Andrew Hansen and Linda Phillips Montana State University Curt Flather Colorado State University

Andrew Hansen and Linda PhillipsAndrew Hansen and Linda PhillipsMontana State UniversityMontana State University

Curt FlatherCurt FlatherColorado State UniversityColorado State University

Biophysical and Land-use Controls on Biophysical and Land-use Controls on Biodiversity: Regional to Continental Biodiversity: Regional to Continental

ScalesScales

Joint Workshop on NASA Biodiversity, Terrestrial Ecology, and Related Applied Sciences

May 1-2, 2008

Human Land Human Land UseUse

(Land use,Home

density)

Current Current Biodiversity Biodiversity

ValueValue

Biophysical Biophysical PotentialPotential

(i.e. Energy,Habitat

structure)

Conservation Conservation Priority/StrategiesPriority/Strategies

Research Questions:Research Questions:

1: Which biophysical predictor variables are most strongly related to 1: Which biophysical predictor variables are most strongly related to bird biodiversity potential in areas without intense human land use?bird biodiversity potential in areas without intense human land use?

2: How are these patterns of biodiversity modified due to land use?2: How are these patterns of biodiversity modified due to land use?

3: What geographic areas are highest priorities for conservation based 3: What geographic areas are highest priorities for conservation based on biodiversity modification resulting from land use change?on biodiversity modification resulting from land use change?

Ecosystem Energy as a Framework for Conservation?Ecosystem Energy as a Framework for Conservation?

Hawkins et al. 2003

Key HypothesisKey HypothesisPrimary productivity, and the factors that drive it (climate, soils, topography), ultimately influence:

disturbance and succession resources for organisms species distributions and demographies community diversity responses to habitat fragmentation, land use, exotics effectiveness of conservation

Conservation CategoryConservation Category Low EnergyLow Energy Medium EnergyMedium Energy High EnergyHigh Energy

Conservation Zones Protect high energy places Protect more natural areas Protect low energy places

Disturbance Use fire, flooding, logging judiciously in hotspots

Similar to “Descending” Use disturbance to break competitive dominance Use shifting mosaic harvest pattern Maintain structural complexity

Landscape Pattern Maintain connectivity due to migrations

Manage for patch size and edge

Sensitive Species Many species with large home ranges and low population sizes due to energy limitations

Forest interior species

Exotics High exotics likely due to productivity and high land use

Protected Area Size Large Smaller Smaller

Land Use Low overall High overall Moderate overall

Focused on hot spots Emphasize “backyard” conservation

More random across landscape

Plan development outside of hotspots

Apply restoration

Framework for Classifying Ecosystems for ConservationFramework for Classifying Ecosystems for Conservation

Focus of This TalkFocus of This Talk

1. Which biophysical predictor variables are most strongly related to bird biodiversity potential in areas without intense human land use?

Which MODIS energy products best explain patterns of bird diversity across North America?

Does the relationship between birds and energy (slope and sign) differ between places of low, medium, and high energy?

History of Predictor Variables Used to Explain History of Predictor Variables Used to Explain Species Energy PatternsSpecies Energy Patterns

NDVI = (NIR - red) / (NIR + red)NDVI = (NIR - red) / (NIR + red)

Latitude (MacArthur 1972)

Evapotranspiration (Currie 1987, Hawkins et al 2003)

Ambient temperature (Acevedo and Currie, 2003)

Water/Energy Balance (Hawkins et al 2003)

1960’s

1970’s

1980’s

Remote Sensing advances

1990’s

1999

present

MODIS Land Surface Product Development

NDVINDVIEVIEVIGPP (simulated from fpar, climate, cover type)GPP (simulated from fpar, climate, cover type)NPPNPP

AVHRR

Thematic Mapper

Precipitation (Chown et al., 2003)

----------------------------------

Strength of relationship

with bird richness

Vegetated coverLow vegetation Dense vegetation

Vegetation index

Vegetation productivityhigh

low

Phillips, L.B., Hansen, A.J. & Flather, C.H. (in press), Remote Sensing of Environment

Not complete vegetation cover(backscatter)

Dense vegetation(saturation)

Does NDVI have limitations that higher order products address?Does NDVI have limitations that higher order products address?

GPP NPP

What is the shape of the species energy relationship?What is the shape of the species energy relationship?

What is the shape of the relationship?What is the shape of the relationship?Why?Why?

energy energy

richn

ess

richn

ess

Hypothesis:

More individuals hypothesis More individuals hypothesis (Wright, 1983, Preston, 1962; MacArthur & Wilson,

1963, 1967)

Hypothesis:

Competitive exclusion Competitive exclusion (MacArthur and Levins, 1964, 1967; Grime, 1973 1979,

Rosenzweig 1992)

Energy as a framework for conservationEnergy as a framework for conservation

Energy

Bio

div

ersi

ty

Energy

Bio

div

ersi

ty

Energy

Bio

div

ersi

ty

Identify and manage Identify and manage hotspots judiciouslyhotspots judiciously

Protect harsh placesProtect harsh places

But most of But most of landscape is high in landscape is high in diversity, so more diversity, so more options for multiple options for multiple use such as shifting use such as shifting mosaic approach to mosaic approach to forest management; forest management;

If slope and sign vary among energy levels, conservation strategies should differ among low, intermediate, and high energy places.

Response dataResponse data Bird richness from BBS data for years 2000-2005, estimated Bird richness from BBS data for years 2000-2005, estimated richness using COMDYNrichness using COMDYN

Subset of routes (1838) to represent terrestrial natural routes Subset of routes (1838) to represent terrestrial natural routes (exclude human dominated land uses, water impacted)(exclude human dominated land uses, water impacted)

MethodsMethods

• Survey unit is a roadside route• 39.4 km in length• 50 stops at 0.8 km intervals• Birds tallied within 0.4 km• 3 minute sampling period

• Water birds, hawks, owls, and nonnative species excluded in this analysis

Predictor dataPredictor data Calculate both Calculate both breeding season averages breeding season averages for NDVI, EVI and for NDVI, EVI and

GPP and GPP and annual averages annual averages of NDVI, EVI, and GPP, NPPof NDVI, EVI, and GPP, NPP

MethodsMethods

Annual Average MODIS GPPAnnual Average MODIS GPP

NDVI

Enhanced Vegetation Index

Gross Primary Production

Net Primary Production

MODIS MODIS products usedproducts used

0% h

erba

ceou

s

100%

her

bace

ous

0% bare ground

100% bare ground

0% tree 100% tree50%

50%50%

Statistical analysisStatistical analysis Stratify BBS routes by vegetation life from and density Stratify BBS routes by vegetation life from and density

(MODIS VCF)(MODIS VCF)Perform correlation analyses between predictors across Perform correlation analyses between predictors across

vegetative strata and regression analysis between predictor vegetative strata and regression analysis between predictor and response variables across strata.and response variables across strata.

MethodsMethods

Statistical analysisStatistical analysis Perform regression analysis with linear, polynomial, spline Perform regression analysis with linear, polynomial, spline

and breakpoint spline modelsand breakpoint spline models

Perform simple linear regression analysis of four quartiles Perform simple linear regression analysis of four quartiles of GPP to determine slopes and significanceof GPP to determine slopes and significance

Assess and control for effects of spatial correlation on Assess and control for effects of spatial correlation on significance levels and coefficients using generalized least significance levels and coefficients using generalized least squares analyses.squares analyses.

MethodsMethods

variable time modeloverall

rankdelta aic-R from

best overall r2 adj r2

GPP annual quadratic 1 31.625 0.5353 0.5346

NDVI annual quadratic 2 72.939 0.5212 0.5205

NPP annual quadratic 3 96.321 0.513 0.5123

EVI annual quadratic 4 180.654 0.4824 0.4816

NDVI BS linear 5 288.095 0.3561 0.3556

NDVI BS quadratic 6 309.786 0.4406 0.4398

NDVI annual linear 7 331.438 0.4219 0.4215

NPP annual linear 8 374.89 0.4035 0.4031

EVI BS linear 9 395.62 0.4296 0.4292

EVI BS quadratic 10 395.62 0.3954 0.3945

EVI annual linear 11 410.694 0.3878 0.3874

GPP annual linear 12 411.244 0.3876 0.3872

GPP BS linear 13 416.29 0.376 0.3756

GPP BS quadratic 14 416.29 0.3863 0.3854

Results: Best Predictor?Results: Best Predictor?

0

5

10

15

20

0.05 0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85

NDVI

GP

P

Vegetation classesClass 1 Class 2Class 3Class 4Class 5Class 6Class 7Class 8Class 9Class 10

Correlation between NDVI and GPP across vegetation classes


0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 2 3 4 5 6 7 8 9 10

Per

cen

t o

f m

axim

um

R2

0

10

20

30

40

50

60

70

80

90

100

Co

rrelation

coefficien

t

NDVI

GPP

NPP

NDVI/GPP

NDVI/NPP

GPP/NPP

Vegetated surface gradientbare cover----------------------------------------------- herbaceous cover----------------------------------------- forested cover



with bird richness


Vegetation index


low

GPP NPP

Intrepertation: Best PredictorIntrepertation: Best Predictor

1. Annual formulation better than breeding season for all predictors

2. Results suggest that GPP better represents primary productivity and bird richness than NDVI in low and high vegetation areas

3. GPP should be used especially in desert areas (bare ground) and dense forests (SE and PNW)

4. Results help explain differences in past studies on predictors and strength of relationships: will depend on vegetation density of samples.

Vegetation Coninuous fieldsVegetation Coninuous fields

Blue gradient - bare groundRed gradient - forest cover Green gradient - herbaceous cover

R2 = 0.5346

0

0.5

1

1.5

2

2.5

0 5 10 15 20 25

GPP (gC/m2)

Bir

d r

ich

ne

ss

(lo

g)

Results: Slope and Shape?Results: Slope and Shape?


a=0.083p<.001

a=0.013p<.001

a=0.005p<.036

a= - 0.018p<.001


Variable (annual) model overall rank

delta aic-R from best overall r2 adj r2

GPP Spline (cubic) 1 0 0.5484 0.5464

GPP Breakpoint (linear) 2 26.304 0.5384 0.5371

GPP Quadratic 3 31.625 0.5353 0.5346

NDVI Quadratic 4 72.939 0.5212 0.5205

NDVI Spline (cubic) 5 74.416 0.5234 0.5214

NDVI Breakpoint (linear) 6 83.322 0.519 0.5176

NPP Spline (cubic) 7 91.33 0.5176 0.5155

NPP Quadratic 8 96.321 0.513 0.5123

NPP Breakpoint (linear) 9 101.476 0.5126 0.5112

EVI Spline (cubic) 10 180.609 0.4854 0.4832

EVI Quadratic 11 180.654 0.4824 0.4816

EVI Breakpoint (linear) 12 186.134 0.4818 0.4803



0.7

0.9

1.1

1.3

1.5

1.7

1.9

2.1

0 2 4 6 8 10 12 14 16 18 20

GPP (gC/m2)

rich

nes

s (l

og

)

Results: Slope and Results: Slope and Shape?Shape?

0.7

0.9

1.1

1.3

1.5

1.7

1.9

2.1

0 2 4 6 8 10 12 14 16 18 20

GPP (gC/m2)

rich

nes

s (l

og

)

energy energy

richn

ess

richn

ess

More Individuals Hypothesis

Predicts higher habitat heterogeneity in areas

of high richness

Competitive Exclusion Hypothesis

Predicts high canopy cover in overstory and

lower habitat heterogeneity

Interpretation: Interpretation: Slope and ShapeSlope and Shape

5000 10000 15000

510

15

Annual Gross Primary Productivity

Bir

d S

pe

cie

s R

ich

ne

ss

Coast RangeSpringfieldCle ElumGoldforkYellowstone

Breakpoint GPP=12266; R-squared = 0.46Quadratic Model; R-squared = 0.44Quadratic 95% confidence bands

3000 4000 5000 6000 7000 8000 90005

10

15

Breeding Season NDVI

Bir

d S

pe

cie

s R

ich

ne

ss

Coast RangeSpringfieldCle ElumGoldforkYellowstone

Cubic Model; R-squared = 0.49Cubic 95% confidence bandsBreakpoint NDVI=8007; R-squared = 0.48

Disturbance Effects and Ecosystem EnergyDisturbance Effects and Ecosystem Energy

Diversity increases with disturbance under high energy and decreases under low energy.

HighLow

Lo

wH

igh

Divers

ity

Landscape Productivity

Inte

nsi

ty o

f D

istu

rba

nce

SpringfieldSpringfield

Cle ElumCle Elum

Disturbance Frequency

Spec

ies

Div

ersi

ty

Cle Elum

Springfield

High Low

60 70 80 90 100

81

01

21

41

61

8

% of Landscape Occupied by Closed Canopy Forest

Bir

d R

ich

ne

ss

Site: SpringfieldR2=.16P-value <.01

High Low

40 50 60 70 80 90

68

10

12

14

% of Landscape Occupied by Closed Canopy Forest

Bir

d R

ich

ne

ss

Site: Cle ElumR2=.30P-value <.01

High Low

Huston 1994.

McWethy et al. in review.


(Land use,Home

density)


ValueValue



structure)


Next Steps:Next Steps:

1: Which biophysical predictor variables are most strongly related to bird biodiversity potential in areas without intense human land use?

2: How are these patterns of biodiversity modified due to land use?

3: What geographic areas are highest priorities for conservation based on biodiversity modification resulting from land use change?


(Land use,Home

density)


ValueValue



structure)


Next Steps:Next Steps:

1: Which biophysical predictor variables are most strongly related to bird biodiversity potential in areas without intense human land use?

2: How are these patterns of biodiversity modified due to land use?

3: What geographic areas are highest priorities for conservation based on biodiversity modification resulting from land use change?

Vegetation structure from ELVS/GLAS

Balmford et al. 2001

Vertebrates and NPP

This study.

Humans and NPP

Next StepsNext Steps

Environmental gradient

bio

div

ers

ity

Fine scale studies, local scale

Broader scale studies, regional scale

Broadest scale studies, covering entire environmental gradient

Energy thresholds where limiting factors for organismsChange and cause change SER

Does the shape of the relationship vary with Does the shape of the relationship vary with energy levels (geographically)?energy levels (geographically)?

Is the negative portion of the unimodal Is the negative portion of the unimodal relationship real?relationship real?

Nugget .002Sill .006So using GLS, enter (800000, .25)

NDVI = (NIR - red) / (NIR + red)

Do higher order MODIS products help us answer these questions? Do higher order MODIS products help us answer these questions?


with bird richness


Vegetation index


low

Phillips, L.B., Hansen, A.J. & Flather, C.H. (in press), Remote Sensing of Environment

Not complete vegetation cover(backscatter)

Dense vegetation(saturation)

Does NDVI have limitations that higher order products address?Does NDVI have limitations that higher order products address?

GPP NPP


This slide corresponds to green cells in previous slide

Andrew Hansen and Linda Phillips Montana State University Curt Flather Colorado State University

Documents