The NARCliM Project: Model Agreement and Significance of Climate Projections.

NARCliM (New South Wales and Australian Capital Territory Regional Climate Modelling project) is a climate downscaling project for Australia and the surrounding regions. Present and future climate simulations are performed using a 1-way nested dynamical down- scaling approach and span 2 domains. We focus on the inner 10 km domain that extends across south-east Australia. Three regional climate models (RCMs) based on the Weather Research and Forecasting System (WRF) version 3.3 dynamically downscale 4 global climate model (GCM) sim- ulations to finer resolutions. This project complements and improves on already available GCM projections for the region. Our simulations cover 3 epochs: present (1990−2009), near future (2020−2039), and far future (2060−2079). Here, we focus on the mean surface air temperature and precipitation. The RCMs are better able to capture spatial patterns of temperature and precipita- tion and improve the temperature root mean square error (RMSE) compared to the GCMs, at least for the inner domain. The RCMs tend to be biased cold compared to observations and are wetter than the GCMs during warm seasons. The downscaled RCM projections exhibit a weaker warm- ing over land compared to the GCMs. The RCMs project no significant precipitation changes in the far future over most areas. However, Victoria is expected to see significant springtime drying of 15 mm mo−1, which is considerably higher than previous GCM results. This drying is associated with a larger strengthening of the subtropical ridge than modelled previously by the GCMs. In addition, the RCMs project significant precipitation changes in contradicting directions for some inland areas during winter.

Fig. 8. Ensemble mean seasonal mean precipitation change (mm mo−1) from present day (1990−2009) to far future (2060−2079) for the GCMs, the RCMs on domain 01, and the RCMs on domain 02. Stippled (significant agreeing) areas indicate that half or more models show a statistically significant change, with 80% of more significant models changing in the same direction. White (significant disagreeing) areas indicate that half or more models show a statistically significant change, with less than 80% of significant models changing in the same direction