ArticlePDF Available

Abstract and Figures

Land surface phenological cycles of vegetation greening and browning are influenced by variability in climatic forcing. Quantitative spatial information on phenological cycles and their variability is important for agricultural applications, wildfire fuel accumulation, land management, land surface modeling, and climate change studies. Most phenology studies have focused on temperature-driven Northern Hemisphere systems, where phenology shows annually recurring patterns. However, precipitation-driven non-annual phenology of arid and semi-arid systems (i.e., drylands) received much less attention, despite the fact that they cover more than 30% of the global land surface. Here, we focused on Australia, a continent with one of the most variable rainfall climates in the world and vast areas of dryland systems, where a detailed phenological investigation and a characterization of the relationship between phenology and climate variability are missing. To fill this knowledge gap, we developed an algorithm to characterize phenological cycles, and analyzed geographic and climate-driven variability in phenology from 2000 to 2013, which included extreme drought and wet years. We linked derived phenological metrics to rainfall and the Southern Oscillation Index (SOI). We conducted a continent-wide investigation and a more detailed investigation over the Murray–Darling Basin (MDB), the primary agricultural area and largest river catchment of Australia. Results showed high inter- and intra-annual variability in phenological cycles across Australia. The peak of phenological cycles occurred not only during the austral summer, but also at any time of the year, and their timing varied by more than a month in the interior of the continent. The magnitude of the phenological cycle peak and the integrated greenness were most significantly correlated with monthly SOI within the preceding 12 months. Correlation patterns occurred primarily over northeastern Australia and within the MDB, predominantly over natural land cover and particularly in floodplain and wetland areas. Integrated greenness of the phenological cycles (surrogate of vegetation productivity) showed positive anomalies of more than 2 standard deviations over most of eastern Australia in 2009–2010, which coincided with the transition from the El Niño-induced decadal droughts to flooding caused by La Niña.
Content may be subject to copyright.
Biogeosciences, 11, 5181–5198, 2014
© Author(s) 2014. CC Attribution 3.0 License.
Land surface phenological response to decadal climate variability
across Australia using satellite remote sensing
M. Broich1,*, A. Huete1, M. G. Tulbure2, X. Ma1, Q. Xin4, M. Paget3, N. Restrepo-Coupe1, K. Davies1, R. Devadas1,
and A. Held3
1Plant Functional Biology and Climate Change Cluster, University of Technology, Sydney, NSW 2007, Australia
2Centre of Ecosystem Science, School of Biological, Earth and Environmental Sciences,
University of New South Wales, Kensington NSW 2052, Australia
3CSIRO Marine and Atmospheric Research, Pye Laboratory, Acton, ACT, 2600, Australia
4Ministry of Education Key Laboratory for Earth System Modeling, Center for Earth System Science,
Tsinghua University, Beijing 100084, China
*now at: Centre of Ecosystem Science, School of Biological, Earth and Environmental Sciences,
University of New South Wales, Kensington NSW 2052, Australia
Correspondence to: M. Broich (
Received: 31 March 2014 – Published in Biogeosciences Discuss.: 28 May 2014
Revised: 22 August 2014 – Accepted: 22 August 2014 – Published: 29 September 2014
Abstract. Land surface phenological cycles of vegetation
greening and browning are influenced by variability in cli-
matic forcing. Quantitative spatial information on pheno-
logical cycles and their variability is important for agricul-
tural applications, wildfire fuel accumulation, land manage-
ment, land surface modeling, and climate change studies.
Most phenology studies have focused on temperature-driven
Northern Hemisphere systems, where phenology shows an-
nually recurring patterns. However, precipitation-driven non-
annual phenology of arid and semi-arid systems (i.e., dry-
lands) received much less attention, despite the fact that they
cover more than 30% of the global land surface. Here, we
focused on Australia, a continent with one of the most vari-
able rainfall climates in the world and vast areas of dryland
systems, where a detailed phenological investigation and a
characterization of the relationship between phenology and
climate variability are missing.
To fill this knowledge gap, we developed an algorithm to
characterize phenological cycles, and analyzed geographic
and climate-driven variability in phenology from 2000 to
2013, which included extreme drought and wet years. We
linked derived phenological metrics to rainfall and the South-
ern Oscillation Index (SOI). We conducted a continent-
wide investigation and a more detailed investigation over the
Murray–Darling Basin (MDB), the primary agricultural area
and largest river catchment of Australia.
Results showed high inter- and intra-annual variability in
phenological cycles across Australia. The peak of pheno-
logical cycles occurred not only during the austral summer,
but also at any time of the year, and their timing varied
by more than a month in the interior of the continent. The
magnitude of the phenological cycle peak and the integrated
greenness were most significantly correlated with monthly
SOI within the preceding 12 months. Correlation patterns
occurred primarily over northeastern Australia and within
the MDB, predominantly over natural land cover and par-
ticularly in floodplain and wetland areas. Integrated green-
ness of the phenological cycles (surrogate of vegetation pro-
ductivity) showed positive anomalies of more than 2 stan-
dard deviations over most of eastern Australia in 2009–2010,
which coincided with the transition from the El Niño-induced
decadal droughts to flooding caused by La Niña.
Published by Copernicus Publications on behalf of the European Geosciences Union.
5182 M. Broich et al.: Land surface phenological response to climate variability across Australia
1 Introduction
Vegetation phenology refers to the response of vegetation to
inter- and intra-annual variations in climate, specifically ir-
radiance, temperature and water (Myneni et al., 1997; White
et al., 1997; Zhang et al., 2003). Vegetation phenology is a
useful indicator in the study of the response of ecosystems
to climate variability (Zhang et al., 2012; Richardson et al.,
2013), and an important parameter for land surface, climate
and biogeochemical models that quantify the exchange of
water, energy and gases between vegetation and the atmo-
sphere (Pitman, 2003; Eklundh and Jönsson, 2010). A vari-
ety of applications that require the characterization of veg-
etation phenology include crop yield quantification, wildfire
fuel accumulation, vegetation condition, ecosystem response
to climate variability and climate change, and ecosystem re-
silience (Schwartz, 2003; Liang and Schwartz, 2009; Peñue-
las et al., 2009). The phenology of the vegetated land surface
(land surface phenology, hereafter phenology) is “the sea-
sonal pattern of variation in vegetated land surfaces observed
from remote sensing” (Friedl et al., 2006).
In temperature-limited systems, phenological cycles occur
on an annual basis, starting in spring and ending in autumn.
Existing algorithms aiming to characterize phenological cy-
cles from remotely sensed spectral vegetation “greenness”
indices perform well for ecosystems in temperature-driven
mid and high latitudes (Eklundh and Jönsson, 2010; Ganguly
et al., 2010). However, in ecosystems where rainfall is lim-
ited and highly variable, such as semi-arid and arid systems
(i.e., drylands; United Nations, 2011), phenological cycles
may be irregular in their length, timing, amplitude and reoc-
currence interval, occur at any time of the year, or not occur
at all in a given year (Brown and de Beurs, 2008; Ma et al.,
2013; Walker et al., 2014; Bradley and Mustard, 2007).
Despite the fact that drylands cover over 30% of the
global land surface and occur on every continent (United
Nations, 2011), their rainfall-driven phenology that features
non-annual cycles has not been well characterized. Here, we
focused on Australia, a continent where drylands cover more
than 80 % of the land surface. Recent reports by the Intergov-
ernmental Panel on Climate Change highlighted not only the
importance of quantifying vegetation phenology in general
(IPCC, 2007, 2013; Schwartz, 2013), but also pointed to a
lack of phenological studies for Australia and New Zealand
(Keatley et al., 2013; IPCC, 2001, 2007). We developed an
algorithm to characterize phenological cycles, and analyzed
the phenology of Australia as an example of a rainfall-driven
dryland system. Phenology on the landscape-to-continental
scale is typically derived using time series of remotely sensed
vegetation greenness indices such as the normalized differ-
ence vegetation index (NDVI) and the enhanced vegetation
index (EVI) (de Beurs and Henebry, 2008). Several stud-
ies have used NDVI time series recorded by the Advanced
Very High Resolution Radiometer (AVHRR) to investigate
long-term phenological trends induced by climate change
(Moulin et al., 1997; Zhang et al., 2012). More recent studies
used EVI time series recorded by the MODerate-resolution
Imaging Spectroradiometer (MODIS) that has better geomet-
ric correction and increased resolution compared to AVHRR
(Tan et al., 2011). Compared with NDVI, EVI is less sensitive
to residual atmospheric contamination and soil background
variations, and has a larger dynamic range of sensitivity to
vegetation greenness (Huete et al., 2002). EVI time series
measure change in an integrated property commonly referred
to as “greenness” has been found to be correlated with sub-
pixel chlorophyll content and leaf area index (Huete et al.,
Parameters describing phenological cycles (hereafter phe-
nological metrics) can be used to quantify the influence of
climate change and variability on phenological magnitude
(Ma et al., 2013; Brown et al., 2010) and timing (Guan et al.,
2014a). Australia has one of the most variable climates in the
world, subject to high inter-annual rainfall variability due to
the influence of the El Niño–Southern Oscillation (ENSO)
(Nicholls, 1991; Nicholls et al., 1997). Previous studies in-
vestigated the relationship between vegetation index time se-
ries and rainfall globally, and the correlation with soil mois-
ture for Australia (Chen et al., 2014a; Andela et al., 2013).
However, studies quantifying the relationship between phe-
nological magnitude and ENSO-related climate variability,
as shown for example for Africa (Brown et al., 2010; Philip-
pon et al., 2014), are missing. Here, we analyzed phenologi-
cal magnitude responses to climate variability through a pe-
riod of time from 2000 to 2013. This period encompassed the
Australian Millennium Drought from 2001 to 2009 (van Dijk
et al., 2013) and the 2010–2011 La Niña-associated flooding
(Heberger, 2011; Australian Bureau of Meteorology, 2014a),
and focused on one of the most affected areas, the Murray–
Darling Basin (MDB) in the southeast of Australia (van Dijk
et al., 2013; Kirby et al., 2012; Australian Bureau of Meteo-
rology, 2014b).
Particular emphasis was given to the MDB, the catchment
of Australia’s largest river system, and associated ecolog-
ically valuable floodplain and wetland ecosystems and the
primary agricultural area of the continent (Connell, 2007).
The objectives of this study were (1) to characterize the
inter- and intra-annual variability of phenological cycles of
greening and browning, including non-annual cycles across
Australia, a continent with vast areas of dryland ecosys-
tems, and (2) to investigate the relationships between the
derived phenological magnitude and rainfall, as well as be-
tween phenological magnitude and the Southern Oscillation
Index (SOI; Trenberth and Caron, 2000), a proxy of ENSO,
across the entire continent and in more detail for the MDB.
Biogeosciences, 11, 5181–5198, 2014
M. Broich et al.: Land surface phenological response to climate variability across Australia 5183
Figure 1. Land cover map of Australia showing closed and open
tree cover in dark and light green, respectively. The purple colors
that occur predominantly in the southwest and southeast represent
crops and pasture. Brown marks shrubs, orange colors mark tus-
sock grass and light brown colors mark hummock grass cover across
most of the semi-arid and arid interior (land cover classes were ag-
gregated based on Lymburner et al., 2011). The most prominent to-
pographic feature is the Great Dividing Range that runs along the
eastern seaboard. Locations of the 21 OzFlux flux tower sites and
15 additional sites are shown as red and blue circles. We used the
EVI time series at the sites for phenological algorithm development
and testing (site list provided in Table 1). The phenology for the
sites marked by large black circles is presented and discussed in
Sect. 2.2.3. The bottom-left panel shows the extent of the MDB.
2 Methods
2.1 Study area and data used
Australia covers an area of more than 7.6 millionkm2, and
climatic zones range from tropical in the north to temper-
ate in the south (Fig. 1). Average rainfall does not exceed
600mm over 80% of the land area, and is less than 300 mm
over 50% of the land area (Australian Bureau of Meteorol-
ogy, 2014c). Northern Australia is dominated by savanna,
whereas most of the country is covered by grassland and
desert vegetation (Köppen, 1884). Forest occurs at higher
elevations in the temperate southwest and southeast, where
large areas of the lowlands are used for rain-fed agriculture
(Fig. 1; Lymburner et al., 2011). The MDB contains Aus-
tralia’s primary agricultural area, and occupies 14% of Aus-
tralia in the southeast of the continent (Fig. 1).
For algorithm development and testing, we used a set
of EVI time series at 36 sites distributed across Australia
(Fig. 1). These 36 sites represented a range of land cover and
climatic zones (Table 1; Lymburner et al., 2011; Australian
Bureau of Meteorology, 2014c) to ensure that the algorithm
effectively captures the variability in phenology across the
country, and we used them to determine optimized algorithm
parameters. The majority (21) of our test sites were flux
tower sites from the OzFlux network (2014). We selected 15
additional test sites to represent a wider coverage of climate
conditions, vegetation cover and land uses.
As input data for the phenological characterization, we
sourced EVI MOD13C2 and MOD13A1 with a tempo-
ral resolution of 16 days for the 18 February 2000 to
22 April 2013 time period (NASA Land Processes Dis-
tributed Active Archive Center, 2014).
We used the 5.6 km product (MOD13C2) to charac-
terize the biogeographic patterns of vegetation phenology
across the entire Australian continent, and the 500m prod-
uct (MOD13A1) to investigate the phenological patterns in
more detail across the MDB. We chose the 16-day versions
of the products, as they attenuate the noise present in higher
temporal resolution versions (Solano et al., 2012).
To analyze the responses of phenological metrics to rain-
fall variability, we used monthly data from the Tropical Rain-
fall Monitoring Mission Project (TRMM_3B43.v7 product;
Goddard Space Flight Center, 2014) with a 0.25×0.25
spatial resolution for 1999–2012. Instead of using gridded
rainfall data interpolated from widely spaced weather sta-
tions across large areas of the interior, we opted for remotely
sensed rainfall measured by TRMM, which is systematic
across space and time.
To analyze the responses of phenological metrics to
ENSO, we used monthly data of the Southern Oscillation
Index (SOI) obtained from the Australian Bureau of Mete-
orology (2014d). SOI represents the standardized difference
in air pressures between Darwin and Tahiti, and serves as a
proxy of convection in the western Pacific caused by ENSO
sea surface temperature anomalies (Trenberth and Caron,
Across the MDB, we used the Dynamic Land Cover data
set provided by Geoscience Australia (Lymburner et al.,
2011) to investigate the differences between the phenolog-
ical responses to SOI and rainfall over natural and man-
aged land cover types. We derived the natural land cover
class by grouping land cover dominated by trees, shrubs and
grasses. The managed land cover classes encompassed rain-
fed and irrigated agriculture and pasture. Almost a third of
the basin’s area is managed for cropping and pasture (Lym-
burner et al., 2011). We also analyzed the phenological re-
sponse over the ecologically valuable floodplain and wet-
land areas of MDB (Kingsford et al., 2004), and evaluated
the floodplain’s response to SOI as a proxy of ENSO-related
drought and flooding. Biogeosciences, 11, 5181–5198, 2014
5184 M. Broich et al.: Land surface phenological response to climate variability across Australia
Table 1. Names, locations, land cover class (Lymburner et al., 2011) and average annual rainfall amounts (Australian Bureau of Meteorology,
2014c) for the 36 sites shown in Fig. 1
Site name OzFlux Site code Lat. Long. Land cover Average annual
site Fig. 1 (S) (E) classes rainfall (mm)
Nullaboure NU 30.275 127.175 Woody
GBD 29.125 133.075 Woody
Lake Eyre LE 27.425 137.225 Woody
GWW 30.225 120.625 Woody
East of
Shark Bay ESB 24.475 116.325 Woody
CW 24.125 124.175 Woody
IEA 29.425 144.225 Woody
Calperum X CP 34.025 140.375 Woody
wheat belt
WAW 32.125 117.425 Herbaceous
cropping IC 35.275 145.275 Herbaceous
Springs X AS 22.275 133.225 Herbaceous
Desert SD 20.475 124.025 Herbaceous
Hamersley X HA 22.275 115.725 Woody
X GWWF 31.925 120.075 Herbaceous
tussock QTU 21.225 143.075 Herbaceous
Biogeosciences, 11, 5181–5198, 2014
M. Broich et al.: Land surface phenological response to climate variability across Australia 5185
Table 1. Continued.
Site name OzFlux Site code Lat. Long. Land cover Average annual
site Fig. 1 (S) (E) classes rainfall (mm)
Queensland NWQ 19.525 140.025 Woody
Sturt Plain X SP 17.175 133.375 Woody
Riggs Creek X RC 36.625 145.575 Herbaceous
Arcturus X AR 23.875 149.275 Woody
trees, open 800
Gingin X GG 31.375 115.725 Woody
Otway X OT 38.525 142.825 Herbaceous
Wombat X WO 37.425 144.075 Woody
Plain X CU 33.725 150.725 Woody
Dry River X DR 15.275 132.375 Woody
Creek X WC 37.425 145.175 Woody
Daly River
pasture X DRP 14.075 131.375 Woody
trees, open 1200
West of
WNQ 16.275 142.475 Woody
Nimmo X NI 36.225 148.575 Woody
Samford X SA 27.425 152.825 Woody
Tumbarumba X TU 35.675 148.175 Woody
trees, open 1600
Springs X HO 12.475 131.175 Woody
trees, open 1600
peninsula DP 15.125 125.725 Woody
Dargo X DA 37.125 147.175 Herbaceous
Tasmania NWT 41.225 145.175 Woody
Cape Tribula-
tion X CT 16.125 145.375 Woody
Daintree X DT 16.225 145.425 Woody
8000 Biogeosciences, 11, 5181–5198, 2014
5186 M. Broich et al.: Land surface phenological response to climate variability across Australia
Figure 2. Algorithm steps applied to the 14-year MODIS EVI time series (MOD13C2 single 5.6km pixel) for the Alice Springs flux site,
representing semi-arid mulga (Acacia) woodland of the center of Australia. (a) EVI time series after screening out low-quality observations
(brown circles), EVI time series after gap filling and smoothing (blue circles), and flagged minimum and peak-of-cycle points (green dia-
monds). (b) Curves fitted as seven-parameter double logistic functions (red squares) characterizing the phenological cycles, and identifying
start- and end-of-cycle points (yellow circles) delineating the cycles. The timing, length, amplitude, and magnitudes of the phenological
cycles at the site vary inter-annually.
2.2 Phenology metrics and algorithm
2.2.1 Phenology metrics
To account for non-annual vegetation dynamics, we defined a
phenological cycle not as an annually or seasonally recurring
event, but more broadly, as a cycle of EVI-measured greening
and browning that may occur more than once per year or may
skip a year entirely and not occur for one or more years.
We modeled phenological cycle curves and key properties
of each phenological cycle in the form of curve metrics. The
phenological metrics modeled the timing and magnitude of
key transitional points on the cycle’s curve, and included the
timing and magnitude of the minimum points before and af-
ter a phenological cycle, the peak point of the cycle and the
start and end point of the cycle. In addition, we also calcu-
lated the integrated area between the start and end points of a
cycle as a surrogate of vegetation productivity during a cycle
(Zhang et al., 2013). By tracking the phenological cycle met-
rics over time, we characterized the intra- and inter-annual
variability of the phenological cycle, and thereby vegetation
growth patterns.
2.2.2 Data pre-processing
We used the quality assurance flags in the MOD13 prod-
ucts to discard observations with insufficient quality, which
included any observation with either VI usefulness greater
than code “10”, snow cover, high aerosol or climatology
aerosol quantity, mixed or high clouds present, or water in the
Land/Water Flag. For each pixel, we first used cubic spline
interpolation (Dougherty et al., 1989) to temporally gap-fill
the data points discarded in the previous filtering step. Next,
we smoothed the time series for each pixel using a Savitzky–
Golay smoothing filter (Savitzky and Golay, 1964) with a
window width of 15 time steps. This step effectively reduced
the remaining noises in the time series that would otherwise
impact the identification of minimum and maximum points
and the subsequent fitting of a mathematical curve that we
conducted to characterize the phenological cycles in a con-
sistent way.
2.2.3 Curve fitting and phenological metric derivation
We identified local minimum and maximum points of the
per-pixel time series using a moving window of nine time
steps and a >0.01EVI amplitude threshold to identify cy-
cles of greening and browning. We used the identified min-
imum points to define the temporal extent of phenological
cycles in the entire time series. We then fitted the seven-
parameter double logistic model for each identified interval.
We did not expect one or multiple phenological cycles in
fixed intervals of the year. We thus allowed cycles to be char-
acterized at any time to better represent the highly variable
rainfall-driven phenological patterns across Australia’s vast
drylands and dual cycles in cropping and pasture zones. We
fitted seven-parameter double logistic curves to each cycle in
the per-pixel time series, defined as
EVI(t) =VminaVmax Vmina
Sa+Vmax Vminb
where Vminaand Vminbare equal to the first and second
minimum EVI, respectively. Vmax is the high asymptote in
the double logistic model, and Tmidais the time when EVI
reached half of Vmax Vmina.Tmidbis the time when EVI
Biogeosciences, 11, 5181–5198, 2014
M. Broich et al.: Land surface phenological response to climate variability across Australia 5187
Figure 3. Examples of temporal variability of the characterized phenological cycles for the Sturt Plains, Calperum, and Great Western
Woodlands sites (refer to Fig. 1 and Table 1 for the sites’ locations and descriptions, respectively), based on 14 years of MODIS EVI
data after screening out low-quality observations (brown circles), EVI time series after gap filling and smoothing (blue circles), fitting
seven-parameter double logistic functions (red squares) and identifying the start- and end-of-cycle points (yellow circles) delineating the
characterized phenological cycles.
reached half of Vmax Vminb.Saand Sbare the scale pa-
rameters on the increasing and decreasing sides of the curve,
respectively. We identified the start and end points of each
cycle as the points where the EVI reached 20% of the ampli-
tude, between the first minimum and the peak, and the peak
and the second minimum, respectively, as also used in other
studies (Eklundh and Jönsson, 2010; Tan et al., 2011; Jones
et al., 2011; Delbart et al., 2005).
An example of the algorithm processing steps is shown
for the Alice Springs flux tower site (Fig. 2). The site repre-
sents Acacia woodlands in the arid interior of Australia. The
site serves as an example showing how our algorithm derives
phenological metrics to characterize the high temporal vari-
ability in phenological cycles for the interior of Australia.
We provide further examples of how the algorithm char-
acterized the phenological cycles over different land cover
types in different rainfall zones in Fig. 3. The sites’ locations
and descriptions are provided in Fig. 1 and Table 1, respec-
2.3 Analysis of spatial–temporal patterns of phenology
across Australia
After deriving phenological cycles and their metrics from
per-pixel greenness time series, we analyzed the metrics
across Australia at two levels of temporal aggregation: (1) in
the form of summary statistics (mean and standard deviation)
across greenness time series to quantify overall phenological
variability over the 14-year time series; and (2) in the form
of inter-cyclic variability as the difference between a metric
of one cycle and the following cycle over the 14-year time
For a given site, we calculated for example the mean peak
magnitude and the peak magnitude’s standard deviation. An
example of inter-cycle variability of metrics is our analysis Biogeosciences, 11, 5181–5198, 2014
5188 M. Broich et al.: Land surface phenological response to climate variability across Australia
of peak timing for all peaks across the time series. We also
analyzed the deviation of an individual phenological cycle
integral relative to the expected variability. For this purpose,
we calculated the standardized anomaly of each cycle’s inte-
gral as the difference of the cycle’s integral from the mean
integral divided by the standard deviation of the integrals.
2.4 Analysis of spatial–temporal patterns of Australian
phenology in response to rainfall and SOI
We further analyzed the statistical relationship between phe-
nological cycle peak magnitude and cycle-integrated green-
ness with TRMM rainfall and SOI (four combinations of
correlation analyses) across Australia and in more detail
for the MDB. The cycle peak magnitude represents maxi-
mum greenness, while the cycle-integrated greenness serves
as a proxy of ecosystem productivity (Zhang et al., 2013).
We used non-parametric Spearman rank correlation tests
(Lehmann and D’Abrera, 1975), hereafter Spearman rho, to
determine the strength and significance of monotonic rela-
tionships between rainfall and each of the two phenology
metrics, as well as SOI and the two phenology metrics. We
evaluated relationships between rainfall and SOI as the ex-
planatory variables binned over different intervals and with
different lead times to the phenological cycle integral and
peak magnitude, which were used as the response variables.
We binned rainfall accumulation for intervals of 1 to 12
months and average SOI values for periods of 1 to 12 months
up to 12 months prior to the phenological cycle peak.
The underlying assumption for investigating Spearman
rho correlations between phenology and rainfall or SOI was
that a significant and strong monotonic relationship between
a phenological metric and preceding rainfall or SOI sug-
gested that the phenology metric (peak magnitude and in-
tegrated greenness) is likely driven by the respective climate
Aiming to identify correlation patterns and how these pat-
terns change as a function of binning interval (1–12 months)
and lead times (up to 12 months), we extracted for each pixel
and binning interval the most significant test result. For each
potential driver and binning interval, we analyzed the lead
time, correlation and significance value. We illustrated the
results only for areas that were significant (pvalue less than
0.05) and had a rho value of greater than 0.6.
Using the above methodology, we conducted a continent-
wide analysis and a higher-resolution analysis investigat-
ing the relationship of SOI with phenology metrics for the
MDB in southeastern Australia. Within the MDB, we fur-
ther investigated the relationship between SOI and phe-
nology (differences in correlation patterns) over natural
and managed land cover types, as well as the catchment’s
floodplain and wetlands.
TS1: new figures with non cab letters below.
TS2: ok thanks.
Fig 4
Figure 4. Mean of peak magnitude (a), mean of minimum mag-
nitude (b), standard deviation of peak magnitude (c) and standard
deviation of minimum magnitude (d). A map of the dominant land
cover type is provided in Fig. 1.
3 Results
3.1 Mean and variability of peak and minimum
magnitude as well as start and end of cycle
timing across 14 years
We evaluated the mean and variability of the peak and min-
imum magnitude across the 14-year time series to investi-
gate the inter-annual variations in vegetation phenology. The
highest mean peak magnitude occurred in a narrow area cov-
ered predominantly by evergreen humid tropical forest along
the northeastern coast (areas with high EVI in Fig. 4a and b).
The same area also had the highest mean minimum magni-
tude values, indicating that greenness was persistently high
(light colored areas in Fig. 4b). Other areas with high levels
of persistent greenness (areas with high mean peak magni-
tude and high mean minimum magnitude) included temper-
ate grasslands in coastal locations of southeastern Australia,
temperate broadleaf forest in the southeast and southwest of
the continent, and across most of Tasmania (light colored ar-
eas in Fig. 4a and b). The largest mean seasonal amplitude
(peak minus minimum magnitude) occurred in areas used for
crop cultivation and grazing in the southwest and southeast.
Areas of low mean peak amplitude were found across large
parts of the interior (darker toned areas in both Fig. 4a and
b), with the exception of the desert river beds.
The highest level of variability in peak magnitude occurred
over cropped areas in the southeast and southwest of Aus-
tralia (light colored areas in Fig. 4c). High variability of
peak magnitude over natural vegetation cover was observed
for example for regions predominantly covered with tropical
Biogeosciences, 11, 5181–5198, 2014
M. Broich et al.: Land surface phenological response to climate variability across Australia 5189
Fig 5
Figure 5. Mean Julian day of the start of the phenological cycles
(a1), standard deviation of the start of the phenological cycles in
number of days (b1), mean Julian day of the end of the phenological
cycles (a2) and standard deviation of the end of the phenological
cycles in number of days (b2) across the 14-year time series.
tussock grasses in the inland north and northeast, as well as
areas with predominant chenopod woody shrub cover along
the Great Australian Bight along the southern coast of Aus-
tralia (light colored areas in Fig. 4c). High variability in mini-
mum magnitude occurred at higher elevations of the southern
Great Dividing Range in the southeast of Australia (light col-
ored areas in Fig. 4d) and around the center of the arid Lake
Eyre, which is the lowest point of the continent.
We also evaluated the mean and variability of the start and
end of cycle timing across the 14-year time series. Across
western and southeastern Australia, the mean start of cycles
occurred during the first half of the year, and the mean end of
the cycle occurred in the second half of the year (Fig. 5a1).
Across northern and eastern Australia, the mean start of cy-
cles occurred during the second half of the year, and the mean
end of cycle occurred in the first half of the following year
(Fig. 5a2). The variability in the start and end of the cycle
was highest across interior Australia, with the area of high
variability being higher for the end of cycle timing (Fig. 5b1
and 2).
3.2 Inter-cycle variability in peak timing
The timing of the first cycles’ peak within each year showed
large variation from one year to another across most of Aus-
tralia (Fig. 6). Variations in peak timing were observed over
most of interior Australia. Peak timing was later than aver-
age in 2001, 2004 and 2005 (Fig. 6), but earlier in 2010–
2012 over interior Australia (Fig. 6). The peak timing in the
wet tropical savannas of the Northern Territory and for most
of the southwestern wheat belt was relatively stable (Fig. 6).
The center of the continent showed an earlier-than-average
peak in 2002 and 2009.
Over interior Australia, peak timing varied by over a
month from one year to another. Areas for which no peak
was observed in a given year (shown in gray in Fig. 6) oc-
curred primarily in the drylands of the continent’s interior,
where phenological cycles may not follow an annually re-
curring pattern. For example, areas with no peak over interior
Australia in Fig. 6 for 2005 and 2008 can also be traced in
Fig. 2, where the phenology of the Alice Springs site did not
show a peak in those years.
3.3 Variability of cycle-integrated greenness
Greenness integrated between the start and end of a pheno-
logical cycle can provide a first approximation of vegetation
productivity (Ponce Campos et al., 2013; Zhang et al., 2013).
Standardized anomalies of integrated greenness highlight the
deviation of an individual value from the mean, relative to the
expected level of variability (the standard deviation). Stan-
dardized anomalies of integrated greenness were highly vari-
able across time (Fig. 7). Negative standardized anomalies of
integrated greenness (red tones in Fig. 7) occurred across the
continent in most areas in 2002, and in vast areas of the con-
tinent in 2008 and 2009. Large areas of negative anomalies
also occurred in 2001, 2003 and from 2004 to 2007. Large ar-
eas of positive standardized anomalies (green tones in Fig. 8),
with an increased greening of 1–2 standard deviations, oc-
curred in 2010, a year of particularly high rainfall.
When relating the cycles’ standardized anomalies of inte-
grated greenness to the phenology at the Alice Springs tower
site, the widespread negative standardized anomaly over in-
terior Australia in 2008 (Fig. 7) was not represented on the
site’s curve (Fig. 2), where no cycle started or ended in 2008
and 2009. Conversely, the positive standardized anomalies
of cycles that started in 2010 and 2011 over large areas of
eastern and interior Australia can also be seen on the Al-
ice Springs curve in the form of larger-than-average integrals
(Fig. 2).
3.4 Analysis of spatial–temporal patterns of Australian
phenology relative to rainfall and SOI variability
We conducted a correlation analysis relating two climate
drivers (SOI and rainfall) and two phenological metrics (first
peak magnitude and cycle integral of each year), respectively
(four combinations). Each of the four analyses included cli-
mate drivers binned over periods between 1 and 12 months
within the 12-month period leading up to the phenological
peak. We found that areas with significant correlations be-
tween SOI and phenology or rainfall and phenology were
most widespread for a binning interval of 1 month. Areas
with significant correlations shrank as we increased the bin-
ning interval of SOI or rainfall from 1 to 12 months. Biogeosciences, 11, 5181–5198, 2014
5190 M. Broich et al.: Land surface phenological response to climate variability across Australia
Fig. 6 2
Figure 6. Inter-annual variation in the peak timing. The Julian day of the phenological cycles’ peak is displayed in the calendar year when
the peak occurred. The mean (¯x) and standard deviation (σ) of the cycle peak timing are provided for reference. The scale is cyclic. Areas
where no peak was observed during a given calendar year are shown in gray.
The spatial pattern of significant correlations (areas sig-
nificantly correlated, correlation strength, and lead times)
was generally similar for all four combinations of vari-
ables. However, the patterns of significant correlation be-
tween peak magnitude and climate variables covered a larger
area compared to patterns of significant correlation between
cycle integral and climate variables. The patterns of signifi-
cant SOI-driven correlation with phenology covered a larger
and more concentrated area compared to the rainfall-driven
correlation patterns. Given the above similarities and the
largest extent of significant correlation patterns at a single-
month binning interval, we limit the presentation of results
to the most significant monthly SOI and – cycle peak mag-
nitude and the most significant monthly rainfall – cycle
peak magnitude correlation.
The most significant correlations between monthly SOI
and cycle peak magnitude and monthly rainfall and cy-
cle peak magnitude were most widespread in northeastern
Australia (Fig. 8c). Lead times between the most signifi-
cantly correlated driver month and the phenological cycle
peak were 1–6 months for northeastern Australia and 7–12
months for the eastern Australian interior, representing an
increase in lead time along a gradient of decreasing rainfall
(Fig. 8a and b). These correlation patterns extended into the
Australian interior along desert river drainage lines such as
the Cooper Creek. The floodplain of the middle reach of the
Cooper Creek can be clearly distinguished in the correlation
pattern, indicating a strong response of the floodplain veg-
etation to, for example, SOI variability (Fig. 9). Additional
correlation patterns with a shorter lag time behind SOI (1–3
months) were observed near the western coast of Australia,
with longer lag times of 5–8 months behind rainfall (Fig. 8a).
In the MDB, correlation patterns between monthly SOI
and cycle peak magnitude occurred primarily over natural
vegetation cover, as opposed to areas used for agriculture
or pasture (managed land cover). The percentage of all sig-
nificant relationships over natural land cover was 83.6%, as
opposed to 15.9%, the percentage of all significant relation-
ships over managed land cover (Table 2). These percentages
were disproportional to areal percentages of natural and man-
aged land cover within the MDB (71.8 and 28.2%, respec-
tively). The highest percentage of significantly correlated ar-
eas within each land cover class and the highest mean rho
values were found in areas dominated by shrubs, trees and
Biogeosciences, 11, 5181–5198, 2014
M. Broich et al.: Land surface phenological response to climate variability across Australia 5191
Fig. 7 2
Figure 7. Mean of the cycles’ integral greenness across the time series (top-left panel in day units) and standardized anomaly of each cycle’s
integrated greenness. The standardized anomalies of the cycles are shown in the year when the cycle started. For example, for a site with six
phenological cycles across the time series that started in 2001, 2002, 2003, 2005, 2008 and 2010, the cycles’ standard deviations are shown
in 2001, 2002, 2003, 2005, 2008 and 2010. All other years are shown in gray, as no phenological cycle start was detected for those years.
The white circle in the top-left panel marks the OzFlux site shown in Fig. 2.
Table 2. Percentage distribution of the most significant correlation relationships between monthly SOI and phenological peak magnitude
per land cover class across the MDB. Shown are percentages of the MDB occupied by different land covers, the percentage of basin-wide
significantly correlated areas per land cover, the percentage of significantly correlated land cover classes and the average rho value per land
Aggregated land % of % of the areas of significant % of each LCC where a significant Average rho of significant
cover classes (LCCs) basin covered correlations between monthly SOI and correlation between monthly SOI correlations within LCC
by each LCC peak magnitude within each LCC and peak magnitude occurred
Trees 43.0 48.7 5.2 0.71
Shrubs 9.8 12.2 5.7 0.74
Grasses 19.0 22.7 5.4 0.72
Rain-fed agriculture 28.1 15.9 2.6 0.69
and pasture
Irrigated agriculture 0.1 <0.0 0.9 0.69
and pasture
grasses. Irrigated agriculture and pasture had the smallest
percentage of correlated areas (Table 2) compared to other
land cover classes.
The ecologically valuable floodplains and wetlands of the
MDB made up 10.9% of the basin area and were of mixed
land cover composition. The percentage of all areas with sig-
nificant correlations between monthly SOI and phenological
cycle peak magnitude in floodplains and wetlands was dis-
proportionately higher (14.8%) than the percentage of the
area occupied by this zone (10.9%). In addition, 6.1 % of
the floodplain and wetlands area showed significant relation-
ships with monthly SOI, which is higher than for any of the
individual land cover classes in Table 2. Biogeosciences, 11, 5181–5198, 2014
5192 M. Broich et al.: Land surface phenological response to climate variability across Australia
4 Discussion
4.1 A phenological characterization of Australia that
accommodates non-annual phenological cycles
Our research characterized the cycles and variability of non-
annual vegetation phenology across Australia, and identi-
fied their relationships with variability in rainfall and ENSO-
related large-scale atmospheric circulation. We provide a
characterization of annual and non-annual phenological cy-
cles of vegetation greening and browning for Australia based
on MODIS EVI data.
We used an enhanced phenology model to characterize
rainfall-driven phenology across the Australian continent,
which includes large dryland regions. Very few studies have
previously quantified the land surface phenology of dryland
systems (Walker et al., 2014), likely due to the fact that
the phenology of these systems is more complex than that
of most temperature-limited regions (Walker et al., 2014;
Primack and Miller-Rushing, 2011). Dryland phenology re-
sponds to a variable rainfall regime where the timing and
magnitude of precipitation events varies inter-annually (Loik
et al., 2004; Brown et al., 1997).
We identified and characterized rainfall-driven phenolog-
ical cycles at any time of the year over a 14-year time se-
ries rather than within a predefined interval of every calen-
dar year. This is important, as the timing of phenological
cycles varied, and not every phenological cycle metric oc-
curred in every year. We first identified points demarcating
phenological cycles from the entire EVI time series, and then
characterized the cycles using mathematical curves. For ex-
ample, we did not identify a cycle peak for every year and
every pixel (areas shown in gray in Fig. 6). However, this
does not imply that no cycle occurred, but that the vegeta-
tion at these sites and points in time could be greening up to-
wards a peak in the following year, browning down towards
an end-of-cycle point, or in a phase between cycles. For ex-
ample, the absence of peaks over interior Australia in 2005
and 2008 (Fig. 6) is also reflected in Fig. 2, where the veg-
etation at the Alice Springs site in interior Australia was in
between phenological cycles. Phenological cycles thus need
to be analyzed in the temporal context of multiple years.
While most studies of phenology attempted to fit phenolog-
ical curves within a predefined interval every calendar year,
certain authors have proposed methods that include iterating
the curve fitted to the vegetation index time series or by fit-
ting a curve of vegetation index versus accumulated moisture
(Tan et al., 2011; Brown and de Beurs, 2008). Our approach
to characterize non-annual phenology can be applied to other
areas with rainfall-driven phenology, and thus contributes to
our understanding of non-annual, rainfall-driven phenolog-
ical dynamics globally. While the results presented in this
work focus on the phenological metrics of the first cycles
of each year, a second cycle was not detected over most of
Australia. For example, two peaks during a calendar year oc-
curred over only 25% of the Australian land surface. Within
the 14 years of the study, two peaks per year occurred no
more than three times across 96% of Australia. Areas with
two peaks per year occurred mostly in cropping or pasture
land use (Fig. 10). An alternative method for identifying the
number of cycles for broad regions can be found in Guan et
al. (2014b).
4.2 Phenology of Australia’s interior
For the interior of Australia, we identified a low phenologi-
cal peak and minimum magnitude and the associated small
amplitude (darker toned areas in both Fig. 4a and b), and
high variability in magnitude, timing and the cycle integral.
In addition, a peak was not identified in every year for large
areas of the interior. Most areas of the interior are dryland
systems with sparse vegetation cover and where vegetation
phenology is driven by highly irregular rainfall timing and
amounts (Australian Bureau of Meteorology, 2014c, e), and
hydrologic regimes can be difficult to predict (Young and
Kingsford, 2006). Thus, we do not see a strong phenologi-
cal response (low amplitude); however, we interpret the high
variability in the start of cycle and peak timing (Figs. 4 and 5)
as a fast response to rainfall pulses, and the missing cy-
cles (Fig. 5) were interpreted as dormant periods during dry
years (Loik et al., 2004). We interpret these patterns of vari-
able phenological cycles over interior Australia, where a cy-
cle may vary in timing and length, or may skip a year en-
tirely, to occur as a function of high climate variability. De
Jong et al. (2012) identified frequent trend breaks of green-
ing and browning over Australia that may be related to the
non-annual phenological cycles identified here.
Desert river beds in the interior of the continent had a low
minimum but moderate peak magnitude. The elevated peak
magnitudes are caused by flooding driven by high amounts
of distant rainfall (Young and Kingsford, 2006). The center
of the arid Lake Eyre basin showed high variability in the
minimum magnitude. Lake Eyre is the center of a sparsely
vegetated, close drainage basin, and the fact that we identi-
fied high variability was in line with known flooding patterns,
as this salt lake is reached by flooding only once in a century
(McMahon et al., 2005). We interpret the positive anomaly in
2010 (Fig. 7) as a function of the La Niña floods (Australian
Bureau of Meteorology, 2014a).
Conversely, large variability of peak timing and cycle-
integrated greenness from one to another phenological cycle
was found not just in the interior of Australia, but across most
of the continent (Fig. 6 and Fig. 7). High inter-annual vari-
ability in water availability across most of Australia rather
than for the continent’s interior has also been demonstrated
by the Australian Water Availability Project (2014).
Biogeosciences, 11, 5181–5198, 2014
M. Broich et al.: Land surface phenological response to climate variability across Australia 5193
Figure 8. Statistically significant relationships between monthly SOI and phenological cycle peak magnitude (top row) and monthly rainfall
and phenological cycle peak magnitude (bottom row). (a) SOI and rainfall month most significantly correlated with peak magnitude. (b)
Lead time of SOI and rainfall month relative to phenological peak and (c) Spearman’s rho. Areas with p > 0.05 shown in white. The black
box in the top-right panel marks the extent of the area shown in Fig. 9, centered on the Cooper Creek floodplain in interior eastern Australia.
4.3 Australia’s phenology, the 2001–2009 Millennium
Drought and a La Niña high precipitation event
in 2010
The years with widespread negative standard anomalies of
cycle-integrated greenness coincided with the Millennium
Drought from 2001 to 2009 (Heberger, 2011; Fig. 7). Dry-
land vegetation is subject to environmentally marginal con-
ditions, and is therefore highly sensitive to climate variability
(Hufkens et al., 2012; Brown et al., 1997).
However, the spatial extent of negative anomalies in cer-
tain years that extend beyond the dry interior suggested tem-
porary yet severe drought-related water limitations also in
the monsoonal north and the temperate areas of southeast-
ern and southwestern Australia (Fig. 7). The large positive
standardized anomalies of cycle-integrated greenness iden-
tified in this work across most of eastern Australia in 2010
(1–2 standard anomalies; Fig. 7) coincided with a strong
La Niña event and associated high rainfall and floods that
broke the Millennium Drought (Australian Bureau of Mete-
orology, 2014a; Heberger, 2011). This pattern includes the
desert rivers extending from northeastern Australia to Lake
Eyre, which experienced a major flood in 2010.
While the relationship between ENSO cycles and rainfall
variability primarily over eastern Australia has been investi-
gated before (van Dijk et al., 2013; Risbey et al., 2009), our
research has quantified vegetation responses across Australia
to the transition from a strong El Niño drought to La Niña wet
periods. While the positive vegetation response to the 2010
La Niña occurred over eastern Australia, which is also influ-
enced by ENSO cycles (van Dijk et al., 2013; Nicholls, 1991;
Nicholls et al., 1997), the negative vegetation responses dur-
ing the Millennium Drought cover a larger area, and occurred
across the continent. Biogeosciences, 11, 5181–5198, 2014
5194 M. Broich et al.: Land surface phenological response to climate variability across Australia
Figure 9. Significant Spearman rho correlations (shown in green)
between monthly SOI and phenological cycle peak magnitude over
a region in central Australia. The Cooper Creek floodplain of the
middle reach of the Cooper Creek is visible in the center. Only areas
with p < 0.05 and rho >=0.6 are shown.
4.4 Spatially explicit relationship between phenology
and climatic variability
We found that SOI-driven patterns of correlation with phe-
nology covered a larger area compared to rainfall-driven pat-
terns, likely because SOI is a more generic proxy of climatic
variability that influences temperature, incoming solar radi-
ation and rainfall rather than rainfall alone (Risbey et al.,
2009; Australian Bureau of Meteorology, 2014f), and be-
cause ecosystems of Australia are limited not only by water
availability, but also by temperature and radiation (Nemani
et al., 2003).
The spatial extent of areas where we detected a correlation
between SOI or rainfall and phenological metrics shrank with
longer binning intervals of the climatic drivers. This sug-
gested that relationships between climatic drivers and pheno-
logical variability were strongest for driver variability within
a specific month of the year (e.g., SOI in September), as op-
posed to driver variability within for example a 6-month pe-
riod (e.g., mean SOI across 6 months starting in April). This
falls in line with the findings by Stone et al. (1996), who
identified relationships between short-term SOI dynamics at
specific times of the year and rainfall. Previous studies (e.g.,
Brown et al., 2010) using seasonal or longer temporal aggre-
gation of driver variables may therefore not have identified
the full spatial extent of correlation patterns.
We found the most concentrated significant correlation
patterns between SOI and peak magnitude in northeastern
Australia, which is in the proximity of the western Pacific
convection variability indicated by SOI. We observed a sim-
ilar yet less concentrated pattern for the rainfall–peak mag-
nitude correlation. We interpret this latter pattern as primar-
ily the effect of the large-scale atmospheric circulation pat-
terns indicated by SOI. The lag times of correlations over
northeastern Australia varied between 1 and 6 months fol-
Figure 10. Number of years within the 14-year time series where
two peaks were detected, mostly associated with cropping or pas-
ture land (Fig. 1).
lowing SOI or rainfall. Shorter lag time (1–3 months) cor-
relation patterns with SOI were observed near the west-
ern coast of Australia, yet lag times following rainfall were
longer (5–8 months). These patterns are spatially remote
from the variability in convection over the western Pacific
(northeast of Australia) indicated by SOI. They may be re-
lated to the influence of the Indian Ocean Dipole (IOD) and
the interaction between SOI and IOD (Risbey et al., 2009),
which may explain the difference in lead time of the SOI
and rainfall drivers. Over northeastern Australia and the east-
ern Australian interior, the identified 3–6 and 7–12 months’
lag times of phenological cycle peak magnitude were simi-
lar for the SOI and the rainfall driver. The lag times identi-
fied here fell within the range of aggregation found by An-
dela et al. (2013), who related NDVI to rainfall. A study by
Chen et al. (2014b) identified short lags (predominantly 1
month) between soil moisture and NDVI, which are shorter
than most of the lags we identified here. Soil moisture in
the previous month may provide the most direct relationship
with vegetation response (as it represents water available to
vegetation), but the climatic conditions that drive soil mois-
ture may precede the soil moisture by a few months (Philip-
pon et al., 2014). The identified increase in lag time be-
tween SOI and phenological peak magnitude and rainfall and
phenological peak magnitude along a gradient of decreas-
ing rainfall was in agreement with the findings by Andela
et al. (2013). However, these findings contradict the concept
that rainfall pulses drive rapid phenological responses (Loik
et al., 2004). We interpret our findings as the dominating
space–time relationship between large-scale atmospheric cir-
culation pattern variability and phenological response. How-
ever, these patterns are unlikely to represent responses to in-
dividual storm events, and less significant relationships with
Biogeosciences, 11, 5181–5198, 2014
M. Broich et al.: Land surface phenological response to climate variability across Australia 5195
a different SOI and rainfall month and lag time were also
present, suggesting that vegetation responds to climatic vari-
ability on multiple timescales. A more in-depth analysis of
the relationship between climatic drivers and phenological
responses across multiple temporal scales should be investi-
gated in future research.
The proportion of areas for which we identified signifi-
cant correlations was generally smaller than those identified
in other studies (e.g., Andela et al., 2013 and Chen et al.,
2014a). This could be related to the relatively short time se-
ries we used and consequently the smaller power of our cor-
relation analysis. Nonetheless, the spatial pattern of correla-
tion was most widespread in northeastern Australia and along
desert river beds (e.g., Cooper Creek) in the interior. These
patterns agreed spatially with what would be expected from
the SOI-approximated moisture source over the western Pa-
cific and the associated progression of rainfall and runoff into
interior Australia.
We conducted a higher spatial resolution correlation anal-
ysis for the MDB to investigate the sensitivity of the area’s
vegetation to SOI variability. The MDB contains the primary
agricultural area of Australia, and the basin’s agriculture was
severely impacted by the Millennium Drought (van Dijk et
al., 2013; Kirby et al., 2012; Heberger, 2011). We identi-
fied correlation patterns between SOI and peak magnitude
primarily over natural vegetation cover, as opposed to ar-
eas used for dryland agriculture or pasture. As expected, ir-
rigated agriculture had the lowest percentage of area, with
significant correlations between SOI and phenological peak
magnitude. The lowest percentage of areas with significant
correlations over managed land may be explained by the ef-
fort that land managers and irrigators make to archive max-
imum production regardless of climatic variability (e.g., fer-
tilization, use of pesticides, crop rotation, livestock density,
movement and irrigation), whereas landscapes with natural
vegetation cover may respond directly to climatic variabil-
ity. In the context of climatic influence on agriculture in the
MDB, van Dijk et al. (2013) suggested that the Millennium
Drought impact on dryland wheat yields was offset by steady
increases in cropped area and plant water use efficiency as
well as possibly CO2fertilization. As a zone of special in-
terest within the MDB, we focused on floodplains and wet-
lands. These ecosystems were strongly impacted by the Mil-
lennium Drought and the 2010 La Niña floods (Australian
Bureau of Meteorology, 2014b; Leblanc et al., 2012). Across
the MDB’s floodplains and wetlands, we identified the high-
est percentage of areas (6.1%) with a significant correlation
between SOI and phenological peak magnitude compared to
other natural or managed land cover, highlighting the sensi-
tivity of these ecosystems to ENSO-related climatic variabil-
ity. We attributed the low percentage tolimited test power as
a function of the relatively short time series (14 years) used
here. For example, Brown et al. (2010) found between 10
and 27% of certain areas in Africa to be significantly corre-
lated with atmospheric indices using a 25-year AVHRR time
4.5 Limitations and future work
Several caveats toour work should be noted. When interpret-
ing the phenological cycles characterized here, it should be
noted that the sub-pixel composition of vegetation and back-
ground as well as multi-layer vegetation structure is unknown
and may change over time (Zhang et al., 2009; Walker et al.,
2012, 2014). Various methods for validating remotely sensed
metrics of phenological cycles with ground-based observa-
tions have been discussed, including flux tower productiv-
ity time series, ground-based radiation sensor time series,
phenocam time series as well as crowd-sourced citizen sci-
ence (Richardson et al., 2007; Liang and Schwartz, 2009;
Restrepo-Coupe et al., 2013). Validation of the phenological
metrics developed here is currently underway.
The phenological metrics derived and described here rep-
resent different stages of vegetation growth. They have been
made freely available in contribution to the Australian Ter-
restrial Ecosystem Research Network (TERN), and can be
downloaded from the AusCover TERN Sydney node1(http:
//, providing opportunities for a range of
In this work, we traced phenological cycles over time,
quantified cycles’ inter-annual variability, and investigated
their relationship with rainfall and ENSO, thereby advancing
phenological research for Australia, a country with extensive
drylands. The phenological metrics provided here can be fur-
ther used for characterizing the effect of anthropogenic dis-
turbances on phenology and unraveling this effect from the
influence of climatic forcing related to ENSO. Other oppor-
tunities for future work are the reanalysis of trends and trend
breaks in vegetation phenological magnitude dynamics and
climatic drivers (Donohue et al., 2009; de Jong et al., 2012;
Chen et al., 2014a) and the relationship between vegetation
phenological timing and climate control (Guan et al., 2014a).
5 Conclusions
We characterized vegetation phenological cycles that we de-
rived from time series of Earth-observing satellite images
from 2000 to 2013, across Australia, the driest inhabited
continent. The precipitation-driven, non-annual phenology of
Australia’s drylands has not been previously studied in detail,
and the relationship between phenology and climatic drivers
including rainfall and SOI has not been previously quantified.
1The Australian Phenology Product is scheduled to migrate per-
manently to the Australian Research Data Storage Infrastructure
(RDSI) that is funded through the Australian Government’s Super
Science Initiative and sourced from the Education Investment Fund
(EIF). Biogeosciences, 11, 5181–5198, 2014
5196 M. Broich et al.: Land surface phenological response to climate variability across Australia
We found the phenology of Australia’s drylands to be
highly variable across the time series, with shifts in pheno-
logical cycle peak timing of more than 1 month in the in-
terior of the continent. Cycle-integrated greenness, a surro-
gate of vegetation productivity, shifted from negative to pos-
itive anomalies over most of eastern Australia with the tran-
sition from the El Niño-induced decadal droughts to flood-
ing caused by La Niña. We related phenological magnitude
response variability to the variability in rainfall and SOI
across the continent and at a higher spatial resolution for the
MDB, the main agricultural basin of Australia. We found the
most widespread correlation patterns with single-month as
opposed to multi-month aggregated drivers, suggesting that
rainfall and SOI at a specific point in time is of primary im-
portance in driving phenology. Correlation patterns between
phenological magnitude response with rainfall and SOI oc-
curred primarily over northeastern Australia and within the
MDB, predominantly over natural land cover and particularly
in floodplain and wetland areas, highlighting the sensitivity
of these ecosystems to ENSO-related climatic variability.
A more in-depth analysis of the relationship between cli-
matic drivers and phenological magnitude response across
multiple temporal scales and including temperature and radi-
ation drivers and driver combinations should be investigated
in future research. Furthermore, the analysis of the relation-
ship between phenological timing and climatic drivers should
also be investigated.
Our approach could be valuable for other areas of rainfall-
driven systems, and thus contributes to our understanding of
non-annual phenological dynamics globally. The quantified
spatial–temporal variability in phenology across Australia in
response to climate variability presented here advances re-
search of dryland phenology and provides important infor-
mation for land management and climate change studies.
The phenological metrics derived represent different stages
of vegetation growth. They have been made freely available
in contribution to the Australian Terrestrial Ecosystem Re-
search Network (TERN), and can be downloaded from the
AusCover TERN Sydney node, providing opportunities for a
range of applications. The phenological metrics can be fur-
ther used for characterizing the effect of anthropogenic dis-
turbances on phenology and unraveling this effect from the
influence of climatic forcing components and large-scale at-
mospheric circulation indices.
Acknowledgements. This research was supported by an Australian
Research Council Discovery grant (DP1115479) entitled “Integrat-
ing remote sensing, landscape flux measurements, and phenology to
understand the impacts of climate change on Australian landscapes”
(A. Huete, CI), and by funding from the AusCover facility of the
Australian Terrestrial Ecosystem Research Network (TERN). Cal-
culations were performed at the University of Technology, Sydney
eResearch high-performance computing facility.
M. G. Tulbure was partially funded through an Australian
Research Council Discovery Early Career Researcher Award
Edited by: K. Guan
Andela, N., Liu, Y. Y., van Dijk, a. I. J. M., de Jeu, R. a. M., and
McVicar, T. R.: Global changes in dryland vegetation dynamics
(1988–2008) assessed by satellite remote sensing: comparing a
new passive microwave vegetation density record with reflective
greenness data, Biogeosciences, 10, 6657–6676, doi:10.5194/bg-
10-6657-2013, 2013.
Australian Bureau of Meteorology: The 2010–2011 La Niña: Aus-
tralia soaked by one of the strongest events on record, available
shtml, 2014a.
Australian Bureau of Meteorology: El Niño - Detailed Australian
Analysis and La Niña – Detailed Australian Analysis, avail-
able at:; http://www., 2014b.
Australian Bureau of Meteorology: Climate Data Online: Average
annual, seasonal and monthly rainfall (mm) and Rainfall variabil-
ity (index of variability), available at:
data/index.shtml, 2014c.
Australian Bureau of Meteorology: Southern Oscillation Index
Data, available at:
shtml, 2014d.
Australian Bureau of Meteorology: Rainfall variability, avail-
able at:
rainfall-variability/index.jsp, 2014e.
Australian Bureau of Meteorology: ENSO impacts – tempera-
ture, available at:
ln-2010-12/ENSO-temperature.shtml, 2014f.
Bradley, B. A. and Mustard, J. F.: Comparison of phenology trends
by land cover class: a case study in the Great Basin, USA, Glob.
Change Biol., 14, 334–346, 2007.
Brown, J. H., Valone, T. J., and Curtin, C. G.: Reorganization of an
arid ecosystem in response to recent climate?change, Proc. Natl.
Ac. Sci., 94, 9729–9733, 1997.
Brown, M. E. and de Beurs, K. M.: Evaluation of multi-sensor semi-
arid crop season parameters based on NDVI and rainfall, Remote
Sens. Environ., 112, 2261–2271, 2008.
Brown, M. E., de Beurs, K., and Vrieling, A.: The response of
African land surface phenology to large scale climate oscilla-
tions, Remote Sens. Environ., 114, 2286–2296, 2010.
Chen, B., Xu, G., Coops, N. C., Ciais, P., Innes, J. L., Wang, G., My-
neni, R. B., Wang, T., Krzyzanowski, J., Li, Q., Cao, L., and Liu,
Y.: Changes in vegetation photosynthetic activity trends across
the Asia–Pacific region over the last three decades, Remote Sens.
Environ., 144, 28-=41, 2014a.
Chen, T., de Jeu, R. a. M., Liu, Y. Y., van der Werf, G. R., and
Dolman, a. J.: Using satellite based soil moisture to quantify the
water driven variability in NDVI: A case study over mainland
Australia, Remote Sens. Environ., 140, 330–338, 2014b.
Connell, D.: Water politics in the Murray-Darling basin, Federation
Press, 2007.
Biogeosciences, 11, 5181–5198, 2014
M. Broich et al.: Land surface phenological response to climate variability across Australia 5197
de Beurs, K. M., and Henebry, G. M.: Spatio-temporal statistical
methods for modeling land surface phenology, in: Phenological
Research, edited by: Hudson, I. L. and Keatley, M. R., Springer,
Dordrecht, 2008.
de Jong, R., Verbesselt, J., Schaepman, M. E., and Bruin, S.: Trend
changes in global greening and browning: contribution of short-
term trends to longer-term change, Glob. Change Biol., 18, 642–
655, 2012.
Delbart, N., Kergoat, L., Le Toan, T., Lhermitte, J., and Picard, G.:
Determination of phenological dates in boreal regions using nor-
malized difference water index, Remote Sens. Environ., 97, 26–
38, 2005.
Donohue, R. J., McVicar, T. R., and Roderick, M. L.: Climate-
related trends in Australian vegetation cover as inferred from
satellite observations, 1981–2006, Glob. Change Biol., 15, 1025–
1039, 2009.
Dougherty, R. L., Edelman, A., and Hyman, J. M.: Nonnegativity-,
Monotonicity-, or Convexity-Preserving Cubic and Quintic Her-
mite Interpolation, Mathem. Computat., 52, 471–794, 1989.
Eklundh, L. and Jönsson, P.: TIMESAT 3.0 Software Man-
ual, 1–74. (
SoftwareManual.pdf), 2010.
Friedl, M., Henebry, G., Reed, B., Huete, A., White, M., Morisette,
J., Nemani, R., Zhang, X., and Myneni, R.: Land surface
phenology, A Community White Paper requested by NASA,
10 April 2006.
Ganguly, S., Friedl, M. A., Tan, B., Zhang, X., and Verma, M.:
Remote Sensing of Environment Land surface phenology from
MODIS : Characterization of the Collection 5 global land cover
dynamics product, Remote Sens. Environ., 114, 1805–1816,
Goddard Space Flight Center: Tropical Rainfall Monitoring Mis-
sion Project TRMM_3B43.v7 product, USGS/Earth Resources
Observation and Science (EROS) Center, Sioux Falls, South
Dakota,, 2014.
Guan, K., Medvigy, D., Wood, E. F., Caylor, K. K., Li, S., and
Jeong S. J.: Deriving vegetation phenological time and trajectory
information over Africa using SEVIRI daily LAI, IEEE Ttrans.
Geosci. Remote Sens., 53, 1113–1130, 2014a.
Guan, K., Wood, E. F., Caylor, K. K., Medvigy, D., Sheffield, J.,
Pan, M., Kimball, J., Xu, X., and Jones, M. O.: Terrestrial hydro-
logical control on vegetation phenology of African savannas and
woodlands, J. Geophys. Res. Biogeosciences, accepted, 2014b.
Heberger, M.: Australia’s Millennium Drought: Impacts and Re-
sponses, in: The World’s Water, edited by: Gleick, P. H., Island
Press/Center for Resource Economics, 97–125, 2011.
Huete, A., Didan, K., Miura, T., Rodriguez, E. P., Gao, X., and Fer-
reira, L. G.: Overview of the radiometric and biophysical perfor-
mance of the MODIS vegetation indices, Remote Sens. Environ.,
83, 195–213, 2002.
Huete, A., Miura, T., Yoshioka, H., Ratana, P., and Broich, M.: In-
dices of Vegetation Activity, in, edited by: Hanes, J., Springer,
Berlin Heidelberg, 2014.
Hufkens, K., Friedl, M., Sonnentag, O., Braswell, B. H., Milliman,
T., and Richardson, A. D.: Linking near-surface and satellite re-
mote sensing measurements of deciduous broadleaf forest phe-
nology, Remote Sens. Environ., 117, 307–321, 2012.
IPCC: Climate Change 2001: impacts, adaptation, and vulnerabil-
ity, Contribution of working group II to the third assessment re-
port of the Intergovernmental Panel on Climate Change (IPCC),
Cambridge University Press, Cambridge, 2001.
IPCC: Climate change 2007 – impacts, adaptation and vulnerability,
Contribution of Working Group II to the Fourth Assessment Re-
port of the IPCC, Cambridge University Press, Cambridge, 2007.
IPCC: Climate Change 2013: The Physical Science Basis, Contri-
bution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change, Cambridge Univer-
sity Press, Cambridge, 2013.
Jones, M. O., Jones, L. a., Kimball, J. S., and McDonald, K.
C.: Satellite passive microwave remote sensing for monitoring
global land surface phenology, Remote Sens. Environ., 115,
1102–1114, 2011.
Keatley, M. R., Chambers, L. E., and Phillips, R.: Australia and New
Zealand, in: Phenology: an Integrative Environmental Science,
edited by: Schwartz, M. D., Springer, Dordrecht, 23–52, 2013.
Kingsford, R. T., Brandis, K., Thomas, R. F., Crighton, P., Knowles,
E., and Gale, E.: Classifying landform at broad spatial scales: the
distribution and conservation of wetlands in New South Wales,
Australia, Mar. Fresh. Res., 55, 17–31, 2004.
Kirby, M., Connor, J., Bark, R., Qureshi, E., and Keyworth, S.:
The economic impact of water reductions during the Millennium
Drought in the Murray-Darling Basin, AARES conference, 2012,
7–10, 2012.
Köppen, W.: The thermal zones of the Earth according to the dura-
tion of hot, moderate and cold periods and of the impact of heat
on the organic world. (translated and edited by Volken, E. and S.
Brönnimann), Meteorologische Zeitschrift, 1, 351–360, 1884.
Leblanc, M., Tweed, S., Van Dijk, A., and Timbal, B.: A review of
historic and future hydrological changes in the Murray-Darling
Basin, Glob. Planet. Change, 80/81, 226–246, 2012.
Lehmann, E. L. and D’Abrera, H. J. M.: Nonparametrics: statistical
methods based on ranks, Holden-Day, 1975.
Liang, L. and Schwartz, M.: Landscape phenology: an integra-
tive approach to seasonal vegetation dynamics, Lands. Ecol., 24,
465–472, 2009.
Loik, M., Breshears, D., Lauenroth, W., and Belnap, J.: A multi-
scale perspective of water pulses in dryland ecosystems: clima-
tology and ecohydrology of the western USA, Oecologia, 141,
269–281, 2004.
Lymburner, L., Tan, P., Mueller, N., Thackway, R., Lewis, A.,
Thankappan, M., Randall, L., Islam, A., and Senarath, U.: The
National Dynamic Land Cover Dataset, Geoscience Australia,
Symonston, Australia, p. 105, 2011.
Ma, X., Huete, A., Yu, Q., Coupe, N. R., Davies, K., Broich, M.,
Ratana, P., Beringer, J., Hutley, L. B., Cleverly, J., Boulain, N.,
and Eamus, D.: Spatial patterns and temporal dynamics in sa-
vanna vegetation phenology across the North Australian Tropical
Transect, Remote Sens. Environ., 139, 97–115, 2013.
McMahon, T. A. T. A., Murphy, R., Little, P., Costelloe, J. F., Peel,
M. C. M. C., Chiew, F. H. S., Hayes, S., Nathan, R. J. R. J.,
Kandel, D. D., (Firm), S. K. M., Engineering, U. o. M. D. o. C. a.
E., and Heritage, A. D. o. t. E. a.: Hydrology of Lake Eyre Basin,
edited by: McMahon, T. A., Murphy, R., Little, P., Costelloe, J.
F., Peel, M. C., Chiew, F. H. S., Hayes, S., Nathan, R., Kandel,
D. D., Canberra, Australian Capital Territory, Natural Heritage
Trust, 2005. Biogeosciences, 11, 5181–5198, 2014
5198 M. Broich et al.: Land surface phenological response to climate variability across Australia
Moulin, S., Kergoat, L., Viovy, N., and Dedieu, G.: Global-Scale
Assessment of Vegetation Phenology Using NOAA/AVHRR
Satellite Measurements, J. Climate, 10, 1154–1170, 1997.
Myneni, R. B., Keeling, C. D., Tucker, C. J., Asrar, G., and Nemani,
R. R.: Increased plant growth in the northern high latitudes from
1981 to 1991, Nature, 386, 698–702, 1997.
Nemani, R. R., Keeling, C. D., Hashimoto, H., Jolly, W. M., Piper,
S. C., Tucker, C. J., Myneni, R. B., and Running, S. W.: Climate-
Driven Increases in Global Terrestrial Net Primary Production
from 1982 to 1999, Science, 300, 1560–1563, 2003.
Nicholls, N.: The El Niño/Southern Oscillation and Australian veg-
etation, Vegetatio, 91, 23–36, 1991.
Nicholls, N., Drosdowsky, W., and Lavery, B.: Australian rainfall
variability and change, Weather, 52, 66–72, 1997.
OzFlux: Australian and New Zealand Flux Research and Monitor-
ing,, 2014.
Peñuelas, J., Rutishauser, T., and Filella, I.: Phenology Feedbacks
on Climate Change, Science, 324, 887–888, 2009.
Philippon, N., Martiny, N., Camberlin, P., Hoffman, M. T., and
Gond, V.: Timing and patterns of ENSO signal in Africa over
the last 30 years: insights from Normalized Difference Vegeta-
tion Index data, J. Climate, 27, 2509–2532, 2014.
Pitman, A. J.: The evolution of and revolution in, land surface
schemes designed for climate models, Int. J. Climatol., 23, 479–
510, 2003.
Ponce Campos, G. E., Moran, M. S., Huete, A., Zhang, Y., Bresloff,
C., Huxman, T. E., Eamus, D., Bosch, D. D., Buda, A. R., Gunter,
S. A., Scalley, T. H., Kitchen, S. G., McClaran, M. P., McNab, W.
H., Montoya, D. S., Morgan, J. A., Peters, D. P. C., Sadler, E. J.,
Seyfried, M. S., and Starks, P. J.: Ecosystem resilience despite
large-scale altered hydroclimatic conditions, Nature, 494, 349–
352, 2013.
Primack, R. B. and Miller-Rushing, A. J.: Broadening the study of
phenology and climate change, New Phytologist, 191, 307–309,
Restrepo-Coupe, N., Huete, A., Broich, M., and Davies, K.: Phe-
nology Validation, in: Terrestrial Ecosystem Research Network,
Richardson, A., Jenkins, J., Braswell, B., Hollinger, D., Ollinger, S.,
and Smith, M.-L.: Use of digital webcam images to track spring
green-up in a deciduous broadleaf forest, Oecologia, 152, 323–
334, 2007.
Richardson, A. D., Keenan, T. F., Migliavacca, M., Ryu, Y., Sonnen-
tag, O., and Toomey, M.: Climate change, phenology, and phe-
nological control of vegetation feedbacks to the climate system,
Agr. Forest Meteorol., 169, 156–173, 2013.
Risbey, J. S., Pook, M. J., McIntosh, P. C., Wheeler, M. C., and
Hendon, H. H.: On the Remote Drivers of Rainfall Variability in
Australia, Month. Weather Rev., 137, 3233–3253, 2009.
Savitzky, A. and Golay, M. J. E.: Smoothing and Differentiation of
Data by Simplified Least Squares Procedures, Analyt. Chem., 36,
1627–1639, 1964.
Schwartz, M.: Introduction, in, edited by: Schwartz, M., Tasks for
Vegetation Science, Springer, Netherlands, 3–7, 2003.
Schwartz, M. D.: Preface, in: Phenology: An Integrative Environ-
mental Science, edited by: Schwartz, M. D., Springer, Dordrecht,
Stone, R. C., Hammer, G. L., and Marcussen, T.: Prediction of
global rainfall probabilities using phases of the Southern Oscil-
lation Index, Nature, 384, 252–255, 1996.
Tan, B., Morisette, J. T., Wolfe, R. E., Gao, F., Ederer, G. A.,
Nightingale, J., and Pedelty, J. A.: An Enhanced TIMESAT
Algorithm for Estimating Vegetation Phenology Metrics From
MODIS Data, IEEE J. Select. Top. Appl. Earth Observat. Re-
mote Sens., 4, 361–371, 2011.
Trenberth, K. E. and Caron, J. M.: The Southern Oscillation Revis-
ited: Sea Level Pressures, Surface Temperatures, and Precipita-
tion, J. Climate, 13, 4358–4365, 2000.
United Nations: Global Drylands: A UN system-wide response,
Geneva, Switzerland, 2011.
van Dijk, A. I. J. M., Beck, H. E., Crosbie, R. S., de Jeu, R. a. M.,
Liu, Y. Y., Podger, G. M., Timbal, B., and Viney, N. R.: The Mil-
lennium Drought in southeast Australia (2001–2009): Natural
and human causes and implications for water resources, ecosys-
tems, economy, and society, Water Resour. Res., 49, 1040–1057,
Walker, J. J., de Beurs, K. M., Wynne, R. H., and Gao, F.: Evalua-
tion of Landsat and MODIS data fusion products for analysis of
dryland forest phenology, Remote Sens. Environ., 117, 381–393,
Walker, J. J., de Beurs, K. M., and Wynne, R. H.: Dryland vege-
tation phenology across an elevation gradient in Arizona, USA,
investigated with fused MODIS and Landsat data, Remote Sens.
Environ., 144, 85–97, 2014.
White, M. A., Thornton, P. E., and Running, S. W.: A continental
phenology model for monitoring vegetation responses to interan-
nual climatic variability, Global Biogeochem. Cy., 11, 217–234,
Young, W. J. and Kingsford, R. T.: Flow variability in large unreg-
ulated dryland rivers, in: Ecology of Desert Rivers, edited by: R.
T., Kingsford, Cambridge University Press, Cambridge 2006.
Zhang, X., Friedl, M. A., and Schaaf, C. B.: Monitoring vegetation
phenology using MODIS, Remote Sensing of Environment, 84,
471–475, 2003.
Zhang, X., Friedl, M. A., and Schaaf, C. B.: Sensitivity of vegetation
phenology detection to the temporal resolution of satellite data,
Int. J. Remote Sens., 30, 2061–2074, 2009.
Zhang, X. Y., Friedl, M. A., and Tan, B.: Long-term detection of
global vegetation phenology from satellite instruments, in: In-
Tech, edited by: Zhang, X., InTech, 2012.
Zhang, Y., Susan Moran, M., Nearing, M. A., Ponce Campos, G. E.,
Huete, A. R., Buda, A. R., Bosch, D. D., Gunter, S. A., Kitchen,
S. G., Henry McNab, W., Morgan, J. A., McClaran, M. P., Mon-
toya, D. S., Peters, D. P. C., and Starks, P. J.: Extreme precipita-
tion patterns and reductions of terrestrial ecosystem production
across biomes, J. Geophys. Res. Biogeosciences, 118, 148–157,
Biogeosciences, 11, 5181–5198, 2014
... Additionally, as C 4 grasses tend to produce a larger amount of standing litter (Kollmann et al., 2009), increases in C 4 grass cover would intensify the severity of wildfires through greater fuel loads (Kollmann et al., 2009), which pose a real threat to ecosystems with highly flammable vegetation (Watson et al., 1996). In contrast to reported impacts of climate warming on Northern Hemisphere ecosystems (Parmesan, 2006;Piao et al., 2008), much less is known about Southern Hemisphere grasslands, where moisture limitations and climate extremes are expected to play a more important role (Broich et al., 2014), e.g., little is known about the changes in grassland functional type composition in Australia where climate has warmed by just over 1 • C from 1910 to 2018 and a higher proportion of annual rainfall has come from heavy rain events in recent decades (NSW Office of Environment and Heritage, 2014;CSIRO and Bureau of Meteorology, 2015). The warming climate could offset C 3 photosynthetic advantages offered by elevated atmospheric CO 2 concentrations (Morgan et al., 2011), however, it remains unclear how grassland ecosystems respond to the changing rainfall patterns. ...
... Numerous studies have characterized vegetation phenology (i.e. the timing of recurring biological cycles, such as emergence of first leaf, senescence, growing season length) by exploiting time-series of remotely-sensed spectral vegetation indices (Zhang et al., 2006;Ganguly et al., 2010;Henebry and de Beurs, 2013;Broich et al., 2015;Wu et al., 2016;Richardson et al., 2018). Satellite data have also offered the opportunity to capture years with extreme high and low productivity, including dormant growing seasons for plants in environments with extreme climate variability like Australia (Broich et al., 2014). Using a time series phenology approach, it is possible to identify and map the distribution of C 3 and C 4 grasses and assess seasonality and productivity of grasslands (Foody and Dash, 2007;Adjorlolo et al., 2012;Crabbe et al., 2019). ...
... Shifts in grassland compositions can also result in lower quality forage for livestock/wild animals (Chamaillé-Jammes and Bond, 2010), changes in habitat quality (Sivicek and Taft, 2011), shifts in biogeochemical cycling (Milesi et al., 2005), etc. Other factors such as CO 2 (Reyes-Fox et al., 2014), water availability (Mahdavi and Bergmeier, 2018), soil properties (Nord et al., 2015), and terrain effects are also important constraints on the composition of grasslands at local scales (Broich et al., 2014). ...
Species composition is a key determinant of grassland ecosystem function and resilience. Climate change is predicted to alter the distribution of cool season (C3) and warm season (C4) grasses, however, the lack of spatial distributions and temporal variations of grass functional type information severely limits our understanding of climate impacts on grasslands. This study classified C3 and C4 grasses per pixel according to the peak of growing season generated from Enhanced Vegetation Index time series. From 2003 to 2017, the C3-C4 composition of Australian rain-fed grasslands and pastures was mapped at 500 m resolution on an annual basis across a wide geographical range (10°S – 45°S), and revealed extreme inter-annual fluctuations. Over the 15-year period, the satellite-derived ratio of C4 to C3 grasses significantly increased (p < 0.05), indicating a long-term shift in community composition that was confirmed with 182,911 Atlas of Living Australia ground observations. The most pronounced changes occurred in mid-latitude transitional areas where C3 and C4 grasses co-dominate. Our climate analysis indicated the inter-annual fluctuations of C4/C3 grass ratios were significantly associated (p < 0.05) with warm/cool season rainfall ratios, and not with temperature or annual rainfall. This suggests that an increase in the warm/cool season rainfall ratio favors C4 grasses and a decrease in the warm/cool season rainfall ratio favors C3 grasses. Our findings reveal spatially-detailed dynamics of grasslands and demonstrate large-scale grassland compositional changes over 15 years. The grass composition maps should help improve ecological forecasting of grass distributions and enable researches on grassland ecosystem responses to climate change that are relevant to both adaptation of rangeland agricultural and fire management practices. Our study should also help predict grass distribution under future climate conditions, and assist in the accurate modelling of global water, carbon, and energy exchanges between the land surface and atmosphere.
... The Enhanced Vegetation Index (EVI) has been shown to be well correlated with LAI, biomass, canopy cover, and the fraction of absorbed photosynthetically active radiation [26,27], and is therefore useful for monitoring seasonal, inter-annual, and long-term variation of the vegetation structure and function [28]. EVI has been used instead of the Normalized Difference Vegetation Index (NDVI) because it reduces sensitivity to soil and atmospheric effects and remains sensitive to variation in canopy density where NDVI becomes saturated [29][30][31]. Given these characteristics, modelers have begun using EVI data to predict net primary production in ecosystem modelling applications [24,29]. ...
... The transformation of summer leaves and the appearance of winter leaves during autumn result in an increase in LAI, which remains almost stable during winter, until next spring. Even though this phenological/physiological cycle is repeated every year, the extent and/or the exact date for each particular phenological event seem to depend on the microenvironmental conditions [31]. The main physiological advantage of the semi-deciduous habit is the decrease in the transpiring leaf area during the summer dry months, resulting in more efficient water economy. ...
... Especially in semi-arid systems and drylands, the precipitation during water-deficit periods is significantly more important in driving phenology than the rain during favorable moisture conditions. Broich et al. [31] suggested that the stronger correlation patterns with single-month compared with multi-month aggregated drivers indicate that rainfall at a specific time-point determines most phenological events. Corroborating this conclusion, our findings suggest that 7 out of 13 phenology metrics were influenced by single or double-month precipitation related parameters in the drier site of Araxos. ...
Full-text available
A 21-year Enhanced Vegetation Index (EVI) time-series produced from MODIS satellite images was used to study the complex phenological cycle of the drought semi-deciduous shrub Phlomis fruticosa and additionally to identify and compare phenological events between two Mediterranean sites with different microclimates. In the more xeric Araxos site, spring leaf fall starts earlier, autumn revival occurs later, and the dry period is longer, compared with the more favorable Louros site. Accordingly, the control of climatic factors on phenological events was examined and found that the Araxos site is mostly influenced by rain related events while Louros site by both rain and temperature. Spring phenological events showed significant shifts at a rate of 1–4.9 days per year in Araxos, which were positively related to trends for decreasing spring precipitation and increasing summer temperature. Furthermore, the climatic control on the inter-annual EVI fluctuation was examined through multiple linear regression and machine learning approaches. For both sites, temperature during the previous 2–3 months and rain days of the previous 3 months were identified as the main drivers of the EVI profile. Our results emphasize the importance of focusing on a single species and small-spatial-scale information in connecting vegetation responses to the climate crisis.
... Apply the Breaks for Additive Seasonal and Trend (BFAST) to the complete timeseries VPR-Residuals to identify potential breakpoints [12,28]. The BFAST method is a widely used method for assessing variable trends in time series data in ecosystem assessment, which detects the phenological cycles that vegetation skipped without influencing ecological health as breakpoints [29]. A Chow test was used to check the breakpoints in VPR and residual trends, and then to assess whether these breakpoints have a significant impact [12,30]. ...
Full-text available
Satellite remote sensing has witnessed a global widespread vegetation greening since the 1980s. However, reliable observation-based quantitative knowledge on global greening remains obscure due to uncertainties in model simulations and the contribution of natural variability is largely unknown. Here, we revisit the attribution of global vegetation changes using the Time Series Segment and Residual Trend (TSS-RESTREND) method. Results showed global vegetation significantly greening over 40.6% of the vegetated grids, whereas vegetation significantly browning over 11.6% of the vegetated grids. The attribution results based on the TSS-RESTREND method show that CO2 fertilization (CO2) plays an influential role in vegetation changes over 61.4% of the global vegetated areas, followed by land use (LU, 23.5%), climate change (CC, 7.3%), and climate variability (CV, 1.5%). The vegetation greening can be largely attributed to CO2 fertilization while the vegetation browning is mainly caused by LU. Meanwhile, we also identify positive impacts of LU and CC on vegetation change in arid regions but negative impacts in humid regions. Our findings indicate spatial heterogeneity in causes behind global vegetation changes, providing more detailed references for global vegetation modeling.
... Decisions trees are popular supervised classification methods for wetland applications that also have applicability to a wide range of other data problems such as ranking, probability estimation, regression, and clustering and to a variety of remote sensing land cover mapping applications, including wetlands [276][277][278][279][280]. Decision trees are based on a series of logical decisions that are easily interpreted; however, their high expressivity results in a tendency to overfit models [281]. ...
Full-text available
The Prairie Pothole Region (PPR) of North America is an extremely important habitat for a diverse range of wetland ecosystems that provide a wealth of socio-economic value. This paper describes the ecological characteristics and importance of PPR wetlands and the use of remote sensing for mapping and monitoring applications. While there are comprehensive reviews for wetland remote sensing in recent publications, there is no comprehensive review about the use of remote sensing in the PPR. First, the PPR is described, including the wetland classification systems that have been used, the water regimes that control the surface water and water levels, and the soil and vegetation characteristics of the region. The tools and techniques that have been used in the PPR for analyses of geospatial data for wetland applications are described. Field observations for ground truth data are critical for good validation and accuracy assessment of the many products that are produced. Wetland classification approaches are reviewed, including Decision Trees, Machine Learning, and object versus pixel-based approaches. A comprehensive description of the remote sensing systems and data that have been employed by various studies in the PPR is provided. A wide range of data can be used for various applications, including passive optical data like aerial photographs or satellite-based, Earth-observation data. Both airborne and spaceborne lidar studies are described. A detailed description of Synthetic Aperture RADAR (SAR) data and research are provided. The state of the art is the use of multi-source data to achieve higher accuracies and hybrid approaches. Digital Surface Models are also being incorporated in geospatial analyses to separate forest and shrub and emergent systems based on vegetation height. Remote sensing provides a cost-effective mechanism for mapping and monitoring PPR wetlands, especially with the logistical difficulties and cost of field-based methods. The wetland characteristics of the PPR dictate the need for high resolution in both time and space, which is increasingly possible with the numerous and increasing remote sensing systems available and the trend to open-source data and tools. The fusion of multi-source remote sensing data via state-of-the-art machine learning is recommended for wetland applications in the PPR. The use of such data promotes flexibility for sensor addition, subtraction, or substitution as a function of application needs and potential cost restrictions. This is important in the PPR because of the challenges related to the highly dynamic nature of this unique region.
... One possible reason is that vegetation phenology is largely temperature-driven in the Northern Hemisphere and likely precipitation-driven in arid and semi-arid systems (e.g., Australia in the Southern Hemisphere). Broich et al. (2014) showed that vegetation phenological cycles across Australia have large inter-and intra-annual variations and the peak of phenological cycles could occurred at any time in a calendar year. It is necessary to strengthen vegetation phenology studies in the Southern Hemisphere to enhance our understanding of vegetation activities under global climate change. ...
Full-text available
The land surface phenology (LSP) associated with vegetation dynamics plays an important role in influencing the land surface processes and land–atmosphere interactions. Satellite observations have been widely used in studies for the monitoring of LSP across large areas based on different phenology retrieval algorithms and the routine production of LSP from remote sensing data has yet come to fruition. Here we used six phenology retrieval methods, including the amplitude threshold (AT), first-order derivative (FOD), second-order derivative (SOD), relative changing rate (RCR), third-order derivative (TOD), and curvature change rate (CCR), to retrieve the start of the growing season (SOS) and the end of the growing season (EOS) from the Advanced Very High Resolution Radiometer (AVHRR) data. We improved the curve fitting method to reduce uncertainties owing to data preprocessing. The results indicated that both SOS and EOS retrieved by six different methods had similar spatial distribution and the retrieved dates could vary largely at the pixel level. In the Northern Hemisphere, from 1982 to 2018, the trends of SOS retrieved vary across methods and only the EOS extracted by the relative change curvature method had a significant advanced trend. In the Southern Hemisphere, from 1982 to 2018, SOS results derived from four methods (i.e., AT, SOD, TOD, and CCR) showed significantly delayed trends, EOS results extracted by all the methods demonstrated insignificant trends. The phenology retrieval methods were assessed using the field observation data from the Pan European Phenology Project (PEP725) and from time series of leaf area index (LAI) measured at flux towers. The satellite-retrieved dates of both SOS and EOS were positively correlated with field observation and the relationships are largely dependent on how field phenology metrics are defined. We presented longer time series (1982–2018) data of phenology metrics with fewer gaps and multiple phenology retrieving methods as compared to the MODIS land cover dynamics product. Based on our assessments, one might use the SOS generated by FOD and the EOS generated by RCR as they provide results the most consistent with field data among all the tested methods. If studies aim to use the earliest SOS (or latest EOS) in a year, one might use the data retrieved based on TOD or CCR. The global dataset is delivered for uses in studies and applications associated with LSP.
... FSDAF effectively reduces the amount of data input and is easily operated. At present, global land cover and land surface phenology products (MOD09GA) are available at 500 m spatial resolution from MODIS [22,23], but the 500 m spatial resolution is too coarse for most crop fields. This resolution often results in mixed pixels of different vegetation or crop types, which may have very different phenological growth cycles [24,25]. ...
Full-text available
Research on fusion modeling of high spatial and temporal resolution images typically uses MODIS products at 500 m and 250 m resolution with Landsat images at 30 m, but the effect on results of the date of reference images and the ‘mixed pixels’ nature of moderate-resolution imaging spectroradiometer (MODIS) images are not often considered. In this study, we evaluated those effects using the flexible spatiotemporal data fusion model (FSDAF) to generate fusion images with both high spatial resolution and frequent coverage over three cotton field plots in the San Joaquin Valley of California, USA. Landsat images of different dates (day-of-year (DOY) 174, 206, and 254, representing early, middle, and end stages of the growing season, respectively) were used as reference images in fusion with two MODIS products (MOD09GA and MOD13Q1) to produce new time-series fusion images with improved temporal sampling over that provided by Landsat alone. The impact on the accuracy of yield estimation of the different Landsat reference dates, as well as the degree of mixing of the two MODIS products, were evaluated. A mixed degree index (MDI) was constructed to evaluate the accuracy and time-series fusion results of the different cotton plots, after which the different yield estimation models were compared. The results show the following: (1) there is a strong correlation (above 0.6) between cotton yield and both the Normalized Difference Vegetation Index (NDVI) from Landsat (NDVIL30) and NDVI from the fusion of Landsat with MOD13Q1 (NDVIF250). (2) Use of a mid-season Landsat image as reference for the fusion of MODIS imagery provides a better yield estimation, 14.73% and 17.26% higher than reference images from early or late in the season, respectively. (3) The accuracy of the yield estimation model of the three plots is different and relates to the MDI of the plots and the types of surrounding crops. These results can be used as a reference for data fusion for vegetation monitoring using remote sensing at the field scale.
Full-text available
In 2020, the Australian and New Zealand flux research and monitoring network, OzFlux, celebrated its 20th anniversary by reflecting on the lessons learned through two decades of ecosystem studies on global change biology. OzFlux is a network not only for ecosystem researchers, but also for those ‘next users’ of the knowledge, information and data that such networks provide. Here, we focus on eight lessons across topics of climate change and variability, disturbance and resilience, drought and heat stress and synergies with remote sensing and modelling. In distilling the key lessons learned, we also identify where further research is needed to fill knowledge gaps and improve the utility and relevance of the outputs from OzFlux. Extreme climate variability across Australia and New Zealand (droughts and flooding rains) provides a natural laboratory for a global understanding of ecosystems in this time of accelerating climate change. As evidence of worsening global fire risk emerges, the natural ability of these ecosystems to recover from disturbances, such as fire and cyclones, provides lessons on adaptation and resilience to disturbance. Drought and heatwaves are common occurrences across large parts of the region and can tip an ecosystem's carbon budget from a net CO2 sink to a net CO2 source. Despite such responses to stress, ecosystems at OzFlux sites show their resilience to climate variability by rapidly pivoting back to a strong carbon sink upon the return of favourable conditions. Located in under-represented areas, OzFlux data have the potential for reducing uncertainties in global remote sensing products, and these data provide several opportunities to develop new theories and improve our ecosystem models. The accumulated impacts of these lessons over the last 20 years highlights the value of long-term flux observations for natural and managed systems. A future vision for OzFlux includes ongoing and newly developed synergies with ecophysiologists, ecologists, geologists, remote sensors and modellers.
Vegetation growth drives many of the interactions between the land surface and atmosphere including the uptake of carbon through photosynthesis and loss of water through transpiration. In arid and semi-arid regions water is the dominant driver of vegetation growth. However, few studies consider the fact that water can move laterally across the landscape as runoff via streams and floodplains, termed hydrologic connectivity. Using multiple observations alongside models and a hydromorphology dataset for Australia, we examine how ecosystems with high hydrologic connectivity differ in their vegetation response to water availability, soil properties, and interannual variability and extremes in vegetation productivity. We find that the average interannual variability of vegetation productivity is 21–34% higher in ecosystems with high hydrologic connectivity, with skewed annual anomalies showing larger extremes in carbon uptake. This is driven by a higher average and more variable surface soil moisture and significantly higher soil available water capacity and soil depth. These spatially small ecosystems, covering 14% of the study region, contribute 15–22% (median = 17%) to regional-scale carbon uptake through higher rates of gross photosynthesis, especially evident during wet years, and 3–37% (median = 19%) to annual anomalies. Current global land surface models do not reproduce the observed spatial patterns of interannual variability in carbon uptake over regions where hydrologic connectivity is high as they lack the mechanism of connectivity of water between discrete land surface elements. This study highlights the significant role of riparian and floodplain vegetation on the interannual variability and extremes of the regional carbon cycle.
Full-text available
The Mediterranean region is one of the most vulnerable regions to climate change. The majority of climate models forecast a rise in temperatures and less rainfall, which have been observed in recent decades. These changes will affect several vegetation properties, especially phenological dynamics and traits, by increasing drought intensity and recurrence. In this climate change context, the present study aims to assess the evolution of vegetation state and its relation with the climate dynamics in the Mediterranean forest region of northeast Tunisia using Land Surface Phenology (LSP) metrics and the vegetation index (NDVI) analysis from 2000 to 2017. To conduct this work, we used precipitation and temperature data from the two closest weather stations and 16-day NDVI composite images from the MODIS satellite source, with 250-m spatial resolution. Three phenological metrics— start of season (SOS), end of season (EOS), and length of season (LOS) — were obtained and compared for different vegetation types. The LSP variation in response to climatic metrics was also analyzed. The results showed that the LSP in our study area changed significantly during the 2000–2017 period, which includes an average 7.8 days delay in the SOS, an average advance in the EOS by 5 days, and LOS shortened by an average 12.8 days. Autumn (Pr_9) and spring (Pr_3 and P3_4) precipitations, as well as maximum temperature (Tx9+10), represent the best climate parameters to explain the changes in LSP. Both the NDVI and SPEI showed a significant high correlation (p < 0.001) on longer time scales. LSP and NDVI proved useful tools for monitoring the vegetation state according to climate for better planning territory purposes.
Full-text available
As this edition of The World’s Water goes to press in early 2011, eastern Australia is recovering from devastating floods that claimed more than 20 lives and destroyed hundreds of homes. The heavy rains of 2009 and 2010 that caused so much destruction also marked the end of Australia’s decade-long Millennium Drought. Beginning in about 1997, declines in rainfall and runoff had contributed to widespread crop failures, livestock losses, dust storms, and bushfires. Such are the vagaries of water on the continent with the world’s most uncertain and variable climate.
Full-text available
Drylands, covering nearly 30% of the global land surface, are characterized by high climate variability and sensitivity to land management. Here, two satellite-observed vegetation products were used to study the long-term (1988-2008) vegetation changes of global drylands: the widely used reflective-based Normalized Difference Vegetation Index (NDVI) and the recently developed passive-microwave-based Vegetation Optical Depth (VOD). The NDVI is sensitive to the chlorophyll concentrations in the canopy and the canopy cover fraction, while the VOD is sensitive to vegetation water content of both leafy and woody components. Therefore it can be expected that using both products helps to better characterize vegetation dynamics, particularly over regions with mixed herbaceous and woody vegetation. Linear regression analysis was performed between antecedent precipitation and observed NDVI and VOD independently to distinguish the contribution of climatic and non-climatic drivers in vegetation variations. Where possible, the contributions of fire, grazing, agriculture and CO2 level to vegetation trends were assessed. The results suggest that NDVI is more sensitive to fluctuations in herbaceous vegetation, which primarily uses shallow soil water, whereas VOD is more sensitive to woody vegetation, which additionally can exploit deeper water stores. Globally, evidence is found for woody encroachment over drylands. In the arid drylands, woody encroachment appears to be at the expense of herbaceous vegetation and a global driver is interpreted. Trends in semi-arid drylands vary widely between regions, suggesting that local rather than global drivers caused most of the vegetation response. In savannas, besides precipitation, fire regime plays an important role in shaping trends. Our results demonstrate that NDVI and VOD provide complementary information and allow new insights into dryland vegetation dynamics.
Full-text available
[1] Precipitation regimes are predicted to shift to more extreme patterns that are characterized by more heavy rainfall events and longer dry intervals, yet their ecological impacts on vegetation production remain uncertain across biomes in natural climatic conditions. This in situ study investigated the effects of these climatic conditions on aboveground net primary production (ANPP) by combining a greenness index from satellite measurements and climatic records during 2000–2009 from 11 long-term experimental sites in multiple biomes and climates. Results showed that extreme precipitation patterns decreased the sensitivity of ANPP to total annual precipitation (PT) at the regional and decadal scales, leading to decreased rain use efficiency (RUE; by 20% on average) across biomes. Relative decreases in ANPP were greatest for arid grassland (16%) and Mediterranean forest (20%) and less for mesic grassland and temperate forest (3%). The cooccurrence of heavy rainfall events and longer dry intervals caused greater water stress conditions that resulted in reduced vegetation production. A new generalized model was developed using a function of both PT and an index of precipitation extremes and improved predictions of the sensitivity of ANPP to changes in precipitation patterns. Our results suggest that extreme precipitation patterns have substantially negative effects on vegetation production across biomes and are as important as PT. With predictions of more extreme weather events, forecasts of ecosystem production should consider these nonlinear responses to altered extreme precipitation patterns associated with climate change.
Full-text available
A more complete picture of the timing and patterns of the ENSO signal during the seasonal cycle for the whole of Africa over the three last decades is provided using the normalized difference vegetation index (NDVI). Indeed, NDVI has a higher spatial resolution and is more frequently updated than in situ climate databases, and highlights the impact of ENSO on vegetation dynamics as a combined result of ENSO on rainfall, solar radiation, and temperature. The month-by-month NDVI–Ni~no-3.4 correlation patterns evolve as follows. From July to September, negative correlations are observed over the Sahel, the Gulf of Guinea coast, and regions from the northern Democratic Republic of Congo to Ethiopia. However, they are not uniform in space and are moderate (;0.3). Conversely, positive correlations are recorded over the winter rainfall region of South Africa. In October– November, negative correlations over Ethiopia, Sudan, and Uganda strengthen while positive correlations emerge in the Horn of Africa and in the southeast coast of South Africa. By December with the settlement of the ITCZ south of the equator, positive correlations over the Horn of Africa spread southward and westward while negative correlations appear over Mozambique, Zimbabwe, and South Africa. This pattern strengthens and a dipole at 188S is well established in February–March with reduced (enhanced) greenness during ENSO years south (north) of 188S. At the same time, at ;28N negative correlations spread northward. Last, from April to June negative correlations south of 188S spread to the north (to 108S) and to the east (to the south of Tanzania).
In this chapter we explain satellite-based vegetation indices (VIs) as dynamic spectral measures of vegetation activity. VIs are among the most widely used satellite products in monitoring ecosystems and agriculture, resource management, and estimations of many biophysical canopy properties. A theoretical basis for their formulation is presented and we describe how VIs are processed and composited from satellite imagery. Recent trends in their validation and quality assessment using in situ tower measurements are also discussed. Finally, a cross section of major findings involving the use of satellite VIs in ecological and climate science is presented and we conclude with research challenges and environmental issues that will drive future uses of satellite VIs.
In attempting to analyze, on digital computers, data from basically continuous physical experiments, numerical methods of performing familiar operations must be developed. The operations of differentiation and filtering are especially important both as an end in themselves, and as a prelude to further treatment of the data. Numerical counterparts of analog devices that perform these operations, such as RC filters, are often considered. However, the method of least squares may be used without additional computational complexity and with considerable improvement in the information obtained. The least squares calculations may be carried out in the computer by convolution of the data points with properly chosen sets of integers. These sets of integers and their normalizing factors are described and their use is illustrated in spectroscopic applications. The computer programs required are relatively simple. Two examples are presented as subroutines in the FORTRAN language.
This paper presents a continental-scale phenological analysis of African savannas and woodlands. We apply an array of synergistic vegetation and hydrological data records from satellite remote sensing and model simulations to explore the influence of rainy season timing and duration on regional land surface phenology and ecosystem structure. We find that: (i) the rainy season onset precedes and is an effective predictor of the growing season onset in African grasslands. (ii) African woodlands generally have early green-up before rainy season onset, and have a variable delayed senescence period after the rainy season, with this delay correlated non-linearly with tree fraction. These woodland responses suggest their complex water use mechanisms (either from potential groundwater use by relatively deep roots or stem-water reserve) to to maintain dry season activity. (iii) We empirically find that the rainy season length has strong non-linear impacts on tree fractional cover in the annual rainfall range from 600-1800 mm/year, which may lend some support to the previous modeling study that given the same amount of total rainfall, the tree fraction may first increases with the lengthening of rainy season unitl reaching an “optimal rainy season length”, after which tree fraction decreases with the further lengthening of rainy season. This non-linear response is resulted from compound mechanisms of hydrological cycle, fire and other factors. We conclude that African savannas and deciduous woodlands have distinctive responses in their phenology and ecosystem functioning to rainy season. Further research is needed to address interaction between groundwater and tropical woodland as well as to explicitly consider the ecological significance of rainy season length under climate change.