Simulations of present-day (1961-1990) climate from 19 regional climate model experiments
have been compared to the observed baseline climate for Ireland. These simulations, driven
by global climate models, are obtained through the EC PRUDENCE (Prediction of Regional
scenarios and Uncertainties for Defining EuropeaN Climate change risks and Effects)
project. The ability to represent the statistics of Irish climate has been assessed, both
temporally (comparisons of meteorological year, seasonal mean and seasonal variance) and
spatially (seasonal covariation and pattern-analysis) for two key meteorological parameters,
that of temperature and precipitation.
For the average meteorological year (30-year averages of each month), mean
temperatures are found to be within 1.5Ë C of observations, except in winter, when
temperatures are overestimated by up to 2.5Ë C. Temporal variation is also not
well-represented by some models in winter. Conversely, temporal variation in precipitation is
most poorly simulated in summer. Seasonal variance is the area in which greatest inter-model
variability is shown. Ratio of observed variance to modeled variance ranges from
weak (0.5 or less) to very strong (greater than 0.7) in both seasons, and for both
parameters.
Spatially, temperature is over-estimated throughout Ireland, by up to 4.6Ë C in winter and
2.7Ë C in summer in individual grid cells from some models. Precipitation is found to be both
under and over-estimated, with grid cell biases ranging from –2.5 mm/day to 4.2
mm/day in winter and from –1.2 mm/day to 1.8 mm/day in summer. While skill at
representing the spatial precipitation pattern is found to be very strong in 16 out of 19
experiments in winter, only 2 of those experiments show the same level of skill in
summer.
Errors are identified in all individual models, both systematic and random. While using an
ensemble average overcomes some of these deficiencies, the optimal approach is to correct
systematic errors and minimize the effect of random errors using a weighted ensemble. |