Transcript Intro

Intro
NordLaM Nordic Workshop: Deriving Indicators from Earth Observation
Data - Limitations and Potential for Landscape Monitoring,
22nd - 23rd October, Drøbak, Norway
The potential of landscape metrics from
Remote Sensing data as indicators in forest
environments
Niels Chr. Nielsen, M.Sc.
[email protected]
JRC Project:
Development and evaluation of remote sensing based spatial indicators for
the assessment of forest biodiversity and sustainability, using landscape
metrics derived from high- to medium resolution sensors
Lancaster University thesis under way:
Development and test of spatial metrics derived from EO data for indicators
of sustainable management of forest and woodlands at the landscape level
Structure of presentation:
Definitions of indicators for different purposes
Landscape ecology – spatial metrics
Land Cover and forest maps, data needs
and potential outputs
Processing chain, combining with GIS
Limitations to monitoring, examples from
study of fragmentation
Conclusions, perspectives for monitoring
Convention framework for development of
indicators:
Helsinki (93) – Lisbon (98): Ministerial Conference on Protection of
Forests in Europe
Convention on Biological Diversity
IUFRO working group on Sustainable Forest Management (SFM)
European Landscape Convention (Firenze 2000)
“Natura 2000” network (linked to the EU habitats directive)
Activities somehow related:
Timber Certification
BEAR project on forest biodiversity + indicators of same
GAP analysis
Kyoto protocol (forests as carbon pool)
These processes could use indicators as tool for
monitoring and reporting of state and progress!
SFM Hierarchy
Sustainable Forest Management (SFM)
hierarchy:
PRINCIPLES (Universal)
CRITERIA (General)
ARE THE GOALS ACHIEVED?
INDICATORS (Adapted to local
conditions)
ADJUSTING
+VALIDATION
VERIFIERS (Basic observations,
comparable, can be threshold values )
Helsinki process (MCPFE) criteria:
1.MAINTENANCE AND APPROPRIATE ENHANCEMENT OF FOREST
RESOURCES AND THEIR CONTRIBUTION TO GLOBAL CARBON
CYCLES: Area, Age structure
2.MAINTENANCE OF FOREST ECOSYSTEM HEALTH AND VITALITY :
Burned area, Storm damage
3.MAINTENANCE AND ENCOURAGEMENT OF PRODUCTIVE
FUNCTIONS OF FORESTS (WOOD AND NON-WOOD):
Balance Growth - Removals
4.MAINTENANCE, CONSERVATION AND APPROPRIATE
ENHANCEMENT OF BIOLOGICAL DIVERSITY IN FOREST
ECOSYSTEMS (Natural forest types)
5.MAINTENANCE AND APPROPRIATE ENHANCEMENT OF
PROTECTIVE FUNCTIONS IN FOREST MANAGEMENT (NOTABLY
SOIL AND WATER)
6.MAINTENANCE OF OTHER SOCIO-ECONOMIC FUNCTIONS AND
CONDITIONS
BEAR biodiversity indicators
areas where application of RS data is possible:
NATIONAL SCALE
Structural factors
Indicators
Total area of forests
Total area (ha)
Area in relation to total land area (%)
Afforestation (yearly rate)
Deforestation (yearly rate)
Natural regeneration (by 10 years)
Area of ‘ancient’ woodland
Percentage of total area
Compositional factors
Fire/lightning
Storms
Number, size and area (% of forest) and age of forest
affected
Average annual area of damage
Silvicultural regimes
Clearcuts (number and area)
Age class frequency in relation to felling area
Agriculture/grazing/browsing
Area transformation from agriculture to forestry and
vice versa
BEAR biodiversity indicators, landscape and stand scale
LANDSCAPE SCALE
Distribution of tree species in different age
classes
All species in 20yr age classes up to 250+ years
Representativity of forest biodiversity types
Area and percentage of the biodiv. forest types
Old growth forest guild habitat connectivity
Spatial pattern of habitat type
Declining trees forest guild habitat connectivity
Spatial pattern of habitat type
Recently disturbed forest guild habitat
connectivity
E.g. for boreal forest: area of ground with trees that
was burned
Patch size distribution
Mean value and st.dev. of patch size
Reasons for stand renewal, abiotic
Fire
Wind
STAND SCALE
Large trees
Basal area and/or density
Size of stand
In Ha
Shape of stand
Area and perimeter (+more advanced?)
Core concepts from Landscape Ecology :
- Flows of matter, energy, information (across landscapes, soilvegetation-air)
- Importance of spatial structure and terrain
- Disturbance – regeneration (shifting mosaic in natural systems)
- Holistic approach – analysis at “landscape level” – the landscape as a
system, hierarchical, multifunctional approach
- Core areas – ecotones
-Island biogeography: species/area-curves
--
Later: Metapopulation ecology
-‘Ancillary’ assumptions:
-Richness of biotope types = richness of habitats
-Interspersion promotes co-habitation of species and
movement of indivduals
LE concepts
Landscape concepts
LANDSCAPE
MATRIX
CORRIDOR
CLASS
PATCH
STEPPING
STONES
Example 1 : Patterns of forest in the landscape
Natural
Managed
Shape e.g. edge/area measures
Connected
Fragmented
Number of patches, distance measures
Example 2 : Patterns of patches in the forest
More - less DIVERSE (area presence, distribution measures)
More - less INTERSPERSED (edge length, neighborhood-juxtaposition measures)
Examples of spatial metrics :
”Information Hierarchy” of Spatial Metrics
Spatial
information type
Describing..
Output units
Area
Land cover classes or patches
m2 , ha, km2, %
Count
Objects, patches (richness of)
Number
Shape
Structure: from patches to
landscapes
Any (m-1, FD
normally unitless)
Position, distance
Relative placement of patches
m, km
Topology
Context – connectivity,
relative edge type proportions
(weighted edge indices)
Unit-less number
ADVANCED
less
more
What is possible with Landscape Ecology?
Concept
“BASIS”
Widely accepted as
facts/possible
“POTENTIALS”
Under investigation/
discussion
“LIMITS”
Not accepted/at the
moment not seen as
possible
Land cover
Mapping land cover types
Mapping habitat types
Mapping species presence
using EO
Species/
area curves
Species/area relationships
exist
Mathematical formulation
of S/A relations
Predicting
presence/absence of a
single species in specific
habitat(?)
Landscape
structure
Influence of landscape
structure on taxonomic
diversity
Prediction of single
species presence solely
from landscape diversity
information
Landscape
Metrics
Calculation of landscape
metrics
Structural diversity as
surrogate for taxonomic
diversity, causal links
between measures of
(abiotic) landscape
diversity and taxonomic
diversity
Meaning of landscape
metrics
Relating landscape metric
values to abundance of a
certain species or directly
to taxonomic diversity
What is possible with Landscape Ecology2?
“BASIS”
Widely accepted as
facts/possible
“POTENTIALS”
Under investigation/
discussion
Scale
Influence of
measurement scale on
mapping accuracy,
metrics values etc.
Also on spatial
perception by individual
animals
Mathematical (spatial
statistics) processes
influencing spatial metrics,
ecological scaling
mechanisms governing
results from measurement
of (local) extinctions
dispersal of animal and
plants (sampling issues)
‘Grand unifying theory’ of
scaling behaviour, reliable
prediction of metrics values
between imagery at very
different scales (?)
PatchCorridorMatrix
(PCM)
model
PCM model can be
applied in agricultural
landscapes
Applicability of PCM model
inside forests
Delineation of functional
‘habitat patches’ in forests
(only/purely) from EO data
Corridors
Definitions and
mapping of corridors in
open/high contrast
lands
Roles of corridors in
landscapes (for specific
species), managing for
biodiversity by creating
corridors
Measuring influence of
corridors on taxonomic
diversity in landscapes
Concept
“LIMITS”
Not accepted/at the
moment not seen as
possible
Who needs forest information ?
* International organisations, NGO’s and
environmental organisations
* National ministries
* Research and academic institutes
* Forest Industry
* Forest owners
Forest processes, spatio-temporally
Forest management information needs 1
Function, type and level of
information
Variable / data type
Forest protection
Stand
Forest area (actual/potential ratio)
Species Composition
Structure (horizontal, vertical)
Site
Soil
Vegetation types
Topography (elevation, aspect, slope)
Climate
Stability
Forest condition, Quality, health
Management
Value of protected infrastructure
Water resources
Objectives
Forest management information needs 2
Ecosystem /
environment
Variable / data type
Carbon Cycle
Woody and herb biomass
Soil organic matter
Climate
Biodiversity –
Ecosystem
Vegetation type
Vegetation cover
Pattern of vegetation
Naturalness; management history, age, exotic
species Management objectives
Forest condition (rate of change)
Biodiversity - Species
Species composition (including rare species)
Species richness (indicator species)
Pattern (corridors / networks)
Threats to sp. diversity; human disturbance,
pollutant deposition, exotic species
Sustainability
Management objectives / history / planning and
Land use change
Similarities RS - LE
Similarities RS – Landscape Ecology approaches:
* Different processes at different levels;
different scales of observation are relevant
* Integrated (holistic) view
* Pattern does matter(!) – studies of vegetation patterns
* Dealing with spatial heterogeneity..
* Search for Self-similarity, as reflected in truly fractal patterns
* Analysis of scaling effects
* Minimum mapping unit: Grain = Pixel
What can RS do for forest ecology?
From RS to landscape monitoring and valuation
Adding value, refinement and compression of information
Process
steps:
Image
acquisition
Data
types:
Atmospheric
correction, geometric
correction, illumination
correction
“raw
images”
Derived
information:
Segmentation /
vectorisation / on-screendigitisation, Land Cover
classification
“orthophotos”
etc.,
rectified,
georeferenced
imagery
Change detection,
(based on spectral
Extent of rapid / disastrous
characteristics)
processes, such as active fires,
clear-cutting, oil spills etc.
Land
cover
maps
Applying criteria,
using knowledge
Landscape
type maps,
habitat type
maps,
“diversity
maps”
Area statistics,
Spatial metrics,
(input to) GIS
analysis
Habitat
suitability,
change
sensitivity
How to get to land cover maps 1
Vectorise/digitise
Aerial photo with shape file outline
Dominant vegetation type
assigned to each polygon
How to get to land cover maps 2
classify raster images
Landsat TM
bands 3,4,5
Forest/nonforest mask
Selected spatial metrics for measuring
fragmentation
The test case:
One land cover type, the rest “background”
Fragmentation the issue - edge, shape, patch number
M  10*
number of runs betweenforest and other cover type pixels
(number of forest pixels)* (total number of pixels)
PPU 
m
(n *  )
4* A
SqP  1 
P
1 
[2]
[3]
”Moving Windows” Approach
As implemented with calculation of Fragstats-derived
and other spatial metrics for “sub-landscapes”
INPUT: “cover type” map(1)
OUTPUT: metrics/index value map(2)

Calculate
(e.g.)
Patch type
Richness
1 2 3 4 5
Map 1:
Applied to
Grain = pixel size = 30m
Extent = 30*30 pix = 900*900 m
Window (user choice):
Size (extent) = 9 pixels = 270 m
Step = 3 pixels = 90 m
Determines
Map 2:
Grain = pixel size = 90 m
Extent = 8*8 pixels = 720*720 m
Spatial metrics maps from regional forest map
Maps of spatial metrics, from application of “moving windows”
Base = GIS layer of physiological
type from regional forest mapping:
Edge Density
Total (forest)area
Values of spatial metrics:
Core Area (TCAI)
LOW
HIGH
Diversity (SHDI)
Forest maps from satellite at different resolutions
Results, satellite images, land cover classification
TM, pixel size 25 m
WiFS, pixel size 200 m
Detected forest cover 44.9%
Detected forest cover 54.9%
50 km
Displaying landscape metrics
Large area maps...
WiFS
based
FMERS project
CORINE land cover
CORINE land cover reclassified to FMERS
nomenclature (6 forest classes)
Maps of metrics
Matheron
’fragmentation’ index
LOW
HIGH
Shannon – Simpson diveristy indices
Umbria, faunal observations
RED=low no. of species
GREEN= high no. of species
Observation per
species
BRIGHT
Summarization
over gridcells
Landscape metrics
calculated for
relevant cells, where
species are observed
Combination with RS based maps
Presence/non-presence in grid net
CORINE (100m pixels)
FMERS - WiFS (200m pixels)
Watershed-polygon-statistics example:
Umbria, mid-Italy, N and E of Assisi, the selected two 2nd order
catchments are part of the Tevero (Tiber) catchment (5th order).
Watershed mapping 1
Statistics from 2nd order watersheds
Region
Bands
AREA/km2
non-classified
Coniferous
Broadleaved Decid.
Broadl. evergreen
Mixed
OWL Coniferous
OWL Broadleaved
Other Land
Water
Cloud/Snow
0
1
2
3
4
5
6
7
8
9
2nd1014.shp
16821 pixels included
D:\Geodata\fmers_forestmap\FMERS_Central_other.img
672,84
%
128
0,76
766
4,55
2342
13,92
34
0,20
1226
7,29
157
0,93
688
4,09
11.480
68,25
0
0,00
0
0,00
0
1
2
3
4
5
6
7
8
9
2nd1042.shp
10458 pixels included
D:\Geodata\fmers_forestmap\FMERS_Central_other.img
418,32
%
16
0,15
1195
11,43
1040
9,94
9
0,09
899
8,60
695
6,65
1646
15,74
4958
47,41
0
0,00
0
0,00
Region
Bands
AREA/km2
non-classified
Coniferous
Broadleaved Decid.
Broadl. evergreen
Mixed
OWL Coniferous
OWL Broadleaved
Other Land
Water
Cloud/Snow
..calculated indices can be written
’back’ as parameter of WS polygon
Watershed mapping 2
Region
Bands
AREA/km2
non-classified
Coniferous
Broadleaved Decid.
Broadl. evergreen
Mixed
OWL Coniferous
OWL Broadleaved
Other Land
Water
Cloud/Snow
0
1
2
3
4
5
6
7
8
9
1st5224.shp
1006 pixels included
D:\Geodata\fmers_forestmap\FMERS_Central_other.img
40,24
%
0
0,00
19
1,89
3
0,30
0
0,00
112
11,13
0
0,00
73
7,26
799
79,42
0
0,00
0
0,00
Region
Bands
AREA/km2
non-classified
Coniferous
Broadleaved Decid.
Broadl. evergreen
Mixed
OWL Coniferous
OWL Broadleaved
Other Land
Water
Cloud/Snow
0
1
2
3
4
5
6
7
8
9
1st5230.shp
869 pixels included
D:\Geodata\fmers_forestmap\FMERS_Central_other.img
34,76
%
1
0,12
27
3,11
0
0,00
0
0,00
9
1,04
16
1,84
21
2,42
795
91,48
0
0,00
0
0,00
Region
Bands
AREA/km2
non-classified
Coniferous
Broadleaved Decid.
Broadl. evergreen
Mixed
OWL Coniferous
OWL Broadleaved
Other Land
Water
Cloud/Snow
0
1
2
3
4
5
6
7
8
9
1st5217.shp
3166 pixels included
D:\Geodata\fmers_forestmap\FMERS_Central_other.img
126,64
%
1
0,03
523
16,52
338
10,68
0
0,00
524
16,55
224
7,08
750
23,69
806
25,46
0
0,00
0
0,00
http://www.europa.eu.int/comm/agriculture/publi/landscape/ch4.htm
Further work..
* Apply the spatial metrics land cover maps derived using more
sophisticated methods, e.g. edge preserving smoothing,
segmentation and/or neural networks.
* Multiple regression of metrics such as the ones studied here or
other parameters describing ecological conditions.
* Verify how indices derived from classifications of aerial photos of
the area (preferably ~1 m resolution), relate to satellite data.
* Comparison with CORINE land cover data, taking into
account that:
- Coverages are not regularly updated (not to be used for
monitoring)
- The dataset is originally vector based, some information
is lost when converted to raster format, not intended to be
used as a pixel based land-cover mask.
Conclusions
- Remote sensing provides synoptic images at different scales,
potentially making it a powerful tool for applications in multiscale landscape analysis.
- The role of Remote Sensing and other Earth Observation
techniques concerning forest management is to complement
other information sources and inventories done by specialised
researchers on the ground.
- GIS is an adequate tool for combining information stored
in data-bases, map information and EO-data.
- Moving Windows approaches can provide information on
landscape sturcture and forest diversity over large areas –
illustrating distributions and highlighting ’hot-spots’.
- (-Infinitely) Many spatial metrics can be calculated from
EO-data, but connections with ecological conditions must be
established and their use verified.
RS – spatial metrics
Do spatial metrics fit in somewhere?
Sketch of Terra satellite ©NASA, 2000
Digital
Imagery
Thematic
Maps
Land Use
Planning
Decision
making
Administration
Spatial
Metrics
Indicators
Earth
Observation
Monitoring
Inventories
Trad. Forestry /
Ecological Environmental
Land/Forest
Management
Future options
& research needs
* Development of methods for detection of areas
threatened or in need of special management
techniques/consideration.
* Satellites with higher spatial resolution +
satellites with multi-spectral sensors – extended
spatial and spectral domains.
* Still a need for better understanding of how to
relate spatial/textural measures/information from
high resolution to medium scale spectral and/or
spatial information.
* Watersheds as natural regions for calcultaion
and reporting of spatial/structural landscape
properties...