Andrew Hansen and Linda Phillips

1
Biophysical and Land-use Controls on
Biodiversity Regional to Continental Scales
Andrew Hansen and Linda Phillips Montana State
University Curt Flather Colorado State
University
Joint Workshop on NASA Biodiversity, Terrestrial
Ecology, and Related Applied Sciences May 1-2,
2008
2
Research Questions 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?
Biophysical Potential (i.e. Energy, Habitat
structure)
Human Land Use (Land use, Home density)
Current Biodiversity Value
Conservation Priority/Strategies
3
Ecosystem Energy as a Framework for Conservation?

• Key Hypothesis
• Primary 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

Hawkins et al. 2003
4
Framework for Classifying Ecosystems for
Conservation
Conservation Category Low Energy Medium Energy High 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
Use disturbance to break competitive dominance Use shifting mosaic harvest pattern Maintain structural complexity
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
5
Focus of This Talk

• 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?

6
History of Predictor Variables Used to Explain
Species Energy Patterns
-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --

Latitude (MacArthur 1972)
1960s 1970s 1980s Remote Sensing
advances 1990s 1999 present
Evapotranspiration (Currie 1987, Hawkins et al
2003)
Ambient temperature (Acevedo and Currie, 2003)
Precipitation (Chown et al., 2003)
Water/Energy Balance (Hawkins et al 2003)
AVHRR
NDVI (NIR - red) / (NIR red)
Thematic Mapper
MODIS Land Surface Product Development
NDVI EVI GPP (simulated from fpar, climate, cover
type) NPP
7
Does NDVI have limitations that higher order
products address?
GPP NPP
Not complete vegetation cover (backscatter)
Dense vegetation (saturation)
Phillips, L.B., Hansen, A.J. Flather, C.H. (in
press), Remote Sensing of Environment
8
What is the shape of the species energy
relationship?
9
What is the shape of the relationship? Why?
richness
richness
energy
energy
Hypothesis More individuals hypothesis
(Wright, 1983, Preston, 1962 MacArthur
Wilson, 1963, 1967)
Hypothesis Competitive exclusion (MacArthur
and Levins, 1964, 1967 Grime, 1973 1979,
Rosenzweig 1992)
10
Energy as a framework for conservation
If slope and sign vary among energy levels,
conservation strategies should differ among low,
intermediate, and high energy places.
Protect harsh places But most of landscape is
high in diversity, so more options for multiple
use such as shifting mosaic approach to forest
management
Identify and manage hotspots judiciously
11
Methods

• Response data
• Bird richness from BBS data for years 2000-2005,
  estimated richness using COMDYN
• Subset of routes (1838) to represent terrestrial
  natural routes (exclude human dominated land
  uses, water impacted)

• 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

12
Methods

• Predictor data
• Calculate both breeding season averages for
  NDVI, EVI and GPP and annual averages of NDVI,
  EVI, and GPP, NPP

Annual Average MODIS GPP
MODIS products used
NDVI Enhanced Vegetation Index Gross Primary
Production Net Primary Production
13
Methods

• Statistical analysis
• Stratify BBS routes by vegetation life from and
  density (MODIS VCF)
• Perform correlation analyses between predictors
  across vegetative strata and regression analysis
  between predictor and response variables across
  strata.

14
Methods

• Statistical analysis
• Perform regression analysis with linear,
  polynomial, spline and breakpoint spline models
• Perform simple linear regression analysis of
  four quartiles of GPP to determine slopes and
  significance
• Assess and control for effects of spatial
  correlation on significance levels and
  coefficients using generalized least squares
  analyses.

15
Results Best Predictor?
variable time model overall rank delta 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
16
Results Best Predictor?
Correlation between NDVI and GPP across
vegetation classes
17
Results Best Predictor?
18
Intrepertation 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.

GPP NPP
19
Vegetation Coninuous fields
Blue gradient - bare ground Red gradient - forest
cover Green gradient - herbaceous cover
20
Results Slope and Shape?
R2 0.5346
21
Results Slope and Shape?
22
Results Slope and Shape?
a - 0.018 plt.001
a0.083 plt.001
a0.005 plt.036
a0.013 plt.001
23
Results Slope and Shape?
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
24
Results Slope and Shape?
25
Results Slope and Shape?
26
Interpretation Slope and Shape
richness
richness
energy
energy
Competitive Exclusion Hypothesis Predicts high
canopy cover in overstory and lower habitat
heterogeneity
More Individuals Hypothesis Predicts higher
habitat heterogeneity in areas of high richness
27
(No Transcript)
28
Disturbance Effects and Ecosystem Energy
Huston 1994.
High
Springfield
Intensity of Disturbance
Diversity
Low
McWethy et al. in review.
Low
High
Diversity increases with disturbance under high
energy and decreases under low energy.
Landscape Productivity
Cle Elum
29
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?
Biophysical Potential (i.e. Energy, Habitat
structure)
Human Land Use (Land use, Home density)
Current Biodiversity Value
Conservation Priority/Strategies
30
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
Biophysical Potential (i.e. Energy, Habitat
structure)
Human Land Use (Land use, Home density)
Current Biodiversity Value
Conservation Priority/Strategies
31
Next Steps
Vertebrates and NPP
Humans and NPP
This study.
Balmford et al. 2001
32
(No Transcript)
33
Does the shape of the relationship vary with
energy levels (geographically)?
34
Is the negative portion of the unimodal
relationship real?
Nugget .002 Sill .006 So using GLS, enter
(800000, .25)
35
(No Transcript)
36
Do higher order MODIS products help us answer
these questions?
NDVI (NIR - red) / (NIR red)
37
Results Best Predictor?
Does NDVI have limitations that higher order
products address?
GPP NPP
Not complete vegetation cover (backscatter)
Dense vegetation (saturation)
Phillips, L.B., Hansen, A.J. Flather, C.H. (in
press), Remote Sensing of Environment
38
This slide corresponds to green cells in previous
slide