How to read a climate projection
Everything on this platform is a projection: a statement about the statistics of a future climate under an assumed emissions pathway. Reading one correctly means knowing what it is not.
Four different questions
These sound similar and are answered by completely different science. Conflating them is the most common error in public discussion of climate.
- What will the temperature in Rawalpindi be tomorrow?
- Weather. An initial-value problem — you need to know the current state of the atmosphere very precisely. Skill collapses after about ten days.
- What is Rawalpindi's typical temperature?
- Climate. The statistics of weather over decades — a distribution, not an event.
- How could Rawalpindi's climate change by 2040–2059?
- Climate projection. A boundary-value problem — you need to know the forcing, not today's weather. Useful for decades precisely because it ignores individual days.
- What happens under lower versus higher emissions?
- Scenario analysis. The comparison itself is the finding; no single pathway is the prediction.
What a scenario actually is
A Shared Socio-economic Pathway pairs a story about how the world develops — population, technology, inequality, energy — with the radiative forcing that story produces by 2100. The number after the dash is that forcing in watts per square metre. SSP2-4.5 is the “middle of the road” storyline reaching 4.5 W/m².
SSP1-1.9
Sustainability — taking the green road1.0–1.8 °C by 2100The most ambitious pathway in CMIP6. Net-zero CO₂ around 2050, deep and immediate cuts across all sectors. Roughly consistent with holding warming near 1.5 °C.
SSP1-2.6
Sustainability — taking the green road1.3–2.4 °C by 2100Strong, sustained mitigation with net-zero CO₂ in the second half of the century. Broadly the successor to RCP2.6 and roughly aligned with the upper end of the Paris goals.
SSP2-4.5
Middle of the road2.1–3.5 °C by 2100Development follows historical patterns; emissions stay near current levels to mid-century then decline. Often treated as the closest analogue to stated national policies.
SSP3-7.0
Regional rivalry — a rocky road2.8–4.6 °C by 2100Resurgent nationalism, weak international cooperation and slow technological change. Emissions roughly double by 2100. High aerosol and land-use forcing.
SSP5-8.5
Fossil-fuelled development — taking the highway3.3–5.7 °C by 2100Rapid, energy-intensive growth built on abundant fossil fuels. The high end of the CMIP6 range; now widely regarded as a low-likelihood upper bound rather than business as usual.
None of these is a prediction, and they are not equally likely. The IPCC does not assign probabilities to them. SSP5-8.5 in particular was designed as a high-end sensitivity case and is now widely regarded as implausible as a business-as-usual trajectory — it is most useful as an upper bound for stress-testing, not as an expectation.
RCP and SSP are not the same thing
CMIP5 — the model generation behind IPCC AR5 — used Representative Concentration Pathways, which specify a forcing trajectory and nothing else. CMIP6 replaced them with SSPs, which add a socio-economic storyline that also determines aerosols and land use. Two scenarios with the same forcing number can therefore produce noticeably different regional climates, especially over South Asia where aerosol loading strongly affects the monsoon.
| CMIP5 (RCP) | W/m² | Nearest CMIP6 (SSP) |
|---|---|---|
RCP2.6 Peak-and-decline pathway with substantial negative emissions. | 2.6 | SSP1-2.6 |
RCP4.5 Stabilisation without overshoot by 2100. | 4.5 | SSP2-4.5 |
RCP6.0 Stabilisation reached after 2100. | 6 | no close counterpart |
RCP8.5 Rising radiative forcing throughout the century. | 8.5 | SSP5-8.5 |
The right-hand column matches forcing levels only. It is a reading aid for older literature, not a licence to treat RCP4.5 results as SSP2-4.5 results.
Why the models disagree
Thirty modelling groups have built independent representations of the same physical system. They agree on the fundamentals and differ on the hard parts — clouds above all, then aerosols, convection and land surface processes. The result is a genuine range in how much the planet warms for a given amount of CO₂.
The usual summary of that is equilibrium climate sensitivity: the long-run warming from doubling CO₂. Across the models used here it runs from 1.8 °C to 5.6 °C. A single model is not wrong for sitting at either end — it is one internally consistent answer to a question that is not yet settled.
This is why the multi-model ensemble median is the default everywhere on this platform, and why the 10th–90th percentile range is shown alongside it rather than tucked into a footnote. Choosing one model and reporting its number as the projection is the fastest way to mislead with entirely real data.
Why twenty-year windows
Individual years swing far more than the underlying trend. Averaging over twenty years suppresses that natural variability enough for the forced signal to show through, while still being short enough to matter for planning. Every projection here is one of these windows, and every change is measured against the same 1995–2014 baseline.
Hazard is not risk
A climate model produces climate variables. It does not produce flood damages, crop failures or deaths. Getting from one to the other needs hydrology, exposure and vulnerability — separate models with their own assumptions and their own, usually larger, uncertainties.
This platform stops at the climate variable and names the mechanism that would carry it further. It does not cross the line into impact claims it cannot support.
Heat and human tolerance
- ↓Mean and maximum temperature rise
- ↓More days cross physiological thresholds (35 °C, 40 °C)
- ↓Overnight minima stay high, so the body cannot shed accumulated heat
- ↓Heat stress concentrates in outdoor workers, infants and the elderly
Stops here without: An exposure and vulnerability model — who is outdoors, who has cooling, who has power — to translate this into health outcomes.
Humid heat and the survivability limit
- ↓Warming over a warm, moist surface raises wet-bulb temperature
- ↓Above roughly 31 °C wet-bulb, sweating stops cooling the body effectively
- ↓Above 35 °C wet-bulb, healthy people cannot survive extended exposure even at rest in shade
- ↓The lower Indus valley is one of the few regions worldwide already near these values
Stops here without: Station-level humidity validation; gridded wet-bulb estimates carry larger errors than dry-bulb temperature.
Rainfall intensity and flood hazard
- ↓Warmer air holds roughly 7% more moisture per degree of warming
- ↓When it rains, more water is available to fall in a short window
- ↓One-day and five-day maxima increase faster than annual totals
- ↓Drainage and river systems designed for historical intensities are exceeded more often
Stops here without: A hydrological model with catchment routing, plus channel capacity and land-use data, to turn rainfall into flood extent or depth.
A more erratic monsoon
- ↓Total monsoon rainfall changes only modestly in most models
- ↓But it arrives in fewer, heavier bursts separated by longer dry spells
- ↓Both flood and drought exposure can rise at the same time
- ↓Annual rainfall totals hide this entirely
Stops here without: Sub-daily and event-scale analysis; the 20-year climatologies here average over exactly the variability that matters.
The indicators
Mean temperature is the headline, but it is rarely the number that matters for a decision. Threshold counts, extremes and dry-spell lengths carry far more of the consequence.
Temperature
Mean and extreme air temperature.
Heat & heat stress
Threshold exceedances, warm spells and humid-heat stress.
Precipitation
How much rain falls, and when.
Heavy rainfall
Rainfall intensity and extremes — inputs to flood hazard, not flood risk itself.
Dryness & drought
Dry-spell length and moisture balance.
Snow & cold
Snowpack, frost and ice — the Indus headwaters story.
Agriculture
Growing season and evaporative demand.
Energy demand
Degree-day proxies for cooling and heating load.