Modelling the risk of cyclone wave damage to coral reefs using GIS: a case study of the Great Barrier Reef, 1969-2003

Tropical cyclones (hurricanes and typhoons) produce high winds that Call generate waves capable of damaging coral reefs. As cyclones frequently pass through northeast Australia's Great Barrier Reef (GBR), it is important to understand how the spatial distribution of reef damage changes over time. However, direct measurements of wave damage, or even wave heights or wind speeds, are rare within the GBR. An important factor in estimating whether cyclone damage was possible is the magnitude and duration of high-energy wind and waves. Thus, before the spatio-temporal dynamics of past cyclone damage call be modelled, it is necessary to reconstruct the spread, intensity, and duration of high-energy conditions during individual cyclones. This was done every hour along the track taken by each of 85 cyclones that passed near the GBR from 1969 to 2003, by implementing a cyclone wind hindcasting model directly within a raster GIS using cyclone data available from the Australian Bureau of Meteorology. Three measures of cyclone energy (maximum wind speed-MAX, duration of gales-GALES, and continuous duration of gales-CGALES) were derived from these data. For three cyclones, where field data documenting actual reef damage from cyclone-generated waves were available, the predictive ability of each measure was assessed statistically. All three performed better in predicting reef damage at sites Surveyed along the high-energy reef front than those SUrveyed along the more protected reef back. MAX performed best for cyclone Joy (r(2)=0.5), while CGALES performed best for cyclones Ivor (r(2)=0.23) and Justin (r(2)=0.48). Using thresholds for MAX and GALES obtained via comparison with field data of damage, it was possible to produce a preliminary prediction of the risk of wave damage across the GBR from each of the 85 cyclones. The results suggest that while up to two-thirds of the GBR was at risk from some damage for 30-50% of the time series (similar to 18 out of 35 years), only scattered areas of the region were at risk more frequently than that.