Collaborative Research on the Ecology and Management of the ‘Wulo’ Monsoon Rainforest in Wunambal Gaambera Country, North Kimberley, Australia

Abstract

Indigenous groups are increasingly combining traditional ecological knowledge and Western scientific approaches to inform the management of their lands. We report the outcomes of a collaborative research project focused on key ecological questions associated with monsoon vine thickets in Wunambal Gaambera country (Kimberley region, Western Australia). The study mapped monsoon rainforests and analysed the environmental correlates of their current distribution, as well as the historical drivers of patch dynamics since 1949. Remote sensing was used to chart the effectiveness of an intervention designed to re-instate Aboriginal fire regimes according to customary principles. We identified the most vulnerable patches based on size, distance from neighbouring patches, and fire frequency. More than 6000 rainforest patches were mapped. Most were small ( < 1 ha), occurring predominantly on nutrient-rich substrates (e.g., basalt) and fire-sheltered topographic settings (e.g., slopes and valleys). Rainforests with low fire frequency and no cattle were more likely to expand into surrounding long-unburnt savannas. Frequent fires and cattle did not cause substantial contraction, although the latter affected rainforest understories through trampling. Fire management performed by Aboriginal rangers effectively shifted fire regimes from high-intensity late dry season fires to early dry season fires, particularly in areas with clusters of vulnerable rainforests. The remote sensing methods developed in this project are applicable to the long-term monitoring of rainforest patches on Aboriginal-managed land in North Kimberley, providing tools to evaluate the impacts of fire management, feral animal control, and climate change. The study confirmed the importance of the cattle-free and rarely burnt Bougainville Peninsula as one of the most important rainforest areas inWestern Australia.

Full text

land
Article
Collaborative Research on the Ecology and
Management of the ‘Wulo’ Monsoon Rainforest in
Wunambal Gaambera Country,
North Kimberley, Australia
Tom Vigilante 1,2 ,*, Stefania Ondei 3, Catherine Goonack 1, Desmond Williams 1, Paul Young 2
and David M. J. S. Bowman 3ID
1Uunguu Healthy Country Team, Wunambal Gaambera Aboriginal Corporation, PMB 16 Kalumburu via
Wyndham, WA 6740, Australia; info@wunambalgaambera.org.au (C.G.);
adminmanager@wunambalgaambera.org.au (D.W.)
2Science and Conservation Team, Bush Heritage Australia, Melbourne, VIC 3000, Australia;
paul.young@bushheritage.org.au
3School of Biological Sciences, University of Tasmania, Sandy Bay, TAS 7005, Australia;
stefania.ondei@utas.edu.au (S.O.); david.bowman@utas.edu.au (D.M.J.S.B.)
*Correspondence: infohc@wunambalgaambera.org.au; Tel.: +61-427-163-080
Received: 14 August 2017; Accepted: 22 September 2017; Published: 5 October 2017
Abstract:
Indigenous groups are increasingly combining traditional ecological knowledge and
Western scientific approaches to inform the management of their lands. We report the outcomes
of a collaborative research project focused on key ecological questions associated with monsoon
vine thickets in Wunambal Gaambera country (Kimberley region, Western Australia). The study
mapped monsoon rainforests and analysed the environmental correlates of their current distribution,
as well as the historical drivers of patch dynamics since 1949. Remote sensing was used to chart the
effectiveness of an intervention designed to re-instate Aboriginal fire regimes according to customary
principles. We identified the most vulnerable patches based on size, distance from neighbouring
patches, and fire frequency. More than 6000 rainforest patches were mapped. Most were small (<1 ha),
occurring predominantly on nutrient-rich substrates (e.g., basalt) and fire-sheltered topographic
settings (e.g., slopes and valleys). Rainforests with low fire frequency and no cattle were more
likely to expand into surrounding long-unburnt savannas. Frequent fires and cattle did not cause
substantial contraction, although the latter affected rainforest understories through trampling. Fire
management performed by Aboriginal rangers effectively shifted fire regimes from high-intensity late
dry season fires to early dry season fires, particularly in areas with clusters of vulnerable rainforests.
The remote sensing methods developed in this project are applicable to the long-term monitoring
of rainforest patches on Aboriginal-managed land in North Kimberley, providing tools to evaluate
the impacts of fire management, feral animal control, and climate change. The study confirmed the
importance of the cattle-free and rarely burnt Bougainville Peninsula as one of the most important
rainforest areas in Western Australia.
Keywords:
aboriginal natural resource management; Australian monsoon tropics; biodiversity;
feral cattle; fire regimes; traditional ecological knowledge; tropical savanna
1. Introduction
The important role of Indigenous People in biodiversity conservation programs has been
increasingly recognized internationally [
1
–
5
]. This includes the declaration of Indigenous and
Community Conservation Areas (ICCAs), whereby indigenous or other communities voluntarily
Land 2017,6, 68; doi:10.3390/land6040068 www.mdpi.com/journal/land

Land 2017,6, 68 2 of 20
conserve their lands and waters as protected areas. In Australia, over 70 Indigenous Protected Areas
(IPAs), formally recognised by the Australian Government and counted in the National Reserve
System, have been declared by Indigenous groups in recent decades, and now cover 65 million
hectares, which represents 7% of Australia and 40% of the National Reserve System [
6
,
7
]. IPAs, like
other protected areas, require management planning, monitoring, and evaluation, and a variety of
approaches have been adopted by practitioners [
8
]. Implementation of management on indigenous
lands also requires the integration of the knowledge systems of Indigenous people (including
Traditional Ecological Knowledge) and non-indigenous partner agencies and the development of
participatory community-based approaches to monitoring and management (including Participatory
Monitoring and Management) [9–12].
The Uunguu Indigenous Protected Area was declared by Wunambal Gaambera Aboriginal
Corporation (WGAC, Kalumburu, Australia) in 2011 and now extends over 800,000 ha of Wunambal
Gaambera country (WGC) in the Kimberley region of Western Australia. The management plan for the
Uunguu IPA and wider WGC was developed [
13
] (and later reviewed [
14
]) using a local form of the
Open Standards for the Practice of Conservation called ‘Healthy Country Planning’ [
15
–
18
] and identified
10 important conservation ‘Targets’ (assets) to be monitored and managed. In implementing the plan,
WGAC has utilised collaborative research projects with external partners to provide Traditional
Owners with the opportunity to integrate Indigenous Knowledge and Western scientific methods and
approaches, to address key research questions such as monitoring target species (e.g., marine turtles
and dugong [19]).
Rainforests, known as ‘Wulo’ in Wunambal and Gaambera languages, are one of the 10 ‘Targets’
identified as important to Traditional Owners in the Wunambal Gaambera Healthy Country Plan. They
are an important source of plant food (e.g., yams, Dioscorea spp.) and animal food (particularly bats and
birds) and materials such as trees for making canoes (e.g., Bombax ceiba) [
20
]. Some rainforest patches
are associated with important cultural sites and feature in cultural stories including links to Gwion
rock art [
13
,
21
]. These rainforests, classified as monsoon vine thickets and dry rain-green forests, are
found across Northern Australia, occurring in regions with strongly seasonal climate at the driest end
of the Australian rainforest climatic range. They are characterised by low stature trees, a high density
of shrubs and vines, and are a type of tropical monsoon forest [
22
,
23
]. Rainforests were first formally
reported in Western Australia in 1965 [24], with some initial descriptive studies [25,26] followed by a
series of biological surveys in the region [
27
–
30
] and finally a dedicated Kimberley Rainforest Survey
(KRS) [
24
]. Despite the relatively small extent they occupy, rainforests contribute 25% of the floristic
biodiversity of North Kimberley [
31
], with some 453 species of vascular plants (93 of which were
rainforest specialist trees) and one endemic shrub (Hibiscus peralbus Fryxell). Unusually for that period,
the KRS team included an indigenous Wunambal Gaambera elder, Geoffrey Mangglamarra (now
deceased), who accompanied scientists to a number of rainforests in the WGC and contributed an
ethnoecological paper to the study [20].
Key threats to the health of rainforests identified in the Healthy Country Plan included unmanaged
wildfires and invasive species. Unmanaged wildfires have been identified as a key threat to natural
and cultural values in the Australian monsoon tropics [
32
–
35
]. Changes in fire regimes, as observed in
Australia with the switch from Traditional Aboriginal burning to European colonisation, are likely
to pose a threat to rainforests due to the increased extent and frequency of wildfires [
36
,
37
]. This is
in spite of the rainforest expansion trends observed in the Australian monsoon tropics in the past
years [
38
,
39
], which have been linked with regional wetting trends and possibly atmospheric CO
2
enrichment [
40
–
44
]. Intense fires can cause direct damage to entire rainforest patches or portions of
them, since rainforest species generally present limited fire protection (thick bark) and the ability to
resprout after a fire compared with savanna species [
45
–
47
]. It can also influence the dynamics of
woody vegetation in savanna adjacent to rainforest patches and affect the expansion and contraction of
rainforests over time [
38
,
40
]. Early rainforest surveys at the Mitchell Plateau identified wildfire impacts
and suggested that degradation had resulted in fewer bird records [
48
]. For example, abandoned

Land 2017,6, 68 3 of 20
Orange-footed Scrubfowl (Megapodius reinwardt) mounds on the rainforest edge can be an indicator of
change associated with fire as they are usually located under closed canopy [
49
,
50
]. Of the 95 patches
visited in as part of the KRS in 1987 KRS, 40 showed evidence of fire damage [35].
The presence of mega-herbivores, such as cattle and buffalos, can result in decreased fire intensity
by reducing fuel load through grazing [
51
]. Nonetheless, unmanaged grazers have a negative impact
on rainforests, by increased tree—and particularly seedling—death through wallowing and root
compaction, consequently reducing plant recruitment and understorey diversity [
52
,
53
]. In the
Australian monsoon tropics, feral pigs (Sus scrofa) and cattle (Bos Taurus/indicus) can damage rainforests
through trampling vegetation, particularly affecting small rainforest patches [
52
,
54
]. The KRS provided
anecdotal evidence of structural changes to a rainforest patch vegetation, including the loss of lower
level species, through trampling by cattle [
48
] and suggested that cattle may have facilitated fire damage
through the loss of canopy cover and spread of savanna grass species into the rainforest edge [55].
Invasive grasses have the potential to increase fuel loads and associated fire severity (e.g., gamba
grass Andropogon gayanus [
56
,
57
]), whilst vines such as Siam weed (Chromolaena odorata), rubber vine
(Cryptostegia spp.), and stinking passionflower (Passiflora foetida), as well as woody weeds such as neem
(Azadirachta indica), have the potential to alter the structure and composition of vine thickets [
58
,
59
].
Stinking passionflower is currently widespread across the Kimberley; its presence has been recorded in
North Kimberley since 1984 [
26
]. In 1987, stinking passionflower was recorded in half of the surveyed
patches [
60
]. Weeding programs, aimed to contain the spread of exotic invasive species in strategic
locations, have been carried out by WGAC and partner agencies since 2010.
This paper reviews the outcomes of a collaborative research project between WGAC and the
University of Tasmania, aimed at combining customary and scientific knowledge systems to answer
key ecological questions surrounding the management and monitoring of ‘Wulo’ or monsoon vine
thicket rainforests. Drawing together the results of joint field operations, the outputs of a series of
research papers [
47
,
61
,
62
] and additional unpublished analysis of remote sensing data, we discuss
(Table 1):
(a)
the location and characteristics of rainforest patches, mapped via a combination of an analysis of
remote sensing data and targeted field validation [62];
(b)
decadal scale rainforest boundary dynamics, detected using historical imagery and field
surveys [61];
(c)
the effects of climate change, fire frequency, and feral livestock occurrence on rainforest and
savanna historical expansion trends, inferred using natural landscape ecology experiments [
61
];
(d)
the use of a field experiment to investigate the effect of fire on rainforest and savanna seedling
survival and growth [47]; and
(e) recent changes in fire regimes following an intervention designed to restore elements of traditional
Aboriginal fire management.
Table 1.
Summary of methods presented in this paper. Some approaches were undertaken across all of
Wunambal Gaambera country (WGC), while others were limited to comparison of two locations or a
single location.
Aim Method Analysis Extent
References for Full
Method Description
and Results
Mapping rainforest
distribution
Creation of an ex novo map of rainforest patches from
2005 orthophotos using a 30 m ×30 m lattice
WGC
Ondei et al. 2017 [62]
Patch connectivity index estimated by calculating patch
area and average distance from the 5 nearest patches This paper
Determining
rainforest change
Rainforest patch change was estimated by applying
30 m ×30 m lattice to available historic aerial
photography (1949 and 1969) and comparing the output
with the 2005 orthophoto results.
Two locations: the
Bougainville Peninsula &
Mitchell Plateau
Ondei et al. [61]

Land 2017,6, 68 4 of 20
Table 1. Cont.
Aim Method Analysis Extent
References for Full
Method Description
and Results
Boundary
dynamics and
threats to patch
health
Florastic and vegetational data collected in 20 transects
(each with five 10 m ×20 m plots) spanning
rainforest–savanna boundaries. These data were
associated with rainforest change data (obtained from
the analysis of aerial photos and orthophotos).
Feral herbivore occupancy sampled with camera traps
placed along vegetation transects.
Two locations: the
Bougainville Peninsula &
Mitchell Plateau
Ondei et al. 2017 [61]
Unpublished data
Sapling survival
and growth
Controlled experiment on sapling survival rates and
resprouting strategies following fire.
One location: the
Mitchell Plateau Ondei et al. 2016 [47]
Fire regime change
GIS analysis comparing occurrence of fire (MODIS fire
scar data) adjacent to rainforest patches before and after
fire management intervention.
‘Vulnerability index’ determined using GIS analysis
WGC This paper
We combined the abovementioned results with literature on local traditional knowledge and
comments from Traditional Owners to consider implications and land management options for
Traditional Owners.
2. Materials and Methods
2.1. Study Area
The WGC occupies about 1.2 million ha of land and sea in the North Kimberley region of Western
Australia, which is located toward the driest portion of the rainforest range in the Australian tropics [
63
],
and listed as one of the 15 Australian National Biodiversity Hotspots [
64
]. The WGC is defined by the
Wanjina Wunggurr Uunguu Native Title Determination (Goonack v State of Western Australia [2011]
FCA 516), representing the traditional lands of Wunambal Gaambera Aboriginal people (Figure 1). The
majority of WGC has been declared as the Uunguu Indigenous Protected Area. Other conservation
estates include the Lawley R National Park, Camp Creek Conservation Park, and parts of Mitchell
River National Park, Laterite Conservation Park, and the Prince Regent River National Park.
Land 2017, 6, 68 4 of 20
threats to
patch health
savanna boundaries. These data were associated with
rainforest change data (obtained from the analysis of
aerial photos and orthophotos).
Feral herbivore occupancy sampled with camera traps
placed along vegetation transects.
Peninsula &
Mitchell Plateau
Sapling
survival and
growth
Controlled experiment on sapling survival rates and
resprouting strategies following fire.
One location: the
Mitchell Plateau Ondei et al. 2016 [47]
Fire regime
change
GIS analysis comparing occurrence of fire (MODIS fire
scar data) adjacent to rainforest patches before and
after fire management intervention.
‘Vulnerability index’ determined using GIS analysis
WGC This paper
We combined the abovementioned results with literature on local traditional knowledge and
comments from Traditional Owners to consider implications and land management options for
Traditional Owners.
2. Materials and Methods
2.1. Study Area
The WGC occupies about 1.2 million ha of land and sea in the North Kimberley region of
Western Australia, which is located toward the driest portion of the rainforest range in the
Australian tropics [63], and listed as one of the 15 Australian National Biodiversity Hotspots [64].
The WGC is defined by the Wanjina Wunggurr Uunguu Native Title Determination (Goonack v State
of Western Australia [2011] FCA 516), representing the traditional lands of Wunambal Gaambera
Aboriginal people (Figure 1). The majority of WGC has been declared as the Uunguu Indigenous
Protected Area. Other conservation estates include the Lawley R National Park, Camp Creek
Conservation Park, and parts of Mitchell River National Park, Laterite Conservation Park, and the
Prince Regent River National Park.
Figure 1. The extent of Wunambal Gaambera country as defined by the Wanjina Wunggurr Uunguu
Native Title Determination boundary. The Uunguu Indigenous Protected Area and co-existing State
Parks including 1. Mitchell River National Park; 2. Lawley River National Park; 3. Camp Creek
Conservation Park; 4. Prince Regent National Park.
The WGC lies within the Kimberley, a dissected plateau of deeply weathered Precambrian
sandstones with interspersed areas of Precambrian basalt, occasionally capped by mid-Cainozoic
laterite [65]. The climate is tropical monsoonal, characterised by a short summer wet season
Figure 1.
The extent of Wunambal Gaambera country as defined by the Wanjina Wunggurr Uunguu
Native Title Determination boundary. The Uunguu Indigenous Protected Area and co-existing State
Parks including 1. Mitchell River National Park; 2. Lawley River National Park; 3. Camp Creek
Conservation Park; 4. Prince Regent National Park.

Land 2017,6, 68 5 of 20
The WGC lies within the Kimberley, a dissected plateau of deeply weathered Precambrian
sandstones with interspersed areas of Precambrian basalt, occasionally capped by mid-Cainozoic
laterite [
65
]. The climate is tropical monsoonal, characterised by a short summer wet season (November
to April) during which most of precipitation occurs, while the rest of the year is virtually rain-free [
65
].
Mean annual rainfall currently ranges from 1000 to 1400 mm, and average maximum temperatures are
equal or higher than 30
◦
C all year, while average minimum temperatures vary from 20
◦
C or more
in summer to as low as 10
◦
C in winter [
66
]. A wetting and warming trend has been detected since
the beginning of the 21st century, with an increment in average annual rainfall of 40–50 mm/10 years,
and a temperature rise of 0.15
◦
C/10 years since the late 1940s [
66
]. WGC falls within the Gardner
botanical district classification [
65
] and the Mitchell sub-region of the North Kimberley Bioregion [
67
].
Its vegetation is predominantly biodiverse tropical savannas: Eucalyptus tetrodonta-E. miniata savannas
are found on the hills and laterite mesas, while E. tectifica-E. grandifolia savannas are common on
deeper, clay soils on plains. Small patches of monsoon rainforests are interspersed in the savanna,
typically found in fire-protected locations [
65
]. These monsoon rainforests support a high diversity of
plant species [
22
] and are important refugia for savanna-adapted mammals to avoid climatic extremes,
predators, and fires [
68
]. The monsoon forests of the North Kimberley Coast and islands and the
Kimberley Plateau have been inscribed on the West Kimberley National Heritage List in recognition of
the outstanding national heritage value of its evolutionary refugial role, which has resulted in high
invertebrate richness and endemism [69,70].
Some of the reported analyses focused on two specific locations with high rainforest density: the
Mitchell Plateau and the Bougainville Peninsula, which share similar climate and geology, but different
levels of disturbance. The Mitchell Plateau (754 km
2
) has a high fire frequency (average times burnt:
0.5 year
−1
, data from North Australian Fire Information (NAFI) [
71
], based on a 15-year time period
(2000–2014)), and unmanaged cattle are found throughout the area. By contrast, the Bougainville
Peninsula (298 km
2
) has a much lower fire frequency (average times burnt: 0.08 year
−1
), and cattle
have never been recorded [
35
]. These dynamics are likely to have persisted during the last several
decades; a study performed in 1985 found that half of the surveyed rainforest patches in the Mitchell
Plateau were damaged by fire and/or cyclones, while on the Bougainville Peninsula cyclones were the
only visible cause of damage [72].
In the Kimberley, the pastoral industry began in the late 1800s. Cattle stations adjacent to the
WGC were consolidated only in the 1950s, after the first roads were surveyed and constructed in the far
north. Cattle were first recorded at the Mitchell Plateau in 1976 [
48
], probably moving in from adjacent
pastoral leases established from the 1950s. Mining exploration for bauxite occurred at the Mitchell
Plateau and Bougainville Peninsula in the 1980s, and cut lines and crushing sites caused disturbance
across the laterite plateaus. Mining companies also established a vehicle track and airstrip at the
Mitchell Plateau, which facilitated visitation by tourists and promoted weed incursions along roadsides
and campgrounds (weeds include grader grass (Themeda quadrivalvis), mintweed (Hyptis suavolens),
and annual mission grass (Cenchrus pedicellatus syn. Pennisetum pedicellatum). The tourism industry has
expanded in recent decades including an expedition cruise industry operating along the coast [73].
Traditional burning practices undertaken by Wunambal Gaambera people [
74
] declined when
traditional lifestyles were disrupted by European colonisation. By the 1960s, Aboriginal burning was
limited to areas around communities where Aboriginal people had now settled [
75
]. The cessation
of Aboriginal burning led to the emergence of large destructive late dry season wildfires, set by both
anthropogenic ignitions and lightning, that became a dominant feature of the fire regime of the region
until recent times [76].
2.2. Mapping Distribution and Local Environmental Factors
An ex novo map of rainforest patches in the WGC was developed to investigate the relationship
between rainforest distribution and local environmental factors. The map was generated and
ground-truthed as described in Ondei et al. [
62
], overlaying a 30 m
×
30 m lattice on orthophotos

Land 2017,6, 68 6 of 20
of the area taken in 2005 (most recent available), and manually classifying each cell as ‘rainforest’ or
‘other’ based on the dominant vegetation type in the cell. Adjacent cells classified as ‘rainforest’
were merged to form a rainforest patch. Rainforest density was then calculated as rainforest
area/km
2
, and associated with local factors, such as geology, topography, and distance from the
coastline using thematic layers and ArcMap (v. 10.3, Environmental Systems Research Institute Inc.,
Redlands, CA, USA). A comparison was then made between mainland rainforests exposed to frequent
ambient wildfires and those on islands and peninsulas that have lower fire frequency being naturally
fire protected.
Based on the outcomes of the mapping process, we identified the most isolated patches by
calculating, for each rainforest patch, the average edge-to-edge distance of the nearest five neighbours,
using the ‘average nearest neighbour’ tool in ArcMap. A simple connectivity index was then calculated
as follows:
Patch connectivity =
A
Dist
where Ais the area of the focal patch (ha), and Dist is the average distance to the nearest five patches
(km) [
77
]. Patch isolation increases with the decrease of this index. Further analysis was then
undertaken to identify risk factors associated with isolation (see ‘Vulnerability Index’ below).
2.3. Vegetation Transects
Rainforest patches were sampled at two locations to investigate the influence of patch distribution,
environmental factors and potential threatening processes on rainforest plants. To do so, vegetation
transects, running across the rainforest–savanna boundary, were established in the Mitchell Plateau
and Bougainville Peninsula, and the data analysed [
61
]. Each transect included five 10 m
×
20 m plots,
and for each plot data on canopy cover, grass cover, rock cover, floristic, stand basal area, and number
of adult trees were collected.
Vertebrate species occupancy was also sampled by positioning 40 camera traps (RECONYX,
Holmen, WI, USA, HyperFire
TM
PC800 Professional) along the vegetation transects for 8–9 months
(from July/August 2014 to May/June 2015). For each transect, two cameras were positioned, one on
the rainforest edge and the other in the savanna, at a minimum distance of 500 m. Initial cattle results
are reported in this paper, while other species results are to be published elsewhere.
2.4. Determining Rainforest Change
The next step was to investigate the effects of disturbance on historical rainforest boundary
change and vegetation structure [
61
]. To do so, a landscape-scale natural experiment was performed,
mapping rainforest patches (adopting the same 30 m
×
30 m lattice applied above) in the fire- and
cattle-prone Mitchell Plateau and the low-disturbance Bougainville Peninsula, using available historical
aerial photographs from 1949 to 1969, and comparing the results with the map of current rainforest
distribution. To associate rainforest dynamics with current vegetation characteristics, each plot in
the transects described above was associated to the corresponding vegetation trends, classified as
‘stable rainforest’, ‘rainforest expanded in 1969’, ‘rainforest expanded in 2005’, or ‘stable savanna’.
This approach is explained in more detail by Ondei et al. [61].
2.5. Examining Threats to Rainforest Health
A controlled burning experiment was conducted to investigate the capacity of rainforest species
to survive a low-intensity fire. Rainforest and savanna saplings, located across the rainforest–savanna
boundary, were exposed to either a natural low-intensity fire (for savanna species) or a controlled
experimental burning, calibrated to reproduce the effects of a natural mild savanna fire (for rainforest
species) [
47
]. Differences in fire-protection (bark thickness), survival rates, and resprouting strategies
(basal or aerial) were recorded and compared.

Land 2017,6, 68 7 of 20
The recent fire history of rainforest patches was examined using ‘firescar ’ data created at a pixel
resolution of 250 m based on MODIS satellite imagery, accessed from the North Australian Fire
Information website [
71
,
78
]. A 250 m buffer was created around each individual rainforest patch and
the fire scar data for each year from 2000 to 2015 was analysed using the ‘spatial join’ function in
ArcGIS. Each rainforest patch was considered to have experienced fire if one or more fire scars occurred
within its area. This was intended to be an index of fire regime rather than an absolute measure of
fire impact on rainforest patches and their boundaries given the low resolution of MODIS relative
to rainforest mapping and other limitations with the firescar method [
79
–
81
]. Total fire frequency,
frequency of early dry season fires (EDS, January to July), and frequency of late dry season (LDS) fires
(August to December) were calculated [
82
]. As described earlier, the period from the 1950s to 2000s
had little or no effective fire management, whilst prescribed burning (a combination of ground-lit fires
and aerial burning) had been initiated in 2010 through the Wunambal Gaambera Healthy Country Plan.
Fire scar years were organised into unmanaged ‘baseline’ years (2000–2009) and managed ‘project’
years (2012–2015), aligning with the Wunambal Gaambera Uunguu Fire Project and the Savanna
Burning Methodology [
82
]. The years 2010 and 2011 were excluded from the analysis, as they were
‘transition’ years where WGAC was reaching management capacity. Differences in the frequency of
EDS fires, LDS fires, and total fire frequency were then analysed with a paired t-test. The frequency of
fires was calculated using the baseline period (2000–2009, for which fire scar data were accessible) as a
sample of the pre-management period (1950s–2000s). Pearson product moment correlation coefficient
(r) was employed to examine correlations between fire frequency and distance from the coastline as
well as slope. Fire frequency was also assessed in relation to geology types using the Kruskal–Wallis
test, and the average fire frequency was calculated for each geology type.
In order to identify which patches were most at risk of being completely destroyed by fire, with
lower chance of being recolonised, we calculated a vulnerability index, taking into account patch size,
distance from neighbouring patches, and fire frequency. The ‘Vulnerability index’ was calculated for
each rainforest patch as follows:
Patch connectivity
Total Fire Freqbas
where Patch connectivity includes information on patch size and distance from the nearest five patches,
and total Fire Freq
bas
is the total fire frequency in proximity of the patch recorded during the baseline
years. We used total fire frequency because under unmanaged fire regimes even EDS fires can be
extensive and intense due to fuel load accumulation. Note that lower index values indicate higher
fire risk.
Based on the fire risk index values, we tested for the presence of clusters of patches potentially at
risk, by using the Moran’s I test in ArcMap on (i) all patches and (ii) the 5% most vulnerable patches.
We then evaluated the effectiveness of planned burning in protecting particularly vulnerable patches
by comparing using a paired t-test, changes in LDS fire frequency for the 5% most vulnerable patches.
3. Results
3.1. Rainforest Distribution and Local Environmental Factors
A total of 6460 rainforest patches were mapped in the WGC, ranging from 0.1 to 220 ha in size
(average 1.6 ha
±
0.1 (SE)), although most (75%) were smaller than 1 ha (Figure 2) [
62
]. This was a
substantial improvement from the previous assessment of rainforest occurrence, which detected 1500+
rainforest patches over an area of 170,000 km
2
of the entire Kimberley region [
35
]. The disparity was
most likely due to the lower resolution (80 m pixel size) of the earlier study, so that smaller patches
were not detected.

Land 2017,6, 68 8 of 20
Land 2017, 6, 68 8 of 20
Figure 2. A map of rainforest patches in WGC. Insets show areas of high patch size and density at the
Bougainville Peninsula and Mitchell Plateau, including transect locations. Note that patch size has
been enhanced to better illustrate patch distribution at map scale.
Rainforest density is strongly dependent on rainfall, as shown by the correlations between
rainforest density and mean annual rainfall in the Australian monsoon tropics and distance from
coastline [62]. Rainforests were also preferentially located on relatively nutrient-rich substrates, such
as basalt, and on fire-protected topographic settings (slopes and valleys). These results confirm the
importance of nutrient and water availability, as well as protection from disturbance, as pivotal
factors affecting the dynamics occurring between rainforests and savannas [83,84].
The average distance from a rainforest to the five nearest patches ranged from 0.015 to 10.879
km (average 0.659 km ± 0.011 km), and the patch connectivity index ranged from 0.008 to 7327.150
(average 8.887 ± 1.774) (Figure 3). The frequency of baseline fires was correlated with distance from
the coastline (r = 0.65), but not with slope (r = −0.15). It also differed significantly between geology
types (p < 0.001).
Figure 3. Patch connectivity of rain forests in WGC. Patches with high connectivity (dark shade, up
to 7327.150) are larger and/or closer together than patches with low connectivity (light shade, down
to 0.008).
Figure 2.
A map of rainforest patches in WGC. Insets show areas of high patch size and density at the
Bougainville Peninsula and Mitchell Plateau, including transect locations. Note that patch size has
been enhanced to better illustrate patch distribution at map scale.
Rainforest density is strongly dependent on rainfall, as shown by the correlations between
rainforest density and mean annual rainfall in the Australian monsoon tropics and distance from
coastline [
62
]. Rainforests were also preferentially located on relatively nutrient-rich substrates, such as
basalt, and on fire-protected topographic settings (slopes and valleys). These results confirm the
importance of nutrient and water availability, as well as protection from disturbance, as pivotal factors
affecting the dynamics occurring between rainforests and savannas [83,84].
The average distance from a rainforest to the five nearest patches ranged from 0.015 to 10.879 km
(average 0.659 km
±
0.011 km), and the patch connectivity index ranged from 0.008 to 7327.150
(average 8.887
±
1.774) (Figure 3). The frequency of baseline fires was correlated with distance from
the coastline (r= 0.65), but not with slope (r=
−
0.15). It also differed significantly between geology
types (p< 0.001).
Land 2017, 6, 68 8 of 20
Figure 2. A map of rainforest patches in WGC. Insets show areas of high patch size and density at the
Bougainville Peninsula and Mitchell Plateau, including transect locations. Note that patch size has
been enhanced to better illustrate patch distribution at map scale.
Rainforest density is strongly dependent on rainfall, as shown by the correlations between
rainforest density and mean annual rainfall in the Australian monsoon tropics and distance from
coastline [62]. Rainforests were also preferentially located on relatively nutrient-rich substrates, such
as basalt, and on fire-protected topographic settings (slopes and valleys). These results confirm the
importance of nutrient and water availability, as well as protection from disturbance, as pivotal
factors affecting the dynamics occurring between rainforests and savannas [83,84].
The average distance from a rainforest to the five nearest patches ranged from 0.015 to 10.879
km (average 0.659 km ± 0.011 km), and the patch connectivity index ranged from 0.008 to 7327.150
(average 8.887 ± 1.774) (Figure 3). The frequency of baseline fires was correlated with distance from
the coastline (r = 0.65), but not with slope (r = −0.15). It also differed significantly between geology
types (p < 0.001).
Figure 3. Patch connectivity of rain forests in WGC. Patches with high connectivity (dark shade, up
to 7327.150) are larger and/or closer together than patches with low connectivity (light shade, down
to 0.008).
Figure 3.
Patch connectivity of rain forests in WGC. Patches with high connectivity (dark shade,
up to 7327.150) are larger and/or closer together than patches with low connectivity (light shade,
down to 0.008).

Land 2017,6, 68 9 of 20
3.2. Vegetation and Cattle Patterns
Although not designed to describe the totality of floristic diversity of the area, the rainforest study
captured a representative portion of it: we recorded 82 rainforest and 29 savanna tree species on the
Bougainville Peninsula, 71 rainforest and 31 savanna tree species on the Mitchell Plateau. Fifteen of the
rainforest tree species and 13 of the savanna tree species identified were listed as culturally significant
by Karadada et al., 2011 [
21
]. The Bougainville Peninsula contained regionally interesting floristic
elements in both rainforests and savannas. Eucalyptus oligantha Schauer and Xanthostemon psidioides
(Lindl.) Peter G. Wilson & J.T. Waterh, found in the Bougainville Peninsula’s savannas, are uncommon
in Western Australia, and the latter is considered near threatened in the State (Western Australian
Herbarium, 2016). Other species, such as Acacia drepanocarpa subsp. latifolia Pedley and Vachellia ditricha
(Pedley) Kodela, were thought not to grow in North Kimberley [85].
Rainforests did not present substantial structural differences between the Mitchell Plateau and
Bougainville Peninsula, being characterised by high canopy cover, and absence of grasses and savanna
tree species in both locations. Abrupt changes in vegetation were detected across the rainforest–savanna
boundary in the Mitchell Plateau, including differences in canopy cover, grass cover, species richness,
and the proportion of rainforest trees. Conversely, on the Bougainville Peninsula, these changes were
more gradual, and savannas included a high proportion of adult rainforest trees, despite the high grass
cover. Rainforest species were also recorded in the understorey of each plot located on the Bougainville
Peninsula, signalling ongoing expansion [
61
]. Our vegetation assessment provided similar results to
those of the vegetation survey performed in 1987 [60].
Stinking passionflower was present at only 10% of the vegetation transect plots on the Mitchell
Plateau. All of them were close to the rainforest edge; none was inside the rainforest. In contrast,
more than half of the plots on the Bougainville Peninsula included stinking passionflower, which was
common in savannas and on rainforest edges, but never inside rainforest patches. Where present,
stinking passionflower coverage was extensive.
Cattle were not detected on the Bougainville Peninsula, but at the Mitchell Plateau they were
detected in camera-traps and their impacts were evident in transects. Cattle were detected twice
as often in rainforest as savannas and behaviours included foraging, travelling, and resting (Reid,
unpublished), providing the first photographic evidence of the extensive use of rainforests by cattle in
North Kimberley, thereby corroborating the conclusion of earlier observers [
35
]. Rainforests were used
but cattle not only for grazing, but also as shelter from the heat and sun. Decreased seedling density
was associated with elevated cattle density in both rainforests and savannas, negatively affecting tree
species recruitment [
61
]. Elevated cattle density was not correlated with decreased grass cover in
adjacent savanna.
3.3. Rainforest Change
The historical analysis of rainforest patches showed that rainforests had expanded in all locations
since 1949, probably in response to the wetting trend detected in the area [
66
]. However, there were
substantial differences in expansion rates associated with varying levels of disturbance. On the
fire-protected and cattle-free Bougainville Peninsula, rainforest expanded by 69% (for the 1949–2005
time interval), compared with 9% recorded in the Mitchell Plateau. In areas with low disturbance,
rainforests also expanded in topographic settings that do not provide fire protection, such as flat areas,
while this phenomenon was not observed in the disturbance-prone plateau. Rainforest expansion was
more likely to occur close to the edge of already existing big patches, suggesting that small rainforest
patches may be particularly sensitive to local disturbance. These results suggest that fire activity and
grazing are important factors affecting rainforest distribution and, at a landscape scale, they modulate
the effects of climate-driven vegetation changes.
Aerial photographs from 1949 showed two fire scars that are likely to have been lit by Aboriginal
people in the dry season. One fire scar was on the Bougainville Peninsula, in the vicinity of Seaflower
Bay, and corresponds to an area of savanna with a cycad (Cycas basaltica) population that is known

Land 2017,6, 68 10 of 20
to have been accessed as a food source (Louis Karadada pers. comm.; [
26
]) (Figure 4). Another
fire scar was located at the Mitchell Plateau in an area that is known to have been an important
kangaroo hunting ground (Wilfred Goonack, pers. comm.). In 1949, some indigenous people were still
living Traditional lifestyles and many had returned to the bush following wartime bombing raids on
Kalumburu in 1943.
Land 2017, 6, 68 10 of 20
that is known to have been accessed as a food source (Louis Karadada pers. comm.; [26]) (Figure 4).
Another fire scar was located at the Mitchell Plateau in an area that is known to have been an
important kangaroo hunting ground (Wilfred Goonack, pers. comm.). In 1949, some indigenous
people were still living Traditional lifestyles and many had returned to the bush following wartime
bombing raids on Kalumburu in 1943.
Figure 4. Aerial image of Seaflower Bay, the Bougainville Peninsula, June 1949 with a recent fire scar
in a savanna valley in close proximity to a rainforest patch. The savanna valley contains cycad
populations that were harvested as a food source by Aboriginal people.
3.4. Fire Ecology
Rainforest saplings displayed capacity to survive a single mild fire during the experimental
burning. When exposed to low-intensity experimental fire, most rainforest saplings survived the
treatment (81%), although stem survival was lower (63%), probably due to their thin bark. By
comparison, savanna saplings experienced higher whole plant and stem survival (98% and 88%
respectively). These differences were reflected in their recovery strategies; while savanna saplings
expressed both aerial and basal resprouting, rainforest plants were predominantly characterised by
basal resprouting. This negatively affected growth rates: rainforest trees were, on the average, 43%
shorter after one year, while savanna saplings had regained their pre-fire height [47].
The fire history estimated for the baseline years (2000–2009) showed that total fire frequency
around rainforest patches ranged from 0.000 to 0.900 times burnt·year−1 (average 0.295 ± 0.003; Figure
5a), while during the project years total fire frequency ranged from 0.000 to 1.000 times burnt·year−1
(average 0.333 ± 0.004; Figure 5b) (Table S1). Total fire frequency under unmanaged conditions
(baseline years) was correlated with distance from the coastline (r = 0.65), but not with topographic
slope (r = −0.15). It also differed significantly between geology types (p < 0.001), with alluvium and
colluvium experiencing the highest average frequency (0.36 ± 0.03 times burnt·year−1) and coastal
deposits the lowest (0.16 ± 0.02 times burnt·year−1), reflecting contrasting productivity of these
substrate types.
Figure 4.
Aerial image of Seaflower Bay, the Bougainville Peninsula, June 1949 with a recent fire
scar in a savanna valley in close proximity to a rainforest patch. The savanna valley contains cycad
populations that were harvested as a food source by Aboriginal people.
3.4. Fire Ecology
Rainforest saplings displayed capacity to survive a single mild fire during the experimental
burning. When exposed to low-intensity experimental fire, most rainforest saplings survived
the treatment (81%), although stem survival was lower (63%), probably due to their thin bark.
By comparison, savanna saplings experienced higher whole plant and stem survival (98% and 88%
respectively). These differences were reflected in their recovery strategies; while savanna saplings
expressed both aerial and basal resprouting, rainforest plants were predominantly characterised by
basal resprouting. This negatively affected growth rates: rainforest trees were, on the average, 43%
shorter after one year, while savanna saplings had regained their pre-fire height [47].
The fire history estimated for the baseline years (2000–2009) showed that total fire frequency
around rainforest patches ranged from 0.000 to 0.900 times burnt
·
year
−1
(average 0.295
±
0.003;
Figure 5a), while during the project years total fire frequency ranged from 0.000 to 1.000 times
burnt
·
year
−1
(average 0.333
±
0.004; Figure 5b) (Table S1). Total fire frequency under unmanaged
conditions (baseline years) was correlated with distance from the coastline (r= 0.65), but not with
topographic slope (r=
−
0.15). It also differed significantly between geology types (p< 0.001), with
alluvium and colluvium experiencing the highest average frequency (0.36
±
0.03 times burnt
·
year
−1
)

Land 2017,6, 68 11 of 20
and coastal deposits the lowest (0.16
±
0.02 times burnt
·
year
−1
), reflecting contrasting productivity of
these substrate types.
Land 2017, 6, 68 11 of 20
Figure 5. Heat maps showing total fire frequency in proximity of rainforests during (a) the
unmanaged baseline years and (b) the project years.
Prescribed burning, initiated under the Wunambal Gaambera Healthy Country Plan [13] in
2012, resulted in a significant reduction in LDS fires between the baseline years and the project years
(p < 0.001), and a significant increase of EDS and total fire frequency (p < 0.001 for both; Figure 6).
Average frequency of EDS fires, LDS fires, and total frequency for the baseline years and the project
years are reported in Table 1.
Based on our vulnerability index, which considers patch size, distance from neighbouring
patches, and fire frequency, no clustering was detected when considering all rainforest patches.
However, when only the 5% most vulnerable patches were considered, we detected significant
clustering (p < 0.001; Figure 7). The most vulnerable patches experienced a significant reduction in
LDS fires, with an average reduction of 0.32 times burnt·year−1 during the LDS (p < 0.001), while total
fire frequency did not significantly change.
Figure 5.
Heat maps showing total fire frequency in proximity of rainforests during (
a
) the unmanaged
baseline years and (b) the project years.
Prescribed burning, initiated under the Wunambal Gaambera Healthy Country Plan [
13
] in
2012, resulted in a significant reduction in LDS fires between the baseline years and the project years
(p< 0.001), and a significant increase of EDS and total fire frequency (p< 0.001 for both; Figure 6).
Average frequency of EDS fires, LDS fires, and total frequency for the baseline years and the project
years are reported in Table 1.
Based on our vulnerability index, which considers patch size, distance from neighbouring patches,
and fire frequency, no clustering was detected when considering all rainforest patches. However, when
only the 5% most vulnerable patches were considered, we detected significant clustering (p< 0.001;
Figure 7). The most vulnerable patches experienced a significant reduction in LDS fires, with an
average reduction of 0.32 times burnt
·
year
−1
during the LDS (p< 0.001), while total fire frequency did
not significantly change.

Land 2017,6, 68 12 of 20
Land 2017, 6, 68 12 of 20
Figure 6. Heat maps displaying variations in the frequency of early dry season (EDS) fires and late
dry season (LDS) fires in proximity of rainforests between the baseline years and (2000–2009) and the
project years (2012–2015). Note the heat maps reflect change in fire frequency not actual fire
frequency as per Figure 5.
Figure 7. Location and density of the 5% most vulnerable rainforest patches, based on their size,
distance from neighbouring patches, and fire frequency. The density map (red gradient) shows some
clear clusters of vulnerable patches. Black dots represent single patches.
Figure 6.
Heat maps displaying variations in the frequency of early dry season (EDS) fires and late
dry season (LDS) fires in proximity of rainforests between the baseline years and (2000–2009) and the
project years (2012–2015). Note the heat maps reflect change in fire frequency not actual fire frequency
as per Figure 5.
Land 2017, 6, 68 12 of 20
Figure 6. Heat maps displaying variations in the frequency of early dry season (EDS) fires and late
dry season (LDS) fires in proximity of rainforests between the baseline years and (2000–2009) and the
project years (2012–2015). Note the heat maps reflect change in fire frequency not actual fire
frequency as per Figure 5.
Figure 7. Location and density of the 5% most vulnerable rainforest patches, based on their size,
distance from neighbouring patches, and fire frequency. The density map (red gradient) shows some
clear clusters of vulnerable patches. Black dots represent single patches.
Figure 7.
Location and density of the 5% most vulnerable rainforest patches, based on their size,
distance from neighbouring patches, and fire frequency. The density map (red gradient) shows some
clear clusters of vulnerable patches. Black dots represent single patches.

Land 2017,6, 68 13 of 20
4. Discussion
This collaborative research project between indigenous land managers and a research institution
has been able to fill key ecological knowledge gaps and develop resources and methods for the
long-term management and monitoring of ‘Wulo’ or monsoon vine thickets in WGC.
Rainforest patches have now been effectively mapped and quantified, providing a comprehensive
picture of their distribution. Mapping has confirmed that, while rainforest occur in a wide range
of geologic and topographic settings, they are concentrated in higher rainfall areas and on richer
geographic substrates such as basalt. Patches are typically small (75% are less than 1 ha) and embedded
in a flammable savanna matrix with some exceptions. The Bougainville Peninsula holds the largest
single rainforest patch, 220 ha in size, as well as 6 other patches over 100 ha in size. Its patches also
have a high level of connectivity. Large patches ranging between 10 and 100 ha were also located at
key locations including Lawley River National Park, Crystal Head, Hunter River, the Osborn Island
group, the Institute Island group, and a few inland sites.
Rainforests occur in clusters of higher density in some locations, such as the Bougainville
Peninsula and the Lawley River National Park, while other patches are significantly isolated (such as
in inland areas). The fragmentary distribution of rainforests (including those on offshore islands) has
facilitated speciation of low-dispersal taxa such as landsnails (a value which has been recognized on
the National Heritage Listing of the West Kimberley) [
69
,
86
]. Conversely, volant taxa like fruit-eating
birds and bats are dependent on the network of rainforest patches (patches are typically small and
embedded in a flammable savanna matrix) [
87
–
89
]. In this context, there is a need to manage the entire
network of rainforest patches to prevent individual patches from contracting below the minimum
size needed to support the animal species that currently inhabit them [
90
], with particular attention
to small, isolated rainforests. This is now possible given our detailed mapping of all the rainforests
>0.1 ha in extent.
This study found a 10–60% expansion in rainforest boundaries in the last 50 years despite the
rise of severe wildfire events during that time. The expansion matches similar trends in Northern
Australia attributed to regional wetting trends and perhaps atmospheric CO
2
enrichment over the
same period. Climate models predict significant potential for ecological change in Australia’s northern
savannas [
91
] including a high turnover of plant species by 2050 [
92
]. Long-term monitoring can help
to determine any change in the number and extent of rainforest patches in response to climate change.
Stinking passionflower was the only weed species recorded in the patches, but other weed species
in the region have the potential to change fire regimes and damage both rainforests and savannas [
52
].
Of concern are large grass species already present in the region that can substantially increase fire
severity, with associated savanna and rainforest tree mortality that potentially cause rainforest to
retreat. Exotic vines, such as stinking passionflower, which is now naturalized in North Kimberley,
have the potential to smother and kill rainforest species [
93
]. The continuation and evaluation of
existing weed management programs is essential to the long-term management of healthy rainforest.
In particular, it is important that grass weeds are contained on roadsides and campgrounds and do not
establish in the wider landscape.
Large, long unburnt areas can contain relatively high and continuous fuel loads of grass, litter
and woody debris, which presents a risk of extensive and potentially destructive fires late dry season
wildfire ignited by lightning, as occurred on Bigge Island in 2012 and Middle Osborn Island in 2015 [
71
].
Nonetheless, in some areas, rainforests have expanded over the past 50 years, and this was most
pronounced in areas with low fire frequency such as the Bougainville Peninsula and offshore islands,
despite the abundance of grasses found in the adjacent savannas [
61
]. Small patches were less likely to
expand compared with bigger patches, possibly due to their higher sensitivity to intense fires due to
higher perimeter to core ratio [
52
]. By contrast, patches exposed to frequent fire activity, such as at
the Mitchell Plateau, were found to have stable boundaries, probably because they are restricted to
topographic settings such as slopes and valleys, which, compared with flat locations, provide a higher
abundance of nutrients and water as well as fire protection [61,94].

Land 2017,6, 68 14 of 20
No rainforest patches had contracted or disappeared since 1949, despite having experienced
a prolonged period without fire management during which wildfire events were more likely to
be severe and large in extent. This suggests that rainforest patches are relatively resilient to fire,
somewhat in contrast to other rainforest types such as the Anbinik (Allosyncarpia ternata) rainforests
on sandstone-derived substrates in Arnhem Land, which are sensitive to wildfire events [
95
–
97
].
Edwards and Russell-Smith, in 2009 [
37
], identified a critical ecological threshold when >10% of patch
boundaries are affected by one or more LDS fires. Accordingly, Warddeken rangers undertake an
annual program of protecting 50 ‘at risk’ sites with mineral earth breaks and fire breaks [
98
]. Evidence
presented here suggests that most Wulo rainforests do not appear to have the same sensitivity to fire
and do not require such intensive management of individual patches as Anbinik but could rather be
managed through establishing and maintaining healthy fire regimes in the adjacent savanna matrix
in areas characterized by high fire frequency. However, there may be some benefit in applying more
direct management to isolated inland patches and vulnerable patches such as those in the Lawley
River NP, which are exposed to frequent wildfire.
Rainforest boundaries can be impacted by single severe wildfire events (through inadequate fire
management) as well as recurrent frequent management fires. Using a combination of aerial and
ground-based ignitions Uunguu Rangers have caused a shift from LDS fires to EDS fires in the savanna
matrix adjacent to rainforest patches. This approach is consistent with traditional Aboriginal fire
management [
20
] that limited wildfires that could cause damage to rainforests. Prior to colonization,
Traditional Owners burned some savanna areas on the peninsula to facilitate walking, hunting and
gathering. In particular, they burnt the open woodland on the laterite plateau and some of the larger
savanna areas in the valleys and slopes, including the areas with cycads (Cycas basaltica), a traditional
food resource and important endemic species (Figure 8). Unburnt areas may disadvantage some
plant species such as the cycad population on the Bougainville Peninsula. Traditional Owners, and
Aboriginal Rangers involved in fieldwork during this study, suggest that savanna areas that become
overgrown with vine thicket elements such as ‘arndarn’ or thorny vine species (including Capparis
spp., Protasparagus racemosa, and Smilax australis) should be burned occasionally to ensure that they do
not become unhealthy and impenetrable to people. Clearly, there is a need to carefully balance the
health of the substantial rainforest patches and the health of the adjacent savanna matrix.
Land 2017, 6, 68 14 of 20
No rainforest patches had contracted or disappeared since 1949, despite having experienced a
prolonged period without fire management during which wildfire events were more likely to be
severe and large in extent. This suggests that rainforest patches are relatively resilient to fire,
somewhat in contrast to other rainforest types such as the Anbinik (Allosyncarpia ternata) rainforests
on sandstone-derived substrates in Arnhem Land, which are sensitive to wildfire events [95–97].
Edwards and Russell-Smith, in 2009 [37], identified a critical ecological threshold when >10% of
patch boundaries are affected by one or more LDS fires. Accordingly, Warddeken rangers undertake
an annual program of protecting 50 ‘at risk’ sites with mineral earth breaks and fire breaks [98].
Evidence presented here suggests that most Wulo rainforests do not appear to have the same
sensitivity to fire and do not require such intensive management of individual patches as Anbinik
but could rather be managed through establishing and maintaining healthy fire regimes in the
adjacent savanna matrix in areas characterized by high fire frequency. However, there may be some
benefit in applying more direct management to isolated inland patches and vulnerable patches such
as those in the Lawley River NP, which are exposed to frequent wildfire.
Rainforest boundaries can be impacted by single severe wildfire events (through inadequate
fire management) as well as recurrent frequent management fires. Using a combination of aerial and
ground-based ignitions Uunguu Rangers have caused a shift from LDS fires to EDS fires in the
savanna matrix adjacent to rainforest patches. This approach is consistent with traditional
Aboriginal fire management [20] that limited wildfires that could cause damage to rainforests. Prior
to colonization, Traditional Owners burned some savanna areas on the peninsula to facilitate
walking, hunting and gathering. In particular, they burnt the open woodland on the laterite plateau
and some of the larger savanna areas in the valleys and slopes, including the areas with cycads
(Cycas basaltica), a traditional food resource and important endemic species (Figure 8). Unburnt areas
may disadvantage some plant species such as the cycad population on the Bougainville Peninsula.
Traditional Owners, and Aboriginal Rangers involved in fieldwork during this study, suggest that
savanna areas that become overgrown with vine thicket elements such as ‘arndarn’ or thorny vine
species (including Capparis spp., Protasparagus racemosa, and Smilax australis) should be burned
occasionally to ensure that they do not become unhealthy and impenetrable to people. Clearly, there
is a need to carefully balance the health of the substantial rainforest patches and the health of the
adjacent savanna matrix.
(a) (b)
Figure 8. (a) A rainforest patch in a fire protected gully surrounded by open savanna on the plateau
of the Bougainville Peninsula, and (b) a cycad growing amidst senescent long unburnt grass and
invasive Stinking Passionflower vines on the Bougainville Peninsula.
Although cattle have not yet caused rainforest to retreat they may represent a threat for the
long-term persistence of some rainforests in the WGC, as found for other locations in the Australian
tropics [52]. The local impact of feral cattle on rainforest patches is not easily detectable by remote
sensing, but evidence suggests that cattle are damaging the understorey of rainforest patches. The
high number of cattle detections in rainforests, combined with the reduced number of seedlings
Figure 8.
(
a
) A rainforest patch in a fire protected gully surrounded by open savanna on the plateau of
the Bougainville Peninsula, and (
b
) a cycad growing amidst senescent long unburnt grass and invasive
Stinking Passionflower vines on the Bougainville Peninsula.
Although cattle have not yet caused rainforest to retreat they may represent a threat for the
long-term persistence of some rainforests in the WGC, as found for other locations in the Australian
tropics [
52
]. The local impact of feral cattle on rainforest patches is not easily detectable by remote

Land 2017,6, 68 15 of 20
sensing, but evidence suggests that cattle are damaging the understorey of rainforest patches. The high
number of cattle detections in rainforests, combined with the reduced number of seedlings associated
with their presence, is consistent with the observations during early surveys [
35
], that cattle degraded
patches in the Kimberley through trampling and browsing by cattle. Cattle are currently an important
source of food for Aboriginal People in the WGC [
21
], as in other Aboriginal communities in Northern
Australia [
99
]. Traditional Owners are currently developing programs that balance their food needs
with conservation objectives, and have undertaken some aerial culling with partner agencies. A fence
is also maintained on the narrow neck of the Bougainville Peninsula to keep the area free of cattle and
smaller cultural sites have also been fenced off from cattle. Management will likely focus on ensuring
that some areas remain cattle-free and that numbers are kept at sustainable levels in high conservation
areas such as where there are important rainforest patches. Ongoing monitoring of the structure and
floristics of rainforest patches in areas used by cattle is needed to determine the ecological thresholds
associated with cattle numbers.
A large proportion of rainforest patches are associated with bauxite reserves that have been the
focus of mining exploration in the past. Bauxite is typically extracted by strip mining and requires the
clearing of vegetation. Recent legislation permanently removed mining interests from the Mitchell
Plateau, but the Bougainville Peninsula has an extant mining lease and remains at risk from mining.
Under state legislation, national park tenure provides a degree of protection from mining but requires
negotiation between industry, government, and Traditional Owners to achieve.
The monsoon vine thickets of the Dampier Peninsula in Southwest Kimberley were recently
recognized as a threatened ecological community under the Environmental Protection and Biodiversity
Conservation Act 1999. Given the threats of weeds, cattle, wildfire, and mining, a similar listing of the
monsoon vine thickets of North Kimberley may also be justified.
5. Conclusions
We report the findings of a study that sought to provide baseline data on the distribution and
health of monsoon rainforests, or Wulo, on the WGC. Using historical and contemporary remote
sensing imagery combined with fieldwork, and collaborating with Aboriginal rangers and Traditional
Owners, we were able to significantly improve our understanding of the value, health, and threats
to this regionally restrictive and biogeographically significant community. Of prime importance is
the ongoing management of fire, feral cattle, invasive weeds, and potential mining across the WGC
lands. Particular attention, both in terms of land management and legislation, needs to be paid to the
Bougainville Peninsula as it has the dual distinction of being one of the largest areas of rainforest in
the Australian monsoon tropics but also free from the threat of feral cattle and prevailing wildfires.
This study also provides further proof that collaborative research partnerships can provide substantial
benefits to Traditional Owners responsible for managing remote high conservation lands and seas.
Supplementary Materials:
The following are available online at www.mdpi.com/2073-445X/6/4/68/s1, Table S1:
Fire Activity and Connectivity Values for Each Rainforest Patch Considered in this Study.
Acknowledgments:
This work was supported by funds from the Australian Government’s Biodiversity Fund.
The project was covered by a research agreement between the University of Tasmania and Wunambal Gaambera
Aboriginal Corporation. We would like to acknowledge the following people who assisted with field work
and logistics: Phil Docherty, Deborah Apgaua, Carly Ward, Ernie Boona, Robert Warren, Bevan Stott, and the
Uunguu Rangers. Members of the Uunguu Monitoring & Evaluation Committee provided expert advice on the
project including Norm McKenzie, Andrew Burbidge, Kim Doohan, Rod Kennett, Stuart Cowell, Neil Waina and
Sylvester Mangolamara. David M. J. S. Bowman, Linda Prior, Grant Williamson and Tom Vigilante supervised
Stefania Ondei’s doctoral thesis as part of this project.
Author Contributions:
The research was conceived, designed and implemented as a collaborative project
between WGAC and UTAS. Tom Vigilante and Stefania Ondei drafted this paper with supervision from
David M. J. S. Bowman and GIS advice from Paul Young. The paper utilized work undertaken by Stefania Ondei
as part of a doctoral thesis as well as additional analyses. Catherine Goonack and Desmond Williams advised and
collaborated on field work and Traditional Ecological Knowledge.
Conflicts of Interest: The authors declare no conflict of interest.

Land 2017,6, 68 16 of 20
References
1.
Kothari, A. Recognising and Supporting Territories and Areas Conserved by Indigenous Peoples and Local
Communities: Global Overview and National Case Studies; Technical Series No. 64; Secretariat of the Convention
on Biological Diversity, ICCA Consortium, Kalpavriksh, and Natural Justice: Montreal, QC, Canada, 2012.
2.
Berkes, F.; Reid, W.V.; Wilbanks, T.J.; Capistrano, D. Bridging scales and knowledge systems. In Bridging
Scales and Knowledge Systems: Concepts and Applications in Ecosystem Assessment; Reid, W.V., Berkes, F.,
Wilbanks, T., Capistrano, D., Eds.; Island Press: Washington, DC, USA, 2006; Volume 315.
3.
Gadgil, M.; Olsson, P.; Berkes, F.; Folke, C. Exploring the role of local ecological knowledge in ecosystem
management: Three case studies. In Navigating Social-Ecological Systems: Building Resilience for Complexity
and Change; Berkes, F., Colding, J., Folke, C., Eds.; Cambridge University Press: Cambridge, UK, 2003;
pp. 189–209.
4.
Prado, H.M.; Murrieta, R.S.S.; Adams, C.; Brondizio, E.S. Complementary viewpoints: Scientific and local
knowledge of ungulates in the brazilian atlantic forest. J. Ethnobiol. 2013,33, 180–202. [CrossRef]
5.
Bohensky, E.L.; Butler, J.R.; Davies, J. Integrating indigenous ecological knowledge and science in natural
resource management: Perspectives from Australia. Ecol. Soc. 2013, 18. [CrossRef]
6.
Australian Government. Ownership of Indigenous Protected Areas. Available online: https://www.
environment.gov.au/land/nrs/about-nrs/ownership (accessed on 7 June 2016).
7.
Altman, J.C.; Whitehead, P.J. Caring for Country and Sustainable Indigenous Development: Opportunities,
Constraints and Innovation; Centre for Aboriginal Economic Policy Research, Australian National University:
Canberra, Australia, 2003.
8. Davies, J.; Hill, R.; Walsh, F.J.; Sandford, M.; Smyth, D.; Holmes, M.C. Innovation in management plans for
community conserved areas: Experiences from australian indigenous protected areas. Ecol. Soc.
2013
, 18.
[CrossRef]
9.
Participatory Monitoring and Management Partnership. Available online: http://www.pmmpartnership.
com/ (accessed on 7 June 2016).
10.
Constantino, P.; Carlos, H.S.A.; Ramalho, E.E.; Rostant, L.; Marinelli, C.E.; Teles, D.; Fonseca-Junior, S.F.;
Fernandes, R.B.; Valsecchi, J. Empowering local people through community-based resource monitoring:
A comparison of brazil and namibia. Ecol. Soc. 2012, 17. [CrossRef]
11.
Luzar, J.B.; Silvius, K.M.; Overman, H.; Giery, S.T.; Read, J.M.; Fragoso, J.M.V. Large-scale environmental
monitoring by indigenous peoples. BioScience 2011,61, 771–781. [CrossRef]
12.
Hill, R.; Walsh, F.; Davies, J.; Sandford, M. Our Country Our Way: Guidelines for Australian Indigenous Protected
Area Management Plans; CSIRO Ecosystem Sciences and Department of Sustainability, Water, Environment,
Population and Communities: Cairns, Australia, 2011.
13.
Wunambal Gaambera Aboriginal Corporation. Wunambal Gaambera Healthy Country Plan—Looking after
Wunambal Gaambera Country 2010–2020; Wunambal Gaambera Aboriginal Corporation: Wyndham,
Australia, 2010.
14.
Austin, B.J.; Vigilante, T.; Cowell, S.; Dutton, I.M.; Djanghara, D.; Mangolamara, S.; Puermora, B.;
Bundamurra, A.; Clement, Z. The uunguu monitoring and evaluation committee: Intercultural governance
of a land and sea management programme in the kimberley, Australia. Ecol. Manag. Restor.
2017
,18, 124–133.
[CrossRef]
15.
Moorcroft, H. Wunambal gaambera healthy country plan. In Innovations for 21st Century Conservation;
Figgis, P., Fitzsimons, J., Irving, J., Eds.; Australian Committee for IUCN: Sydney, Australia, 2012; pp. 116–123.
16.
Moorcroft, H.; Ignjic, E.; Cowell, S.; Goonack, J.; Mangolomara, S.; Oobagooma, J.; Karadada, R.; Williams, D.;
Waina, N. Conservation planning in a cross-cultural context: The wunambal gaambera healthy country
project in the kimberley, Western Australia. Ecol. Manag. Restor. 2012,13, 16–25. [CrossRef]
17.
Carr, B.; Fitzsimons, J.; Holland, N.; Berkinshaw, T.; Bradby, K.; Cowell, S.; Deegan, P.; Kock, P.; Looker, M.;
Varcoe, T.; et al. Capitalising on conservation knowledge: Using conservation action planning, healthy
country planning and the open standards in Australia. Ecol. Manag. Restor. 2017,18, 1–14. [CrossRef]
18.
The Nature Conservancy (TNC). Action Planning. Available online: https://www.conservationgatewayorg/
ConservationPlanning/ActionPlanning/Pages/conservation-action-plann.aspx (accessed on 13 May 2015).

Land 2017,6, 68 17 of 20
19.
Jackson, M.V.; Kennett, R.; Bayliss, P.; Warren, R.; Waina, N.; Adams, J.; Cheinmora, L.; Vigilante, T.;
Jungine, E.; Woolagoodja, K. Developing collaborative marine turtle monitoring in the kimberley region of
Northern Australia. Ecol. Manag. Restor. 2015,16, 163–176. [CrossRef]
20.
Mangglamarra, G.; Burbidge, A.A.; Fuller, P.J. Wunambal words for rainforest and other kimberley plants
and animals. In Kimberley Rainforest Australia; McKenzie, N.L., Johnston, R.B., Kendrick, P.G., Eds.; Surrey
Beatty and Sons: Canberra, Australia, 1991; pp. 413–421.
21.
Karadada, J.; Karadada, L.; Goonack, W.; Mangolamara, G.; Bunjuck, W.; Djanghara, B.; Mangolamara, S.;
Oobagooma, J.; Charles, A. Uunguu Plants and Animals: Aboriginal Biological Knowledge from Wunambal
Gaambera Country in the North-West Kimberley, Australia; Wunambal Gaambera Aboriginal Corporation:
Wyndham, Australia, 2011.
22.
Russell-Smith, J. Classification, species richness, and environmental relations of monsoon rain forest in
Northern Australia. J. Veg. Sci. 1991,2, 259–278. [CrossRef]
23. Webb, L.J. A physiognomic classification of Australian rain forests. J. Ecol. 1959,47, 551–570. [CrossRef]
24.
McKenzie, N.L. An ecological survey of tropical rainforests in Western Australia: Background and methods.
In Kimberley Rainforests of Australia; McKenzie, N.L., Johnston, R.B., Kendrick, P.G., Eds.; Surrey Beatty and
Sons: Chipping Norton, UK, 1991; pp. 1–26.
25.
Beard, J.S. The monsoon forest of the admirality gulf, Western Australia. Vegetatio
1976
,31, 177–192.
[CrossRef]
26.
Beard, J.S.; Clyton-Greene, K.A.; Kenneally, K.F. Notes on the vegetation of the bougainville peninsula,
osborn and institut islands, north kimberley district, Western Australia. Vegetatio
1984
,57, 3–13. [CrossRef]
27.
Miles, J.M.; Burbidge, A.A. A Biological survey of the Prince Regent River Reserve, North-West Kimberley, Western
Australia, in August, 1974; Department of Fisheries and Wildlife, Western Australia: Perth, Australia, 1975.
28.
Wilson, B. Biological Survey of Mitchell Plateau and Admiralty Gulf, Kimberley, Western Australia; Western
Australian Museum: Perth, Australia, 1981.
29.
Burbidge, A.A.; McKenzie, N.L. Islands of the North-West Kimberley, Western Australia; Department of Fisheries
and Wildlife: Perth, Australia, 1978; Volume 7, pp. 1–47.
30.
Bowman, D. Monsoon forests in North-Western Australia. II. Forest-savanna transitions. Aust. J. Bot.
1992
,
40, 89–102. [CrossRef]
31.
McKenzie, N.; Start, A.; Burbidge, A.; Kenneally, K.; Burrows, N. Protecting the Kimberley: A Synthesis of
Scientific Knowledge to Support Conservation Management in the Kimberley Region of Western Australia. Part B:
Terrestrial Environments; Department of Environment and Conservation: Perth, Australia, 2009.
32.
Trauernicht, C.; Brook, B.W.; Murphy, B.P.; Williamson, G.J.; Bowman, D.M.J.S. Local and global
pyrogeographic evidence that indigenous fire management creates pyrodiversity. Ecol. Evol.
2015
,5,
1908–1918. [CrossRef] [PubMed]
33.
Williams, R.J.; Cook, G.D.; Gill, A.M.; Moore, P.H.R. Fire regime, fire intensity and tree survival in a tropical
savanna in Northern Australia. Austral Ecol. 1999,24, 50–59. [CrossRef]
34.
Legge, S.; Kingswood, R.; Swan, D.; Murphy, S.; Maher, B. Ecofire: Restoring the biodiversity values of the
kimberley region by managing fire. Ecol. Manag. Restor. 2011,12, 84–92. [CrossRef]
35.
McKenzie, N.L.; Belbin, L. Kimberley rainforest communities: Reserve recommendations and management
considerations. In Kimberley Rainforests of Australia; McKenzie, N.L., Johnston, R.B., Kendrick, P.G., Eds.;
Surrey Beatty and Sons: Chipping Norton, UK, 1991; pp. 453–468.
36.
Russell-Smith, J.; Edwards, A.C.; Price, O.F. Simplifying the savanna: The trajectory of fire-sensitive
vegetation mosaics in Northern Australia. J. Biogeogr. 2012,39, 1303–1317. [CrossRef]
37.
Edwards, A.C.; Russell-Smith, J. Ecological thresholds and the status of fire-sensitive vegetation in western
arnhem land, Northern Australia: Implications for management. Int. J. Wildland Fire
2009
,18, 127–146.
[CrossRef]
38.
Banfai, D.S.; Bowman, D.M.J.S. Forty years of lowland monsoon rainforest expansion in kakadu national
park, Northern Australia. Biol. Conserv. 2006,131, 553–565. [CrossRef]
39.
Bowman, D.M.J.S.; Walsh, A.; Milne, D.J. Forest expansion and grassland contraction within a eucalyptus
savanna matrix between 1941 and 1994 at litchfield national park in the Australian monsoon tropics.
Glob. Ecol. Biogeogr. 2001,10, 535–548. [CrossRef]
40.
Bowman, D.; Murphy, B.P.; Banfai, D.S. Has global environmental change caused monsoon rainforests to
expand in the Australian monsoon tropics? Landsc. Ecol. 2010,25, 1247–1260. [CrossRef]

Land 2017,6, 68 18 of 20
41.
Banfai, D.S.; Bowman, D.M.J.S. Drivers of rain-forest boundary dynamics in kakadu national park, Northern
Australia: A field assessment. J. Trop. Ecol. 2007,23, 73–86. [CrossRef]
42.
Buitenwerf, R.; Bond, W.J.; Stevens, N.; Trollope, W.S.W. Increased tree densities in south african savannas:
>50 years of data suggests CO2as a driver. Glob. Chang. Biol. 2012,18, 675–684. [CrossRef]
43.
Bond, W.J.; Midgley, G.F. A proposed CO
2
-controlled mechanism of woody plant invasion in grasslands and
savannas. Glob. Chang. Biol. 2000,6, 865–869. [CrossRef]
44.
Earl, N.; Simmonds, I. Variability, trends, and drivers of regional fluctuations in Australian fire activity.
J. Geophys. Res. Atmos. 2017,122, 7445–7460. [CrossRef]
45.
Clarke, P.J.; Lawes, M.J.; Midgley, J.J.; Lamont, B.B.; Ojeda, F.; Burrows, G.E.; Enright, N.J.; Knox, K.J.E.
Resprouting as a key functional trait: How buds, protection and resources drive persistence after fire.
New Phytol. 2013,197, 19–35. [CrossRef] [PubMed]
46.
Lawes, M.J.; Midgley, J.J.; Clarke, P.J. Costs and benefits of relative bark thickness in relation to fire damage:
A savanna/forest contrast. J. Ecol. 2013,101, 517–524. [CrossRef]
47.
Ondei, S.; Prior, L.D.; Vigilante, T.; Bowman, D.M.J.S. Post-fire resprouting strategies of rainforest and
savanna saplings along the rainforest–savanna boundary in the Australian monsoon tropics. Plant Ecol.
2016
,
217, 711–724. [CrossRef]
48.
Johnstone, R.B.; Burbidge, A.H. The avifauna of kimberley rainforests. In Kimberley Rainforests of Australia;
McKenzie, N.L., Johnston, R.B., Kendrick, P.G., Eds.; Surrey Beatty and Sons: Chipping Norton, UK, 1991;
pp. 361–391.
49.
Bowman, D.; Woinarski, J.; Russellsmith, J. Environmental relationships of orange-footed scrubfowl
megapodius reinwardt nests in the northern territory. Emu Austral Ornithol. 1994,94, 181–185.
50.
Bowman, D.M.J.S.; Panton, W.J.; Head, J. Abandoned orange-footed scrubfowl (megapodius reinwardt)
nests and coastal rainforest boundary dynamics during the late holocene in monsoonal Australia. Quat. Int.
1999,59, 27–38. [CrossRef]
51.
Werner, P.A. Impact of feral water buffalo and fire on growth and survival of mature savanna trees: An
experimental field study in kakadu national park, Northern Australia. Austral Ecol.
2005
,30, 625–647.
[CrossRef]
52.
Russell-Smith, J.; Bowman, D.M.J.S. Conservation of monsoon rainforest isolates in the Northern Territory,
Australia. Biol. Conserv. 1992,59, 51–63. [CrossRef]
53.
Camargo-Sanabria, A.A.; Mendoza, E.; Guevara, R.; Martínez-Ramos, M.; Dirzo, R. Experimental
Defaunation of Terrestrial Mammalian Herbivores Alters Tropical Rainforest Understorey Diversity. Proc. R.
Soc. Lond. B Biol. Sci. 2014. [CrossRef] [PubMed]
54.
Petty, A.M.; Werner, P.A.; Lehmann, C.E.R.; Riley, J.E.; Banfai, D.S.; Elliott, L.P. Savanna responses to feral
buffalo in Kakadu National Park, Australia. Ecol. Monogr. 2007,77, 441–463. [CrossRef]
55.
Friend, G.R.; Morris, K.D.; McKenzie, N.L. The mammal fauna of kimberley rainforests. In Kimberley
Rainforests of Australia; McKenzie, N.L., Johnson, C.N., Kendrick, P.G., Eds.; Surrey Beatty and Sons:
Chipping Norton, UK, 1991; pp. 393–412.
56.
Setterfield, S.A.; Rossiter-Rachor, N.A.; Hutley, L.B.; Douglas, M.M.; Williams, R.J. Turning up the heat:
The impacts of andropogon gayanus (gamba grass) invasion on fire behaviour in Northern Australian Savannas.
Divers. Distrib. 2010,16, 854–861. [CrossRef]
57.
Government, A. Threat Abatement Plan to Reduce the Impacts on Northern Australia’s Biodiversity by the Five
Listed Grasses; Department of Sustainability, Environment, Water, Population and Communities: Canberra,
Australia, 2012.
58.
Carwardine, J.; O’Connor, T.; Legge, S.; Mackey, B.; Possingham, H.; Martin, T. Priority Threat Management to
Protect Kimberley Wildlife; 0643103066; CSIRO Ecosystem Sciences Brisbane: Canberra, Australia, 2011.
59.
McKenzie, N.; May, J.; Australia, W. A Biodiversity Audit of Western Australia’s Biogeographical Subregions in
2002; 0730755347; Department of Conservation & Land Management: Canberra, Australia, 2002.
60.
Kenneally, K.F.; Keighery, G.J.; Hyland, B.P.M. Floristics and phytogeography of kimberley rainforests.
In Kimberley Rainforests of Australia; McKenzie, N.L., Johnston, R.B., Kendrick, P.G., Eds.; Surrey Beatty and
Sons: Chipping Norton, UK, 1991; pp. 93–131.
61.
Ondei, S.; Prior, L.D.; Vigilante, T.; Bowman, D.M.J.S. Fire and cattle disturbance affects vegetation structure
and rain forest expansion into savanna in the Australian monsoon tropics. J. Biogeogr. 2017. [CrossRef]

Land 2017,6, 68 19 of 20
62.
Ondei, S.; Prior, L.D.; Williamson, G.J.; Vigilante, T.; Bowman, D.M.J.S. Water, land, fire, and forest:
Multi-scale determinants of rainforests in the Australian Monsoon tropics. Ecol. Evol.
2017
,7, 1592–1604.
[CrossRef] [PubMed]
63.
Bowman, D.M.J.S. Australian Rainforests: Islands of Green in a Land of Fire; Cambridge University Press:
Cambridge, UK, 2000.
64.
Government, A. Australia’s 15 National Biodiversity Hotspots. Available online: http://www.environment.
gov.au/biodiversity/conservation/hotspots/national-biodiversity-hotspots#hotspot15 (accessed on
2 October 2017).
65.
Beard, J.S. Vegetation Survey of Western Australia: Kimberley 1:1 000 000 Vegetation Series Sheet 1; University of
Western Australia Press: Perth, Australia, 1979.
66. Bureau of Meteorology. Available online: http://www.bom.gov.au (accessed on 2 October 2017).
67.
Thackway, R.; Cresswell, I.D. An Interim Biogeographic Regionalisation for Australia: A Framework for Setting
Priorities in the National Reserves System Cooperative; Australian Nature Conservation Agency, Reserve Systems
Unit: Canberra, Australia, 1995.
68.
Bowman, D.M.J.S.; Woinarski, J.C.Z. Biogeography of Australian monsoon rainforest mammals: Implications
for the conservation of rainforest mammals. Pac. Conserv. Biol. 1994,1, 98–106. [CrossRef]
69.
Solem, A.; McKenzie, N.L. The composition of land snail assemblages in kimberley rainforests. In Kimberley
Rainforests of Australia; McKenzie, N.L., Johnston, R.B., Kendrick, P.G., Eds.; Surrey Beatty and Sons:
Chipping Norton, UK, 1991; pp. 247–263.
70.
Government, A. National Heritage Places—West Kimberley. Available online: http://www.environment.
gov.au/heritage/places/national/west-kimberley (accessed on 18 July 2017).
71.
NAFI. North Australian Fire Information. Available online: www.firenorth.org.au (accessed on
1 March 2016).
72.
Clayton-Greene, K.; Beard, J. The fire factor in vine thicket and woodland vegetation of the Admiralty Gulf
region, north-west Kimberley, Western Australia. In Proceedings of the Ecological Society of Australia,
Carlton, Australia, 11–16 August 1985; pp. 225–230.
73.
Smith, A.J.; Scherrer, P.; Dowling, R. Impacts on aboriginal spirituality and culture from tourism in the
coastal waterways of the Kimberley region, North West Australia. J. Ecotour. 2009,8, 82–98. [CrossRef]
74.
Vigilante, T. The Ethnoecology of Landscape Burning around Kalumburu Aboriginal Community, North
Kimberley Region, Western Australia: An Examination of the Ecological and Cultural Significance of
Aboriginal Landscape Burning in the North Kimberley Using Experimental, Ethnographic and Historical
Approaches. Ph.D. Thesis, Northern Territory University, Darwin, Australia, 2004.
75.
Vigilante, T.; Bowman, D.M.J.S.; Fisher, R.; Russel-Smith, J.; Yates, C. Contemporary landscape burning
patterns in the far north Kimberley region of north-west Australia: Human influences and environmental
determinants. J. Biogeogr. 2004,31, 1317–1333. [CrossRef]
76.
Fisher, R.; Vigilante, T.; Yates, C.; Russel-Smith, J. Patterns of landscape fire and predicted vegetation response
in the north Kimberley region of Western Australia. Int. J. Wildland Fire 2003,12, 369–379. [CrossRef]
77.
Munguía-Rosas, M.A.; Montiel, S. Patch size and isolation predict plant species density in a naturally
fragmented forest. PLoS ONE 2014, 9. [CrossRef] [PubMed]
78.
Roy, D.P.; Boschetti, L.; Justice, C.O.; Ju, J. The collection 5 MODIS burned area product—global evaluation
by comparison with the MODIS active fire product. Remote Sens. Environ. 2008,112, 3690–3707. [CrossRef]
79.
He, L.; Li, Z. Enhancement of a fire detection algorithm by eliminating solar reflection in the mid-IR band:
Application to AVHRR data. Int. J. Remote Sens. 2012,33, 7047–7059. [CrossRef]
80.
Schroeder, W.; Prins, E.; Giglio, L.; Csiszar, I.; Schmidt, C.; Morisette, J.T.; Morton, D. Validation of GOES
and MODIS active fire detection products using ASTER and ETM + data. Remote Sens. Environ.
2008
,112,
2711–2726. [CrossRef]
81.
Giglio, L.; Schroeder, W.; Justice, C.O. The collection 6 MODIS active fire detection algorithm and fire
products. Remote Sens. Environ. 2016,178, 31–41. [CrossRef]
82.
Hunt, G. Carbon Credits (Carbon Farming Initiative-Emissions Abatement through Savanna Fire Management)
Methodology Determination 2015; Department of Environment and Energy: Canberra, Austrlia, 2015.
83.
Murphy, B.P.; Bowman, D.M.J.S. What controls the distribution of tropical forest and savanna? Ecol. Lett.
2012,15, 748–758. [CrossRef] [PubMed]

Land 2017,6, 68 20 of 20
84.
Hoffmann, W.A.; Geiger, E.L.; Gotsch, S.G.; Rossatto, D.R.; Silva, L.C.R.; Lau, O.L.; Haridasan, M.;
Franco, A.C.; Lloret, F. Ecological thresholds at the savanna-forest boundary: How plant traits, resources
and fire govern the distribution of tropical biomes. Ecol. Lett. 2012,15, 759–768. [CrossRef] [PubMed]
85. Atlas of Living Australia. Available online: http://www.ala.org.au (accessed on 29 June 2016).
86.
Graham, C.H.; Moritz, C.; Williams, S.E. Habitat history improves prediction of biodiversity in rainforest
fauna. Proc. Natl. Acad. Sci. USA 2006,103, 632–636. [CrossRef] [PubMed]
87.
Woinarski, J.C.Z.; Legge, S. The impacts of fire on birds in Australia’s tropical savannas. Emu Austral Ornithol.
2013,113, 319–352. [CrossRef]
88.
Moran, C.; Catterall, C.P. Responses of seed-dispersing birds to amount of rainforest in the landscape around
fragments. Conserv. Biol. 2014,28, 551–560. [CrossRef] [PubMed]
89.
Palmer, C.; Price, O.; Bach, C. Foraging ecology of the black flying fox (pteropus alecto) in the seasonal tropics
of the Northern Territory, Australia. Wildl. Res. 2000,27, 169–178. [CrossRef]
90.
Bierregaard, R.O.; Lovejoy, T.E.; Kapos, V.; Angelo Augusto dos, S.; Hutchings, R.W. The biological dynamics
of tropical rainforest fragments. BioScience 1992,42, 859–866. [CrossRef]
91.
Segan, D.B.; Murray, K.A.; Watson, J.E. A global assessment of current and future biodiversity vulnerability
to habitat loss-climate change interactions. Glob. Ecol. Conserv. 2016,5, 12–21. [CrossRef]
92.
Williams, K.J.; Prober, S.M.; Harwood, T.D.; Doerr, V.A.J.; Jeanneret, T.; Manion, G.; Ferrier, S. Implications
of Climate Change for Biodiversity: A Community-Level Modelling Approach; CSIRO Land and Water Flagship:
Canberra, Australia, 2014.
93.
Grice, A.; Setter, M. Weeds of Rainforests and Associated Ecosystems. Cooperative Research Centre for Tropical
Rainforest Ecology and Management; Rainforest CRC: Cairns, Australia, 2003; p. 116.
94.
Ash, J. The location and stability of rainforest boundaries in North-Eastern Queensland, Australia. J. Biogeogr.
1988,15, 619–630. [CrossRef]
95.
Russell-Smith, J.; Lucas, D.E.; Brock, J.; Bowman, D. Allosyncarpia-dominated rain forest in monsoonal
northern Australia. J. Veg. Sci. 1993,4, 67–82. [CrossRef]
96.
Russell-Smith, J.; Ryan, P.G.; Klessa, D.; Waight, G.; Harwood, R. Fire regimes, fire-sensitive vegetation and
fire management of the sandstone arnhem plateau, monsoonal Northern Australia. J. Appl. Ecol.
1998
,35,
829–846. [CrossRef]
97.
Bowman, D.M.J.S. Preliminary observations on the mortality of allosyncarpia ternata stems on the Arnhem
Land Plateau, Northern Australia. Aust. For. 1994,57, 62–64. [CrossRef]
98. Warddeken Land Management. Available online: www.warddeken.com/about (accessed on 20 July 2017).
99.
Altman, J.C. The dietary utilisation of flora and fauna by contemporary hunter-gatherers at Momega
Outstation, north-central Arnhem Land. Aust. Aborig. Stud. 1984, 35–46.
©
2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).