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 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 – 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?

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

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 agentsLibrary of spectral trajectoriesExploration into textural componentsPulling it all together for national mapping