just filling a spreadsheet.
It is worth being precise about why reconstruction fails. A daily cycle is not random variation around a mean, it is a deterministic shape with a known period. Sample it sparsely and at a fixed hour and you do not get a blurred version of the shape, you get a single phase of it repeated. Statisticians call this aliasing, and it is the reason a fortnightly round can run for a decade without ever describing the river it visits.
- The actual minimum, and when it happens. The number the statutory test is built around, measured rather than inferred.
- The amplitude of the daily swing. A wide swing is an early warning of enrichment long before the annual classification moves, because it says the reach is running a large biological engine.
- The sag after an event. A discharge shows up as dissolved oxygen falling and then recovering over hours. That curve, with its timing, is evidence. A single bottle is an anecdote.
- Recovery time. How long the reach takes to return to its normal curve is a better measure of resilience than any single reading, and it only exists if you were watching continuously.
That third point is where a continuous dissolved oxygen record earns its keep commercially. If you operate an outfall, a works or a site with a consent, a dissolved oxygen trace that runs through the night converts an argument about whether you caused something into a timestamped curve that either implicates you or clears you. Both outcomes are worth paying for.
What continuous monitoring needs on site
Less than people expect at the sensing end, and more than people expect at the maintenance end. The instrument list is short: a dissolved oxygen probe in the water, something to carry the reading back, and power that survives a winter.
- The probe. An optical dissolved oxygen sensor such as the SpecSens SPS-DO, deployed in situ. Where you need chemistry alongside it, the iSPS-X multi-parameter instrument covers several parameters in one body, which matters when the chamber has room for one thing.
- The link. An outdoor LoRaWAN gateway such as the Milesight UG67 at £658.39 ex VAT, or the solar SG50 at £683.83 where there is no mains at the bank. Both prices exclude VAT.
- The interface. If the probe presents a 4-20mA output, a LoRaWAN IO controller reads it directly. That path is set out in our guide to 4-20mA current loops, and it is usually cheaper than replacing an instrument you already own.
The maintenance end is where honesty is owed. Optical dissolved oxygen sensors hold calibration better than the electrochemical parameters sitting next to them, but they still foul. Biofilm, silt and weed change the reading slowly enough that you will not notice from the trace alone, which is why the calibration schedule is part of the specification and not an afterthought.
What it will not do for you
It will not make your instruments self-maintaining, and it will not automatically satisfy a permit that names a certified instrument class. Both of those are worth settling before anything is ordered, because both of them end projects late.
On certification, some permits accept instrument-class continuous data and some specify MCERTS-certified equipment for a named parameter. Those are different requirements with different price tags. We will tell you which one your permit is asking for before quoting, rather than letting you assume coverage that is not there. That is standard practice across our water quality monitoring work.
On maintenance, continuous dissolved oxygen measurement reduces truck rolls, it does not remove them. Grab sampling stops being the daily workload and becomes the verification and calibration visit. That is a real saving, and it is a smaller one than the marketing usually implies.
Nor will it settle every argument on its own. A dissolved oxygen trace tells you what the water did, not who made it do that. Pairing an upstream point with a downstream one is what turns a measurement into an attribution, which is why we usually quote two positions rather than one.
And it will not fix the river. A continuous record tells you the reach dropped to 64% saturation at 04:40, that the swing has widened over three summers, and that the recovery after the last spill took nine hours. What you do with that is a catchment question, not an instrument one.
Frequently asked questions
Should we measure mg/l or percent saturation?
Both, and record temperature with them. The statutory river standard is written in percent saturation, which already accounts for the fact that warm water holds less dissolved oxygen. Concentration in mg/l is what the fish experience. A sensor that gives you both plus temperature lets you answer either question later.
How often should a dissolved oxygen sensor report?
Often enough to resolve the shape of the daily curve and the recovery after an event, which in practice means minutes rather than hours. Fifteen minutes captures the diel cycle comfortably. Anything hourly starts to blunt the minimum, which is the one value you are there for.
Which river type are we?
It is set by altitude and alkalinity in the Directions, with 80 metres as the altitude split and alkalinity bands as calcium carbonate in mg/l. Salmonid rivers take the stricter column regardless. The Environment Agency holds the classification for your water body, and it decides which threshold applies to you.
Can one sensor cover a whole reach?
No. Dissolved oxygen varies along a river as well as through the day, and a single point tells you about that point. Two positions, one upstream and one downstream of whatever you are interested in, answer far more questions than one position ever will, because the difference between them is the effect.
How much does continuous monitoring cost?
The probe dominates, the gateway is a one-off shared across every sensor in range, and the dashboard is annual. We quote the water quality instruments individually because the right probe depends on the parameter and the permit. Prices come back itemised within one working day.
Where this leaves you
The case for continuous dissolved oxygen measurement is not that more data is better. It is that the UK wrote its river dissolved oxygen rule as a 10 percentile, and a 10 percentile is a claim about the hours a sampling round never covers. A monitoring regime that only operates in daylight is not measuring the thing the law is about.
Tell us the water body, the parameters your permit names and whether there is power and network at the bank, and we will specify the probe, the link and the calibration interval, with the certification question answered up front. Start from our water quality monitoring solution, or read the non-revenue water guide if your interest is the network rather than the river.
Next step
Get a priced kit list for your site
Answer three quick questions and tell us where to send it. An engineer replies within one working day with the parts, the prices and the lead time.
Rather talk it through? Call 023 9223 3611
Received. An engineer has it.
You will hear back within one working day with the parts, the prices and the lead time. A confirmation is on its way to your inbox.
If it is urgent, call 023 9223 3611.
