Section 7.5: Comparing Observed and GCM Climates
131
model-generated means of A and C for eaeh month are within the extreme
observed range ealculated over the last 110 years. The models perform best
in late summer and autumn (between August and Oetober), when simulated
A and C types for both MPI and UKHI are elose to the observed mean values.
As for directional types, Wand S are the most important (with annual
means of 75.4 and 59.0 respeetively). UKHI has an annual W total elose
to the real-world but it underestimates in January and overestimates in the
autumn months, produeing a false annual eyele. The MPI annual eyele has
the right shape but W days are generally too low in all but the late spring
months (April and May). Both models simulate the weak annual eyele in S
days very well, and UKHI is remarkably accurate in simulating the eorreet
monthly means. MPI is again eonsistently low. The simulated monthly
means fall outside of the observed range in two winter months (January and
Deeember) and all three summer months.
The other direetional types (N, E and NW) together oeeur on average
about 64 days of the year (18.8, 18.7 and 26.8 respeetively). Annual mean N
days are too high in MPI but generally weH simulated by UKHI. MPI also
overestimates E types whereas UKHI overestimates in late winter and underestimates in autumn (September-November). Both models tend to produee
too few NW days (especiaHy MPI) over the year as a whole though they both
produee a similar weak annual eyele.
7.5.2 Temperature and Precipitation
Figure 7.6 eompares the annual eyeles of temperature and precipitation representing observed average eonditions over the U.K. with those simulated by
eaeh ofthe GCMs. The observed temperature data are the "Central England"
series eompiled by Manley (1974; and updated by Parker et al. , 1992). The
preeipitation reeord is the "England and Wales" series deseribed in Wigley et
al. (1984; updated in Gregory et al., 1991). Both of these reeords are made
up of a number of individual site reeords and are therefore spatially representative and appropriate for eomparison with the model-generated data.
Mean annual temperature in the real-world data is 9.3°C (1881-1990).
The equivalent MPI value is very similar at 9.9°C but the UKHI value is
only 6.9°. The MPI model, in fact, simulates the annual eyele very weH,
the only diserepaney being an overestimate of temperatures in late autumn,
and early winter. Though the shape of the annual eyele is reasonable, the
UKHI model is much too cold (2 - 3°C lower) in all but the three months
October-Deeember, whieh are weH simulated. In eontrast, UKHI simulates
the monthly totals of preeipitation remarkably weH. Eaeh is within the observed range. The MPI monthly totals are accurate in autumn and early
winter (September-Deeember) but are mueh too low in the spring and dramaticaHy low in the summer. In fact MPI produces hardly any rain at all
during the summer months of June-August whereas the observed values for
summer as a whole are over 200 mm (eompared with below 40 mm produeed
by MPI).
131
model-generated means of A and C for eaeh month are within the extreme
observed range ealculated over the last 110 years. The models perform best
in late summer and autumn (between August and Oetober), when simulated
A and C types for both MPI and UKHI are elose to the observed mean values.
As for directional types, Wand S are the most important (with annual
means of 75.4 and 59.0 respeetively). UKHI has an annual W total elose
to the real-world but it underestimates in January and overestimates in the
autumn months, produeing a false annual eyele. The MPI annual eyele has
the right shape but W days are generally too low in all but the late spring
months (April and May). Both models simulate the weak annual eyele in S
days very well, and UKHI is remarkably accurate in simulating the eorreet
monthly means. MPI is again eonsistently low. The simulated monthly
means fall outside of the observed range in two winter months (January and
Deeember) and all three summer months.
The other direetional types (N, E and NW) together oeeur on average
about 64 days of the year (18.8, 18.7 and 26.8 respeetively). Annual mean N
days are too high in MPI but generally weH simulated by UKHI. MPI also
overestimates E types whereas UKHI overestimates in late winter and underestimates in autumn (September-November). Both models tend to produee
too few NW days (especiaHy MPI) over the year as a whole though they both
produee a similar weak annual eyele.
7.5.2 Temperature and Precipitation
Figure 7.6 eompares the annual eyeles of temperature and precipitation representing observed average eonditions over the U.K. with those simulated by
eaeh ofthe GCMs. The observed temperature data are the "Central England"
series eompiled by Manley (1974; and updated by Parker et al. , 1992). The
preeipitation reeord is the "England and Wales" series deseribed in Wigley et
al. (1984; updated in Gregory et al., 1991). Both of these reeords are made
up of a number of individual site reeords and are therefore spatially representative and appropriate for eomparison with the model-generated data.
Mean annual temperature in the real-world data is 9.3°C (1881-1990).
The equivalent MPI value is very similar at 9.9°C but the UKHI value is
only 6.9°. The MPI model, in fact, simulates the annual eyele very weH,
the only diserepaney being an overestimate of temperatures in late autumn,
and early winter. Though the shape of the annual eyele is reasonable, the
UKHI model is much too cold (2 - 3°C lower) in all but the three months
October-Deeember, whieh are weH simulated. In eontrast, UKHI simulates
the monthly totals of preeipitation remarkably weH. Eaeh is within the observed range. The MPI monthly totals are accurate in autumn and early
winter (September-Deeember) but are mueh too low in the spring and dramaticaHy low in the summer. In fact MPI produces hardly any rain at all
during the summer months of June-August whereas the observed values for
summer as a whole are over 200 mm (eompared with below 40 mm produeed
by MPI).