Impact of Higher Spatial Resolution on Precipitation Properties Over Australia

Coarse resolution datasets often poorly capture precipitation properties. It is commonly expected that skill at the local level should increase by refining spatial resolution. Here, we examine the influence of spatial resolution on the accuracy of means and extremes in precipitation by comparing high-resolution dynamically downscaled data against the driving lower-resolution reanalysis. We show that the higher-resolution data are less accurate than lower-resolution data for both mean and extreme precipitation. The quantitative deterioration (increase in the domain averaged relative error) from coarse to high resolution varies typically between 1%-8% and 5%-30% for mean and extreme precipitation, respectively. We also find that the finer-scale variability resolved only by the higher-resolution system and successfully predicted by that system is of negligible magnitude compared to the overall error (less than 1%). We conclude that finer-scale resolution by itself does not necessarily bring a meaningful improvement in local simulation accuracy.