Monitoring Trends in Land Change? The Big Tent – MTLC
A hierarchical land cover and land use change monitoring system that leverages existing projects and programs of work NLCD 2011 Tree Canopy Cover
Research & Development, Quantitative Sciences Remote Sensing Applications Center Rocky Mountain Research Station, FIA Southern Research Station, FIA Northern Research Station, FIA Pacific Northwest Research Station, FIA State and Private Forestry, Forest Health Protection USGS, EROS What is the NLCD and Who are the Clients?
. NLCD is the National Land Cover Database: . Land cover classification layer, percent tree canopy cover layer, and a percent impervious surface layer. Primarily based on LANDSAT (30meter pixels) imagery and ancillary data. . Produced by the Multi-Resolution Land Cover (MRLC) consortium. . Available free from http://www.mrlc.gov/ . The MRLC is a consortium of the following agencies and programs: These are the clients of the NLCD Percent Tree Canopy Cover is Important ! (Example of the NLCD 2001 Percent Tree Canopy Layer)
.An integral part of both international and US forest land definitions . Important both within forest land areas and in areas not traditionally considered forest. .Irrespective of land use , it’s an additional dimension of fragmentation .Knowing where trees are (not just the forest) is an important first step in quantifying carbon and managing tree resources. The Opportunity
. Motivation for Forest Service and FIA Leadership . If it’s related to trees, the Forest Service should be saying it . FIA is a fundamental component of Forest Service research. . FIA is a data rich program: . Consistency between map based and plot based estimates: . If needed, the FIA survey design is easily intensified . How are we positioned ? . Implementation of tree canopy cover estimates on at all sampling locations. . Experience with cutting edge modelling techniques . Biomass map . Forest type map update . Imputation approaches for the Atlas project NLCD Pilot Study Design 4x Intensity Photo-based Sample Locations
105 photo points to estimate % tree canopy cover for model development 2011 General Modelling Approach
= Response developed by photo Interpreting Tree crown cover on NAIP Imagery for ~4160 8100m2 sampling chips per Landsat scene.
Fig. from Homer et al. 2007
Random forests Example modelling Stochastic gradient boosting techniques Support vector machines Pilot Phase - Key Questions . Research on alternative pixel-level modelling techniques, alternative stratification/grouping strategies, using ordinal data for developing model, and model stability under different sampling intensification levels. (Moisen et al.,Tipton et al., Coulston et al.) . Research on the impact of scale of observation on tree canopy cover estimates. Relationship among plot based, PI based, and modeled estimates (Toney et al. 2009) at multiple scales. (Toney et al., Frescino et al., Gatziolis et al.) . Research on the impact of data normalization in the response variables. (Tipton et al.) . Assessment and recommendations on photo interpretation repeatability (Jackson et al.) . Research on modelling approaches for unique landscapes (Sen et al.) . Synthesis (Coulston et al.) Prototype Phase – Study Design Prototype Phase – Key Questions
• How large an area can a single modeling unit encompass? • How many samples (photo interpreted plots) are needed per modeling unit? • Are normalized Landsat mosaic images required by the models or the maps? • What is the minimum set of predictor layers needed by the models? Timelines
Major Milestones Major Milestones 1Q Aug Prototype Kickoff 2Q Sept 2010 3Q Pilot Complete 2010 Oct Pilot Complete 4Q Nov 1Q Dec 2Q Production Process Defined 2011 3Q Production Begins Jan SRS prototype data available 4Q Feb Prototype PI data available 1Q Mar 2Q 2012 Apr Prototype ancillary data available 3Q 2011 May 4Q 1Q Jun Prototype analyses complete 2Q Jul 2013 3Q Aug 4Q CONUS Complete Sept Production Begins 1Q 2Q 2014 3Q 4Q Coastal Alaska Complete 1Q 2Q 2015 3Q 4Q HI, PR, VI Compelete North American Forest Dynamics NAFD Phase 3
University of Maryland Rocky Mountain Research Station, FIA Pacific Northwest Research Station NASA-Goddard NASA-Ames NAFD Science (NASA, PNW, UMD, CONAFOR, Canadian Forest Service, and others) Characterizing disturbance and regrowth patterns on US forests by analyzing a biennial time series of Landsat imagery over a sample of Landsat data cubes spread across US Forests. Objectives include: • Produce nationwide estimates of forest dynamics • Convert data cube reflectance to data cube biomass • Develop nationwide maps of forest biomass dynamics • Begin trials in Canada and Mexico • Quantify forest component of woody encroachment nationally NAFD Applications (NASA, PNW, UMD, all FIA units) Illustrate how FIA data can be combined with temporal disturbance and biomass products to answer management questions relevant to FIA users. Objectives: • Develop FIA monitoring products that take advantage of satellite-derived disturbance and biomass data (storm-related loss, harvest rates across time and ownerships, fragmentation, carbon considerations) • Develop tool kit to enable users to request analyses through FIA NAFD Phase 3: US Forest Disturbance History from Landsat
1) Conduct an annual, wall-to-wall analysis of US disturbance history between1985-2010 2) Undertake a detailed validation of the resultant national disturbance map 3) Examine variation in post-disturbance forest recovery trajectories, using repeat measurements from FIA plot data, 4) Determine disturbance causal agents *** Different types of disturbance have different…
Carbon Consequences Spatial Patterns Fire Fire Clearcut
Temporal Intensities Clearcut We will build a database of forest change processes Web GeoDatabase Browser/Distribution Forest Change Processes
User Community Suburbanizatio Pests and Hurricanes/ Conversi Change Agent Forestry Fires n/Urbanization Pathogens Tornadoes on Timber Decadal Census – Digitized Aerial Ground Landsat Landsat Treatment & # new housing sketches of insect measurements-wind change Data Source NDVI change Removals units damage speed detection
US Forest Health Program U.S. National NLCD MTBS UFSF FIA (Smith http://www.fs.fed.u Hurricane Center Retrofit Data (Theobald 2004) (Eidenshenk Reference et al. 2009) s/r3/resources/hea (Jarvinen et al. Set (Fry et et al. 2007) lth/fid_surveys.sht 1984) al. 2009) ml County polygons polygon <1 ha to 100m grid lines 30m grid 30m Grain or > county Spatial sampled - national sampled - national national national National Extent national
Grain 5-10 year cycles decadal annual annual annual decadal Tempora l Extent varies by region 1940-2030 varies by region 1851-2008 1984-2007 1992-2001 We will explore potential for using Landsat spectral trajectories to classify forest disturbance types Green Leaf Area Forest Structure 0.9 0.35
0.8 0.3
0.7 0.25 0.6 0.2 NDVI 0.5 Forest Stdev 0.15 TM B5 Reflectance 0.4 Forest Avg Clearcut 0.3 0.1 Fire 0.2 0.05 1 3 4 5 7 8 10 1 3 4 5 7 8 10 Years Since Disturbance Years Since Disturbance We will determine how forest type/ecosystem differences impact Landsat spectral classification of disturbance types
Boreal Forest - Canada Temperate/Tropical Forest - Mexico 4500 4500 4000 Clearcut 4000 Clearcut 3500 Fire 3500 Fire 3000 3000 2500 2500 Reflectance Reflectance 2000 2000 Band 5 5 Band 1500 5 Band 1500 TM 1000 TM 1000 500 500 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Time Since Disturbance Time Since Disturbance
Clearcut Clearcut
Fire Fire We will assess the degree to which textural metrics enhance our image-based predictions of disturbance type
Different disturbance processes result in different patterns of landscape structure and fragmentation that are visible in Landsat Imagery.
1984 1987
Harvest Fire Suburbanization
Patch level spatial metrics Continuous Discrete Shape •Homogeneity • Direction •Edge Contrast • Fractal •Heterogeneity • dimension •Texture Area •Range/Mean • •Compactness Timeline
• 3-yr project spanning July 2011 – June 2014 • NAFD Phase 3 Kickoff Meeting – June 2011 • Deliverables for Task 4 – Cause of disturbance Database of forest change agents Library of spectral trajectories Exploration into textural components Pulling it all together for national mapping