Worked example · Geography · Year 8

Modelling cliff retreat at Porth y Rhedyn

DCF: Data and computational thinking → Problem-solving and modelling

What this is

A secure Year 8 response, annotated against the success criteria. It is also your answer key. Mean rate: 36.0 m ÷ 24 years = 1.5 m per year. Year reached at 1.5 m per year: path 2031, caravans 2040, road 2048, houses 2064. More storms (1.875 m per year): 2030, 2037, 2044, 2056. Rock armour (0.6 m per year): 2038, 2060, 2080, 2120. Test: the 2002 to 2010 rate is 1.35 m per year, which predicts 18.6 m for 2026, 3.6 m more than the real 15.0 m. The sheet asks pupils to round years up. If a pupil rounds to the nearest year and says so, accept answers one year earlier where that applies.

Optional spreadsheet route: type the inputs and the Step 4 formulae below into any spreadsheet, then copy row 11 down to row 24. Change B6 to 1.25 for more storms or 0.4 for rock armour.

Step 1: why the cliff retreats

ProcessHow it makes the cliff retreat
Hydraulic actionBig waves from the south-west hit the cliff foot and force water and air into cracks. The pressure breaks lumps of clay off.
AbrasionThe waves throw the beach pebbles at the cliff foot, which scrapes it away and cuts into it.
WeatheringRain soaks into the clay and it wets and dries, so it cracks and weakens from the top, even when the sea is calm.
SlumpingWhen the clay is full of water it is heavy and slippery, and a whole section slides down a curved surface. The sea then removes it and the cycle starts again.

Boulder clay is soft and not held together like solid rock, so waves wear it away quickly. It also soaks up rain and slumps, so it loses metres at a time. Hard rock resists hydraulic action and abrasion and does not slump, so it loses only centimetres a year.

Each process is described accurately and linked to this cliff (the south-westerly waves, the pebbles, the rain). The comparison gives two separate reasons: strength against waves and slumping when wet. Evidences criterion 1.

Step 2: break down the problem

InputsRuleOutputs
The survey distances and years. The distance from the 2026 edge to each feature. A multiplier for each scenario.Rate = total retreat ÷ years. Each year, distance left = last year’s distance left − rate. A feature is reached when the distance left is 0 m or less.The distance left to each feature every year, and the year each feature is reached, for each scenario.

Assumption: the cliff retreats by the same amount every year, but really it can lose 4 m in one night in a slump and almost nothing in a calm year.

Inputs, rule and outputs are kept apart, and the rule is precise enough for a computer to follow, including the stopping test. The assumption is tied to evidence from the survey note. Evidences criterion 2.

Step 3: the retreat rate

BetweenRetreat (m)YearsRate (m per year)
2002 and 200651.0 − 45.8 = 5.245.2 ÷ 4 = 1.3
2006 and 201045.8 − 40.2 = 5.645.6 ÷ 4 = 1.4
2010 and 201440.2 − 31.8 = 8.448.4 ÷ 4 = 2.1
2014 and 201831.8 − 26.6 = 5.245.2 ÷ 4 = 1.3
2018 and 202226.6 − 20.6 = 6.046.0 ÷ 4 = 1.5
2022 and 202620.6 − 15.0 = 5.645.6 ÷ 4 = 1.4

(a) Total retreat = 51.0 − 15.0 = 36.0 m. Total years = 2026 − 2002 = 24. Mean rate = 36.0 ÷ 24 = 1.5 m per year. Check: 5.2 + 5.6 + 8.4 + 5.2 + 6.0 + 5.6 = 36.0 m.

(b) 2010 to 2014, at 2.1 m per year, when every other period is 1.3 to 1.5. The storms of winter 2013 to 2014 caused a slump that took about 4 m in one night, so this period includes a sudden event, not just steady erosion.

Every difference and division is shown, the mean uses total retreat over total years, and the pupil checks the total by adding the six differences. The unusual period is explained with the surveyor’s evidence. Evidences criterion 3.

Step 4: the spreadsheet design

CellFormulaWhat it does
B3=C1-B1Years between the surveys: 24
B4=B2-C2Total retreat: 36.0 m
B5=B4/B3Mean rate: 1.5 m per year
B7=B5*B6Rate for the scenario chosen in B6
A11=A10+1Next year
B11=B10-$B$7Path: last year’s distance minus one year of retreat
C11=C10-$B$7Caravans, the same rule
D11=D10-$B$7Road, the same rule

To run a different scenario I only change B6: 1.25 for more storms or 0.4 for rock armour. Everything else updates itself.

Every formula uses cell references rather than typed numbers, so the model updates when an input changes. The $ signs keep B7 fixed when row 11 is copied down, and one input cell controls the scenario. Evidences criterion 4.

Step 5: running the model

YearPathCaravansRoadYearPathCaravansRoad
20267.020.032.02034-5.08.020.0
20275.518.530.52035-6.56.518.5
20284.017.029.02036-8.05.017.0
20292.515.527.52037-9.53.515.5
20301.014.026.02038-11.02.014.0
2031-0.5 (reached)12.524.52039-12.50.512.5
2032-2.011.023.02040-14.0-1.0 (reached)11.0
2033-3.59.521.5
FeatureDistance (m)Time = distance ÷ rateRounded upYear reached
Road3232 ÷ 1.5 = 21.33 years222026 + 22 = 2048
Houses5656 ÷ 1.5 = 37.33 years382026 + 38 = 2064

Formula for the houses: =56/B7, or =ROUNDUP(56/B7,0) to round up as well. Check with the path: 7 ÷ 1.5 = 4.67, rounded up is 5, and 2026 + 5 = 2031, which matches my table.

The model is run accurately year by year, and the stopping rule is applied the same way in every column. The pupil switches to time = distance ÷ rate when adding rows would be slow, rounds up for a reason, and checks the shortcut against the table. Evidences criteria 4 and 5.

Step 6: two scenarios

(a) More storms: 25% of 1.5 = 0.25 × 1.5 = 0.375. New rate = 1.5 + 0.375 = 1.875 m per year (or 1.5 × 1.25).

(b) Rock armour: 60% of 1.5 = 0.6 × 1.5 = 0.9. New rate = 1.5 − 0.9 = 0.6 m per year (or 1.5 × 0.4).

FeatureMore stormsRock armour
Time (years)YearTime (years)Year
Path (7 m)7 ÷ 1.875 = 3.73, up to 420307 ÷ 0.6 = 11.67, up to 122038
Caravans (20 m)20 ÷ 1.875 = 10.67, up to 11203720 ÷ 0.6 = 33.33, up to 342060
Road (32 m)32 ÷ 1.875 = 17.07, up to 18204432 ÷ 0.6 = 53.33, up to 542080
Houses (56 m)56 ÷ 1.875 = 29.87, up to 30205656 ÷ 0.6 = 93.33, up to 942120

(c) The armour lasts about 50 years, so only until about 2076. The road (2080) and houses (2120) answers assume it works for 54 and 94 years, longer than it lasts. After 2076 the cliff would go back to retreating at 1.5 m per year or faster, so those features would be reached earlier. The armour also does not stop rain soaking the clay, so slumping can carry on.

Both percentage changes are worked out in full, the unrounded 1.875 is kept in every division, and only the final answers are rounded. Part (c) uses the model’s own output to spot where its assumption breaks down. Evidences criteria 5 and 6.

Step 7: testing and evaluating

(a) Rate from 2002 to 2010 = (51.0 − 40.2) ÷ 8 = 10.8 ÷ 8 = 1.35 m per year. From 2010 to 2026 is 16 years, so 16 × 1.35 = 21.6 m of retreat. Prediction for 2026 = 40.2 − 21.6 = 18.6 m. The real survey was 15.0 m, so the model was 3.6 m out: the cliff retreated further than it predicted.

(b) The model is useful for planning but its years are not exact. When I tested it using only the early surveys, it was 3.6 m out by 2026, because the 2013 to 2014 storms were not in the early data. It assumes the cliff retreats at a constant rate, but slumps take several metres at once, so the path could go years before 2031. It also assumes the future is like the past, but rising sea levels and more storms could speed it up, as the 25% scenario shows. It uses one marker post for the whole village, but the cliff may retreat faster in some places. The council should plan using the storm scenario, for example moving the coast path before 2030 and planning a new lifeboat road well before 2044. To improve the model, I would add more marker posts along the cliff and re-run it after every survey with the new data.

The test uses only data the model would have had in 2010, compares the prediction with a real result and explains the error. The report names three assumptions with reasons, gives cautious advice based on the outputs and suggests a practical improvement. Evidences criterion 6.

If you finish early: answers

Why this response is secure