MODELLING SURFACE AIRFLOW FOR FOREST AND RURAL FIRE DANGER RATING: COMPARISONS BETWEEN ACTUAL OBSERVATIONS AND PREDICTED VALUES

The performance of three airflow models (Flowstar, MSFD-PC, and MS-Micro/3) was compared as a means of providing wind speed and wind direction data for use in the assessment of forest and rural fire danger. At Cowal Peninsula, Scotland, and the Manawatu Gorge, New Zealand, six simulations were performed with each model, using data for incident winds from the westerly quarter and incorporating effects of complex terrain. Model performance was similar for both study areas. The mean wind speed prediction error for the best-performing model (MS-Micro/3) was approximately 30%. Without simulation, prediction errors for average wind speed were 67 % at Cowal Peninsula and 50 % at Manawatu Gorge. Root mean square errors for wind direction prediction by models averaged 86 ° at Cowal Peninsula and 26 ° at Manawatu Gorge. With the exception of MS-Micro/3 output for the Manawatu Gorge area, wind direction simulations at both sites produced unacceptable results. Three independent data sets were used to examine the influence of wind speed prediction errors on fire danger rating assessment. Error magnitudes associated with use of airflow modelling of single point source data altered the number of days allocated to Very High and Extreme fire danger classes, differences being much greater for areas with a severe fire climate. The Jackson-Hunt linear microscale models tested can provide wind speed information which will improve the accuracy of fire danger rating