Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Testing the grass‐fire cycle: alien grass invasion in the tropical savannas of northern Australia

Testing the grass‐fire cycle: alien grass invasion in the tropical savannas of northern Australia INTRODUCTION Plant invasions are regarded as one of the most serious threats to global biodiversity and ecosystem functioning ( Heywood, 1989 ; D’Antonio & Vitousek, 1992 ; Vitousek ., 1996 ; Williams & Baruch, 2000 ). Grasses have been identified as a particularly invasive set of plant species ( Heywood, 1989 ) with substantial areas of invasion occurring on almost every continent ( D’Antonio & Vitousek, 1992 ). One of the most significant ecological changes caused by invading alien grasses is the alteration of fire regimes ( D’Antonio & Vitousek, 1992 ). There have been numerous documented cases where alien grass invasion has altered fire characteristics (e.g. Whisenant, 1990 ; Hughes ., 1991 ; Pivello & Coutinho, 1996 ). For example, in Florida, Cogongrass ( Imperata cylindrica (L.) Beauv.) invasion has increased fine‐fuel loads resulting in hotter, more continuous fires ( Lippincott, 2000 ). Alien grass invasion may cause dramatic changes in fire regimes through a positive feedback cycle, which D’Antonio & Vitousek (1992 ) called the ‘grass‐fire cycle’. The grass‐fire cycle occurs when an alien grass invades an area and increases the abundance of fine fuel, which increases fire frequency, area and in some cases intensity. This causes a decline in tree and shrub cover, facilitating further grass invasion, which in turn increases the likelihood of future fire and so the invasion proceeds in a self‐perpetuating cycle. The grass‐fire cycle has been documented in Hawaii, western North America, Central and South America ( D’Antonio & Vitousek, 1992 ; Freifelder ., 1998 ; D’Antonio ., 2000 ). The current spread of the African grass Andropogon gayanus in the tropical savannas of northern Australia has lead to concerns about its possible effect on fire regimes. Introduced to Australia as a pasture grass in 1931 ( Oram, 1990 ), A. gayanus is now established outside pastoral systems in the Northern Territory and Queensland ( Smith, 2001, 2002 ). Concern has been expressed that invasion by A. gayanus could increase fuel loads and promote intense, late dry season fires ( Williams ., 2002 ; Russell‐Smith . in press ). This could potentially initiate a grass‐fire cycle in Australia's savannas. To date, these concerns have been based on speculation and anecdotal evidence, so there is an urgent need for quantitative data to test these claims. To determine whether a grass‐fire cycle has been initiated in the savannas of northern Australia it is necessary to test the assumptions of D’Antonio & Vitousek's (1992 ) model, namely: (1) alien grass invasion alters the fuel load characteristics, for example, by increasing fuel loads or increasing flammability; (2) altered fuel characteristics lead to an increase in fire frequency, area and/or intensity; (3) the altered fire characteristics lead to a decrease in tree cover; (4) there is an increase in alien grass cover in the postfire community. This study tested the first two assumptions, specifically that A. gayanus invasion increases fuel loads and fire intensity. In the Northern Territory, fuel loads and fire regimes have been described in detail for savannas dominated by an overstorey of Eucalyptus tetrodonta (F. Muell) and Eucalyptus miniata (Cunn. Ex Schauer), with an understorey of native annual and perennial grasses ( Andersen ., 1998 ; Williams ., 1998 ; Gill ., 2002 ). These data provide a solid baseline for comparisons with areas that have been invaded by A. gayanus . METHODS This study was undertaken in the mesic savanna communities of the Northern Territory. The major vegetation types at both locations were open forest and woodlands ( sensu Specht, 1981 ) dominated by E. tetrodonta and E. miniata with a native grassy understorey dominated by Pseudopogonatherum contortum (Brongn.) A. Camus., Chrysopogon fallax S.P. Blake. and Aristida spp. The climate is characterized by a distinct wet season (December–March) alternating with an almost rainless dry season (May–October). The average annual rainfall at Darwin, approximately 120 km from the study sites, is 1708.3 mm, and annual rainfall during 2001 was 1443.8 mm (Darwin Airport data 12°39′S, 132°53′E, Bureau of Meteorology 2001). Temperatures are high throughout the year, with the monthly mean maximum temperature ranging from 31.3 °C in July to 35.6 °C in November In this region, the fire season occurs in the dry season. Fires are of relatively low intensity, but there are distinct seasonal differences in fire behaviour, due largely to changes in fire weather, fuel production and curing ( Williams ., 1998 ). The fuel load is primarily fine fuels (< 6 mm diameter) comprising grass and tree leaf litter. The grass component consists of annual and perennial grasses. Annual grasses (e.g. Sorghum spp.) commence curing late in the wet season (March/April) whereas perennials ( Heteropogon triticeus (R. Br) Stapf., Aristida spp.) cure early in the dry season (June). The fuel load increases and the proportion of leaf and twig components in the woodlands increase as the dry season progresses, due to the leaf fall of deciduous and semideciduous trees ( Williams ., 2002 ). Early dry season fires (May–June) occur when fuels are still relatively moist and tend to be very low in intensity (≈ 2000 kW m −1 ), patchy and limited in extent ( Gill ., 1990 ). As the dry season progresses, the understorey vegetation cures and the fire weather becomes more extreme ( Gill ., 1996 ). As a result, late dry season fires are more intense (≈ 8000 kW m −1 ), and extensive, than early dry season fires ( Williams ., 1998 ). Fuel Loads In July 2001 fuel loads were measured at two locations: Wildman Reserve (12°43′S, 131°49′E), approximately 120 km east of Darwin, and Crater Lake (13°02′S, 131°05′E), approximately 120 km south of Darwin. At both locations, A. gayanus has invaded an understorey of native grasses, although the extent of invasion and the density of A. gayanus plants are much greater at Wildman Reserve. Within a 1‐ha site, fine fuel (< 6 mm minimum diameter) was sampled from eight quadrats (1 m 2 ); four within patches of A. gayanus (up to 30 m 2 in size and at least 10 m apart) and four within nearby native grass, in which there was no A. gayanus present. Fuel was sampled at a single site at Crater Lake and, because of the greater extent of infestation, at three sites (approximately 500 m apart) at Wildman Reserve. Samples were oven dried (for 48 h at 80 °C) and weighed. Differences in fuel loads between grass types ( A. gayanus vs. native) were compared using a t ‐test for Crater Lake and a two‐way anova (with factors Grass Type (Fixed) and Sites (Random)) for Wildman Reserve. Fire Characteristics In June 2002, fire characteristics were measured at two sites at Wildman Reserve, approximately 5 km apart. Both sites were densely infested with A. gayanus , beneath a canopy cover of E. tetrodonta and E. miniata . At Site 1, three 1 ha plots were chosen, approximately 100 m apart, and at Site 2, two 1 ha plots were chosen, approximately 400 m apart. All measurements were taken at least 30 m from the road. The savanna communities of Wildman Reserve are burnt frequently (typically annually or biennially) as part of the Reserve's fire management strategy (Alan Anderson pers. comm., 2001). This fire frequency is typical for this region. For example, data on the fire regimes for nearby Kakadu National Park between 1980 and 1994 indicated that approximately 65% of Eucalyptus dominated woodland and 50% of open forest was burnt annually over the 15‐year period ( Russell‐Smith ., 1997 ). Approximately 60% of this occurred in the early dry season. Similarly, fire history mapping over the broader landscape indicated that 50% of the mesic savannas in Australia's Northern Territory were burnt at least 3 years in 6 ( Williams ., 2002 ). Both sites were subject to fuel reduction burns in the first week of June as part of the Reserve's fire management strategy. Fires were lit at 1500 h. Temperature, relative humidity, wind speed, and wind direction were measured immediately prior to ignition. At Site 1, a single‐line back fire (against the wind direction) was lit. At Site 2, a single‐line head fire was lit on the windward side of the savanna, then, as the fire front progressed through the savanna, a second fire line was lit to contain the first. To determine intensity, specific fuel loads were measured by harvesting all above‐ground biomass in three randomly placed 2 × 1 m quadrats within each plot at each site. Samples were oven dried (for 48 h at 80 °C) and weighed. The moisture content of the fuel was determined from small samples of fuel, which were placed in plastic, double sealed, sample bags and stored on ice. Samples were weighed within 3 h of cutting and then oven dried (for 48 h at 80 °C) and re‐weighed. Moisture content (dry weight/fresh weight × 100) of the fuel was determined. At Site 2, the fire's rate of spread was measured at both plots using a series of electronic temperature‐residence‐time meters (TRTMs) following Moore . (1995 ). The TRTMs were buried in a series of near‐equilateral triangles, with sides 10–20 m in length, following Simard . (1984 ). The rate of spread and the fuel load data were used to calculate Byram fire‐line intensity ( Byram, 1959 ), with the heat yield of the fuel assumed to be 20 000 kJ/kg, as is conventionally used in Australian fire studies ( Gill & Knight, 1991 ; Williams ., 1998 ). On the day after the fires, average char and scorch heights were measured on 10 adult trees within each plot, following Williams . (1998 ). Leaf‐scorch height (the maximum height of scorched leaves within a tree canopy) was used as an indicator of fire characteristics such as flame height and fire intensity ( Williams ., 1998 ). Leaf‐char height (the height to which leaves are blackened) was used as a surrogate for flame height ( Gill & Moore, 1994 ). Comparison of results The results obtained at Wildman Reserve were compared to the data collected on fire regimes from 5 years of experimental fires in E. tetrodonta and E. miniata dominated vegetation at Kapalga Research Station (see Andersen ., 1998 ; for a description of the Kapalga Fire Experiment, and Williams ., 1998 ; for a comprehensive description of the fire behaviour). The Kapalga Fire Experiment is the most comprehensive fire study that has been conducted to date in the northern Australian savannas ( Andersen & Braithwaite, 1992 ; Andersen ., 1998 ). The savannas of Kapalga and Wildman Reserve overlie the deeply weathered and partly laterized Late Tertiary sediments of the Koolpinyah surface, which comprises the gently undulating lowland plains that stretch from Darwin to the Arnhem Land escarpment ( Russell‐Smith ., 1995 ). The soils supporting the savanna communities at both locations are typically deep well‐drained red earths ( Day ., 1983 ; Cook, in press ). The savannas at Kapalga are typical of those occurring throughout the Top End of the Northern Territory, and fall into the same vegetation classification ( E. miniata , E. tetrodonta open forest; Wilson ., 1996 ) as those studied at Wildman Reserve. Differences between the Kapalga and Wildman Reserve data sets could not be statistically analysed, due to heterogeneity of variances. RESULTS Fuel load Once established, A. gayanus is an obvious element in the savanna understorey ( Fig. 1 ). At the time of the study, the native tall‐grasses had cured and fallen over and were less than 0.5 m high. By contrast, the A. gayanus tussocks had not fully cured, and were up to 4.1 m tall with basal areas of 0.6 m 2 . 1 Dense infestation of Andropogon gayanus Kunth. (Gamba grass) at Wildman Reserve. A. gayanus now forms the main understorey at this site. At Crater Lake mean fuel loads in A. gayanus were four times higher than in native grass (1.54 ± 0.47 vs. 0.39 ± 0.12 kg m −2 ; t 6 = 3.84, p = 0.009; Fig. 2 ). At Wildman there was a significant difference between grass types ( F 1,2 = 45.20, p = 0.02) and no difference between sites ( F 2,18 ) = 0.93, p = 0.41. The A. gayanus grass fuel load was significantly higher than that of native grasses at all three sites (Tukey's HSD, p < 0.05), but the difference was much greater at Site 3 (1.72 ± 0.29 vs. 0.25 ± 0.02 kg m −2 ) than Sites 1 or 2 (Site 1, 1.12 ± 0.07 vs. 0.38 ± 0.09 kg m −2 ; Site 2, 1.20 ± 0.07 vs. 0.27 ± 0.07 kg m −2 ) which resulted in a significant interaction between grass types and sites ( F 2,18 ) = 5.28, p = 0.15; Fig. 2. 2 Mean (± SE) fuel loads (g m −2 ) measured in 1 m 2 quadrats in savanna understorey dominated by introduced Andropogon gayanus Kunth. (□) and native grasses (▪). Samples were collected from two locations: Crater Lake (CL) and Wildman Reserve (WR). Fire intensity The fires monitored at Wildman Reserve occurred during typical early dry season weather, with low wind speed and low relative humidity ( Table 1a ). The fuel loads at Sites 1 and 2 were 4.4 ± 0.6 t ha −1 and 10.2 ± 1.0 t ha −1 , respectively, which was substantially higher than the fuel loads measured in annually burnt native grass savannas at Kapalga in the early dry season ( Table 1b ). Average moisture content of A. gayanus was more than double that measured at Kapalga in the early dry season, and four times greater than the late dry season ( Table 1 ). 1 Summary of (a) weather characteristics, and (b) fire behaviour of experimental fires at Wildman Reserve and Kapalga Research Station. Rows are mean values (± SE). The Kapalga data are based on five years of fire monitoring. Detailed results are presented in Williams . (1998 ) Wildman Reserve Kapalga Site 1‐Early Site 2‐Early Early Late (a) Weather Characteristics Temperature (°C) 30.1 29.6 30.3 35.1 Relative Humidity (%) 36 34 34 24 Wind Speed (m s −1 ) 1 0.4 2 3 Wind Direction SE SE SE ESE (b) Fire Behaviour Fuel (t ha −1 ) 4.4 (0.6) 10.2 (1) 3.2 (0.6) 5.0 (0.6) Fuel Moisture (% ODW) 44.4 (1.3) 43.5 (2.8) 19.3 (0.8) 11.1 (0.9) Rate of Spread (m s −1 ) N/A 0.72 (0.3) 0.37 (0.1) 0.78 (0.1) Fire Intensity (kW m −1 ) N/A 15700 (6200) 2100 (290) 7700 (290) Char Height (m) 5.2 (0.5) 9.9 (0.6) 1.3 2.8 Scorch height (m) 21 (0.7) 16 (0.6) 11 20 Despite the higher fuel moisture, the mean rate of spread (0.72 ± 0.3 m s −1 ) measured at Wildman Reserve was approximately double that of the early dry season fires at Kapalga (0.37 ± 0.1 m s −1 ) and similar to that measured during the late dry season fires at Kapalga (0.78 ± 0.1 m s −1 ; Williams ., 1998 ). Approximately 90–100% of all fuel was consumed at each of the measurement sites. Mean fire intensity at Site 2 (15 700 ± 6200 kW m −1 ; Fig. 3a ; Table 1 ) was approximately eight times higher than the mean intensity measured for early dry season fires at Kapalga (2100 ± 290 kW m −1 ; Fig. 3b ; Table 1 ). The highest intensity measured (Site 2, Plot 1) was 24 000 kW m −1 , which is greater than any fire intensity measured at Kapalga, including fires that occurred after fuel had built up over several years (1990; 18 000 kW m −1 ; Williams ., 1998 ). Furthermore, the mean intensity of the early dry season fire at Wildman Reserve was over double that of the late dry season fires at Kapalga (means 15700 ± 6200 and 7700 ± 260 kW m −1 , respectively). Fig. 3 (a) Early dry season fire at Wildman Reserve. The mean fire intensity of this fire was ≈ 16000 kW m −1 . Fuel loads here were predominantly Andropogon gayanus Kunth. (Gamba grass). (b) Typical early dry season fire at Kapalga Research Station. Mean fire intensities for these fires was ≈ 2000 kW m −1 . Fuel loads here were predominantly Sorghum spp. Crown fires were not observed at Wildman Reserve, however, ‘torching’ (the ignition of foliage) of certain species, particularly Pandanus spiralis R. Br. and Erythrophleum chlorostachys (F. Muell.) Baill., was observed. Average leaf‐char height was extremely high for northern Australia (means 5.2 ± 0.5 m and 9.9 ± 0.6 m at Site 1 and 2, respectively; Table 1 ). Individual char heights of up to 15 m were recorded, which was more than 10 m higher than the maximum char height at Kapalga in either the early or late dry season. The average char height at Site 2 (9.9 m) was approximately eight times higher than that recorded at Kapalga (1.3 m) in the early dry season. Leaf‐scorch height was observed at the top of canopy of the tallest trees at both sites (means 21 ± 0.7 m and 16 ± 0.6 m at Site 1 and 2, respectively). Based on the char heights recorded, scorch heights at Site 2 would be expected to be higher than that recorded at Site 1. However, at Site 2, the scorch height was limited by the height of the canopy and therefore it did not provide a useful indication of fire intensity and flame height. Despite this, scorch heights at Wildman were twice that recorded at Kapalga in the early dry season (2.8 m), and similar to that measured at Kapalga in the late dry season (20 m; Table 1 ). DISCUSSION The results of this study support the first two assumptions of D’Antonio & Vitousek's (1992 ) grass‐fire cycle; that alien grass invasion alters the fuel characteristics of the savanna, which leads to an increase in fire intensity. Fine‐fuel loads are on average four times greater in savannas heavily invaded by A. gayanus . The characteristics of the fuel bed are also changed following invasion. A. gayanus grows up to 4.75 m (N.A. Rossiter, unpublished data) over the wet season whereas native tall grasses are typically between 1 and 3 m ( Wilson ., 1996 ). A. gayanus cures later and remains erect for longer into the dry season, forming a taller, denser fuel load than the native grasses. This fuel architecture led to a higher rate of spread than predicted by the high fuel moisture content, high humidity and low wind speed ( Cheney & Sullivan, 1997 ). A. gayanus invasion is likely to result in substantial changes to the savanna fire regime. Even in the early dry season, the large fuel loads resulting from A. gayanus invasion can support fires on average eight times more intense than those fuelled by native grasses. A. gayanus invasion may also increase fire frequency because land managers have to burn the invaded sites each year to reduce the large fuel loads that A. gayanus generates and to reduce the potential for high intensity late dry season wildfires (Alan Anderson pers. comm.). Fire frequency may also increase because A. gayanus has the potential to support fire more than once a year. Due to its high tolerance to fire and ability to resprout soon after fire ( Bowden, 1963 ), A. gayanus can produce sufficient biomass to support a second fire in the same season (Chris Howard pers. obs.). Invasion by A. gayanus also has the potential to increase the uniformity and extent of fires that occur in savannas. In native grass savanna, early dry season burns are patchy and limited in extent, due largely to the discontinuous nature of the fuel loads and fire weather ( Williams ., 1998 ). With an increased, continuous fuel load comes an increased probability of uniform fires over large areas. Such an intense, frequent and uniform fire regime is likely to drastically alter the vegetation structure and composition of communities. Although savannas are relatively resistant to the effects of annual low intensity fire ( Stocker & Mott, 1981 ), high intensity fires in the savannas can substantially reduce savanna woody species recruitment ( Setterfield, 2002 ) and cause considerable tree stem mortality ( Williams ., 1998 ). Fires in areas invaded by A. gayanus are therefore likely to cause high tree mortality and reduce tree recruitment, which is the third assumption of the grass‐fire cycle. This study clearly demonstrates that the invasion of A. gayanus creates the initial conditions required to initiate a grass‐fire cycle in the savannas of northern Australia. The effect of A. gayanus on fire intensity in northern Australia is similar to that resulting from the invasion of African grasses in other ecosystems, and in these cases the landscape has been grossly modified by this process ( Smith, 1985 ; D’Antonio & Vitousek, 1992 ; Gordon, 1998 ; Williams & Baruch, 2000 ). As with other fire‐promoting grasses such as Bromus tectorum L. and Melinis minutiflora Beauv, A. gayanus can clearly be described as an ecosystem ‘transformer’ ( sensu Richardson ., 2000 ) with the potential to alter the community structure and the nutrient, water and carbon cycling processes over large areas of Australia's savanna ecosystems. It has been argued that such species ‘demand a major allocation of resources for containment/control/eradication’ ( Richardson ., 2000 ). The concern over the continued spread and invasion of this grass is thus justified, and greater attention from savanna land managers is called for. ACKNOWLEDGMENTS We thank the Parks and Wildlife Commission of the Northern Territory (PWCNT) for access to Wildman Reserve. We would also like to thank the rangers at Wildman Reserve, particularly Alan Anderson and Chris Howard, for their advice on study sites and for conducting the fires. Many thanks to Dick Williams for his advice and use of the timers. We also thank Michael Welch for his assistance, particularly during the fire. Alan Andersen, Gary Cook and Dick Williams, provided helpful comments on the draft manuscript. The Tropical Savannas CRC provided Natalie Rossiter with an Honours scholarship to undertake the project. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diversity and Distributions Wiley

Testing the grass‐fire cycle: alien grass invasion in the tropical savannas of northern Australia

Loading next page...
 
/lp/wiley/testing-the-grass-fire-cycle-alien-grass-invasion-in-the-tropical-owzv4zz52t

References (42)

Publisher
Wiley
Copyright
Copyright © 2003 Wiley Subscription Services, Inc., A Wiley Company
ISSN
1366-9516
eISSN
1472-4642
DOI
10.1046/j.1472-4642.2003.00020.x
Publisher site
See Article on Publisher Site

Abstract

INTRODUCTION Plant invasions are regarded as one of the most serious threats to global biodiversity and ecosystem functioning ( Heywood, 1989 ; D’Antonio & Vitousek, 1992 ; Vitousek ., 1996 ; Williams & Baruch, 2000 ). Grasses have been identified as a particularly invasive set of plant species ( Heywood, 1989 ) with substantial areas of invasion occurring on almost every continent ( D’Antonio & Vitousek, 1992 ). One of the most significant ecological changes caused by invading alien grasses is the alteration of fire regimes ( D’Antonio & Vitousek, 1992 ). There have been numerous documented cases where alien grass invasion has altered fire characteristics (e.g. Whisenant, 1990 ; Hughes ., 1991 ; Pivello & Coutinho, 1996 ). For example, in Florida, Cogongrass ( Imperata cylindrica (L.) Beauv.) invasion has increased fine‐fuel loads resulting in hotter, more continuous fires ( Lippincott, 2000 ). Alien grass invasion may cause dramatic changes in fire regimes through a positive feedback cycle, which D’Antonio & Vitousek (1992 ) called the ‘grass‐fire cycle’. The grass‐fire cycle occurs when an alien grass invades an area and increases the abundance of fine fuel, which increases fire frequency, area and in some cases intensity. This causes a decline in tree and shrub cover, facilitating further grass invasion, which in turn increases the likelihood of future fire and so the invasion proceeds in a self‐perpetuating cycle. The grass‐fire cycle has been documented in Hawaii, western North America, Central and South America ( D’Antonio & Vitousek, 1992 ; Freifelder ., 1998 ; D’Antonio ., 2000 ). The current spread of the African grass Andropogon gayanus in the tropical savannas of northern Australia has lead to concerns about its possible effect on fire regimes. Introduced to Australia as a pasture grass in 1931 ( Oram, 1990 ), A. gayanus is now established outside pastoral systems in the Northern Territory and Queensland ( Smith, 2001, 2002 ). Concern has been expressed that invasion by A. gayanus could increase fuel loads and promote intense, late dry season fires ( Williams ., 2002 ; Russell‐Smith . in press ). This could potentially initiate a grass‐fire cycle in Australia's savannas. To date, these concerns have been based on speculation and anecdotal evidence, so there is an urgent need for quantitative data to test these claims. To determine whether a grass‐fire cycle has been initiated in the savannas of northern Australia it is necessary to test the assumptions of D’Antonio & Vitousek's (1992 ) model, namely: (1) alien grass invasion alters the fuel load characteristics, for example, by increasing fuel loads or increasing flammability; (2) altered fuel characteristics lead to an increase in fire frequency, area and/or intensity; (3) the altered fire characteristics lead to a decrease in tree cover; (4) there is an increase in alien grass cover in the postfire community. This study tested the first two assumptions, specifically that A. gayanus invasion increases fuel loads and fire intensity. In the Northern Territory, fuel loads and fire regimes have been described in detail for savannas dominated by an overstorey of Eucalyptus tetrodonta (F. Muell) and Eucalyptus miniata (Cunn. Ex Schauer), with an understorey of native annual and perennial grasses ( Andersen ., 1998 ; Williams ., 1998 ; Gill ., 2002 ). These data provide a solid baseline for comparisons with areas that have been invaded by A. gayanus . METHODS This study was undertaken in the mesic savanna communities of the Northern Territory. The major vegetation types at both locations were open forest and woodlands ( sensu Specht, 1981 ) dominated by E. tetrodonta and E. miniata with a native grassy understorey dominated by Pseudopogonatherum contortum (Brongn.) A. Camus., Chrysopogon fallax S.P. Blake. and Aristida spp. The climate is characterized by a distinct wet season (December–March) alternating with an almost rainless dry season (May–October). The average annual rainfall at Darwin, approximately 120 km from the study sites, is 1708.3 mm, and annual rainfall during 2001 was 1443.8 mm (Darwin Airport data 12°39′S, 132°53′E, Bureau of Meteorology 2001). Temperatures are high throughout the year, with the monthly mean maximum temperature ranging from 31.3 °C in July to 35.6 °C in November In this region, the fire season occurs in the dry season. Fires are of relatively low intensity, but there are distinct seasonal differences in fire behaviour, due largely to changes in fire weather, fuel production and curing ( Williams ., 1998 ). The fuel load is primarily fine fuels (< 6 mm diameter) comprising grass and tree leaf litter. The grass component consists of annual and perennial grasses. Annual grasses (e.g. Sorghum spp.) commence curing late in the wet season (March/April) whereas perennials ( Heteropogon triticeus (R. Br) Stapf., Aristida spp.) cure early in the dry season (June). The fuel load increases and the proportion of leaf and twig components in the woodlands increase as the dry season progresses, due to the leaf fall of deciduous and semideciduous trees ( Williams ., 2002 ). Early dry season fires (May–June) occur when fuels are still relatively moist and tend to be very low in intensity (≈ 2000 kW m −1 ), patchy and limited in extent ( Gill ., 1990 ). As the dry season progresses, the understorey vegetation cures and the fire weather becomes more extreme ( Gill ., 1996 ). As a result, late dry season fires are more intense (≈ 8000 kW m −1 ), and extensive, than early dry season fires ( Williams ., 1998 ). Fuel Loads In July 2001 fuel loads were measured at two locations: Wildman Reserve (12°43′S, 131°49′E), approximately 120 km east of Darwin, and Crater Lake (13°02′S, 131°05′E), approximately 120 km south of Darwin. At both locations, A. gayanus has invaded an understorey of native grasses, although the extent of invasion and the density of A. gayanus plants are much greater at Wildman Reserve. Within a 1‐ha site, fine fuel (< 6 mm minimum diameter) was sampled from eight quadrats (1 m 2 ); four within patches of A. gayanus (up to 30 m 2 in size and at least 10 m apart) and four within nearby native grass, in which there was no A. gayanus present. Fuel was sampled at a single site at Crater Lake and, because of the greater extent of infestation, at three sites (approximately 500 m apart) at Wildman Reserve. Samples were oven dried (for 48 h at 80 °C) and weighed. Differences in fuel loads between grass types ( A. gayanus vs. native) were compared using a t ‐test for Crater Lake and a two‐way anova (with factors Grass Type (Fixed) and Sites (Random)) for Wildman Reserve. Fire Characteristics In June 2002, fire characteristics were measured at two sites at Wildman Reserve, approximately 5 km apart. Both sites were densely infested with A. gayanus , beneath a canopy cover of E. tetrodonta and E. miniata . At Site 1, three 1 ha plots were chosen, approximately 100 m apart, and at Site 2, two 1 ha plots were chosen, approximately 400 m apart. All measurements were taken at least 30 m from the road. The savanna communities of Wildman Reserve are burnt frequently (typically annually or biennially) as part of the Reserve's fire management strategy (Alan Anderson pers. comm., 2001). This fire frequency is typical for this region. For example, data on the fire regimes for nearby Kakadu National Park between 1980 and 1994 indicated that approximately 65% of Eucalyptus dominated woodland and 50% of open forest was burnt annually over the 15‐year period ( Russell‐Smith ., 1997 ). Approximately 60% of this occurred in the early dry season. Similarly, fire history mapping over the broader landscape indicated that 50% of the mesic savannas in Australia's Northern Territory were burnt at least 3 years in 6 ( Williams ., 2002 ). Both sites were subject to fuel reduction burns in the first week of June as part of the Reserve's fire management strategy. Fires were lit at 1500 h. Temperature, relative humidity, wind speed, and wind direction were measured immediately prior to ignition. At Site 1, a single‐line back fire (against the wind direction) was lit. At Site 2, a single‐line head fire was lit on the windward side of the savanna, then, as the fire front progressed through the savanna, a second fire line was lit to contain the first. To determine intensity, specific fuel loads were measured by harvesting all above‐ground biomass in three randomly placed 2 × 1 m quadrats within each plot at each site. Samples were oven dried (for 48 h at 80 °C) and weighed. The moisture content of the fuel was determined from small samples of fuel, which were placed in plastic, double sealed, sample bags and stored on ice. Samples were weighed within 3 h of cutting and then oven dried (for 48 h at 80 °C) and re‐weighed. Moisture content (dry weight/fresh weight × 100) of the fuel was determined. At Site 2, the fire's rate of spread was measured at both plots using a series of electronic temperature‐residence‐time meters (TRTMs) following Moore . (1995 ). The TRTMs were buried in a series of near‐equilateral triangles, with sides 10–20 m in length, following Simard . (1984 ). The rate of spread and the fuel load data were used to calculate Byram fire‐line intensity ( Byram, 1959 ), with the heat yield of the fuel assumed to be 20 000 kJ/kg, as is conventionally used in Australian fire studies ( Gill & Knight, 1991 ; Williams ., 1998 ). On the day after the fires, average char and scorch heights were measured on 10 adult trees within each plot, following Williams . (1998 ). Leaf‐scorch height (the maximum height of scorched leaves within a tree canopy) was used as an indicator of fire characteristics such as flame height and fire intensity ( Williams ., 1998 ). Leaf‐char height (the height to which leaves are blackened) was used as a surrogate for flame height ( Gill & Moore, 1994 ). Comparison of results The results obtained at Wildman Reserve were compared to the data collected on fire regimes from 5 years of experimental fires in E. tetrodonta and E. miniata dominated vegetation at Kapalga Research Station (see Andersen ., 1998 ; for a description of the Kapalga Fire Experiment, and Williams ., 1998 ; for a comprehensive description of the fire behaviour). The Kapalga Fire Experiment is the most comprehensive fire study that has been conducted to date in the northern Australian savannas ( Andersen & Braithwaite, 1992 ; Andersen ., 1998 ). The savannas of Kapalga and Wildman Reserve overlie the deeply weathered and partly laterized Late Tertiary sediments of the Koolpinyah surface, which comprises the gently undulating lowland plains that stretch from Darwin to the Arnhem Land escarpment ( Russell‐Smith ., 1995 ). The soils supporting the savanna communities at both locations are typically deep well‐drained red earths ( Day ., 1983 ; Cook, in press ). The savannas at Kapalga are typical of those occurring throughout the Top End of the Northern Territory, and fall into the same vegetation classification ( E. miniata , E. tetrodonta open forest; Wilson ., 1996 ) as those studied at Wildman Reserve. Differences between the Kapalga and Wildman Reserve data sets could not be statistically analysed, due to heterogeneity of variances. RESULTS Fuel load Once established, A. gayanus is an obvious element in the savanna understorey ( Fig. 1 ). At the time of the study, the native tall‐grasses had cured and fallen over and were less than 0.5 m high. By contrast, the A. gayanus tussocks had not fully cured, and were up to 4.1 m tall with basal areas of 0.6 m 2 . 1 Dense infestation of Andropogon gayanus Kunth. (Gamba grass) at Wildman Reserve. A. gayanus now forms the main understorey at this site. At Crater Lake mean fuel loads in A. gayanus were four times higher than in native grass (1.54 ± 0.47 vs. 0.39 ± 0.12 kg m −2 ; t 6 = 3.84, p = 0.009; Fig. 2 ). At Wildman there was a significant difference between grass types ( F 1,2 = 45.20, p = 0.02) and no difference between sites ( F 2,18 ) = 0.93, p = 0.41. The A. gayanus grass fuel load was significantly higher than that of native grasses at all three sites (Tukey's HSD, p < 0.05), but the difference was much greater at Site 3 (1.72 ± 0.29 vs. 0.25 ± 0.02 kg m −2 ) than Sites 1 or 2 (Site 1, 1.12 ± 0.07 vs. 0.38 ± 0.09 kg m −2 ; Site 2, 1.20 ± 0.07 vs. 0.27 ± 0.07 kg m −2 ) which resulted in a significant interaction between grass types and sites ( F 2,18 ) = 5.28, p = 0.15; Fig. 2. 2 Mean (± SE) fuel loads (g m −2 ) measured in 1 m 2 quadrats in savanna understorey dominated by introduced Andropogon gayanus Kunth. (□) and native grasses (▪). Samples were collected from two locations: Crater Lake (CL) and Wildman Reserve (WR). Fire intensity The fires monitored at Wildman Reserve occurred during typical early dry season weather, with low wind speed and low relative humidity ( Table 1a ). The fuel loads at Sites 1 and 2 were 4.4 ± 0.6 t ha −1 and 10.2 ± 1.0 t ha −1 , respectively, which was substantially higher than the fuel loads measured in annually burnt native grass savannas at Kapalga in the early dry season ( Table 1b ). Average moisture content of A. gayanus was more than double that measured at Kapalga in the early dry season, and four times greater than the late dry season ( Table 1 ). 1 Summary of (a) weather characteristics, and (b) fire behaviour of experimental fires at Wildman Reserve and Kapalga Research Station. Rows are mean values (± SE). The Kapalga data are based on five years of fire monitoring. Detailed results are presented in Williams . (1998 ) Wildman Reserve Kapalga Site 1‐Early Site 2‐Early Early Late (a) Weather Characteristics Temperature (°C) 30.1 29.6 30.3 35.1 Relative Humidity (%) 36 34 34 24 Wind Speed (m s −1 ) 1 0.4 2 3 Wind Direction SE SE SE ESE (b) Fire Behaviour Fuel (t ha −1 ) 4.4 (0.6) 10.2 (1) 3.2 (0.6) 5.0 (0.6) Fuel Moisture (% ODW) 44.4 (1.3) 43.5 (2.8) 19.3 (0.8) 11.1 (0.9) Rate of Spread (m s −1 ) N/A 0.72 (0.3) 0.37 (0.1) 0.78 (0.1) Fire Intensity (kW m −1 ) N/A 15700 (6200) 2100 (290) 7700 (290) Char Height (m) 5.2 (0.5) 9.9 (0.6) 1.3 2.8 Scorch height (m) 21 (0.7) 16 (0.6) 11 20 Despite the higher fuel moisture, the mean rate of spread (0.72 ± 0.3 m s −1 ) measured at Wildman Reserve was approximately double that of the early dry season fires at Kapalga (0.37 ± 0.1 m s −1 ) and similar to that measured during the late dry season fires at Kapalga (0.78 ± 0.1 m s −1 ; Williams ., 1998 ). Approximately 90–100% of all fuel was consumed at each of the measurement sites. Mean fire intensity at Site 2 (15 700 ± 6200 kW m −1 ; Fig. 3a ; Table 1 ) was approximately eight times higher than the mean intensity measured for early dry season fires at Kapalga (2100 ± 290 kW m −1 ; Fig. 3b ; Table 1 ). The highest intensity measured (Site 2, Plot 1) was 24 000 kW m −1 , which is greater than any fire intensity measured at Kapalga, including fires that occurred after fuel had built up over several years (1990; 18 000 kW m −1 ; Williams ., 1998 ). Furthermore, the mean intensity of the early dry season fire at Wildman Reserve was over double that of the late dry season fires at Kapalga (means 15700 ± 6200 and 7700 ± 260 kW m −1 , respectively). Fig. 3 (a) Early dry season fire at Wildman Reserve. The mean fire intensity of this fire was ≈ 16000 kW m −1 . Fuel loads here were predominantly Andropogon gayanus Kunth. (Gamba grass). (b) Typical early dry season fire at Kapalga Research Station. Mean fire intensities for these fires was ≈ 2000 kW m −1 . Fuel loads here were predominantly Sorghum spp. Crown fires were not observed at Wildman Reserve, however, ‘torching’ (the ignition of foliage) of certain species, particularly Pandanus spiralis R. Br. and Erythrophleum chlorostachys (F. Muell.) Baill., was observed. Average leaf‐char height was extremely high for northern Australia (means 5.2 ± 0.5 m and 9.9 ± 0.6 m at Site 1 and 2, respectively; Table 1 ). Individual char heights of up to 15 m were recorded, which was more than 10 m higher than the maximum char height at Kapalga in either the early or late dry season. The average char height at Site 2 (9.9 m) was approximately eight times higher than that recorded at Kapalga (1.3 m) in the early dry season. Leaf‐scorch height was observed at the top of canopy of the tallest trees at both sites (means 21 ± 0.7 m and 16 ± 0.6 m at Site 1 and 2, respectively). Based on the char heights recorded, scorch heights at Site 2 would be expected to be higher than that recorded at Site 1. However, at Site 2, the scorch height was limited by the height of the canopy and therefore it did not provide a useful indication of fire intensity and flame height. Despite this, scorch heights at Wildman were twice that recorded at Kapalga in the early dry season (2.8 m), and similar to that measured at Kapalga in the late dry season (20 m; Table 1 ). DISCUSSION The results of this study support the first two assumptions of D’Antonio & Vitousek's (1992 ) grass‐fire cycle; that alien grass invasion alters the fuel characteristics of the savanna, which leads to an increase in fire intensity. Fine‐fuel loads are on average four times greater in savannas heavily invaded by A. gayanus . The characteristics of the fuel bed are also changed following invasion. A. gayanus grows up to 4.75 m (N.A. Rossiter, unpublished data) over the wet season whereas native tall grasses are typically between 1 and 3 m ( Wilson ., 1996 ). A. gayanus cures later and remains erect for longer into the dry season, forming a taller, denser fuel load than the native grasses. This fuel architecture led to a higher rate of spread than predicted by the high fuel moisture content, high humidity and low wind speed ( Cheney & Sullivan, 1997 ). A. gayanus invasion is likely to result in substantial changes to the savanna fire regime. Even in the early dry season, the large fuel loads resulting from A. gayanus invasion can support fires on average eight times more intense than those fuelled by native grasses. A. gayanus invasion may also increase fire frequency because land managers have to burn the invaded sites each year to reduce the large fuel loads that A. gayanus generates and to reduce the potential for high intensity late dry season wildfires (Alan Anderson pers. comm.). Fire frequency may also increase because A. gayanus has the potential to support fire more than once a year. Due to its high tolerance to fire and ability to resprout soon after fire ( Bowden, 1963 ), A. gayanus can produce sufficient biomass to support a second fire in the same season (Chris Howard pers. obs.). Invasion by A. gayanus also has the potential to increase the uniformity and extent of fires that occur in savannas. In native grass savanna, early dry season burns are patchy and limited in extent, due largely to the discontinuous nature of the fuel loads and fire weather ( Williams ., 1998 ). With an increased, continuous fuel load comes an increased probability of uniform fires over large areas. Such an intense, frequent and uniform fire regime is likely to drastically alter the vegetation structure and composition of communities. Although savannas are relatively resistant to the effects of annual low intensity fire ( Stocker & Mott, 1981 ), high intensity fires in the savannas can substantially reduce savanna woody species recruitment ( Setterfield, 2002 ) and cause considerable tree stem mortality ( Williams ., 1998 ). Fires in areas invaded by A. gayanus are therefore likely to cause high tree mortality and reduce tree recruitment, which is the third assumption of the grass‐fire cycle. This study clearly demonstrates that the invasion of A. gayanus creates the initial conditions required to initiate a grass‐fire cycle in the savannas of northern Australia. The effect of A. gayanus on fire intensity in northern Australia is similar to that resulting from the invasion of African grasses in other ecosystems, and in these cases the landscape has been grossly modified by this process ( Smith, 1985 ; D’Antonio & Vitousek, 1992 ; Gordon, 1998 ; Williams & Baruch, 2000 ). As with other fire‐promoting grasses such as Bromus tectorum L. and Melinis minutiflora Beauv, A. gayanus can clearly be described as an ecosystem ‘transformer’ ( sensu Richardson ., 2000 ) with the potential to alter the community structure and the nutrient, water and carbon cycling processes over large areas of Australia's savanna ecosystems. It has been argued that such species ‘demand a major allocation of resources for containment/control/eradication’ ( Richardson ., 2000 ). The concern over the continued spread and invasion of this grass is thus justified, and greater attention from savanna land managers is called for. ACKNOWLEDGMENTS We thank the Parks and Wildlife Commission of the Northern Territory (PWCNT) for access to Wildman Reserve. We would also like to thank the rangers at Wildman Reserve, particularly Alan Anderson and Chris Howard, for their advice on study sites and for conducting the fires. Many thanks to Dick Williams for his advice and use of the timers. We also thank Michael Welch for his assistance, particularly during the fire. Alan Andersen, Gary Cook and Dick Williams, provided helpful comments on the draft manuscript. The Tropical Savannas CRC provided Natalie Rossiter with an Honours scholarship to undertake the project.

Journal

Diversity and DistributionsWiley

Published: May 1, 2003

There are no references for this article.