Global Observation of Forest Cover GOFC: Fire

1
Global Observation of Forest Cover (GOFC) Fire
S. E. Asia Fire Cal/Val Workshop A special
session of the APAN Earth Monitoring Working
Group Summary of January 22 session
2
Attendees Pakorn Apaphant John Owens Chaowalit
Silapathong Veerachai Tanpitat Treechada
Tawornmas Horacio T. Cadiz Krisanadej
Jaroensutasinee Peera Chernchit Agus
Hidayat Chanchai Sitabutr Mastura
Mahmud Alias Mohm Sood Khen Rong Den Pisuth
Paiboonrat Do Xuan Lan Soo Chin Liew Panarat
Sareewattana Panjit Musik Renato L. Lapitan Bruno
Margerin
3
Background to GOFC

• Developed originally under the Committee on Earth
  Observation Satellites (CEOS) initially as a
  pilot to test the concept of an Integrated Global
  Observing System
• IGOS goals
• To improve use of Earth Observation data to
  address major problems of global concern
• To improve coordination of national programs
• To improve co-operation between providers and
  users of Earth Observation data for regional and
  global applications.
• Has become one of the Panels of the Global
  Terrestrial Observing System GTOS (FAO GTOS
  Secretariat)

4
The Organization of GOFC
Implementation Teams, Activities and Projects
Fire Monitoring and Mapping??.. Cover
Characteristics and Changes.. Biophysical
Parameters?????
Scientific and Technology Board
Regional Activities and Networks
Collaborations e.g. WGISS and WGCV
5
GOFC Fire Implementation Team

• The GOFC Forest Fire Monitoring and Mapping
  Implementation Team is responding to a critical
  need by fire management agencies, international
  agencies, and policy makers at national,
  regional and global levels, for accurate and
  timely information regarding wildfires in
  forestland and other vegetated areas.
• The Fire IT consists of a number of contributory
  activities from different research groups and
  organizations, which demonstrate the operational
  utility of satellite observations of fire, making
  the case for long-term observations of fire and
  the research and development necessary to
  improve these capabilities to meet user needs.

6
Fire in S.E. Asia

• Fire affects all of the nations in S.E. Asia.
• Nearly all the fires in the region are man
  made.
• During dry years fires can rage uncontrollably.
• Fire is widely used to dispose of agricultural
  residues, short term enhancement of crop growth
  conditions, and in the conversion of forests to
  agricultural land.
• Fire is one of the major factors in the region
  associated with global change.
• Impacts include transboundary smoke, loss of
  forests, loss of biodiversity, land degradation.
• Fire management is now recognized as an
  important function of government, requiring local
  and regional communication and cooperative
  effort.
• Fire management includes prediction of risk,
  prevention, detection, suppression, post fire
  assessment and recovery.

7
Workshop Objectives While a number of resource
management and scientific applications for
satellite fire observations have been identified,
little is known about the accuracy and validity
of the various data sources and products or how
they might be used together for an improved
depiction of fire events. The objective of the
workshop is to initiate and plan the
intercomparison, calibration, and validation of
active fire and burn scar products from primary
sensor systems capable of making frequent wide
area observations of S.E. Asia. In addition, the
topics of building a regional capability to
generate satellite fire products and widening the
use of these products through open access and
interactive overlay of multiple data sources will
be explored.
8
Why Accuracy Assessment? "There are many reasons
for performing an accuracy assessment. Perhaps
the simplest reason is curiosity - the desire to
know how good something is. In addition to the
satisfaction gained from knowledge, we also need
to increase the quality of the map information by
identifying and correcting the sources of errors.
Third, analysts often need to compare various
techniques, algorithms, analysts, or interpreters
to test which is best. Finally, if the
information derived from the remotely sensed data
is to be used in some decision-making process,
then it is critical that some measure of its
quality be known." Quotation from " Assessing
the Accuracy of Remotely Sensed Data Principles
an Practices", Russell G. Congalton and Kass
Green, 1999, Lewis Publishers.
9
Importance of Accuracy Assessment for Fire and
Burn Scar Products Documenting the detection
limits for various sensors. Understanding the
errors for each sensor. Improving the
algorithms. Developing confidence in the use of
the data. Developing calibrations for more
accurate input into models. Others.....
10
Initial set of objectives of GOFC cal/val in S.E.
Asia. A) What are the practical reference data
sources, their collection methods, uses, and
limitations. B) How well do active fire
detections from course resolution environmental
satellites (MODIS, AVHRR, DMSP and others) match
the location and extent of fires found in
reference data. C) How well do burn scar
products from course resolution environmental
satellites match the location and extent of burn
scar found in reference data. D) How well do
fire risk predictions match the fire observations
(active fires and burn scars).
11
Review of satellite data and product sources for
fire and burn scar monitoring in S.E. Asia Who
has what data? What products are being
generated? How are the data distributed or
accessed? Develop an inclusive table
summarizing as many of the data sources and
activities as possible.
12
Positional Accuracy ReferenceU.S. Federal
Geographic Data Committee (FGDC) Geospatial
Positioning Accuracy Standards Part 3 National
Standards for Spatial Data Accuracy (NSSDA) The
NSSDA uses root-mean-square error (RMSE) to
estimate positional accuracy. RMSE is the square
root of the average of the set of squared
differences between dataset coordinate values and
coordinate values from an independent source of
higher accuracy for identical points. http//www.f
gdc.gov/standards/status/sub1_3.html
13
Thematic Accuracy Reference "Assessing the
Accuracy of Remotely Sensed Data Principles an
Practices", Russell G. Congalton and Kass Green,
1999, Lewis Publishers. "The purpose of
quantitative accuracy assessment is the
identification and measurement of map errors.
Quantitative accuracy assessment involves the
comparison of a site on a map against reference
information for the same site. The reference data
is assumed to be correct. Because comparison of
every spatial point is impractical, sample
comparisons are used to estimate the accuracy of
maps."
14
For the thematic accuracy assessment of active
fires what types of reference data could be
used? The best reference data would be
simultaneously (or near simultaneously) satellite
observations at much finer spatial resolution.
Ideally this source exhibit many fires, allowing
detection limits in the coarse resolution data
source to be analyzed. E.g. MODIS ASTER
(simultaneous), 10 am Landsat and 10 am MODIS,
10 pm Landsat and 9 pm DMSP. Accuracy
assessment of cumulative fire detections
performed using the burn scar increment is flawed
due to the fact that much of the burning is
unobserved due the the discrete nature of the
active fire observations (once or twice per
day). Field data
15

• For the thematic accuracy assessment of burn
  scars
• what types of reference data could be used?
• Burn scars derived for the same time period
  from fine
• resolution satellite imagery (ETM, SPOT, ASTER,
  BIRD)
• Field data

16

• Pilot Study - Datasets assembled
• Landsat 7 ETM of Landsat Path/Row 131/47 February
  2, 2000 and March 5,2000
• AVHRR fire detections from World Fire Web,
  Bangkok Node (http//wfw-thailand.acrors.ait.ac.th
  /wfw/)
• ATSR fire detections from World Fire Atlas
  (http//shark1.esrin.esa.it/ionia/FIRE/AF/ATSR/)
• TRMM VIRS fire detections from Louis Giglio
  (NASA-GSFC)
• DMSP-OLS fire detections from NGDC  

17

• Pilot Study Data Analaysis
• Burn scar map produced for each date.
• Color image generated to show new burn scar added
  from February 2 to March 5, burn scar present in
  both dates, and burn scar found in February 2
  image but not in the marhc 5 image.
• Fire pixels overlayed on burn scar image.
• Preliminary Results and Conclusions
• TRMM VIRS and ATSR had very few fire detects, bu
  associated with burn scars, but did appear to be
  associated with burn scar.
• AVHRR had a moderate number of fire detections,
  and were geenerally associated with burn scars.
• DMSP had a large number of fire detects, most of
  which were associated with burn scars.
• Fresh burn scars tend to have high DN in TM
  bands 5 and 7 and low
• DN in bans 1,2,3 and 4.

18
Regional and Technology Presentations CRISP
experiences in active fire and burn scar cal/val
(Dr. Liew) Research results and activities
related to forest fires in Indonesia (Mr. Agus
Hidayat, Indonesia) Fire cal/val activities and
plans for the Huai-Kha-Keang Wildlife Sanctuary,
Thailand (Veerachai Tanpipat and Dr. Kiyoshi
Honda, Thailand) Options for on-line access to
regional fire data (John Owens, University of
Maryland). Satella A tool for peer-to-peer
Landsat data sharing (Mr. Bruno Margerin, USA)
19
Two Sets of GOFC Cal/Val Activities Planned for
2002 Thailand (January March) Indonesia
Malaysia (May September)
20
Thailand (January March) 1.Landsat Path/Row
131/47 (Chang Mai) used in CEOS WGISS Demo and
Pilot Study. Daytime Landsat acquisitions
scheduled on 1/31, 2/16, ¾, and 3/20. Nighttime
Landsat acquisitions on 2/8, 2/24, 3/12 and
3/28. 2. Landsat Path/Row 130/49 (Huai-Kha-Keang
Wildlife Sanctuary). RFD is collecting extensive
field data on active fire and burn scar timing
and location. Daytime Landsat acquisitions
scheduled on 2/7, 2/23, 3/11, 3/27.. Nighttime
Landsat acquisitions on 2/8, 2/24, 3/12 and
3/28. Other requested data ASTER, MODIS fire and
burn scar (Umd/NASA), AVHRR fire detects (AIT),
DMSP fire detects (NOAA-NGDC), fire risk map
(TBD). Analysis plan Active fires from Landsat
and ASTER data will be used to perform an
accuracy assessment of fires detected in MODIS,
AVHRR, and DMSP-OLS. The two best Landsat scenes
will be used to perform an accuracy assessment of
a MODIS burn scar product for the same time
period Data hosting for ftp access AIT?
21
Indonesia - Malaysia (May-September) 1. Riau,
Sumatra 2. Selangor, Malaysia Data sets to be
requested Day and night Landsat ETM, SPOT
(CRISP), ASTER, MODIS fire and burn scar (CRISP),
AVHRR fire detects (multiple sources), DMSP fire
detects (NOAA-NGDC), FDRS fire risk maps, LAPAN
fire risk maps . Analysis plan 1) Active fires
from Landsat and ASTER data will be used to
perform an accuracy assessment of fires detected
in MODIS, AVHRR, and DMSP-OLS. 2) Active fires
detections will be compared against fire risk
maps. 3) The two best Landsat scenes will be used
to perform an accuracy assessment of a MODIS burn
scar product for the same time period Data
hosting for ftp access UKM?
22
Outstanding Issues 1. How can we determine the
accuracy of burn scar maps to be used as
reference data sources? Questions were raised in
the pilot that dry grass and clearings were
mapped as burn scar. 2. How do we know how many
active fires are missed in reference data sets,
especially daytime Landsat and ASTER? 3. Should
we have a follow on workshop. If so, when and
where? APAN 2003?