Probabilistic climate projections have been produced by the combined use of dynamical statistical downscaling techniques for locations within the ENSEMBLES European region. This has been achieved via the use of a number of GCM/RCM climate model simulations and a stochastic weather generator. The climate scenario data (for the period 2021-50) have been sampled, both with and without individual RCM weighting, so as to produce ensemble-based probability density functions (PDFs), for a number of mean and extreme indices of temperature and precipitation. The PDFs quantify both the magnitude of change and the probability of that change occurring. All are conditional on the A1B emissions scenario.
The ENSEMBLES RT2B RCM simulations do not all extend to the end of the 21st Century. Thus the scenario period was fixed at 2021-50 so as to include the maximum number of GCM/RCM simulations. This is also the period of greater interest to many stakeholders. The choice of station locations for PDF construction was guided by a number of considerations – mainly the question of possible links to other aspects of ENSEMBLES analyses.
The choice of station locations was largely driven by the need to complement/relate to other avenues of work within the larger ENSEMBLES project. For example, work by Huth et al. (see ENSEMBLES Final Report, Section 6 – link is now available), has tested different variants of statistical downscaling methods to produce temperature projections for several station locations across the European continent.
In a further example, Déqué has produced PDFs of mean temperature and precipitation projections, for the period 2021-50, at 35 different European capital city locations. See also Sections 6.2.3 and 6.6.2 of the Final Report). A final factor in the choice of location is the availability of reliable daily observed climate data, for the period 1961-90, for training the weather generator. The map (Fig. 1, below) shows the final selection of station locations. Station details (latitude, longitude and elevation) are given in the index of PDF links – below.
|
|
A major modelling effort under ENSEMBLES RT2B has made an extensive array of GCM-driven RCM transient simulations available via the data portal. RCM output data have been downloaded, using the highest resolution (25km) runs available, so as to get the best possible match with the single point station locations chosen for the study. Not all RCM archives now in place were available when the downloading extraction exercise took place. This means that not every GCM/RCM-combination was used in the current study. For example, the HadCM3-Hirham output (from DNMI, Norway) was not included.
A total of 17 GCM-RCM combinations (encompassing seven GCMs and thirteen RCMs) were used in the ensemble PDF generation. For a full list of these, see Table 2 below.
The weather generator (WG) produces internally consistent series of "synthetic" meteorological variables including rainfall and temperature. The system produces series at a daily time resolution, using two stochastic models in series: first, for rainfall RainSim (Kilsby et al., 2007), produces an output series which is then used for a second model generating the other variables dependent on rainfall. The WG is calibrated on observed daily data (ca.1961-1990) and generates output that is statistically similar to the observed. Scenario output is produced by perturbing the WG using monthly change factors calculated from RCM present and future. The output of changes in index values (scenario minus observed), from multiple runs (100 x 30-years in this case) - is sampled with and without weighting (see below) and the samples are used to generate PDFs.
The CRU WG has been developed over a number of years. For more details and examples of previous use of the CRU weather generator, in conjunction with RCM output, towards probabilistic climate projections - see the CRANIUM work and also earlier ENSEMBLES work using PRUDENCE RCM output.
A wide range of indices has been developed and widely used in the field of climate analyses – often applied to the study of extremes. For example, the EU STARDEX Project developed software for the production of an extensive suite of climate indices. The selection of indices used here (see Table 1), falls into two sub-categories. Firstly, a number of indices relating to temperature and precipitation changes have been selected, which should be of particular interest to the user community. Secondly, this selection has been added to, to be consistent with other analyses under ENSEMBLES RT5 which test the abilities of different RCMs to simulate different aspects of observed climate (see WP5 Deliverables D5.23 and D5.31).
| Index | Description | User-friendly name |
|---|---|---|
| tav | Mean Tmean | Average temperature |
| txav | Mean Tmax | Average maximum temperature |
| tnav | Mean Tmin | Average minimum temperature |
| txhw90 | Heat Wave Duration (days) | Longest heatwave |
| txf90 | % days Tmax > 90th percentile | Number of hot days |
| tnf10 | % days Tmin < 10th percentile | Number of cold nights |
| tnf90 | % days Tmin > 90th percentile | Number of warm nights |
| pav | Mean climatological precipitation | Average seasonal rainfall |
| pxcdd | Max number of consecutive dry days | Longest dry period |
| pfl90 | Fraction of total rainfall from events > long-term 90th percentile | Heavy rainfall proportion |
| px5d | Maximum 5-consecutive-day total rainfall | Total rainfall from a wet period |
| Additional extreme indices (also relate to RT5 extremes work) | ||
| txq95 | 95th percentile of daily Tmax | Temperature on hot days |
| tnq05 | 05th percentile of daily Tmin | Temperature on cold nights |
| pq50 | 50th percentile of daily precipitation | Median daily rainfall |
| pq95 | 95th percentile of daily precipitation | Total rainfall on a wet day |
In all of the PDFs plotted (see index of links below), the (1961-90) baseline index values are shown thus allowing the projected changes, in percentage or absolute terms, to be related to the baseline.
It is common practice to use ensembles of climate model output to reduce uncertainty in climate projections. Uncertainty can be addressed further by the use of weights which reflect the abilities of individual ensemble members to replicate different aspects of observed climate. Weights have been calculated using the ERA-40 driven RCM output, with averaging across all European regions, and comparisons with observed climate – see Deliverable D3.2.2. Within this work, model weights have been calculated by different means - which have all been examined. The PDFs presented here are based on the weighting system ('Wprod') which shows the greatest diversity between GCMs.
In this set of figures, changes for the A1B emissions scenario calculated from weather generator output are plotted as Probability Density Functions (PDFs). The magnitude of the change is shown on the horizontal axis and the density (frequency) of the change occurring in any one season on the vertical axis. Each figure has two PDFs – weighted and unweighted. The links to the PDFs are split into two subsets – as indicated in Table 1.
In each seasonal panel, both of the lines represent scenario (weighted and un-weighted 17-member ensemble) changes for the 2030s. Each individual PDF is constructed using 1700 paired changes (i.e., 17 [30-year-means] x 100 weather generator runs). The changes are calculated as the difference between each of the 1700 values for 2021-2050 and the equivalent control-period value.
In order to estimate probabilities of change from these plots, it is necessary to integrate the area under the part of the curve of interest. However, these PDFs do provide a clear visual picture of the most likely change (indicated by the peak or mode of the distribution) and the uncertainty (indicated by the shape and spread of the distribution).
| Average temperature | Temperature extremes | Ave. precip. | Precip. extremes | | Temperature extremes | Precip. extremes | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| txav | tnav | tav | tav +RCM means | txhw90 | txf90 | tnf10 | tnf90 | pav | pav +RCM means | pxcdd | pfl90 | px5d | txq95 | tnq05 | pq50 | pq95 | ||||||
| Karlstad (Sweden) 59.35N, 13.47E, 50m |    |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Karlstad | ||||
| Linköping (Sweden) 58.40N, 15.53E, 100m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Linköping | ||||
| Eskdalemuir (UK) 55.32N, 3.20W, 242m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Eskdalemuir | ||||
| Valentia (Ireland) 51.94N, 10.22W, 9m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Valentia | ||||
| Ringway (UK) 53.35N, 2.27W, 75m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Ringway | ||||
| Coltishall (UK) 52.77N, 1.35E, 20 |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Coltishall | ||||
| Heathrow (UK) 51.48N, 0.45W, 25m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Heathrow | ||||
| Bamberg (Germany) 49.88N, 10.88E, 282m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Bamberg | ||||
| Praha (Czech Rep.) 50.09N, 14.42E, 191m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Praha | ||||
| Basel (Switzerland) 47.55N, 7.58E, 318m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Basel | ||||
| Bologna (Italy) 44.48N, 11.25E, 60m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Bologna | ||||
| Beograd (Serbia) 44.80N, 20.47E, 99m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Beograd | ||||
| Bucuresti (Romania) 44.52N, 26.08E, 90m |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  | Bucuresti | ||||
As an illustrative example, we consider results for average daily temperature (tav) at Heathrow - see index above:
 |
 |
The spread (horizontal range) of the PDFs (both weighted and un-weighted) for all seasons, appear to follow the spread of individual RCM scenarios which are reflected by the complementary graphics showing the individual RCM tav scenario changes (see index). The range amounts to about 2° C for all three-month seasons. The range is less for the annual PDF. In addition, long tails in the PDFs can be seen to emanate from the presence of greater individual RCM scenario mean changes. The use of weighting (black curve PDFs) makes the effects of individual RCM scenarios less clear, however, the effects of weighting are quite significant in this case, especially in the winter, autumn and annual seasonal panels. Individual RCM weights are, as yet, available only to Project partners and thus not in the public domain.
For tav at Heathrow, the indications are that temperatures, in all seasons, will rise by about 1° C by the 2030s. The greatest divergence between weighted and un-weighted PDFs is evident for the winter season. Without weighting, the increase in temperature would be about 1.5° C.
| Institute | Scenario | Driving GCM | RCM | Grid res. | ENSEMBLES Acronym | Acronym indicating GCM/RCM pairs |
|---|---|---|---|---|---|---|
| CNRM | A1B | ARPEGE | Aladin | 25km | CNRM-RM4.5 | ARP-ALADIN |
| KNMI | A1B | ECHAM5-r3 | RACMO | 25km | KNMI-RACMO2 | ECH-RACMO |
| OURANOS | A1B | CGCM3 | CRCM | 25km | OURANOSMRCC4.2.1 | CGC-CRCM |
| SMHI | A1B | BCM | RCA | 25km | SMHIRCA | BCM-RCA |
| A1B | ECHAM5-r3 | RCA | 25km | SMHIRCA | ECH-RCA | |
| A1B | HadCM3Q3 | RCA | 25km | SMHIRCA | HAD-RCA | |
| MPI | A1B | ECHAM5-r3 | REMO | 25km | MPI-M-REMO | ECH-REMO |
| METNO | A1B | BCM | HIRHAM | 25km | METNOHIRHAM | BCM-HIRHAM |
| C4I | A1B | HadCM3Q16 | RCA3 | 25km | C4IRCA3 | HAD-RCA3 |
| UCLM | A1B | HadCM3Q0 | PROMES | 25km | UCLM-PROMES | HAD-PROMES |
| ETHZ | A1B | HadCM3Q0 | CLM | 25km | ETHZ-CLM | HAD-CLM |
| HC | A1B | HadCM3Q0 | HadRM3Q0 | 25km | METO-HC_HadRM3Q0 | HAD-HAD |
| A1B | HadCM3Q3 | HadRM3Q3 (low sensitivity) | 25km | METO-HC_HadRM3Q3 | HAD-HADLO | |
| A1B | HadCM3Q16 | HadRM3Q16 (high sensitivity) | 25km | METO-HC_HadRM3Q16 | HAD-HADHI | |
| DMI | A1B | ARPEGE | HIRHAM | 25km | DMI-HIRHAM5 | ARP-HIRHAM5 |
| A1B | ECHAM5-r3 | DMI-HIRHAM5 | 25km | DMI-HIRHAM5 | ECH-HIRHAM5 | |
| ICTP | A1B | ECHAM5-r3 | RegCM | 25km | ICTP-REGCM3 | ECH-REGCM |
Kilsby, C.G., Jones, P.D., Burton, A., Ford, A.C., Fowler, H.J., Harpham, C., James, P., Smith, A. and Wilby, R.L. 2007: A daily weather generator for use in climate change studies. Environmental Modelling & Software, 22, 12, 1705-1719.