BAMOS Vol 30 No. 3 2017 | Page 18

18 BAMOS Sept 2017 Figure 2. Evaluation of forecasts from the 12-km reanalysis suite (BARRA-R, shown in green), the 1.5-km downscaled reanalysis over the Sydney subdomain (BARRA-SY, 37.88 to 28.05 °S and 147.12 to 153.63 °E, shown in red) and ERA- Interim (shown in blue) against observed screen-level temperature for the period February to March 2010. 6-hour forecasts of temperature are evaluated at 00 and 12 UTC. Left: Root Mean Square Error (RMSE), right: bias. Figure 3. Comparison of model daily rainfall from BARRA-R (left) against gauge-interpolated daily rainfall (AWAP) (right) for 24 August 2010. Model rainfall less than 0.1 mm/day is masked. the NASA-Global Modelling and Assimilation Office. They found that the most extreme changes in wind speed are consistently underestimated in MERRA, with the largest underestimations associated with high altitude stations. The underestimation of wind speeds at high elevation is a result of overly smooth topography, MERRA has a resolution of about 50 km. Preliminary case studies The two case studies below highlight examples of potential benefits from the reanalysis data. The first case derives benefit from the extensive, consistent spatial coverage of the reanalysis data. The second case study derives benefit from the potentially more accurate wind data available with fine-scale coverage in the reanalysis. This is an exciting area of further research. Bushfire in Tasmania – The Giblin River fire Details of the Giblin River Fire event are documented in Marsden-Smedley (2014). The fire was started by lightning in the afternoon of 3 January 2013. Due to the remoteness of the location, no attempts were made suppress the fire. The fire then spread over the following weeks and continued to burn slowly until it was contained almost three weeks later (22 January 2013), having burned an area of approximately 450 km 2 (Figure 4). The orography in this region strongly affects wind conditions during fire events and the reanalysis can provide information that is not readily available on the basis of surface observations alone (Figure 5). Data is shown from the Giblin River fire where a lack of any