Projected changes in precipitation are computed by the Norwegian Water Resources and Energy Directorate (NVE) in collaboration with the Norwegian Centre for Climate Services (NCCS). Projected changes are presented as 30-year averages for five seasons and three scenarios (15 map layers in total): annual, winter, spring, summer, autumn, multiplied with reference period, intermediate emission scenario, RCP4.5 and high emission scenario, RCP8.5. All these 15 map layers are presented for three periods: 19712000, 20312060 and 20712100. In the netcdf file, the first frame presents 19712000, the second frame presents 20312060 and the third frame presents 20712100. For the reference period, the second and third frames therefore contain no data. For the projection periods, the first frame contains no data. The reference period displays absolute values for 19712000, whereas the two projection periods, 20312060 og 20712100, show changes relative to the reference period (e.g, the average of 20712100 minus the average of 19712000). Climate and hydrological projections are uncertain for several reasons. Uncertainties are related to future anthropogenic emissions, natural climate variations, climate models, bias correction methods and hydrological models. This is important to bear in mind in the interpretation of results from any study where the downloaded projections have been used.

No computations should be made on these 30-year averages! Both the reference period and projection periods are computed from ten models. The reference period can therefore differ from observed data. The reference period is subtracted from the projection period towards the middle and end of the century to obtain projected changes. Finally, the median of the ten models are computed (ensemble median). If you wish to do computations of climate projections to e.g. impact research, the background data must first be downloaded from http://nedlasting.nve.no/klimadata/kss/ and the method above should be followed. Performing computations over several grid cells directly on the 30-year averages will not give a correct result.

Due to the systematic biases of the GCM/RCM outputs and their mismatch in scale with impact models data requirement, a post-processing of those outputs is necessary to obtain plausible time series for use in local impact studies. The method is described in NVE report 59-2016 (Wong et al., 2016; https://publikasjoner.nve.no/rapport/2016/rapport2016_59.pdf ) In short, an empirical quantile mapping method (EQM) was used to bias-adjust Euro-CORDEX simulations to Norway by first re-gridding to 1 x 1 km and then bias-adjust against SeNorge version 1.1 as observed data for the bias adjustment procedure for each calendar-month and grid-cell. Available as WMS on https://klimagrid.nve.no/wms/GridWMSServer/.