Simulation of New Zealand climate is examined through the use of a reg
ional climate model nested within the output of the Commonwealth Scien
tific and Industrial Research Organisation nine-level general circulat
ion model (GCM). R21 resolution GCM output is used to drive a regional
model run at 125 km grid spacing over the Australasian region. The 12
5 km run is used in turn to drive a simulation at 50 km resolution ove
r New Zealand. Simulations with a full seasonal cycle are performed fo
r 10 model years. The focus is on the quality of the simulation of pre
sent-day climate, but results of a doubled-CO2 run are discussed brief
ly. Spatial patterns of mean simulated precipitation and surface tempe
ratures improve markedly as horizontal resolution is increased, throug
h the better resolution of the country's orography. However, increased
horizontal resolution leads to a positive bias in precipitation. At 5
0 km resolution, simulated frequency distributions of daily maximum/mi
nimum temperatures are statistically similar to those of observations
at many stations, while frequency distributions of daily precipitation
appear to be statistically different to those of observations at most
stations. Modeled daily precipitation variability at 125 km resolutio
n is considerably less than observed, but is comparable to, or exceeds
, observed variability at 50 km resolution. The sensitivity of the sim
ulated climate to changes in the specification of the land surface is
discussed briefly. Spatial patterns of the frequency of extreme temper
atures and precipitation are generally well modeled. Under a doubling
of CO2, the frequency of precipitation extremes changes only slightly
at most locations, while air frosts become virtually unknown except at
high-elevation sites.