University College London
1770 files

Graphs of post-processed time series of climate projections for the UK

Version 3 2023-02-27, 13:17
Version 2 2023-02-06, 10:46
Version 1 2023-01-30, 11:45
posted on 2023-02-27, 13:17 authored by Richard ChandlerRichard Chandler, Chris BrierleyChris Brierley, Clair BarnesClair Barnes, Ed LowtherEd Lowther, Yagmur OzdemirYagmur Ozdemir

Visualisations of post-processed projections of annual and seasonal mean temperatures and precipitation totals to the year 2080, derived from the UKCP18 and EuroCORDEX regional ensembles under the RCP8.5 emissions scenario and averaged over different regions of the UK land surface.The post-processing uses Bayesian techniques that acknowledge the presence of shared and structured discrepancies between the properties of the real climate system and those of the contributing climate models. The methodology is based on the framework of: Chandler R.E. (2013). Exploiting strength, discounting weakness: combining information from multiple climate simulators. Phil Trans R Soc A 371: 20120388, doi: 10.1098/rsta.2012.0388.There are five components to the file names for the visualisations, as follows:

1. Plot type, which is either of the following:

  • Mimic: this plot shows a time series of observed climate, together with a collection of climate model simulations. Superposed on these is an estimate of the trend in the real climate system, together with a 95% uncertainty envelope for the quantity of interest in each year. The trend and uncertainty envelope are derived from the best estimates obtained from a mimic, which is a statistical characterisation of the structure in both the observations and climate model simulations and the relationships between them (see Chandler, 2013). 
  • PostPred: these plots attempt to provide a more accurate assessment of uncertainty in future climate, by allowing for sampling variation in the calibration of the mimics instead of just taking the best estimates. This is done by sampling from an approximate posterior predictive distribution for quantities of interest in a Bayesian framework. There are two versions of these plots. The first is a static plot which is a direct analogue of the Mimic plots but incorporating the effect of sampling variation. 
  • PPAnim: this is the second version - an animated plot which superimposes a range of potential realisations of the future climate, in such a way that the proportion of time spent in a particular state in the animation is proportional to the posterior probability of being in that state. The animation methodology is that of Bowman (2019). Note, however, that the Bayesian computations used to produce these plots can sometimes be numerically unstable: although every attempt has been made to stabilise them where possible, there are still some instances where the results seem unrealistic. In these cases, the Mimic plots are probably more reliable.

2. Ensemble identifier, which indicates the climate model outputs used to produce the plot and is one of the following:

  • cmip5: outputs from the 10 global climate models used to drive the EuroCORDEX regional ensemble (see below). These were produced as part of the Coupled Model Intercomparison Project Phase 5 (CMIP5) experiment (Taylor et al., 2012), and have coarse spatial resolutions ranging from 0.75 degrees to 2.8 degrees in both latitude and longitude. 
  • eurocordex: outputs from 64 members of the EuroCORDEX regional climate model ensemble (Jacob et al. 2014), each combining one of the CMIP5 global models (see above) with one of thirteen different regional climate models. These outputs are at a finer resolution of 0.11 degrees in both latitude and longitude. 
  • ukcp18-12km: these are regional outputs from the 2018 UK Climate Projections, produced by coupling 12 variants of a regional climate model with 12 variants of a corresponding global model (Murphy et al. 2019). The outputs are at a spatial resolution of 12kmx12km, which is comparable with the resolution of the EuroCORDEX ensemble (see above). 
  • ukcp18-60km: these are the coarse-resolution (60kmx60km) global model outputs used to drive the regional model variants in the ukcp18-12km ensemble. The spatial resolution is roughly comparable with that of most of the CMIP5 models (see above).

3. Region: this is the geographical region of the UK represented in the plot. For a given region, the original climate model outputs have first been regridded to the same resolution as the UKCP18 12km ensemble (see above) and then averaged over all grid cells within the region. The available regions are east-midlands, east-of-england, east-scotland, london, north-east-england, north-scotland, north-west-england, northern-ireland, south-east-england, south-west-england, wales, west-midlands, west-scotland and yorkshire-and-humber. In addition, uk is used for plots considering average values over the entire UK land area. Please note that it has not been possible to produce plots for the Channel Islands, because these are not represented in the HadUK-Grid dataset used as the source of observed climate information in this work.

4. Time scale: this is one of DJF, MAM, JJA, SON or annual and represents the time period or season considered in each plot. In each case, the plot represents averages (for temperature) or totals (for precipitation) over all days within the respective period. The annual values are over meteorological years starting on 1st December and ending on 30th November (meteorological year Y starts on 1st December in year Y-1 so that, for example, meteorological year 2015 starts on 1st December 2014); the others are respectively for winter (December, January, February), spring, summer and autumn. 

5. Quantity considered: this is either tas (surface temperature, in degrees Celsius) or pr (precipitation, in centimetres).

Note: the UK-wide charts have been updated since they were first published, to correct an error with the processing of date information.


Strategic Priorities Fund P108457: CR20/3 Enabling the use and producing improved understanding of EuroCORDEX data over the UK