Vegetation Trends In Australia

1
Vegetation TrendsIn Australia
Peter Briggs Michael Raupach, Edward King,
Michael Schmidt, Matt Paget, Jenny Lovell, Pep
Canadell CSIRO Marine and Atmospheric
Research Acknowledgements Damian Barrett, Susan
Campbell, Dean Graetz, Tim McVicar, Udaya
Senarath, Stephen Plummer and GlobCarbon (ESA)
2
Trends in fAPAR from SeaWiFS, Oct 1999 to Sep 2003
Knorr, W., Scholze, M., Gobron, N., Pinty, B.,
Kaminski, T. (2005). Global scale drought caused
atmospheric CO2 increase. EOS 18(18), 178-181.
3
Decadal Vegetation Greenness TrendsNorthern
Hemisphere Change in NDVI
1980s d(NDVI)/dt Summer 1982-1991
1990s d(NDVI)/dt Summer 1994-2002

• Gains from earlier onset of growing season are
  almost cancelled out by hotter and drier summers
  which depress assimilation
• Suggests a decreasing net terrestrial C sink

Angert et al. 2005 Dai et al. 2005 Buermann et
al. 2005 Courtesy Inez Fung 2005
4
NDVI AnomalyMonthly 1981-2003

• Anomaly as NDVImonth - ltNDVImonthgt
• Veg condition relative to expectations for that
  time of year
• "BPAL" AVHRR data series
• PAL (Pathfinder AVHRR Land, NASA) dataset 1981-94
• CSIRO EOC AVHRR dataset 1992-2003, with BISE
  filtering and sampling to match PAL
• 8-11 day max-NDVI composites, aggregated to
  monthly
• No atmos correctionNo BRDF correctionOnly a
  demonstration
• Shows space and time evolution of the drought
  cycles, e.g.
• Droughts 1982-83, 2002-04
• Wet La Nina 1988

5
NDVI Anomaly 1992-2004 (BPAL CATS1)

• Test marketed on
• Scientific colleagues
• Senior bureaucrats

6
NDVI Anomaly 1992-2004 (BPAL CATS1)

• Test marketed on
• Scientific colleagues
• Senior bureaucrats
• Response
• Holy sh
• Is this real?
• Whats going on?

7
NDVI BPAL AVHRR 1982 to 2000
Mean

• Mean

• Rain is main constraint so NDVI follows rain map
  except
• Northern tropics high VPD and strong seasonality
  of rain limit growth
• Tasmania light and temp-limited in winter

StdDev
Std Dev

• Areas of high SD show agricultural cropping zones
• Desert interior green flush as ephemeral lakes
  and rivers respond to seasonal runoff from the
  north

8

• Confirmation of the 2000-2004 trend in AVHRR NDVI
• Comparison with
• Other AVHRR NDVI treatments
• NDVI from other sensors
• Other products from other sensors
• How the trend varies with bioclimatic region
• Comparison of trends within major drainage
  divisions
• East Australian drought recovery 2002-2005

9
Satellite vegetation time series for Australia

• AVHRR NDVI
• BPAL (5 km, 8-11 day max-NDVI composite,
  BISE-filtered)
• CATS1 (1 km, 10 day max-NDVI composite, no
  BRDF, no cloud)
• CATS2b ( CATS1 with CLAVR cloud clearing)
• CATS2a ( CATS2b with BRDF correction)
• MODIS NDVI
• SeaWifs NDVI
• SPOT-Vgt NDVI
• LAI from GlobCarbon project
• SPOT-Vgt, ATSR2/AATSR, MERIS
• LAI from MODIS

Converted to fraction cover fC 1 ? e?kLAI
with k 0.5
10
Australia vegetation greenness trends (1990-2005)
NDVI MODIS, SeaWiFS, SpotVgt
NDVI BPAL
NDVI, fC
NDVI CATS2a(CLAVR, BRDF)
fraction cover fC from GlobCarbon LAI
11
NDVI 1992-2004NE Coast

• 0Australia
• 1 NE Coast
• 1.1 NE Coast (sea)
• 1.2 NE Coast (Burd-Fitz)
• 2 SE Coast
• 3 Tasmania
• 4 MDB
• 4.1 MDB (wet)
• 4.2 MDB (agric)
• 4.3 MDB (semiarid)
• 5 SA Gulfs
• 6 SW Coast
• 7 Indian Ocean
• 8 Timor Sea
• 9 Carpentaria
• 10 Lake Eyre
• 11 Bulloo-Bancannia
• 12 Western Plateau

12
NDVI 1992-2004Murray-Darling Basin
0Australia 1 NE Coast 1.1 NE Coast (sea) 1.2
NE Coast (Burd-Fitz) 2 SE Coast 3 Tasmania 4
MDB 4.1 MDB (wet) 4.2 MDB (agric) 4.3 MDB
(semiarid) 5 SA Gulfs 6 SW Coast 7 Indian
Ocean 8 Timor Sea 9 Carpentaria 10 Lake
Eyre 11 Bulloo-Bancannia 12 Western Plateau
13
Drought recovery (or not) in SE AustraliaSeaWiFS
Fraction Cover for 5 Decembers(NDVI scaled by
GlobCarbon FC)
Before Drought
Drought Max
6 Months Ago
14
Marginal country slow to recover(if at all yet)
Before Drought
Drought Max
6 Months Ago
15
Annual Rainfall (mm d-1)
Before Drought
Drought Max
6 Months Ago
16
Various NDVIs vs GlobCarbon FC By Drainage
Division, Monthly, 1999-2002

• Up to 36 pts (months) per drainage division
• Periods of known sensor problems removed from
  AVHRR
• For planned rescaling exercise

SeaWiFS
BPAL
?Used in previous maps
MODIS
CATS2a
17
Various NDVIs vs GlobCarbon FC By Drainage
Division, Monthly, 1999-2002

• Well-defined relationship between all sensors
• Saturation of greenness wrt GlobCarbon in all
  cases
• Only MODIS showing stratification wrt biogeography

SeaWiFS
BPAL
MODIS
CATS2a
18
Carbon consequences of vegetation greenness
changes

• Model
• Let biospheric C obey rate equation dC/dt FC ?
  kC, with mean turnover rate k. If NPP changes
  suddenly by dFC, then while Dt ltlt 1/k, the change
  in C is
• Assume NPP green leaf cover fraction
• Then biospheric C change associated with a
  perturbation in green leaf cover is
• Numbers
• Take Dt 1 year FC 1 GtC/y dfGL/fGL
  0.2 (a low value)
• gt DC 0.2 GtC 0.2 PgC 200 MtC 730 Mt
  CO2
• Compare Australian GHG emissions (2002 NGGI)
  were 550 Mt CO2eq

19
Conclusions (1) Trends in vegetation greenness,
are they real?

• Broad agreement from multiple sensors on
  Australian vegetation trends
• Continent-wide decline 2000-2004
• Long AVHRR record suggests decline commenced
  around 1998
• Areas of disagreement
• MODIS and GlobCarbon give substantially different
  estimates of LAI (therefore fraction cover)
• For Australia, GlobCarbon is closer to the level
  we expect (from other RS and flux station work)
• Trend from AVHRR probably too strong modern
  sensors (SeaWiFS, SpotVGT, MODIS) appear to show
  some recovery not seen in AVHRR.
• Due to continued use of at-launch calibrations by
  AVHRR?
• Important to continue to follow trends through
  the recovery
• However
• Areas of agreement are more significant than
  areas of disagreement

20
Conclusions (2) Whats going on?

• Strongest relative decline occurs in semi-arid
  zones (rainfall lt 600 mm)
• Declining trends in heavily forested areas not
  revealed here, but
• Scale, NDVI saturation may be issues
• See Helen Cleugh for MODIS/flux station results
• Trends are too widespread for land use change to
  be a significant driver
• But land use (eg. stocking rates in rangeland)
  may have an effect
• Likely drivers
• Rainfall no major floods in most of the country
  for 10-15 years (until Cyclone Larry)
• Warming 2002-2004 was a hot drought 2005 hot
  and a little wetter
• Possible contributing processes
• Effects of warming and dryness on fire,
  heterotrophic respiration
• Soil evaporation (half of Australian ET) is
  favoured in the competition for soil water,
  causing transpiration to fall

21
Messages

• Loss of terrestrial C in 2002-2004 is similar to
  total anthropogenic Australian GHG emissions
• Terrestrial biospheric C is highly dynamic and
  vulnerable to drought and probably also to
  warming
• Long term satellite environmental time series are
  an important tool for examining trends
• Modern sensors (with their overlap) will be
  invaluable for validating and carrying on the
  series
• Too soon to say what will happen to continental
  vegetation next
• Many parts of the country still waiting for
  drought recovery
• Warming, associated with drought, is particularly
  worrying
• Is this a foretaste of the future Australia?