1 · projection
Climate projection
Climate projections describe how the distribution of weather may develop over decades under a stated emissions pathway. Weather forecasts instead describe conditions expected on a particular day.
A date in a future series represents a weather sequence within the projected climate. It is not a forecast for that calendar date.
2 · scenario
Scenario assumption: SSP3-7.0
Every future series in this product uses , the high-emission CMIP6 pathway selected for the Klima i Norge 2025 modelling chain. The pathway specifies a possible development of society, emissions and radiative forcing through this century.
The scenario is an input to the calculation. It is not assigned a probability here, and the product does not claim that this pathway will occur. Using one scenario keeps that assumption the same across all ten model chains.
3 · model chain
Global–regional model chains
A global climate model represents the large-scale climate system. A regional climate model represents finer-scale processes over Europe and Norway. Each pairing of a global and a regional model is a model chain.
Klima i Norge regrids the regional fields from about 12.5 km resolution to Norway’s 1 km grid using nearest-neighbour selection. Bias adjustment is then applied separately at each 1 km grid cell using Norwegian reference data.
4 · ensemble
Ten-chain ensemble
Climate models represent processes differently and do not produce identical changes. The product retains results from selected global–regional pairings so users can compare their responses.
The range across the ensemble describes differences among these selected chains under the same scenario. It is neither a probability range nor a measure of total uncertainty. The ten chains do not cover every possible model response, and all use the same upstream EQM method. Their range therefore excludes uncertainty associated with the choice of bias-adjustment method.
5 · uncertainty
Sources of uncertainty
Scenario uncertainty asks which emissions pathway the world follows. Model uncertainty asks how the climate responds and how different models represent that response. Natural variability is the chaotic variation between weather years within the same climate.
This product fixes the scenario. Differences among the ten chains show part of the model uncertainty, while variation within each 30-year series represents one realization of year-to-year variability. One realization per chain does not quantify the full range of natural variability.
6 · local fit
Spatial resolution and the 1 km delivery grid
The regional climate models resolve processes at about 12.5 km. Their output is placed on a 1 km Norwegian grid and adjusted separately at each cell, but that does not create new climate-change information at 1 km scale.
The 1 km product provides a locally adjusted daily series. It does not resolve conditions for every slope, street or cloudburst. Differences between nearby cells should not be treated as independent local climate signals, especially for precipitation, which varies more sharply in space than temperature.
7 · upstream adjustment
Detrended empirical quantile mapping
The Klima i Norge production chain uses detrended empirical quantile mapping, or EQM. Month-specific transfer functions fitted over 1985–2014 adjust each model variable’s distribution against the Norwegian reference climate.
The long-term linear trend is removed before mapping and reimposed afterwards. This limits the effect of bias adjustment on the projected change signal. The source evaluation reports that EQM generally preserves seasonal change signals; it does not establish exact preservation for every variable, grid cell or statistic. The adjusted values remain model projections with the associated model uncertainty.
8 · our boundary
Framklima input boundary
Framklima does not run these climate models and does not repeat their EQM. Framklima selects the location’s 1 km series from the supplied EQM product. Subsequent dependence correction and hourly construction use the historical reference and hourly weather inputs maintained by this repository.
Dependence and chronology are handled in the separate 3DBC method. The hourly synthesis method describes how the daily values are converted to hourly weather.
9 · interpretation
Reference and future periods
We publish two 30-year climates: and , compared with the reference. Comparisons across these periods provide climate statistics, design conditions and sensitivity to the selected model chain.
Changing the reference period changes the reported difference, but not the future model run. Comparing with one reference year also includes the weather of that particular year. The 30-year periods support climate comparisons. Individual years provide weather sequences, and the range across chains describes model sensitivity.
The climate models’ historical runs end in 2014, so their 1991–2020 model baseline uses the first six SSP3-7.0 years to complete the current standard normal period. This convention applies to the model baseline. The historical weather reference is a separate observed/reanalysis-based series described in the historical methodology.