
compare competing forecast methods with a weighted decision matrix, use them to predict 2050 carbon dioxide, temperature and sea level, and build a first forecast card that states ranges and their sources of uncertainty.
Phase 4: Build
Do Now. Two classmates forecast your town’s population in 2050. One draws a straight line through the last 50 years; the other uses the last 10 years only. Write one situation in which each would be the better choice.
Carbon dioxide's rate of rise has been increasing. A straight line fitted to 2000–2024 is extended to 2050. Its answer is probably…
The design problem. The planning office needs a forecast it can design to. Our criteria: each quantity has a range and a central value; each range names its method; the forecast has been checked against the past; it shows how choices change the answer; it fits on one page a non-scientist can read. Our constraints: public data only, two class periods, and sources cited.
Three competing methods.
Worked example 1: a weighted decision matrix. Score each method from 1 to 5 against each criterion, multiply by the weight and add.
| Criterion (weight) | A: Long-line extrapolation | B: Recent-rate extrapolation | C: Class carbon model, three scenarios |
|---|---|---|---|
| Reproduces the past (5) | 2 | 3 | 4 |
| Based on physical mechanism (4) | 1 | 1 | 4 |
| Shows effect of human choices (4) | 1 | 1 | 5 |
| Easy for the public to follow (3) | 5 | 5 | 3 |
| Low effort to update (1) | 5 | 5 | 2 |
| Weighted total | 38 | 43 | 67 |
Scores are one class’s sample judgments on a 1–5 scale, not measured values.
Method C: 5×4 + 4×4 + 4×5 + 3×3 + 1×2 = 20 + 16 + 20 + 9 + 2 = 67. The matrix favors the model, but A and B are still useful as simple checks: if the model disagrees wildly with the recent trend, something is wrong.
Worked example 2: temperature and sea level. Temperature: the 2015–2024 average was about 1.2 °C above the late 1800s; extending 0.20 °C per decade for three decades adds about 0.6 °C, giving about 1.8 °C in 2050. Sea level at the sample harbor: 2.0 mm per year adds 5 cm by 2050, 2.4 mm per year about 6 cm, and the IPCC’s recent global rate of 3.7 mm per year about 9 cm. Because the rise is accelerating, a careful range runs from about 5 to 15 cm, with extrapolation of a past rate as the low end.
Sources of uncertainty. Name them on the card: future emissions (the biggest by far), how much the ocean and land will keep absorbing, natural swings such as El Niño, gaps or bias in the data (a single gauge, a moved station), and what the models leave out.
Prototype: Forecast Card V1. One page: three panels (CO2, temperature, harbor sea level), each with a small graph, a 2050 range, the method, a confidence word (high, medium, low) and one sentence on its main uncertainty. A footer lists sources.
Score, compute and draft in your team.
Build Forecast Card V1.
Make the card: three panels with a small graph each (organize the data you used, with units), a 2050 range and central value, the method, a confidence word and one sentence naming the main source of uncertainty. Keep it to one page.
Under the card, write two sentences explaining your team's decision matrix result and why you kept a simple extrapolation as a check.
Version 1 exists. Next lesson it faces the past, a gallery of critics and a rewrite.
How confident are you that you can I can compare forecast methods and state a range with its sources of uncertainty.
Exit ticket: which panel of your card is least certain, and what is the main reason?
Materials: Mauna Loa table, sample gauge table, model runs from 3.05, card templates (or blank paper), colored pencils, devices optional.
Facts to keep careful: the 2.27 ppm per year slope and about 482 ppm are from a least-squares fit to annual means, as Poly gave them; the fill-blank’s simpler extension from 2024 gives about 484. Matrix scores are sample judgments.
Listen for: teams scoring the matrix to justify a choice made in advance; ranges with no method; temperature ranges that confuse the global anomaly with the harbor’s air temperature.
Support / stretch: give a card template with labeled boxes. Stretch: change one weight and show whether the matrix winner changes.