You are on page 1of 10

agricultural water management 95 (2008) 123–132

available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/agwat

A distributed agro-hydrological model for irrigation


water demand assessment

M. Minacapilli a,*, M. Iovino a, G. D’Urso b


a
Dipartimento di Ingegneria e Tecnologie Agro-Forestali, University of Palermo, Italy
b
Dipartimento di Ingegneria Agraria e Agronomia del Territorio, University of Naples Federico II, Italy

article info abstract

Article history: The actual irrigation water demand in a district in Sicily (Italy) was assessed by the spatially
Received 6 February 2007 distributed agro-hydrological model SIMODIS (SImulation and Management of On-Demand
Accepted 12 September 2007 Irrigation Systems). For each element with homogeneous crop and soil conditions, in which
Published on line 7 November 2007 the considered area can be divided, the model numerically solves the one-dimensional
water flow equation with vegetation parameters derived from Earth Observation data. In
Keywords: SIMODIS, the irrigation scheduling is set by means of two parameters: the threshold value of
Irrigation scheduling soil water pressure head in the root zone, hm, and the fraction of soil water deficit to be re-
Distributed hydrological model filled, D. This study investigated the possibility of identifying a couple of irrigation para-
Remote sensing meters (hm, D) which allowed to reproduce the actual irrigation water demand, given that the
On-demand irrigation system study area was adequately characterized with regard to the spatial distribution of the soil
hydraulic properties and the vegetation conditions throughout the irrigation season. The
spatial distribution of the soil and vegetation properties of the study area, covering an
irrigation district of approximately 800 ha, was accurately characterized during the summer
of 2002. The soil hydraulic properties were identified by an intensive undisturbed soil
sampling, while the vegetation cover was characterized in terms of leaf area index, surface
albedo and fractional soil cover by analysing multispectral LandSat TM imageries. Irrigation
volumes were monitored at parcel scale.
A reference scenario with hm = 700 cm and D = 50% (corresponding to a mean actual to
potential transpiration ratio of 0.95) allowed to reproduce the spatial and temporal dis-
tribution of the actual irrigation demand at the district scale. The spatial variability of the
crop conditions in the considered area had much more influence to assess the irrigation
water demand than the soil hydraulic spatial variability. The proposed approach showed
that, under the agro-climatic conditions typical for the Mediterranean region, SIMODIS may
be a valuable tool in managing irrigation to increase water productivity.
# 2007 Elsevier B.V. All rights reserved.

1. Introduction the development of physically based models for simulating


the process of exchange of mass and energy in the soil–plant–
The issue of efficient water use for irrigation is of utmost atmosphere (SPA) system (Feddes et al., 1978; Bastiaanssen
importance in the semi-arid Mediterranean regions, where et al., 2004). In particular, deterministic models have been
water scarcity often causes severe damages in the fragile agro- proposed to simulate all the components of the water balance
ecosystems. In the last two decades, this evidence has induced in great detail, including crop growth, irrigation and solute

* Corresponding author. Tel.: +39 091 7028112; fax: +39 091 484035.
E-mail address: minacap@idra.unipa.it (M. Minacapilli).
0378-3774/$ – see front matter # 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.agwat.2007.09.008
124 agricultural water management 95 (2008) 123–132

transport (Vanclooster et al., 1994; van Dam et al., 1997; the simulated and measured irrigation volumes for the
Droogers et al., 2000; Ragab, 2002). These models have been considered area.
developed for site-specific applications but they have
seldom been applied to large areas, due to the complexity
in the acquisition of all required input data, often 2. The SIMODIS procedure
characterized by spatial and temporal variability. To
overcome this problem, techniques have been proposed SIMODIS was conceived as a Decision Support System for the
which involve the use of geostatistical approaches to simulation of the operation of an irrigation network based on
spatialize the measured soil hydraulic properties (Warrick the following aspects: (i) spatial and temporal variability of the
and Nielsen, 1980; Wosten et al., 1986) integrated with data soil water deficit; (ii) water resources availability; (iii) struc-
derived from Earth Observation to gather quantitative tural capability of the conveyance and distribution irrigation
information on the temporal and spatial distribution of network.
selected vegetational parameters (Choudhury et al., 1994; The first issue is modelled by considering distinct sepa-
Moreno et al., 2002). rated elements with homogeneous crop, soil and climate
The knowledge of the spatial distribution of the soil and conditions. Assuming that the soil water flow in each element
vegetation parameters is of crucial importance for the is mainly a one-dimensional process, the SWAP model (van
assessment of crop water use and for the definition of Dam et al., 1997, Kroes et al., 2000) is applied in distributed
management strategies aiming at improving the efficiency form. In each element, the vertical distributions of the water
of irrigation. The spatial distribution of the irrigation water use content, u (L3 L3), and pressure head, h (L), along the soil
is influenced by two components: a first one is determinis- profile at various times, are obtained by numerically solving
tically linked to the physical processes involved, which the equation:
themselves are depending on the spatial and temporal
  
variability of climate, soil and vegetation; a second stochastic @h @ @h
CðhÞ ¼ KðhÞ þ 1 þ SðhÞ (1)
component is represented by the farmers’ behaviour in the @t @z @z
irrigation practice.
At temporal scales larger than a few days, i.e. one week, under prescribed upper and lower boundary conditions. In
and in presence of an on-demand irrigation scheduling, the Eq. (1) z (L) is the vertical coordinate, assumed positive
first component is largely predominant on the second one. upwards, t (T) is time, C (L1) is the differential moisture
On the condition that the spatial variability of climate can be capacity, K(h) (L T1) is the soil hydraulic conductivity function
neglected compared to that of soil and vegetation, the and S (T1) is a root uptake term that, for the case of uniform
irrigation scheduling can be described by means of two root distribution, is defined by the following equation:
parameters: the threshold value of soil water pressure head
in the root zone, hm, and the fraction of soil water deficit to Tp
SðhÞ ¼ aw ðhÞ (2)
be re-filled with irrigation, D. The first parameter, hm, is crop jzr j
dependent, since it describes the crop tolerance to water
stress; the second, D, is related to the soil hydraulic in which Tp (L T1) is the potential transpiration, zr (L) the
properties. Both parameters, to a certain extent, are site- rooting depth and aw is a h-dependant reduction factor which
specific, in the sense that they reflect the average perception accounts for water deficit and oxygen stress (Feddes et al.,
of farmers to the soil water dynamics and their attitude to 1978). The solution of Eqs. (1) and (2) requires the knowledge of
induce or not water stress for quantitative and qualitative the soil hydraulic properties, i.e. the soil water retention curve,
production issues. This aspect has particular relevance u(h), and the soil hydraulic conductivity function, K(h); there-
when determining irrigation strategies for vineyards where fore, detailed field and/or laboratory investigations are
drought is controlled during fruit ripening in order to needed.
improve the wine quality. Applications of agro-hydrological Boundary conditions are defined according to the proce-
models to simulate the soil water balance for grapes and dure described by D’Urso et al. (1999). In particular, the upper
other fruit-tree crops were carried out by several authors boundary condition depends on rainfall and potential
(Bastiaanssen et al., 1996; Droogers et al., 2000; D’Urso, 2001; evapotranspiration rate, ETp (L T1), that is obtained by the
Ben-Asher et al., 2006). product of the crop coefficient Kc and the Penman-Monteith
Assessment of hm and D represents the basis for the reference evapotranspiration rate, ET0 (L T1) (Allen et al.,
simulation model SIMODIS (SImulation and Management of On- 1998). In SIMODIS, ET0 is used as a lumped input, while the
Demand Irrigation Systems), developed by D’Urso (2001). In this spatial distribution of the crop coefficient Kc is obtained using
study, the SIMODIS procedure was applied to simulate the a remote sensing approach based on the analytical relation-
spatial and temporal variability of the water demand of ship proposed by Stanghellini et al. (1990). It requires the
vineyards covering an area of approximately 800 ha located knowledge of canopy parameters such as the leaf area index,
in the south-west coast of Sicily (Italy). The increasing LAI, the fractional vegetation cover, the surface albedo, and
diffusion of the wine-specialized, high-income grape culti- crop height. According to Allen et al. (1998), a standard
vation in this area emphasized the need of irrigation minimum stomatal resistance of 100 s m1 was used in this
strategies aimed to optimize available water during drought study. The LAI is also used to fraction the ETp into potential
periods. Available water consumptions for the season 2002 transpiration rate, Tp (L T1) and potential soil evaporation
allowed to compare the spatial and temporal distributions of rate, Ep (L T1) (Choudhury et al., 1987). To this aim, assuming
agricultural water management 95 (2008) 123–132 125

that the net radiation inside the canopy decreases according 3. Materials and methods
to an exponential function and the soil heat flux can be
neglected, the following equations can be adopted (Ritchie, 3.1. The study area
1972):
The SIMODIS procedure was applied to simulate seasonal
Ep ¼ ETp expðkgr LAIÞ (3) irrigation water use of 136 grape fields distributed within a
850 ha irrigation district of the ‘‘Basso Belice’’ Consortium
Tp ¼ ETp  Ep (4) (Fig. 1) located on the southern coast of Sicily. The climate in
this area is typically Mediterranean: in 2002 the total annual
where kgr is the extinction coefficient for global solar radiation rainfall was 654 mm and the reference evapotranspiration
(Campbell and Norman, 1998). 1077 mm. Apart from wine grapes that occupy 46% of the
Several studies showed the possibility of detecting the total area, other crops are olives, citrus, artichoke and
above-mentioned canopy parameters from remote sensing wheat. The soils of the area are mainly alluvial soils formed
data with different spatial and temporal resolution. In from river deposits which overlay a deep and highly
particular, the surface reflectance in different bands of visible conductive calcarenitic aquifer. A specifically conducted
and infrared spectrum can be related to the LAI (Baret and soil survey recognized eight soil units (Fig. 1). Association of
Guyot, 1991). In the SIMODIS procedure, D’Urso (2001) brown soils, occurring in about 57% of the area, are located
suggested the use of a semi-empirical relationship for the on a flat fluvial terrace in the west; the textures range from
retrieval of the LAI from the Weighted Difference Vegetation Index sandy loam to loam. Vertisols (43% of the total area) occur in
(WDVI) (Clevers, 1989). Using remote sensing data and image- the hilly eastern part of the area and the texture is mainly
processing techniques, the spatial distributions of ETp can be silty clay loam. Irrigation water is delivered by means of a
derived for a given set of climatic variables. Then, these maps pressurised pipe network on the basis of an ‘‘on demand’’
can be used to set the upper boundary condition of the soil system. Drip irrigation is the most used irrigation system for
water flow model in each elementary unit in which the grape.
irrigation system can be divided.
Different options are available in SIMODIS to schedule 3.2. Data collection
irrigation (i.e. determining irrigation times and specific
volumes); for the purpose of this study, as mentioned in the The data set for the application of SIMODIS includes four main
introduction, two parameters needed to be defined: the information types: (i) soil hydraulic parameters; (ii) ground-
threshold value of soil water pressure head in the root zone, water data for the lower boundary condition; (iii) vegetation
hm, as timing criterion, and the fraction of soil water deficit, D, parameters for the upper boundary condition; (iv) irrigation
to be re-filled with irrigation water as volume criterion. volumes.

Fig. 1 – Map of the study area with indication of the soil units and sampling locations.
126 agricultural water management 95 (2008) 123–132

Table 1 – Soil characteristics and hydraulic parameters according to van Genuchten (1980)
Soil units Na Cl (%) Si (%) Sa (%) BD ln Ks uS (m3 m3) ur (m3 m3) n a (cm1)
(g cm3) (Ks in cm d1)

Calcareous 13 8.4 35.2 56.4 1.293 3.9(24) 0.464(10) 0.095(45) 1.317(9) 0.091(45)
brown soils
Vertisols 13 30.7 45.6 23.8 1.270 1.8(36) 0.491(6) 0.067(101) 1.150(5) 0.123(66)
Idromorphic 8 9.6 23.4 66.9 1.336 3.6(16) 0.445(10) 0.101(27) 1.546(8) 0.039(58)
brown soils
Vertex brown soils 7 38.1 40.1 21.8 1.302 1.6(104) 0.498(10) 0.021(260) 1.123(3) 0.089(85)
Lixiviate soils 2 24.9 19.9 55.2 1.205 3.7 0.410 0.138 1.295 0.087
Soils’ complex 2 12.6 43.7 43.7 1.164 3.3 0.490 0.090 1.271 0.091

Reported data are averages of the results obtained for the N samples collected within each soil unit. The coefficients of variation (%) are given
in parentheses.
a
N: Sample number; Cl: clay; Si: silt; Sa: sand; BD: soil bulk density; see text for other symbols.

The distribution of soil hydraulic properties within the area water content, us (L3 L3) and ur (L3 L3), and two empirical
was deduced from laboratory measurements performed on parameters, a (L1) and n, affecting the shape of retention
undisturbed soil cores (D’Urso et al., 2005) collected according curve, with a > 0 and n > 1) and the measured Ks values. A
to information retrieved from the soil survey map (Fig. 1). For representative set of parameters was obtained for each soil
each soil core (diameter d = 8.5 cm, height L = 5 cm), saturated unit by simply averaging the soil hydraulic parameters
hydraulic conductivity, Ks (L T1) was determined by the estimated for the cores collected in that soil unit (Table 1).
standard constant head technique (Reynolds et al., 2002); This choice was imposed by both the limited variability of
water contents, u, corresponding to pressure head values estimated/measured parameters and the restricted number of
ranging from 5 to 15,300 cm were determined by both a samples within some soil units that prevented a more detailed
hanging water column apparatus (Burke et al., 1986) and a spatial analysis. A profile depth of 3 m, with uniform soil
pressure plate apparatus (Dane and Hopmans, 2002). The hydraulic properties was considered for each field.
water retention function of van Genuchten (1980) was fitted to Monitoring of three wells within the irrigated area during
the measured u–h values using the RETC (RETention Curve) 2002 showed that the water table was always deeper than
code (van Genuchten et al., 1991). According to the Mualem- 4.5 m from the soil surface. Therefore, a gradient of 1 was
van Genuchten model (van Genuchten, 1980), the unsaturated accepted as the lower boundary condition in SIMODIS.
hydraulic conductivity relationship was described by the The definition of upper boundary condition in terms of
water retention parameters (i.e. the saturated and residual potential evapotranspiration fluxes was performed by a

Fig. 2 – Spatial distribution of LAI derived from the LandSat TM image of 07/07/2002.
agricultural water management 95 (2008) 123–132 127

Table 3 – Mean values of the specific irrigation volume, V,


and total irrigation volume, I, supplied by each water
application with the most widely applied irrigation
strategies in the study area
Water I3 strategy I4 strategy
application
DOYa V (mm) I (m3) DOY V (mm) I (m3)

1 183 20.6 158 179 18.9 180


2 197 19.2 144 190 14.4 162
3 218 20.0 158 212 17.6 130
4 – – – 224 12.8 144
a
Day of the year.

with a* = 0.255 and WDVI1 = 0.57, the LAI map with a resolu-
tion of 30 m (pixel size) was obtained for each available Land-
Sat TM image (Fig. 2). The resulting pixel-scale LAI values were
Fig. 3 – Variation of the mean crop coefficient values, Kc, for averaged within each field to calculate the mean Kc and ETp
the 136 grape fields considered. values. The spatial variability of Kc for each date was higher
during the peak growth (Fig. 3). The missing Kc value at the end
of the cropping season (day of the year, DOY = 320) was
derived from literature data (Allen et al., 1998). Others crop
multi-temporal analysis of LandSat TM images at three times parameters required by SIMODIS, i.e. kgr, zr and the critical
during the growing season: (a) 13 February 2002, initial stage; pressure head values defining the reduction factor aw in Eq. (2)
(b) 27 May 2002, prebloom stage; (c) 7 July 2005, fruit stage. The were taken from specific literature (Taylor and Ashcroft, 1972;
image processing produced maps of the canopy parameters of Doorenbos and Kassam, 1979; Wesseling et al., 1991) with
the entire area. The simplified model CLAIR (Clevers Leaf Area some minor adaptations based on local observations and
Index Relationship) (Clevers, 1989) was used to derive the knowledge. The above-mentioned crop parameters for the
spatial distribution of the leaf area index. The variation of the case study are reported in Table 2. GIS tools were used to
radiation absorption and reflection in a canopy that partially associate to each field the combination of crop and soil para-
covers the soil is described by the following semi-empirical meters according to the input scheme of SIMODIS (D’Urso,
relationship: 2001).
Irrigation water volumes were monitored in all the 136
 
1 WDVI selected grape fields during the 2002 season. From a
LAI ¼  ln 1  (5)
a WDVI1 preliminary analysis of this data set, irrigation applications
in the grape fields occurred from the end of June (around DOY
in which WDVI is derived from LandSat TM3 and TM4 bands 179) to the first decade of August (around DOY 224). In
(Clevers, 1989); WDVI1 is the asymptotical value of WDVI for particular, in 73% of the cases, two typical irrigation strategies
LAI!1, and a* is an extinction coefficient, expressing the
increase of LAI for a unit increase of WDVI, that has to be
estimated from simultaneous measurements of LAI and
WDVI. In our case, the calibration of Eq. (5) was performed
by in situ LAI measurements collected with the portable
instrument LAI2000 (Li-Cor, Lincoln, NE). A complete descrip-
tion of the analysis conducted for the study area is given in
Minacapilli et al. (2005, 2006). By using the calibrated Eq. (5)

Table 2 – Main crop parameters used for the wine grape


of the study area
Parameters Value

Critical pressure heads (cm)


h2 (h below which optimum water 30
uptake starts in the root zone)
h3 (h below which optimum water 600 Fig. 4 – Relative transpiration, T/Tp, obtained from
uptake reduction starts in the root zone)
preliminary simulations. Dots indicate the mean values of
h4 (wilting point, no water 8000
uptake at lower pressure heads)
the specific irrigation volume for the most widely used
irrigation strategies. Vertical and horizontal bars indicate
kgr (extinction coefficient) 0.5
variability in specific volume and irrigation time,
zr (rooting depth) (cm) 45
respectively.
128 agricultural water management 95 (2008) 123–132

Table 4 – Irrigation scheduling parameters selected for


the different simulated scenarios
Scenario hm (cm) D (%) RMSEa (m3) Ev (%)

1. (T/Tp = 0.97) 450 45 3754 25


2. (T/Tp = 0.95) 700 50 2814 8
3. (T/Tp = 0.90) 1000 55 4163 32
4. (T/Tp = 0.85) 1250 58 4403 36
a
RMSE: Root mean square error of simulated daily irrigation
volume compared to the measured value; Ev: difference between
simulated and measured seasonal cumulative irrigation volume.

The daily values of relative transpiration, i.e. the actual to


potential transpiration ratio, T/Tp, for the two irrigation
strategies are plotted in Fig. 4. These plots evidence a similar
temporal distribution of daily T/Tp values with a mean
seasonal value of about 0.95 in both cases. It should be noted
that a value of T/Tp = 0.95 was adopted by Droogers et al. (2000)
in their application of SWAP made on grape in Turkey as the
lower threshold to start irrigation. Our results seem to
confirm that, in the Mediterranean areas, a value of T/
Tp = 0.95 may be assumed as characteristic to schedule
irrigation for grape.
The relationships between the daily T/Tp values and the
two irrigation parameters hm and D are shown in Fig. 5a and
b. From the plot in Fig. 5a can be noticed that the relative
transpiration values strongly depend on hm in the entire
range of pressure head values from 0 to 2500 cm. Relative
transpiration dependence on the water deficit is otherwise
limited to D values greater than 50%. For D < 50%, T/Tp is
independent from D and equal, on average, to 0.95. For both
irrigation strategies, most of the variability of hm is
comprised between 300 and 1300 cm; in correspondence
of the same interval, the value of D varies between 40
and 60%.
On the basis of this preliminary analysis, four different
scenarios were simulated, each one with a different couple of
Fig. 5 – Daily values of relative transpiration, T/Tp, vs. (a) irrigation parameter values (hm, D), corresponding to a range
mean soil pressure head values, hm, and (b) fraction of of mean seasonal T/Tp values between 0.85 and 0.97. The four
water deficit values, D, both computed in the root zone of water management scenarios, that were used in SIMODIS to
the soil profile. simulate the farmers’ water demand at district scale, are
summarized in Table 4.

(named I3 and I4) were practiced, respectively with 3 and 4


water applications (Table 3).

3.3. Application of SIMODIS

In order to identify the values of the irrigation scheduling


parameters hm (mean soil pressure head in the root zone)
and D (fraction of the water deficit in the root zone), a
preliminary simulation was performed on an ideal grape
field with soil hydraulic functions corresponding to those of
the prevailing soil type in the study area and averaged
vegetation parameters. By using the irrigation dates and
supply volumes derived from the observed data, the daily
values of hm and D were calculated. For the sake of simplicity
only the two most widely applied irrigation strategies Fig. 6 – Comparison between measured and simulated
(Table 3) were considered. daily irrigation volumes.
agricultural water management 95 (2008) 123–132 129

Fig. 7 – Comparison between measured and simulated


cumulative irrigation volumes.

Fig. 8 – Comparison between measured and simulated


seasonal irrigation volumes at secondary unit level.
4. Results and discussion

The simulations with SIMODIS were made for each scenario of


Table 4, represented by a couple of irrigation parameter the average behaviour of farmers in irrigation scheduling, and
values, hm and D. Then, the simulated daily irrigation volumes they can be varied in order to simulate alternative scenarios of
at the district scale were compared with the measured data. water management.
The scenario resulting in the lowest value of the root mean The monthly and annual water balance components for the
square error (RMSE) between simulated and measured daily 136 grape fields obtained using the reference scenario are
irrigation volumes was scenario 2 (RMSE = 2814 m3), corre- summarized in Table 5. The water use efficiency could be
sponding to a value of T/Tp = 0.95 (Fig. 6), in agreement with the evaluated from the components of the water balance distin-
findings of the preliminary analysis. For this scenario, that we guishing between, for instance, soil evaporation and crop
can now consider as the ‘‘reference’’ one, the estimated mean transpiration, groundwater recharge and capillary rise. In our
value of the seasonal specific irrigation volume was 619 m3/ha, case, the actual transpiration varied from April to the end of
near to the mean measured value of 574 m3/ha. A good September according the phenological phases of grape. The
agreement was also found with regard to the temporal total simulated irrigation supply of 60 mm was distributed
evolution of the cumulative water consumption at district during the months of June, July and August in agreement with
scale (Fig. 7), with a difference between simulated and the observed data, while significant downward fluxes through
measured seasonal cumulated irrigation volumes less than the bottom of the profile occurred only in December because of
8%. The couple of irrigation parameter values for the reference the high rainfall (194 mm). No runoff occurred as, for the
scenario (hm = 700 cm; D = 50%) can be adopted to represent observed rainfall intensities, water always infiltrated in the soil.

Table 5 – Monthly and total annual water balance components computed by SIMODIS for the 136 grape fields using
scenario 2
Precipitation ET0 Irrigation Actual transpiration Actual Downward flux at profile
(mm) (mm) (mm) (mm) evaporation (mm) bottom (mm)

January 43.4 25.8 0.0 0.0 5.1 3.0 (78)


February 22.3 47.3 0.0 0.0 7.8 1.9 (65)
March 9.8 77.9 0.0 0.5 7.7 1.7 (59)
April 45.8 94.6 0.0 3.8 (32) 9.2 1.4 (55)
May 64.2 140.3 0.0 19.7 (38) 21.5 1.3 (54)
June 8.0 171.9 8.6 (194) 34.0 (24) 10.2 (14) 16.2 (56)
July 18.0 176.4 37.4 (31) 35.7 (18) 17.4 (15) 11.2 (56)
August 11.2 138.8 14.0 (133) 28.3 (18) 12.6 (28) 8.2 (41)
September 30.8 89.2 0.0 14.7 (15) 10.9 (16) 12.0 (22)
October 70.0 60.5 0.0 4.0 (20) 14.7 (4) 1.8 (91)
November 111.8 39.1 0.0 0.0 6.8 1.8 (79)
December 194.4 17.8 0.0 0.0 2.6 102.2 (39)

Total 630 1079 60 141 123 163

Coefficients of variation (%) are given in parentheses.


130 agricultural water management 95 (2008) 123–132

The reliability of the results obtained at the district level study area. These plots are useful to distinguish sub-areas
using scenario 2 combined with the spatial distributed with different hydrological response to irrigation manage-
approach for both soil and plant characterization were also ment.
confirmed at the secondary unit level, that is a cluster of plots The reliability of these results shows the importance of the
served by a common node of the pipeline network. By the knowledge of spatial soil and vegetation distribution in
considered spatial distributed approach, the prediction of simulating the spatial and temporal variation of water
seasonal irrigation volumes at secondary unit level was very demand at district scale. In order to evaluate the relevance
satisfactory resulting in a significant determination coefficient of the distributed approach used to define the soil hydraulic
R2 = 0.84 between simulated and measured values (Fig. 8). This parameters and the upper boundary conditions, two further
result could not be confirmed at field scale where the scenarios were simulated, characterized by two different
differences between simulated and measured irrigation levels of soil–crop information:
volumes were attributed to different management behaviours
depending on water availability as well as on farmers’ 1) lumped crop, with a unique Kc value for the entire irrigation
subjectivity to recognize the crop water requirement. district, varying in time according to the mean distribution
In terms of spatial distributions, the outputs of SIMODIS reported in Fig. 3, combined with spatially distributed soil
allowed to obtain a set of thematic maps useful to irrigation hydraulic parameters (Table 1);
management. In Fig. 9, the spatial distributions of the annual 2) lumped soil, with soil hydraulic properties averaged for the
values of actual transpiration, actual evaporation, irrigation entire district area, combined with the spatially distributed
and flux at the bottom of the soil profile are mapped for the Kc values obtained from remote sensing.

Fig. 9 – Spatial distributions of annual values of selected components of the water balance computed by SIMODIS using
scenario 2: (a) downward flux at profile bottom; (b) irrigation; (c) actual evaporation; (d) actual transpiration.
agricultural water management 95 (2008) 123–132 131

scheduling in SIMODIS was set on the basis of the simulated


threshold values of the soil water pressure head in the root
zone, hm, and the application volume, D, taken as a percentage
of the corresponding soil water deficit. The findings of this
study confirmed that the two irrigation parameters (hm, D) can
be tuned to adequately reproduce the spatial and temporal
evolution of crop water use. A simple methodology based on
the evaluation of the daily values of the relative transpiration,
T/Tp, was proposed to identify the hm, and D values resulting in
the best agreement between measured and simulated irriga-
tion volumes.
The spatial distributed approach adopted in SIMODIS
allowed a reliable simulation of the spatial and temporal
variation of water demand at district scale. For the case study,
the spatial variability of crop conditions was more significant
Fig. 10 – Daily irrigation volumes simulated by the
than the soil hydraulic spatial variability, showing the
‘‘reference’’ scenario 2 and the lumped soil and vegetation
potentiality of remote sensing techniques to gather spatial
scenarios.
and temporal distributions of vegetation parameters, like Kc.
With these characteristics, the proposed approach can be
effectively used for supporting the decision making process in
The first simulation corresponds to a qualitatively high the management of irrigation water resources, and for
soil information, compared to a poor crop information, improving the efficiency of the irrigation systems.
based solely on land-use; this refers to a simulation in which
vegetation parameters are extracted from tabulated data on
crop phenology, independently from their actual develop- Acknowledgement
ment. The second simulation refers to a limited information
on soil properties combined with a high-quality level The authors are grateful to Dr. Claudio Gucciardi for his help in
information on the spatio-temporal distribution of crop collecting irrigation data and setting up the GIS tools.
properties. The comparison among these two simulations
and the ‘‘reference’’ scenario 2 is shown in Fig. 10. The plot
evidences that, for this case study, the spatial variability of references
vegetation parameters, expressed by Kc, has greater influ-
ence on simulated water requirement than the soil
hydraulic characteristics. Indeed, when we adopted a Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop
lumped crop coefficient, even with the same set of irrigation evapotranspiration. FAO Irrigation and Drainage Paper 56,
Food and Agricultural Organization of the United Nations,
parameters of the reference scenario, we obtained a
Rome, 300 pp.
different temporal distribution of irrigation volumes that Baret, F., Guyot, G., 1991. Potential and limits of vegetation
resulted in an overprediction around DOY 190, and also in indices from LAI and APAR assessment. Remote Sens.
different seasonal cumulative volumes. In contrast, adop- Environ. 35, 161–173.
tion of a lumped soil characterization resulted in a much Bastiaanssen, W.G.M., El Deen Milad Soliman, K., Mirabile, C.,
smaller difference with the reference scenario. These results 1996. Data management related to the application of two
evidenced that, in the agro-climatic conditions of the study crop–water–environment models in Argentina and Egypt. In:
Ragab, R., El Din-El Quosy, D., Van den Broek, B. (Eds.), Crop-
area, a good knowledge of the spatial distribution of crop
water-environment models: selected papers of the 16th ICID
properties is likely to be more important than an accurate congress, Cairo, Egypt, 17 September 1996, pp. 54–68.
soil hydraulic characterization. Bastiaanssen, W.G.M., Allen, R.G., Droogers, P., D’Urso, G.,
Steduto, P., 2004. In: Feddes, R.A., de Rooij, G.H., van Dam,
J.C. (Eds.), Inserting Man’s Irrigation and Drainage Wisdom
5. Conclusions into Soil Water Flow Models and Bringing it Back Out: How
Far Have we Progressed? Unsaturated-zone Modeling:
Progress, Challenges and Applications. Kluwer Academic
The main aim of the study was the assessment of the water
Publishers, pp. 263–299.
management options in an on-demand irrigation system to Ben-Asher, J., van Dam, J.C., Feddes, R.A., Jhorar, R.K., 2006.
optimize available water in a droughty Mediterranean Irrigation of grapevines with saline water. II: Mathematical
environment. The model SIMODIS was used to simulate the simulation of vine growth and yield. Agric. Water Manage.
spatial and temporal variability of farmers’ water demand for 83, 22–29.
a set of 136 irrigated grape fields in an 850 ha irrigation district. Burke, W., Gabriels, D., Bouma, J., 1986. Soil Structure
Assessment. Balkema, Rotterdam, The Netherlands.
The spatial distributions of vegetation parameters, used in
Campbell, G.S., Norman, J.M., 1998. An Introduction to
SIMODIS to fix the upper boundary conditions, were obtained Environmental Biophysics. Springer, New York, 286 pp.
from Earth Observation data, while soil hydraulic properties Choudhury, B.J., Idso, S.B., Reginato, R.J., 1987. Analysis of an
were derived from laboratory measurements performed on 45 empirical model for soil heat flux under a growing wheat
undisturbed soil cores and from soil survey maps. Irrigation crop for estimating evaporation by infrared-temperature
132 agricultural water management 95 (2008) 123–132

based energy balance equation. Agric. For. Meteorol. 39, Moreno, J., Menenti, M., Richter, R., 2002. Assessment of inputs
283–297. to land surface processes models derived from
Choudhury, B.J., Ahmed, N.U., Idso, S.B., Reginato, R.J., hyperspectral multiangular data. In: Owe, M., D’Urso, G.
Daughtry, C.S.T., 1994. Relations between evaporation (Eds.), Remote Sensing for Agriculture, Ecosystems and
coefficients and vegetation indices studied by model Hydrology. III SPIE Proceedings Series No. 4542, Bellingham,
simulations. Remote Sens. Environ. 50, 1–17. pp. 223–236.
Clevers, J.G.P.W., 1989. The application of a weighted infrared- Ragab, R., 2002. A holistic generic integrated approach for
red vegetation index for estimating leaf area index by irrigation, crop and field management: the SALTMED
correcting for soil moisture. Remote Sens. Environ. 29, model. Environ. Model. Software 17, 345–361.
25–37. Reynolds, W.D., Elrick, D.E., Youngs, .G., Booltink, H.W.G.,
D’Urso, G., 2001. Simulation and management of on-demand Bouma, J., 2002. Saturated and field-saturated water flow
irrigation systems: a combined agrohydrological and parameters: laboratory methods. In: Dane, J.H., Topp, G.C.
remote sensing approach. Monography, Wageningen (Eds.), Methods of Soil Analysis. Part 4. Physical Methods.
University. ISBN 90-5808-399-3,174 pp. Book Series No. 5. Soil Science Society of America, Madison,
D’Urso, G., Iovino, M., Minacapilli, M., 2005. Applicazione della WI, pp. 802–817.
procedura SIMODIS per la gestione dell’irrigazione in un Ritchie, J.T., 1972. A model for predicting evaporation from a
comprensorio irriguo siciliano (Application of SIMODIS row crop with incomplete cover. Water Resour. Res. 8, 1204–
procedure for irrigation management in a Sicilian irrigated 1213.
area). In: Proceedings of the Congress ‘‘L’ingegneria agraria Stanghellini, C., Bosma, A.H., Gabriels, P.C.J., Werkhoven, C.,
per lo sviluppo sostenibile dell’area mediterranea’’. AIIA, 1990. The water consumption of agricultural crops: how
Associazione Italiana Ingegneria Agraria, 27–30 June 2005, crop coefficients are affected by crop geometry and
Catania, Italy (in Italian with English abstract). microclimate. Acta Hortic. 278, 509–515.
D’Urso, G., Santini, A., Menenti, M., 1999. Regional application of Taylor, S.A., Ashcroft, G.M., 1972. Physical Edaphology. The
one-dimensional water flow models for irrigation Physics of Irrigated and Non-irrigated Soils. Freeman, San
management. Agric. Water Manage. 40, 291–302. Francisco, CA, 563 pp.
Dane, J.H., Hopmans, J.W., 2002. Water retention and storage. van Dam, J.C., Huygen, J., Wesseling, J.G., Feddes, R.A., Kabat, P.,
Laboratory: pressure plate extractor. In: Dane, J.H., Topp, van Walsum, P.E.V., Groenendijk, P., van Diepen, C.A., 1997.
G.C. (Eds.), Methods of Soil Analysis. Part 4: Physical Theory of SWAP version 2.0. Simulation of water flow,
Methods. Book Series No. 5. Soil Science Society of America, solute transport and plant growth in the soil–water–
Madison, WI, pp. 688–690. atmosphere–plant environment. Technical Document 45.
Doorenbos, J., Kassam, A.H., 1979. Yield response to water. FAO Wageningen Agricultural University and DLO Winand
Irrigation and Drainage Paper 33, Food and Agricultural Staring Centre, The Netherlands.
Organization of the United Nations, Rome. van Genuchten, M.T., 1980. A closed form equation for
Droogers, P., Bastiaanssen, W.G.M., Beyazgul, M., Kayam, Y., predicting the hydraulic conductivity of unsaturated soils.
Kite, G.W., Murray-Rust, H., 2000. Distributed agro- Soil Sci. Soc. Am. J. 44, 892–898.
hydrological modeling of an irrigation system in western van Genuchten, M.T., Leij, F.J., Yates, S.R., 1991. The RETC code
Turkey. Agric. Water Manage. 43, 183–202. for quantifying the hydraulic functions of unsaturated soils.
Feddes, R.A., Kowalik, P.J., Zaradny, H., 1978. Simulation of field Report No. EPA/600/2-91/065, US Environmental Protection
water use and crop yield. Monographs, Pudoc (Centre for Agency, Office of Research and Development, Washington,
Agricultural Publishing and Documentation), Wageningen, DC.
189 pp. Vanclooster, M., Viane, P., Diels, J., Christiens, K., 1994. Wave. A
Kroes, J.G., Wesseling, J.G., van Dam, J.C., 2000. Integrated mathematical model for simulating water and
modelling of the soil–water atmosphere–plant system using agrochemicals in the soil and vadose environment. In:
the model SWAP 2.0, an overview of theory and an Reference and User’s Manual (release 2.0), Institute for Land
application. Hydrol. Process. 14, 1993–2002. and Water Management, Katholieke Universiteit Leuven,
Minacapilli, M., D’Urso, G., Qiang, L., 2005. Applicazione e Leuven, Belgium.
confronto dei modelli SAIL e CLAIR per la stima dell’indice Warrick, A.W., Nielsen, D.R., 1980. In: Hillel, D. (Ed.), Spatial
di area fogliare da dati iperspettrali MIVIS (A comparison Variability of Soil Physical Properties in the Field.
between SAIL and CLAIR models for LAI estimation from Application of Soil Physics. Academic Press, NY, pp. 319–
hyperspectral MIVIS data). Rivista Italiana di 344.
Telerilevamento 33, 15–25 (in Italian with English abstract). Wesseling, J.G., Elbers, J.A., Kabat, P., van den Broek, B.J., 1991.
Minacapilli, M., Iovino, M., D’Urso, G., 2006. Crop and irrigation SWATRE: Instructions for Input. Internal Note. Winand
water management using high resolution remote sensing Staring Centre, Wageningen, The Netherlands.
and agrohydrological models. In: D’Urso, G., Osann Jochum, Wosten, J.H.M., Bannik, M.H., De Gruijter, J.J., Bouma, J., 1986. A
M.A., Moreno, J. (Eds.), Earth Observation for Vegetation procedure to identify different groups of hydraulic-
Monitoring and Water Management. American Institute of conductivity and moisture-retention curves for soil
Physics, (0-7354-0346-5/06), pp. 99–105. horizons. J. Hydrol. 86, 133–145.

You might also like