Greening of the Earth and

its drivers

Shilong Piao, Zaichun Zhu

Sep 22, 2015

Changes in CO2

(Keeling et al., 1996. Nature)

Changes in vegetation index in northern high lattitudes

(Myneni et al., 1997. Nature)

Pioneer research:

- Inferred from changes in seasonal cycle of atmospheric CO2 concentration

- Direct evidence from satellite observations.

Background

Relationship between vegetation index and temperature

(Zhou et al., 2001)

Changes in global NPP during 1982-1999

(Piao et al., 2006)

Following research:

- Statistical analyses between vegetation and environmental factors (Zhou et al., 2001)

- Linking satellite observations and ecosystem models (Lucht et al., 2001;Piao et al.,

2006)

Background

Relationship between the constructed vegetation index

and observation (Los et al., 2013)

Dominant driving factors in LAI

(Mao et al., 2013)

Recent research:

- Attribution of global vegetation based on statistical method (Nemani et al.,

2003; Los et al., 2013)

- Attribution of global vegetation trends based on ecosystem models (Mao

et al., 2013)

Background

Main limitations:

Statistical methods – assume factors are independent

Ecosystem models – large model uncertainties

Uncertainties in satellite observations

Lack of nitrogen deposition and land cover change

processes

Background

Main data

Models S1 S2 S3 S4

CLM4.5 yes yes no yes

LPJ yes yes no yes

LPJ-GUESS yes yes no yes

LPX-Bern yes yes no yes

OCN yes yes no yes

ORCHIDEE yes yes no yes

VISIT yes yes no yes

CLM4 yes yes yes no

CALBE yes yes yes no

VEGAS yes yes no no

Simulations: S1: varying CO2

S2: varying CO2 and climate

S3: varying CO2, climate and nitrogen deposition

S4: varying CO2, climate and land cover change

CO2 fertilization: S1

Climate change: S2-S1

Nitrogen deposition: S3-S2

Land cover change: S4-S2

Satellite observations:

GIMMS LAI

GLOBMAP LAI

GLASS LAI

Freeze/Thaw data

Study period:

1982-2009

Schematic diagram of growing season

AVHRR GIMMS LAI3g - Global coverage

- 8 km

- 15 days

- 1982 – 2009

FT-ESDR（Freeze/Thaw data） - Global

- 25 km

- daily

- 1982 – 2009

Time line Start date (SG，20%) End date (SG，20%)

Thawed Periods

Growing Season

……

Note： - Evergreen needle

leaf forests

Definition of growing season

Growing Season:

Overlapping period of thawed period and SG based growing season.

Comparison between growing season

integrated leaf area index (LAI) and GPP

product from Beer et al. (2010) for 83

terrestrial eco-regions of the world (Olsen,

2001).

Definition of growing season

Trend Analysis

Spatial distribution of greening:

GIMMS LAI: ~25%

GLOBMAP LAI: ~43%

GLASS LAI: ~50%

Browning:

<4%

Optimal Fingerprints

Models CTL

years (chunks)

ACCESS1-0 500(17)

ACCESS1-3 500(17)

bcc-csm1-1 500(17)

bcc-csm1-1-m 400(14)

CCSM4 1051(37)

CESM1-BGC 500(17)

CESM1-CAM5 319(11)

CESM-FASTCHEM 222(7)

CESM1-WACCM 200(7)

GFDL-CM3 500(17)

GFDL-ESM2G 500(17)

GFDL-ESM2M 500(17)

HadGEM2-CC 240(8)

HadGEM2-ES 577(20)

Inmcm4 500(17)

MPI-ESM-LR 1000(35)

MPI-ESM-MR 1000(35)

MPI-ESM-P 1156(41)

Y = 𝛽𝑖𝑥𝑖 + 𝜀

𝑛

𝑖=1

Internal variability (18 models)

CO2 fertilization: S1

Climate change: S2-S1

Nitrogen deposition: S3-S2

Land cover change: S4-S2

GIMMS LAI3g

GLOBMAP LAI

GLASS LAI

Satellite observations

Model simulated signals

Attribution

CO2 fertilization:

70.1±29.4%, 0.048±0.020 m2m-2yr-1

Climate change:

8.1±20.6%, 0.006±0.014 m2m-2yr-1

Nitrogen deposition:

8.8±11.8%, 0.006±0.008 m2m-2yr-1

Land cover change:

3.7±14.7%, 0.003±0.010 m2m-2yr-1

Global scale:

Simple Conceptual Model

W =𝐴

𝐸=𝐶𝑎1.6𝑣1 −𝐶𝑖𝐶𝑎

𝑑𝑊

𝑊=𝑑𝐴

𝐴−𝑑𝐸

𝐸=𝑑𝐶𝑎𝐶𝑎−𝑑𝑣

𝑣+𝑑 1 −

𝐶𝑖𝐶𝑎

1 −𝐶𝑖𝐶𝑎

𝑑𝑊

𝑊=𝑑𝐴

𝐴−𝑑𝐸

𝐸=𝑑𝐶𝑎𝐶𝑎−1

2

𝑑𝑣

𝑣

The water use efficiency of photosynthesis is defined as the ratio of assimilation (A)

and transpiration (E) rates per unit of leaf area:

Here, v, Ci and Ca refer to leaf-to-air water vapor pressure deficit, intercellular and atmospheric

CO2 concentrations, respectively.

The quantity 1 − 𝐶𝑖 𝐶𝑎 has been modeled and observed as being proportional to 𝑣 (𝐹𝑎𝑟𝑞𝑢ℎ𝑎𝑟 𝑒𝑡 𝑎𝑙. 1993, 𝑀𝑒𝑑𝑙𝑦𝑛 𝑒𝑡 𝑎𝑙. 2011). Thus,

Simple Conceptual Model

E = 1.6𝑔𝑠𝑣

𝑑𝐸

𝐸=𝑑𝑔𝑠𝑔𝑠+𝑑𝑣

𝑣

𝑑𝑊

𝑊=𝑑𝐴

𝐴−𝑑𝑔𝑠𝑔𝑠−𝑑𝑣

𝑣=𝑑𝐶𝑎𝐶𝑎−1

2

𝑑𝑣

𝑣

𝑑𝐴

𝐴=𝑑𝐶𝑎𝐶𝑎+1

2

𝑑𝑣

𝑣+𝑑𝑔𝑠𝑔𝑠

E can be written as

The relative effects of changes in gs and v can be expressed as:

With that, Equation 3 can be written as

And,

Simple Conceptual Model

𝑑𝐴

𝐴=𝑑𝐶𝑎𝐶𝑎+1

2

𝑑𝑣

𝑣+𝑑𝑔𝑠𝑔𝑠

341ppm to 387ppm (~46 ppm) (Tans 2015), i.e. 𝑑𝐶𝑎

𝐶𝑎=13.5%

Climate Research Unit (CRU), we calculated that 𝑑𝑣

𝑣=2.25%

Experiment measurements suggests that 𝑑𝑔𝑠

𝑔𝑠= -5.0 ~ -3.0%

We can estimate: 𝑑𝐴

𝐴=9.7~11.7%

or 𝑑𝐴𝑐𝑜2

𝐴𝑐𝑜2=8.5~10.5%

FACE experiment: 𝑑𝑁𝑃𝑃

𝑁𝑃𝑃= 6.1~9.4%

𝑑𝐿𝐴𝐼

𝐿𝐴𝐼= 0.3~11.1%

Model simulation: 𝑑𝐺𝑃𝑃

𝐺𝑃𝑃=5.2~8.3%

𝑑𝑁𝑃𝑃

𝑁𝑃𝑃=5.2~9.0%

𝑑𝐿𝐴𝐼

𝐿𝐴𝐼=4.7~9.5%

The generally comparable relative changes of global vegetation growth estimated

from the simple conceptual models, the models, and the FACE experiments lend

credibility to our estimates of the response of global vegetation to elevated Ca

during 1982-2009.

Attribution

CO2 CLI NDE LUC

CLM4.5 0.054** -0.004 -- 0.007**

LPX-Bern 0.044** 0.003 -- 0.009

OCN 0.040** 0.016** -- -0.004**

LPJ 0.069** -0.002 -- -0.008

LPJ-GUESS 0.047** -0.010** -- 0.005

ORCHIDEE 0.036** -0.006 -- -0.008**

VISIT 0.089** 0.026** -- 0.018**

CLM4 0.034** 0.025** 0.011** --

CABLE 0.053** -0.006** 0.001** --

VEGAS 0.013** 0.013** -- --

MMEM 0.048**±0.020 0.006±0.014 0.006**±0.008 0.002±0.010

Individual Model Results:

Attribution

Continental scale: Africa, Asia

Attribution

Continental scale: Europe, Oceania

Attribution

Continental scale: South America, North America

CO2 fertilization effects

Generally uniformly distributed

The largest magnitudes of LAI trends

due to CO2 is found in the tropics

Relative change in LAI trends due to

CO2 is posited over semiarid regions

Global climate regions

Pixel scale

Climate change effects

Strong heterogeneity

Over-sensitive to decrease in precipitation (Piao et al, 2013)

NDE and LCC effects

Strong NDE effects were observed in

Southeast Asia

Slight decrease in Europe

Large uncertainties - only two models

were available to assess the NDE effects

Strong LCC effects were observed in

Southeast China and Southeast U.S.

Slight decrease in tropics

Large discrepancies among models,

sometimes differ in sign

Dominant driving factors

Climate change dominates 28% of the

vegetated area, followed by CO2(23%),

LCC(10%) and NDE(1%)

Other factors (OF) dominates 25%,

mainly found in regions influenced by

intensive ecosystem management, such

as northeast China, India and Europe.

Conclusion

We show a persistent and widespread greening over 25 to 50% of

the global vegetated area, whereas less than 4% of the globe shows

browning.

CO2 fertilization effects explains 70% of the observed greening

trend, followed by nitrogen deposition (9%), climate change (8%)

and land cover change (LCC) (4%).

CO2 fertilization and climate change effects explain most of the

greening trends in the tropics and the high latitudes, respectively.

LCC explains the regional browning observed in Southeast China

and Eastern United States.

Effects of non-modeled factors suggest further research is required

to reduce uncertainties in ecosystem models, especially their

modeling of climate change and land management.

Acknowledgements

This study was supported by the National Basic Research Program of China (Grant No.

2013CB956303), National Natural Science Foundation of China (41125004), Chinese

Ministry of Environmental Protection Grant (201209031), the 111 Project(B14001), and

the European Research Council Synergy grant ERC-SyG-610028 IMBALANCE-P.

We thank all people and institutions who provide data used in this study, in particular, the

TRENDY modelling group.

RBM is funded by NASA Earth Science. JGC thanks the support from the Australian

Climate Change Science Program. AA and TMP acknowledge support through EC FP7

grant LUC4C (grant 603542) and the Helmholtz Association ATMO programme.