The GLOBAL LAND COVER FACILITY: OVERVIEW John Townshend Department of Geography and Institute for Advanced Computer Studies University of Maryland
The GLCF… is a resource of information and data to researchers interested in land cover change. offers an increasingly comprehensive range of data products facilitating further land cover research. is a partnership between the Earth science and computer science communities – Earth sciences benefiting from advances in high performance computing. has in-house expertise in the analysis of remote sensing data making us a science data center not simply a purveyor of data. Who are we?
The GLCF emphasis has been: 1) supplying data to the science community ASAP, 2) providing a ‘one-stop-shop’ for finding land cover data, 3) developing novel tools for exploring and manipulating these data “Provide value-added Earth science products to the widest possible user base…” The GLCF Mission
GLCF » Currently the GLCF operates with a staff of employees, representing a mixture of full-time, part-time and student staff. GLCFGLCF » The GLCF is an end-to-end shop -- GLCF positions include full-time Earth science faculty, programmers, network developers, customer service representatives, data ingestion staff, data base developers, etc. The GLCF Staff
GLCF Collaborators A global network of science and technology partners
Global Land Cover Facility (G LCF) Creation and distribution of enhanced Earth Science Products: MODIS: Process, store and distribute 250m data and derived products User-defined custom products: from AVHRR GAC data archive Land cover data sets and related remotely sensed data: archive and distribute Landsat TM & ETM, MODIS, GAC products. Provide access to GOES derived products: Satellite Estimates of Radiative Fluxes Serving specific science groups: archive and distribute data sets for global modeling communities, EOS Validation sites for EOS scientists. Research activities: Advanced land cover products - global deforestation, global land cover classified data sets, continuous fields tree cover data, coastal marsh health data. Computer Science technologies – interoperability (MOCHA), data mining, high performance computing. Expanding user communities: RESAC - Broker and facilitate the acquisition of data for the Mid-Atlantic RESAC Field campaigns Support (data and RS expertise) for LBA-E, CARPE, and Safari2000 Products and services for UN Organizations, NGOs and Government Departments (e.g. FAO, UNEP, UNDP, UNDCP, CEOS, GOFC, USDA, USFS, WRI).
GLCF Hits/month
Distribution of products
Web 12 month summary of domains
FTP 12 month summary domains
Types of Data Downloaded
GLCF: Project Lessons Learned » Increased collaborations will lead to advances in the quality and quantity of earth science research and projects that the GLCF can support. » Challenges remain when using and distributing very large amounts of remote sensing data. » The needs of user groups vary and one must be engaged with these communities to adequately support their data needs. » Ensure that the project ‘knowledge base’ can be easily transferred in and among GLCF staff. » Not every technology or implementation strategy we use works as expected.
Carbon related activities John Townshend
What is a forest? FAO 1990: tree canopy cover > 20% in developed countries and > 10% in developing countries FAO 2000: tree canopy cover > 10% IGBP: tree canopy cover > 60% Forestry agencies: harvestable lands, actually or potentially
23 March 2001, p
Methodology for Global % Tree Cover Use continuous training data from high-resolution data sets scaled to MODIS resolutions Create temporal metrics describing vegetation phenology Employ regression tree using percent cover training as dependent variable and metrics as independent variables Current layers include tree, tree leaf type, tree leaf longevity, bare ground and herbaceous covers
Prototype AVHRR Continuous Field of Tree Cover % tree cover derived from km AVHRR (DeFries et al, 2000)
Comparison of FAO and UMD Forest Cover Estimates by Country
FAO estimate km DRC Namibia <10% >80% Tree cover % tree cover FAO estimate 1995 Possibility of standardizing global forest statistics % tree cover
Training from Landsat MODIS 250m U.S. Tree Cover Prototype 0% 100% Tree Cover For details see For details see
EPA Region 3 MRLC 30 meter map, green is forest, beige non-forest MODIS 250 meter map AVHRR 1km, Tree cover 0% 100% 100km The MODIS result compares quite favorably with the MRLC and is clearly superior to AVHRR heritage products.
Using Continuous Fields to Detect Forest Cover Change 0% >80% Tree Canopy Cover +30% -30% Percent tree cover ~1985Percent tree cover ~1995 Change in tree cover
Use of Landsat data for fine-scale mapping of forest disturbance EOS, 1 May 2001
Country forest statistics 200 km 0% 100% Tree canopy cover Argentina Paraguay Brazil Interior Atlantic Forest Ecoregion South American Ecoregions 1) 2) 3) Atlantic Ocean 200 km cyan = protected areas % Tree Cover Derived from AVHRR Data to Assess Forest Degradation within Ecoregions Brazil Argentina Paraguay R. DeFries and M. Hansen
25km UMD, GSFC and USFS MODLand Rapid Response collaboration: Low latency (1 to 6 hours from acquisition) provision of Active Fire and Reflectance Builds upon IMAPP L0-to-L1B software developed by Univ. of Wisconsin Operational prototyping of Linux system to create standard MODIS L2+ products R&D for enhanced VCC Burn Severity and Smoke / Aerosol products Assist USFS in procurement of MODIS Direct Broadcast receiving stations Transfer of Rapid Response software to USFS Remote Sensing Applications Center (Salt Lake City) Fire Science Lab (Missoula) 2001 bridging activity to provide MODIS products while DB stations are built Daily Reflectance and Active Fire mapped to standard NIFC geographic regions VCC Burned Area, ID/MT border 2000