Session 5 Forestry and Change Detection CORSE 2000 June 26-29, 2000 University of Southern Mississippi Gulf Park Conference Center Daniel L. Civco LERIS / NRME University of Connecticut Storrs CT 06269 dcivco@canr.uconn.edu June 2000 Volume 98 Number 6 •Remote Sensing and Forestry: Collaborative Implementation for a New Century of Forest Information Solutions •Foresters' Roles in Remote Sensing •From Pixels to Decisions: Digital Remote Sensing Technologies for Public Land Managers Remote Sensing Data Sources and Techniques •Aerial Photography in the Next Decade •Digital Imaging Basics for Natural Resource Managers •Videography for Foresters •The Earth Observing System and Forest Management •Intermediate Multispectral Satellite Sensors •Selecting and Interpreting High-Resolution Images •Forest Information from Synthetic Aperture Radar •Lidar Remote Sensing for Forestry •Using Hyperspectral Data to Assess Forest Structure •Map Data in Support of Forest Management •Image and Spatial Analysis Software Tools •Field Applications for Statistical Data and Techniques •Integrating Data and Information for Effective Forest Management http://www.safnet.org/pubs/jof/index.html •Remote Sensing and Forestry: Collaborative Implementation for a New Century of Forest Information Solutions •Kathleen Bergen, John Colwell, & Frank Sapio “ ... new forest management paradigms and rapid technological advances together create an organizational and technological challenge, as well as a great opportunity for advancing forestry.” http://www.safnet.org/pubs/jof/index.html •From Pixels to Decisions: Digital Remote Sensing Technologies for Public Land Managers •Henry Lachowski, Paul Maud, and Norm Roller “… forest managers need information about the geospatial infrastructure, including the location, amount, and condition of the forest’s natural and cultural resources.” http://www.safnet.org/pubs/jof/index.html Deforestation is the permanent destruction of forests and woodlands. The increasing population requires greater food production - deforestation occurs as the forests are converted for agricultural and urban uses. In the past three decades one fifth of all tropical forests were lost. Currently, 12 million hectares of forests are cleared annually. Most deforestation occurs in the moist forests and open woodlands of the tropics. Deforestation http://ps.ucdavis.edu/classes/ire001/env/deforest.htm http://www.wri.org/forests/index.html Deforestation in the Tropics Landsat Image Overview Map Degree of Deforestation http://www.dpi.inpe.br/Amazonia/pg13.html Fragmentation • Habitat Fragmentation, Modification or Loss Sources of habitat fragmentation include: – Agriculture: Conversion of prairie and forest areas to intensive agriculture eliminates nesting cover. – Forestry: Harvesting and regeneration modify the forest landscape and alter the structural and plant species diversity. – Urbanization: Urban sprawl to accommodate a growing human population progressively consumes natural areas. – Linear development: Roads, pipelines and hydro rights-of-way open up previously difficult-to-access territory to human use. – Climate change: When growing conditions are altered, habitat availability is affected, especially for species at the edge of their range http://www.cws-scf.ec.gc.ca/canbird/pif/habitat.htm The Northeast Landscape In the beginning, there was forest... The Northeast Landscape After near total conversion to farmland, much forest has returned... The Northeast Landscape Now, farm and forest are being converted to developed land, particularly subdivisions. The Northeast Landscape Is urban sprawl, deforestation, and habitat fragmentation the future of the Northeast? The Power to Visualize wa*ter*shed n. 1. An area of land draining to a common outlet. ZZZZZ... The Power to Visualize wa*ter*shed n. 1. An area of land draining to a common outlet. HMMM... The Power to Visualize AWESOME ! AWESOME ! AWESOME ! The Power to Confuse What the…?! ? ? ? ? ? Picasso What is a Watershed? A Watershed is an area of land that drains to a single outlet. 3-D to drive home the point 3D Visualizations ADAR TM DEM Make the Obvious Even More Obvious Thematic Mapper Band 6, Thermal, Resampled to 30 Meter Resolution By enhancing visualization ... 3-D Surface of Temperature Differences Warmer Cooler What are heat sinks? What will reduce thermal gains? Forest Fragmentation Roads are built. Forest Fragmentation Developed areas follow. Forest Fragmentation Patches of contiguous forest become smaller. Forest Fragmentation Forest resources are fragmented. “Let me see this fragmentation” Multiresolution Comparison Impervious 10 80 1 1 30 130 meterADAR meter meter meter meter 10 80 meter 1 meter meter meter DOQ SPOT ADAR MSS overlay TM TMDOQ SPOT w/ w/ MSS w/ 7 w/ 4band impervious band from impervious impervious panchromatic impervious multispectral panchromatic multispectral multispectral planimetric overlay overlay overlay overlay data Steamboat Willie In 1928, Disney made history with the release of first talkie animation film Steamboat Willie featuring Mickey Mouse. .. and haven’t we come a long way since? Mt Fuji volcano flyby created completely from ASTER data http://terra.nasa.gov/Gallery/ Remote Sensing in Action: The ASTER Sensor Aboard Terra http://terra.nasa.gov/Gallery/ Remote Sensing in Action: The MODIS Sensor Aboard Terra http://terra.nasa.gov/Gallery/ Global Normalized Difference Vegetation Index (NDVI) http://terra.nasa.gov/Gallery/ Global Normalized Difference Vegetation Index (NDVI) http://terra.nasa.gov/Gallery/ Mount St. Helens April 1980 May1980 June 1980 http://edcwww.cr.usgs.gov/earthshots/slow/MtStHelens/MtStHelens Mount St. Helens 1973 1983 1988 1992 1996 Landsat MSS http://edcwww.cr.usgs.gov/earthshots/slow/MtStHelens/MtStHelens EarthShots http://edcwww.cr.usgs.gov/earthshots/slow/tableofcontents Rondonia, Brazil: 1975-1992 1975 1986 1992 http://edcwww.cr.usgs.gov/earthshots/slow/Rondonia/Rondonia Fires in Wyperfeld National Park, Victoria, Southeast Australia 1975 1985 1999 http://edcwww.cr.usgs.gov/earthshots/slow/Wyperfeld/Wyperfeld Fires in Wyperfeld National Park 1979 to 1997 http://edcwww.cr.usgs.gov/earthshots/slow/Wyperfeld/Wyperfeld Clearcutting Near Olympic National Park, WA http://svs.gsfc.nasa.gov/imagewall/LandSat/olympic.html Clearcutting Near Olympic National Park, WA 1986 1987 1988 1991 http://svs.gsfc.nasa.gov/imagewall/LandSat/olympic.html Clearcutting Near Olympic National Park, WA http://svs.gsfc.nasa.gov/imagewall/LandSat/olympic.html Clearcutting Near Olympic National Park, WA 1984 to 1995 http://svs.gsfc.nasa.gov/imagewall/LandSat/olympic.html Deforestation near Santa Cruz, Bolivia from 1984 to 1998 150 miles 200 miles http://svs.gsfc.nasa.gov/imagewall/LandSat/santa_cruz.html Urban Growth in the DC Area: 1973-1996 http://svs.gsfc.nasa.gov/imagewall/LandSat/dc_growth.html Urban Growth Histories BaltimoreWashington Corridor http://www.ncgia.ucsb.edu/projects/gig/ Urban Growth History Marlborough Subdivision Growth resac.uconn.edu Urban Growth Projections San Francisco Bay Area Eastern Pennsylvania http://www.essc.psu.edu/~dajr/chester/animation/ Even WE cause fragmentation Click photo File / Open Movie / Uconn from Towers.mov Resize Windows Pan Right-Left http://www.sp.uconn.edu/~wwwucimt/pano/images/towers.mov NAUTILUS Research Objectives Better land cover Sprawl metrics Forest fragmentation metrics Better impervious cover Landsat ETM+ Data 30 meter multispectral 15 meter panchromatic 60 meter thermal High Resolution Airborne & Satellite Data ADAR 5500 and IKONOS IKONOS coverage ADAR coverage IKONOS Acquired through the NASA Scientific Data Purchase Program ADAR 5500 Other Satellite Data ASTER MODIS Earth Observer 1 Hyperion ALI SPOT Basic Land Cover Characterization and Change Objectives • To identify and quantify general land cover change over a 25 year period • Perform classifications using traditional classification techniques Basic Land Cover Characterization and Change Procedures, Salmon River Watershed Identify best signatures from 1973MSS & 1985TM, Perform Maximum Likelihood Classifier MSS & TM ISODATA clustering into 200 clusters (for each date) 1973 1976 1978 1981 1983 1985 1988 1993 1995 HECTARES Urb/Bare Ag/Grass Forest 2484 3516 31298 2479 3917 30594 2612 3060 32173 3674 3024 30667 2407 3851 30890 3927 4151 29116 5292 3673 28152 5024 2483 29573 5110 4072 28059 Calculate category areas Label clusters into 7 classes Adjust classifications to remove unlikely changes due to classification error (i.e. urban to forest) Land Cover (7 classes) Basic Land Cover Characterization and Change Data, Salmon River Watershed April 24, 1973 resampled MSS May 5, 1976 resampled MSS October 31, 1978 resampled MSS March 4, 1981 resampled MSS April 18, 1983 resampled MSS May 4, 1988 TM May 8, 1995 TM April 26, 1985 TM April 25, 1993 TM Basic Land Cover Characterization and Change Results to Date, Salmon River Watershed April 24, 1973 resampled MSS May 5, 1976 MSS October 31, 1978 MSS March 4, 1981 MSS April 18, 1983 MSS May 4, 1988 TM May 8, 1995 TM April 26, 1985 TM April 25, 1993 TM Forest Fragmentation and Urban Sprawl Objectives • To develop a practical method for assessing forest fragmentation and urban sprawl • Use ArcView GIS and extensions exclusively Forest Fragmentation and Urban Sprawl Procedures ISODATA clustering into 100 clusters Label clusters into 7 classes (ArcView Image Analyst) Spring & Summer TM Images Total Class Area (ha) Class 1985 1995 URBAN 1,495 2,578 FOREST 30,565 29,751 GRASS 3,909 3,713 Number of Patches (n) 1985 1995 812 1,856 2,524 2,785 3,823 3,463 Derive Spatial Statistics Mean Patch Size (ha) 1985 1995 2.42 1.68 11.97 10.52 0.99 1.05 (ArcView Spatial Analyst) Land Cover (7 classes) Forest Fragmentation and Urban Sprawl Data, Salmon River Watershed Landsat TM August 28, 1995 Landsat TM August 9, 1985 Landsat TM May 8, 1995 Landsat TM April 26, 1985 Forest Fragmentation and Urban Sprawl Results to Date, Salmon River Watershed 1985 Land Cover 1995 Land Cover Detail Areas of Change 1985 1995 Land Cover 1995 1985 Land Cover Forest Fragmentation and Urban Sprawl Results to Date, Salmon River Watershed Urban Land Gains: 1985 to 1995 (Hectares) 42.9 7.7 Water 353.4 Forest Wetland Grassland 36.4 892.4 Barren Total urban land gain 1,333 ha - Change to Urban Land - Other Change - No Change Forest Fragmentation and Urban Sprawl Results to Date, Salmon River Watershed Forest Land Losses: 1985 to 1995 (Hectares) 63.9 68.7 345.5 Water Wetland 853.0 Urban Grassland Barren 892.4 Total forest land loss 2,223 ha - Change from Forest Land - Other Change - No Change Forest Fragmentation and Urban Sprawl Results to Date, Salmon River Watershed Grassland Losses: 1985 to 1995 (Hectares) 47.7 2.8 Water Forest 353.4 Wetland Urban 5.9 801.5 Barren Total grassland loss 1,211 ha - Change from Grassland - Other Change - No Change Forest Fragmentation and Urban Sprawl Results to Date, Salmon River Watershed Total Class Area (ha) Class 1985 1995 URBAN 1,495 2,578 FOREST 30,565 29,751 GRASS 3,909 3,713 Number of Patches (n) 1985 1995 812 1,856 2,524 2,785 3,823 3,463 Mean Patch Size (ha) 1985 1995 2.42 1.68 11.97 10.52 0.99 1.05 Forest Fragmentation and Urban Sprawl Results to Date, Salmon River Watershed GRASSLAND 1985, 10.1 % of the total area 1995, 9.6 % of the total area FOREST 1985, 79.2 % of the total area 1995, 77.1 % of the total area URBAN 1985, 3.9 % of the total area 1995, 6.7 % of the total area Image Visualization Research Temporal Image Sequencing Objectives • To create a visual demonstration of actual change occurring in the landscape through the use of animations Image Visualization Research Temporal Image Sequencing Procedures, Histogram Matching 1995 TM Linear Transforms 1995 TM Non-linear transform Histogrammatched 1985 TM ER Mapper Algorithm Original 1985 TM Image Visualization Research Temporal Image Sequencing Procedures, Interpolation and Movie Creation 1990 TM .GIF Movie Creator 1991 Interpolated TM Formula Editor [((1990 * 4) + (1995 * 1))/5] = 1991 1995 TM .AVI Format Movie Image Visualization Research Temporal Image Sequencing Data, springtime Landsat imagery April 24, 1973 resampled MSS May 5, 1976 resampled MSS April 18, 1983 resampled MSS April 26, 1985 TM May 4, 1988 TM May 8, 1995 TM Image Visualization Research Temporal Image Sequencing Data, summertime Landsat imagery August 9, 1985 TM August 30, 1990 TM August 28, 1995 TM August 31, 1999 TM Image Visualization Research Temporal Image Sequencing Springtime 1973-1995 Animation Image Visualization Research Temporal Image Sequencing Summertime 1985-1999 Animation Intertnet Sources of Forestrelated Information http://www.wri.org/gfw/ http://www.fanweb.org/index.shtml/ http://www.forestwatch.sr.unh.edu/ Where Can You Find Additional Educational Resources on Remote Sensing? … how about the Internet? ASPRS Remote Sensing Core Curriculum http://research.umbc.edu/~tbenja1/index.html NASA On-Line Remote Sensing Tutorial http://rst.gsfc.nasa.gov/ Canada Center for Remote Sensing Fundamentals http://www.ccrs.nrcan.gc.ca/ccrs/eduref/tutorial/indexe.html Where Can You Find Digital Remote Sensing Image Data For Forest Characterization ? … how about the Internet? North American Landscape Characterization (NALC) 1973 1980 1990 Landsat MSS Triplicates http://edcdaac.usgs.gov/pathfinder/pathpage.html#nalc North American Landscape Characterization (NALC) 1990 DEM Landsat MSS and DEM http://edcdaac.usgs.gov/pathfinder/pathpage.html#nalc TerraServer http://www.terraserver.com/ Global Land Information System Landsat TM August 20, 1998 http://edcwww.cr.usgs.gov/webglis March http://landsat7.usgs.gov/order.html http://edcimswww.cr.usgs.gov/pub/imswelcome/ 7, 2000 ETM+ Other Sites for Data http://images.jsc.nasa.gov/iams/html/ http://terra.nasa.gov/ http://svs.gsfc.nasa. gov/imagewall.html … or Visit NASA’s RESAC at UConn http://resac.uconn.edu … where you’ll find ... .. and ... http://resac.uconn.edu Research & Education Watersheds Presumpscot SuAsCo Salmon Stonybrook A range of land covers and issues Salmon River Watershed, CT • Focused watershed for NAUTILUS research • CES program and research already existing in watershed • Rapid Urbanization • State highway connects with major Hartford market • 5 out of 7 watershed towns are listed as the fastest growing towns in the State • Key component of the lower Connecticut River Watershed 140 Sq. Miles Stony Brook Millstone Watershed, NJ • Has a strong Watershed Association in existence • Between New York City and Philadelphia • Increased development pressures • Loss of agriculture land to urban sprawl 265 Sq. Miles SuAsCo Watershed, MA Sudbury Assabet Concord Watershed • Has a strong Watershed Coalition in existence • Between the Boston metropolitan region and Worcester • Rapid development • New residential development replacing forests • All river segments are Class B waters 377 Sq. Miles Presumpscot Watershed, ME • Existing NEMO Program • Focus is on the lower portion of Presumpscot River • Most urban and rapidly developing region in the State • Significant water quality problems in the lower portion • Adjacent Harraseeket River coastal watershed, which contains Freeport, is included in NAUTILUS Project 200 Sq. Miles Alternative/Emerging Classification Approaches Knowledge Based Expert Systems, Procedures Decision Tree Source Data and Derivatives Final Classifications Alternative/Emerging Classification Approaches Neural Networks, Procedures Input Layers Output Layer Neural Network Training May 8, 1995 TM Back-propagation Neural Network Classification Principal Components Neural Network Based Land Cover Change Proof of Concept Study Backpropagation neural network Forest to nonforest land cover change …. In tomorrow’s breakout session we’ll look at ... Land Cover Mapping and Change Detection Using ArcView Session 5 Forestry and Change Detection CORSE 2000 June 26-29, 2000 University of Southern Mississippi Gulf Park Conference Center Daniel L. Civco LERIS / NRME University of Connecticut Storrs CT 06269 dcivco@canr.uconn.edu