Productivity responses of desert vegetation to precipitation patterns

Productivity responses of desert vegetation to precipitation patterns

J Plant Res (2015) 128:283–294
DOI 10.1007/s10265-014-0685-4
REGULAR PAPER
Productivity responses of desert vegetation to precipitation
patterns across a rainfall gradient
Fang Li · Wenzhi Zhao · Hu Liu Received: 5 May 2014 / Accepted: 27 September 2014 / Published online: 23 January 2015
© The Botanical Society of Japan and Springer Japan 2014
Abstract The influences of previous-year precipitation
and episodic rainfall events on dryland plants and communities are poorly quantified in the temperate desert region
of Northwest China. To evaluate the thresholds and lags
in the response of aboveground net primary productivity
(ANPP) to variability in rainfall pulses and seasonal precipitation along the precipitation-productivity gradient in three
desert ecosystems with different precipitation regimes, we
collected precipitation data from 2000 to 2012 in Shandan (SD), Linze (LZ) and Jiuquan (JQ) in northwestern
China. Further, we extracted the corresponding MODIS
Normalized Difference Vegetation Index (NDVI, a proxy
for ANPP) datasets at 250 m spatial resolution. We then
evaluated different desert ecosystems responses using statistical analysis, and a threshold-delay model (TDM). TDM
is an integrative framework for analysis of plant growth,
precipitation thresholds, and plant functional type strategies that capture the nonlinear nature of plant responses
to rainfall pulses. Our results showed that: (1) the growing season NDVIINT (INT stands for time-integrated) was
largely correlated with the warm season (spring/summer) at
our mildly-arid desert ecosystem (SD). The arid ecosystem
(LZ) exhibited a different response, and the growing season
NDVIINT depended highly on the previous year’s fall/winter precipitation and ANPP. At the extremely arid site (JQ),
the variability of growing season NDVIINT was equally correlated with the cool- and warm-season precipitation; (2)
F. Li · W. Zhao (*) · H. Liu Linze Inland River Basin Research Station, Key Laboratory
of Inland River Basin Ecohydrology, Cold and Arid
Regions Environmental and Engineering Research Institute,
Chinese Academy of Sciences, 320 Dong‑gang West Road,
Lanzhou 730000, China
e-mail: [email protected]
some parameters of threshold-delay differed among the
three sites: while the response of NDVI to rainfall pulses
began at about 5 mm for all the sites, the maximum thresholds in SD, LZ, and JQ were about 55, 35 and 30 mm
respectively, increasing with an increase in mean annual
precipitation. By and large, more previous year’s fall/winter precipitation, and large rainfall events, significantly
enhanced the growth of desert vegetation, and desert ecosystems should be much more adaptive under likely future
scenarios of increasing fall/winter precipitation and large
rainfall events. These results highlight the inherent complexity in predicting how desert ecosystems will respond to
future fluctuations in precipitation.
Keywords Aboveground net primary productivity ·
Previous precipitation · Rainfall events · Temperate desert
regions · Threshold-delay model
Introduction
Water is the primary limiting resource for terrestrial biological activity in arid and semi-arid regions, and water
availability mediates the responses of plant communities
and entire ecosystems (Fay et al. 2003; Fay et al. 2008;
Huxman et al. 2004a; Jenerette et al. 2008; Zeppel et al.
2007). In water-limited ecosystems, complex linkages
exist between rainfall and vegetation behavior, as rain is
intermittent, and unpredictable, and exhibits lag-effects
(Li et al. 2013; Ospina et al. 2012). Global climate change
is predicted to further increase rainfall variability, with
increases in drought intensity and duration, and changes
in seasonality (less summer and more winter precipitation)
(IPCC 2007). Exactly how different desert ecosystems will
respond to these changes is unclear and requires research at
13
284
the interface of ecology and hydrology, especially in arid
regions (Plaut et al. 2013; Roca et al. 2004).
Aboveground net primary productivity (ANPP), a key
ecosystem process, exhibits spatial and temporal variability, and reflects floral and community responses to fluctuations in annual and seasonal precipitation (Bradford
et al. 2006; Shafran-Nathan et al. 2012). In all terrestrial
ecosystems, plant functions, including ANPP, respond by
constraining gas exchange during dry periods, and altering
structural and physiological attributes that affect the ability of plants to utilize soil water following precipitation
(Blackman et al. 2009; Brodribb et al. 2010; Resco et al.
2009).
The normalized difference vegetation index (NDVI) provides information on plant density and growing conditions
on the ground (Berger et al. 2013), and it is widely used as
a proxy for productivity in arid ecosystems (Gamon et al.
2013; Horion et al. 2013; Svoray and Karnieli 2011). NDVI
is closely and positively correlated with the fraction of photosynthetically active radiation absorbed by green vegetation which tends to be in synchrony with canopy development, when the inter-annual variability of incident radiation
and radiation use efficiency is small (Camberlin et al. 2007;
Paruelo et al. 1997). These requirements were met in our
study sites, and thus most of the inter-annual variability in
ANPP was accounted for by changes in green-leaf biomass
(Fabricante et al. 2009).
Precipitation seasonality exerts an important control
over ecosystem processes, particularly in dry ecosystems,
and affects productivity (Gao and Reynolds 2003; Xiao and
Moody 2004), vegetation structure, and composition (Robinson et al. 2013). Desert ecosystems have a high potential to respond to changes in seasonality (Robinson et al.
2013). The relationship between seasonally-accumulated
rainfall and ANPP during the growing season has been
studied extensively on a regional scale, but fall/winter climate dynamics also play an important role in determining
the structure and function of arid-land ecosystems (Hamerlynck et al. 2013; Hamerlynck et al. 2012a). Previous-year
precipitation controls a significant fraction of the variability in current-year production (Hamerlynck et al. 2013;
Reichmann et al. 2013; Sala et al. 2012; Yahdjian and Sala
2006). In arid-land ecosystems fall/winter rainfall, especially that of low intensity, long duration, and large areal
extent, is critical for plant springtime activity (Hamerlynck
et al. 2012a; Reichmann et al. 2013). At the same time, the
spring response of plant physiology and production to fall/
winter precipitation also affects the response to summer
rainfall threshold values (Emmerich and Verdugo 2008).
However, within-season variation in ANPP, especially
that from July to September, is controlled by a variety of
interacting factors, mainly the magnitude and frequency of
precipitation events (Muldavin et al. 2008). In fact, high
13
J Plant Res (2015) 128:283–294
rates of shrub production are triggered by water pulses during warm periods (Sponseller et al. 2012). ANPP generally
increases with rainfall event size, but remains within the
limits of the upper and lower thresholds of the size of
events that effectively contribute to ANPP (Ogle and Reynolds 2004; Swemmer et al. 2007). For short-grass prairie,
the greatest response of ANPP to an individual precipitation event was an event of 15–30 mm (Lauenroth and Sala
1992). However, responses of different plant functional
types to rain events differ across different biomes (Reynolds et al. 2004). For example, ANPP was reduced by 18 %
in a tallgrass prairie (Fay et al. 2003; Heisler-White et al.
2009; Knapp et al. 2002) and by 16 % in an arid grassland
(Zhang et al. 2013b), but increased by 70 % in a mixedgrass grassland (Heisler-White et al. 2009) following
extreme precipitation regimes.
The objective of this study was to evaluate different
responses of desert ecosystems to precipitation patterns
along a natural rainfall gradient ranging from 84 to 198 mm
of mean annual precipitation. We evaluated 13 years of precipitation and corresponding NDVI datasets in three desert
ecosystems, selected individual precipitation events during
the summer, and evaluated different ecosystem responses
using the threshold-delay model and statistical analysis.
Specifically, we wanted to answer the following questions:
(1) How does ANPP in desert ecosystems respond to different-sized rainfall events? (2) Does ANPP respond differently to warm-season than to cool-season precipitation in
desert ecosystems? Understanding the responses to variations in precipitation patterns will assist in assessing how
desert ecosystems may change under likely future scenarios of more extreme precipitation regimes.
Materials and methods
Study area and datasets
In this study, we selected three desert ecosystems in northwestern China: Shandan (SD, 38°45′–38°49′ N, 100°46′–
100°50′ E, 1664 m), Linze (LZ, 39°24′–39°26′ N, 100°05′–
100°08′ E; 1445 m), and Jiuquan (JQ, 39°58′–40°02′ N,
98°36′–98°41′ E; 1477 m), which are about 100–200 km
apart. These three sites exhibited different climate, vegetation and soil properties. The climate of the SD desert is
alpine semi-arid temperate continental desert; at LZ, it is
arid to semiarid temperate continental desert; and at JQ, it
is arid temperate continental desert. Mean annual temperatures are 6.5, 7.6 and 7.9 °C; the potential evaporation values are 1,700, 2,390 and 2,560 mm; the relative humidity
values are 49, 46 and 43 %; and mean annual precipitation
values are 198, 110 and 87 mm, at SD, LZ and JQ, respectively, about 60 % of which falls during June–August.
J Plant Res (2015) 128:283–294
285
summer, decreased, in the 2000- to- 2012 period, compared
to 1953 to 2012, and changed from 39, 43, and 46 % to
35, 40, and 41 % at SD, LZ, and JQ, respectively (Fig. 1).
Furthermore, the mean annual temperature increased about
0.41, 0.24 and 0.19 °C per 10 years at SD, LZ and JQ from
1952 to 2012.
Zonal vegetation is temperate desert steppe and consists
of super xerophytic shrubs as well as xerophytic and salttolerant sub-shrubs at the three sites; however each site
had different vegetation community characteristics. Vegetation cover percentages were 19, 11, and 7 % at SD, LZ
and JQ, respectively. Vegetation densities were 0.93, 0.35,
and 0.19 plants m−2, and numbers of species were 18, 11
and 8, at SD, LZ and JQ, respectively. The plant community at SD was dominated by a C3 sub-shrub—Asterothamnus alyssoides, with an average height of 20–50 cm, a C4
sub-shrub—Salsola passerina, with an average height of
15–30 cm, and annual plants that were abundant in the
late growing season and hand relatively complex structures. Only one C3 sub-shrub—Reaumuria soongorica, at
an average height of 15–40 cm, and a few annual plants in
the late growing season with relatively simple structure,
dominated at LZ. A C3 sub-shrub—Reaumuria soongorica,
with average height of 10–30 cm, and a C4 sub-shrub—Salsola passerina, with average height of 15–25 cm, and few
annual plants, dominated JQ.
The zonal soil type is aridisol with soil texture of 53 and
78 % sand, 42 and 19 % silt, and 5 and 3 % clay, at SD
and LZ, respectively; no data were available for JQ. Average annual soil moisture at 0-60 cm soil depth is 3.7, 2.5,
and 1.9 % at SD, LZ, and JQ, respectively (Zhang et al.
2007). Bulk density was 1.32 and 1.57 gm−2, and pore
space was 50 and 41 % at SD and LZ, respectively; no
data were available for JQ (Wang and Zhao 2009). Roots
of Asterothamnus alyssoides were concentrated near the
soil surface at 0–10 cm depth, and accounted for 60 % of
total roots, with Reaumuria soongorica roots making up
the remaining 40 %. Soil freezing occurs from October
to March–April at all three sites. In these systems, water
availability depends largely on precipitation, supplemented
with infrequent mountain snowmelt. The depth of snow in
cold season at SD, LZ, and JQ is detailed in Table 1.
We used 13 years of growing season NDVI data for SD,
LZ, and JQ to evaluate patterns of plant growth in desert
ecosystems across a gradient of MAP. We acquired surface
reflectance 8-day composited data with reduced influence
of weather conditions and clouds from MODIS Terra maintained by the NASA Land Processes Distributed Active
Archive Center at the USGS (http://ladsweb.nascom.nasa.
gov/data/search.html.). Data were at 5-level, 3-global,
250 m spatial resolution (MOD09Q1) in HDF format, for
the growing season from May to September.
The images were treated with projection transformation
and image registration, and were clipped using the study
area borders executed in ENVI 4.7 Software. MOD09Q1
include red and near-infrared spectral bands that are used
to obtain NDVI. Because NDVI values for an 8-day period
could be recorded on any date within that period, we
assumed that the NDVI value corresponded to the middle
day of the NDVI compositing period (Zhang et al. 2013a).
In addition, we obtained growing-season scales (total-,
early- and late-) NDVI (NDVIINT, where INT stands for
time-integrated) by summing NDVI for each 8-day period
from 1 May to 30 September (total growing season), from
1 May to 30 June (early growing season) and from 1 July to
30 September (late growing season) to determine the interannual variation at the three sites. The midseason date of
1 July was determined based on personal communication
with principal investigators.
Frequency distributions of rainfall events
Rainfall data
Precipitation patterns at the three sites in our study can be
characterized as typical of arid regions. Annual rainfall
averaged 198, 110, and 84 mm in SD, LZ, and JQ from
1953 to 2012, but appeared to be increasing to 215, 113 and
92 mm at the three sites, respectively, in the last thirteen
years (2000–2012) (Table 1). Precipitation was unimodal,
with the majority of precipitation occurring in the summer (June–August). Most of the rainfall events were small
(<5 mm). Large rainfall events (>10 mm) made up 15, 10
and 7 % of total precipitation at SD, LZ, and JQ, respectively, over the 60-year period from 1953 to 2012; however,
these proportions increased to 18, 12 and 9 % at the three
sites in the last thirteen years (Fig. 1). Also, the frequency
of rainfall events in fall/winter/spring increased; and in
Meteorological data (2002–2012) for the LZ study area
were measured at an open-field weather station located
approximately 5 km from the study site. Daily rainfall data for years 1967–2002 at LZ, and 1953–2012 at
SD and JQ, were acquired from http://www.cma.gov.
cn/2011qxfw/2011qsjgx/, maintained by the National
Meteorological Information Center (NMIC).
Continuous rainy-day data were treated as individual
rainfall events. Those events totaling less than 3 mm were
not used, as this amount of rain was assumed to be negligible for plant activity. We then selected 37, 33, and 38
independent precipitation events at JQ, LZ, and SD, respectively. The selected rainfall events met the following conditions (Wang et al. 2003): (1) no other major rainfall events
NDVI data
13
286
J Plant Res (2015) 128:283–294
Table 1 Climate, vegetation and soil characteristics of three desert ecosystems located within the He-xi corridor of Northwest China
Climate properties
Characteristics
Shandan (SD)
Linze (LZ)
Jiuquan (JQ)
Study area (km2)
Altitude (m)
Topography
Climatic types
56
1,664
Mildly undulating
Alpine semi-arid temperate
desert
198
39.4
45
1,445
Flat
Arid to semiarid temperate
desert
110
29.1
52
1,477
Mildly undulating
Arid temperate desert
84
26.5
68.7
5.9
49
1,700
<12
19 ± 7
24 ± 8
0.1319 ± 0.016
0.93 ± 0.31
18
73.7
7.6
46
2,390
<10
11 ± 5
19 ± 9
0.0752 ± 0.008
0.35 ± 0.11
11
46.9
7.9
43
2,560
<14
5 ± 4
14 ± 9
0.0652 ± 0.005
0.19 ± 0.17
8
Reaumuria soongorica
Sandy–silt–clay* (%)
Bulk density* (gm−3)
Pore space* (%)
Swcg (0–60 cm) (%)
Swc (60–180 cm) (%)
Asterothamnus alyssoides,
Salsola passerina
53–42–5
1.32
50
3.7 ± 1.67
–
78–19–3
1.57
41
2.45 ± 1.33
1.89 ± 0.67
Reaumuria soongorica, Salsola passerina
–
–
–
1.87 ± 0.92
–
MSFDh (cm)
110
100
143
MAPa (mm)
MNb (n)
MEc (mm)
MATd (°C)
RHe (%)
Evaporation (mm)
Snow depth(mm)
Vegetation properties Cover (%)
Height (cm)
Average NDVIf
Density (N m−2)
Species richness
Dominant plant species
Soil properties
a
MAP is the mean annual precipitation from 1953 to 2012
b
MN is the mean number of precipitation events per year from 1953 to 2012
c
ME is the maximum number of precipitation events during 1953–2012
d
MAT is the mean annual temperature from 1953 to 2012
e
RH is relative humidity from 1953 to 2012
f
Average NDVI is the average value over the growing season (May–September) from 2000 to 2012
g
Swc is average soil water content during April–October from 2002 to 2012 at the Linze site
h
MSFD is the maximum depth of soil freezing
*The data of soil physical properties were compiled from the literature (Wang and Zhao 2009)
– Indicates no data available
occurred during the preceding 15 days, the antecedent soil
moisture was constant, and 15 days after the pulse, soil
moisture returned to pre-pulse conditions (Hamerlynck
et al. 2012b); (2) the event occurred close to the middle
of the growing season (June–August), when temperatures
were relatively high. In addition, we calculated hydrological annual precipitation as the sum of all precipitation from
the previous September to August, based on a plant-centric
precipitation. We then sub-divided the hydrological annual
precipitation data into four seasons: autumn (September–
November), winter (December–February), spring (March–
May) and summer (June–August). Additionally, we defined
the cool and warm season precipitation as that falling in
13
the previous autumn and winter and in spring and summer,
respectively.
Statistical associations between variables
We examined multiple regression analyses with forward
selection to assess the sensitivity of ANPP (total, early-,
and late-growing season NDVIINT) to seasonal patterns
of precipitation using cool- (previous autumn and winter)
and warm- (current spring and summer) season precipitation as predictors of ANPP. Moreover, we used adjusted
R2 and partial (r2) to evaluate the explanatory power of our
analyses. The analyses were considered to be statistically
J Plant Res (2015) 128:283–294
287
Fig. 1 Frequency distributions of rainfall events. a represents the average from 1953 to 2012 throughout all panels; b represents the average
from 2000 to 2012 throughout all panels
supported when they had a P < 0.01. We performed all statistical analyses in SPSS and mapping in Origin 8.0.
The threshold‑delay model
We used the threshold-delay model (Ogle and Reynolds
2004) to analyze and compare the rate, and duration of
response, and the thresholds below or above which no, or
no further, response was observed.
The threshold-delay model is an integrative framework for
analysis of plant growth, precipitation thresholds, and plant
functional type strategies that capture the nonlinear nature of
plant responses to rainfall pulses. It is based on six parameters, including lower and higher precipitation thresholds
(RL or RU), lags (τ), potential responses (δmax), maximum
response rates (ymax), and the reduction in the response variables over time (k). A large (small) k-value means that the
duration of the plant response to a precipitation pulse should
be longer (shorter). The model can be expressed as follows:
yt = k × yt −1 +δt
(1)
yt −1
∗
δt = Min ymax × (1 − k), δt 1 −
ymax
(2)
 δ
· (Rt−τ − RL )
 RUmax
−RL
∗
δt = 0

δmax
RL < Rt - τ <RU
Rt−τ ≤ RL
Rt−τ ≥ RU
(3)
13
288
J Plant Res (2015) 128:283–294
In this study, the response variable (yt) is NDVI; yt−1
is the antecedent value of this NDVI; ymax is the maximum value of NDVI; δt is the magnitude of the increase
in NDVI; δ*t is the potential increase; δmax is the maximum
potential increase; RL is the lower threshold and RU is the
upper threshold of rainfall; Rt−τ is the effective rainfall
event; τ is the time lag; t is the response time; and k is the
reduction rate.
We estimated rainfall thresholds using regression models of rainfall size and NDVI increments induced by different-sized rainfall events. NDVIs following the response
were selected as the maximum response values because
the response lasted for a significant period of time. When
NDVI increments increased or decreased negligibly, the
corresponding rainfall size was used as the upper threshold.
The other parameters of threshold delay were determined
by multiple regression based on Eqs. 1, 2, and 3 (Zeppel
et al. 2007; Zhao and Liu 2010).
Results
Relationship between growing season NDVIINT
and precipitation in different periods
Current warm-season precipitation explained 43, 15 and
29 % of the variation in the growing season NDVIINT at
SD, LZ and JQ, respectively. Hydrological annual precipitation significantly increased the proportion of the
explained growing season NDVIINT to 68, 46 and 68 %,
at the three sites, respectively (Fig. 2). These results indicated that the previous autumn and winter precipitation
cannot be ignored in explaining the inter-annual variation
of the growing season NDVIINT. In addition, the rain-use
efficiency (RUE, growing season NDVIINT/precipitation)
both that in the warm season and the hydrological annual
value, did not differ significantly across the three sites
(Fig. 2a, F2, 36 = 2.034 P = 0.146; Fig. 2b, F2, 36 = 0.692
P = 0.507).
The relationship between the growing season (early- and
late-) NDVIINT and the two precipitation seasons (previous
cool and warm) was expressed as a linear function within
our three sites (Table 2). However, large within-site differences were observed. The relationship between the growing season NDVIINT and the warm-season precipitation
at SD was stronger (rw = warm2 = 0.45) than that with the
cool-season precipitation (rc = cool2 = 0.32). Also, the lategrowing-season NDVIINT was correlated with the warm(rw2 = 0.51) rather than with the cool-season precipitation
(rc2 = 0.18). In contrast, both total- and late-growing-season NDVIINT were more highly correlated with cool- than
with warm-season precipitation at LZ, with rw2 = 0.17 and
rc2 = 0.29 (growing-season) and rw2 = 0.15 and rc2 = 0.27
13
Fig. 2 Relationships between precipitation and NDVIINT in desert
ecosystems at SD, LZ, and JQ; the relationships were significant
a Relationship between growing season precipitation (accumulated precipitation from March to August) and NDVIINT: R2 = 0.65,
P < 0.001; b Relationship between annual precipitation (accumulated
precipitation from previous September to current August) and NDVIINT. The slopes were not significantly different among the three sites
(late-season). At JQ, the variability of growing-season
NDVIINT was equally correlated with the cool- (rc2 = 0.33)
and warm-season precipitations (rw2 = 0.32). In addition,
the early-growing-season NDVIINT was much more correlated with the cool-season precipitation than with spring
precipitation, at all sites. Spring precipitation showed some
response relationship to vegetation growth over the early
growing season, but this relationship was not significant.
Previous NDVIINT significantly increased the proportion
of the explained variance of the model considering only
the precipitation of cool and warm seasons, at LZ and JQ,
but not at SD (Table 3). At LZ, between 25 and 36 % of
the growing season (early- and late-) NDVIINT inter-annual
variation was associated with the previous NDVIINT. In
contrast, 27 % of early-growing-season NDVIINT interannual variation was related to the previous NDVIINT, but
growing- and late-season NDVIINT values were not significant at JQ. These results indicated that the response of current NDVIINT to previous-year production differed across
these three sites.
J Plant Res (2015) 128:283–294
289
Table 2 Regression statistics for the relationships between the entire- (May–September), early- (May–June), and late-growing-season (July–
September) NDVIINT, and precipitation at different times of the year in three desert ecosystems
Growing season
Early
Site
Formula
R2
r2
SD
pt = 1.337 + 0.007xc + 0.006xw
R2 = 0.77
rc2 = 0.32; rw2 = 0.45
LZ
pt = 1.037 + 0.003xc + 0.003xw
R = 0.46
JQ
pt = 1.071 + 0.003xc + 0.002xw
R2 = 0.65
SD
LZ
pt = 0.065 + 0.002xc + 0.002xs
pt = 0.492 + 0.001xc + 0.001xs
rc2 = 0.29; rw2 = 0.17
rc2 = 0.33; rw2 = 0.32
2
rc2 = 0.34; rs2 = 0.06
2
rc2 = 0.23; rs2 = 0.04
2
R = 0.40
R = 0.26
SD
pt = 0.798 + 0.004xc + 0.005xw
R = 0.67
rc2 = 0.26; rs2 = −0.01
rc2 = 0.18; rw2 = 0.51
LZ
pt = 0.604 + 0.002xc + 0.002xw
R2 = 0.42
rc2 = 0.22; rw2 = 0.20
JQ
Late
2
JQ
pt = 0.457 + 0.001xc + 0.0004xs
pt = 0.662 + 0.002xc + 0.001xw
R = 0.25
2
2
R = 0.64
rc2 = 0.33; rw2 = 0.31
Multiple regressions were performed with cool and warm season precipitation as predictors of ANPP. Both partial (r2) and overall model (R2)
coefficients are shown. x, cumulative precipitation; the subscripts c, w, and s represent cool season, warm season, and spring season, respectively. pt represents current year growing season NDVIINT. Bold font represents significant level (P < 0.01)
Table 3 Regression statistics for the relationships between the entire growing season NDVIINT (early, late) and hydrological annual precipitation and the previous NDVIINT in three desert ecosystems
Growing season
Early
Late
Site
Formula
R2
r2
SD
pt = 1.349 + 0.006xh + 0.082 pt−1
R2 = 0.65
LZ
pt = 0.384 + 0.003xh + 0.527 pt−1
R = 0.73
rh2 = 0.72; rt−12 = –0.08
rh2 = 0.40; rt−12 = 0.34
JQ
pt = 0.840 + 0.003xh + 0.200 pt−1
R2 = 0.71
rh2 = 0.65; rt−12 = 0.07
2
2
SD
pt = 0.289 + 0.003xc + 0.135 pt−1
R = 0.44
LZ
pt = 0.245 + 0.001xc + 0.192 pt−1
R2 = 0.47
rh2 = 0.48; rt−12 = –0.04
rh2 = 0.20; rt−12 = 0.27
JQ
pt = 0.132 + 0.001xc + 0.245 pt−1
R2 = 0.53
rh2 = 0.30; rt−12 = 0.23
2
LZ
pt = 0.140 + 0.002xh + 0.378 pt−1
R = 0.70
rh2 = 0.64; rt−12 = –0.07
rh2 = 0.34; rt−12 = 0.36
JQ
pt = 0.577 + 0.001xh + 0.075 pt−1
R2 = 0.72
rh2 = 0.69; rt−12 = 0.02
SD
pt = 1.009 + 0.005xh– 0.025 pt−1
R = 0.57
2
Multiple regressions were performed with hydrological year precipitation and previous year ANPP as predictors of current year ANPP. Both
partial (r2) and overall model (R2) coefficients are shown. xh, hydrological year precipitation; pt, the NDVI integrated over the current growing
season; pt−1, the NDVI integrated over the previous growing season; xc, cool season precipitation. Bold font indicates significant level (P < 0.01)
Response relationship of NDVI to precipitation pulse
Increments of NDVI could be a direct response of the
ANPP to the size of rainfall events (Fig. 3). Increments of
NDVI differed across the three ecosystems, as did the relationships between NDVI and rainfall event size classes, and
the SD site maintained higher overall rates of NDVI change
than the other two ecosystems. At all three sites, increments in NDVI could be expressed as a sigmoidal function
of rainfall event size classes. The minimum rainfall thresholds in the three ecosystems were similar, at about 5 mm;
however, there were large differences among the maximum
rainfall thresholds, which increased with increasing MAP,
and the maximum rainfalls for SD, LZ, and JQ were about
55, 35, and 30 mm, respectively (Fig. 3).
The values of k were different across the three sites,
with the lowest at SD, and the highest at JQ (Table 4),
indicating that the duration of desert habitat response to a
precipitation pulse decreased from SD to JQ. The magnitudes of the increase in response (δt) and potential response
(δ*t ) values were highest at SD, and lowest at JQ, indicating that the NDVI response values at SD were significantly
greater than those at JQ after the same precipitation pulses.
The maximum potential response variable values (ymax)
and, correspondingly, the NDVI values, were highest at SD
(0.4826), and lowest at LZ (0.1491).
Discussion
The rainfall-productivity relationship at the ecosystem
scale differed from that at the temporal scale. In this study,
we analyzed productivity responses to both individual precipitation events, and cumulative precipitation at specific
temporal scales. Our results highlighted the importance of
variability in precipitation to desert ecosystems.
13
290
J Plant Res (2015) 128:283–294
Fig. 3 Relationships between NDVI increments and rainfall event
sizes at three desert ecosystems in northwestern China. JQ: n = 30;
LZ: n = 33; SD: n = 37; here, the NDVI increment is the maximum
response to a rainfall event
Responses of the three desert ecosystems’ ANPP to cool
and warm season precipitation
Our results indicated that early-growing-season NDVIINT
was related mainly to the cool-season precipitation, which
accounted for 25–40 % at the three sites (Table 2), but was
not significantly affected by spring precipitation. One possible reason for this result is that increased temperature
leads to enhanced evaporation and hence depletion of soil
moisture, in spring time in arid regions (Yang et al. 2011;
Zeppel et al. 2014). Thus, temperature might account for
some of the unexplained variation of NDVIINT. However,
previous cool-season precipitation could be related to the
carryover effects of soil water from the melting of frozen
water in the preceding autumn, and in the snowpack accumulated over the winter. Our results were consistent with
those from previous research in that previous cool-season
precipitation was the critical resource needed for springtime plant activity (Hamerlynck et al. 2013; Hamerlynck
et al. 2012a; Reichmann et al. 2013). In addition, previous growing-season production accounted for 23–27 % of
inter-annual variation in early-growing-season NDVIINT at
LZ and JQ (Table 4). These results may be explained by
a carry-over effect of the structure of the vegetation, ranging from canopy cover and plant density to species composition, which determined the density of meristems where
plant growth occurs (Yahdjian and Sala 2006). This result
agreed with the observation that plant density is the main
characteristic of vegetation structure that constrains primary productivity in the Patagonian steppe when a wet year
follows a dry year (Yahdjian and Sala 2006). It is also similar to that of the spring response of plant physiology and
production, both of which would amplify their response to
summer rainfall (Heisler-White et al. 2008; Muldavin et al.
2008).
Our results also indicated that variation in NDVIINT
during the growing season responded to both cool-season
(fall/winter) and warm-season (spring/summer) precipitation at the scale of years; this response was observed in
all three desert ecosystems studied (Table 2; Fig. 2), and
confirmed long-term data for many grassland sites (Ogle
and Reynolds 2004; Reichmann et al. 2013). However, our
three sites differed markedly in the sensitivity of ANPP to
precipitation changes in different seasons; thus, NDVIINT
was largely correlated with warm-season (spring/summer)
precipitation at SD, highly dependent on autumn/winter
precipitation and previous production at LZ, and equally
correlated with cool- and warm-season precipitation at JQ.
The result at LZ supported the conclusion that production
Table 4 Parameters of the threshold-delay model for changes in NDVI in response to rainfall pulses at SD, LZ, and JQ
Desert
RL (mm)
RU (mm)
k
δt
ymax
δ*t
δmax
SD
LZ
5.0
5.0
55
35
0.9484
0.9564
0.0249
0.0065
0.4826
0.1491
0.0337
0.0139
0.1369
0.0670
JQ
4.0
30
0.9751
0.0048
0.1927
0.0072
0.0205
RL and RU are the lower and maximum thresholds of rainfall; k is the reduction rate; ymax is the maximum response variable value; δt is the
response increase; and δt* is the potential response increase
13
J Plant Res (2015) 128:283–294
was sensitive to changes in cool-season precipitation, but
only slightly sensitive to changes in warm season precipitation at arid and semiarid steppes and semi-desert regions in
Patagonia (Milchunas et al. 1994; Oesterheld et al. 2001).
In contrast, it has been reported that January-July precipitation was mainly responsible for the fluctuations of community biomass production at a semiarid temperate steppe site
in Inner Mongolia, in China (Bai et al. 2004; Zeppel et al.
2014). These contrasting results are likely to be a consequence of a complex interplay among plant-available water,
plant functional type, and resultant influences on plant phenology, growth, and water relations (Zeppel et al. 2014).
The differing sensitivities of ANPP to changes in cooland warm-season precipitation at our three sites may also
be partly explained by vegetation structure (Byrne et al.
2013). In our study, the SD desert site was dominated by
Asterothamnus alyssoides and Salsola passerina; 60 % of
roots of A. alyssoides and 70 % of roots of S. passerina
occurred in the top 0–10 cm (Wang et al. 2013). In August,
the ground was also covered with many annual plants, all
of which are adapted to efficiently utilize episodic summer
rains with the use of shallow root systems and the ability to
rapidly respond with high rates of photosynthesis following rainfall (Nobel 1980). However, vegetation at LZ and
JQ is dominated by Reaumuria soongorica, a plant with
40 % of roots occurring in the top 0–10 cm of soil (Wang
et al. 2013); the plant can respond to moisture in shallow
soil in the early growing season, and that in deep soil in
the late growing season (Zhou et al. 2013). The relatively
deep soil moisture carryover from previous precipitation
provided vegetation with an ability to mediate the growing season ANPP response (Reichmann et al. 2013; Robertson et al. 2009; Thomey et al. 2011). Moreover, previous research at the LZ desert indicated that NDVI for sites
dominated by with Reaumuria soongorica responded to
summer rainfall events of <30 mm, and had no significant
increase (Fig. 3b). At JQ, vegetation is dominated by Reaumuria soongorica (40 %, 0–10 cm) and Salsola passerina
with shallower roots, with 70 % of roots occurring in the
top 0–10 cm (Wang et al. 2013), and a larger response to
cool-season and warm-season precipitation. This explains
why the growing season NDVIINT responded equally to the
cool- and warm-season precipitations.
In addition, the growing season ANPP was closely
related to the previous production at LZ; this can be
explained with an increase in resource exploration following two or more consecutive wet years, and attributed
to biotic mechanisms mediating the ecosystem response
to antecedent environmental conditions (Reichmann et al.
2013). The lag in response of growing season ANPP to
previous precipitation was pronounced at LZ compared
with other two sites (Oesterheld et al. 2001); additionally,
flat topography at LZ allowed deeper infiltration and more
291
water storage (Zeppel et al. 2014). Our results were similar
to those from the Chihuahuan Desert where tiller density
accounted for an extra 13 % of variability in ANPP (Reichmann et al. 2013). These results indicate that the correlation of NDVI and precipitation was modified (buffered or
amplified) by vegetation structure (Fabricante et al. 2009;
Milchunas et al. 1994; Yang et al. 2011).
However, Yang et al. (2011) revealed that climate and
soil factors play important roles in regulating plant community structure and composition in response to precipitation fluctuation in arid-semiarid ecosystems. For example,
warming may indirectly affect species richness and production of grass species by altering soil water availability,
which would, in turn, buffer the response of precipitation
to ANPP (Fatichi and Ivanov 2014; Yang et al. 2011). Also,
the projected change in timing of snowfall, in combination with rising temperatures, may lead to altered snowmelt and modify the seasonal availability of soil water;
these changes would affect plant growth (Yang et al. 2011).
In addition, the effects that soil moisture from redistributed precipitation may have on plant processes is likely to
depend upon soil properties, microclimate, plant functional
types, and the nature of the change in timing (Yang et al.
2011; Zeppel et al. 2014). Moreover, variation in soil water
content and soil characteristics often have a substantial
effect on plant production and soil processes, with increasingly large fluctuations in soil water content following
redistributed precipitation driving physiological responses,
community structure, and the distribution of plant functional types (Zeppel et al. 2014). However, variation in soil
moisture storage, carbon and water fluxes among species
and precipitation regime in the early growing season were
minor compared to the variation observed for these factors
following late-growing-season precipitation (Potts et al.
2006). Therefore these factors should be incorporated into
future research in models predicting terrestrial vegetation
dynamics under precipitation fluctuations.
Responses of the three desert ecosystems’ ANPP to rainfall
pulses
A precipitation event of >5 mm has a pronounced effect
on NDVI of both desert ecosystems and grasslands, independent of the actual precipitation amount (P < 0.001)
(Heisler-White et al. 2009; Jankju 2008; Li et al. 2013;
Swemmer et al. 2007). This rainfall threshold represents
an ecologically significant rainfall event that interacts with
plant water-use patterns of utilizing soil moisture pulses at
particular infiltration depths or durations (Schwinning and
Sala 2004; Zhao and Liu 2011). For example, in the North
American short-grass steppe, rain events as small as 5 mm
improved water conditions and increased soil water potential (Loik 2007; Sala and Lauenroth 1982). Furthermore,
13
292
for the desert shrubs Nitraria sphaerocarpa and Elaeagnus
angustifolia, the lower rainfall thresholds were 5 mm for
sap flow response (Zhao and Liu 2010) as an indicator
of the potential for shrub growth and water use; by comparison, in a temperate Australian woodland, the threshold needed for rain events to elicit an increase in sap flow
exceeded 20 mm (Zeppel et al. 2007). Therefore, rainfall
events of 5 mm appear to be ecologically significant for
ANPP responses in desert ecosystems.
However, large rainfall events resulted in significantly
higher pulse-response soil moisture in the top 16 cm of
soil, extended the duration of the water pulse for up to
6 days longer (Thomey et al. 2011), and also resulted in
a higher plant growth response. Our results indicate that
NDVI responses were increasing in response to precipitation events >5 mm and less than the maximum threshold.
Compared with a 5 mm rain event, shrub species with deep
roots would continue to take up water from deeper soil after
a large event (Reynolds et al. 2004; Schwinning and Sala
2004); this could trigger a large productivity increase, but
when precipitation reached an upper threshold, the response
would not increase. In this study, these upper thresholds
were 55, 35, and 30 mm, for SD, LZ, and JQ, respectively.
The effective response of plants to water inputs is optimum
at moderate event size, while precipitation at larger events
is less effective and therefore reduces RUE (Huxman et al.
2004b) and vegetation production (Swemmer et al. 2007).
Compared to previous research in grasslands, this value is
much higher than the one for a short-grass steppe, in which
precipitation events of 15–30 mm contributed most of the
effect on ANPP (Lauenroth and Sala 1992). The difference between these ecosystems was most likely the result
of higher water demand in the desert ecosystem, caused by
lower holding capacities and hydraulic conductivity, higher
evaporation rates, and different soil moisture dynamics
(Muldavin et al. 2008). In addition, biodiversity structure,
life forms, and the ability of root systems to exploit moisture at varying depths differed greatly between these two
types of ecosystems (Huxman et al. 2004b).
We used the threshold-delay model to analyze desert
ecosystem responses to precipitation pulses, and ultimately estimated several model parameters that quantitatively demonstrated the response process. The parameter
k represented complex interactions among root profiles,
density, structure and morphology, and soil water dynamics (Ogle and Reynolds 2004). In this study, the three
desert ecosystems exhibited different k values in the order
JQ > LZ > SD, indicating that the durations of response in
JQ and LZ were greater than that in SD (Table 3) due to
relatively deep and low-density root systems of the dominant species Reaumuria soongorica at JQ and LZ. In contrast, the mainly horizontal root system of Asterothamnus
alyssoides and Salsola passerina, the dominant species in
13
J Plant Res (2015) 128:283–294
SD, combined with an abundance of annual plants in the
late growing season, resulted in a smaller k value. Smaller
k means a large and rapid response to each pulse event
that rapidly diminishes thereafter. This is consistent with
smaller k values of annual plants and shallowly-rooted
woody plants reported by others (Ogle and Reynolds 2004;
Zeppel et al. 2007).
The parameter ymax reflects differences in physiology
and growth strategies (Ogle and Reynolds 2004). The maximum values of a potential response variable (ymax) in this
study were highest at SD, lowest at LZ, and intermediate at
JQ, indicating that the physiology and ANPP response to
precipitation were much higher at SD than at the other two
sites; this result was due partly to plant diversity, in particular of annual plants, which responded during the growing
season and with root systems better able to exploit available moisture. Also, the structure of the vegetation, ranging
from plant density to species composition, determines the
density of meristems where plant growth occurs, and may
result in fluctuations in production corresponding to fluctuations in precipitation (Yahdjian and Sala 2006). Second in
ANPP response level was the JQ desert; due to the mixture
of C3 and C4 plants, Reaumuria soongorica and Salsola
passerina, exhibiting a greater response than other desert
plant species to precipitation (Throop et al. 2012).
Our study demonstrated a direct NDVI response to the
precipitation regime. However, we recognize the following limitations: (1) sufficient effective precipitation events
are rare in arid regions, and differences in the June–August
response varied with environmental variables (Li et al.
2013); (2) the selected rainfall events occurred mainly in
summer, but to some extent, fall, winter, and spring precipitations will affect vegetation response to summer rainfall,
and these interactions were difficult to determine in this
study. However, our results highlight the inherent complexity, as well as the importance to desert ecosystems, of precipitation variability at different temporal scales.
Conclusions
In this study, we focused on understanding how desert ecosystems responded to variations in previous and current
precipitation and rainfall events along a rainfall-productivity gradient. Our results showed that current-year production was positively related with hydrological annual precipitation (R2 = 0.77, 0.46 and 0.65) in SD, LZ, and JQ,
of which previous cool-season precipitation explained 32,
29, and 33 % of variation, respectively. Furthermore, NDVI
response thresholds of minimum and maximum rainfall
events were 5 and 30–55 mm respectively, in these sites.
These results showed that previous precipitation and large
rainfall events significantly enhance the growth of desert
J Plant Res (2015) 128:283–294
vegetation, which should be much more adaptive under
expected future scenarios of increasing autumn/winter
precipitation and large rainfall events. At the same time,
these results highlight the inherent complexity in predicting
desert ecosystem responses to fluctuations in precipitation.
Our study provides a better model for revealing the
nature of the temporal variation of growing season (earlyand late-) ANPP at three desert sites in northwestern China,
providing information to develop early warnings of ANPP
anomalies that could appear several months in advance
or even earlier. Thus, managers may assess the potential
response of desert vegetation to fluctuations in precipitation, and, further, make decisions related to livestock grazing or wildlife management.
Acknowledgments This study was supported by the National Science Foundation for Distinguished Young Scholars of China (Grant
No. 41125002) and the general program of the National Natural Science Foundation of China (Grant No. 41071019). We thank Dr. Kathryn Piatek and Dr. Marian Rhys for language assistance. We thank
the anonymous reviewers for their constructive comments, which
improved the overall quality of the manuscript.
References
Bai YF, Han XG, Wu JG, Chen ZZ, Li LH (2004) Ecosystem stability and compensatory effects in the inner Mongolia grassland.
Nature 431:181–184
Berger KA, Wang Y, Mather TN (2013) MODIS-derived land surface
moisture conditions for monitoring blacklegged tick habitat in
southern New England. Int J Remote Sen 34:73–85
Blackman CJ, Brodribb TJ, Jordan GJ (2009) Leaf hydraulics and
drought stress: response, recovery and survivorship in four
woody temperate plant species. Plant Cell Environ 32:1584–1595
Bradford JB, Lauenroth WK, Burke IC, Paruelo JM (2006) The influence of climate, soils, weather, and land use on primary production and biomass seasonality in the US Great Plains. Ecosystems
9:934–950
Brodribb TJ, Bowman DMJS, Nichols S, Delzon S, Burlett R (2010)
Xylem function and growth rate interact to determine recovery rates after exposure to extreme water deficit. New Phytol
188:533–542
Byrne KM, Lauenroth WK, Adler PB (2013) Contrasting effects of
precipitation manipulations on production in two sites within the
central grassland region, USA. Ecosystems 16:1039–1051
Camberlin P, Martiny N, Philippon N, Richard Y (2007) Determinants
of the interannual relationships between remote sensed photosynthetic activity and rainfall in tropical Africa. Remote Sens Environ 106:199–216
Emmerich WE, Verdugo CL (2008) Precipitation thresholds for CO2
uptake in grass and shrub plant communities on walnut gulch
experimental watershed. Water Resour Res 44:5
Fabricante I, Oesterheld M, Paruelo JM (2009) Annual and seasonal
variation of NDVI explained by current and previous precipitation across Northern Patagonia. J Arid Environ 73:745–753
Fatichi S, Ivanov VY (2014) Interannual variability of evapotranspiration and vegetation productivity. Water Resour Res 50:3275–3294
Fay PA, Carlisle JD, Knapp AK, Blair JM, Collins SL (2003) Productivity responses to altered rainfall patterns in a C4-dominated
grassland. Oecologia 137:245–251
293
Fay PA, Kaufman DM, Nippert JB, Carlisle JD, Harper CW (2008)
Changes in grassland ecosystem function due to extreme rainfall events: implications for responses to climate change. Global
Change Biol 14:1600–1608
Gamon JA, Huemmrich KF, Stone RS, Tweedie CE (2013) Spatial
and temporal variation in primary productivity (NDVI) of coastal
Alaskan tundra: decreased vegetation growth following earlier
snowmelt. Remote Sens Environ 129:144–153
Gao Q, Reynolds JF (2003) Historical shrub-grass transitions in the
northern Chihuahuan Desert: modeling the effects of shifting
rainfall seasonality and event size over a landscape gradient.
Global Change Biol 9:1475–1493
Hamerlynck EP, Scott RL, Barron-Gafford GA, Cavanaugh ML,
Moran M, Huxman TE (2012a) Cool-season whole-plant gas
exchange of exotic and native semiarid bunchgrasses. Plant Ecol
213:1229–1239
Hamerlynck EP, Scott RL, Stone JJ (2012b) Soil moisture and ecosystem function responses of desert grassland varying in vegetative cover to a saturating precipitation pulse. Ecohydrology
5:297–305
Hamerlynck E, Scott R, Barron-Gafford GA (2013) Consequences
of cool-season drought-induced plant mortality to Chihuahuan
Desert grassland ecosystem and soil respiration dynamics. Ecosystems 16:1178–1191
Heisler-White JL, Knapp AK, Kelly EF (2008) Increasing precipitation event size increases aboveground net primary productivity in
a semi-arid grassland. Oecologia 158:129–140
Heisler-White JL, Blair JM, Kelly EF, Harmoney K, Knapp AK
(2009) Contingent productivity responses to more extreme rainfall regimes across a grassland biome. Global Change Biol
15:2894–2904
Horion S, Cornet Y, Erpicum M, Tychon B (2013) Studying interactions between climate variability and vegetation dynamic using a
phenology based approach. Int J Appl Earth Obs 20:20–32
Huxman TE, Cable JM, Ignace DD, Eilts JA, English NB, Weltzin J,
Williams DG (2004a) Response of net ecosystem gas exchange to
a simulated precipitation pulse in a semi-arid grassland: the role
of native versus non-native grasses and soil texture. Oecologia
141:295–305
Huxman TE et al (2004b) Precipitation pulses and carbon fluxes in
semiarid and arid ecosystems. Oecologia 141:254–268
IPCC (2007) Climate change 2007: the physical science basis. Contribu-tion of Working Group 1 to the fourth assessment report
of the Intergovernmental Panel on Climate Change. Cambridge,
United Kingdom and New York, Cambrid ge University Press
Jankju M (2008) Individual performances and the interaction between
arid land plants affected by the growth season water pulses. Arid
Land Res Manag 22:123–133
Jenerette GD, Scott RL, Huxman TE (2008) Whole ecosystem metabolic pulses following precipitation events. Funct Ecol 22:924–930
Knapp AK et al (2002) Rainfall variability, carbon cycling, and plant
species diversity in a mesic grassland. Science 298:2202–2205
Lauenroth WK, Sala OE (1992) Long-term forage production of
north-American shortgrass steppe. Ecol Appl 2:397–403
Li F, Zhao W, Liu H (2013) The response of aboveground net primary
productivity of desert vegetation to rainfall pulse in the temperate
desert region of northwest china. PLoS One 8:e73003
Loik ME (2007) Sensitivity of water relations and photosynthesis to
summer precipitation pulses for Artemisia tridentata and Purshia
tridentata. Plant Ecol 191:95–108
Milchunas DG, Forwood JR, Lauenroth WK (1994) Productivity of
long-term grazing treatments in response to seasonal precipitation. J Range Manage 47:133–139
Muldavin EH, Moore DI, Collins SL, Wetherill KR, Lightfoot DC
(2008) Aboveground net primary production dynamics in a northern Chihuahuan Desert ecosystem. Oecologia 155:123–132
13
294
Nobel PS (1980) Water-vapor conductance and co2 uptake for leaves
of a C4 desert grass, Hilaria–Rigida. Ecology 61:252–258
Oesterheld M, Loreti J, Semmartin M, Sala OE (2001) Inter-annual
variation in primary production of a semi-arid grassland related
to previous-year production. J Veg Sci 12:137–142
Ogle K, Reynolds JF (2004) Plant responses to precipitation in desert
ecosystems: integrating functional types, pulses, thresholds, and
delays. Oecologia 141:282–294
Ospina S, Rusch GM, Pezo D, Casanoves F, Sinclair FL (2012) More
stable productivity of semi natural grasslands than sown pastures
in a seasonally dry climate. PLoS ONE 7:e35555
Paruelo JM, Epstein HE, Lauenroth WK, Burke IC (1997) ANPP estimates from NDVI for the Central Grassland Region of the United
States. Ecology 78:953–958
Plaut JA, Wadsworth WD, Pangle R, Yepez EA, McDowell NG, Pockman WT (2013) Reduced transpiration response to precipitation
pulses precedes mortality in a piñon-juniper woodland subject to
prolonged drought. New Phytol 200:375–387
Potts DL et al (2006) Antecedent moisture and seasonal precipitation influence the response of anopy-scale carbon and water
exchange to rainfall pulses in a semi-arid grassland. New Phytol
170:849–860
Reichmann LG, Sala OE, Peters DPC (2013) Precipitation legacies in
desert grassland primary production occur through previous-year
tiller density. Ecology 94:435–443
Resco V, Ewers BE, Sun W, Huxman TE, Weltzin JF, Williams DG
(2009) Drought-induced hydraulic limitations constrain leaf gas
exchange recovery after precipitation pulses in the C-3 woody
legume, Prosopis velutina. New Phytol 181:672–682
Reynolds JF, Kemp PR, Ogle K, Fernandez RJ (2004) Modifying
the ‘pulse-reserve’ paradigm for deserts of North America: precipitation pulses, soil water, and plant responses. Oecologia
141:194–210
Robertson TR, Bell CW, Zak JC, Tissue DT (2009) Precipitation timing and magnitude differentially affect aboveground annual net
primary productivity in three perennial species in a Chihuahuan
Desert grassland. The New phytol 181:230–242
Robinson TMP, La Pierre KJ, Vadeboncoeur MA, Byrne KM, Thomey
ML, Colby SE (2013) Seasonal, not annual precipitation drives
community productivity across ecosystems. Oikos 122:727–738
Roca AL et al (2004) Mesozoic origin for West Indian insectivores.
Nature 429:649–651
Sala OE, Lauenroth WK (1982) Small rainfall events: an ecological
role in semiarid regions. Oecologia 53:301–304
Sala OE, Gherardi LA, Reichmann L, Jobbagy E, Peters D (2012)
Legacies of precipitation fluctuations on primary production: theory and data synthesis. Philos T Roy Soc B 367:3135–3144
Schwinning S, Sala OE (2004) Hierarchy of responses to resource
pulses in and and semi-arid ecosystems. Oecologia 141:211–220
Shafran-Nathan R, Svoray T, Perevolotsky A (2012) The resilience of
annual vegetation primary production subjected to different climate change scenarios. Clim Change 118:227–243
Sponseller RA, Hall SJ, Huber DP, Grimm NB, Kaye JP, Clark CM,
Collins SL (2012) Variation in monsoon precipitation drives spatial and temporal patterns of Larrea tridentata growth in the Sonoran Desert. Funct Ecol 26:750–758
Svoray T, Karnieli A (2011) Rainfall, topography and primary production relationships in a semiarid ecosystem. Ecohydrology
4:56–66
13
J Plant Res (2015) 128:283–294
Swemmer AM, Knapp AK, Snyman HA (2007) Intra-seasonal precipitation patterns and above-ground productivity in three perennial
grasslands. J Ecol 95:780–788
Thomey ML, Collins SL, Vargas R, Johnson JE, Brown RF, Natvig
DO, Friggens MT (2011) Effect of precipitation variability on net
primary production and soil respiration in a Chihuahuan Desert
grassland. Global Change Biol 17:1505–1515
Throop HL, Reichmann LG, Sala OE, Archer SR (2012) Response
of dominant grass and shrub species to water manipulation: an
ecophysiological basis for shrub invasion in a Chihuahuan Desert
Grassland. Oecologia 169:373–383
Wang H, Zhao WZ (2009) Change of soil physical properties in process of oasisization. J Desert Res 29:1109–1115
Wang J, Rich PM, Price KP (2003) Temporal responses of NDVI to
precipitation and temperature in the central great plains, USA. Int
J Remote Sen 24:2345–2364
Wang M, Su Y, Yang R, Yang X (2013) Allocation patterns of aboveand belowground biomass in desert grassland in the middle
reaches of Heihe River, Gansu Province, China. Acta Phytoecol
Sinica 37:209–219
Xiao JF, Moody A (2004) Photosynthetic activity of US biomes:
responses to the spatial variability and seasonality of precipitation and temperature. Global Change Biol 10:437–451
Yahdjian L, Sala OE (2006) Vegetation structure constrains primary
production response to water availability in the patagonian
steppe. Ecology 87:952–962
Yang H, Wu M, Liu W, Zhang ZHE, Zhang N, Wan S (2011) Community structure and composition in response to climate change in a
temperate steppe. Global Change Biol 17:452–465
Zeppel M, Macinnis-Ng CMO, Ford CR, Eamus D (2007) The
response of sap flow to pulses of rain in a temperate Australian
woodland. Plant Soil 305:121–130
Zeppel MJB, Wilks JV, Lewis JD (2014) Impacts of extreme precipitation and seasonal changes in precipitation on plants. Biogeosciences 11:3083–3093
Zhang B, Zhang H, Zhang K, Zhang MJ, Lin Q, Lu AX, Guo ZG
(2007) Study on spatial diversification of soil moisture content of
oasis and oasis-desert ecotone in the middle reaches of the Heihe
River. Geographical Res 26:321-327Zhang G, Zhang Y, Dong J,
Xiao X (2013a) Green-up dates in the Tibetan Plateau have continuously advanced from 1982 to 2011. PNAS 110:4309–4314
Zhang G, Zhang Y, Dong J, Xiao X (2013a) Green-up dates in the
Tibetan Plateau have continuously advanced from 1982 to 2011.
PNAS 110:4309–4314
Zhang Y et al (2013b) Extreme precipitation patterns and reductions
of terrestrial ecosystem production across biomes. J Geophys Res
118:148–157
Zhao W, Liu B (2010) The response of sap flow in shrubs to rainfall pulses in the desert region of China. Agr Forest Meteorol
150:1297–1306
Zhao W, Liu H (2011) Precipitation pulses and ecosystem responses
in arid and semiarid regions: a review. The J of Appl Ecol
22:243–249
Zhou H, Zheng X-J, Tang L-S, Li Y (2013) Differences and similarities between water sources of Tamarix ramosissima, Nitraria
sibirica and Reaumuria soongorica in the southeastern Junggar
Basin. Chinese J of Plant Ecol 37:665–673