Mapping tropical cyclone disturbance of the Great Barrier Reef.

Tropical cyclones periodically cross the Great Barrier Reef (GBR). The large waves they generate can damage coral reefs. This paper will describe the challenges of mapping cyclone disturbance across such a vast geographic region, and how cartographic visualisation can help us understand the implications of these challenges on the quality of the results.

**********

Australia's Great Barrier Reef (GBR) forms the largest coral reef system in the world, including nearly 3000 individual reefs that stretch for more than 2000 km along the Queensland coast and cover an area of over 348 000 [km.sup.2] (CRC Reef Research Centre, 2002). In recognition of its outstanding natural beauty and bio-diversity, the GBR has been protected within a Marine Park since 1975 and a World Heritage Area since 1981. Further, the GBR region supports a billion-dollar tourism industry as well as several economically important fisheries. However, a range of natural and human-caused disturbances threaten the region, including tropical cyclones, coral bleaching, crown-of-thorns starfish and reduced water quality.

Over time, tropical cyclones track throughout the entire GBR region (Figure 1). The large waves they generate break along shallow reef areas, resulting in impacts ranging from broken corals to removal of entire sections of reef structure. Severe disturbance, where long-lived, large coral colonies are destroyed or major damage is widespread (Figure 2), can require centuries or longer for reefs to fully recover. Over time, repeated widespread impacts have the potential to significantly alter coral reef community structure, which may reduce the value of particular reefs as tourist destinations or fisheries, or as habitat for other resident species. In combination with other disturbances, this can cause permanent changes. For example, in Jamaica, hard corals now struggle to compete with algae after damage from two hurricanes was exacerbated by overfishing (Hughes, 1994). Thus, to provide for "the protection, wise use, understanding and enjoyment of the Great Barrier Reef in perpetuity", the Great Barrier Reef Marine Park Authority needs to understand which reefs are likely to be affected by cyclones and how often.

[FIGURES 1-2 OMITTED]

Mapping the risk of cyclone disturbance across the GBR over time poses many challenges. While the region is vast and cyclones affect large areas at a time, damage surveys and reef vulnerability to impacts take place over short time periods and across small areas (square metres)--creating a mismatch of scales. Further, the positional and attribute uncertainty inherent in key data sets, such as cyclone paths and reef polygons, is unquantifiable but estimated to be significant.

Given these difficulties, is it really necessary to map cyclone disturbance across the entire GBR? While local-scale studies of cyclone disturbance of particular reefs or groups of reefs are valuable, it is not possible to build a regional picture from these studies alone. Such studies presently cover only a limited extent of the GBR and a short time span of the cyclone history (Table 1).

Moreover, although cyclones track through the far north (between 10-110 S) of the GBR less frequently than elsewhere, there are no sections of the GBR that are completely free of cyclones (Puotinen et al 1997) and that could be eliminated from consideration. In fact, the maximum distance of most reefs in the region to the nearest cyclone from 1969-1997 was within 30 km (Figure 3).

[FIGURE 3 OMITTED]

Given that studies indicate cyclone damage to reefs is possible up to 100 to 200 km (Done 1992, Connell et al 1997) away from a cyclone's path, every reef in the GBR has been located within striking distance of at least one cyclone over the past 30 years. Further, researchers have found that the coral larvae produced on one reef can be transported to other reefs by ocean currents, though the spatial extent of this connectivity has yet to be established. This means that the ability of coral species to recolonise a particular reef after a major cyclone disturbance could depend on the extent and severity of disturbance sustained by nearby reefs throughout the GBR.

MAPPING CYCLONE DISTURBANCE

Very few direct observations of cyclone wind and wave speeds exist in Australia, as instruments are expensive and tend to fail during the extreme conditions typically experienced. Also, because cyclone paths are virtually impossible to predict, it is difficult to know when and where to deploy instruments to measure cyclone energy. However, basic cyclone characteristics (the location of the cyclone along its path, its intensity and its speed and direction of forward motion) are typically recorded by the Australian Bureau of Meteorology in their tropical cyclone database. Meteorological models can be used to reconstruct the distribution of cyclone energy from these basic characteristics (Holland 1980). The degree to which a particular reef is affected by a cyclone, however, depends just as much on that reef's vulnerability to impact as it does on the intensity of the cyclone. Thus, a combination of hindcast cyclone energy and reef vulnerability measures was used with actual field observations of cyclone damage to build a predictive model of the spatial distribution of cyclone disturbance across the GBR. This model was then used to map cyclone disturbance for each cyclone passing near the GBR from 1969 to 2003.

Issues and Challenges

The ability to map cyclone disturbance is limited by a general lack of key data sets, as well as high levels of uncertainty in the data that is available (Table 2).

Eye Positions

Obviously it is vital to know where a particular cyclone tracked in order to map the disturbance it caused. Cyclones are tracked using a combination of satellite and radar imagery and land and ship-based observations. The timing of the observation varies depending on the proximity of the cyclone to land-based communities. The quality of the observation depends on the combination of methods used. Radar produces the most accurate observations, though many cyclones track outside radar range.

While cyclone eye positions have been recorded in Australia since about 1910, the reliability of these positions is largely unacceptable before the advent of widely available satellite imagery in 1969 (Holland 1981). Researchers at the Australian Bureau of Meteorology Research Centre estimate that errors in initial cyclone positions can exceed 400 km in any direction (Figure 4A), though most are within 100 km (Woodcock 1995). While this does not reveal the level of error remaining in the revised cyclone tracks, it suggests a high level of positional uncertainty.

[FIGURE 4 OMITTED]

Intensity

The magnitude and extent of the winds and waves generated by a particular cyclone depends largely on its intensity. The central air pressure, in hPa, of a cyclone is commonly used as a measure of intensity. As central pressure falls, the difference in air pressure between the cyclone and the external environment increases, powering the storm. Central pressure is rarely measured directly and is typically estimated using satellite imagery (Dvorak 1975). Initial central pressure measurements have also been found to be in error (Figure 4B), suggesting considerable uncertainty in cyclone intensity estimates.

Eye Width

Every cyclone contains a relatively calm region at the centre of the storm circulation called the eye. Wind speeds are near zero within the eye, while they are at their maximum around the boundary, called the eye wall. Defining the diameter of the eye is thus important to accurately reconstructing the cyclone's energy. However, cyclone eyes can be very dynamic over short time periods, sometimes with a double eye structure (Willoughby 1990). In addition, the presence of upper atmospheric cloud can make it difficult to measure eye width from satellite imagery (Figure 5). Subsequently, eye width is not always recorded in the tropical cyclone database.

[FIGURE 5 OMITTED]

Model Scale

Another problem is that the meteorological models used to estimate cyclone energy are designed to operate across broad spatial scales with a resolution of 1-10 km. The processes that control the finer scale dynamics of cyclones have not yet been adequately modelled.

Damage Observations

While basic data about the cyclones that have tracked through the GBR during the last three decades is available, observations of reef damage from these cyclones are very limited in both spatial and temporal distribution (Table 1). Broad scale field surveys were conducted only for cyclones Ivor and Joy in 1990 and Justin in 1997. Some data can be gleaned from crown-of-thorns starfish survey records (Althea in 1971) and from incidental reports (Celeste in 1996). The only long term data set available is a study of reefs near Heron Island over three decades (Connell et al 1997). Field surveys in the marine environment are expensive and time consuming. It is simply not feasible to visit every reef that could have been affected by a cyclone across 100s of square kilometres. Thus, researchers typically target surveys in areas where cyclone damage seems most likely (Table 3). However, this creates a bias in the data, which reduces the potential strength of predictive models that are developed.

Reef Vulnerability

As previously mentioned, the likelihood of cyclone damage to a reef depends just as much on the vulnerability of that reef to impact as it does on the intensity of cyclone energy. Vulnerability is controlled by a range of highly variable factors, such as the history of disturbance at a site and the predominant size and types of corals growing there. This information is poorly known for much of the GBR, most of which has never been directly surveyed (Figure 6). Vulnerability also depends on the relative exposure of sites to waves during a cyclone versus ambient conditions, which can be modelled based on the relative positions of reefs and other wave blocking obstacles such as the coastline and islands. Although the boundaries of the reefs have been mapped as vector polygons, this was done using satellite imagery and was verified by field surveys for only a small fraction of the reefs. Further, the definition of reef that is used, as well as the scale and timing of the observations, can make a big difference to the resultant reef polygons (Spalding and Grenfell, 1997).

[FIGURE 6 OMITTED]

In summary, much of the data needed to map cyclone disturbance of the GBR is missing. The data that is available contains unknown, but potentially large, uncertainties in both position and attribute. Further, there is a mismatch of spatial scales between the cyclone data from 1 km to 10 or more and the reef vulnerability data in square metres. Finally, the study area is vast but must be modelled at a high resolution (500 m pixel) in order to adequately depict the basic outlines of most of the reefs.

CREATIVE SOLUTIONS

Although the major uncertainties inherent in this project are unavoidable, the work is still worthwhile as long as some estimate of the confidence of the results, and thus how they should be appropriately used, can be made. For example, cartographic visualisation can be used to assess the implications of large uncertainties in the position of the cyclone eye to the modelling based on those positions. The cyclone database records the methods used to estimate each cyclone eye position. They involve direct, radar and satellite observations, or a combination of these. As previously mentioned, Holland (1981) and Woodcock (1995) have estimated the level of potential uncertainty associated with using each of these methods. Although this is not an actual measure of error, it does provide a 'worst case scenario' of the maximum error likely in each eye position.

To visualise this potential error, I constructed uncertainty circles around each eye position for cyclones Ivor and Justin, with the radius of each circle equal to the likely maximum positional error (Figure 7). Thus, each cyclone eye position could be located anywhere within its uncertainty circle. Interestingly, eye positions for cyclone Ivor were more uncertain when the cyclone was closer to land. This occurred because it was relatively weak at that stage, and thus was tracked less diligently. In contrast, cyclone Justin's eye positions were most uncertain when it was located far out to sea. This occurred because it was out of radar range and posed no immediate threat to the Queensland coast at that time.

[FIGURE 7 OMITTED]

To get an idea of how this uncertainty varied throughout each entire cyclone, I counted the number of times each position across the GBR was located inside an uncertainty circle, and thus could have a cyclone eye passing directly through it (Figure 8). Where the colours are more pink than red, the uncertainty of the cyclone eye positions is low, such as when cyclone Justin tracked southward down the Queensland coast. Where the colours are more red than pink, the uncertainty is high, such as where cyclone Ivor tracked down the coast or where cyclone Justin was located far out to sea. Reefs of interest, such as those for which damage data is available, can be overlaid with the uncertainty zones to assess the confidence with which the relevant cyclone energy measures should be used. For example, uncertainty in the location of cyclone Justin's eye was highest while the cyclone was located in the middle of the Coral Sea (Figure 8), which is when most of the reef sites surveyed after the cyclone were probably damaged. This suggests that reconstructed cyclone conditions for this period are likely to be in error and that the predictive model for cyclone Justin is likely to be poor. In addition, animated movies of this uncertainty could be used to show how it changes over time.

[FIGURE 8 OMITTED]

CONCLUSION

Mapping cyclone disturbance of the GBR is difficult due to a general lack of necessary data and potentially large uncertainties in both the location and attributes of the data that is available. However, it is possible to build a predictive model of cyclone disturbance by linking field observations of reef damage from cyclones to hindcast cyclone energy and reef vulnerability measures. Creative use of cartographic visualisation, as demonstrated in the previous section, is essential for estimating the implications of data uncertainty on the model results and the confidence with which they should be used.

Table 1. Cyclone damage observations available for the Great
Barrier Reef.

Year    Cyclone   # Sites   # Reefs      Region            Source

1971    Althea      27        10         Swains      Hartcher 2001
                                        Cairns -
1990    Ivor        63        33        Cooktown     Done 1992
                                        Cairns -
1990    Joy         186       33        Cooktown     Ayling 1991
1971-
1992    Various      6         1      Heron Island   Connell et al 1997
1996    Celeste      7         6      Whitsundays    DPI 1996
                                      Townsville,
1997    Justin      54        13      Whitsundays    Puotinen 1997
                                        Swains -
1997    Justin      55        35        Cooktown     Puotinen 1997

        TOTAL       398       131

Table 2. Challenges associated with mapping cyclone disturbance
of reefs.

Issues          Explanation

Cyclone Eye     Estimated positional error [+ or -] 20-200 km.
Positions       Largely unreliable before 1969.

Cyclone         Cyclone movement between eye
Paths           positions unknown and presumed
                linear.

Cyclone         Estimated intensity error [+ or -] 30 hPa.
Intensity       Largely unreliable before 1969.

Cyclone Eye     Often difficult to measure, often not
Width           recorded, very important to cyclone
                energy models.

Cyclone         Mesoscale in nature (10s of km)--don't
Models          incorporate local scale effects such as
                rain bands

Reef damage     Very limited number available. Biased
observations    towards damaged sites. Measured at a
                very local scale (meters).

Reef            Very little data exists. Highly variable
vulnerability   over short time periods and small
                distances.

Table 3. Percentage of field survey observations
found to be damaged versus the number of reefs
potentially damaged by cyclone waves.

                          % Surveyed   % GBR Sites
Year       Cyclone          Sites      Potentially
                           Damaged       Damaged

1990         Ivor             75           74
1990         Joy              63           12
1996       Celeste            63            8
1997    Justin - field
            survey            56            7
1997       Justin -
        questionnaire         60            7

REFERENCES

Ayling, T (1991). Unpublished data.

Connell, JH, Hughes, TP and Wallace, CC (1997). 'A 30-year study of coral abundance, recruitment and disturbance at several scales in space and time'. Ecological Monographs, 67(4): 461-488.

CRC Reef Research Centre (2002). The Great Barrier Reef Fact Sheet. http://www.reef.crc.org.au

Done, TJ (1992). 'Effects of tropical cyclone waves on ecological and geomorphological structures on the Great Barrier Reef'. Continental Shelf Research 12, 859-872.

Department of Primary Industries (1996). Unpublished data.

Dvorak, V F (1975). 'Tropical cyclone intensity analysis and forecasting from satellite imagery'. Monthly Weather Review, 103, 420-430.

Hartcher, M (2001). Unpublished data.

Holland, GJ (1981). 'On the quality of the Australian tropical cyclone database'. Austl. Met. Mag. 29, 169-181.

Holland, G J (1980). 'An Analytic Model of the Wind and Pressure Profiles in Hurricanes'. Monthly Weather Review 108 (August), 1212-1218.

Hughes, TP (1994). 'Catastrophes, phase shifts, and large scale degradation of a Caribbean coral reef'. Science 265: 1547-1551.

Puotinen, ML, Done, TJ and Skelly, WC, (1997): 'An atlas of tropical cyclones in the Great Barrier Reef Region, 1969-1997'. CRC Reef Research Centre, Technical Report No. 19. Townsville; CRC Reef Research Centre, 203 pp.

Puotinen, M (1997). Unpublished data.

Spalding, MD and Grenfell, AM (1997). 'New estimates of global and regional coral reef areas'. Coral Reefs, 16:225-230.

Willoughby, H E, (1990), 'Temporal changes of the primary circulation in tropical cyclones'. Journal of the Atmospheric Sciences 47(15 January), 242-264.

Woodcock, F (1995). Australian Bureau of Meteorology Research Centre. Unpublished data.

Marji Puotinen, Lecturer in GIS, School of Tropical Environmental Studies and Geography, James Cook University, Townsville, QLD, 4811. [e-mail]