Global gross primary production estimation
using satellite-derived
canopy conductance
Marta Yebra, Albert Van Dijk, Ray Leuning and Juan Pablo Guerschman
Marta Yebra | PhD
Ozfux Meetin 2014
Tuesday 30 September
Allice Springs
Physiological nexus GPP-ET
Animation from Joe Berry
2
Global Gc (2001-2011)
Yebra, M., Van Dijk, A., Leuning, R., Huete, A., Guerschman, J.P., 2013. Evaluation of optical remote sensing to estimate actual
evapotranspiration and canopy conductance. Remote Sensing of Environment. 129, 250-261.
3
Global Gc model
Inverted Penman-Monteith (PM) Equation
 E Ga
Gs 
 A  (  1) E   c p D / 
(Monteith 1964)
Yebra, M., Van Dijk, A., Leuning, R., Huete, A., Guerschman, J.P., 2013. Evaluation of optical remote sensing to estimate actual
evapotranspiration and canopy conductance. Remote Sensing of Environment. 129, 250-261.
Evaluation against tower ET
Yebra, M., Van Dijk, A., Leuning, R., Huete, A., Guerschman, J.P., 2013. Evaluation of optical remote sensing to estimate actual
evapotranspiration and canopy conductance. Remote Sensing of Environment. 129, 250-261.
5
Gc vs Da has a hyperbolic form
Your satellite Gc
estimates do not consider
the effect of VPD!!!!!!
(Leuning 1995)
6
Improved global Gc model
1
𝐺𝑐
𝐺𝑐𝑅𝑆
=
𝐶0
1+𝐷 𝐷
50
Lohammar et al (1980)
Two parameters model
Da
D50=0.70 kPa
C0=1.94
Yebra et al, In preparation
7
Objective
GPP
• Couple GPP to ET though remotelybased estimates of Gc
• Simple model formulations
• Small number of ‘free’ parameters
• Global scale
• No-land cover dependant
ET
8
Light use efficiency models
• Primary drivers of GPP (F) are vegetation
cover and radiation
• Simplest approach to estimate F
F =ɛ 𝑓𝑃𝐴𝑅 𝑄
(Sims et al 2008;
Running et al 1999;
Sjostrom et al 2011 and
more and more)
What happens where other
environmental conditions limit
photosynthesis?
9
Climatic controls on GPP
33%
27%
Running et al 2004, Bioscience
40%
10
Our two parameter-model
F = min {Fr , Fc,}
Radiation limited
Fr=ɛ fPAR Q
Conductance limited
Fc=cg Gcw (1-R0)Ca
Ɛ=Ɛmax EVI (Light use efficiency)
cg conversion coefficient
fPAR Linear ramp function
between (NDVImin,0) and
(NDVImax,0.95) (Donohue et al.
2008)
Gcw (m s-1) canopy conductance to
water vapour
Q incident PAR (mol photons)
R0=Ci/Ca (minimum)
Ci internal [CO2]
Caexternal [CO2]
Only 2 Parameters! R0=0.76 and ɛmax=0.045
11
R0 decreases linearly with Da
Are you assuming R0 is
constant?!!!!!
(Leuning 1995)
12
R0 decreases linearly with Da
1.00
AU-How
AU-Tum
FI-Hyy
IT-Ro1
IT-Ro2
NL-Loo
US-Bo1
US-Ha1
US-Ho1
US-MMS
US-Ne1
US-Ne2
US-Ne3
US-Ton
US-Var
US-WCr
R0
0.90
0.80
Optimized
R0=0.76
0.70
0.60
0
1000
2000
3000
4000
5000
VPD
Da (Pa)
ENF, DBF/EBF, CRO/GRA, WSA
Yebra et al, In preparation
13
Da for Periods when F=Fr and F=Fc
F=Fr
F=Fc
14
Other factors to explain GPP
• Optimum temperature; most enzymemediate reactions have a
• Da may have additional influence on ε or
[CO2 ]
• At High radiations photosynthesis
becomes limited by biochemical capacity
(light-saturated)
• Diffuse/direct radiation; ε has been
observed to be greater under cloudy
conditions
Yebra et al, In preparation
15
Other factors explain the residuals
1.00
R0
Ftower/Fmod
0.90
0.80
T (ºC)
Da (Pa)
0.70
0.60
0
1000
2000
3000
VPD (Pa)
Q (W m-2)
Yebra et al, In preparation
Q / Qx
4000
AU-How
AU-Tum
FI-Hyy
IT-Ro1
IT-Ro2
NL-Loo
US-Bo1
US-Ha1
US-Ho1
US-MMS
US-Ne1
US-Ne2
US-Ne3
US-Ton
US-Var
US-WCr
5000 ENF
DBF/EBF
CRO/GRA
WSA
16
Evaluation against flux tower data
𝐸=
2
2
𝜎𝑡𝑜𝑤𝑒𝑟
− 𝜎𝑟𝑒𝑠
2
𝜎𝑡𝑜𝑤𝑒𝑟
Nash-Sutcliffe
𝑃𝐸 = 𝐹𝑡𝑜𝑤𝑒𝑟 − 𝐹𝑚𝑜𝑑
𝑅𝑃𝐸 =
𝐹𝑡𝑜𝑤𝑒𝑟 −𝐹𝑚𝑜𝑑
𝐹𝑡𝑜𝑤𝑒𝑟
Gc tower and local meteorology
(R0=0.76 and ɛmax=0.045)
Yebra et al, In preparation
17
Evaluation against flux tower data
● F Tower
○ F modelled
Gc tower and local
meteorology
(R0=0.76 and ɛmax=0.045)
Yebra et al, In preparation
18
Global estimate
• Global daily meteorological data 20002011
– Daily values of solar radiation, specific humidity and air pressure from Sheffield
et al. (2006), (1º)
• MODIS data 2000-2011
– 8-day/5km (0.05º) MOD42 from Distributed Active Archive Center of NASA
fPAR, EVI, Gc
• Compared to MOD17 (Zhao and Running
2010) and MPI(Jung et al 2009)
19
2000–2011 annual average
107 PgCy-1
Yebra et al, In preparation
4%< MOD17 (112 PgC y-1)
14% < MPI (122 PgC y-1)
20
2000–2011 annual average
21
2000–2011 annual average
22
GPP per different latitudes
800
GPP (gC m-2 y-1)
MOD17
MPI
600
this study
400
200
0
90
60
30
0
-30
-60
-90
Latitude ( N)
Yebra et al, In preparation
23
GPP per biome category
evergreen needleleaf forest
evergreen broadleaf forest
decidous and mixed forest
savannas
grasslands
croplands
MOD17
shrublands (low latitudes)
MPI
shrublands (high latitudes)
this study
other
0
10
20
30
40
GPP (PgC y-1)
Yebra et al, In preparation
24
Conclusions and future work
• The spatial and temporal patterns in our GPP
estimates compare favourably with other data
sets.
• Encouraging result, given
– the simplicity of our two-parameter model
– the lack of biome- or land-cover specific parameters
– the simple but explicit coupling between ET and GPP
• Test model performance fully
independently  OZFLUX!!
25
THANKS!
Marta Yebra PhD.
Research Fellow
Fenner School of Environment & Society
Forestry Building (48), room F15-G
The Australian National University
Linnaeus way, Canberra ACT 0200
+61-2-612 54107
[email protected]
www.anu.edu.au
https://researchers.anu.edu.au/researchers/yebra-m
http://www.researcherid.com/rid/B-5122-2011
Adjunct Research Scientist | CSIRO Land and Water
[email protected]
26
Local vs global inputs
Gc and meteo data from Tower
Yebra et al, In preparation
GcRS and global meteo data
27
AU-How
AU-How
40
Ftower
35
30
Fr
25
20
Fobs
15
Fr
10
Fc
5
Fc
AU-Tum
0
97 % Fr<Fc
40
19344610213243
7 193041
5 1635461021324310
273912213042 6 19283711223140
6 152434461324
35
2001
2001
2002
2003
2002
2004
2003
2005
2006
30
2004
78% Fr<Fc
25
2005 2006
20
15
10
5
0
8 172740 2 1120293847 9 19283746 8 243443 5 14233241 6 15
3912213042
6
19283711223140
62002152434461324
(R0=0.76 and ɛmax=0.045)
2001
2003
2004
2005
2006
28