
439f342dcfa9ba1af4c6c4284816b5e1.ppt
- Количество слайдов: 50
OPPORTUNITIES TO DEVELOP A 30 -METER SPATIAL DATABASE FOR THE USA JOINT MEETING OF THE MIDWEST FOREST MENSURATIONISTS AND THE ANNUAL FIA SCIENCE SYMPOSIUM Ray Czaplewski Forest Inventory and Analysis Program Fort Collins, CO
FIA’s Technical Vision for Remote Sensing 1. Chronology 2. Implementation 1992 1995 1997 1998 1999 2000 2001 2
First Blue Ribbon Panel on FIA should … have a • “FIA (should) preeminent position in all implement … remote federal efforts to sensing to accomplish inventory and a more inventory in monitor forest resource and cost-effective conditionsmanner; … ” efficient at the regional and national levels. '‘ 1992 3
MRLC 1992 Data Buy • MMulti-Resolution USGS builds a Land Characterization consortium of federal (MRLC) consortium programs to acquire standardized Landsat • UUSGS, EPA, NOAA, -5 imagery for the BLM, NASA, NPS, USA USDA, Forest Service 1992 4
National Land Cover Data 1992 • USGS EROS Data Center begins first successful effort to map the forest and land cover of the USA with Landsat satellite data from MRLC 1992 1995 2000 5
National Land Cover Data 1992 • Forested Upland – Deciduous Forest – Evergreen Forest – Mixed Forest • Woody Wetlands • Non-native Woody • Shrubland • Grasslands 1992 1995 • 14 categories for other types of land cover • 2 water classes • 3 barren classes • 3 urban classes • 5 agricultural classes • 1 other wetland class 2000 6
MRLC v. NLCD • MRLC: Multi-Resolution Land Characterization – Consortium to buy Landsat data and post-process into standardized national imagery • NLCD: National Land Cover Data – Consortium to produce national land cover map using standardized MRLC imagery 1992 1995 2000 7
MRLC 1992 & NLCD 1992 Forest Service joins MRLC 1992 to buy Landsat data FIA data from periodic surveys too out-ofdate in many States to use as training data for Landsat classifications. FIA does not join NLCD 1992 1995 2000 8
Office of Science and Technology Policy • “Increased availability and affordability of remotely sensed data requires improved collaboration and communication between agencies and programs …” 1992 1995 1997 9
Second Blue Ribbon Panel on FIA • FIA should “(i)mprove that Panel reemphasizes … efficiency … through a “FIA should … have cooperation with agencies preeminent position in all which have the remote federal efforts to inventory sensing expertise not and monitor forest available within the FIA the resource conditions at organization. ” national regional and levels. ” 1992 1995 1997 1998 10
1998 Farm Bill • “The Secretary shall … submit to Congress a strategic plan … that shall describe in detail … the process for employing … remote sensing … , and (its) subsequent use. ” 1992 1995 1997 1998 11
Rand Corporation policy analysis • “Decentralized forest monitoring efforts greatly complicate the data delivery process. ” • “Integrat(e) … FIA … and Multi-Resolution Land Characterization (MRLC)” consortium. 1992 1997 1998 1999 12
“Adopting an annual inventory system: user perspectives” • “More ambitious applications of remote sensing are … hindered by … lack of a farreaching technical vision (that) might lead to radical changes in the FIA system and yield significant improvements in information quality and costeffectiveness. ” December, 1999 97(12)11 -14 1992 1997 1998 1999 13
FIA Vision for Remote Sensing • FIA adopts performance standard to “go … operational with (satellite) remote sensing … by the end of … 2003. ” 1992 1995 1997 1998 1999 2000 14
FIA Vision for Remote Sensing • Traditional FIA statistics remain essential for strategic analyses Million of acres U. S. A North South West Timber land 159 201 143 Reserved 52 8 4 40 Other forest 1992 504 191 3 9 179 1995 1997 1998 1999 2000 15
FIA Vision for Remote Sensing • “At the national to However, “key regional scale …, analytical outputs distinguishing today are maps, amonglayers, 20 map 15 to or forest cover types is other spatial important. representations of information and complex models. ” 1992 1995 1997 1998 1999 2000 16
FIA Vision for Remote Sensing 1992 1995 1997 1998 1999 2000 17
FIA Vision for Remote Sensing • “(D)ata from various … “Reduc(e) … human • The “process must programs… linkages to intervention … provide can be (to combined more effectively achieve) the other spatial data sets, to answer a much broader cheapest and than such asquestionsfastest census range of way to produce demographics or (the any agency could tackle remote sensing) digital elevation alone. ” product models. ”. ” 1992 1995 1997 1998 1999 2000 18
FIA joins MRLC 2000 Consortium • FIA contributes $200, 000 in a federal partnership to build a Landsat 7 image-set for the entire USA. 1992 1995 1997 1998 1999 2000 19
FIA joins MRLC 2000 Consortium • Three seasons of Landsat 7 imagery 2000 -2001 – – – 1992 1995 1997 1998 Early season (green up) Peak greenness (summer) Late season (brown up) 1999 2000 20
FIA joins MRLC 2000 Consortium • • 1992 1995 1997 1998 Terrain corrected, one pixel spatial accuracy Radiometric calibration for seamless data across scenes 1999 2000 21
FIA joins MRLC 2000 Consortium • 1992 1995 1997 1998 $45 per CD, including 3 dates of calibrated Landsat 7 data and Digital Elevation Model (DEM) used for terrain correction 1999 2000 22
National Land Cover Data (NLCD 2000) • FIA considers working with USGS EROS Data Center to replace NLCD 1992 forest cover map of USA using MRLC 2000 new Landsat 7 ETM+ data • Annualized FIA data more valuable as training data for classification of Landsat data 1992 1995 1997 1998 1999 2000 23
FIA partnership with USGS • FIA and USGS conduct pilot studies on highly automated digital classification of forest types with MRLC 2000 database and FIA plot data. 1992 1995 1997 1998 1999 2000 2001 24
NLCD 2000 Classification Process MRLC Input Database Brightness Elevation Greeness Slope Wetness Position Texture Shape Aspect Soils Land cover classification rule-sets • Apply CART • If seasons of Digital <45 30 -m Digital Elevationthe to wall Derivativeswithin a STATSGO 3 bright >segmentation (digital art Coarse 30 & greenstate ofmodel resolution Variability of <25 Image • & wet > 30 & texture Model =for each pixel MRLC Input Elevation&Model. NRCS into Landsat ofslope >5 data soilsmoving window Database • & shape 3 7 adjacent pixels grouping ETM+ 5 x 5 data from • & the USA in elevation > 4500 FRAGSTATS “polygons”) Landsat over soil 3 &and soil 30 -m canopy>10 • & position < texture Tassel=Cap&Transformation to E. g. of polygon 2 indices • & aspect soil pixels north Forest= shape assigned compress data quantity depth pixel • Then Deciduous to each • CART model Regression to Training % Tree E. g. a square polygon is land cover predict more equation data Canopy likely a corn field than from Landsat type a forest predicting crown FIA and geospatial data stand cover from NRI Landsat IKONOS NLCD 2000 Land Cover 25
FIA partnership with USGS • Pilot study using FIA plots from NE and SRS FIA Units – 1100+ non-forest plots – 535 forested plots for training digital classifier – 134 forested plots for validation 1992 1995 1997 1998 1999 2000 2001 26
FIA partnership with USGS Preliminary accuracy results (Chesapeake Bay) Classification detail Accuracy Forest v. nonforest 95% +2% 3 MRLC forest type groups 80% + 2% 6 FIA forest type groups 65% + 5% 1992 1995 1997 1998 1999 2000 2001 27
FIA partnership with USGS • FIA decides to station an FIA scientist at USGS EROS Data Center in Sioux Falls SD to ? ? ? ? ? ? – Assure FIA has preeminent role in national mapping of forest cover – Improve coordination on NLCD 2000 database – Guard confidentiality of FIA plot locations 1992 1995 1997 1998 1999 2000 2001 28
Western Governors’ Association • 10 -year comprehensive strategy to improve prevention and suppression of wildfires, and reduce hazardous fuels 1992 1995 1997 1998 1999 2000 2001 29
Western Governors’ Association • Produce a geospatial database for largearea assessments of – – 1992 Communities at risk Current vegetative conditions Likelihood of severe wildland fire Threats to • • Key habitats Water quality, such as post-fire erosion Air quality Local economies 1995 1997 1998 1999 2000 2001 30
LANDFIRE Database • Produce higher-resolution maps to support more cost-effective implementation of the National Fire Plan and Western Governors’ 10 -year Comprehensive Strategy. • Better prepare for and allocate firefighting resources 1992 1995 1997 1998 1999 2000 2001 31
LANDFIRE Database • Map and prioritize areas for fuel reduction efforts • Determine potential impacts of these fire treatments on wildlife, fish and riparian areas. 1992 1995 1997 1998 1999 2000 2001 32
LANDFIRE Database Develop geospatial database at the 30 -m scale for the entire USA • Vegetation type • Forest structural-stage • Stand density 1992 1995 1997 1998 • • • Fire history Fuel loadings Wildlife habitats 1999 2000 2001 33
LANDFIRE Database • Proposal to develop 30 -m database with MRLC and FIA partnership as its foundation • LANDFIRE could fund considerably more detail on forest conditions in 30 -m geospatial database for the USA 1992 1995 1997 1998 1999 2000 2001 34
• • NLCD 2000 becomes a 30 -m database for USA Landsat imagery and derived indices Land cover and forest type classification Digital Elevation Models Coarse-scale geospatial data – Ecoregions – Climate – Soils (STATSGO) 35
GIS themes in 30 -m database Map more detailed types (15 -20) of forest cover using • NLCD data base • Landforms • FIA training data(? ) 36
GIS themes in 30 -m database Coarse-scale (e. g. , 1 -km) information to geospatial database • Bailey’s Ecoregion Sub-sections • Climate (DAYMET) daily means – – Precipitation Minimum and maximum surface air temperature Surface air humidity Incident shortwave radiation • Land ownership • Bureau of the Census population and housing density 37
GIS themes in 30 -m database • Soils (STATSGO) – Available water capacity – Soil organic carbon – Soil suitability for agriculture – Soil Texture – Soil Depth 38
GIS themes in 30 -m database Land-use • Probable urban areas – Buffered road network – Nighttime Lights of the World • Protected Areas Database • Urban/wildland interface zones 39
GIS themes in 30 -m database Biophysical Settings Model • 30 -m version of Potential Natural Vegetation – 30 -m Digital Elevation Model – Soil depth and texture (STATSGO) • Better identify areas with similar – Fire frequencies – Climatic regimes – Geological and topographical characteristics 40
GIS themes in 30 -m database Historical Natural Fire Regimes • GIS model predicting – Fire frequency – Fire severity • Inputs – Digital Elevation Model – Biophysical Settings 41
GIS themes in 30 -m database Subdivide general forest cover types into additional 30 -50 detailed forest types • Separate spectrally similar forest types using Biophysical Settings model (PNV) • >65% accuracy 42
GIS themes in 30 -m database • Remotely sensed estimates of tree/shrub cover • Structural stages – Open stands, small trees – Open stands, large trees – Closed stands, small trees – Closed stands, large trees • >70% accuracy 43
GIS themes in 30 -m database Model predictions of fuel loading • Climatic regime • % tree/shrub cover • Detailed forest type • Structural stage (open v. closed; largev. small trees) • Potential Natural Vegetation • 30 -m Digital Elevation Model 44
GIS themes in 30 -m database Value-added products from other partner programs using same database • GIS analyses of risk from insects and diseases • GAP wildlife habitat mapping (USGS BRD) • % imperviousness surfaces (EPA) 45
Caveat Intended for large assessment areas • National (e. g. , fuel treatment priorities) • Regional (e. g. , multiple States) • Large-areas – River basins – Ecological Provinces 46
Caveat • National 30 -m spatial database is a starting point for priority setting among smaller geographical areas • Local datasets are usually more accurate for local analyses 47
Conclusion • FIA mission “Make and keep current a comprehensive inventory and analysis of the present and prospective conditions of and requirements for the renewable resources of the forest and rangelands of the United States. " Partnerships among agencies and USFS programs for remote sensing and geospatial databases provide costeffective support to the FIA mission and users of FIA data. 48
Poster Session • A comparison of stratification effectiveness between the National Land Cover Data set (NLCD 1992) and photo-interpretation in western Oregon by Paul Dunham, Dale Weyermann and David Azuma (PNW-FIA) • Synergistic use of FIA plot data and Landsat 7 ETM+ images for large area forest mapping by Chengquan Huang (USGS EROS Data Center) and Andrew Lister (NE-FIA) 49
50
439f342dcfa9ba1af4c6c4284816b5e1.ppt