Conclusion and evaluation
A finished analysis. Both of these sections are judgements about work that already exists, so writing them before the analysis is settled means writing them twice.
- A conclusion that answers the fieldwork question, in numbers
- Strengths and weaknesses that are specific to your investigation
- Improvements, each with its likely effect explained
- An honest position on how far your findings can be trusted
Two marks for a conclusion built from your findings, three for evaluating everything from your methods to your choice of location, and they are the easiest five to lose.
The conclusion is short and mechanical: answer the question, using your data, adding nothing new. The evaluation is where reports collapse into generalities, and the mark scheme is unusually explicit about what it wants instead.
Summarise your findings and make a clear, concise statement answering the fieldwork question, consistent with the analysis. Quote data. Introduce nothing new, no fresh ideas, no new evidence, no first appearance of an argument.
Two marks: 1 if the conclusion is partially supported by the analysis, 2 for a clear conclusion supported by it.
Review the methodology. Weigh the strengths and weaknesses of the data collection methods, the fieldwork question itself, the presentation of the data, and the choice of location: the river itself, and the number and placement of sites. Suggest realistic improvements, and explain what each would actually change.
Three marks: listed and superficial · outlined and mostly relevant · explained, with the impact of improvements explained.
Everything below is how we suggest you actually do it.
The conclusion: answer the question you asked
Two marks, 200 words, and a narrow job. State what you found, answer the fieldwork question directly, and stop.
Answer it in numbers, which means your own. Name the variables that answered the question, give the values at each end of the reach and the direction of travel, and say plainly whether that answers the fieldwork question or only part of it. A conclusion built from figures you have already presented is a conclusion supported by the analysis, and that is the whole of the top band. One built from adjectives is not.
Then be honest about scope. If one sub-hypothesis was supported and another was not, say so here; a conclusion that quietly reports only the successes is not consistent with the analysis, and consistency is what the descriptor asks for.
No new ideas. Not a new explanation, not a new source, not a limitation making its first appearance. Everything in the conclusion should already have been argued upstream.
Overclaiming is the tempting mistake, because the overclaimed version always sounds more impressive. Three sentences, each of which sounds like a conclusion to this investigation, and none of which the data can carry:
"Discharge increases downstream, which proves the Bradshaw model." Group 5's discharge rises from 0.0015 to 0.234 m³ s⁻¹ between sites 1 and 8, then loses 96% of it in the 200 m to site 9. One variable, broadly consistent until it is not, on one river on one day, proves no model. "Consistent with" is as far as the sentence can go, and site 9 has to appear in it.
"Site 9 must be a measurement error." Perhaps, but the conclusion cannot say so, because nothing was re-measured. Calling a reading an error is a claim about the methods, and it needs the evidence trail from step 4, the map, the photographs, the field notes, not a trend the reading happens to spoil.
"The renaturation works have reduced the river's discharge." The scheme is real and sites 8 and 9 sit in its reach, which is what makes this sound informed. But one day's readings contain no before, no after and no unmodified comparison, so the design cannot carry a causal claim about the works. What the data licenses is smaller and better: a candidate explanation for site 9, argued in the analysis, not announced in the conclusion.
Restating the analysis. A 200-word paraphrase of the section above it earns nothing and costs 200 words.
Overclaiming. "This proves the Bradshaw model." Nine sites on one river on one day do not prove a model. "Consistent with" is the honest phrase, and it costs you nothing.
Hedging into meaninglessness. "The results were somewhat inconclusive but generally showed some change." If the data answered the question, say so.
The evaluation: specific, or it does not count
The bottom band is listed, mostly superficial, not appropriate or not relevant. That is a precise description of the evaluation almost everyone writes first: human error, the weather, more sites, better equipment, more time.
Those are not wrong. They are unfalsifiable, true of every fieldwork investigation ever conducted, which means they say nothing about yours. The fix is to attach every weakness to something that actually happened.
"There may have been human error when measuring. More data could have been collected and better equipment used. The weather could have affected the results."
Could have been written before the fieldwork day. Could have been written about any river.
"Wetted perimeter was measured with a tape laid along the bed, which lifted off between boulders at the upper sites where the bed is coarsest. This systematically underestimates wetted perimeter where the bed is roughest, and since hydraulic radius is area divided by wetted perimeter, it overestimates hydraulic radius at exactly the sites where the analysis claimed it was lowest, weakening the strongest part of the velocity explanation. A chain pressed into the bed profile would follow the gaps the tape bridged, and re-measuring one coarse site with it would show how large the underestimate actually is."
Names the technique, the sites, the direction of the error, what it does to a specific conclusion, and an improvement whose effect is stated, which is what the band asks for.
Notice what makes the second version work: it states the direction of the error and follows it through to a claim. Not "this may have affected accuracy" but "this makes the number too high, at these sites, which weakens this argument".
If you are stuck for one evidenced limitation, start with the flowmeter. Geopacks quote it at ±5% and give it a working range of 0.5 to 1.5 m s⁻¹ (step 3 has the reference).
That gives you all three things the top band wants in one move: a number, a source, and a direction. Say which of our sites fell outside the working range, and say which comparison between sites is too close to survive ±5%.
The descriptor says strengths and/or weaknesses, and an evaluation that is nothing but self-criticism is as unbalanced as one with none.
You have real strengths to name: all nine sites measured on the same day under the same flow conditions, repeated velocity trials, a wide altitude range, and every member of the group recording every reading independently. Say what each one lets you trust.
The four things the top band names
The top band of Criterion F (Evaluation) lists four targets. Give each one a short paragraph and you have structured the evaluation for yourself.
The one everybody writes. Be specific: which technique, at which sites, in which direction, affecting which conclusion.
Rarely written, and easy marks. "Fluvial characteristics" is broad, eight variables answering it thinly. A question naming two or three would have been sharper. You were given this question, and saying how you would refine it is exactly what the descriptor asks.
Which figure did not work, and what would have shown it better? A proportional-symbol map that made small values invisible; cross-sections at a scale that hid the shallow sites.
The river itself, and the number and placement of sites. Nine, placed by public access rather than systematically, with a 2.3 km gap at the top and two just 40 m apart in the middle; step 4 showed what n = 9 does to a statistical test. And the engineered reaches, where distance from the source is not the only thing controlling the channel.
The Asse gives you a genuinely strong location paragraph, because it cuts both ways.
For: 298 m of fall across 8.9 km, from close to the source into the lower course in a single day, under identical flow conditions.
Against: the lower reaches are exactly where the flood-protection and renaturation scheme is reshaping the channel, so the sites that should show the clearest downstream trend are also the ones whose form is partly a design decision. Distance from the source is not the only independent variable acting on them.
A location weakness that is really about your own results is worth far more than "a longer river could have been used". Say which sites are affected, and what it does to the confidence you can place in the trend.
Improvements: say what each one would change
The top band asks for improvements outlined and the potential impact of these improvements explained. That second half is where the mark is, and it is the half that gets left off.
So never stop at the suggestion. Every improvement needs a because.
| Improvement | …and what it would change |
|---|---|
| A chain rather than a tape for wetted perimeter | It follows the bed's roughness instead of bridging it, removing the systematic underestimate at the coarse upper sites and making hydraulic radius comparable along the whole river |
| Five velocity trials rather than three | A more reliable mean at sites with turbulent flow, and a wider spread to quote; three readings tell you little about how variable the flow really was |
| Sites at fixed 1 km intervals | Even coverage, no clustered pair at effectively the same distance, and a Spearman's result not weighted towards wherever access happened to be easy |
| Repeat measurements in a different season | Separates the downstream trend from the particular discharge on the day, turning a snapshot into a comparison |
Keep the improvements realistic. "A gauging station at every site" is not an improvement to this investigation; it is a different investigation with a different budget.
The booklet also asks how the study could be extended, and the Asse offers an obvious one: compare an engineered reach with a natural one at a similar distance from the source, holding distance roughly constant and varying human modification instead.
That turns the thing that complicated this investigation into the subject of the next one, which is a much better answer than "measure more sites".
It can argue with a limitation you have already identified. Describe the weakness you found, say why you think it matters, and let it push back; that is a conversation about your own thinking.
It cannot generate the limitations. Ask a tool to evaluate a study it cannot see and it will produce human error, weather, and more sites, which is, almost word for word, what the bottom band means by "superficial, not appropriate, or not relevant". The generic answer is the low-scoring answer.
Ready for step 8?
0 of 13Read your evaluation and cross out every sentence that would be true of any river fieldwork anywhere. What is left is the part that scores.
Step 8 turns the sections into one report: the word count and what escapes it, referencing, being honest about AI, and what actually gets submitted.
Anything marked “our river” is specific to the fieldwork we do together on the River Asse at Nyon: the sites, the equipment and the way we collect the data. Site coordinates and altitudes are from our own fieldwork records and the exact list can change from year to year, so check them against the sheet you are given on the day. Photographs are the author’s own, taken at the river.