Interfacing Vegetation Databases with ecological theory and

practical analysis.Mike Austin, Margaret Cawsey and

Andre Zerger

CSIRO Sustainable EcosystemsCanberra Australia

Examples of Current Vegetation Databases

• Purpose:Vegetation classification– TurboVeg: Phytosociological relevees– Vegbank: General vegetation classification

• Purpose: Vegetation Analysis– Minimalist: minimum data set– Biograd: Regional prediction and mapping

Purposeand

Product

Statistical methods

model

Ecological theory model

DataMeasurement

model

RelationalDatabase

GeographicInformation

System(GIS)

Topics

• Interface between vegetation databases theory and analysis

• Interface between data and practical applications for conservation evaluation

Biograd Database

• Grew from minimalist database– Location, plot data, co-occurrence of canopy species,

slope, aspect, elevation.– Current size 10027 plots.

• Used software packages and GIS to derive environmental variables– Temperature, rainfall, radiation, soil properties.

• Predicted potential vegetation from species environmental models

Application to Theory

• Pattern of Species Density in relation to climate.

Plot Tree Species Density in response to Temperature

0

5

10

15

20

25

30

0 2 4 6 8 10 12 14 16 18 20

Annual Mean Temperature

Plot

Spe

cies

Den

sity

Mean Species Density in response to Mean Annual

Temperature in one degree classes

0

1

2

34

5

6

7

8

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.

5

11.5 12.

5

13.

5

14.

5

15.

5

16.

5

17.

5

18.

5

19.

5

Mean Annual Temperature Classes

Mea

n Spe

cies

Den

sity

Number of plots in each temperature class

0

200

400

600

800

1000

1200

1400

1600

1800

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5

Temperature class (midpoint)

Num

ber

of p

lots

Species Density and Mean Annual Temperature by Lithology

0

2

4

6

8

10

12

14

16

2.5 4.5 6.5 8.5 10.5 12.5 14.5 16.5 18.5 20.5

Mean Annual Temperature classes

Sp

ecie

s d

ensi

ty Volcanics

Hard Seds

Soft Seds

Granites

Other liths

Quat Seds

Species Density and Mean Annual Temperature by Topographic position on Soft Sediments

0

2

4

6

8

10

12

0 5 10 15 20

Mean annual Temperature

Spe

cies

Den

sity ridge

slope

lowslope

gully

flat

Questions•What is a suitable statistical method for species/environment modelling

•What environmental variables predict species density?

•What is their relative importance?

•Does their importance vary with mean annual temperature?

•What does this say about models of species density determinants?

•What are the Database requirements for this type of analysis?

Some Suggested Answers

• Statistical modelling using Generalized Additive Modelling (GAM)

• Predictors: use both climatic and local variables ( 7 variables used)

• Importance: GAM gives relative measure

• Hypothesis: Behaviour of tree species density differs above and below 12ºC :- split data.

Mean annual temperature

Mean annual rainfall

slope

Mean annual temperature

Mean annual rainfall

slope

>=12 degrees<12 degrees

Species density responses to environmental predictors for two models <12 and >12 degrees

topography

aspect

>=12 degrees<12 degrees

Species density responses to environmental predictors for two models <12 and >12 degrees

topography

aspect

4=gully1=ridge

lithology

relative heat load

>=12 degrees<12 degrees

Species density responses to environmental predictors for two models <12 and >12 degrees

relative heat load and lithology are not

included in this model

>=12 degrees model <12 degrees model

Relative contribution of environmental predictors

Purposeand

Product

Statistical methods

model

Ecological theory model

DataMeasurement

model

RelationalDatabase

GeographicInformation

System(GIS)

Application to conservation evaluation

• Problem of aggregating data into classes for inclusion in a data base

• How many soil types should be recognised?

• What are the implications for predicting species distribution?

Predicting Spatial Distribution of Acacia pendula

• Acacia pendula occurs on floodplain soils under low rainfall conditions (<600mm mean annual rainfall) in the Central Lachlan region of New South Wales, Australia.

• GAM models of 135 tree and shrub species including A. pendula were used to predict potential vegetation on cleared areas in the region.

Condobolin

Tullamore

Parkes

Forbes

GrenfellCowra

147 º 148 º 150 º-32.5 º

-33 º

-33.5 º

-34 º

Selected study area

The central Lachlan region

Study area

1:100,000 mapsheet boundary

NSW

...

..

.

Relational Database

Geographical InformationSystems (GIS) data

DigitalElevation

Model (DEM)

Data Collectionand Management

SurveyClassificationand Mapping

Products

Soil landscapedata frommanuals

EnvironmentalStratification

Survey

DigitalTerrain

Models (DTM)

Climaticattributes

Soil landscapes

Multivariatepattern analysis

Statisticalmodelling of

individual species

SpeciesPrediction

SpeciesPrediction

SpeciesPrediction

SpeciesPrediction

SpeciesPredictions

Spatial allocation tovegetation communities

PredictedVegetation

Vegetationplot dataPlant species

data

Plot location &environmental

data

Plotvegetation

data

An integrated approach to vegetation mapping

Drainage

Individual species predictionsMean

Temperature

Geology

TopographicPosition

ArcViewGrasp script

Great Soil Group

Soil Depth

Soil pH

Soil Fertility

TemperatureSeasonality

RainfallSeasonality

Annual MeanRainfall

S-PlusGrasp

Species LookupTables

Species LookupTables

Species LookupTables

Species LookupTables

Species LookupTables

PlotData

SpeciesModels

SpeciesPrediction

SpeciesPrediction

SpeciesPrediction

SpeciesPrediction

SpeciesPredictions

Spatial Prediction of Acacia pendula using original Great Soil Groups

Masked mean annual rainfall > 568mm

Spatial Prediction of Acacia pendula using reaggregated Great Soil Groups

Masked mean annual rainfall >568mm

Spatial Prediction of Acacia pendula

Difference between model predictions

Conclusions

• Small changes in attribute classification can have a marked impact on outcomes

• Attributes in a database should be kept at as disaggregated a level as possible

• How cost-effective are databases where numerous attributes are kept which may not be used?

• Is this best done with “in-house” or commercial software

Parkes

Forbes

Cowra

Grenfell

Condobolin

Location map of central Lachlan region

Predicted vegetation map for the central Lachlan region

Parkes

Forbes

Cowra

Grenfell

Condobolin

Location map of central Lachlan region

Current remnant distribution of predicted vegetation communities

Remaining area for different communities(based on M305 mapping of woody vegetation)

Red < 10 % remaining Green > 30 % remaining

Alliance Community Potential woodedarea (km2)

Area remaining(%)

1 E. melliodora / E. microcarpa 1552 3 2 E. melliodora 22 8 3 E. camaldulensis / E. melliodora 260 10

Eucalyptus melliodora

4 E. albens / E. melliodora 262 4 6 E. goniocalyx / E. blakelyi / E. melliodora 755 92 Eucalyptus melliodora /

E. blakelyi 7 E. bridgesian / E. blakelyi / E. melliodora 172 88 E. microcarpa / Callitris glaucophylla 547 2110 Allocasuarina luehmanii / E. microcarpa 59 3

Eucalyptus microcarpa

13 E. microcarpa / Casuarina cristata 277 2 Callitris glaucophylla 15 Callitris glaucophylla / E. albens 67 2 Eucalyptus populnea 18 E. populnea / Callitris glaucophylla 5202 7

23 Callitris endlicheri / E. sideroxylon 1557 20 Callitris endlicheri24 E. dealbata/C. endlicheri/A. doratoxylon 92 25

Eucalyptus blakelyi / E. macrorhyncha

28 E. blakelyi / Callitris endlicheri 381 5

32 Callitris endlicheri / E. macrorhyncha 393 5733 Calytrix tetragona / C. endlicheri / E. macrorhyncha

1232 49 Callitris endlicheri / E. macrorhyncha

34 C.endlicheri / Baeckea cunninghamiana / E. sideroxylon

1063 26

36 E. macrorhyncha / E. goniocalyx 76 48 E. macrorhyncha37 E. polyanthemos / E. macrorhyncha / E. albens

75 13

43 E. viminalis / Acacia melanoxylon 107 23 E. pauciflora/E. viminalis44 E. pauciflora / Acacia dealbata 35 73

Purposeand

Product

Statistical methods

model

Ecological theory model

DataMeasurement

model

RelationalDatabase

GeographicInformation

System(GIS)

Final

Vegetation plots in “good” condition

(Good condition is defined as greater than 50% native plant cover in the lower vegetation layer)

Community Arearemaining

(%)

Number ofplots

surveyed

Proportion > 50 %

native cover

Area with“modest”

condition (%) 1 E. melliodora / E. microcarpa 3 30 0.17 0.5 2 E. melliodora 8 39 0.03 0.2 3 E. camaldulensis / E. melliodora 10 12 0.08 0.8 4 E. albens / E. melliodora 4 57 0.05 0.2 6 E. goniocalyx / E. blakelyi / E. melliodora 9 47 0.08 0.7 7 E. bridgesian / E. blakelyi / E. melliodora 8 31 0.03 0.2 8 E. microcarpa / Callitris glaucophylla 21 63 0.38 8.010 Allocasuarina luehmanii / E. microcarpa 3 18 0.39 1.218 E. populnea / Callitris glaucophylla 7 105 0.31 2.223 Callitris endlicheri / E. sideroxylon 20 30 0.30 6.024 E. dealbata/C. endlicheri/A. doratoxylon 25 14 0.86 12.028 E. blakelyi / Callitris endlicheri 5 20 0.15 0.834 C.endlicheri / Baeckea cunninghamiana / E. sideroxylon

26 12 0.75 19.5

36 E. macrorhyncha / E. goniocalyx 48 18 0.06 2.948 E. sideroxylon / E. dwyeri 34 23 0.30 10.249 E. sideroxylon / E. microcarpa 13 19 0.47 6.162 E. dwyeri /Callitris endlicheri / A. doratoxylon

68 38 0.42 28.6

71 E. camaldulensis 10 82 0.16 1.675 E. albens / E. microcarpa 6 103 0.01 0.6

Area and condition estimates for communities

Red < 10 % in “modest” condition

COMMUNITY AS AN AREAL CONCEPT

RECOGNITION OF COMMUNITIES DEPENDS ON THE FREQUENCY OF ENVIRONMENTAL COMBINATIONS IN THE LANDSCAPE

Topographic distribution of “communities” as indicated in previous slide

Altered topographic distribution of “communities” with the lowest bench at 170m and the highest bench at 430m

Frequency of species co-occurrences as a function of landscape