The Oldham projectA flow duration curve for the Tame
Download the daily flows, check what is really there, then rank the days.
The learner downloads every daily mean flow for the Uppermill gauge: 10,284 rows from 10 July 1998 to 25 September 2026. The first check finds only 21 calendar days with no row at all, which looks reassuring. The second check is the one that matters. Another 714 rows exist but are flagged Missing and hold no value, including one unbroken run of 636 days from November 2016 to August 2018. A program that only counts rows would believe the record is almost complete. Each row also carries a quality flag: 8,504 Good, 923 Unchecked, 123 Estimated and 20 Suspect.
Tame at Uppermill, flows exceeded for a share of days, water years 1999 to 2025, our Python run, 28 September 2026| Measure | Flow exceeded (cubic metres a second) | Meaning |
|---|
| Q95 | 0.211 | Exceeded on 95% of days, a low flow |
| Q90 | 0.239 | Exceeded on 90% of days |
| Q70 | 0.382 | Exceeded on 70% of days |
| Q50 | 0.609 | The median day |
| Q10 | 2.433 | Exceeded on only 10% of days |
| Q5 | 3.51 | Exceeded on only 5% of days, a high flow |
To build the curve the learner keeps whole water years, October to September, from 1999 to 2025, takes the Good and Estimated days, 8,544 of them, and sorts the flows from highest to lowest. Each flow gets a rank, and the rank divided by the number of days gives, near enough, the share of time that flow is equalled or beaten. Here sits the classic trap. Q95 means the flow exceeded 95% of the time, which is the 5th percentile, not the 95th. Ask a library for the 95th percentile and label it Q95 and you get 3.51 instead of 0.211, turning the river's lowest flows into its highest. The mean, 1.076, is about 1.8 times the median, and only 30.3% of days are above it, because a few wet days pull the average up.
Then comes the subtle part. Dropping the Unchecked days feels cautious, but they are not a random sample. In those water years there are 581 of them, mostly between May and September, and their median flow is 0.165, below the Q95 of the checked days. Including them lowers Q95 to 0.181 and the median to 0.568. So the exclusion rule shifts exactly the low-flow end that abstraction planners care about. The learner reports both versions side by side and writes down why, rather than choosing the one that looks tidier.
Ages 8 to 11
Record a week of rain in a jar, sort the days and find the middle one.
Ages 11 to 15
Sort the Tame flows in Python and read off Q50 and Q95.
Ages 15 and up
Handle flags and gaps, plot the curve, and test each exclusion rule.
Environment Agency data, our curve
Daily flows and quality flags come from the Environment Agency's Hydrology service; station details come from the National River Flow Archive. The filtering, the ranking and every figure in the tables are our own calculation.
