Interpreting a Student Stream-Quality Comparison
A ninth-grade Earth science class of 72 students samples a creek at two sites: Site U, 300 m upstream of a wastewater treatment plant outfall, and Site D, 150 m downstream. Each of three class periods visits once on a different day. On the first day, 2.1 cm of rain fell before sampling. Students measure turbidity and pH with handheld meters at one spot per site, one reading each, between 9:30 and 10:15 a.m. at Site U and 1:45 and 2:30 p.m. at Site D. Results: mean turbidity 12 NTU upstream and 45 NTU downstream; mean pH 7.4 upstream and 7.8 downstream. The class concludes the plant raises turbidity and pH. The principal asks whether the conclusion is defensible and how to improve the investigation for next year.
1. Evaluate whether the class's conclusion is supported by the data. Identify at least three design problems and explain how each could produce a misleading result.
Read the worked answer
The conclusion is not defensible as stated. The comparison is confounded by time: upstream was sampled in the morning and downstream in the afternoon, so temperature-driven algal or sediment changes and discharge changes could explain the difference. The rain event adds runoff and suspended sediment, and it may affect the two sites differently. One grab sample per site gives no estimate of local variability or meter error, and there are no field duplicates or blanks. The upstream site is not a true before-treatment control because it is 300 m away and may receive other inputs. The pH difference of 0.4 units is small relative to typical meter uncertainty and could be caused by timing or calibration. A defensible claim needs paired sampling at the same time, repeated visits across wet and dry conditions, and replicate readings.
2. Propose a revised sampling plan that would let the class test the plant's effect more fairly. Include specific numbers of sites, sampling times, replicates, and a comparison strategy.
Read the worked answer
Sample both sites on the same day within 30 minutes, alternating the order, on at least six dates: three after at least 1 cm of rain and three after 72 dry hours. At each site collect three replicate readings at three evenly spaced points across the channel, at 10 cm depth if safe, and record temperature, conductivity, an estimated discharge, and time. Add a third reference site 300 m upstream of Site U to detect background trends. Use the same calibrated meter for all readings and record a field blank and a duplicate. Compare paired downstream-minus-upstream differences by date rather than pooled means. This design reduces diurnal and storm confounding and gives an uncertainty estimate.
3. The class has one handheld turbidity meter with a stated accuracy of plus or minus 2 NTU and a calibration check that is two months overdue. How should this affect interpretation, and what should they do before reusing it?
Read the worked answer
The meter's plus or minus 2 NTU accuracy is smaller than the 33 NTU difference, so random meter error alone cannot explain the result, but a missed calibration check can create a systematic bias that affects one site or one day. If the meter reads 5 NTU high after drifting, both sites are shifted, and the difference may be less affected; if drift changes during the day, the morning/afternoon comparison is biased. Before reuse, the class should run a 0 NTU and a known standard, record the check, recalibrate if outside tolerance, and repeat a few samples with a second meter. Until then, report the turbidity difference as suggestive, not proof.