Transcript Monitoring Trends in Land Change? The Big Tent – MTLC A hierarchical land cover and land use change monitoring system that leverages existing.

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
Example modelling
techniques
Random forests
Stochastic gradient boosting
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
2010
1Q
Aug
2Q
Sept
3Q
Pilot Complete
2010
4Q
2012
2Q
Production Process Defined
3Q
Production Begins
SRS prototype data available
4Q
Feb
Prototype PI data available
1Q
Mar
2Q
Apr
3Q
2011
Jun
2Q
Jul
3Q
2Q
3Q
Coastal Alaska Complete
1Q
2015
2Q
3Q
4Q
Prototype analyses complete
Aug
CONUS Complete
1Q
4Q
Prototype ancillary data available
May
1Q
4Q
2014
Dec
Jan
4Q
2013
Pilot Complete
Nov
1Q
2011
Oct
Prototype Kickoff
HI, PR, VI Compelete
Sept
Production Begins
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…
Spatial Patterns
Carbon Consequences
Fire
Clearcut
Fire
Temporal Intensities
Clearcut
We will build a database of forest change processes
Web
Browser/Distribution
GeoDatabase
Forest
Change
Processes
User Community
Change Agent
Forestry
Data Source
Timber
Treatment &
Removals
Suburbanizatio Pests and
n/Urbanization Pathogens
Hurricanes/
Tornadoes
Fires
Decadal Census – Digitized Aerial
Ground
Landsat
# new housing
sketches of insect measurements-wind
NDVI change
units
damage
speed
Conversi
on
Landsat
change
detection
UFSF FIA (Smith
et al. 2009)
(Theobald 2004)
US Forest Health
Program
http://www.fs.fed.u
s/r3/resources/hea
lth/fid_surveys.sht
ml
Grain
County polygons
or >
100m grid
polygon <1 ha to
county
lines
30m grid
30m
Extent
sampled national
national
sampled - national
national
national
National
decadal
annual
annual
annual
decadal
1940-2030
varies by region
1851-2008
1984-2007
1992-2001
Reference
Spatial
Grain 5-10 year cycles
Tempora
l
Extent varies by region
U.S. National
Hurricane Center
(Jarvinen et al.
1984)
NLCD
MTBS
Retrofit Data
(Eidenshenk
Set (Fry et
et al. 2007)
al. 2009)
We will explore potential for using Landsat spectral
trajectories to classify forest disturbance types
Green Leaf Area
0.9
0.8
0.6
0.5
Forest Stdev
Forest Avg
Clearcut
Fire
0.4
0.3
0.2
TM B5 Reflectance
0.3
0.7
NDVI
Forest Structure
0.35
0.25
0.2
0.15
0.1
0.05
1
3
4
5
7
Years Since Disturbance
8
10
1
3
4
5
7
Years Since Disturbance
8
10
We will determine how forest type/ecosystem differences
impact Landsat spectral classification of disturbance types
Temperate/Tropical Forest - Mexico
Boreal Forest - Canada
4500
4000
Clearcut
4000
Clearcut
3500
Fire
3500
Fire
TM Band 5 Reflectance
TM Band 5 Reflectance
4500
3000
2500
2000
1500
1000
3000
2500
2000
1500
1000
500
500
0
0
1 2
3 4 5 6 7 8
9 10 11 12 13 14 15 16
Time Since Disturbance
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Time Since Disturbance
Clearcut
Fire
Clearcut
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
Harvest
1987
Fire
Suburbanization
Patch level spatial metrics
Continuous
•Homogeneity
•Edge Contrast
•Heterogeneity
•Texture
•Range/Mean
Discrete
•Shape
•Direction
•Fractal
dimension
•Area
•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