Professional Documents
Culture Documents
Karl Vanderlinden
Harry Vereecken
Horst Hardelauf
Michael Herbst
Gonzalo Mar nez
Michael H. Cosh
Yakov A. Pachepsky*
This paper reviews the current knowledge base on the temporal stability
of soil water contents, highlights the
interac ons of the temporal stability controls, and proposes research
avenues which can advance mul ple
applica ons that the temporal stability of soil water contents currently
has in environmental monitoring,
modeling, and management.
K. Vanderlinden, IFAPA, Centro Las TorresTomejil, 41200 Alcal del Rio, Spain. Harry
Vereecken, H. Hardelauf, and M. Herbst,
Agrosphere (IBG-3), Ins tute of Bio- and
Geosciences, Forschungszentrum Jlich
GmbH, 52428 Jlich, Germany. G. Marnez, Dep. of Agronomy, Univ. of Cordoba,
14071, Cordoba, Spain. M.H. Cosh, USDAARS, Hydrology and Remote Sensing Lab.,
Beltsville, MD, 20705, USA. Y.A. Pachepsky,
USDA-ARS-EMFSL, Beltsville, MD, 20705, USA.
K. Vanderlinden and G. Mar nez, Currently
at USDA-ARS-EMFSL, Beltsville, MD, 20705
USA. Received 23 Nov. 2011. *Corresponding
author (Yakov.pachepsky@ars.usda.gov).
Vadose Zone J.
doi:10.2136/vzj2011.0178
Received 23 Nov. 2011.
The concept of TS in soils was first introduced by Vachaud et al. (1985). They observed
that at specific locations within a field, the field averaged soil water storage was preserved
in time. In addition, they found that locations could be spotted where soil was consistently
wetter or consistently dryer than average across the field leading to a persistent bias. The TS
SWC has also been reported both at scales finer than the field, e.g., within a plot (Rolston
et al., 1991; Pachepsky et al., 2005; Herbst et al., 2009), and at scales coarser than the
watershed, e. g. regional scale (Martnez-Fernndez and Ceballos, 2003). Temporal stability is often called time stability, and other terms such as rank stability or order stability
have been proposed and discussed (Chen, 2006).
Information on TS SWC has found multiple applications in environmental monitoring,
modeling, and management. It has been shown to be useful, for example, in locating time
stable sites viewed as the most representative locations (Vachaud et al., 1985; Kachanoski
and de Jong, 1988; Rolston et al., 1991) and in up- and downscaling soil water contents
(Cosh et al., 2004; Jacobs et al., 2004; Guber et al., 2008; de Rosnay et al., 2009). The TS
SWC was instrumental in developing methods to infi ll missing data from malfunctioning probes (Pachepsky et al., 2005; Dumedah and Coulibaly, 2011). Data on TS found
direct use in hydrologic modeling (Brocca et al., 2009; Heathman et al., 2009) and in
SWC monitoring data assimilation in soil water flow modeling (Pan et al., 2012). Field
management zones were delineated based on TS SWC (da Silva et al., 2001; Starr, 2005).
It was hypothesized that TS SWC information can be useful in improving hydrologic
modeling by accounting for differences in antecedent soil water contents (Gmez-Plaza et
al., 2001; Zehe et al., 2010; Minet et al., 2011), to design adequate treatment experiments
with replicated plots (Kamgar et al., 1993; Reichardt et al., 1993; Cassel et al., 2000; Rocha
et al., 2005) and to design sensor networks and optimize the number of sensors (Mohanty
and Skaggs, 2001; Heathman et al., 2009).
www.VadoseZoneJournal.org
Currently the number of publications on TS SWC exhibits accelerated growth. Yet, the basic questions about TS SWC and its
controls remain unanswered. Moreover, the evidence found in literature with respect to TS SWC controls remains contradictory.
The practically important assumption that an increase in spatial
scale affects the time stability remains untested. An inventory of
existing examples of the TS SWC analysis seems to be in order
to begin answering these questions in a systematic manner. Such
inventory, based on available studies in literature, was the purpose
of this work. Specific objectives were (i) to analyze available literature to identify the key controls affecting TS SWC, (ii) to evaluate
the effect of measurement techniques on TS SWC, (iii) to evaluate
time and scale effects on TS SWC, and (iv) to suggest an identifiable spatiotemporal model of the TS SWC.
6 Quan
fying TS SWC
ij
[1]
1
Nt
j= N t
RDij ,
[2]
j =1
1 j= N t
(RDij MRDi )2 .
N t 1 j =1
[3]
[4]
www.VadoseZoneJournal.org
rS ( j1 , j2 ) = 1
( N s 1) N s ( N s +1)
[5]
[6]
and select the location with smallest RMSE. The mean absolute
bias error, or MABE (Hu et al., 2010b) was also proposed in
the form
MABE i =
j= N t
RDij MRDi
j =1
1 + MRDi
[7]
www.VadoseZoneJournal.org
Mediterranean conditions that approximately 1 yr of measurements was required to determine the representative point, with
MRD closest to zero and smallest SDRD, both at extents of 600
m and 40 km.
[8]
1
Ns
i= N s
MRDi ,
[9]
i =1
1 i= N s
(MRDi MRD
N s 1 i=1
[10]
Non-zero MRD
Several datasets showed MRD values substantially different
from zero. This contradicts the definition of MRD, because
the sum of the MRD is equal to zero at any observation time,
and therefore also MRD has to be zero [Eq. 12 and Eq. 24].
Small deviations of MRD from zero could appear during
the digitizing process due to the limited quality of the graphs.
About 11, 20, and 29% of the cases had |MRD| > 0.05, 0.02,
and 0.01, respectively.
Assuming that no errors have been made in the computation of
MRD in the original papers, the only explanation for non-zero
MRD can be a difference between numbers of observations used
on different observation times due to sensor malfunctioning or other
possible causes of data loss. The effect of the incomplete observation
on MRD is illustrated with the synthetic example of Table 2. The
ideal TS SWC is assumed for three locations. The top part of the
table summarizes the case when the data are complete, for which
MRD = 0. The bottom part of the table shows the case when two
measurementstime 2 at location (1) and time (3) at location 3
are absent. The absence of two observations leads to the non-zero
MRD = 0.025.
www.VadoseZoneJournal.org
Table 1. Descriptive statistics of the available data on temporal stability of soil water content (TS SWC), (see also supplementary information) presented in terms of mean relative differences (MRD), standard deviation of relative difference (SDRD), and measurement-related parameters. Digitized
data from 19 graphs, corresponding to four different papers, were excluded from the analysis.
N
Min.
Max.
Mean
SD
Skewness
138
0.166
0.098
0.022
0.032
1.403
Sample MRD
138
0.052
0.097
0.000147
0.014668
1.828
Fitted SDMRD
138
0.013
0.866
0.213
0.151
0.876
Sample SDMRD
138
0.012
0.778
0.217
0.148
0.744
R2
138
0.841
0.998
0.969
0.030
2.043
KMRD
138
2.009
7.313
0.482
1.149
3.258
SMRD
138
0.948
2.438
0.253
0.536
0.853
SDRD
137
0.010
0.650
0.120
0.090
2.181
SDSDRD
137
0.004
0.387
0.062
0.058
2.656
44000
4869
10006
2.3
Measurement-related parameters
Extent (m)
138
Spacing (m)
131
.5
10000.0
1318.0
3070.9
2.3
Nt (days)
136
87840
3078
14993
5.5
Ns
138
210
45
38
1.4
Period (days)
137
11
1292
234.5
277.1
1.9
Depth (m)
138
0.025
3.500
0.414
0.618
2.9
Th ickness (m)
135
0.040
1.50
0.36
0.36
1.3
R 2:
Fig. 1. Overview of experimental designs in studies of temporal stability of soil water content (TS SWC). Frequency distributions are given for (a) spatial extents of observations defined as is the maximum distance between two sampling locations, (b) the number of sampling locations Ns, (c) duration
of observation periods, (d) depths of the center of the layer in which soil water content (SWC) or soil water storage were recorded, (e) thickness of the
layer in which SWC or soil water storage were recorded, and (f ) position of the measurement layer in soil profiles. Surface: from 0 to 0.2 m, Profile:
from 0 to 0.3 m and deeper, Subsurface: at specific depths, with varying thickness.
www.VadoseZoneJournal.org
Table 2. Synthetic example of the effect of the incomplete measurement set on the relative difference (RD), the mean relative difference (MRD), and
the average mean relative difference MRD.
Measured SWC (cm3cm3)
Time
Loc. 1
Loc. 2
RD
Loc. 3
Loc. 1
Loc. 2
Loc. 3
MRD
Complete measurements
1
0.250
0.310
0.190
0.250
0.000
0.240
0.240
0.280
0.340
0.220
0.280
0.000
0.214
0.214
0.310
0.370
0.250
0.310
0.000
0.194
0.194
0.160
0.220
0.100
0.160
0.000
0.375
0.375
MRD
0.000
0.667
0.667
0.250
0.000
0.240
0.240
0.250
0.310
0.190
ND
0.340
0.220
0.280
ND
0.214
0.214
0.310
0.370
ND
0.340
0.088
0.088
ND
0.160
0.220
0.100
0.160
0.000
0.375
0.375
MRD
0.029
0.229
0.276
0.025
www.VadoseZoneJournal.org
Fig. 3. (a) Histogram of R2 for the Gaussian fit of the mean relative difference (MRD) data and (b)
relationship between R2 and the sample standard deviation of the mean relative difference (SDMRD).
www.VadoseZoneJournal.org
www.VadoseZoneJournal.org
Sensors
Fig. 6. Time dependency of the Spearman rank correlation coefficient
(rs) for three fields in the Little Washita Watershed (Mohanty and
Skaggs, 2001).
www.VadoseZoneJournal.org
Fig. 8. Box and Whisker plots for the sample standard deviation of
the mean relative differences (SDMRD) according to surface (from
0 to 0.2 m), profile (from 0 to 0.3 m and deeper), and subsurface (at
specific depths, with varying thickness) soil water content (SWC)
measurements. The caps at the end of each box indicate the minimum
and maximum, the lower and upper quartiles delimit the box, and the
line in the center of the box is the median. Black dots are outliers, more
than 1.5 times the inter-quartile range larger than the upper quartile.
Different letters represent different averages at p < 0.05.
Spa al Scale
Data were grouped into three extent categories: <200m, 2002000
m and >2000 m. Figure 10 shows box-and-whisker plots of sample
SDMRD for the three categories. Excluding the outlier for the
200- to 2000-m category (Fig. 10), the averages standard errors
of SDMRD were 0.07 0.01, 0.24 0.01, and 0.28 0.02, respectively, being the average of the first category significantly different
(p < 0.001) from the others. This result could be expected since for
www.VadoseZoneJournal.org
Fig. 10. Box and Whisker plots of sample standard deviation of the
mean relative differences (SDMRD) for three categories of extent
(<200, 2002000, >2000 m). The caps at the end of each box indicate
the minimum and maximum, the lower and upper quartiles delimit the
box, and the line in the center of the box is the median. Black dots are
outliers, more than 1.5 times the inter-quartile range larger than the
upper quartile. Different letters represent different averages at p < 0.001.
Fig. 11. Sample standard deviation of mean relative difference (SDMRD) versus extent
for the Walnut Creek watershed, distinguishing between surface (from 0 to 0.2 m), profile
(from 0 to 0.3 m and deeper), and subsurface (at specific depths, with varying thickness)
soil water content (SWC) measurements. Data from Cosh et al., 2004 (1); Jacobs et al.,
2004 (2a2d); Choi and Jacobs, 2007 (3a3j); Choi and Jacobs, 2011 (4a4f ); and Joshi
et al., 2011 (5a5e).
www.VadoseZoneJournal.org
Fig. 12. Sample standard deviation of mean relative difference (SDMRD) versus extent
for the Little Washita watershed, distinguishing between surface (from 0 to 0.2 m),
profile (from 0 to 0.3 m and deeper), and subsurface (at specific depths, with varying
thickness) soil water content (SWC) measurements. Data from Mohanty and Skaggs,
2001 (1a1c); Starks et al., 2006 (2a2j); Heathman et al., 2009 (3a3k); Joshi et al.,
2011 (4a4d).
Li le Washita Watershed
Data on the TS SWC for the Little Washita Watershed are
summarized in Fig. 12. The field and catchment data show an
increasing SDMRD with extent. At 600 m an average SDMRD
of 0.15 was observed whereas at the scale of 35 km the average
value was 0.32. Important variation in the SDMRD was observed
at each of the observation scales (Fig. 12). Mohanty and Skaggs
(2001) presented the TS analysis of surface Theta-probe-measured
SWC at three fields during SGP97, using a regular measurement
grid of 49 points with a spacing and extent of 100 and 600 m,
respectively. The SDMRD for the flat LW13 field (Fig. 12, point
1b) was smaller than the SDMRD for fields LW03 and LW21 (Fig.
12, points 1a and 1c), which showed both a rolling topography and
identical SDMRD. For soil surface measurements at fields LW12,
LW13, LW21, and LW45, during the extremely wet remote sensing Cloud and Land Surface Interaction Campaign (Heathman
et al., 2009), substantially lower SDMRD were obtained (Fig. 12,
points 3a3d). In each field eight points were monitored, with an
approximate spacing of 200 m and an extent of 600 m.
Soil water content was measured by Starks et al. (2006) at 15-cm
depth-intervals down to 60 cm using TDR probes at eight sites,
with an average spacing and extent of approximately 10 and 21 km,
respectively. During SGP97 measurements were repeated 29 times
and during SMEX03 six times. Standard deviation of MRD was
highly consistent between both sampling campaigns and across
depth. The first two topsoil layers (Fig. 12, points 2a and 2e; and
2b and 2f) showed similar, but higher SDMRD than compared
to the deeper layers (Fig. 12, points 2c and 2 g; and 2d and 2h).
Considering the entire profi le (60 cm), almost identical, SDMRD
were obtained for both campaigns (Fig. 12, points 2i and 2j).
Joshi et al. (2011) analyzed the same Theta-probe measured soil surface data as Mohanty and Skaggs (2001), for fields LW03, LW13,
and LW21 (Fig. 12, points 4a, 4b, and 4c, respectively). For field
LW21 also MRD data from the CLASIC2007 experiment were
provided (Fig. 12, point 4d). Except for field LW03, for which the
SDMRD was twice as large as for the other fields, similar SDMRD
were obtained for both studies. Overall, for this watershed, published MRD data showed an increasing SDMRD with extent.
Spa al Density
The spatial density, defi ned as the ratio Ns/Extent, was used to
summarize the data with respect to spacing. Figure 13 demonstrates that SDMRD is most variable at intermediate densities,
roughly from 0.02 to 0.3 m1, and is generally higher for smaller
densities. Standard deviation of MRD appears to be generally less
than 0.2 for increasing density from 0.3 m 1. Datasets with these
larger densities were obtained with a spacing and extent ranging
from 0.5 to 22 m, and from 3 to 164 m, respectively, with Ns ranging from 10 to 108. We hypothesize that for these high densities
and fine scale studies, the effects of non-local controls, including
rainfall, are possibly minimized, and that non-local effects are the
dominant controlling factor for the smallest densities, leading to
increasing SDMRD with decreasing density. At intermediate densities local and non-local controls interplay, leading to a wide range
of possible SDMRD values.
Both the extent and the resolution of the sampling scheme had
an impact on the mean of the distribution of SWC in the work of
Petrone et al. (2004). In terms of the distribution of soil moisture,
the extent of sampling within a grid was not as significant as the
density, or spacing, of the measurements in this work. We note that
it is not known how the spatial configuration of sampling locations
affects the manifestation of TS SWC. The spatial configuration
www.VadoseZoneJournal.org
that vegetation controls affected the time stability of soil moisture in the 0 to 20 cm layer making
it difficult to estimate soil moisture values from
earlier measurements using the time persistence
concept as expressed in Eq. [4]. Manfreda and
Rodrguez-Iturbe (2006) found that the length of
the measurement method affected the variance of
the long-term spatial mean daily SWC, especially for
shallow rooted vegetation. The SDMRD might be
seen as the TS equivalent of this variability.
Soil Proper es
Soil properties are commonly implicated in the existence and extent of TS SWC since the ability of soil
to retain and to transmit water is the obvious reason
for differences in SWC. The effect of soil properties
Fig. 13. Dependence of the sample standard deviation of the mean relative difference
and local terrain attributes on SWC patterns was
(SDMRD) on the spatial density of measurements [number of spatial observations (Ns)/
referred to by Grayson et al. (1997) as local controls.
Extent], distinguishing between surface (from 0 to 0.2 m), profile (from 0 to 0.3 m and
They considered the dry SWC state to be dominated
deeper), and subsurface (at specific depths, with varying thickness) soil water content
(SWC) measurements.
by vertical fluxes, which depend mainly on local soil
properties, while lateral fluxes are minimized due to
was shown to be an important factor in characterizing the spatial
the low hydraulic conductivity. Soil texture was sugvariability of SWC (Manfreda and Rodrguez-Iturbe, 2006).
gested to affect the TS by Vachaud et al. (1985). In their work, two
locations that always showed the largest and smallest water storage
Temporal Scale
had a total percentage of particle size smaller than 20 m by weight
Both measurement frequency and duration of the observation
in the first meter of 60 and 49%, respectively. Field water retention
period can be considered as components of the temporal scale to
was strongly related to the silt+clay content in their work. Better
define the TS. Most campaigns included in the dataset were shortstability, i.e., higher rs, was observed in the sandy loam soils than
term for the purposes of remote sensing validation studies. Several
in silt loam soil in the work of Mohanty and Skaggs (2001). Soil
studies had high temporal frequency provided by sensor networks.
type, as characterized by bulk density, clay and sand content, was
responsible for nearly 50% of spatial variability of MRD in the
Degradation of TS SWC with time was demonstrated by Rolston
work of Cosh et al. (2008). Soil thickness was shown to affect RD
et al. (1991). In their work, single sampling locations identified
and to be a factor of TS SWC by Zhu and Lin (2011).
during 1 yr gave estimates of mean storage during the following
year with some increase in error. However, use of the same sampling
Relatively low variations in soil texture and structure across study
locations for more than 2 yr increased the error in storage estimates
areas was suggested as the reason for the absence of significant difin this work. On the other hand, Schneider et al. (2008) found
ferences between MRD in different locations across the site with
time-stable locations with a low deviation from mean field SWC
volcanic soil (Comegna and Basile, 1994). However, measurements
and low standard deviation for research sites in China. Although
were done under the developing barley crop, and root activity might
the time stability characteristics of some points varied between
decrease the TS manifestation as suggested by Cassel et al. (2000).
years, the selected points were appropriate to predict mean SWC
of the sites for multiple years. The authors question the feasibility
Vegeta on
of finding temporally stable locations with low SDRD when time
Land cover and crop development can affect the TS SWC. Root
series longer than 2 yr are considered. In a large-scale study, Cosh
activity was suggested as the factor weakening the TS as measured
et al. (2006) found for half-hour SWC measurements that TS was
by the Spearman rank correlation coefficient (Kamgar et al., 1993;
maintained throughout the different seasons of a 21-mo measureCassel et al., 2000). Shallow root activity or no root activity was
ment period. In another large-scale study, Martnez-Fernndez
observed in most of the cases when the depth effect on TS was not
and Ceballos (2003) found that TS SWC was maintained, i.e.,
detected. In the work of Pachepsky et al. (2005), a similar degree
points maintain their rank in the MRD curve, throughout a 3-yr
of TS SWC was observed at different depths in a loamy soil withmonitoring period.
out vegetation. Temporal persistence was found to depend on
grazing management and the related plant cover at the field scale
TS SWC can be changed during the growing period due to the
by Schneider et al. (2008). At larger scales, land cover or vegetaeffect of root activity. Gmez-Plaza et al. (2001) used rs to show
tion do not necessarily affect the spatial distribution of soil water
www.VadoseZoneJournal.org
Topography
Topography as a control has been analyzed in the majority of
works on TS SWC. Results have been inconclusive partly because
different topographic descriptors along with different designs of
measurement campaigns were used in different landscapes. The
differences in topographic position were shown to impose the TS
SWC (Tomer and Anderson, 1995; Biswas and Si, 2011c). GmezPlaza et al. (2000) studied the TS SWC at three different transects
with slopes varying between 33 and 41%. They showed that topographic effects or local topography were the mean causes of TS
SWC in a sense that the bias of each location with respect to the
mean SWC value was maintained. Th is can be explained by the
fact that dry and wet positions at the top and bottom of the slope,
respectively, remain stable in time. No substantial effect of topographic variables on TS was found in relatively flat areas (Kaleita
et al., 2004).
The effect of climate and seasonality on TS SWC has to our knowledge not directly been analyzed. Therefore we use SWC dynamics
here as a proxy for climate and seasonality. Earlier studies showed
better time stability for the drier sites in terms of the SDRD and
the RMSE, although this was not confirmed by Choi and Jacobs
(2007). Several studies found that TS SWC is lower during dry
periods, based on a comparison of Spearman rank correlation
for wet and dry periods (Kachanoski and de Jong, 1988; GmezPlaza et al., 2000). Martnez-Fernndez and Ceballos (2003) used
data of the REMEDHUS network to analyze the effect of wet
and dry conditions on the temporal persistence of soil moisture.
They found that dry points are more time-stable than wet points
in terms of SDRD. During wet and dry periods, similar Spearman
rank correlation coefficients between subsequent measurements
were obtained, indicating similar temporal persistence. Drying
periods typically preserve temporal persistence whereas transitional periods from consistently dry to consistently wet periods
may break it (Kachanoski and de Jong 1988). Soil moisture studies
are generally performed in areas where soils are not overwetted,
so the average of the surface is dry which gives a short interval for
moisture at the low end, and more of an interval from the average
to the saturation stage. Overall, this leads to the conclusion that
SWC itself is one of the factors controlling its TS.
The effects of weather, topography, soil properties, and vegetation on TS SWC are probably intertwined and complementary.
Choi and Jacobs (2007) analyzed the effect of soil properties on
SWC variability and noted that the principal component analysis
www.VadoseZoneJournal.org
Z(u , t ) = M (u , t ) + R(u , t )
[11]
[12]
[13]
or as
6 Sta
where p(u) and q(t) are functions of space and time, respectively.
A similar decomposition can be obtained for M(u,t) and R(u,t).
According to Kyriakidis and Journel (1999), the adoption of one
of these decompositions is a modeling decision, rather than a
data-based hypothesis, since measurements of M(u,t), p(u), and
q(t) are generally not available. They suggest basing this decision
on secondary information, using deterministic physically-based
relationships between the trend and the observations. An additive
model of particular interest with respect to TS analysis, reported
by Kyriakidis and Journel (1999), is
m(u , t ) = m + v(u ) + w(t ) ,
Spatiotemporal models have been extensively explored in geostatistical literature. Kyriakidis and Journel (1999) reviewed
geostatistical space-time models for environmental data, including SWC. They recalled that features such as time order-past,
present, and future- and isotropy are only meaningful in the time
and space domain, respectively, while distance units in space and
time are unrelated. Therefore, time cannot be considered as just
an extra dimension and models do generally not account for the
full space-time dependence. One way they suggested to represent
environmental data was by using a nonstationary spatiotemporal
random function model, Z(u, t) , which is decomposed into either
a non-stationary stochastic mean, M(u, t) , or a deterministic mean,
m(u, t) , and a zero-mean stationary residual random function component R(u, t) , which both depend on the location, u, within a
two-dimensional space A, and the moment, t, during a period of
time T:
[15]
where m is the stationary space-time mean and v(u) and w(t) are
smooth functions of u and t for which the following conditions
apply over the Ns spatial locations and the Nt observation times,
Ns
v (u ) = 0
[14]
and
=1
Nt
w ( t ) = 0 ,
[16]
=1
www.VadoseZoneJournal.org
[17]
(u , t ) w(t ) m
v(u )
=
,
w(t ) + m
w(t ) + m
[18]
1
T
v(u )
T RD(u , t )dt = mT T
1
dt
w(t )
1+
m
[19]
1
SDRD(u ) =
T
12
2
2
RD(
u
,
t
)
d
t
MRD(
u
)
[23]
v
u
(
)
1
1
1
R (u , t )
1
1
+
dt du
A A mT T 1 + w(t )
[24]
m
1
1
dt v(u )du
=
w (t ) A
mTA T
1+
m
1
1
+
dt
R ( u , t ) dt du
w (t ) A T
mTA T
1+
m
Both terms at the right-hand side are equal to zero. Thus the mean
value, MRD, across the domain is also zero. The variance of
MRD(u), SDMRD2 is then equal to
SDMRD2 =
1
MRD(u ) 2 du .
A A
[25]
MRD(u )
v(u )
m
[20]
v(u ) + R(u , t )
w (t ) + m
1
T
[21]
T RD(u , t )dt
v(u )
1
1
dt +
mT T 1 + w(t )
mT
m
R (u , t )
dt .
w(t )
1+
m
[22]
SDMRD2 =
1
1
1
v(u )
dt +
A
T
(
)
w
t
A mT
mT
1+
or also
2
v ( u ) + R ( u , t ) dt 2
1
1
2
SDMRD =
dt
du [27]
T
A
w (t )
A
mT
1+
Th is shows that the variance of MRD(u) is determined by a temporal and a spatial component.
If the residual term, R(u,t) is small with respect to v(u), and |w(t)|
<< m we obtain:
v (u ) 2
1
du
SDMRD =
A A m
2
2
R (u , t )
T w(t ) dt du [26]
1+
www.VadoseZoneJournal.org
[28]
This means that if the average SWC remains rather constant during
the observation period, the variance is mainly determined by the
spatial variation of SWC. If the SWC variation is large during
the observation period, the term w(t) influences the variance of
MRD(u). In case that only the residual term is small we find:
v (u ) 2
1
1
du
SDMRD2 =
dt
w (t ) A mT
A T
1+
[29]
The model Eq. [17] is additive as the spatial and temporal effects
are separated. Multiplicative models, of the type shown in Eq. [14]
can be developed in which the deterministic component in SWC
can be obtained by multiplication of the same spatial pattern by
the time dependent component. The EOF modeling of the TS
SWC presents one example of multiplicative models (Perry and
Niemann, 2007; Korres et al., 2010).
1
Ns Nt
v(u ) =
w(t ) =
1
Nt
Ns Nt
(u , t )
[30]
=1=1
Nt
(u , t ) m
[31]
1 Ns
(u , t ) m
N s =1
[32]
=1
Research Avenues
www.VadoseZoneJournal.org
Overall, the results of this review call for a focused research effort
directed towards identifying the interactions and effects of measurement design, topography, soil, vegetation and climate on TS
SWC. More detailed studies are needed to identify the effect of
local and non-local controls on TS SWC. Soil moisture networks
providing spatially and temporally highly resolved soil moisture
data should be combined with detailed on-site characterization.
This may contribute to identifying the major controls and to study
the temporal dynamics of the rank stability curves in more detail.
A matter related to this is the development of better methods for
defining the spatial properties of soils and vegetation as they affect
the TS SWC.
Acknowledgments
This research was made possible thanks to funding provided by the Spanish Ministry
of Science and Innovation and FEDER through grant AGL2009-12936-C03-03 and
by the Junta de Andaluca though grant AGR-4782. Also mobility grants PR20100191 and of EX2009-0433 the Spanish Ministry of Education are acknowledged.
References
Biswas, A., and B.C. Si. 2011a. Depth persistence of the spa al pa ern of soil
water storage in a Hummocky landscape. Soil Sci. Soc. Am. J. 75:10991109.
doi:10.2136/sssaj2010.0399
Biswas, A., and B.C. Si. 2011b. Scales and loca ons of me stability of soil water
storage in a hummocky landscape. J. Hydrol. 408:100112. doi:10.1016/j.
jhydrol.2011.07.027
Biswas, A., and B.C. Si. 2011c. Iden fying scale specific controls of soil water
storage in a hummocky landscape using wavelet coherency. Geoderma
165:5059. doi:10.1016/j.geoderma.2011.07.002
Bosch, D., V. Lakshmi, T. Jackson, M. Choi, and J. Jacobs. 2006. Large scale measurements of soil moisture for valida on of remotely sensed data: Georgia
soil moisture experiment of 2003. J. Hydrol. 323:120137. doi:10.1016/j.
jhydrol.2005.08.024
Brocca, L., F. Melone, T. Moramarco, and R. Morbidelli. 2009. Soil moisture
temporal stability over experimental areas in Central Italy. Geoderma
148:364374. doi:10.1016/j.geoderma.2008.11.004
Brocca, L., F. Melone, T. Moramarco, and R. Morbidelli. 2010. Spa al-temporal
variability of soil moisture and its es ma on across scales. Water Resour.
Res. 46:W02516. doi:10.1029/2009WR008016
Canton, Y., A. Solebenet, and F. Domingo. 2004. Temporal and spa al pa erns
of soil moisture in semiarid badlands of SE Spain. J. Hydrol. 285:199214.
doi:10.1016/j.jhydrol.2003.08.018
Cassel, D.K., O. Wendroth, and D.R. Nielsen. 2000. Assessing spa al variability
in an agricultural experiment sta on field: Opportuni es arising from spaal dependence. Agron. J. 92:706714. doi:10.2134/agronj2000.924706x
Chen, Y. 2006. Le er to the editor on rank stability or temporal stability. Soil
Sci. Soc. Am. J. 70:306. doi:10.2136/sssaj2005.0290l
Choi, M., and J.M. Jacobs. 2007. Soil moisture variability of root zone profiles
within SMEX02 remote sensing footprints. Adv. Water Resour. 30:883896.
doi:10.1016/j.advwatres.2006.07.007
Choi, M., and J.M. Jacobs. 2011. Spa al soil moisture scaling structure during soil moisture experiment 2005. Hydrol. Processes 25(6):926932.
doi:10.1002/hyp.7877
Comegna, V., and A. Basile. 1994. Temporal stability of spa al pa erns of
soil water storage in a cul vated Vesuvian soil. Geoderma 62:299310.
doi:10.1016/0016-7061(94)90042-6
Coppola, A., A. Comegna, G. Dragone , N. Lamaddalena, A. Kader, and V.
Comegna. 2011. Average moisture satura on eects on temporal stability
of soil water spa al distribu on at field scale. Soil Tillage Res. 114:155164.
doi:10.1016/j.s ll.2011.04.009
Cosh, M.H., T.J. Jackson, R. Bindlish, and J.H. Prueger. 2004. Watershed scale
temporal and spa al stability of soil moisture and its role in valida ng
satellite es mates. Remote Sens. Environ. 92:427435. doi:10.1016/j.
rse.2004.02.016
Cosh, M.H., T.J. Jackson, S. Moran, and R. Bindlish. 2008. Temporal persistence
and stability of surface soil moisture in a semi-arid watershed. Remote
Sens. Environ. 112:304313. doi:10.1016/j.rse.2007.07.001
Cosh, M., T. Jackson, P. Starks, and G. Heathman. 2006. Temporal stability of
surface soil moisture in the Li le Washita River watershed and its applicaons in satellite soil moisture product valida on. J. Hydrol. 323:168177.
doi:10.1016/j.jhydrol.2005.08.020
da Silva, A.P., A. Nadler, and B. Kay. 2001. Factors contribu ng to temporal stability in spa al pa erns of water content in the llage zone. Soil Tillage Res.
58:207218. doi:10.1016/S0167-1987(00)00169-0
De Lannoy, G.J.M., P.R. Houser, N.E.C. Verhoest, V.R.N. Pauwels, and T.J. Gish.
2007. Upscaling of point soil moisture measurements to field averages at
the OPE3 test site. J. Hydrol. 343:111. doi:10.1016/j.jhydrol.2007.06.004
de Rosnay, P., C. Gruhier, F. Timouk, F. Baup, E. Mougin, P. Hiernaux, L. Kergoat, and V. LeDantec. 2009. Mul -scale soil moisture measurements at
the Gourma meso-scale site in Mali. J. Hydrol. 375:241252. doi:10.1016/j.
jhydrol.2009.01.015
Dumedah, G., and P. Coulibaly. 2011. Evalua on of sta s cal methods for infilling missing values in high-resolu on soil moisture data. J. Hydrol. 400:95
102. doi:10.1016/j.jhydrol.2011.01.028
Famiglie , J.S., D. Ryu, A.A. Berg, M. Rodell, and T.J. Jackson. 2008. Field observa ons of soil moisture variability across scales. Water Resour. Res. 44:116.
Gmez-Plaza, A., J. Alvarez-Rogel, J. Albaladejo, and V.M. Cas llo. 2000. Spa al
pa erns and temporal stability of soil moisture across a range of scales
in a semi-arid environment. Hydrol. Processes 14:12611277. doi:10.1002/
(SICI)1099-1085(200005)14:7<1261::AID-HYP40>3.0.CO;2-D
Gmez-Plaza, A., M. Mar nez-Mena, J. Albaladejo, and V. Cas llo. 2001. Factors regula ng spa al distribu on of soil water content in small semiarid
catchments. J. Hydrol. 253:211226. doi:10.1016/S0022-1694(01)00483-8
Grant, L., M. Seyfried, and J. McNamara. 2004. Spa al varia on and temporal
stability of soil water in a snow-dominated, mountain catchment. Hydrol.
Processes 18:34933511. doi:10.1002/hyp.5798
Grayson, R.B., and A.W. Western. 1998. Towards areal es ma on of soil water
content from point measurements: Time and space stability of mean response. J. Hydrol. 207:6882. doi:10.1016/S0022-1694(98)00096-1
Grayson, R.B., A.W. Western, F.H.S. Chiew, and G. Blschl. 1997. Preferred
states in spa al soil moisture pa erns: Local and nonlocal controls. Water
Resour. Res. 33:28972908. doi:10.1029/97WR02174
Guber, A.K., T.J. Gish, Y.A. Pachepsky, M.T. van Genuchten, C.S.T. Daughtry, T.J.
Nicholson, and R.E. Cady. 2008. Temporal stability in soil water content
pa erns across agricultural fields. Catena 73:125133. doi:10.1016/j.catena.2007.09.010
Heathman, G.C., M. Larose, M.H. Cosh, and R. Bindlish. 2009. Surface and profile soil moisture spa o-temporal analysis during an excessive rainfall period in the Southern Great Plains, USA. Catena 78:159169. doi:10.1016/j.
catena.2009.04.002
Herbst, M., N. Prolingheuer, A. Graf, J.A. Huisman, L. Weihermller, and J.
Vanderborght. 2009. Characteriza on and understanding of bare soil
respira on spa al variability at plot scale. Vadose Zone J. 8:762771.
doi:10.2136/vzj2008.0068
Hu, W., M.A. Shao, F.P. Han, K. Reichardt, and J. Tan. 2010a. Watershed
scale temporal stability of soil water content. Geoderma 158:181198.
doi:10.1016/j.geoderma.2010.04.030
Hu, W., M.A. Shao, and K. Reichardt. 2010b. Using a new criterion to iden fy
sites for mean soil water storage evalua on. Soil Sci. Soc. Am. J. 74:762
773. doi:10.2136/sssaj2009.0235
Hu, W., M. Shao, Q. Wang, and K. Reichardt. 2009. Time stability of soil water storage measured by neutron probe and the eects of calibra on
procedures in a small watershed. Catena 79:7282. doi:10.1016/j.catena.2009.05.012
Ibrahim, H.M., and D.R. Huggins. 2011. Spa o-temporal pa erns of soil water storage under dryland agriculture at the watershed scale. J. Hydrol.
404:186197. doi:10.1016/j.jhydrol.2011.04.029
Jacobs, J.M., B.P. Mohanty, E. Hsu, and D. Miller. 2004. SMEX02: Field scale variability, me stability and similarity of soil moisture. Remote Sens. Environ.
92:436446. doi:10.1016/j.rse.2004.02.017
Jawson, S.D., and J.D. Niemann. 2007. Spa al pa erns from EOF analysis of
soil moisture at a large scale and their dependence on soil, land-use, and
topographic proper es. Adv. Water Resour. 30:366381. doi:10.1016/j.advwatres.2006.05.006
Joshi, C., B.P. Mohanty, J.M. Jacobs, and A.V.M. Ines. 2011. Spa otemporal analyses of soil moisture from point to footprint scale in two dierent hydroclima c regions. Water Resour. Res. 47:W01508. doi:10.1029/2009WR009002
Kachanoski, R.G., and E. de Jong. 1988. Scale dependence and the temporal
persistence of spa al pa erns of soil water storage. Water Resour. Res.
24:8591. doi:10.1029/WR024i001p00085
Kaleita, A.L., L.F. Tian, and M.C. Hirschi. 2004. Iden fica on of Op mal Sampling Loca ons and Grid Size for Soil Moisture Mapping. ASAE Paper No.
041146. h p://asae.frymul .com/azdez.asp?JID=5&AID=16171&CID=can2
004&T=2 (accessed 13 Sept. 2012).
Kamgar, A., J.W. Hopmans, W.W. Wallender, and O. Wendroth. 1993. Plot size
and sample number for neutron probe measurements in small field trials.
Soil Sci. 156:213224. doi:10.1097/00010694-199310000-00001
Kirda, C., and K. Reichardt. 1992. Comparison of neutron moisture gauges with
nonnuclear methods to measure field soil water status. Sci. Agric. (Piracicaba, Braz.)
49:111121. doi:10.1590/S0103-90161992000400015
Korres, W., C.N. Koyama, P. Fiener, and K. Schneider. 2010. Analysis of surface
soil moisture pa erns in agricultural landscapes using Empirical Orthogo-
www.VadoseZoneJournal.org
Rocha, G.C., P.L. Libardi, L. Alves de Carvalho, and A.C. Rodrigues Cruz. 2005.
Temporal stability of the spa al distribu on of water storage in a soil under citrus cul va on. Rev. Bras. Cienc. Solo 29:4150. doi:10.1590/S010006832005000100005
Rolston, D.E., J.W. Biggar, and H.I. Nigh ngale. 1991. Temporal persistence of
spa al soil-water pa erns under trickle irriga on. Irrig. Sci. 12:181186.
doi:10.1007/BF00190521
Ruiz-Sinoga, J.D., J.F. Mar nez-Murillo, M.A. Gabarrn-Galeote, and R. GarcaMarn. 2011. The eects of soil moisture variability on the vegeta on pattern in Mediterranean abandoned fields (Southern Spain). Catena 85:111.
doi:10.1016/j.catena.2010.11.004
Schneider, K., J. Huisman, L. Breuer, Y. Zhao, and H. Frede. 2008. Temporal
stability of soil moisture in various semi-arid steppe ecosystems and its
applica on in remote sensing. J. Hydrol. 359:1629. doi:10.1016/j.jhydrol.2008.06.016
Starks, P., G. Heathman, T. Jackson, and M. Cosh. 2006. Temporal stability of soil moisture profile. J. Hydrol. 324:400411. doi:10.1016/j.jhydrol.2005.09.024
Starr, G. 2005. Assessing temporal stability and spa al variability of soil water
pa erns with implica ons for precision water management. Agric. Water
Manage. 72:223243. doi:10.1016/j.agwat.2004.09.020
Tallon, L., and B.C. Si. 2003. Representa ve soil water benchmarking for environmental monitoring. J. Environ. Inf. Arch. 1:581590.
Teuling, A.J., and P.A. Troch. 2005. Improved understanding of soil moisture
variability dynamics. Geophys. Res. Le . 32:710.
Teuling, A.J., R. Uijlenhoet, F. Hupet, E.E. van Loon, and P.A. Troch. 2006. Es ma ng spa al mean root-zone soil moisture from point-scale observa ons.
Hydrol. Earth Syst. Sci. 10:755767. doi:10.5194/hess-10-755-2006
Thierfelder, T.K., R.B. Grayson, D. van Rosen, and A.W. Western. 2003. Inferring the loca on of catchment characteris cs soil moisture monitoring
sites. Covariance structure in the temporal domain. J. Hydrol. 280:1332.
doi:10.1016/S0022-1694(03)00077-5
Tomer, M.D., and J.L. Anderson. 1995. Varia on of soil water storage across
a sand plain hill slope. Soil Sci. Soc. Am. J. 59:10911100. doi:10.2136/
sssaj1995.03615995005900040021x
Vachaud, G., A. Passerat De Silans, P. Balabanis, and M. Vauclin. 1985. Temporal
stability of spa ally measured soil water probability density func on. Soil Sci.
Soc. Am. J. 49:822828. doi:10.2136/sssaj1985.03615995004900040006x
Venkatesh, B., N. Lakshman, B.K. Purandara, and V.B. Reddy. 2011. Analysis
of observed soil moisture pa erns under dierent land covers in Western
Ghats, India. J. Hydrol. 397:281294. doi:10.1016/j.jhydrol.2010.12.006
Yoo, C., and S. Kim. 2004. EOF analysis of surface soil moisture field variability.
Adv. Water Resour. 27:831842. doi:10.1016/j.advwatres.2004.04.003
Zehe, E., T. Grae, M. Morgner, A. Bauer, and A. Bronstert. 2010. Plot and field
scale soil moisture dynamics and subsurface wetness control on runo
genera on in a headwater in the Ore Mountains. Hydrol. Earth Syst. Sci.
14:873889. doi:10.5194/hess-14-873-2010
Zhu, Q., and H. Lin. 2011. Influences of soil, terrain, and crop growth on soil
moisture varia on from transect to farm scales. Geoderma 163:4554.
doi:10.1016/j.geoderma.2011.03.015
www.VadoseZoneJournal.org