Transcript Modeling biodiversity response to habitat

Modeling biodiversity response to habitat
heterogeneity in agricultural lands of
Eastern Ontario using multi spatial and
temporal remote sensing data
Niloofar Alavi
Supervisor: Dr. Doug King
Background: Biodiversity and Habitat
Heterogeneity

Biodiversity:

The variability among living organisms from all sources including,
terrestrial, marine and other aquatic ecosystems and the ecological
complexes of which they are part
-International Convention on Biodiversity (1992)


Biodiversity has declines due to human activities such as
intensive agriculture
Landscape heterogeneity:

Variation in the horizontal dimension of the landscape
-August (1983)


Compositional and configurational heterogeneity
Landscape Heterogeneity Hypothesis:

heterogeneous and complex habitats promote biodiversity by providing
more available sources for species
-Simpson (1949)
-MacArthur & Wilson (1967)
Using Earth Observation Data in Biodiversity
Modeling



Spatial resolution
Temporal resolution
Spectral resolution


Spectral Heterogeneity
Discrete thematic maps vs. continuous metrics
Research Questions and Goal

What is the response of biodiversity to habitat
heterogeneity at multiple spatial and temporal
scales in mixed agricultural landscapes?

Can high spatial and low temporal resolution
imagery, and low spatial and high temporal
resolution imagery be used to create robust
biodiversity models that reflect the response of
biodiversity to habitat heterogeneity?
To fill out the gap in the literature by creating a
multi-spatial and multi-temporal biodiversity
model using multiple farmland taxa.

Study Site




An agricultural region within Eastern Ontario
Approximately 15,000 km2
Mainly covered by maize (21%), soybean (19%),
forage crops (alfalfa, clover, hay; 30%), and
wheat (3%)
Approximately 100 1 km2 sample landscapes




low crop diversity and low mean field size
low crop diversity and high mean field size
high crop diversity and low mean field size
high crop diversity and high mean field size
Study Site
Pasher et al., 2013
Data

Biodiversity Data:




surveyed in the cropped portions of 93 of the
sample landscapes
46 landscapes were surveyed in 2011
47 landscapes were surveyed in 2012
seven species group: birds, plants, butterflies,
syrphids, bees, carabids and spiders

Alpha and Gamma indices
Girard et al., 2012
Data


Remote sensing data
Time series of:

MODIS 1999


MODIS 16-day and 7-day NDVI
Landsat1982
40-cm resolution aerial photos
(2011-2012)

Biodiversity Modeling Variables


Discrete variables:
Compositional





Crop type variability
Patch richness
Amount of suitable species habitat
Amount of natural and semi-natural patches
Configurational



Mean patch size
Edge density
Mean patch shape variability
Biodiversity Modeling Variables

Continuous remote sensing variables


Original spectral band reflectance,
Band combinations


Band transformation



Vegetation indices
Principal Component Analysis
Tasseled Cap Transformation
Fractions

Spectral Mixture Analysis
Temporal Analysis


Coarse Scale Temporal Analysis
 MODIS within seasonal
Fine Scale Temporal Analysis
 Landsat inter-seasonal and inter-annual
Expected Results




We expect to find a correlation between the landscape
heterogeneity metrics and the response of different
species groups.
We expect that the biodiversity models derived from the
selected continuous landscape metrics can display this
correlation and create robust models that explain the
response of different taxa to these metrics.
We expect that MODIS and Landsat time series will be
able to detect the temporal trajectories of past
phenological changes in mixed agricultural landscapes.
We expect to identify the optimal operational and
conservational scale at which each species group
responses to habitat heterogeneity in mixed agricultural
landscapes.
Questions and discussion
Spectral mixture analysis




In spectral mixture analysis the spectral signatures of the constituent substances present in a mixed pixel are referred to as
endmembers, and the fractional area coverage of each endmember in a pixel is called its abundance.
spectral mixture analysis is the process of decomposing the acquired spectrum of a mixed pixel into a set of endmembers and their
corresponding factional abundances
3 directions of spatial spectral mixture analysis:

Endmember extraction

Directly from the remote sensing images (image endmember)

Measure in the field or laboratory (reference endmember) ×

Selecting endmember combination

Limitations:
 The spectral signature of each endmember is assumed unchanged: Endmember variability problem
 The number of endmembers in the entire scene is assumed unchanged: Too many endmemebers for each pixel causes
error

Methods:
 Per-pixel
 Per-field

Abundance estimation
Methods:

Linear mixture model: spectral signature of a mixed pixel is represented by the weighted sum of the endmember spectra and that
the weights associated with the endmembers are given by their corresponding proportional area coverage in the pixel.

Non-linear mixture model: When the ground materials depict an intimate mixture where the light is multiply scattered between at
least two components.

The importance of extracting exact number of endmembers

Too few endmembers

Too many endmembers
Band transformation

Principal Component
Analysis



A linear transformation
technique
The original set of
potentially correlated
numerical variables with
high covariance  smaller
and uncorrelated sets of
variables
Reduces redundancy


In mixed landscapes usually
the first three principal
components represent
most of the variance of the
original dataset.
Measuring the mean
Euclidean distance between
spectral clusters derived
from PCA is a measure of
spectral heterogeneity.
Band transformation




Tasseled Cap Transformation
A linear transformation technique
The original spectral bands new sets of bands
The first three TCTs :





Brightness
Greenness
Wetness
The Brightness component is by definition a positive
value,
The Greenness depends on the contrast between the
visible and near-infrared
Scale in modeling biodiversity

Biodiversity: various forms of life on earth.
 genetic
 Species


species richness
species evenness or abundance


Ecosystems
Habitat heterogeneity

Compositional heterogeneity

Configurational heterogeneity

Habitat heterogeneity hypothesis

Heterogeneity vs. fragmentation


Extinction threshold
Intermediate heterogeneity hypotheses
Fahrig et al. 2011
Fahrig et al. 2003
Species-specific factors


Some studies noted species-specific factors that affect the biodiversity response to habitat
heterogeneity.
Species mobility

Oliver et al. (2010)



36 British butterfly species
166 study sites
3 different spatial scales:



1 km
2 km
5 km
from the center of the study sites.



Trophic level
body size
habitat specialization

Steckel et al. (2014)


bees, wasps and their antagonists
3 regions in Germany



Local scale (sampling plots)
Landscape scale (8 radii between 250 m to 2000 m around the sampling plots)
Regional scale (each study region).
Biodiversity responses to habitat



Loss of species
Species replacement
Stability in species assemblage
Burel et al. 2004
Spatial and Temporal Scales in Biodiversity
Modeling

Spatial scale:



Spatial scales in modeling biodiversity in
agricultural landscapes:




Extent
Grain
between-farms (regional)
between-fields (landscape)
within-field (local)
Spatial-temporal trade-off
Turner et al. 2001
Landscape matrix vs. patchiness

Landscape heterogeneity:

The variation in the
horizontal dimension of the
landscape (August, 1983).
 1) Landscape matrix





landscape is divided into:

“habitat”
“non-habitat”
2) Heterogeneous landscapes
Fahrig et al. 2011
This view has been
challenged by many
authors:

The species perceive the
landscape in a much more
complex manner and they
use more than one land
cover type as resources to
provide their requirements
The heterogeneous
landscape view replaced
the habitat matrix view to
explain the complexity of
species response to
habitat patchiness.