Diagnostic metrics for evaluation of annual and diurnal cycles

• Published in: Climate dynamics 2011-09, Vol.37 (5-6), p.941-955
• Main Authors: Wang, Bin, Kim, Hyung-Jin, Kikuchi, Kazuyoshi, Kitoh, Akio
• Format: Article
• Language: English
• Subjects: Annual variations; Atmospheric circulation; Climate science; Climatology; Climatology. Bioclimatology. Climate change; Coastal zone; Correlation coefficient; Diurnal variations; Earth and Environmental Science; Earth Sciences; Earth, ocean, space; Exact sciences and technology; External geophysics; Geophysics/Geodesy; Hydrologic cycle; Marine; Meteorology; Monsoons; Oceanography; Precipitation; Weather analysis and prediction; Weather forecasting; Wind
• ISSN: 0930-7575; 1432-0894
• DOI: 10.1007/s00382-010-0877-0

Two sets of diagnostic metrics are proposed for evaluation of global models’ simulation of annual and diurnal cycles of precipitation. The metrics for the annual variation include the annual mean, the solstice and equinoctial asymmetric modes of the annual cycle (AC), and the global monsoon precipitation domain and intensity. The metrics for the diurnal variation include the diurnal range, the land–sea contrast and transition modes of the diurnal cycle (DC), and the diurnal peak propagation in coastal regions. The proposed modes for the AC and DC represent faithfully the first two leading empirical orthogonal functions and explain, respectively, 82% of the total annual variance and 87% of the total diurnal variance over the globe between 45°S and 45°N. The simulated AC and DC by the 20-km-mesh MRI/JMA atmospheric general circulation model (AGCM) are in a wide-ranging agreement with observations; the model considerably outperforms any individual AMIP II GCMs and has comparable performance to 12-AMIP II model ensemble simulation measured by Pearson’s pattern correlation coefficient. Comparison of four versions of the MRI/JMA AGCM with increasing resolution (180, 120, 60, and 20 km) reveals that the 20-km version reproduces the most realistic annual and diurnal cycles. However, the improved performance is not a linear function of the resolution. Marked improvement of the simulated DC (AC) occurs at the resolution of 60 km (20 km). The results suggest that better represented parameterizations that are adequately tuned to increased resolutions may improve models’ simulation on the forced responses. The common deficiency in representing the monsoon domains suggests the models having difficulty in replicating annual march of the Subtropical Highs that is largely driven by prominent east-west land–ocean thermal contrast. Note that the 20-km model reproduces realistic diurnal cycle, but fails to capture realistic Madden-Julian Oscillation.