Evaluation of ACCESS-TC Precipitation Forecasts using CRA

The Australian Community Climate and Earth-System Simulator - Tropical Cyclone version (ACCESS-TC) has been operational since 2011. How does it perform on tropical cyclone (TC) rainfall forecasts? Does it have system bias on rainfall location error? What is the error contribution from different components, such as volume, pattern and location? To answer these three questions, we use Contiguous Rain Areas (CRA) methods to verify ACCESS-TC rainfall forecasts, against TRMM 3B42 satellites estimates.

Under the CRA methodology, a rainfall forecast will be shifted to a location where the forecast has the best match with satellite estimates. By comparing the statistics between the original and shifted forecasts, we can calculate the displacement error of the original model forecast. The total precipitation error can also be decomposed into four components: displacement, rotation, volume and pattern. For the limited cases considered, pattern error and displacement error were found to be major error contributors.

This validation study can identify ACCESS-TC rainfall forecasting biases and limitations, and guide future efforts towards model development and improvement of TC rainfall forecasts.