2009 Volume 87 Issue 3 Pages 473-495
Reproducibility of land-surface air temperatures and land precipitation in the twentieth century by an atmospheric general circulation model (AGCM) MJ98 was investigated focusing on long term trends and year-to-year variability. The MJ98 model was jointly developed by the Meteorological Research Institute (MRI) and the Japan Meteorological Agency (JMA), and has a 270-km horizontal grid spacing (T42) with 30 vertical levels. Forcing the MJ98 model with observed historical sea surface temperatures (SST) and observed historical CO2 concentrations, six-member ensemble integrations were conducted for 130 years from 1872 to 2001. Simulated land-surface air temperature and land precipitation were validated against observational data of the Climate Research Unit (CRU) from 1872 to 2001 and from 1951 to 1997, respectively.
The model reproduces the observed positive trend of annual mean global average land-surface air temperature as well as decadal variability and year-to-year variability. The model simulates the observed positive trend of global average land-surface air temperature for all four seasons and the annual mean, though the magnitude is underestimated. The seasonality of the simulated trend is weak compared with that of the observation. At each grid point, the model generally reproduces positive trends of annual mean temperature over land. However, the simulated trends are underestimated especially over the middle and higher latitudes of the Northern Hemisphere, which can be partly attributed to the inability of model to simulate the increasing boreal wintertime trend of the North Atlantic Oscillation (NAO) and the Arctic Oscillation (AO) over the last few decades. The model's ability to reproduce the year-to-year variability of the annual mean temperature is relatively higher in coastal regions than in inland regions.
In the case of annual mean land precipitation, the model simulates the observed negative trend for the global average, its negative trend for the Northern Hemisphere, and its positive trend for the Southern Hemisphere, although these observed trends are not statistically significant. The model fails to reproduce year-to-year variability. The model generally reproduces the distribution of trend of global annual mean land precipitation, but large discrepancies between observation and simulation are found over Asia, Australia and southern Africa.