Influence of environmental variables on suitability for buffel grass

advertisement
Supplementary material
S1.
Table S1. Key environmental variables considered in the BN and the spatial layer used.
Key environmental
Source of data
Variable & spatial layer
Values used
Description of map
in BN
and assumptions
used
made when scaling
to 25km2 grid
Soil Moisture
Bioclim
Bioclim layers
Too wet >1.7
Original data at 0.01
1. mean moisture index of
generated for current
Too dry <
degree resolution.
highest quarter
climate and 2070
0.01; Just
Resampled to 25 km
2. mean moisture index of
medium and high
right 0.04-
grid via bilinear
the driest quarter
sensitivity scenarios
0.75
interpolation.
As above
Temperature
Original data at 0.01
1. min temp for coldest
too cold
degree resolution.
period
<2.5ºC
Resampled to 25 km
2. max temp for warmest
Temperature
grid via bilinear
period
too hot >
interpolation.
Temperature
Bioclim
43ºC
Rainfall
Bioclim mean annual
As above
rainfall
Low <
Original data at 0.01
300mm;
degree resolution.
Moderate
Resampled to 25 km
300-750;
grid via bilinear
High >750
interpolation.
Suitable
Original data at 0.01
≥16ºC
degree resolution.
Temperature wettest quarter
Bioclim mean temperature
As above
of wettest quarter
Resampled to 25 km
Unsuitable
grid via bilinear
<16ºC
interpolation.
Grazing Intensity
ACRIS, stocking rates
Livestock density
Assumed
Coverage only for
DSE/km2
(DSE/sq km) for
equal
rangelands.
rangeland IBRAs -
probability of
1992 to 2004 (Bastin,
low,
AcRIS Management
moderate
committee 2008)
and high
http://www.environment
grazing
.gov.au/land/publication
s/acris/report08.html
Fire Frequency
Australia Fire
High fire
Original data source
Frequency, 1km
frequency
at 1 km resolution.
AVHRR maps for 1997-
calculated as
Resampled to 25km
2008 covering the
an area burnt
grid using nearest
whole of Australia.
3 times or
neighbour
http://138.80.128.152/n
greater
assignment.
afi2/about/ausfrqdownlo
during a 12
ad.htm
year period
NVIS - reclassified to
Australia – Present
Tree cover
Original data source
suitable (vegetation with
Major Vegetation
considered
at 1km resolution.
less than >70% tree cover)
Subgroups – NVIS
unsuitable
Reclassified to
and unsuitable (more than
Stage 1, Version 3.1 –
when greater
classes ‘suitable’ and
70% tree cover and water)
Albers,
than 70%
‘unsuitable’ based on
Tree cover
http://www.environment
major vegetation
.gov.au/erin/nvis/mvg/in
subgroup
dex.html#mvs
descriptions
(ESCAVI, Australian
Vegetation Attribute
Manual version 6,
2003). Resampled to
25km grid via nearest
neighbour
assignment.
Distance to source
Current distribution of
http://www.ersa.edu.au/
Close
Point shapefile
buffel grass from
avh/
<100km
buffered at 500 km
Australian Virtual
Mid 100-
and 1000 km.
Herbarium
500km;
Converted to raster
Far >500km
25km grid.
Soil Surface Salinity
Forecasted areas
Australia Dryland
low 0-2 dS/m
Original data at
containing land of high
Salinity Assessment
= suitable
1:2,500,000.
hazard or risk of dryland
Spatial Data
medium 3-8
Resampled to 25km
salinity from 2000 to 2050
(1:2,500,000) - NLWRA
dS/m =
grid via nearest
2001.
partially
neighbour
Spatial coverage very
suitable
assignment. Current
poor. Where gaps
high >8 dS/m
layer covers <10% of
assumed equal
= unsuitable
Australia, there for
probability of suitable
dS/m=deci-
did not use.
vs unsuitable
Siemens/m
Soil pH for Australian
Soil Ph < 5.0
Original data at 0.001
Areas of Intensive
considered
degree resolution.
Agriculture of Top Soil,
strongly
Resampled to 25km
NLWRA,
acidic and
grid via bilinear
http://www.indexgeo.ne
unsuitable
interpolation.
Soil pH
t/asdd/anrdl/summary/s
phaar9nnd_01611_.ht
ml
Soil Type
Northcote key
Atlas of Australian Soils
Expert knowledge
(Northcote et al. 1960-
from John McIvor
1968), ASRIS.
Original data source
http://www.asris.csiro.a
at 1:2,000,000
u/themes/Atlas.html
resolution,
Resampled to 25km
grid via nearest
neighbour
assignment.
S2. Description of elicited relationships within the BN
The relationships between variables within the BN were determined by an expert panel and
validated with empirical data where available.
Influence of climate on suitability for buffel grass colonization
According to experts, buffel grass establishment and persistence also responds directly to ‘soil
moisture’. Ideal ‘soil moisture’ conditions for buffel grass ranged between 0.04 and 0.75 (see Fig 1
main paper, link 1). A value of 1 indicated complete soil saturation and values greater than 1.7 and
less than 0.01 impeded buffel grass establishment and persistence (Lawson et al. 2004). Buffel
grows in areas with average annual rainfall between 100-1000mm, but ideal conditions are between
300-750mm/year (http://www.tropicalforages.info/, accessed 17/02/2014; (Cox et al. 1988)). Nodes
‘Rainfall’ (Fig 1, link 2) and ‘grazing intensity’ combined to form a grazing index (discussed below).
‘Temperature’ affects buffel persistence directly through cold and heat stress (Fig 1, link 3). Below
2.5ºC and above 43ºC persistence of buffel is impeded (http://www.tropicalforages.info/, accessed
17/02/2014). Being a warm season growing grass (C4), the Kranz leaf anatomy of buffel grass is
specialized for hot climates. Where buffel grows well, the growing period coincides with summer
rains. In southern Australia, where buffel is currently absent, the rains fall in winter – when it is
currently too cold for buffel establishment. In the future as winter temperatures increase, this
impediment may no longer exist. We therefore used the BIOCLIM layer mean temperature of
wettest quarter to capture this feature (Fig 1, Temperature wettest quarter, link 4).
Influence of environmental variables on suitability for buffel grass colonization
Evidence suggests that there is a positive feedback between fire and buffel grass colonization with
its ability to withstand very hot fires and germinate quickly following fire, outcompeting other
plants, in particular spinifex, which is killed in frequent hot fires (Fairfax, Fensham 2000) (Fig 1, link
5). High ‘fire frequency’ was calculated as an area burnt three times or greater during a 12 year
period and was derived from a national fire frequency spatial layer (Table S1). According to experts,
buffel grass responds positively to moderate to high grazing pressure by reducing competition from
other less grazing tolerant native grasses (Fig 1, link 6). Nodes ‘grazing intensity’ and ‘rainfall’
combined to form a ‘grazing index’ node which was defined as; low - history of low stocking rates
relative to landscape, understory intact; moderate - understory effects obvious - loss of some shrub
structure, and; high - loss of understory structure, loss and decline of native perennial tussock
grasses. The node ‘soil quality’ influenced buffel grass establishment and persistence (Fig 1, link 10)
and was made up of ‘soil surface salinity’, ‘soil pH’ and ‘soil type’. The Australia Dryland Salinity
Assessment Spatial Data (Table S1) was used to classify soil salinity into three categories; low 0-2
dS/m, medium 3-8 dS/m and high >8 dS/m (dS/m=deci-Siemens/m), where low was deemed
suitable, medium partially suitable and high unsuitable (Fig 1, link 7). Similarly soil pH was classified
using the Australian Areas of Intensive Agriculture of Top Soil (Fig 1, link 8, Table S1) as either
unsuitable = strongly acidic pH water < 5.0 or suitable pH 7-8 (http://www.tropicalforages.info).
Finally, suitable soil types were derived from expert knowledge (J. McIvor), and mapped using data
from the Atlas of Australian Soils and ASRIS (http://www.asris.csiro.au/themes/Atlas.html) (Table S1,
Fig 1, link 9). In general, buffel grass prefers lighter textured soils, although is well-adapted to deep,
freely draining sandy loam, loam, clay loam, and well-structured red and dark clay soils (Cox et al.
1988). The node ‘distance to source’ of current buffel grass sites was used with a binary node
‘seeded’ delineating whether buffel grass had been deliberately seeded for pasture or not (Fig 1, link
12) to determine the invasion requirement ‘introduction’ (Fig 1, link 15). Distance from known buffel
records was based on the Australian Virtual Herbarium (Table S1), and was used to classify current
buffel records as close ~ <100 km, mid~ 100-500 km and far ~ >500 km. Based on their knowledge of
dispersal ability experts deemed the category close to have a 15% chance of introduction and both
mid and far a 1% chance (Fig 1, link 11). ‘Tree cover’ above 70% was deemed by the experts to
impede buffel establishment (Fig 1, link 13). Together the ‘fire frequency’, ‘grazing index’ and ‘soil
quality’ nodes influenced ‘plant competition’, which in turn influenced buffel grass ‘establishment’
(Fig 1, link 14). The node ‘introduction’ influenced landscape ‘susceptibility’ to buffel grass (Fig 1, link
15), and nodes ‘establishment’ and ‘persistence’ influenced ‘suitability’ (Fig 1, links 16 & 17). Finally
nodes ‘suitability’ along with ‘introduction’ influenced ‘susceptibility’ (Fig 1, link 18).
Download