Satellite Remote Sensing Tools in Support of Forest Monitoring

Stennis Space Center

Satellite Remote Sensing Tools in Support of Forest Monitoring

Presented by Joe Spruce and Jerry Gasser

NASA Stennis Space Center

Eastern Forest Environmental Threat Assessment Center Review

July 6, 2009


Background on activity

• NASA Stennis is helping the USFS EFETAC to integrate NASA satellite data products into a national forest threat early warning system

• Most of work regards MODIS time series data– MODIS NDVI data for vegetation phenology product

development

– MODIS NDVI products for monitoring historical and current forest disturbance

• Some work is being done with other NASA satellite data and software products

Eastern Forest Environmental Threat Assessment Center Review – July 6, 2009 - 2 -


Development of MODIS Vegetation Phenology Products

Presented by Jerry Gasser



Objectives for developing CONUS vegetation phenology products

• Compute and deliver the highest possible quality vegetation phenology geospatial data base for the conterminous United States (CONUS )– Using 250 meter MODIS data from Aqua and Terra sensors

• Evaluate the utility of the phenological database as input to the EWS for monitoring forest threats– Employ products in assessing known forest disturbances

– Validate against available reference data



MODIS vegetation phenology product development method

• Data Downloading– Primary data set - MODIS MOD 13 NDVI data

– 250 meter 16-day composite product

– Downloaded 9 years of CONUS coverage (2000 – 2008)

• Data Processing– The Time Series Product Tool (TSPT) software is used to generate

filtered time series data products

– The Phenological Parameters Estimation Tool (PPET) software uses the TSPT output to compute phenological parameters for each pixel

• Output Products– Phenological parameter data tiles are mosaicked into CONUS products

that are in a standard map projection



MODIS tiles for CONUS coverage

CONUS Products are Based on 14 Tiles of MOD13 Data



MODIS vegetation phenology product – overall work flow

Output Products• Phenology Parameter Products• Cumulative Integral Products

MODIS Data InputsTime Series Product Tool (TSPT)

Phenological Parameter Estimation Tool (PPET)

Create Geographic Gridded Data from MODIS Swath or Tile

TSPT Parameter FileInputs

Output Products• Filtered time series• Time series plots• Sequence of 2-D

single date images



TSPT processing overview

Remove Cloudy and Low Quality Pixels

Remove Outliers

Fuse AQUA and TERRA Data

Calculate NDVI

MODIS Data Inputs

Temporally FilterOutput Products• Filtered time

series data



PPET processing overview

Output Products• Phenology Data• Integral Data

TSPT Filtered Time Series data

Locate Seasonal Transition Points

Calculate Season and Cumulative Integrals

Identify Seasonal Values and Dates

Apply Curve Fit Algorithm



Vegetation phenology parameter products

• NDVI Value and Date of Occurrence for: – Left Season Minimum

– Season Start

– Onset of Max Greenness

– Season Max NDVI

– Onset of Greenness Decrease

– Season End

– Right Season Minimum

• Additional Products– Season Integral Products

– Cumulative Integral Products

– Quality Control (QC) flags

Left

Min

Right

Min

Start

Season

End

Season

Onset of Max

Greenness

Onset of Greenness

Decrease

Peak

Season

Center

Growing Season



MODIS CONUS vegetation phenology product deliverables

• Version 1.0– Derived from MOD13 Collection 4 data– CONUS phenological parameter and cumulative integral products for

2000-2006

• Version 1.5– Derived from MOD13 Collection 5 data– Generated using the average of AQUA and TERRA products– CONUS phenological parameters and cumulative integrals for 2006-

2008

• Version 2.0 (scheduled for October 2009)– Will utilize MOD13 Collection 5 data– Generated using Day of Year (DOY) acquisition information– CONUS phenological parameter and cumulative integral products for

2000-2008



MODIS CONUS phenology product validation - observations

• Compared field data to PPET results and alternative MODIS-based phenology products (MOD12Q2) derived at 1 km resolution

• Also compared PPET results directly to the MOD12Q2 product for mapped northern hardwood forest in the northeastern U.S.

• Efforts to quantify “accuracy” of satellite phenology product results are confounded by differences in product resolution, differences in product development methods, in-availability of field data relevant to MODIS scales of observation, and in-availability of good, detailed forest cover type maps

• The PPET phenology products do appear to be useful as indices of forest growing season conditions that can be monitored over time– Each 250 meter pixel equates to 6.25 hectares (~15 acres) on the ground



Use of MODIS Products for Forest Disturbance Monitoring

Presented by Joe Spruce



Background - MODIS forest disturbance monitoring efforts

• Since 2007, NASA SSC and the USFS have worked to assess MODIS time series data products for monitoring regional forest disturbance from multiple threats

• Even though MODIS is spatially coarse, its broad area coverage and daily collection offer advantages for monitoring regional forest conditions

• We are conducting case studies to assess, develop, and demonstrate MODIS NDVI products for tracking forest disturbance from specific threats

• Such studies have been done in hindcast and more recently nowcast modes



Forest threat assessment case studies in support of EWS development

Gypsy moth forest defoliation Forest mortality from mountain pine beetle

Hurricane-induced forest damage Ice-storm related forest damage



Gypsy Moth Defoliation Detection Studiesfor 2001 and 2007



Assessing gypsy moth forest defoliation with MODIS NDVI data

• The forest threat EWS requires wall to wall regional detection of forest defoliation, preferably in near real time

• An initial feasibility study was done in the Central Appalachian highlands of the eastern US, an area prone to gypsy moth (GM) defoliation

• GM defoliation was detected from classifications of 48 day temporal composites of MODIS NDVI products from the defoliation time frame (June 10-July 27) for 2000-2006

MOD02 Maximum NDVIDefoliation Time Frame (DTF)

for 2000-2006 (Red)Maximum NDVI DTF

2001 (Green)

MOD02 Defoliation RGB - Basedon Maximum NDVI DTF for All

Years vs. 2001Maximum NDVI DTF

2001 (Blue)



Example MODIS 2001 gypsy moth defoliation detection products

MOD13 (16-day) 250 m MOD02 (daily) 250 m MOD02 (daily) 500 m

Defoliated Forest Non-defoliated Forest Non-forest

Non-defoliated forest and non-forest components are from 2001 NLCD LULC product- 18 -


Validation of MODIS 2001 gypsy moth defoliation detection products

• Defoliation detection products were validated by image interpretation of random sample locations on Landsat and ASTER satellite displays

Note: PA = % Producer’s Agreement (# correct/total), UA = % User’s Agreement (# correct/total), Kappa = Kappa Statistic, OA = % Overall Agreement (# correct/total), and OK = Overall Kappa.

Eastern Forest Environmental Threat Assessment Center Review – July 6, 2009

(Omission) (Commission) (Scaled 0-1) (Omission) (Commission) (Scaled 0-1) (Scaled 0-1)

- 19 -


Example of 2007 gypsy moth forest defoliation from MOD13 data

Total Land ExtentShown

~ 44 Million Acres

PA

VA

WVA

NJ

MD

DE

Healthy Non-Defoliated ForestGypsy Moth Forest Defoliation Non-Forest Water State Boundaries

NY


Use of MODIS NDVI for Near Real Time (NRT) Monitoring of Insect Defoliation


Example – Eastern US


2009 Near Real Time (NRT) Forest Defoliation Detection – Michigan

Attribution of Causal Agent is Preliminary and Based on Either 2009 News Accounts and/or from 2008 USFS FHM Sketch Maps

Forest Tent CaterpillarDefoliation

2007 Tornado-InducedForest Mortality

Detection Product from MODIS NDVI Time Series Data2009 Component of Product is from June 10 - 23


MODIS 2009 NRT Forest Defoliation Detection – Western Pennsylvania

Forest Tent CaterpillarDefoliation (Potter County, PA)

Fall CankerwormDefoliation

EarlyGypsy MothDefoliation?

Attribution of Causal Agent is Preliminary and Based on Either 2009 News Accounts and/or from 2008 USFS FHM Sketch Maps

Product Shows CountiesWith Widespread Forest Defoliation from June 10 – 23 of 2009


MODIS-Based Assessment of Forest Damage from Hurricanes


Main Example – Hurricane Katrina


MODIS products for assessing forest damage from Hurricane Katrina

Hattiesburg

Laurel

Multi-Date MODIS Satellite Imagery MODIS Vegetation Damage Classification

State BoundariesHurricane Storm Track Moderate Vegetation Damage

Low Vegetation Damage

High Vegetation Damage- 25 -


View of hurricane impact zones from 20% left maximum NDVI products

RGB: 2004 20% Left Maximum NDVI in Red; 2006 20% Left Maximum NDVI in Blue and Green

HurricaneRita

Hurricane Katrina

HurricaneIvan

NOTE – WE NEED TO REVISIT THIS WITH 2007 to

2009 Data


Use of MODIS Products in Assessing Western US Forest Mortality and Decline

Example - Northern Colorado



Objectives – Northern Colorado Case Study

• Assess MODIS phenology products for monitoring forest disturbance events in the western US Rocky Mountains– Cumulative Integral NDVI, Peak Growing Season NDVI, and 80%

Right Growing Season NDVI

• Assess MODIS MOD13 temporal composite products for monitoring mountain pine beetle mortality

• Focus initial work in Colorado, using sketch maps, Landsat, and aerial reference data



Western US view of forest conditions in 2006 vs. baseline cumulative NDVI

2002-2006 Maximum NDVI CI-23 in red color gun – 2006 CI-23 in blue and green

2 date NDVI RGB from end of year MODIS NDVI cumulative integral (CI-23) data

Potential forest decline is shown on this in pink to red tones, whereas

healthier forests maintained high CI-23

NDVI values

Location of next slide

Some range land alsoshows drops in CI-23 NDVI

- 29 -


USGS 2002-2006 fire perimeter maps overlain onto MODIS RGB

2000

2 date NDVI RGB from end of year MODIS NDVI cumulative integral (CI-23) data2002-2006 Maximum NDVI CI-23 in red color gun – 2006 CI-23 in blue and green

Most forest with reduced CI-23 values were not

affected by major wildfires in 2002-2006

2002 2003 2004

Fires of 2000 – show low CI-23 for 2002-2006, suggesting forest loss

Fire of 2003 showed high CI-23 value drop

compared to baseline, suggesting high intensity

burn

2001, 2005, and 2006 – no wildfires (source: USGS) - 30 -


2006 USFS MPB mortality sketch map overlain onto MODIS RGB

Mountain pine beetle forest mortality was rarely noted in the eastern flank, where there was less drop cumulative NDVI in 2006

versus the baseline (maximum cumulative NDVI

of 2002-2006)

2006 Sketch Map of MPB Mortality

2 date NDVI RGB from end of year MODIS NDVI cumulative integral (CI-23) data2002-2006 Maximum NDVI CI-23 in red color gun – 2006 CI-23 in blue and green

National Park LandNational Forest Land

Sketch mapped MPB mortality tends to occur

within the pink to red tones

- 31 -


True color summer 2006 Landsat data overlain onto MODIS RGB

Eastern Forest Environmental Threat Assessment Center Review – July 6, 2009

Foreground Image - Landsat 5 Thematic Mapper Data Acquired July 28, 2006

Dead red forest evident in multiple areas on Landsat

data

NEXTSLIDE

- 32 -


MODIS RGB versus summer 2006 Landsat true color RGB

MODIS 250-Meter RGB – from CI-23 Products Landsat Data – Acquired 7-28-2006

Red tones on MODIS RGB (left) often correspond to dead red forest on 2006 Landsat true color image (right).

- 33 -Eastern Forest Environmental Threat Assessment Center Review – July 6, 2009

Subset Area is ~11 miles NW of Kremmling, CO


Participation in this work by NASA Stennis contractors (e.g., Science Systems & Applications, Inc. and Lockheed Martin Corporation) at the John C. Stennis Space

Center, Mississippi was enabled under NASA Task Order NNS04AB54T.


Documents

Satellite Remote Sensing Tools in Support of Forest Monitoring