Evaluating the Coupled Dynamic Vegetation-Fire-Emissions model,

LPJ-GUESS-SPITFIRE, against EO-based Tropical Tree Biomass

Allan Spessa 1, Matthew Forrest 2, Thomas Hickler 2

1. Dept Environment, Earth & Ecosystems, Open University (Milton Keynes, UK)

2. Biodiversity and Climate Research Centre, Senkenberg Museum & Goethe

University (BiK-F) (Frankfurt, Germany)

Ignitions

Area Burned

Fuel Moisture & Fire Danger

Index

Rate of Spread

& Fire Duration

Fire Intensity

Emissions (trace greenhouse gases + aerosols)

Human-caused

Lightning-caused

Plant Mortality

Fuel Load & Fuel Structure

Wind speed

Temperature

Relative Humidity

Rainfall

LPJ Dynamic Global Vegetation

Model

Population Density & land-use

SPITFIRE- original

Fuel Consumed

Thonicke, Spessa, Prentice et al (2010) Biogeosc.

LPJ-GUESS simulates ecological succession

Smith et al (2001) GEB

LPJ-GUESS-SPITFIRE simulates

differential fire-induced tree

mortality

Lehsten et al (2008) Biogeosciences

LPJ-GUESS LPJ-DGVM

Fuel Consumed

Observed Area Burned

Fuel Moisture & Fire Danger

Index

Rate of Spread

& Fire Duration

Fire Intensity

Plant Mortality

Fuel Load & Fuel Structure

Wind speed

Temperature

Relative Humidity

Rainfall

LPJ-GUESS Dynamic

Vegetation Model

SPITFIRE- driven by observed burnt area

Emissions (trace greenhouse gases + aerosols)

Princeton Climate Reanalysis Data (1948-2008) ( after Sheffield et al 2006 J. Climate)

Burnt Area Data:

• Mouillot , 1901-1996 (Mouillot & Field GCB (2005)

• GFEDv4, 1997-2008 (Giglio et al 2013 JGR)

Two recent EO-based datasets of pan-tropical tree biomass @ 1 sq km

Saatchi et al (2011) PNAS

Baccini et al (2011) Nature Climate Change

ESA GLOBCOVER 2009 @ 1sq km

GUESS-SPITFIRE simulated tree carbon versus two EO-based datasets (Saatchi et al 2011, Baccini et al 2012)

(mean 1997-2008) (GC 2009 land cover corrected) (kgC.m-2)

Five run scenarios:

I. Default model (ca. 2010 version) with no fire.

II. Default model with fire.

III. Default model with fire and new cambial kill as a function of bark thickness (Hoffman et al 2009 Ecology).

IV. Default model with fire and new allometry for tropical trees (Feldpausch et al 2011 Biogeosc.) and tropical

savanna trees (Dantos & Pausas 2012 J. Ecol).

V. Default model with fire and new cambial kill and new tree allometry.

GUESS-SPITFIRE simulated tree carbon versus two EO-based datasets (Saatchi et al 2011, Baccini et al 2012)

(mean 1997-2007) (GC 2009 land cover corrected) (kgC.m-2)

(Spessa , Forrest, Hickler et al)

C[GUESS-SPITFIRE_No_Fire – C[Baccini] C[GUESS-SPITFIRE_No_Fire – C[Saatchi]

C[GUESS-SPITFIRE_Fire] – C[Saatchi] C[GUESS-SPITFIRE_Fire] – C[Baccini]

GUESS-SPITFIRE simulated tree carbon versus two EO-based datasets (Saatchi et al 2011, Baccini et al 2012)

(mean 1997-2007) (GC 2009 land cover corrected) (kgC.m-2)

(Spessa , Forrest, Hickler et al)

C[GUESS-SPITFIRE_Fire] – C[Saatchi] C[GUESS-SPITFIRE_Fire] – C[Baccini]

C[GUESS-SPITFIRE_Fire + new tree allometry + new cambial kill] – C[Saatchi]

C[GUESS-SPITFIRE_Fire + new tree allometry + new cambial kill] – C[Baccini]

Conclusions 1. LPJ-GUESS-SPITFIRE tree biomass simulations improved when observed fire assimilated into vegetation-fire

model.

2. Further improvement evident by new formulations for tree allometry in tropical forest and savanna trees, and

fire-induced mortality due to cambial damage.

3. Focus areas for further work…

• Compare simulated versus observed tree cover (MODIS).

• Simulated residence time of fires in LPJ-GUESS-SPITFIRE- is this realistic?

• Comparison with CASA biomass (CASA driven by EO-based FAPAR, and underpins GFEDv4 emissions

database).

• How does biomass uncertainty affect emissions?

EXTRAS

Global FWI Seasonality, 1980-2012

CPC Drought Code, 1982-83 fires in Indonesia

• Sufficient spatial resolution to capture drought driving transition in burning from southern Sumatra and southern Kalimantan in late 1982 to East Kalimantan in early 1983

Cambial kill and new bark thickness relationships

• Pm: probability of cambial kill based on ratio of i) critical time to

cambial kill to ii) residence time of fire.

• tau_l: residence time of fire (function of total fuel load W and amount

consumed C ).

• tau_c: critical time (mins) to cambial kill as a function of bark thickness

BT (cm)

• Peterson & Ryan 1986; Hoffman et al 2009, 2012; Lawes et al 2013.

New tree allometry relationships

• ↓ tree height ↑ stem diameter ↑ ↑ crown area per tree ↓ density of trees (less biomass per unit area)

• ↓ tree height ↑ stem diameter ↑ ↑ bark thickness ↓ cambial kill (more biomass per unit area)

New bark thickness relationships

• Pm: probability of cambial kill

• tau_l: residence time of fire (function of total fuel load W and amount

consumed C ).

• tau_c: critical time (mins) to cambial kill as a function of bark thickness

BT (cm)

• Peterson & Ryan 1986; Hoffman et al 2009, 2012; Lawes et al 2013.

New tree allometry relationships

• ↓ tree height ↑ stem diameter ↑ ↑ crown area per tree ↓ density of trees (less biomass per unit area)

New tree allometry II

PFT-based vs Patch-based Approaches to Vegetation Modelling

Bare Gd

C4

TrBlRg tree

TrBlEg tree

1 y.o. 5 y.o.

15 y.o.

30 y.o. 60 y.o.

90 y.o.

LPJ-DGVM, TRIFFID, SDGVM etc LPJ-GUESS Patch-based tile structure. PFT-based tile structure.