Académique Documents
Professionnel Documents
Culture Documents
available at www.sciencedirect.com
Centre for Advanced Numerical Simulation (CANSIM), Department of Civil and Geological Engineering,
University of Saskatchewan, 57 Campus Drive, Saskatoon, SK, Canada S7N 5A9
Received 6 March 2008; received in revised form 1 August 2008; accepted 11 August 2008
KEYWORDS Summary Soil moisture is a key variable that defines the land surface-atmosphere
Soil moisture content; (boundary layer) interactions, by contributing directly to the surface energy balance
Higher-order neural and water balance. This paper investigates the utility of the widely adopted data-driven
networks; model, namely artificial neural networks (ANNs), for modeling the complex soil moisture
Modeling; dynamics. Datasets from three experimental soil covers (D1, D2, and D3), with thickness
Reconstructed of 0.50 m, 0.35 m, and 1.0 m, comprising a thin layer of peat mineral mix over varying
watersheds; thickness of till, are considered in this study. Volumetric soil moisture contents at both
Conceptual models the peat and the till layers were modeled as a function of precipitation, air temperature,
net radiation, and ground temperature at different layers. Initial simulations illustrated
that, in the absence of time-lagged meteorological variables, the ground temperature
is the most influential state variable for characterizing the soil moisture, highlighting
the strong link between the soil thermal properties and the corresponding moisture status.
With the objective of extracting the maximum information from the most influential state
variables (ground temperature), a higher-order neural networks (HONNs) model was
developed to characterize the soil moisture dynamics. The HONNs resulted in relatively
higher correlation coefficient, than traditional ANNs, for some of the soil moisture simu-
lations. Time-lagged inputs were used to improve the model performance and obtain opti-
mum results. The ANN models performed better than a previously developed conceptual
model for estimating the depth-averaged soil moisture content. Results from the study
indicate that modeling of soil moisture using ANNs is challenging but achievable, and its
performance is largely influenced by the structure and formation of the soil covers, which
in turn governs the dynamics of soil moisture variability.
ª 2008 Elsevier B.V. All rights reserved.
0022-1694/$ - see front matter ª 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2008.08.012
2 A. Elshorbagy, K. Parasuraman
weight vector. For the same problem, the HONNs use range of 0.05–0.80 m, and nine TDR sensors for
higher-order correlations to learn the higher-order proper- watershed D3 installed over a depth range of 0.05–1.70 m.
ties in the input space. In general, the objectives of this Hourly values of soil temperature of peat (STp) and till (STt)
study can be stated as: (i) identifying the additional gains, layers are measured using thermisters buried in each of the
if any, of considering HONNs for modeling soil moisture con- watersheds at the depth ranges corresponding to the TDR
tents; (ii) evaluating the ability of neural networks in mod- sensors in each watersheds. Consequently, D1, D2, and D3
eling and characterizing the depth-averaged soil moisture have eight, seven, and nine soil temperature sensors, respec-
values of reconstructed watersheds (soil covers). tively. A weather station located in the mid-slope of D2 mea-
sures air temperature (AT), and precipitation (P). Similarly,
Study area and field data Bowen station located at the mid-slope of D2 measures net
radiation (NR) and energy fluxes. All the meteorological vari-
Large scale mining in the Athabasca basin, Alberta, Canada, ables are measured in an hourly scale. More details on the
involves stripping of large amount of organic and glacial field instrumentation program and the data collected can
deposits and a layer of saline/sodic cretaceous shale to gain be found in Boese (2003) and Elshorbagy et al. (2007).
access to the oil sands. Approximately 120,000 m3 of shale Average daily values of precipitation, air temperature,
overburden is removed daily and is placed in storage areas soil temperature (STp and STt), net radiation (NR), and soil
that will eventually cover a combined area of about moisture (SMp and SMt) are considered for simulation pur-
80 km2 (Boese, 2003). Prior to mining, the shale overburden poses. The ground temperature and soil moisture contents
is stable since it is over consolidated, confined, and is ex- are depth-averaged. As the soil stratum is frozen during
posed only to saline/sodic pore fluids. However, once the the winter, only summer (May–September) time data are
shale is placed on the surface, it is exposed to fresh water considered. Neglecting the year 1999, when the dataset
and oxygen and is susceptible to weathering, which in the was incomplete, soil moisture data between 2000 and
long run affects the stability of the pile (Barbour et al., 2005 are considered. As the reconstructed watersheds
2001). In order to overcome this problem, the piles are re- evolve over time to achieve hydro-sustainability, the
contoured and capped with sufficient soil cover so that freeze–thaw cycles and decomposition of highly organic
the amount of precipitation percolating below the root zone peat layer increases the porosity of the soil and conse-
can be minimized while maintaining enough moisture for quently increasing infiltration rates (Haigh, 2000). Hence,
vegetation. In this way, the overburden can be restored to modeling the moisture dynamics of such evolving water-
its natural state by supporting vegetation. Over the years, sheds would be adding to the already challenging task of
several large scale soil cover (reconstructed watersheds) modeling soil moisture. The following section, in brief, out-
experiments are being conducted to assess the performance lines the ANN models adopted in this study.
of different reclamation strategies by studying the basic
mechanisms that control the moisture movement within Neural networks
these covers. In particular, three experimental soil covers
(D1, D2, and D3) were established in the year 1999. The Since a plethora of information on ANNs is available in the
experimental covers were constructed over the saline-sodic literature (e.g. Haykin, 1999; ASCE Task Committee on
overburden with thickness of 0.50 m, 0.35 m, and 1.0 m, Application of Artificial Neural Networks in Hydrology,
comprising a thin layer of peat mineral mix over varying 2000), the description of ANNs herein is brief, and limited
thickness of secondary (glacial/till) soil. The layout of the to the needs of this study. A three layered feed-forward
soil covers are shown in Fig. 1. Cover D1 consists of 20 cm neural networks (FFNNs) model is conventionally adopted
of peat overlying 30 cm of till; cover D2 consists of 15 cm in modeling hydrological processes. The three layered
of peat overlying 20 cm of till, and cover D3 consists of FFNNs consist of ‘j’ input neurons, ‘k’ hidden neurons,
20 cm of peat overlying 80 cm of till (Fig. 1). and ‘l’ output neurons. These neurons are connected with
The three soil covers are contoured in such a way that each other by means of connection weights and bias. The
these soil covers can be considered as individual sub-water- hidden layer neurons usually employ sigmoidal activation
sheds. Each of the sub-watersheds has an area of 1 ha function and the output layer neurons use linear activation
(approximately 200 m long and 50 m wide), with a 5:1 slope function. These activation functions help in nonlinearly
(5 horizontal to 1 vertical). These reconstructed watersheds, transforming the inputs to the desired output. Symbolically,
compared to the natural watersheds, are not stable during the three layered FFNNs can be represented as ANN(j, k, l).
their initial stages, and hence evolve over time to achieve hy- The development of a neural network model demands two
dro-sustainability. In order to track the evolution (hydrolog- operations, namely, (i) training and (ii) testing. Training is
ical changes) of the watersheds, intensive instrumentations a process by which the connection weights between differ-
were installed in these watersheds. Each watershed has an ent layers and the bias values of the neural networks are
individual soil station located at the middle of the slope, optimized by minimizing the cost function. Once trained,
which measures the volumetric soil moisture content of the the neural network model can be tested on an independent
peat (SMp) and the till (SMt) layers, twice a day. Soil moisture dataset that has not been used during the training process.
is measured using TDR principles with model CS615 In ANN modeling of hydrological processes, the signifi-
(Boese, 2003). The TDR sensors were installed laterally cant input variables for characterizing a particular process
into the soil profile. Watershed D1 has eight TDR sensors is usually determined by linear cross-correlation (Bowden
installed over a depth range of 0.05–1.00 m. Similarly, seven et al., 2005). Linear cross-correlation can only detect linear
TDR sensors were installed in watershed D2 over a depth dependence between two variables, and is not suited for
4 A. Elshorbagy, K. Parasuraman
w1
x1
x2 w2
1st order
.
correlation .
xn .
x1
wn
x1
a
w11 w0
x1 2
x2 x2 x1x2
w12
2nd order z y
correlation
. ∑ ϕ ( .)
. .
. xn2 .
. .
wnn
. .
. .
xn .
. .
xn . .
. .
.
. W1...1
. x1N
x1N-1x2 W1...2
Nth order .
correlation .
x1N .
Wn...n
can be computed externally and fed to the ANNs. Both ways hidden neurons and user adjustable parameters, with the
lead to the same results, however, if derived inputs are objective of minimizing the cost function (Eq. (6)).
computed internally, then the inputs will consist of all pos- !
sible combinations. External computations allow for con- 1X n
2 1 XN
MSE REG ¼ y i y 0i þ ð1 þ aÞ W 2j ð6Þ
structing partial HONNs (Input consists of some, not all, n i¼1 N j¼1
second-order inputs). The neural network models adopted
in this study employ tan-sigmoidal activation function in
the hidden layer neurons and linear activation function in Soil moisture modeling
the output layer neurons. The neural network models were
trained using the Bayesian-regularization back propagation In order to account for the evolving nature of the water-
algorithm (Demuth and Beale, 2001). Unlike other training sheds in neural network modeling, the available dataset
algorithms, where the cost function involves minimizing (2000–2005) was apportioned in such a way that every other
the mean sum of squares of the network errors (MSE) alone, data instance (alternating instances) appears in the training
the cost function of the Bayesian-regularization back prop- dataset, while the remaining instances make up the testing
agation algorithm involves minimizing both the MSE and the dataset. The number of instances considered for each of
mean of sum of squares of the network bias and connection training and testing were 384. The correlation statistics be-
weights (MSW). The cost function used in the Bayesian-reg- tween the independent variables and the dependent vari-
ularization algorithm is presented in Eq. (6), where yi and y 0i ables for sub-watersheds D1, D2, and D3, are presented in
represent the measured and computed counterparts; a rep- Table 1. Although there are no strong correlations between
resents the regularization parameter; n, and N represents each of the independent and the dependent variables (Table
the number of training instances and the number of network 1), the most correlated variables are not the same for all
parameters, respectively. By minimizing both the MSE and the three sub-watersheds (soil covers). For sub-watershed
MSW, the Bayesian-regularization algorithm improves the D1, NR has the highest correlation with SMp, and AT has
generalization property of the ANN model by developing the highest correlation with SMt; for sub-watershed D2,
networks with smaller weights and bias, and thus a smooth- STt and STp have the highest correlation with SMp and SMt,
er response that is less likely to over fit the data (Demuth respectively, signifying the link between the moisture status
and Beale, 2001). Conventionally, the optimal number of of one layer and the thermal properties of the other layer;
hidden nodes and the network model parameters are deter- for sub-watershed D3, STt has the highest correlation with
mined by trial-and-error method. Hence, this study adopted both SMp and SMt, indicating the influence of the large stor-
a similar approach to arrive at the optimal network config- age and the thermal properties of till layer on both SMp and
uration, by performing a systematic search on number of SMt of the thickest cover; D3 (Table 1).
6 A. Elshorbagy, K. Parasuraman
Table 1 Correlation statistics between the independent and the dependent variables for the three sub-watersheds during
training and testing
Sub-watershed Variable Training Testing
SMp SMt SMp SMt
D1 P 0.11 0.06 0.00 0.01
AT 0.05 0.22 0.03 0.21
NR 0.16 0.15 0.14 0.18
STp 0.02 0.16 0.02 0.15
STt 0.11 0.02 0.12 0.00
D2 P 0.09 0.01 0.04 0.00
AT 0.12 0.16 0.12 0.16
NR 0.03 0.09 0.09 0.20
STp 0.34 0.21 0.34 0.22
STt 0.35 0.05 0.36 0.06
D3 P 0.02 0.09 0.03 0.08
AT 0.18 0.16 0.18 0.15
NR 0.05 0.07 0.05 0.11
STp 0.19 0.20 0.17 0.20
STt 0.21 0.23 0.20 0.22
Several experiments (models) were considered to esti- evolution of the reconstructed watersheds to reach sustain-
mate the soil moisture content of the peat (SMp) and the till ability constantly alters the hydraulic conductivity of the
(SMt) layers, using different combinations of the indepen- soil strata, which in turn influences the phase lag. Because
dent variables, with the objective of finding the optimal in- of the soil effect as a low-pass filter, it is hypothesized that
put combination for modeling SMp and SMt. The independent the cumulative effect of the meteorological variables is
variables considered in this study are P, AT, NR, STp, and more important than the incremental daily effect. In other
STt. Since precipitation is the main source of soil moisture, words, for example, instead of considering the preceding
it would have been coherent to use the precipitation of pre- week daily precipitation as seven input variables
vious days (Pt1, Pt2, . . . , Ptn) in estimating the soil mois- (Pt, Pt1, . . . , Pt6), the summation of all values (R Pt7) can
ture of a particular day. However, the first category of be considered as one input. The favorable impact of reduc-
experiments in this study does not consider the preceding ing the number of inputs on the model parameter estima-
precipitation for estimating soil moisture. It is hypothe- tion and robustness cannot be overemphasized. The
sized, and will be proved subsequently, that the effect of performances of models developed in this study were eval-
this phase lag can be implicitly accounted for by considering uated based on the root-mean squared error (RMSE), the
soil temperature as one of the inputs to the ANN model. mean absolute relative error (MARE), and correlation coef-
Eltahir (1998) indicated that a positive feedback mechanism ficient (R). RMSE, MARE, and R are calculated using Eqs.
exists between soil moisture and rainfall as soil moisture (7)–(9), respectively, where N represents the number of in-
conditions themselves reflect past occurrence of rainfall, stances presented to the model; yi and y 0i represent mea-
and the feedback mechanism driven by the control of soil sured and computed counterparts; and y represents the
moisture on surface albedo and Bowen ratio. This implies mean of the corresponding variable.
that an increase in soil moisture decreases the surface albe-
do and Bowen ratio, which in turn increases the net radia- " #0:5
1X N
tion, and decreases the ground and surface temperature RMSE ¼ ðy y 0i Þ2 ð7Þ
(Eltahir, 1998). Due to the link between rainfall, soil mois- N i¼1 i
ture, and soil temperature, the phase lag between the
N
occurrence of precipitation and its manifestation as soil 1 X 0
y i y i
MARE ¼ ð8Þ
moisture would be reflected in the phase of soil tempera- N i¼1 y i
ture. The phase of soil moisture and soil temperature would
be nearly identical due to the strong link between the soil P
N
thermal properties and soil moisture. ðy i y i Þ y 0i y 0i
i¼1
The second category of experiments considers preceding R ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð9Þ
PN PN 2
meteorological variables for estimating soil moisture. The
ðy i y i Þ2 y 0i y 0i
soil acts as a low-pass filter where there is a phase lag be- i¼1 i¼1
tween the occurrence of precipitation and other variables
and their manifestation as soil moisture. This phase lag is
influenced by the atmospheric forcing on the soil strata, Results and analysis
and as a result, keeps shifting over time. This shifting may
be more pronounced in the case of reconstructed water- To understand the relationship between soil moisture con-
sheds, where in addition to the atmospheric forcing, the tents at different layers (depths), and the arrangement
On the relevance of using artificial neural networks for estimating soil moisture content 7
(thickness) of the corresponding soil layers, scatter plots training and testing (Table 2), signifying the strong correla-
between SMp and SMt for the three watersheds are pre- tion between soil thermal properties and the moisture sta-
sented in Fig. 3. As indicated earlier, D3 is the thickest cov- tus. For D1, the model M3 performed better than M2
er, followed by D1 and D2. The correlation between SMp and (Table 2) indicating that the STp and STt alone can explain
SMt for D1, D2, and D3, are 0.7, 0.6, and 0.2, respectively. the variance in SMp. Comparing different models, during
As expected, the correlation between SMp and SMt was the testing, M4 outperformed the other models in estimating
least for the thickest cover. Nevertheless, interestingly, in- SMp. This illustrates the ability of the HONNs in extracting
stead of D2, which is the thinnest cover, D1 resulted in the the maximum information from the most significant vari-
highest correlation between SMp and SMt. This illustrates ables in estimating SMp for sub-watershed D1 Fig. 4 shows
that the relationship between the moisture dynamics of the correlation plot between the measured and simulated
the peat and till layers is not just driven by the thicknesses SMp by different models during testing, for sub-watershed
of the two layers, but by a more complex, and probably D1. Relatively, M4 has less scatter than the other models
poorly understood, mechanism. (Fig. 4). Table 3 indicates the performance of different
models in estimating SMt for watershed D1. In general, the
Experiment category I performance of models in estimating SMt (Table 3) is similar
to the performance of models in estimating SMp (Table 2).
For each sub-watershed, various neural network models Inclusion of STp and STt as inputs to the ANN models helped
(M1, M2, M3, and M4) were developed, as the first category in significantly improving the performance of the models.
of experiments, with the objective of estimating the soil During training, M2 performed better than other models.
moisture (SMp and SMt). These models differed from each However, during testing, though M2, M3, and M4 performed
other with regards to the input combinations considered on par in terms of RMSE and MARE, M4 resulted in relatively
for modeling the soil moisture patterns. While model M1 better correlation statistics (Table 3). Correlation plots be-
(FONNs) uses only climatic data (P, AT, and NR) as the input tween the measured and simulated till moisture for sub-wa-
vector, models M2, M3, and M4 use soil temperature tershed D1 are presented in Fig. 5. To a less extent, the
(STp, STt) or/and its variants (ST2p ; ST2t , STpSTt) as or part of same phenomenon could be observed with SMp (Fig. 4). This
the input vector. Models M2, M3, and M4 can be used to could be attributed to the possibility of lack of precipitation
adjudicate the degree of relevance of using soil tempera- for extended periods. During such dry periods, the evapo-
ture data in soil moisture modeling. Model M2 (FONNs) con- transpiration dynamics becomes the driving variable that af-
siders P, AT, NR, STp and STt in the input vector; Model M3 fects the available moisture. Most probably, these will be
(FONNs) uses STp and STt alone as inputs; and Model M4 the most challenging periods for the ANNs models to learn
(HONNs) employs STp, STt, ST2p , ST2t , STpSTt as inputs. For the soil moisture patterns.
the three watersheds (D1, D2, and D3), the optimal number
of hidden neurons for each of the models (M1, M2, M3, and Sub-watershed D2
M4) varied between four and six. For brevity, the number of For sub-watershed D2, Table 4 shows the performance of
hidden neurons for individual models is not presented. different models in estimating SMp; and Table 5 shows the
performance of models in estimating SMt. Similar to sub-wa-
Sub-watershed D1 tershed D1, the performance of models in estimating SMp
For sub-watershed D1, the performance of different models and SMt improved significantly when STp and STt were used
in terms of RMSE, MARE, and R in estimating the SMp is pre- as inputs to the ANN models. However, in estimating SMp for
sented in Table 2. The correlation between the measured sub-watershed D2, contrary to sub-watershed D1, the per-
and the simulated soil moisture indicate that the models formance of models M3 and M4 were inferior to the perfor-
were able to account for only a moderate proportion of mance of model M2 (Table 4). This could be attributed to
the variability in the soil moisture data. Models M2, M3, the thinner depth of the peat layer in D2, and also the thin-
and M4 performed significantly better than M1 during both ner depth of the entire D2 soil cover. Hence, the peat layer
SMt
SMt
Figure 3 Scatter plots between SMp and SMt for sub-watersheds: (a) D1, (b) D2, and (c) D3.
8 A. Elshorbagy, K. Parasuraman
Table 2 Performance measure of different models in estimating peat moisture content (cm/cm) of sub-watershed D1
Model Training Testing
RMSE MARE R RMSE MARE R
M1 0.051 0.098 0.18 0.051 0.099 0.128
M2 0.040 0.073 0.644 0.046 0.083 0.507
M3 0.043 0.081 0.547 0.043 0.081 0.552
M4 0.042 0.076 0.597 0.042 0.077 0.581
M1, M2, and M3 are FONN models, whereas M4 is an HONN model.
a 0.6 b 0.60
0.55
0.5 0.50
Simulated
Simulated
0.45
0.4 0.40
0.35
0.3 0.30
0.25
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
Measured Measured
c 0.6 d 0.60
0.55
0.5 0.50
Simulated
Simulated
0.45
0.4 0.40
0.35
0.3 0.30
0.25
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
Measured Measured
Figure 4 Comparison of measured vs. simulated peat moisture of cover D1 by (a) M1; (b) M2; (c) M3; and (d) M4 – testing.
Table 3 Performance measure of different models in estimating till moisture content (cm/cm) of sub-watershed D1
Model Training Testing
RMSE MARE R RMSE MARE R
M1 0.018 0.050 0.236 0.018 0.050 0.209
M2 0.015 0.0430 0.624 0.016 0.043 0.532
M3 0.016 0.043 0.552 0.016 0.043 0.539
M4 0.016 0.043 0.573 0.016 0.043 0.556
of D2 is more dynamic and responsive to the meteorological 0.061, 0.138, and 0.575, respectively, M2 outperformed the
variables. During testing, with RMSE, MARE, and R values of other models. The better performance of the M2 reiterates
On the relevance of using artificial neural networks for estimating soil moisture content 9
a 0.32 b 0.32
0.30 0.30
Simulated
Simulated
0.28 0.28
0.26 0.26
0.24 0.24
0.22 0.22
0.22 0.24 0.26 0.28 0.30 0.32 0.22 0.24 0.26 0.28 0.30 0.32
Measured Measured
c 0.32
d 0.32
0.30 0.30
Simulated
Simulated
0.28 0.28
0.26 0.26
0.24 0.24
0.22 0.22
0.22 0.24 0.26 0.28 0.30 0.32 0.22 0.24 0.26 0.28 0.30 0.32
Measured Measured
Figure 5 Comparison of measured vs. simulated till moisture of cover D1 by (a) M1; (b) M2; (c) M3; and (d) M4 – testing.
Table 4 Performance measure of different models in estimating peat moisture content (cm/cm) of sub-watershed D2
Model Training Testing
RMSE MARE R RMSE MARE R
M1 0.072 0.173 0.265 0.1 0.171 0.297
M2 0.062 0.139 0.566 0.061 0.138 0.575
M3 0.064 0.146 0.519 0.063 0.146 0.525
M4 0.063 0.144 0.529 0.063 0.144 0.537
Table 5 Performance measure of different models in estimating till moisture content (cm/cm) of sub-watershed D2
Model Training Testing
RMSE MARE R RMSE MARE R
M1 0.023 0.067 0.167 0.023 0.067 0.179
M2 0.014 0.039 0.807 0.015 0.041 0.774
M3 0.014 0.0140 0.792 0.014 0.041 0.783
M4 0.014 0.039 0.805 0.014 0.040 0.789
the strong influence of atmospheric forcing on the peat low-ranges of SMp (Fig. 6), it can be concluded that most of
moisture dynamics of sub-watershed D2. This underscores the models were unable to learn the moisture dynamics at
the need for considering P, AT, NR, STp, and STt for charac- low-ranges. For watershed D2, the models were able to esti-
terizing SMp of watershed D2 Fig. 6 shows the correlation mate SMt more accurately than SMp (Tables 4 and 5). This
plot between the measured and simulated values of SMp dur- may be attributed to the reason that the upper peat layer
ing testing, for watershed D2. From the large scatter in the acts like a buffer zone that trims the effect of the meteoro-
10 A. Elshorbagy, K. Parasuraman
a 0.6
b 0.6
0.5 0.5
Simulated
Simulated
0.4 0.4
0.3 0.3
0.2 0.2
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
Measured Measured
c 0.6
d 0.6
0.5 0.5
Simulated
Simulated
0.4 0.4
0.3 0.3
0.2 0.2
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
Measured Measured
Figure 6 Comparison of measured vs. simulated peat moisture of cover D2 by (a) M1; (b) M2; (c) M3; and (d) M4 – testing.
logical variables on the till layer, and thereby rendering SMt soil moisture of the thick cover D3. Due to the relative
less vulnerable to atmospheric forcing. During testing, incoherency between STp and STt, the effect is more pro-
compared to other models, the HONN model (M4) resulted nounced in M4, where higher-order derivatives of the above
in the highest R statistics. Nevertheless, models M2, M3, variables were adopted for estimating SMp (Table 6). The
and M4 performed on par in terms of RMSE and MARE (Table correlation plots between measured and simulated SMp for
5) Fig. 7 shows the correlation plots between the measured sub-watershed D3, are presented in Fig. 8. From the plots,
and simulated SMt by different models. Disregarding M1, all it is evident that none of the models was able to learn the
the other models were able to estimate SMt with dynamics of SMp. In modeling SMt for sub-watershed D3, it
reasonable accuracy over the entire modeling range is observed that inclusion of STp and STt as inputs to the
(Fig. 7). ANN models was able to improve the performance of the
models in terms of R statistics. However, the performances
Sub-watershed D3 of the models are comparable in terms of RMSE and MARE
The performance of different models in estimating SMp and statistics (Table 7) Fig. 9 shows the correlation between
SMt for sub-watershed D3 is presented in Tables 6 and 7, the measured and the simulated SMt by different models.
respectively. For sub-watershed D3, similar to sub-water- As can be seen from the plots, though M3 and M4 resulted
sheds D1 and D2, inclusion of STp and STt as inputs to the in less scatter in a particular modeling range (Fig. 9), the
ANN models improved the performance of the models in overall performance of the models are not convincing. In
estimating soil moisture (SMp and SMt). However, the general, it can be concluded that all ANN models, within
improvement in the performance of the models for sub-wa- this first category of experiments, encountered difficulties
tershed D3 is not as significant as compared to sub-water- in learning the moisture patterns of sub-watershed D3.
sheds D1 and D2. This may be attributed to the weak Since D3 is the thickest cover among the three covers con-
correlation (R = 0.2) existing between SMp and SMt. Due to sidered in this study, it implies that sub-watershed D3 re-
the association between soil thermal properties and mois- sults in the highest phase-lag between the occurrence of
ture status, the weak correlation between SMp and SMt im- precipitation and its manifestation as soil moisture. In other
plies that the information content between STp and STt words, sub-watershed D3 has more storage effects than the
would also be relatively weak. Hence, inclusion of STp and other two sub-watersheds, and thereby is prone to more
STt as inputs to the ANN models was not able to appreciably topographic forcing. This makes modeling the soil moisture
improve the performance of the models in estimating the dynamics of sub-watershed D3, more intricate than model-
On the relevance of using artificial neural networks for estimating soil moisture content 11
a 0.34
b 0.34
0.32 0.32
0.30 0.30
Simulated
Simulated
0.28 0.28
0.26 0.26
0.24 0.24
0.22 0.22
0.20 0.20
0.20 0.22 0.24 0.26 0.28 0.30 0.32 0.34 0.20 0.22 0.24 0.26 0.28 0.30 0.32 0.34
Measured Measured
c 0.34
d 0.34
0.32 0.32
0.30 0.30
Simulated
Simulated
0.28 0.28
0.26 0.26
0.24 0.24
0.22 0.22
0.20 0.20
0.20 0.22 0.24 0.26 0.28 0.30 0.32 0.34 0.20 0.22 0.24 0.26 0.28 0.30 0.32 0.34
Measured Measured
Figure 7 Comparison of measured vs. simulated till moisture of cover D2 by (a) M1; (b) M2; (c) M3; and (d) M4 – testing.
Table 6 Performance measure of different models in estimating peat moisture content (cm/cm) of sub-watershed D3
Model Training Testing
RMSE MARE R RMSE MARE R
M1 0.044 0.092 0.275 0.044 0.093 0.279
M2 0.044 0.091 0.293 0.044 0.091 0.296
M3 0.043 0.087 0.328 0.044 0.089 0.326
M4 0.044 0.092 0.215 0.045 0.093 0.211
Table 7 Performance measure of different models in estimating till moisture content (cm/cm) of sub-watershed D3
Model Training Testing
RMSE MARE R RMSE MARE R
M1 0.013 0.024 0.177 0.013 0.024 0.165
M2 0.013 0.024 0.237 0.013 0.025 0.228
M3 0.012 0.023 0.400 0.012 0.023 0.383
M4 0.012 0.023 0.405 0.012 0.023 0.386
ing the soil moisture dynamics of sub-watersheds D1 and D2. averages over larger depths is severely limited by the noise
This finding is in line with Entekhabi et al. (1994), where it is introduced by the micro-topography and roughness of vege-
shown that the ability of models to predict soil moisture tation cover.
12 A. Elshorbagy, K. Parasuraman
a 0.50 b 0.50
0.45 0.45
Simulated
Simulated
0.40 0.40
0.35 0.35
0.30 0.30
0.25 0.25
0.25 0.30 0.35 0.40 0.45 0.50 0.25 0.30 0.35 0.40 0.45 0.50
Measured Measured
c 0.50
d 0.50
0.45 0.45
Simulated
Simulated
0.40 0.40
0.35 0.35
0.30 0.30
0.25 0.25
0.25 0.30 0.35 0.40 0.45 0.50 0.25 0.30 0.35 0.40 0.45 0.50
Measured Measured
Figure 8 Comparison of measured vs. simulated peat moisture of cover D3 by (a) M1; (b) M2; (c) M3; and (d) M4 – testing.
a 0.38
b 0.38
Simulated
0.34 0.34
0.32 0.32
0.30 0.30
0.30 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.30 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38
Measured Measured
c 0.38
d 0.38
0.36 0.36
Simulated
Simulated
0.34 0.34
0.32 0.32
0.30 0.30
0.30 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.30 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38
Measured Measured
Figure 9 Comparison of measured vs. simulated till moisture of cover D3 by (a) M1; (b) M2; (c) M3; and (d) M4 – testing.
Table 8 Performance measure of different models in estimating peat moisture content (cm/cm) of all sub-watersheds – testing
D1 (50 cm cover) D2 (35 cm cover) D3 (100 cm cover)
RMSE MARE R RMSE MARE R RMSE MARE R
Ref. ANNs 0.04 0.08 0.58 0.06 0.14 0.58 0.042 0.09 0.40
M6 0.05 0.09 0.42 0.06 0.13 0.64 0.04 0.08 0.54
M7 0.05 0.08 0.49 0.06 0.12 0.66 0.03 0.065 0.73
M8 0.03 0.07 0.69 0.05 0.13 0.68 0.04 0.075 0.62
M9 0.03 0.07 0.65 0.05 0.11 0.72 0.03 0.064 0.74
SDW model 0.06 0.13 0.60 0.08 0.14 0.42 0.04 0.08 0.68
M6, M7, M8, and M9 are all FONN models.
Ref. ANNs represent the best model from Experiment category I.
statistic is the best indication in this case to reflect the cover has large storage potential, which is captured by
improvement compared to the reference ANNs. For the the ANN models through the precipitation history. The cover
thickest cover, D3, it is interesting to note that when RP6 can store enough moisture during the wet days so that
was included (M6), a good improvement was achieved. How- evapotranspiration (represented by air temperature) does
ever, even more significant improvement was achieved by not become the most important control of the soil moisture.
replacing RP6 with RP13 (M7) compared to adding RAT6 In case of the thinnest cover D2, one can observe that
(M8). Similar conclusions can be drawn regarding the till including RAT6 helps achieve reasonable improvement in
layer of D3 (Table 9). This reflects the fact that the thick the prediction accuracy; M9 is better than M7 in both Table
14 A. Elshorbagy, K. Parasuraman
Table 9 Performance measure of different models in estimating till moisture content (cm/cm) of all sub-watersheds – testing
D1 (50 cm cover) D2 (35 cm cover) D3 (100 cm cover)
RMSE MARE R RMSE MARE R RMSE MARE R
Ref. ANNs 0.016 0.04 0.56 0.014 0.04 0.79 0.012 0.02 0.39
M6 0.016 0.043 0.56 0.014 0.04 0.78 0.011 0.023 0.45
M7 0.016 0.042 0.56 0.014 0.04 0.80 0.01 0.022 0.56
M8 0.013 0.037 0.70 0.011 0.036 0.84 0.011 0.023 0.52
M9 0.012 0.036 0.70 0.011 0.036 0.84 0.01 0.022 0.59
SDW model 0.021 0.08 0.67 0.10 0.29 0.08 0.02 0.05 0.50
8 and Table 9. Unlike D3, moving from M6 to M8 is better and runoff on a daily basis. A combination of physically
than moving from M6 to M7. Even though precipitation is based formulations (e.g. Green-Ampt for infiltration and soil
still an important factor, evapotranspiration is a major con- moisture redistribution) and fitted-parameter formulations
trol of moisture on D2. The cover is thin and lacks the stor- (e.g. actual evapotranspiration based on simulated soil
age ability; therefore, it cannot store moisture from far in moisture index) was employed in formulating the SDW mod-
the past. Evapotranspiration could really cause D2 to dry el. More details on the model can be found in Elshorbagy
out. In case of the intermediate cover D1, both the peat et al. (2007). The model was extensively evaluated for the
(Table 8) and the till layers (Table 9) behaves in a way that purposes of design and hydrologic performance assessment
show a balancing effect of both the storage and the precip- of reconstructed watersheds (Elshorbagy and Barbour,
itation. Models M8 and M9 are significantly better than M6 2007) as well as for management decision making (Elshor-
and M7; reflecting the need of presenting balanced memory bagy, 2006).
of both the storage and the precipitation. Even though it is infeasible to compare the input struc-
Noting that all experiments of category II (M5–M9) are ture of mechanistic and data-driven models, the input vari-
based on FONN models, an additional HONN model (M10) ables and data needed for the SDW model are provided here
was tested in this category. M10 was designed to be fully to facilitate the evaluation of the performances of the ANN
second-order ANN model based on the best performance re- models relative to the SDW model. The current version of
ported in Table 8 and Table 9; i.e., M10 is a second-order the SDW model requires daily precipitation (P), average dai-
ANNs of M8 for the peat layer of D1 cover, and a second-or- ly air temperature (AT), daily net radiation (NR), relative
der ANNs of M9 for the peat layer of D2 and D3, as well as for humidity (RH), average daily temperatures of both peat
the till layer of all three covers. Other possibilities and sce- and till layers (STp and STt), and soil physical properties that
narios of fully and partial HONNs are endless, however, allow for the estimation of the soil water characteristic
experiment M10 is considered sufficient for the purpose of curve (SWCC). However, it should be noted that by defini-
this study. The only noticeable improvement achieved by tion of mechanistic models, the SDW model has the ability
the use of HONNs (M10) over the FONNs is the increase in to keep ‘‘indefinite’’ memory of past values (time-lagged)
the R statistic in the case of the peat layer of the thick cover of input and simulated parameters through is tank (stock)
D3 (slight increase from 0.74 to 0.77) as well as the till layer component. At the same time, the SDW model can provide
(considerable increase from 0.59 to 0.70). It can be con- soil moisture content, actual evapotranspiration, and runoff
cluded that second-order ANNs played a positive role in as outputs. The SDW model has nine different calibration
improving the prediction of soil moisture in thicker cover. parameters related to the processes of infiltration, evapo-
It might be possible that even higher-order ANNs could im- transpiration, snowmelt, and runoff (Elshorbagy et al.,
prove the results further. This should be subject to future 2007).
investigation. The performance of the SDW model in estimating the soil
moisture in the peat layer of the three sub-watersheds is
Comparison between the ANNs and the conceptual presented in Table 8. Clearly, the best ANN models per-
SDW model formed reasonably better, and sometimes much better as
in the case of D2, than the SDW model. The R statistic
Even though the second category of experiments signifi- should be noted in all covers. Similar conclusions can be
cantly improved the ANN model performance with regard drawn with regard to the till layer (Table 9). It is interesting
to the estimation of the soil moisture content, it is desirable to note that the poor performance of the SDW model ap-
to measure such performance against a validated concep- plied to the thinnest cover D2 was significantly improved
tual model. Elshorbagy et al. (2005, 2007) developed a and overcome by the ANN models. Elshorbagy et al. (2007)
site-specific system dynamics watershed (SDW) model that reported that the presence of macro pores and the steep
simulates the various hydrological processes occurring in slope of the covers make the influence of the surface runoff
the reconstructed watersheds; D1, D2, and D3. The model and the interflow dynamics more prominent with the thin
is a lumped mechanistic model that was conceptualized as cover (D2). Such dynamics were difficult to capture within
a control volume that simulates the water balance compo- the conceptual SDW model, whereas the data-driven ANNs
nents among the two soil layers (peat and till) as well as were less affected by such dynamics and phenomena. For vi-
the underlying shale layer including evapotranspiration sual assessment of the second category of ANN experiments,
On the relevance of using artificial neural networks for estimating soil moisture content 15
Discussion
0.5
Simulated
mating the peat and till layers’ moisture content, was bet-
ter than the SDW model in all cases, one can observe that 0.4
the ANN model superiority is less in case of D1 (50 cm) than
in case of D2. The least improvement of the ANN models
over the SDW model was achieved with the thickest cover
0.3
D3 (100 cm) as shown in Tables 8 and 9. Clearly, the storage
effect, which is more prominent in thicker soil layers, was
the most difficult phenomenon to be captured by the ANNs.
Considering meteorological and hydrological parameters 0.2
representing a longer memory from the past could help even 0.2 0.3 0.4 0.5 0.6
Simulated
a 0.60
0.28
0.55
0.50 0.24
S i mulated
0.45
0.20
0.40 0.20 0.24 0.28 0.32 0.36
Measured
0.35
Figure 11 Comparison of measured vs. simulated moisture of
0.30 cover D2 by M9 (a) peat layer; and (b) till layer – testing.
0.25
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 such as wavelet analysis could play a major role in this task
as well as in evaluating the localized performance of various
Measured
models.
0.32 The classical root mean squared error (RMSE) and the
b mean absolute relative error (MARE) were used to test the
overall performance of the models. However, in cases like
0.30 predicting the soil moisture content (cm/cm or percent),
the average error value could be deceivingly low (i.e., good
performance) but the predicted values do not match well
Simulated
0.28
with the observed values in terms of pattern and trend. In
this case, the correlation coefficient (R) is strongly recom-
mended as a model performance statistic. This study shows
0.26
that the R statistic was the most important error measure,
among the three adopted, in revealing the improvement in
0.24 soil moisture prediction. The ANN models and experiments
conducted in this study will help the authors develop a gen-
eric ANN-based model that can be used by the oil sands min-
0.22 ing industry to obtain estimates of the soil moisture
0.22 0.24 0.26 0.28 0.30 0.32 contents of various reclamation soil covers using available
Measured meteorological and hydrological parameters.
As seen from the above discussions, modeling soil mois-
Figure 10 Comparison of measured vs. simulated moisture of ture using neural networks is challenging, and their perfor-
cover D1 by M8 (a) peat layer; and (b) till layer – testing. mance is largely influenced by the structure and formation
16 A. Elshorbagy, K. Parasuraman
0.40 models, developed in this study, are not only good from a
predictive point of view but also conceptually meaningful.
0.35
Summary and conclusions
Table 10 Performance measure of HONNs (experiment category II) models in estimating peat and till moisture content (cm/cm)
of all sub-watersheds – testing
D1 (50 cm cover) D2 (35 cm cover) D3 (100 cm cover)
RMSE MARE R RMSE MARE R RMSE MARE R
Peat layer
M8/M9 0.03 0.07 0.69 0.05 0.11 0.72 0.03 0.064 0.74
M10 0.03 0.06 0.70 0.05 0.10 0.74 0.03 0.06 0.77
Till layer
M8/M9 0.012 0.036 0.70 0.011 0.036 0.84 0.01 0.022 0.59
M10 0.01 0.037 0.70 0.01 0.036 0.82 0.01 0.02 0.70
On the relevance of using artificial neural networks for estimating soil moisture content 17
some cases, the correlations between the measured and the Boese, K., 2003. The Design and Installation of a Field Instrumen-
estimated soil moisture for both the peat and the till layers tation Program for the Evaluation of Soil-atmosphere Water
are not very strong, the results are encouraging considering Fluxes in a Vegetated Cover Over Saline/Sodic Shale Overbur-
the large variance that exists between the independent and den, M.Sc. Thesis, University of Saskatchewan, Saskatoon,
Sask.
the dependent variables. The better performance of the
Bowden, G.J., Dandy, G.C., Maier, H.R., 2005. Input determina-
ANN models considering soil temperature as one of the in- tion for neural network models in water resources applica-
puts reiterates the strong link between the soil thermal tions. Part 1 – background and methodology. J. Hydrol. 301,
properties and the soil moisture status. This finding would 75–92.
be of great importance to orient the future monitoring pro- Bowden, G.J., Dandy, G.C., Maier, H.R., 2006. An evaluation of
gram of the reconstructed watersheds, where it is recom- methods for the selection of inputs for an artificial network
mended to increase the intensity of soil temperature based river model. In: Recknagel, F. (Ed.), Ecological Informat-
measurement network system, and thereby possibly im- ics. Springer, Berlin, Heidelberg, pp. 275–292.
prove the accuracy of soil moisture estimates. Dawson, C.W., Wilby, R., 1998. An artificial neural network
approach to rainfall–runoff modeling. Hydrol. Sci. J. 43, 47–
A second category of experiments were time-lagged in-
66.
puts were considered revealed the need to consider variable
Demuth, H., Beale, M., 2001. Neural Network Toolbox Learning For
time window to obtain optimum inputs and results for each Use with MATLAB. The Math Works Inc, Natick, Mass.
sub-watershed. The optimum inputs were found to match Elshorbagy, A., 2006. Multicriterion decision analysis approach to
with the conceptual understanding of the significance/insig- assess the utility of watershed modeling for management
nificance of the storage effect, which is based on the soil decisions. Water Resour. Res. 42, W09407. doi:10.1029/
layering and thickness. For the three sub-watersheds, it is 2005WR00426.
hard to generalize the best combination of inputs for char- Elshorbagy, A., Barbour, L., 2007. Probabilistic approach for design
acterizing soil moisture. This illustrates that the atmo- and hydrologic performance assessment of reconstructed water-
spheric forcing and the soil moisture response to such sheds. J. Geotech. Geoenv. Eng. ASCE 133 (9), 1110–1118.
forcing is influenced by the type, depth, and structure of Elshorbagy, A., Jutla, A., Kells, J., 2007. Simulation of the
hydrological processes on reconstructed watersheds using sys-
the soil strata. In general, the results from this study indi-
tem dynamics. Hydrol. Sci. J. 52, 538–562.
cate that modeling the soil moisture using ANNs is challeng- Eltahir, E.A.B., 1998. A soil moisture-rainfall feedback mechanism,
ing but achievable, and the model performance is largely 1: Theory and observations. Water Resour. Res. 34 (4), 765–
influenced by the structure of the soil strata. The results re- 776.
ported in this study are unique for three reasons: (i) they Entekhabi, D., Nakamura, H., Njoku, E.G., 1994. Solving the inverse
established a link between readily available climatic vari- problem for soil moisture and temperature profiles by sequential
ables and soil-moisture, using ANNs; (ii) they evaluated assimilation of multifrequency remotely sensed observations.
the potential of adopting the ANNs for estimating soil mois- IEEE Trans. Geosci. Remote Sensing 32 (2), 438–448.
ture of reconstructed watersheds, where the dynamics of Entekhabi, D., Rodriguez-Iturbe, I., Castelli, F., 1996. Mutual
atmospheric forcing and the soil moisture response to such interaction of soil moisture state and atmospheric processes.
J. Hydrol. 184, 3–17.
forcing change based on the soil cover layering and thick-
French, M.N., Krajewski, W.F., Cuykendall, R.R., 1992. Rainfall
ness; and (iii) they show that ANNs can perform as good forecasting in space and time using a neural network. J. Hydrol.
as, and possibly better, than conceptual watershed model 137, 1–31.
in estimating depth-averaged soil moisture content. Gautam, M.R., Watanabe, K., Saegusa, H., 2000. Runoff analysis in
humid forest catchment with artificial neural network. J.
Hydrol. 235, 117–136.
Georgakakos, K.P., Bae, D.H., 1994. Climatic variability of soil
Acknowledgements water in the American midwest: 1, Spatio-temporal analysis. J.
Hydrol. 162, 379–390.
The authors gratefully acknowledge the financial support of Georgakakos, K.P., Smith, G.F., 1990. On improved operational
the National Science and Engineering Research Council hydrologic forecasting: results from a WMO real-time forecasting
(NSERC) of Canada, the CEMA consortium, and the Depart- experiment. J. Hydrol. 114, 17–45.
ment of Civil and Geological Engineering, University of Georgakakos, K.P., Bae, D.H., Cayan, D.R., 1995. Hydroclimatology
of continental watersheds, temporal analyses. Water Resour.
Saskatchewan.
Res. 31 (3), 655–675.
Grayson, R.B., Bloschl, G., Moore, I.D., 1995. Distributed param-
References eter hydrologic modelling using vector elevation data: Thales
and TAPES-C. In: Singh, V.P. (Ed.), Computer Models of
Anctil, F., Tape, D.G., 2004. An exploration of artificial neural Watershed Hydrology. Water Resources Pub. Highlands Ranch,
network rainfall–runoff forecasting combined with wavelet Colorado, pp. 669–695.
decomposition. J. Environ. Eng. Sci., S121–S128. doi:10.1139/ Gupta, M.M., Jin, L., Homma, N., 2003. Static and Dynamic Neural
S03-07. Networks. John Wiley and Sons, Inc., Hoboken, New Jersey.
ASCE Task Committee on Application of Artificial Neural Networks in Haigh, M.J., 2000. The Aims of Land Reclamation, Land Recon-
Hydrology, 2000. Artificial neural networks in hydrology. I: struction and Management, vol. 1. A.A. Balkema Publishers,
Preliminary concepts. J. Hydrol. Eng. 5 (2), 115–123. Rotterdam, The Netherlands, pp. 1–20.
Barbour, S.L., Boese, C., Stolte, B., 2001. Water balance for Haykin, S., 1999. Neural Networks: A Comprehensive Foundation,
reclamation covers on oilsands mining overburden piles. In: second ed. MacMillan, New York.
Proceedings of the 54th Canadian Geotechnical Conference, Hsu, K.L., Gupta, H.V., Sorooshian, S., 1995. Artificial neural
Calgary, Alta., pp. 16–19 September 2001. Canadian Geotech- network modeling of the rainfall–runoff process. Water Resour.
nical Society, Alliston, Ont. pp. 313–319. Res. 31 (10), 2517–2530.
18 A. Elshorbagy, K. Parasuraman
Karunanithi, N., Grenney, W.J., Whitley, D., Bovee, K., 1994. Redlapalli, S.K., 2004. Development of Neural Units with Higher-
Neural networks for river flow prediction. J. Comput. Civil Eng. 8 order Synaptic Weights Operations and Their Applications to
(2), 201–220. Logic Circuits and Control Problems, M.Sc. Thesis, University of
Kim, T.W., Valdés, J.B., 2003. Nonlinear model for drought Saskatchewan, Saskatoon, SK, Canada.
forecasting based on a conjunction of wavelet transforms and Shamseldin, A.Y., 1997. Application of a neural network technique
neural networks. J. Hydrol. Eng. 8 (6), 319–328. to rainfall–runoff modeling. J. Hydrol. 199, 272–294.
Kitanidis, P.K., Bras, R.L., 1980. Real time forecasting with a Taylor, J.G., Commbes, S., 1993. Learning higher order correla-
conceptual hydrologic model, 2. Applications and results. Water tions. Neural Networks 6 (3), 423–428.
Resour. Res. 16 (6), 1034–1044. Thirumalaiah, K., Deo, M.C., 1998. River stage forecasting using
Kuczera, G., 1983. Improved parameter inference in catchment artificial neural networks. J. Hydrol. Eng. 3 (1), 26–32.
models; 2. Combining different kinds of hydrologic data and Wood, E.F., Lettenmaier, D.P., Zartarian, V.G., 1992. A land-
testing their compatibility. Water Resour. Res. 19 (5), 1163–1172. surface hydrology parameterization with subgrid variability for
Makkeasorn, A., Chang, N.B., Beaman, M., Wyatt, C., Slater, C., general circulation models. J. Geophys. Res. 97, 2717–2728.
2006. Soil moisture estimation in a semiarid watershed using Wooldridge, S.A., Kalma, J.D., Walker, J.P., 2003. Importance of
RADARSAT-1 satellite imagery and genetic programming. Water soil moisture measurements for inferring parameters in hydro-
Resour. Res. 42, W09401. doi:10.1029/2005WR00403. logic models of low-yielding ephemeral catchments. Environ.
Parasuraman, K., Elshorbagy, A., 2007. Cluster-based hydrologic Model. Software 18 (1), 35–48.
prediction using genetic algorithm-trained neural networks. J. Zhang, B., Govindaraju, S., 2000. Prediction of watershed runoff
Hydrol. Eng., ASCE 12 (1), 52–62. using Bayesian concepts and modular neural networks. Water
Parasuraman, K., Elshorbagy, A., Carey, S., 2006. Spiking-modular Resour. Res. 36 (3), 753–762.
neural networks: A neural network modeling approach for Zhang, M., Fulcher, J., Scofield, R.A., 1997. Rainfall Estimation
hydrological processes. Water Resour. Res. 42, W05412. Using Artificial Neural Network Group Neurocomputing, vol. 16.
doi:10.1029/2005WR00431. pp. 97–115.