Meeting Forest Carbon Planning Needs with Forest Service Data and Satellite Imagery

Meeting Forest Carbon Planning Needs with Forest Service Data and

Satellite Imagery

Sean Healey, Gretchen MoisenRMRS Inventory, Monitoring, and Analysis Program

Greg Jones, Dan LoefflerRMRS Human Dimensions Program

Shawn UrbanskiRMRS Fire, Fuel, and Smoke Program

Todd MorganU. MT Bureau of Business and Economic Research

Jim Morrison, Barry Bollenbacher, Renate Bush, Keith Stockman

National Forest System, Region 1

Montana

Idaho

The Forest Carbon Management Framework (ForCaMF) has been piloted in Ravalli County, MT, and is currently being applied across the NFS Northern region

Managers and planners need comprehensive information about carbon stocks and flows

• Big Picture: How much carbon is the landscape storing or emitting?

• What are the immediate and long-term effects of natural disturbance on carbon storage?

• How does carbon accumulation in undisturbed parts of the landscape compare with disturbance losses?

• What is the magnitude of harvest effects vs. “natural processes”?

The Forest Service maintains a stand dynamics tool (Forest Vegetation Simulator - FVS) that is used in

ForCaMF to govern carbon accumulation and emission across the landscape.

Mid-1980s imagery is used spatially represent FIA estimates from the same era. The landscape matches FIA in the following ways:• Area of forest•Area of forest by forest type•Mean volume•Distribution of volume (right number of low-, medium-, and high-volume pixels)

Ravalli County (MT) forest volume, 1985

50 km

Estimation of ecosystem flux Starting Point

Non-forest/Background

Fir, Spruce

Cut

No disturbanceBurn

Doug Fir

Lodgepole

PonderosaNon-forestGreyscale: low to high

1985 Forest Volume Forest Type Disturbance

Estimation of ecosystem flux Starting Point

•Spatial representations of reference data are prepared using satellite imagery

1985 Forest Volume

Forest Type

Disturbance

1985 Carbon

1987 Carbon

1989 Carbon

1985 Carbon

1985 Carbon

FVS-derived carbon dynamics are applied according to the spatial inputs to create the best available spatial representation of carbon sequestration over time

However, we know that there is uncertainty involved with each of the inputs …

Forest Type

Forest Volume

Starting-Point Forest Condition Maps

FVS Carbon Simulations

Lookup tables linking the starting landscape variables and disturbance history of each SU with appropriate FVS carbon simulations

% Cover Loss

Volume harvested

Spatial disturbance data

10-ha Simulation Units (SU) are developed representing homogeneous groupings of pixels with identical combinations of starting conditions and disturbance parameters

Spatial inputs of each SU are altered probabilistically to represent their random error and potential bias

Inventory Data

Plot-Level Model Calibration

Population-Level Model Constraint

Plot-Level Basis for Simulation

ENDPOINT: Probability Density Function of Stock or Flux of interest

Probabilistic Treatment of Spatial Inputs (PTSI)

Disturbance Type

+Stocks and fluxes estimated within

and summed across Simulation Units

Probability Density Function (PDF)•Used in ForCaMF to describe and simulate uncertainty of inputs due to random error and bias as well as uncertainty•Also used to describe ForCaMF outputs

Figure from wikipedia.org

Source Type PDF σ RationaleStarting Volume Bias 0.08 of mapped volume

Landscape is matched to FIA's estimate of forest volume; σ is taken from the standard

error of this estimate

Starting Volume Random Error 1611 ft/acre Root mean square error of independent test

set

Forest Type Bias from 0.12 to 0.25, depending upon type

Landscape is matched to FIA's estimates of area by forest type; σ is taken from the

standard errors of these estimates

Forest Type Random Error

PDF not used - 30% chance of error, assumed to be distributed

evenly among typesError structure drawn from error matrix of

independent test set

Area Disturbed by Year Bias 0.15 of mapped area Conservative estimate drawn from literature

involving similar products

% Cover Loss due to Fire

Random Error 26% loss

Taken from predicted vs. observed pair-wise cover differences from the independent test

set

% Volume Removal due to Harvest

Random Error 26% removal Arbitrarily set to uncertainty associated with

cover loss

FVS Link Function Random Error

0.27 of carbon stocks projected via FVS lookup table

Represents the average variation in carbon stocks among simulations binned within each

cell of the lookup table

Uncertainty built into the simulations is estimated from the best available sources, including FIA

Inputs, such as disturbance history, may be changed to derive estimates for alternative scenarios

Bars represent standard deviation of 2000 simulations

1985-86

1986-88

1988-91

1991-93

1993-95

1995-97

1997-98

1998-99

1999-01

2001-03

2003-05

0

50,000

100,000

150,000

200,000Ravalli County, Montana, Fire Emissions (tonnes C)

Bars indicate standard deviation of 2000 simulations

1.9 million (±.4 million)

Average Annual Fossil Fuel Emissions

Unlike standard FIA carbon stock estimates, we can isolate individual processes contributing to overall carbon flux

We see that the net effect of fire on carbon stores actually increases for decades after the fire

Estimated stand carbon in forest population affected by fire in the year 2000 in Ravalli County, MT

Tonn

es

Carb

on

Preliminary programming has occurred to embed PTSI in a decision support tool for

the NFS Northern Region

Time

Land

scap

e Ca

rbon

Exc

hang

e (t

onne

s C)

Sequestration

Emission

Framework

Growth – undisturbed forests

Growth – recovering forests

Combustion emissions

Fossil fuel combustion – road building

Fossil fuel combustion – timber haul

For each time period, PTSI-based ecosystem flux estimates may be combined with non-ecosystem flux estimates

Net of all considered factors

Time

Land

scap

e Ca

rbon

Exc

hang

e (t

onne

s C)

Sequestration

Emission

The basic function of the system is to monitor (with uncertainty estimates) forest carbon flux over time.

Alternative scenarios will be discussed later.

From: Healey and others, Carbon Balance and Management 4:9.

Haul distances can be translated to fossil fuel emissions associated with timber transport

Transport emissions for Ravalli County timber

Road construction activityCarbon dioxide

emissions (pounds/foot)

Carbon dioxide emissions

(pounds/mile)

Carbon equivalent (pounds/foot)

Carbon equivalent (pounds/mile)

Pioneering 0.31345 1,655 0.08548 451

Clearing and grubbing 1.40834 7,436 0.38409 2,028

Sub-grade excavation with sidecasting

0.81769 4,317 0.22301 1,177

Total of all activities 2.53947 13,408 0.69258 3,657

Source: Loeffler, Jones, Vonessen, Healey, Chung. 2008. Estimating Diesel Fuel Consumption and Carbon Dioxide Emissions from Forest Road Construction. In: Forest Inventory and Analysis (FIA) Symposium; October 21-23, 2008; Park City, UT. Proc. RMRS-P-56CD.

We can also estimate carbon emissions related to forest road-building over time

0

2000

4000

6000

8000

10000

12000

14000

16000

1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007

Emis

sion

(ton

nes

C)

Cut 1945-1979, Dump

Cut 1945-1979, Landfill

Cut 1980-2007, Landfill

Using dynamics in the forest product life cycle literature with harvest records, we can track emissions from

historically harvested timber

More on this method: Healey et al., 2008: http://www.treesearch.fs.fed.us/pubs/33355

Time

Land

scap

e Ca

rbon

Exc

hang

e (t

onne

s C)

Sequestration

Emission

Flux Diagnosis

Growth – undisturbed forests

Growth – recovering forests

Combustion emissions

Fossil fuel combustion – road building

Fossil fuel combustion – timber haul

Net of all considered factors

Time

Land

scap

e Ca

rbon

Exc

hang

e (t

onne

s C)

Sequestration

Emission

Time

Land

scap

e Ca

rbon

Exc

hang

e (t

onne

s C)

Alternative disturbance scenarios drive different flux trends

Summary

• Probabilistic treatment of spatial inputs (PTSI) allows us to link satellite and inventory data with FVS to understand landscape carbon dynamics and associated uncertainties

• We can combine ecosystem and non-ecosystem fluxes to comprehensively track effects of disturbance and management on forest carbon storage, using both observed and hypothetical scenarios

Questions?

Sean [email protected], 801-625-5770