Regional Environmental Change

, Volume 11, Issue 1, pp 19–27

Late 20th century mangrove encroachment in the coastal Australian monsoon tropics parallels the regional increase in woody biomass

Authors

    • School of Plant ScienceUniversity of Tasmania
  • Guy S. Boggs
    • Tropical Spatial Sciences GroupCharles Darwin University
  • David M. J. S. Bowman
    • School of Plant ScienceUniversity of Tasmania
Original Article

DOI: 10.1007/s10113-010-0109-5

Cite this article as:
Williamson, G.J., Boggs, G.S. & Bowman, D.M.J.S. Reg Environ Change (2011) 11: 19. doi:10.1007/s10113-010-0109-5

Abstract

In Kakadu National Park, a World Heritage property in the Australian monsoon tropics 250 km to the east of Darwin, a number of recent studies have shown that woody encroachment (expansion of woody communities) and densification (increased biomass in woody communities) has occurred in the last 40 years. The cause of this increase in woody biomass is poorly understood, but possibly associated with the control of invasive Asian water buffalo, trend to higher rainfall, and increased frequency of fires. Mangroves provide an important context to understand these landscape changes, given that they are unaffected by fire or feral water buffalo. We examine change in mangrove distribution in a series of coastal tropical swamps fringing Darwin, Northern Territory, Australia over a 30-year period using a series of 7 aerial photographs spanning 23 years from 1974 and a 2004 high-resolution satellite image. In late 1974, Darwin was impacted by an intense tropical cyclone. Vegetation at 3,000 randomly placed points was manually classified, and a multinomial logistic model was used to asses the impact of landscape position (coastal, intertidal, and upper-tidal) and swamp on mangrove change between 1974 and 2004. Over the study period, there was instability and slight mangrove loss at the coast, stability in the intertidal zone, and mangrove gain in the upper-tidal zone, with an overall increase in mangrove presence of 16.2% above the pre-cyclone distribution. A swamp that was impacted by drainage works for mosquito control and the construction of a sewage treatment plant showed a greater mangrove increase than the two unmodified swamps. The mangrove expansion is consistent with woody encroachment observed in nearby but ecologically distinct systems. Plausible causes for this change include changed local hydrology, changes in sea level, and elevated atmospheric CO2 concentrations.

Keywords

MangrovesAerial photographyClimate changeCoastal ecologyLandscape changeVegetation dynamicsWoody vegetation

Introduction

In many landscapes throughout the world, there is a trend toward encroachment of woody vegetation into adjoining treeless areas and increased cover of woody vegetation in wooded communities, sometimes called densification. These processes have been studied in great detail in Kakadu National Park, a World Heritage property in the Australian monsoon tropics 250 km to the east of Darwin, using sequences of orthorectified aerial photography. These studies have revealed that since the 1960s, monsoon forest patches are expanding into tropical savanna (Banfai et al. 2007; Banfai and Bowman 2007), tropical savannas are increasing their canopy cover (Lehmann et al.2008), and encroachment of fringing woody vegetation is occurring on treeless freshwater floodplains (Bowman et al. 2008). There remains uncertainty and debate about the cause of this expansion, given that a number of changes have occurred to potential drivers over the last 40 years. For example, feral Asian water buffalo underwent a population irruption in the 1970s, which was not brought under control until the 1990s (Petty et al. 2007). Fire regimes have changed from a preponderance of fire in the late dry season in the late 1950s to sustained burning throughout the late dry season (Bowman et al. 2007). Correlative studies have failed to link the landscape scale woody expansion to either proxies of fire activity or buffalo density, although these studies show that at the local scale these disturbances can influence both encroachment and densification (Banfai and Bowman 2007; Bowman et al. 2008; Lehmann et al. 2008). Collectively, researchers point to the importance of the recent trend for increasing rainfall, or more controversially, increased atmospheric CO2, in driving encroachment and densification in Kakadu National Park.

In this context, studies of mangrove communities provide an important perspective on these processes, given these systems are unaffected by landscape fire or Asian water buffalo impacts. Nonetheless, mangroves are highly dynamic systems that respond to slight changes in sediment accumulation, salinity, water depth, and recurrence of tidal inundation (Bridgewater and Cresswell 1999). For example, mangroves bordering salt flats can show localized dieback or expansion in response to slight changes in the hydrology of the system (Duke et al.1998).

Aerial photography has been shown to be useful for assessing vegetation change over medium-term timescales (Fensham and Fairfax 2002) and has been used to study mangrove change (Jones et al.2004; McTainsh et al. 1986; Saintilan and Wilton 2001). For example, a recent study conducted in Florida with aerial photographs spanning 60 years has found a dramatic shift in mangrove vegetation up to 3.3 km inland, that appears to be associated with a combination of sea level rise and local hydrological effects (Ross et al. 2000).

Our aim is to identify changes in mangrove distribution and examine correlates of this change in the period 1974–2004 in three contiguous coastal swamps in the Northern Territory of Australia. All three swamps were severely impacted by Cyclone Tracy, a category 4 tropical cyclone that occurred in December 1974 and destroyed the city of Darwin and damaged the surrounding native vegetation (Stocker 1976). Leanyer Swamp has been impacted by urbanization in its catchment and the construction of drains in the early 1980s in order to increase the drainage of stormwater and to prevent to pooling of water from high tides, a preferred habitat of saltwater mosquitoes such as Aedes vigilax (Medical Entomology Sect. 1983). The other swamps have been minimally impacted by urbanization. There is some evidence that nearby mangrove systems, for example, in the Mary River region (Mulrennan and Woodroffe 1998) have responded to sea level change over the last 50 years. Therefore, swamps provide an opportunity to chart mangrove recovery following cyclone damage, assess the impacts of urbanization, and contextualize background changes associated potentially with global change including increased rainfall, sea level rise, and elevated CO2.

To examine changes in mangrove vegetation, we use a sequence of high resolution remotely sensed imagery (7 aerial photographs spanning 23 years that were georeferenced to a QuickBird satellite image acquired in 2004).

Study area

The study system comprises three contiguous swamps, Leanyer, Holmes Jungle, and Micket, developed around two tidal creeks, Buffalo Creek and Micket Creek, to the east of Darwin in the Northern Territory of Australia (12°22′S, 130°55′E, Fig. 1). This region experiences a monsoonal climate, with 90% of the average 1,660 mm annual rainfall occurring during the wet season from November to April, and a mean daily maximum temperature of 32°C that varies little throughout the year. The study covers an area of approximately 30 km2 with a catchment area of 35 km2. The area has a low elevation gradient and a maximum tidal range of close to 8 m, with the tide reaching over 4 km from the coast along the tidal creeks at high tide. Vegetation along each swamp is dominated by a gradient in salinity and water availability, ranging from freshwater reed swamps, to brackish reeds and sedges, to mangrove forest, with grassland and salt flats more distant from the tidal creeks. Comparable vegetation patterns, in nearby Howard’s Swamp, have been described by Wilson and Bowman (1987).
https://static-content.springer.com/image/art%3A10.1007%2Fs10113-010-0109-5/MediaObjects/10113_2010_109_Fig1_HTML.gif
Fig. 1

Map of 2004 vegetation communities in the study. The coastal, inter-tidal, and upper-tidal zones used in the analysis are shown. The zones were defined in respect of tidal regimes and vegetation types (Table 2)

Materials and methods

Aerial photographs taken at low tide and during the austral winter dry season (May to August), covering the study area for the years 1974, 1976, 1979, 1983, 1985, 1990, and 1997, were obtained from the Northern Territory Department of Planning and Infrastructure. A high-resolution Quickbird Satellite image, also captured at low tide, was used for the 2004 coverage. The photographs obtained included color, black and white, and color-infrared prints for various years. The photographs were registered to the 2004 satellite image in ArcMap 9.1, using a 3rd-order polynomial transform, with RMS error values of around five meters. The 2004 satellite image was manually digitized and classified into polygons representing a set of eight vegetation or landscape units at a scale of 1:3000 following ground reconnaissance. The creek network was digitized using the 2004 satellite image. The aerial photographs were all taken on the low-tide cycle, thus enabling manual mapping of the coastal line as evidenced by boundary between sand and mud flats.

Elevation data in the form of a digital contour map were obtained from the Northern Territory Government, but it was found that the 1-m intervals were too coarse to adequately represent the slope of the salt flat and low-lying mangrove areas. A field survey was conducted using a differential GPS system at 120 randomly placed points in the low-lying salt pan and brackish swamp areas, to fill in gaps in the contour map to a vertical resolution of under 10 cm. A digital elevation model with a cell size of 100 m was produced from the combined data using ordinary kriging with a spherical semi-variogram model using the Spatial Analysis extension in ArcGIS 9.1.

A total of 3,000 sample points were randomly placed across the vegetation map, and a number of habitat attributes were derived from the 2004 vegetation map and preceding sequences of aerial photography. For each aerial photographic coverage, the points were manually classified into one of the 8 vegetation/landscape units by examining the vegetation type inside circles of 10 m diameter at each point. For the purposes of the analysis, a generic mangrove formation was scored as present if >50% cover of the mangrove vegetation unit was observed at a point. Data for the elevation, distance to creek, and distance to coast were spatially joined to the point samples. Following field surveys on the ground and by helicopter during different times in the tidal cycle, and following advice from the Medical Entomology Branch of the Northern Territory Government who conduct frequent assessments of tidal extent for the purposes of mosquito control, the swamps were a priori classified into three landscape zones (coastal, intertidal, and upper-tidal) based on distance from coast and vegetation type and the sample points were assigned to these three landscape position zones. While we recognize a continuous gradient in tidal and saline influence exists, these three zones tend to represent distinct vegetation associations and tidal regimes.

The sample points were recoded as gain, loss, or stable mangrove based on the difference in mangrove presence at the point between 1974 and 2004, and a multinomial logistic model, performed in R 2.8.1 (R Development Core Team 2008), was used to determine the effect of swamp and landscape position on mangrove change.

Darwin tidal gauge and rainfall data used in interpretation were obtained from Australian National Tidal Centre and the Australian Bureau of Meteorology, respectively.

Results

A vegetation map and map of point classification into landscape position zones are presented in Fig. 1. There was clear zonation in vegetation types from the coast inland. In the coastal zone, the elevation was around 6 m Australian Height Datum rising up to around 7 m in the upper-tidal zone some 3 km inland (Table 1). Distance from creek lines increases across this gradient from 100 to 300 m in the coast and intertidal zones to almost 1 km in the upper-tidal zone (Table 1). The coastal zone is dominated by mangrove vegetation, the intertidal zone by salt flats, and the upper-tidal zone has the highest proportion of grass, sedge, and reed vegetation (Table 2). Vegetation patterns are related to salinity gradients and frequency and depth of tidal inundation. For example, in the coastal zone, mangroves (including Bruguiera parviflora, Rhizophora stylosa, and Avicennia marina) are dominant, occupying areas subject to frequent tidal inundation. Mangroves also extend through the intertidal zone along tidal creeks although much of this zone is comprised of sparsely vegetated salt flats, that due to hypersaline conditions and extreme evaporation rates only support mangrove vegetation close to frequently flushed creeks The upper-tidal zone comprises a salinity gradient from low mangrove forests dominated by Ceriops tagal, Lumnitzera racemosa, Avicennia marina, and Sonneratia lanceolata, through brackish Schoenoplectus littoralis reeds to Eleocharis dulcis in areas subject primarily to freshwater inundation.
Table 1

Mean and standard deviation of environmental variables for three landscape zones in the three swamps

Swamp

Coastal

Intertidal

Upper-tidal

Elevation (m)

 Leanyer

6.95 ± 0.47

6.92 ± 0.48

7.46 ± 0.15

 Holmes Jungle

6.51 ± 0.31

6.92 ± 0.21

7.05 ± 0.29

 Micket

6.62 ± 0.36

6.92 ± 0.70

7.40 ± 0.43

Creek distance (km)

 Leanyer

0.32 ± 0.22

0.27 ± 0.25

0.91 ± 0.51

 Holmes Jungle

0.19 ± 0.14

0.43 ± 0.33

0.66 ± 0.49

 Micket

0.18 ± 0.14

0.07 ± 0.06

0.53 ± 0.40

Shore distance (km)

 Leanyer

0.65 ± 0.37

1.78 ± 0.39

3.58 ± 0.59

 Holmes Jungle

0.57 ± 0.34

2.04 ± 0.67

4.04 ± 0.74

 Micket

0.62 ± 0.37

2.19 ± 0.50

3.48 ± 0.43

Elevation is based upon Australian Height Datum, creek distance is the distance from nearest tidal creek, shore distance is the nearest distance to the low-tide point

Table 2

Change in the percentage of all vegetation/landscape units for the three swamps broken down by coastal, intertidal, and upper-tidal zones between 1974 and 2004

Swamp

Position

Salt flat (%)

Water (%)

Grass/sedge (%)

Eleocharis (%)

Typha (%)

Schoenoplectus (%)

Mangrove (%)

Woodland (%)

Holmes Jungle

Coast

20.8 → 28.3

0.5 → 0.5

3.7 → 2.7

0.0 → 0.0

0.0 → 0.0

0.0 → 0.0

72.2 → 64.2

2.7 → 4.3

1974 → 2004

Inter

58.4 → 71.2

1.5 → 0.3

14.8 → 6.2

0.0 → 0.0

0.0 → 0.0

0.0 → 0.0

24.9 → 21.4

0.3 → 0.9

 

Upper

21.2 → 18.0

0.0 → 0.0

41.7 → 39.0

9.9 → 8.7

0.9 → 1.7

5.2 → 2.6

20.3 → 28.2

0.9 → 1.7

Leanyer

Coast

32.7 → 28.4

0.5 → 1.1

5.2 → 6.1

0.0 → 0.0

0.0 → 0.0

0.0 → 0.0

54.7 → 58.4

6.9 → 6.1

1974 → 2004

Inter

44.4 → 36.2

0.3 → 0.6

14.3 → 7.7

0.0 → 0.0

0.0 → 0.0

0.0 → 0.0

35.2 → 51.0

5.7 → 4.6

 

Upper

13.8 → 27.1

0.0 → 0.0

26.9 → 44.8

17.8 → 0.0

0.0 → 0.0

28.4 → 0.0

5.5 → 22.7

7.6 → 5.4

Micket

Coast

8.2 → 10.6

8.8 → 15.3

0.0 → 0.0

0.0 → 0.0

0.0 → 0.0

0.0 → 0.0

82.9 → 74.1

0.0 → 0.0

1974 → 2004

Inter

37.5 → 32.9

3.9 → 5.4

3.9 → 4.9

0.0 → 0.0

0.0 → 0.0

0.0 → 0.0

51.9 → 54.2

2.8 → 2.6

 

Upper

19.8 → 16.7

0.0 → 0.0

30.6 → 30.6

14.4 → 14.9

0.0 → 0.0

14.0 → 6.8

11.7 → 24.8

9.5 → 6.3

Salt flat includes all vegetation-free areas inundated by the highest tides. Typha, Eleocharis and Schoenoplectus are graminoid-dominated vegetation types that fall along a gradient of increasing salinity, while the grass/sedge category comprises the remaining treeless vegetation. Woodland comprises all non-mangrove woody vegetation, including savanna and coastal thicket

Across the 3,000 sample points, there was an overall increase in mangrove presence of 16.2% between 1974 and 2004, amounting to an increase of approximately 123 ha from the 1974 coverage of 761 ha. The upper-tidal zone showed a consistent increase in mangrove cover over the 30-year study period, and little mangrove loss after the 1974 cyclone. In contrast, the coastal zone showed some mangrove loss over the study period, particularly following the cyclone (Fig. 2). The intertidal zone in Micket and Holmes Jungle swamps showed little change in total mangrove area at the end of the study period, though an increase in this zone was seen in Leanyer swamp that was modified by drainage works (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs10113-010-0109-5/MediaObjects/10113_2010_109_Fig2_HTML.gif
Fig. 2

Proportion of sample points with mangroves over time for the coastal, intertidal, and upper-tidal zones across the three swamps for all intervals between 1974 and 2004

From the beginning to end of the study period, the negligible transitions away from mangrove vegetation were mostly to open water or salt flat (Table 3). The transition matrix shows a large proportion of the points transitioning to mangrove vegetation were originally brackish Schoenoplectus sedgelands characteristic of the upper-tidal zone, and fewer mangrove transitions were from salt flats or open water. In the upper-tidal zone of Leanyer swamp, in particular, Eleocharis and Schoenoplectus sedgelands, sample points showed a significant conversion to mangroves as a result of drain construction for mosquito control (Table 2).
Table 3

Vegetation transition matrix, showing proportion of points in each category converting to other categories between 1974 and 2004

 

2004

 

Grass/sedge (%)

Mangrove (%)

Salt flat (%)

Schoenoplectus

Woodland (%)

Eleocharis (%)

Water (%)

Urban (%)

Typha (%)

1974

Grass/sedge

71.0

4.2

20.6

0.0

3.5

0.5

0.0

0.2

0.0

Mangrove

1.0

87.7

7.2

0.2

0.5

0.0

3.2

0.1

0.0

Salt Flat

5.0

18.7

75.4

0.1

0.5

0.0

0.1

0.2

0.0

Schoenoplectus

9.4

44.1

28.3

12.6

0.8

4.7

0.0

0.0

0.0

Woodland

21.5

10.3

4.7

0.0

57.0

5.6

0.0

0.9

0.0

Eleocharis

27.0

7.0

13.9

4.3

1.7

42.6

0.0

0.9

2.6

Water

0.0

33.3

10.3

0.0

0.0

0.0

56.4

0.0

0.0

Urban

0.0

10.0

0.0

0.0

30.0

0.0

0.0

60.0

0.0

Typha

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

100.0

Vegetation/landscape units are as in Table 2

The swamp × landscape position interaction models of mangrove increase and decrease between the start and end of the study period showed a trend toward greater conversion to mangrove in the upper-tidal zone (Fig. 3a) and greater loss of mangroves at the coastal zone (Fig. 3b). There was a statistically significant difference between swamps (p < 0.05), with Leanyer swamp having particularly high rates of conversion to mangrove and low rates of conversion away from mangrove.
https://static-content.springer.com/image/art%3A10.1007%2Fs10113-010-0109-5/MediaObjects/10113_2010_109_Fig3_HTML.gif
Fig. 3

Mean proportion of points showing mangrove increase (a) and decrease (b) between 1974 and 2004. Error bars indicate standard deviation reported by multinomial logistic regression

Mean sea level and rainfall (Fig. 4) in this region have fluctuated during the study period, and although a number of years stand out as exceptional, no trend can be discriminated at the available temporal resolution.
https://static-content.springer.com/image/art%3A10.1007%2Fs10113-010-0109-5/MediaObjects/10113_2010_109_Fig4_HTML.gif
Fig. 4

Preceding 5-year mean sea level and mean annual rainfall for study intervals in Darwin

Discussion

Despite a catastrophic cyclone in 1974, there has been a significant increase in the coverage in mangrove distribution in the swamps adjacent to Darwin over the 30-year study period. Immediately following the tropical cyclone, mangrove cover declined, although there was recovery within 20 years back to the original distributional area of the mangroves. There was both loss and gain of mangrove cover in the coastal zone over the study period, consistent with the continuously shifting substrates found previously in this system (Woodroffe and Grime 1999). In contrast, the upper-tidal zone mangroves underwent expansion and densification, replacing treeless brackish vegetation and, to a lesser extent, the hypersaline flats. The contrasting response of mangroves with different positions along the elevation gradient from the coast to the inland reflects the geomorphologies of the landscape settings: the coastal zone is highly dynamic while in the intertidal zone favors more stable vegetation patterns, given the relatively fixed salinity gradients between creeklines and treeless salt flats and floodplains (Hollins and Ridd 1997), unless impacted by drainage works.

Increased sedimentation as the result of urbanization and changing precipitation regimes are often cited as drivers of mangrove expansion globally (Schwarz 2003; Alongi et al. 2005; Jupiter et al. 2007). The significantly different patterns of mangrove increase in Leanyer swamp appear to be related to hydrological changes caused by drainage works. This swamp was subject to the construction of a sewage treatment plant, disrupting the flow of fresh and tidal water, and the construction of a system of drains for mosquito control (Department of Construction and A.A.Heath 1978), which allowed more rapid drainage of the swamp after heavy rainfall or high tides (Medical Entomology Branch 1982). The pattern of mangrove expansion in Leanyer around the drains is visible in the satellite imagery (Fig. 5), a change that has also been observed in mosquito control channels in eastern Australia (Breitfuss et al. 2003; Jones et al. 2004). Harty (2004) notes that urbanization, with the associated increased nutrient and sediment loads and changes to hydrology through engineering works can also promote mangrove establishment consistent with our finding for Leanyer swamp. However, the fact that the two swamps not subject to human influence also showed strong trends in mangrove in the upper-tidal zone signals some profound changes to the ecology of these swamps.
https://static-content.springer.com/image/art%3A10.1007%2Fs10113-010-0109-5/MediaObjects/10113_2010_109_Fig5_HTML.gif
Fig. 5

Change in extent of mangrove vegetation in the upper-tidal areas of the three swamps between 1974 and 2004

At the boundary between temperate and subtropical climates, changing temperatures and reduced frequency of frosts have been proposed as a mechanism for mangrove expansion into treeless salt marsh communities in the United States (Stevens et al. 2006). But changes in temperature are unlikely to impact significantly on mangrove communities in northern Australia, which experiences a tropical climate with no frosts. Herbert (2007) suggests that mangrove expansion into salt marsh habitat at the landward fringe in the Hunter estuary in New South Wales is due to increased tidal range resulting from the construction of tidal barriers and channel dredging. However, such engineering works have not taken place in the region of the Darwin study, and any alteration in mean sea level or tidal amplitude is presumed to be the result of global change rather than local anthropogenic impacts.

Elevated atmospheric CO2 concentration has frequently been proposed as a driver of the expansion of woody species at the expense of non-woody vegetation (Bazzaz 1990; Archer et al. 1995; Bond and Midgley 2000; Eamus and Palmer 2007), as higher CO2 concentrations are expected to increase photosynthetic rate and water use efficiency more in woody plants than non-woody plants. Mangrove species have shown responses in growth to elevated CO2 (Farnsworth et al. 1996) although this is mediated by salinity. Controlled mangrove growth experiments under elevated CO2 have found little growth enhancement in high-salinity conditions but more growth enhancement in low-salinity conditions in a less salt-tolerant species (Ball et al. 1997). This effect could lead to expansion into brackish or freshwater areas, as we have observed in this system. However, controlled environment experiments growing mangroves and salt marsh species alone and in competition under elevated CO2 found that competition reduced the impact of elevated CO2 on mangrove growth, suggesting other factors may be required to explain mangrove expansion (McKee and Rooth 2008). Without further experimental data, we cannot reject the CO2 fertilizer effects as a plausible contributor to the mangrove expansion. However, experiments comparing the growth of the two invading mangrove species at the study site (Avicennia marina and Sonneratia lanceolata) with and without the presence of dominant sedge (Schoenoplectus littoralis) under elevated CO2 are required to more fully evaluate the importance of the putative fertilizer effect in this system. A further question arises as to the timescale over which any CO2 fertilization effect will be apparent. Simulations of forest dynamics in response to elevated CO2 in temperate forests have found significant increases in basal area over a time span of 50–150 years (Bolker et al. 1995). Free-air carbon enrichment (FACE) experiments have demonstrated increased carbon flux and photosynthesis in forests in response to CO2 enrichment, an effect that appears generally stronger in trees and C3 species, but structural changes in woody ecosystems have generally not been observed over the period of time these studies have been running (Ainsworth and Long 2005). It is possible that the rapid-growing, disturbance-tolerant mangrove communities may be capable of showing a rapid response to CO2 enrichment, and the application of FACE experiments to these systems would be informative.

The discontinuous landward expansion and seaward contraction is broadly correlated with corresponding fluctuations of sea level over the study period. Therefore, this study shows general agreement with previous studies qualitatively linking mangrove distribution changes with trends in sea level rise (McTainsh et al. 1986; Ross et al.2000) or that have been based on analysis of aerial photographs (Alongi 2008; Dowling 1978; Gilman et al. 2007). Like our study, these latter authors found mangrove invasion of brackish swamps at the limit of the inland tidal extent, and a coincident loss of seaward mangroves. An obvious explanation for the landward expansion of mangroves is the increased landward penetration of seawater (Lacerda et al. 2007). Simulation modeling of mangrove habitat based on projected sea level in Florida over the next 100 years showed significant replacement of freshwater marsh and swamp environments by mangroves (Doyle et al. 2003), and a similar outcome was predicted for Homebush Bay, New South Wales based on fuzzy set modeling (Zeng et al. 2007). In our study, direct time-series modeling of mangrove distribution against sea level or rainfall trends was not possible due to the low number of available repeat images and fluctuating tidal gauge data over the last decades (Church et al. 2006). Such variability combined with the sporadic sequence of imagery makes it difficult to directly link mangrove dynamics to sea level rise, although saline intrusion, possibly as the result of sea level rise, is a clear driver of inland mangrove expansion in nearby areas (Woodroffe and Mulrennan 1993; Bell et al.2001; Winn et al. 2006).

Northern Australia is showing a trend in increasing rainfall, particularly during the wet season (Smith 2004, Taschetto and England 2009), although this trend is not evident in the mean annual rainfall data obtained for the years of this study (Fig. 4). Increased rainfall is expected to lead to increased growth rates in mangroves (Ellison 2000). Periods of high rainfall are also expected to contribute to mangrove establishment in the hypersaline flats due to dilution of the salts in the soil (Field 1995; Duke 1997; Saintilan and Williams 1999; Harty 2004). However, expansion into hypersaline salt flats is limited at our study site (Table 3). Indeed, the major area of mangrove expansion was into brackish swamps that are inundated by the highest spring tides in October (at the end of the dry season) before being filled by freshwater in the wet season. This suggests mangrove expansion into these brackish areas is driven by increasing rather than decreasing salinity.

The expansion of mangroves observed in these coastal swamps is temporally and spatially consistent with the woody densification and expansion observed in nearby but ecologically distinct savannah and rainforest ecosystems. Densification in these systems has often been attributed to changes in fire regime or feral animal populations (Banfai et al. 2007; Bowman et al. 2008), factors that should not affect the mangroves. Conversely, mangrove expansion is often attributed to sea level rise, a factor unable to influence inland forests. Little is known about the two potential effects of those environmental drivers shared by both coastal mangrove forests and inland forests: atmospheric CO2 concentration and changing rainfall patterns. Studies examining mangrove tree growth across a wider area of northern Australia, so local differences in substrate types, disturbance history, and rainfall can be statistically evaluated would be a useful future direction for research, as would paired comparison of mangrove and rainforest tree growth within local climatic zones.

Conclusion

Despite severe damage from a tropical cyclone, the coastal swamps examined in this study all show landward mangrove expansion over a 30-year period, primarily replacing brackish reed swamps in the upper-tidal zone. Expansion rate is particularly high in the swamp that has undergone hydrological changes as the result of human engineering works, but given the expansion is also occurring in the other swamps, direct human interference alone cannot be established as the cause. While the observed changes are similar to those expected to be seen with sea level rise, this cannot be confirmed as the primary driver of change, given the fragmentary aerial photographic record. The mangrove expansion is consistent with densification trends in other ecologically distinct ecosystems in northern Australia, including savannas and rainforests, suggesting regional-scale factors are driving woody expansion. Plausible candidates for this change include changed local hydrology, changes in sea level, and elevated atmospheric CO2 concentrations.

Acknowledgments

Funding for this project was provided by the Australian Research Council (ARC) Linkage Grant (No. LP0667619), Northern Territory Department of Health and Community Services, Australian Bureau of Meteorology, Northern Territory Research and Innovation Fund, Australian Department of Defence, and Charles Darwin University. The authors would like to thank the Medical Entomology Branch of the Northern Territory Health Department, especially Peter Whelan for expertise and assistance, the Northern Territory Department of Planning and Infrastructure for assistance with spatial data and equipment, and Lubomir Bisevac and Dimity Boggs for assistance in the field.

Copyright information

© Springer-Verlag 2010