CIVL6001 Chap.4 Climate Models, Scenarios and Projections
Climate Models, Scenarios and Projections
Global climate models represent physical processes in the atmosphere, ocean and land to explore system responses under specified forcing pathways. Their future results are projections conditioned on scenarios, not single unconditional predictions. This chapter keeps pathway, baseline, period and model set attached to every value, then explains why downscaling can add local detail without eliminating uncertainty.
Models divide the Earth system into computational cells and solve approximations to conservation and process relationships through time. Processes smaller than the grid require parameterisation, and different model structures can produce different responses to the same forcing. Ensembles expose this spread rather than turning it into an error that one preferred model can remove.
Future results also depend on emissions or concentration pathways and on internal variability, particularly at regional scales and over shorter periods. A projection is therefore incomplete unless it identifies the variable and metric, historical baseline, future window, pathway, model or ensemble, spatial method and uncertainty range. Comparing unlike baselines or seasons can manufacture a disagreement where none exists.
Downscaling translates coarse information toward a regional or local scale. Statistical methods rely on relationships between large scale predictors and local observations; dynamical methods nest a regional physical model within broader boundary conditions. Neither approach guarantees skill simply because the map has finer pixels.
Evaluation should target the location, season, variable and physical process relevant to the asset. For decision use, a range should be connected to consequence and flexibility. A short lived adjustable measure may track central conditions, while an inflexible high consequence asset may require stress testing toward a severe but credible bound. Transparent labels preserve those choices for review.
What this chapter covers
- 01
global climate models
- 02
prediction and projection
- 03
forcing scenarios
- 04
model ensembles
- 05
projection ranges
- 06
statistical and dynamical downscaling
Worked example · free
Two scenario conditioned model ranges
- 1Report the lower range as 4 to 9 percent.
- 1Report the higher range as 11 to 19 percent.
- 1Keep pathways separate.
- 1Check metric and scale suitability.
- 1Test the asset across both ranges.
Key terms
- Climate projection
- A modelled climate response conditional on a specified future forcing scenario.
- Model ensemble
- A collection of model runs used to examine response spread and common behaviour.
- Downscaling
- Translation of large scale model information to regional or local scales.
Climate Models, Scenarios and Projections FAQ
Why is a projection not a prediction?
A projection describes a possible response conditional on a scenario that may or may not occur. A prediction identifies an outcome judged likely under a forecasting setup. Engineering reports should use conditional language for scenario based results.
Does finer resolution mean a better answer?
Not automatically. Finer grids can add useful regional detail, but reliability depends on process representation, observations, boundary conditions and the downscaling method. Resolution does not remove scenario or model uncertainty.
Should model results be averaged?
An average can summarise an ensemble when its purpose is clear, but it can hide scenario separation, tail outcomes and threshold failures. Test the asset across the range before selecting a value for design.
Exam move
Create a projection label template and use it until the qualifiers become automatic: variable, metric, baseline, future period, pathway, model set, spatial method and uncertainty. Rewrite any statement that says a climate value will occur when the evidence is scenario conditioned. The goal is not cautious wording for its own sake; the condition is part of the scientific result. Redraw the projection fan from memory.
Keep the historical path single, then branch the future by pathway and model response. Explain what each source of spread represents. Scenario uncertainty concerns the forcing pathway, model uncertainty concerns the simulated response and internal variability concerns fluctuations within the system. Do not collapse these into the sentence that models are uncertain.
Repeat the two range example and resist averaging lower and higher pathways together. Describe how an inflexible, high consequence asset might use the range differently from a short lived adjustable asset. If one value is finally adopted, label whether it is a central estimate, percentile, envelope or policy allowance. Review downscaling by making a two column comparison.
Statistical methods relate local observations to large scale predictors; dynamical methods use a regional model driven by larger scale conditions. For both, ask whether the process, variable, season and location relevant to the hazard have been evaluated. Fine grids are not a substitute for skill. Take a high resolution map and identify the evidence needed before trusting one small catchment cell.
Prepare an oral response around one projected rainfall change. State the scenario and period, describe the model or ensemble, name the spatial method, identify the uncertainty that most affects the asset and explain how the option remains safe or adaptable. End with a reversal condition. This routine prevents a projection from becoming a detached percentage and turns it into a transparent input to engineering judgement.