Comparison of BARRA and ERA5 in replicating mean and extreme precipitation over Australia

Abstract

Reanalysis datasets are critical in climate research and weather analysis, offering consistent historical weather and climate data crucial for understanding atmospheric phenomena and validating climate models. However, biases exist in reanalysis datasets that would affect their applications under circumstances. This study evaluates BARRA, which is a high-resolution reanalysis for the Australian region, and ERA5 in simulating mean precipitation and six selected precipitation extremes for their climatology, temporal correlation, coefficient of variation, and trend. Both datasets reproduce daily timescale probability density distributions and spatial patterns of mean precipitation well with minor biases. ERA5 shows stronger temporal correlations, superior interannual precipitation accuracy, and lower biases in coefficient of variation compared to BARRA, especially in Northern Australia. However, both models exhibit substantial biases in trend, underestimating increasing trends in Northern Australia. ERA5 underestimates dry days and heavy rainfall, while BARRA tends to overestimate these extremes. Temporal correlations for extreme precipitation indices are weaker compared to mean annual precipitation. Notable differences exist in variability biases, with BARRA showing larger biases, especially for heavy precipitation in inland regions and Northern Australia. While both datasets replicate the main trends, biases persist. Overall, the evaluation results support application of both datasets for climatology analyses, but caution is advised for variability and trend analyses, particularly for specific extremes.