Transcript From Pixels to Processes: Detecting the Evolution of

Department of Geography, SUNY Bufallo,
February 2007
From Pixels to Processes:
Detecting the Evolution of
Agents in a Landscape
Gilberto Câmara
Director
National Institute for Space Research
Brazil
Knowledge gap for spatial data
source: John McDonald (MDA)
The way remote sensing data is used

Exctracting information from remote sensing imagery


Recipe analogy





Most applications use the “snapshot” paradigm
Take 1 image (“raw”)
“Cook” the image (correction + interpretation)
All “salt” (i.e., ancillary data)
Serve while hot (on a “GIS plate”)
But we have lots of images!

Immense data archives (Terabytes of historical images)
The challenge of remote sensing data
mining

How many cutting-edge applications exist for extracting
information in large image databases?

How much R&D is being invested in spatial data mining
in large repositories of EO data?

How do we put our image databases to more effective
use?
Land remote sensing data mining:
A GIScience view

A large remote sensing image database is a collection of
snapshots of landscapes, which provide us with a unique
opportunity for understanding how, when, and where
changes take place in our world.

We should search for changes, not search for content

Research challenge: How do model land change for data
extracted from a land remote sensing database?
MSS – Landsat 2 – Manaus(1977)
TM – Landsat 5 – Manaus (1987)
Can we avoid that this….
Source: Carlos Nobre (INPE)
Fire...
….becomes this?
Source: Carlos Nobre (INPE)
Dynamic areas (current and future)
New Frontiers
INPE 2003/2004:
Intense Pressure
Future expansion
Deforestation
Forest
Non-forest
Clouds/no data
Modelling Land Change in Amazonia

How much deforestation is caused by:






Soybeans?
Cattle ranching?
Small-scale setllers?
Wood loggers?
Land speculators?
A mixture of the above?
Agent-based models

Recent emphasis on agent-based modeling for
simulation of social processes.

Simulations can generate patterns similar to real-life
situations

How about real-life modelling?

We need to be able to describe the types of agents that
operate in a given landscape.
Extracting Land Change Agents from
Images

Land change agents can be inferred from land change
segments extracted from remote sensing imagery.

Different agents can be distinguished by their different
spatial patterns of land use.

This presentation


Description of methodology
Case studies in Amazonia
Research Questions

What are the different land use agents present in
the database?

When did a certain land use agent emerge?

What are the dominant land use agents for each
region?

How do agents emerge and change in time?
Challenge: How do people use space?
Soybeans
Loggers
Competition for
Space
Small-scale Farming
Source: Dan Nepstad (Woods Hole)
Ranchers
What Drives Tropical Deforestation?
% of the cases
 5% 10% 50%
Underlying Factors
driving proximate causes
Causative interlinkages at
proximate/underlying levels
Internal drivers
*If less than 5%of cases,
not depicted here.
source:Geist &Lambin
Different agents, different motivations

Intensive agriculture (soybeans)



export-based
responsive to commodity prices, productivity and transportation
logistics
Extensive cattle-ranching


local + export
responsive to land prices, sanitary controls and commodity
prices
photo source: Edson Sano (EMBRAPA)
Large-Scale Agriculture
Agricultural Areas (ha)
1970
Legal Amazonia
Brazil
1995/1996
%
5,375,165
32,932,158
513
33,038,027
99,485,580
203
Source: IBGE - Agrarian Census
photo source: Edson Sano (EMBRAPA)
Cattle in Amazonia and Brazil
Unidade
Amazônia Legal
Brasil
Fonte: PAM - IBGE
1992
29915799
154,229,303
2001
51689061
176,388,726
%
72,78%
14,36%
Cattle in Amazonia and Brazil
Unidade
Amazônia Legal
Brasil
1992
2001
%
29,915,799
51,689,061
72,78%
154,229,303
176,388,726
14,36%
Different agents, different motivations

Small-scale settlers




Wood loggers



Associated to social movements
Responsive to capital availability, land ownership, and land
productivity
Can small-scale economy be sustainable?
Primarily local market
Responsive to prime wood availability, official permits,
transportation logistics
Land speculators


Appropriation of public lands
Responsive to land registry controls, law enforcement
Landscape Analysis: Land units associated to
agents
Space Partitions in Rondônia
…linking human activities
to the landscape
Agent Typology: A simple example
Is it enough
to describe
Amazonian
land use
patterns?
Tropical Deforestation Spatial Patterns: Corridor,
Diffuse, Fishbone, Geometric (Lambin, 1997)
Landscape Ecology Metrics


Patterns and differences are immediately recognized by
the eye + brain
Landscape Ecology Metrics allow these patterns in
space to be described quantitatively
Source: Phil Hurvitz
23
Fragstats (patch metrics)
(image from Fragstats manual)
24
Some patch metrics

PARA = perimeter/area ratio

SHAPE = perimeter/ (perimeter for a compact region)

FRAC = fractal dimension index

CIRCLE = circle index (0 for circular, 1 for elongated)

CONTIG = average contiguity value

GYRATE = radius of gyration
25
1975
1986
Increased
fragmentation
1992
on Rondonia, Brazil
Region-growing segmentation
Remote sensing image mining
Patterns of tropical deforestation (example
1)
Patch metrics for example 1
Decision tree classifier

C4.5 decision tree classifier (Quinlan 1993).

Each node matches a non-categorical attribute and each
arc to a possible value of that attribute.

Each node is associated the numerical attribute which is
most informative among the attributes not yet considered
in the path from the root.
Decision tree for patterns
metrics are: perimeter/area ratio (PARA) and fractal dimension (FRAC)
Validation set for decision tree (ex 1)
Validation showed 81% correctness
Incra settlement projects
Small, medium and large farms
Started in the 70’s
Case Study 1:Rondônia
Different spatial and temporal patterns
Lots size of 25 ha to 100 ha – Farms from 500
ha.
Cattle ranching
Objective: To capture patterns and to
characterize and model land use change
processes
Escada, 2003.
Prodes (INPE, 2000)
TM/Landsat, 5, 4, 3 (2000)
Spatial patterns in the Vale do Anari
irregular, linear, regular
Land use
patterns
Spatial
distrib
ution
Clearing
size
Actors
Main land use
Description
Settlement parcels less
than
50
ha.
Deforestation
uses
linear
patterns
following government
planning.
Linear (LIN)
Roadside
Variable
Small
households
Subsistence
agriculture
Irregular
(IRR)
Near main
Settlement
main
roads
Small
(< 50 ha)
Small farmers
Cattle ranching Settlement parcels less
and
than 50 ha. Irregular
subsistence
clearings near roads
agriculture
following settlement
parcels.
Regular
(REG)
Near main
Settlement
main
roads
Mediumlarge
(> 50 ha)
Midsized and
large farms
Cattle ranching
Patterns produced by land
concentration.
Decision tree for Vale do Anari
Changes in Incra parcels
configuration by (Coy, 1987; Pedlowski e Dale,
1992; Escada 2003):
• Fragmentation
• Transference
• Land concentration
Vale do Anari – 1982 -1985
REG
Patterns/Typology
IRR: Irregular – Colonist parcels
LIN: Linear – roadside parcels
REG: Regular agregation parcels
Pereira et al, 2005
Escada, 2003
Vale do Anari – 1985 - 1988
REG
Pereira et al, 2005
Escada, 2003
Vale do Anari – 1988 - 1991
REG
Pereira et al, 2005
Escada, 2003
Vale do Anari – 1991 - 1994
Pereira et al, 2005
Escada, 2003
Vale do Anari – 1994 - 1997
REG
Pereira et al, 2005
Escada, 2003
Vale do Anari – 1997 - 2000
REG
Pereira et al, 2005
Escada, 2003
Vale do Anari – 1985 - 2000
REG
REG
Confirmed by
field work
Pereira et al, 2005
Escada, 2003
Marked land concentration
Government plan for settling many colonists in the area has failed.
Large farmers have bought the parcels in an illicit way
Case study 2: Xingi-Iriri watershed in the state of
Pará
Spatial patterns in the Xingu-Iriri region
linear, small irregular, irregular, medium regular, large regular
Land use
patterns
Spatial
distribution
Clearing
size
Variable
Actors
Main land
use
Small
Subsistence
household agriculture
s
Description
Linear
(LIN)
Roadside
Small
irregular
(SMALL)
Near main Small
settlements (< 35 ha)
and
main
roads
Irregular
(IRR)
Near main Small
Small
settlements (35 – 190 farmers
and
main ha)
roads
Cattle
ranching
Associated to small
family households
Medium
Regular
(MED)
Isolated or 190 – 900 Medium
near
ha
farmers
secondary
roads
Cattle
ranching
Associated
medium to
farms
Large
Regular
(LARGE)
Isolated or Large
at the end of (> 900 ha)
secondary
roads
Cattle
ranching
Isolated, may have
airstrips
Small
farmers
Large
farmers
Roadside clearings,
following
main
roads
Family
Near main roads
labour and and settlements up
cattle
to 10 Km.
ranching
to
large
Decision tree for Terra do Meio spatial patterns
Trend towards land concentration
where large farms dominate over small settlements.
Conclusions

Pattern classification in maps extracted from images of
distinct dates enables associating land change objects to
causative agent

Pattern classification techniques associated to remote
sensing image interpretation are a step forward in
understanding and modelling land use change.

Next step: develop agent-based models for deforestation
in Amazonia
References

Mining Patterns of Change in Remote Sensing Image
Databases.
Marcelino Silva, Gilberto Camara, Ricardo Souza, Dalton Valeriano,
Isabel Escada.
Fifth IEEE International Conference on Data Mining. Houston,TX,
USA, November 2005.

"Remote Sensing Image Mining: Detecting Agents of
Land Use Change in Tropical Forest Areas“
Marcelino Silva, Gilberto Câmara, Ricardo Souza, Dalton Valeriano,
Isabel Escada.
International Journal of Remote Sensing, under review (manuscript
available from the author).