monitoring trends in land change? · nafd applications (nasa, pnw, umd, all fia units) illustrate...
TRANSCRIPT
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 SciencesRemote Sensing Applications Center
Rocky Mountain Research Station, FIASouthern Research Station, FIANorthern Research Station, FIA
Pacific Northwest Research Station, FIAState 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 fragmentationKnowing 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 Design4x Intensity Photo-based
Sample Locations
105 photo points to estimate% tree canopy cover for
modeldevelopment
2011 General Modelling Approach
=Response developed by photo
Interpreting Tree crown cover on NAIP Imagery for ~4160 8100m2
sampling chips per Landsat scene.
Random forestsStochastic gradient boosting
Support vector machines
Example modellingtechniques
Fig. from Homer et al. 2007
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 – 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?
TimelinesMajor Milestones
2010
Aug Prototype Kickoff
SeptOct Pilot Complete
NovDec
2011
Jan SRS prototype data available
Feb Prototype PI data available
MarApr Prototype ancillary data available
MayJun Prototype analyses complete
JulAugSept Production Begins
Major Milestones
2010
1Q
2Q
3Q Pilot Complete
4Q
2011
1Q
2Q Production Process Defined
3Q Production Begins
4Q
2012
1Q
2Q
3Q
4Q
2013
1Q
2Q
3Q
4Q CONUS Complete
2014
1Q
2Q
3Q
4Q Coastal Alaska Complete
2015
1Q
2Q
3Q
4Q HI, PR, VI Compelete
North American Forest Dynamics NAFD Phase 3
University of MarylandRocky Mountain Research Station, FIA
Pacific Northwest Research StationNASA-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 ***
Carbon Consequences
Temporal Intensities
Spatial Patterns
Different types of disturbance have different…
Fire Clearcut Fire
Clearcut
GeoDatabaseForest
ChangeProcesses
User Community
Change Agent Forestry Suburbanization/Urbanization
Pests and Pathogens
Hurricanes/Tornadoes Fires Conversi
on
Data SourceTimber
Treatment & Removals
Decadal Census –# new housing
units
Digitized Aerial sketches of insect
damage
Ground measurements-wind
speed
Landsat NDVI change
Landsat change
detection
Reference UFSF FIA (Smith et al. 2009) (Theobald 2004)
US Forest Health Program
http://www.fs.fed.us/r3/resources/health/fid_surveys.sht
ml
U.S. National Hurricane Center (Jarvinen et al.
1984)
MTBS (Eidenshenk et al. 2007)
NLCD Retrofit Data Set (Fry et al. 2009)
SpatialGrain County polygons
or > 100m grid polygon <1 ha to county lines 30m grid 30m
Extent sampled -national national sampled - national national national National
Temporal
Grain 5-10 year cycles decadal annual annual annual decadal
Extent varies by region 1940-2030 varies by region 1851-2008 1984-2007 1992-2001
Web Browser/Distribution
We will build a database of forest change processes
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1 3 4 5 7 8 10
NDV
I
Years Since Disturbance
Forest StdevForest AvgClearcutFire
0.05
0.1
0.15
0.2
0.25
0.3
0.35
1 3 4 5 7 8 10
TM B
5 Re
flect
ance
Years Since Disturbance
Green Leaf Area Forest Structure
We will explore potential for using Landsat spectral trajectories to classify forest disturbance types
Boreal Forest - Canada
0
500
1000
1500
2000
2500
3000
3500
4000
4500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
TM B
and
5 Re
flect
ance
Time Since Disturbance
Clearcut
Fire
0
500
1000
1500
2000
2500
3000
3500
4000
4500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
TM B
and
5 Re
flect
ance
Time Since Disturbance
Clearcut
Fire
Temperate/Tropical Forest - Mexico
Fire
Clearcut
Fire
Clearcut
We will determine how forest type/ecosystem differences impact Landsat spectral classification of disturbance types
We will assess the degree to which textural metrics enhance our image-based predictions of disturbance type
19871984
Patch level spatial metricsContinuous Discrete
•Homogeneity •Edge Contrast•Heterogeneity
•Texture•Range/Mean
Harvest Fire Suburbanization
Different disturbance processes result in different patterns of landscape structure and fragmentation that are visible in Landsat Imagery.
•Shape•Direction•Fractal
dimension•Area
•Compactness