40
Challenges of monitoring natural disturbance processes using remotely sensed data in North Coast and Cascades Network: comparison of approaches Natalya Antonova, NCCN Catharine Thompson, NCCN Robert Kennedy, OSU* LandTrendr slides provided by Robert Kennedy

Natalya Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

  • Upload
    faxon

  • View
    44

  • Download
    0

Embed Size (px)

DESCRIPTION

Challenges of monitoring natural disturbance processes using remotely sensed data in North Coast and Cascades Network: comparison of approaches. Natalya Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*. LandTrendr slides provided by Robert Kennedy. NCCN Monitoring Goals. - PowerPoint PPT Presentation

Citation preview

Page 1: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Challenges of monitoring natural disturbance processes using

remotely sensed data in North Coast and Cascades Network:

comparison of approaches

Natalya Antonova, NCCNCatharine Thompson, NCCN

Robert Kennedy, OSU*

LandTrendr slides provided by Robert Kennedy

Page 2: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

NCCN Monitoring Goals

• Document landscape changes• When, where, what and magnitude

• Status and trends• Prepare for and manage for landscape responses

to climate change• Develop prediction tools

• Test hypotheses

Page 3: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

NCCN Monitoring Goals

Monitoring goal Type 1: Monitor yearly Avalanche chute clearing

Landslides Fire Insect/disease defoliation in forest Windthrow Riparian disturbance Clearcuts Rural development

Type 2: Monitor decadally Alpine tree encroachment

Hardwood/conifer forest composition Forest structure

Page 4: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*
Page 5: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Protocol for Landsat-Based Monitoring of Landscape Dynamics at NCCN Parks – Kennedy et al.

1. Two different images

2. Select large changes in spectral values to indicate change

Subtract

1994 2004

Probabilities of Change

Page 6: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Brightness: RedGreenness: GreenWetness: Blue

Brt+Grn: Yellow/OrangeBrt+Wet: MagentaGrn+Wet: Cyan

Tasseled-cap transformation of Landsat image

Astoria

Page 7: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Snow and iceMixed

Open: Dark

Water/Deep shade

Closed-canopy coniferDense broadleaf/

grassBroadleaf tree/shrub

Conifer/Broad-leaf Mix

Increasing TC Brightness

Incr

easi

ng

TC

Gre

enn

ess

Open: Bright

Change in Probability of Membership

Time 1

Time 2

Page 8: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Probability Thresholding

All spectral changesAll spectral changes

ArtifactsArtifacts

Uninteresting* changeUninteresting* change

Real changeReal change

Sensor degradation, atmospheric contamination,

geometric misregistration, sun angle variation

Sensor degradation, atmospheric contamination,

geometric misregistration, sun angle variation

Seasonality of vegetation (phenology), clouds, agricultural practices

Seasonality of vegetation (phenology), clouds, agricultural practices

Sustained change in land cover or

condition

Sustained change in land cover or

condition

Mapped “change”Mapped “change”Mapped “no-change” Mapped “no-change” Th

resh

old

Thre

shol

d

FALSE POSITIVESFALSE POSITIVES

FALSE NEGATIVESFALSE NEGATIVES

Page 9: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

North Cascades National Park

Complex

July 29, 2005-Aug 17, 2006

Page 10: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Mount Rainier

National Park

Aug 14, 2005-Aug 17, 2006

Page 11: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Olympic National Park

July 24, 2004-June 28, 2006

Page 12: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Validation - Errors of Omissiona) b)

c) d)

e)

TC 2005 TC 2006

Change image

2006 NAIP Aerial Photo

Polygons outlined in the validation process compared to change detected by the algorithm

Page 13: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Validation - Errors of Commissiona) b)

c) d)

e)

TC 2005 TC 2006

Changeimage

Polygons outlined in the validation process compared to change detected by the algorithm

Change image from east side of the study area

Page 14: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

125 m

Subalpine Environments, Avalanche Chutes, Tree line, and River Disturbances

2004

2006

Increase in conifer Increase in broadleaf Increase in vegetation Decrease in conifer

Page 15: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Summary: Current Protocol

• Can detect change• Detected too much false change (clouds, shadows,

agricultural dynamics) to provide meaningful results

• Threshold level not sensitive enough to detect annual regrowth or low intensity, slow disturbance

• Difficult to see change along narrow, long features of interest, due to misregistration errors

• Upper elevation areas appear as pure speckle due to variable landcover and annual variation in phenology

Page 16: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Landsat-based Detection of Trends in Disturbance and Recovery (LandTrendr)

Rather than look for disturbance EVENTS, look for disturbance TRAJECTORIES

Kennedy, R.E., Cohen, W.B., & Schroeder, T.A. (2007). Trajectory-based change detection for automated characterization of forest disturbance dynamics. Remote Sensing of Environment, 110, 370-386

Page 17: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Segmentation

• Goodness of fit to idealized curves

• Allows for lower threshold levels

• Greatly reduces amount of background noise

Page 18: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Cloud/Shadow Screening

CloudCloud Shadow CloudCloud Shadow

Merge

Page 19: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Poor-quality Images

19961996 1998199819971997

Olympic Peninsula Olympic Peninsula

Page 20: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Outputs

Disturbance and recovery maps

• Intensity/Magnitude• Year of onset• Duration

Page 21: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Current protocol vs. LandTrendr

Page 22: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Original protocol detected ~100,00 ha of change between 2004 and 2006 within the OLYM study area

Current protocol vs. LandTrendr

∑ = ~ 30,000 ha

Page 23: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

LandTrendr – Clearcuts: Forestlands north of Cle Elum, WA

Page 24: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

20+ yr20+ yr

10+ yr starting 1990s10+ yr starting 1990s

RecentRecent

LandTrendr - Insect disease/defoliation: Olimpic N.P.

Page 25: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

LandTrendr - Avalanches

Page 26: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

LandTrendr –Windthrow

Page 27: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

LandTrendr - Fire

Page 28: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

LandTrendr - Landslides

Page 29: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

LandTrendr- Pros

• Captures Pacific Northwest landscape dynamics well

• Captures smaller changes that are still of interest• Already has long time series

• 25 years of change• Provides additional products like intensity and

regeneration• Includes Canada• Works for small and large parks

Page 30: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

LandTrendr - Cons

• Expensive to implement• Still need to interpret results (ascribe agent of

change)• Develop methodology

Subsampling? Modeling? Validate every polygon in park?

• Developed for forested areas • results have not been evaluated for subalpine

vegetation

Page 31: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Existing Tools: C-CAP Data

• NOAA- Coastal Change Analysis Program • Classified Landsat TM data• Every five years (1996, 2001, 2006 …)• Products:

• Map of 21 classes• Map of change between classes

• Accuracy of change classes varies between 75 and 95%• Focus on coastal areas

High Intensity DevelopedMedium Intensity DevelopedLow Intensity DevelopedDeveloped Open Space

CultivatedPasture/Hay

GrasslandDeciduous ForestEvergreen ForestMixed ForestScrub/ShrubBare LandWaterSnow/Ice

Palustrine Emergent WetlandPalustrine Forested WetlandPalustrine Scrub/Shrub WetlandEstuarine Emergent Wetland

Unconsolidated ShorePalustrine Aquatic BedEstuarine Aquatic Bed

Page 32: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

C-CAP Data Analysis - Example from SAJH

Page 33: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

C-CAP vs. LandTrendr

Page 34: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

C-CAP vs. LandTrendr (acres)Landtrendr CCAP

OLYM_AOI 22933.34 69735.51OLYM 1733.72 79.62

MORA_AOI 10273.08 14581.35MORA 1294.99 11.88

NOCA_AOI 6807.26 6735.60NOCA 914.92 42.21

Page 35: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

C-CAP vs. LandTrender – Rural Development

Page 36: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

C-CAP vs. LandTrender - Fire

Page 37: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

C-CAP vs. Landtrendr - Riparian

Page 38: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

C-CAP -Pros• Free• Simple analysis to get results• Could provide “big picture” change detection

outside park, particularly reductions in forest cover

Page 39: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

C-CAP - Cons

• Misses certain change types Slow increase or decrease in vegetation, narrow

features like riparian• Accuracy unknown, errors propagate• Long time delay for results (01-06 change

available in 09)• 5 year interval too long for some types of

change Rivers, avalanche chutes

• No control over product• Doesn’t cover Canada• Still need to ascribe agent to change

Page 40: Natalya  Antonova , NCCN Catharine Thompson, NCCN Robert Kennedy, OSU*

Current Efforts

•Automatically assign disturbance agent based on:• Trajectory label• Location on

landscape• Proximity to stream• Aspect• Elevation• Geology• Soil Type