Water productivity analysis of irrigated crops in Sirsa district, India

Agricultural Water Management 82 (2006) 253–278 www.elsevier.com/locate/agwat

Water productivity analysis of irrigated crops in Sirsa district, India R. Singh, J.C. van Dam *, R.A. Feddes Department of Environmental Sciences, Wageningen University and Research Centre, Nieuwe Kanaal 11, NL-6709 PA Wageningen, The Netherlands Accepted 24 July 2005 Available online 19 September 2005

Abstract Water productivity (WP) expresses the value or benefit derived from the use of water, and includes essential aspects of water management such as production for arid and semi-arid regions. A profound WP analysis was carried out at five selected farmer fields (two for wheat–rice and three for wheat– cotton) in Sirsa district, India during the agricultural year 2001–02. The ecohydrological soil–water– atmosphere–plant (SWAP) model, including detailed crop simulations in combination with field observations, was used to determine the required hydrological variables such as transpiration, evapotranspiration and percolation, and biophysical variables such as dry matter or grain yields. The use of observed soil moisture and salinity profiles was found successful to determine indirectly the soil hydraulic parameters through inverse modelling. Considerable spatial variation in WP values was observed not only for different crops but also for the same crop. For instance, the WPET, expressed in terms of crop grain (or seed) yield per unit amount of evapotranspiration, varied from 1.22 to 1.56 kg m
3 for wheat among different farmer fields. The corresponding value for cotton varied from 0.09 to 0.31 kg m
3. This indicates a considerable variation and scope for improvements in water productivity. The average WPET (kg m
3) was 1.39 for wheat, 0.94 for rice and 0.23 for cotton, and corresponds to average values for the climatic and growing conditions in Northwest India. Including percolation in the analysis, i.e. crop grain (or seed) yield per unit amount of evapotranspiration plus percolation, resulted in average WPETQ (kg m
3) values of 1.04 for wheat, 0.84 for rice and 0.21 for cotton. Factors responsible for low WP include the relative high amount of evaporation into evapotranspiration especially for rice, and percolation from field irrigations. Improving agronomic practices such as aerobic rice cultivation

* Corresponding author. Tel.: +31 317 484825; fax: +31 317 484885. E-mail address: [email protected] (J.C. van Dam). 0378-3774/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.agwat.2005.07.027

254

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

and soil mulching will reduce this non-beneficial loss of water through evaporation, and subsequently will improve the WPET at field scale. For wheat, the simulated water and salt limited yields were 20– 60% higher than measured yields, and suggest substantial nutrition, pest, disease and/or weed stresses. Improved crop management in terms of timely sowing, optimum nutrient supply, and better pest, disease and weed control for wheat will multiply its WPET by a factor of 1.5! Moreover, severe water stress was observed on cotton (relative transpiration < 0.65) during the kharif (summer) season, which resulted in 1.4–3.3 times lower water and salt limited yields compared with simulated potential yields. Benefits in terms of increased cotton yields and improved water productivity will be gained by ensuring irrigation supply at cotton fields, especially during the dry years. # 2005 Elsevier B.V. All rights reserved. Keywords: Water use efficiency; Soil water flow; Salinization; Simulation modelling; Field scale; Wheat; Rice; Cotton

1. Introduction Water scarcity threatens food security for millions of people, particularly in the arid and semi-arid regions. More than 65% of Haryana State (India) has an arid or semi-arid climate, where crop production is not possible without supplemental irrigation. Current wheat grain yields in Haryana are around 4.2 t ha
1 under irrigated conditions. A major constraint to increase the food grain production is limited surface water availability (Aggarwal et al., 2001). Furthermore, the current irrigation systems in Haryana State are causing problems of rising or declining groundwater levels, waterlogging and salinization (Aggarwal and Roest, 1996). Sirsa district, located in the western corner of Haryana State, represents the typical problems in Haryana of canal water scarcity, poor groundwater quality, rising or declining groundwater levels, waterlogging and secondary salinization, and limited crop production. These water management issues are very complex, and must be addressed by better planning and management. In order to improve water management and its productivity, we need to reveal the cause– effect relationships between hydrological variables such as evaporation, transpiration, percolation or capillary rise, and biophysical variables such as dry matter and grain yields under different ecohydrological conditions. Water productivity, a concept expressing the value or benefit derived from the use of water, includes various aspects of water management and is very relevant for arid and semi-arid regions (Molden and Sakthivadivel, 1999; Molden et al., 2001; Droogers and Bastiaanssen, 2002; Kijne et al., 2003). It can be expressed in terms of grain (or seed) yield per amount of water used in different processes such as transpiration, evapotranspiration and percolation, and provides a proper diagnosis of where and when water could be saved. Measurements of the required hydrological variables under field conditions are difficult, and need sophisticated instrumentation or installation of lysimeter. Moreover, field experiments yielding site-specific information are very expensive, laborious and time consuming to conduct for all ecohydrological conditions, especially if they should be representative for a sequence of years. However, ecohydrological models like the soil– water–atmosphere–plant (SWAP) model in combination with field experiments offer the opportunity to gain detailed insights into the system behaviour in space and time. Based on

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

255

the physical characterizations of the main hydrological, chemical and biological processes, the ecohydrological SWAP model (Van Dam et al., 1997) describes one-dimensional variable saturated water flow, solute transport and heat flow in top soils in relation to crop growth. SWAP includes both a simple and detailed crop module. The detailed crop module is based on the World Food Studies (WOFOST) model (Spitters et al., 1989; Supit et al., 1994), which simulates crop growth and its production accounting for water and salt stress of the crop (Van Dam et al., 1997; Kroes and Van Dam, 2003). Being capable to integrate any change to the system, SWAP can evaluate different management options like irrigation scheduling, cropping pattern, conjunctive use, etc. for effects on WP. However, the accuracy of these predictive models depends upon the proper identification of input parameters. In this paper, we present a methodology for water productivity analysis of irrigated crops using field observations and a simulation model. As an example, a profound analysis of input parameters and predicted results of the used SWAP model at farmer fields in Sirsa district (India) is presented. Most of the input parameters were measured directly in field experiments with high accuracy, some remained uncertain. Inverse modelling was used to determine indirectly the remaining uncertain soil hydraulic parameters (Jhorar, 2002; Ritter et al., 2003). The observed soil moisture and salinity profiles were used as system response in inverse modelling. SWAP was calibrated and validated using the observations at different farmer fields representing various combinations of soil, crop, and irrigation amount and quality. Finally, the water productivity and its variation for irrigated crops (wheat, rice and cotton) was simulated and analysed through the calibrated and validated SWAP with field data during the agricultural year 2001–02.

2. Materials and methods 2.1. Monitoring of farmer fields Farmer fields (F) at different sites (S) in Sirsa district were monitored in the framework of the Water Productivity (WATPRO) project during the agricultural year 2001–02 (Malik et al., 2003). Field observations at two wheat–rice fields (denoted as S1F1 and S2F5), and at three wheat–cotton fields (denoted as S3F11, S4F16 and S5F20) are used in this study. The monitored wheat–rice fields were located in the main wheat–rice belt along the Northern Ghagger canal downstream of the Ottu weir (Fig. 1). The heavy soil texture (i.e. clay loam to silt clay loam) and good quality groundwater are providing suitable growing conditions for wheat–rice in this area. The predominant wheat–cotton combination is cultivated on the light soils i.e. sandy loam to loamy sand. Early sowing (in October) of wheat is practised at wheat–rice fields, while late sowing (in November) at wheat–cotton fields. In year 2002, the sowing of kharif crops (cotton/ rice) was delayed by 15–20 days due to a late start of monsoon (rain). The late sowing resulted in the late harvesting. Based on the recorded sowing and harvesting dates, the experimental period for wheat–rice fields is defined from Oct 1, 2001 to Oct 15, 2002, which is further divided into two crop seasons: rabi (wheat) from Oct 1, 2001 to Apr 30, 2002, and kharif (rice) from May 1, 2002 to Oct 15, 2002. Similarly, for wheat–cotton

256

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Fig. 1. Location of farmer fields (F) at different sites (S), which were monitored in Sirsa district (Haryana), India during the agricultural year 2001–02.

fields, it is defined from Nov 1, 2001 to Nov 15, 2002 comprising the rabi (wheat) season from Nov 1, 2001 to Apr 30, 2002, and the kharif (cotton) season from May 1, 2002 to Nov 15, 2002. Table 1 gives an overview of the data collected for calibration and validation of SWAP, and for water productivity analysis of irrigated crops at field scale. The input parameters of SWAP can be categorized into meteorological, soil, water and crop parameters. The meteorological data, including minimum and maximum temperature, relative humidity, vapour pressure, sunshine hours, wind speed and rainfall, were collected from the meteorological station installed at the Indian Council of Agricultural Research-Cotton Research Institute (ICAR-CRS) (lat. 298350 N; long. 758080 E) in Sirsa district. The monitored farmer fields were in a range of 20–35 km from the meteorological station. Soil samples at the monitored fields were taken from five different soil depths: 0–15, 15–30, 30–60, 60–90 and 90–120 cm (Fig. 2). These samples were analysed for basic physio-chemical properties such as soil texture, bulk density, saturated hydraulic conductivity, saturation percentage, pH, electrical conductivity (EC) and organic matter. The saturation percentage ranges from 50 to 60% for heavy soils in wheat–rice fields, and from 31 to 40% for light soils in wheat–cotton fields. The bulk density ranges from 1.36

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

257

Table 1 Overview of the data collected for calibration and validation of SWAP model at farmer fields in Sirsa district (Malik et al., 2003) Data

Method/source

Frequency

Purpose

Meteorological data

Meteorological station

Daily

Input derivation

International pipette method (Piper, 1966), USDA classification Core method Constant water head method (Klute and Dirksen, 1986) Saturation paste method

Once

Input derivation

Once Once

Input derivation Input derivation

Once

Input derivation

Before and after irrigation Before and after irrigation Before and after irrigation Before sowing

Calibration and validation General Calibration and validation Input derivation

Current meter/co-ordinate method/volumetric method

3–4 times

Input derivation

Field observation Irrigation depth were calculated by multiplying the discharge and duration of irrigation and then divided by field area Conductivity meter

Each irrigation Each irrigation

Input derivation Input derivation

Each irrigation

Input derivation

Field observation

4–5 times

Input derivation

Field observation Field observation Field observation

4–5 times 4–5 times 4–5 times

Leaf area

Leaf area meter

4–5 times

Photo synthetically active radiation (PAR) Rooting depth

Sunscan canopy analysis system Field observation, Auger method Field observation

4–5 times

Input derivation Input derivation Calibration and validation Calibration and validation Input derivation

2–3 times

Input derivation

At harvest

Calibration and validation

Soil physio-chemical properties Texture Bulk density Saturated hydraulic conductivity Saturation percentage/moisture Soil moisture pH Electrical conductivity Organic carbon Irrigation regime Discharge of irrigation source, i.e. canal water or tubewell water Duration of irrigation Irrigation depth

Irrigation quality Crop growth parameters Crop development stage (in days after sowing), i.e. emergence, panicle initiation, anthesis, maturity and harvest Plant density and tillers Plant height Dry matter partitioning

Crop yields

Gravimetric method In soil–water suspension of 1:2 by pH meter In soil–water suspension of 1:2 by conductivity meter Wet digestion method (Jackson, 1973)

258

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Fig. 2. Soil information of the selected farmer fields in Sirsa district. The symbol C stands for clay, Si for silt, S for sand, and L for loam. Accordingly soil texture CL means clay loam, and SL sandy loam. BD is the bulk density (g cm
3) and OM is the percentage organic matter.

to 1.46 g cm
3 in wheat–rice fields, and from 1.52 to 1.67 g cm
3 in wheat–cotton fields. Most of the soils are low in organic matter, and sodic with a pH ranging from 8.0 to 9.0. The selected fields were also observed intensively in terms of soil moisture and salinity profiles before and after each irrigation event, mainly during the rabi season. The source (canal or tubewell), amount and quality of each irrigation were recorded. The total irrigation varied from 391 to 568 mm for wheat, 301 to 737 mm for cotton, and 1062 to 1250 mm for rice crop. Irrigation water at the selected fields came mainly from tubewells (groundwater). The groundwater quality at most of the fields is good (<2 dS m
1), except at S3F11 where it is marginal (3.73 dS m
1). Crop growth in terms of density (number of tillers per unit area), height, leaf area index, dry matter and its partitioning, and rooting depth at different crop development stages was measured. The Sunscan canopy analysis system was used for the measurement of light interception, i.e. photosynthetically active radiation (PAR) absorbed by the canopy. Total dry matter, grain (or seed) and straw yields were measured at harvest time of the crop. 2.2. Soil–water–atmosphere–plant (SWAP) model The soil–water–atmosphere–plant (SWAP) is an ecohydrological model based on the deterministic and physical laws for essential hydrological, chemical and biological processes occurring in the soil–water–plant–atmosphere continuum (Van Dam et al.,

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

259

1997). It simulates the vertical soil water flow and salt transport in close interaction with crop growth. Richards’ equation (Richards, 1931) is applied to compute transient soil water flow: 
 @h @ @h þ 1
Sa ðzÞ Cw ðhÞ ¼ (1) KðhÞ @t @z @z where Cw is the differential soil water capacity [L
1], h the soil water pressure head [L], K the hydraulic conductivity [L T
1], Sa the root water extraction rate [T
1], and z the vertical coordinate [L] (positive upward). The numerical solution of Eq. (1) is subject to specified initial and boundary conditions, and requires known relationships between the soil hydraulic variables moisture u, pressure head h and hydraulic conductivity K. The following relations between these variables have been used (Van Genuchten, 1980; Mualem, 1976): uðhÞ ¼ ures þ

usat
ures ½1 þ jahjn ðn
1Þ=n

Þðn
1Þ=n 2 KðuÞ ¼ Ksat Sle ½1
ð1
Sn=n
1 e

(2) (3)

where ures is the residual water content [L3 L
3], usat the saturated water content [L3 L
3], Se = (u
ures)/(usat
ures) the relative saturation [–], a an empirical shape factor [L
1], n an empirical shape factor [–], Ksat the saturated hydraulic conductivity [L T
1], and l an empirical coefficient [–]. For salt transport, the convection–dispersion equation is applied (Van Genuchten and Cleary, 1979; Boesten and Van der Linden, 1991):
 @uC @ @C @qC ¼ qLdis
@t @z @z @z

(4)

where q is the water flux density [L T
1], C the salt concentration [M L
3], and Ldis the dispersion length [L]. SWAP includes both a simple and detailed crop growth module. In the simple crop module, crop growth is described by the measured leaf area index, crop height and rooting depth as a function of crop development stage. The detailed crop growth module is based on the World Food Studies (WOFOST) model (Spitters et al., 1989; Supit et al., 1994), which simulates the crop growth and its production based on the incoming photosynthetically active radiation (PAR) absorbed by the crop canopy and the photosynthetic characteristics of leaf, and accounts for water and salt stress of the crop (Kroes and Van Dam, 2003). In addition to the water and salt stress, nutrient deficiency, weeds, pests and diseases may affect the crop production in actual field conditions. This reduces the water and salt limited production further to the actual production (Fig. 3). The effects of nutrient deficiency, pests, weeds, and diseases on crop growth and its production are not implemented in the present version of SWAP 3.03. However, this detailed crop growth module has the advantage of giving a feedback between crop growth and different water and salt stress conditions. The details of the light interception and CO2

260

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Fig. 3. Production hierarchy in the crop production system (adapted after Lo¨venstein et al., 1995).

assimilation as growth driving processes and the crop phenological development as growth controlling process included in SWAP 3.03 are described in Kroes and Van Dam (2003). To distinguish the simulations in this study, SWAP 3.03 when used with the simple crop growth module is called SWAP hereafter, and when used with the detailed crop growth module is called SWAP-WOFOST. The detailed crop module, SWAP-WOFOST, simulates the potential and water and salt limited crop yields, which cannot be simulated by the simple crop model SWAP. 2.3. Input parameters The input parameters of SWAP can be categorized into parameters required to define the upper boundary, crop, soil, and lower boundary and initial conditions. The potential evapotranspiration ETp and rainfall P and irrigation I fluxes define the upper boundary of the soil profile. The ETp is estimated by the Penman–Monteith equation (Monteith, 1965, 1981; Smith, 1992; Allen et al., 1998) using the daily meteorological data. The required meteorological data were obtained from the ICAR-CRS, Sirsa. The obtained meteorological data contained some missing data and errors. Therefore, a comparison with data from the meteorological station of the CCS Haryana Agricultural University, Hisar (about 90 km from Sirsa) was made. If needed, corrections were made using multiple regression relations between the data of Sirsa and Hisar (Roelevink, 2003). In the case of wind speed, data from Hisar were used. The solar radiation values were calculated from measured sunshine hours using the Angstrom formula (Angstrom, 1924)

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

261

Fig. 4. Measured daily values of rainfall, air vapour pressure, air temperature and solar radiation in Sirsa district during the agricultural year 2001–02.

with coefficients a = 0.29 and b = 0.41, which are specific for Sirsa. Fig. 4 shows the measured temperature, radiation, rainfall and vapour pressure in Sirsa district during the agricultural year 2001–02. The daily vapour pressure ranged from 0.4 to 3.4 kPa with an average value of 1.8 kPa. The daily maximum temperature reached to 46 8C on individual days during the kharif (summer) season. The daily solar radiation varied from 6552 to 26,463 kJ m
2 with an average value of 17,680 kJ m
2. The total rainfall amounted 188 mm only, out of which 177 mm was received during the kharif (monsoon) season.

262

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Table 2 Main crop parameters specified for SWAP-WOFOST model in Sirsa district Parameter Temperature sum from emergence to anthesis, TSUMEA (8C) Temperature sum from anthesis to maturity, TSUMAM (8C) Minimum canopy resistance, rcrop (S m
1) Critical pressure heads, h (cm) h1 h2 h3l h3h h4 Light extinction coefficient, Kgr Light use efficiency, e (kg ha
1 h
1/J m2 s
1) Maximum CO2 assimilation rate, A (kg ha
1 h
1) Salinity Critical level, ECmax (dS m
1) Decline per unit EC, ECslope (% dS m
1) a

Wheat a

Cotton

Rice

1480 890 70

2390 760 70

2060 620 70


1
22
1000
2200
16,000


1
22
1200
7500
16,000

100 55
160
250
16,000

0.375 0.45 40

0.450 0.40 50

0.338 0.45 47

6.0 7.1

7.7 5.4

5.0 9.0

For wheat crop at wheat–rice fields, TSUMEA = 1680 8C and TSUMAM = 1015 8C.

Most of the crop parameters of WOFOSTwere the same as summarized by Bessembinder et al. (2003). The simulation of dry matter production in WOFOST is sensitive to crop parameters such as the light use efficiency and the maximum CO2-assimilation rate, which are crop specific. Often they are calibrated and validated using site-specific crop measurements. In this study, we adjusted manually the light use efficiency e and the maximum CO2-assimilation rate A using the crop measurements at the farmer fields. The input parameters used for the simulated crops (wheat, cotton and rice) are summarized in Table 2. In case of the simple crop module SWAP, the measured leaf area index, crop height and rooting depth were described as a function of crop development stage, which was assumed to be linear in time from emergence to harvest. A soil profile of 300 cm depth was specified during the simulations. The soil profile was divided into one to three layers according to the measured profile description upto 120 cm soil depth (Fig. 2). The properties of the soil layer between 120–300 cm depth were taken equal to those of the last observed layer. The soil domain of 300 cm depth was further discretized into a total of 44 compartments with a nodal distance of 1 cm for the top 10 compartments, followed by 5 cm for the next 10 compartments and 10 cm for the remaining soil profile. This soil domain specification is acceptable as the actual soil evaporation under field conditions is controlled by only the top few centimetres of a soil (Van Dam and Feddes, 2000). In addition to Darcy’s law, a coefficient of 0.35 cm d
1 according to Black et al. (1969) was used to limit the soil evaporation rate. For salt transport in irrigated field soils, the dispersion length Ldis (Eq. (4)) was set to 5 cm (Nielsen et al., 1986). 2.4. Inverse modelling: a technique for parameter estimation Water flow and salt transport is very sensitive to the used soil hydraulic functions u(h) and K(u) (Eqs. (2) and (3)). The parameters describing these functions were based on the

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

263

measured texture (Fig. 2) and so-called pedotransfer functions (Wo¨sten et al., 1998), which relate soil texture to u(h) and K(u). The accuracy of pedotransfer functions is limited for sitespecific water flow and salt transport. Therefore, we might calibrate the soil hydraulic parameters either manually or automatically. We performed automatic calibration, which is also known as inverse modelling. Reliable estimation of soil hydraulic parameters by inverse modelling requires other measured inputs with high accuracy, and field observations which characterize the system behaviour. In addition, the parameters which are optimized should be sufficiently sensitive to these field observations. The observed soil moisture and salinity profiles were used as system response for the calibration of soil hydraulic parameters. SWAP using the measured crop growth was used in the calibration process. A non-linear parameter estimation program PEST (Doherty et al., 1995) was linked with the SWAP model (Fig. 5). The objective function quantifies the differences between model results and observations. If the observation errors follow a multivariate normal distribution with zero mean, no correlation, and constant variance for each observation type, maximization of the probability of reproducing the observed data leads to the weighted least squares objective function F(b):

FðbÞ ¼

N
 X

2 Wu ðuobs ðti Þ
usim ðb; ti ÞÞ



2 

þ WEC ðECobs ðti Þ
ECsim ðb; ti ÞÞ

i¼1

(5)

Fig. 5. Communication between the ecohydrological model, SWAP and the parameter estimation program, PEST.

264

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

where uobs(ti) and ECobs(ti) are the observed soil moisture and salinity at time ti, N the number of observations, and usim(b, ti) and ECsim(b, ti) are the simulated values of u and EC using an array with parameter values b. In case of random observation errors only, according to Maximum Likelihood the weighting factor for a particular observation should be equal to the inverse of the standard deviation of the observation error of that particular observation type. Gribb (1996) weighted each different data type by the inverse of the mean values. We used Wu = 1 and WEC = 10% of average uobs/average ECobs. In this way, we accounted for observation unit differences of u and EC, and at the same time gave relatively more weight to the moisture content observations. The inverse problem should be well posed in order to achieve unique and stable parameter values. In general, a well-posed inverse problem can be realized by a small number of fitting parameters (Kool and Parker, 1988). Of the parameters describing the soil hydraulic functions (Eqs. (2) and (3)), usat (cm3 cm
3) and Ksat (cm d
1) have a clear physical meaning, and can be measured directly. So the values of these parameters were taken from the measurements at the corresponding field. The ures (cm3 cm
3), which might be assigned a value near to zero (Russo, 1988), and empirical shape parameter l [–] were derived from the pedotransfer functions. Both parameters show less sensitivity to soil water flow and salt transport. Two parameters remain uncertain: a (cm
1) and n [–]. As the fields considered in this analysis have one to three soil layers (Fig. 2), the total number of parameters to be optimised is 2–6. In case of regular measurements at ordinary field conditions, 4–8 hydrological parameters could be estimated uniquely with a low coefficient of correlation and variation (Van Dam, 2000). Pedotransfer functions (Wo¨sten et al., 1998) were used to derive reasonable initial estimates of the soil hydraulic parameters for the optimisation process. In general, the farmers puddle the soil before rice transplantation in the field. The purpose of soil puddling is to reduce the percolation below the root zone, and thereby to maintain water ponding on the soil surface for optimal rice growing conditions. Because of this management practice, the hydraulic conductivity of the puddled layer is decreased. In order to capture the reduced percolation in the simulation of water flow during the rice growing period, the Ksat of the upper 30 cm soil layer of the wheat–rice fields (S1F1 and S2F5) was reduced to 20% (Singh et al., 2001). In addition, the reduction in the soil evaporation rate according to the empirical function of Black et al. (1969) was turned off. The groundwater level at the selected farmer fields was deeper than 3 m below soil surface. Therefore, the free drainage condition was applied as lower boundary. Initial salinity profiles were derived from the observations at the corresponding field. Unfortunately, the initial soil moisture contents were not observed. The initial soil moisture profiles at each field, therefore, were generated by running SWAP for 1 year in advance with the same inputs, and using the final pressure heads as initial condition. 2.5. Calculation of water productivity In agricultural production systems, water productivity accounts for crop production per unit amount of water used (Molden, 1997). Furthermore, water productivity can be defined in different ways referring to different types of ‘crop production’, i.e. dry matter or grain yield, and ‘amount of water used’, i.e. transpiration, evapotranspiration and irrigation

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

265

Table 3 Water productivity WP (kg m
3) expressed as crop production (kg m
2) per unit amount of water used (m3 m
2) WP WPT WPET WPETQ

Definitiona Yg/T Yg/ET Yg/ETQ

Unit

Field scale
3

kg m kg m
3 kg m
3

T E+T E + T + Qbot

a Yg is the crop grain (or seed) yield, T is the actual transpiration, E is the actual soil evaporation, and Qbot is the percolation at irrigated fields.

(Molden et al., 2001). This flexibility in defining water productivity provides useful indicators to evaluate the water utilization, and to identify where and when water can be saved. We used the following definitions of water productivity (Table 3). WPT is expressed in crop grain (or seed) yield Yg per unit amount of transpiration T, and sets the lower limit of water used by crop: the crop transpiration only. WPT depends on the crop type, e.g. C3 or C4 and its variety, and presents the physiological performance of a certain crop. WPT, also know as ‘transpiration efficiency’, is related to the diffusion rates of CO2 and H2O molecules, which vary proportionally to the size of stomatal aperture of leaves. Continuously changing environmental conditions, in terms of the CO2 concentration in air, radiation, temperature and vapour pressure deficit, also affect the diffusion rates of CO2 and H2O molecules for a certain crop. Improved crop varieties, well chosen sowing dates and sufficient water supply are expected to increase WPT values for a crop. The actual evapotranspiration ET represents the actual amount of water used in crop production, which is no longer available for reuse in the agricultural production system. It must be used as productive as possible, and it is logic to express WPET in terms of Yg per unit amount of ET. The inevitable loss of water due to evaporation decreases the water productivity from WPT to WPET. Therefore, relative low values of WPET as compared to WPT suggest the need to reduce evaporation by agronomic measures such as soil mulching and conservation tillage. Similarly, including percolation Qbot enlarges the denominator in expression of water productivity, and hence decreases it from WPET to WPETQ. Whether Qbot should be considered as a loss, it depends on the groundwater quality of the region. For instance in the good quality groundwater regions, Qbot recharges to groundwater, and can be recycled through groundwater pumping. If groundwater quality is poor, recycling of this water may not be possible, and any Qbot should be considered as lost water.

3. Results and discussion 3.1. Parameter estimation The soil moisture and salinity profiles observed during the rabi season were used for the calibration and validation of soil hydraulic parameters. The calibration process was performed with the first part of observations (Jan–Feb), and the second part of observations (Mar–Apr) was used for the validation. Soil hydraulic parameters a and n of the different soil layers of stratified soil profile (Fig. 2) were optimized simultaneously. The optimized

266

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Table 4 Derived soil hydraulic parameters at different farmer fields Field number

Soil layer (cm)

Wheat–rice fields S1F1 0–300 S2F5

0–30 30–300

Wheat–cotton fields S3F11 0–30 30–60 60–300

Texture

Soil hydraulic parameters Ksat (cm d
1)

a (cm
1)

l (
)

n (
)

0.57

1.57

0.005


2.57

1.93

0.01 0.01

0.50 0.58

2.63 1.87

0.010 0.005


2.53
2.37

1.40 1.77

SL LS SL

0.01 0.01 0.01

0.34 0.33 0.38

61.82 73.81 60.58

0.011 0.052 0.005


1.55
1.35
1.58

1.42 1.19 1.58

ures (cm3 cm
3)

usat (cm3 cm
3)

CL

0.01

SCL CL

S4F16

0–30 30–300

SL LS

0.01 0.01

0.31 0.32

101.71 120.87

0.014 0.036


1.67
0.87

1.29 1.19

S5F20

0–30 30–300

SL LS

0.01 0.01

0.34 0.31

138.69 141.62

0.041 0.024


1.56
0.80

1.20 1.16

Parameters a and n were optimised.

values of a and n together with the other soil hydraulic parameters (ures, usat, Ksat and l), which were input to the model, are given in Table 4. Repetition of the optimization process with different initial values of a and n resulted in the same values, which showed the uniqueness of the solution. The coefficient of variation and correlation coefficients of the optimized parameters also should be small during proper calibration. Table 5 lists the coefficient of variation (i.e. ratio of standard deviation and mean) and the correlation between the optimised parameters a and n. The coefficient of variation was relatively low for parameter n as compared to parameter a. This is attributed to the higher sensitivity of parameter n to the soil water flow. Table 5 also shows that the correlation coefficients were acceptably small. Table 5 Coefficients of variation and correlation matrix of optimized parameters a and n Field number S2F5

Soil layer (cm) 0–30 30–300

S5F20

0–30 30–300

Parameter

Optimized value

Coefficient of variation

a1 n1

0.010 1.40

0.271 0.06

1.00 0.15

1.00

a2 n2

0.005 1.77

0.504 0.02

0.59 0.26

0.86 0.38

a1 n1

0.041 1.20

1.474 0.10

1.00
0.77

1.00

a2 n2

0.024 1.16

1.182 0.01

0.53 0.23

0.12
0.09

Two typical examples: fields S2F5 (wheat–rice) and S5F20 (wheat–cotton).

Correlation coefficient a1

n1

a2

n2

1.0 0.4

1.0

1.0 0.2

1.0

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

267

The root mean square error (RMSE) (Willmott, 1982) is useful to quantify the difference between the observed and simulated data with the optimised parameters: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u 2 N
u1 X RMSE ¼ t Obsðti Þ
Simðti ; bÞ Ni ¼ 1

(6)

where Obs(ti) and Sim(ti, b) are the observed and simulated values for a output variable at time ti, and N is the total number of observations. As a typical example, Fig. 6 shows the observed and simulated soil moisture contents u and salinity concentrations EC1:2 of the field S5F20 during the rabi season. During the calibration period, the average RMSE at this field was 0.022 cm3 cm
3 for u and 0.09 dS m
1 for EC1:2. The average RMSE was also small during the validation period: 0.022 cm3 cm
3 for u and 0.07 dS m
1 for EC1:2. Table 6 lists the RMSE values for both the calibration and validation period at different fields. The RMSE of u ranged from 0.016 to 0.033 cm3 cm
3, and of EC1:2 from 0.09 to 0.31 dS m
1. These small values reveal that soil water flow and salt transport were well simulated by SWAP at different fields. As no systematic under- or over-estimation of u and EC1:2 was observed, the differences between the observed and simulated u and EC1:2 are contributed to the spatial heterogeneity and observation errors, which are inevitable under field conditions. 3.2. Water and salt balances The optimised soil hydraulic parameters (Table 4) in combination with other input data (Tables 1 and 2) were used in both SWAP and SWAP-WOFOST to simulate the water and salt balances at different fields. The water and salt balances simulated by SWAP-WOFOST were well comparable with those obtained from the calibrated and validated SWAP (Tables 7 and 8). Doorenbos and Kassam (1979) mentioned a range of evapotranspiration ET from 450 to 650 mm for wheat. In this study, the ET for wheat simulated by the calibrated and validated SWAP ranged from 338 to 355 mm at wheat–cotton fields during the period from Nov 1, 2001 Table 6 Number of observations N and root mean square error RMSE of soil moisture contents u and salinity concentrations EC1:2 for both the calibration (Jan–Feb) and the validation period (Mar–Apr) during the rabi (wheat) season of the agricultural year 2001–02 Field number

Calibration u (cm3 cm
3)

S1F1 S2F5 S3F11 S4F16 S5F20

Validation ECl:2 (dS m
1)

u (cm3 cm
3)

EC1:2 (dS m
1)

N

RMSE

N

RMSE

N

RMSE

N

RMSE

15 15 20 25 30

0.032 0.016 0.025 0.022 0.022

10 15 20 25 25

0.18 0.20 0.25 0.15 0.09

13 15 20 20 30

0.023 0.027 0.033 0.026 0.022

15 15 20 20 25

0.20 0.25 0.31 0.10 0.07

268

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Fig. 6. The observed and simulated soil moisture u and salinity EC1:2 profiles at the field S5F20 during the rabi (wheat) season of the agricultural year 2001–02. The calibration was performed with the first part of observations (Jan–Feb), and the second part of observations (Mar–Apr) was used for the validation.

to Apr 30, 2002. These values were in agreement with a mean ET of 360 mm with a standard deviation of 15 mm over the wheat areas in Sirsa district during the same period (Bastiaanssen et al., 2003). The simulated ET for wheat at wheat–rice fields ranged from 425 to 452 mm, which was higher than the simulated ETat wheat–cotton fields (Table 7). This was mainly due to the 1 month longer growing season at wheat–rice fields, i.e. from Oct 1, 2001 to Apr 30, 2002, as compared to wheat–cotton fields, i.e. from Nov 1, 2001 to Apr 30, 2002.

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

269

Table 7 Simulated water and salt balancesa at farmer fields in Sirsa district during the rabi (wheat) season of the agricultural year 2001–02 Componentb

SWAP model

SWAP-WOFOST model

S1F1c

S2F5c

S3F11

S4F16

S5F20

S1F1 c

S2F5c

S3F11

S4F16

S5F20

Water balance (mm) P 13 I 343 Icw 0 Igw 343 T 363 ET 452 Qbot
329 DW
426

13 424 0 424 326 425
195
185

11 430 50 370 244 338
77 23

11 391 0 391 253 351
6 42

11 568 0 568 245 355
171 52

13 343 0 343 312 405
334
387

13 424 0 424 291 397
200
162

11 430 50 370 215 313
86 40

11 391 0 391 223 329
7 64

11 568 0 568 245 353
160 63

24
75
51

102
19 83

25
3 22

20
49
30

20
31
11

24
77
53

102
21 81

25
3 22

20
46
27

Salt balance (mg cm
2) ICi 20 QbotCbot
30 DC
11

SWAP refers to the simple crop module, and SWAP-WOFOST refers to the detailed crop growth module. a Height soil column considered is 300 cm. b P is the rainfall, I the irrigation, Icw the canal irrigation, Igw the groundwater irrigation, T the actual transpiration, ET the actual evapotranspiration, Qbot the percolation (positive upward), DW the change in soil water storage, and C the salt concentration. c Wheat–rice fields. Table 8 Simulated water and salt balancesa at farmer fields in Sirsa district during the kharif (cotton/rice) season of the agricultural year 2001–02 Componentb

SWAP model S1F1

c

S2F5

SWAP-WOFOST model c

S3F11

S4F16

55F20

S1F1 c

S2F5c

S3F11

S4F16

S5F20

Water balance (mm) P 177 I 1250 Icw 0 Igw 1250 T 457 ET 862 Qbot
121 DW 440

177 1062 0 1062 536 960
98 175

177 301 162 139 277 427
86
37

177 554 0 554 582 745
25
44

177 737 0 737 685 827
132
51

177 1250 0 1250 472 858
133 430

177 1062 0 1062 546 949
100 184

177 301 162 139 317 456
96
78

177 554 0 554 586 740
41
58

177 737 0 737 632 758
159
11

Salt balance (mg cm
1) ICi 74 QbotCbot
11 DC 63

61
41 19

33
22 11

36
11 24

26
38
12

74
13 61

61
38 21

33
25 7

36
19 17

26
46
20

SWAP refers to the simple crop module, and SWAP-WOFOST refers to the detailed crop growth module. a Height soil column considered is 300 cm. b P is the rainfall, I the irrigation, Icw the canal irrigation, Igw the groundwater irrigation, T the actual transpiration, ET the actual evapotranspiration, Qbot the percolation (positive upward), DW the change in soil water storage, and C the salt concentration. c Rice fields.

270

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Doorenbos and Kassam (1979) also mentioned a range of ET from 450 to 700 mm for rice, and from 700 to 1300 mm for cotton. Bastiaanssen et al. (2003) estimated a mean ET of 769  50 mm over the rice areas, and of 690  39 mm over the cotton areas in Sirsa district during the period from May 1, 2002 to Oct 31, 2002. In this study, the ET simulated by the calibrated and validated SWAP ranged from 862 to 960 mm at rice fields, and from 427 to 827 mm at cotton fields. The simulated ET at rice fields was slightly higher than the values mentioned by Doorenbos and Kassam (1979) and Bastiaanssen et al. (2003). This is attributed to heavy irrigation applied at these fields: 1062 mm at S2F5 and 1250 mm at S1F1. Further, the simulated ET at cotton field S3F11 was significantly low as compared to other fields and above-mentioned values. This was a result of the low amount of irrigation applied (301 mm only) at this field during the kharif (cotton) season (Table 8). At the rice fields the heavy irrigations and decreased saturated hydraulic conductivity due to soil puddling, resulted into the simulation of saturated conditions during the rice growing season. As a typical example, Fig. 7 shows the irrigation amounts and simulated soil moisture in the upper soil layer (0–15 cm) of the field S1F1 during rice growing period. The percolation Qbot at wheat–rice fields was higher than those at wheat–cotton fields. At wheat–rice fields, the irrigation I during the kharif (rice) season (Table 8) was about 2.5–3.6 times higher than the rabi (wheat) season (Table 7). Despite this, the Qbot during the kharif season was relatively lower than the rabi season. In kharif season, the whole soil profile at the rice fields became almost saturated (change in water storage DW varied from 175 to 440 mm; Table 8), and thereby resulting in the less percolation. The saturated soil profile left after rice and application of two heavy irrigations about 100 mm each in the early stage (Oct–Nov) of the wheat crop resulted into higher Qbot during the rabi (wheat) season. The heavy irrigations also resulted into a high Qbot at the wheat–cotton field S4F20, and indicate over irrigation. The use of poor quality groundwater results into a salt build-up in the soil profile. For example, the change in salt storage DC at field S3F11 during the rabi season was high despite a significant Qbot

Fig. 7. Simulated soil moisture u (usat = 0.57) in the upper soil layer (0–15 cm) of field S1F1 during rice growing period.

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

271

(Table 7). This was due to the use of poor quality groundwater (3.73 dS m
1), which supplied a large amount of salts (102 mg cm
2). In case of poor quality groundwater, conjunctive use of canal and groundwater is beneficial in terms of salt build-up. For instance, the conjunctive use of canal Icw (162 mm) and groundwater Igw (139 mm) for irrigation at same field S3F11 resulted into a relatively low DC during the kharif season (Table 8). 3.3. Water productivity The water productivity for wheat, rice and cotton was analysed through both SWAP and SWAP-WOFOST. First, we calculated the water productivity values (Table 3) using the simulated water balance components T, ET and Qbot by SWAP and the actual (measured) grain (or seed) yields Yg at the selected farmer fields (Table 9). The average WPT, expressed as Yg/T (kg m
3), was 1.88 for wheat, 1.73 for rice and 0.29 for cotton. This presents wheat as the most efficient crop in terms of physical crop production in Sirsa district. The differences in WPT for different crops are due to the differences in the chemical composition, harvest index and evaporative demands during the respective seasons. In Sirsa district, temperatures and vapour pressure deficit are high during the kharif (summer) season, which results into high evaporative demands. Consequently, the WPT, WPET and WPETQ of summer crops (cotton and rice) are lower than those of winter crop (wheat). Based on a review of 82 literature sources with results of experiments in the last 25 years, Zwart and Bastiaanssen (2003) established global benchmark values of WPET, expressed as Yg/ET (kg m
3), at 1.08 for wheat, 1.09 for rice and 0.63 for cotton. Droogers and Kite (2001) mentioned a value of WPET from 0.16 to 0.39 for cotton at basin to field level in Turkey. Similarly, a value of WPET of about 0.27 for cotton is mentioned in a study on crop water productivity in Pakistan during 1970s (PARC, 1982). Tuong and Bouman (2003) summarized a range of WPET from 0.4 to 1.1 for rice in farmer fields and irrigation systems of Northwest India. Hussain et al. (2003) gave a WPET value of 1.36 for wheat in Haryana region. In our analysis, the average WPET at the selected farmer fields in Sirsa district was about 1.39 for wheat, 0.94 for rice and 0.23 for cotton (Table 9), which corresponds to average values for the climatic and growing conditions in Northwest India. To improve the WPET for a crop, the fraction of soil evaporation E in evapotranspiration ET is important (Table 3). During the rice cultivation, the high evaporative demands and continuously surface water ponding result in high soil evaporation. The fraction of E in ET at rice fields, therefore, was as high as 0.44 at field S2F5 and 0.47 at field S1F1 (Table 8). Consequently, the average WPET for rice was 45% lower than the average WPT. Also, the WPET for wheat and cotton was 17–35% lower than the WPT at different fields (Table 9). Improving agronomic practices such as soil mulching and especially aerobic rice cultivation can reduce this non-beneficial loss of water through soil evaporation E, and subsequently will improve the WPET. Reducing water inputs from continuous flooded conditions to soil saturation or alternate wet/dry conditions (aerobic rice) will slightly decrease the rice yields, but will substantially increase the water productivity (Bouman and Tuong, 2001).

Water productivity (kg m
3)/ crop yields (t ha
1)

Rice WPT (Yg/T) WPET (Yg/ET) WPETQ (Yg/ETQ) Yg a YFM b Potential Yg a Potential YFMb Cotton WPT (Yg/T) WPET(Yg/ET) WPETQ(Yg/ETQ) Yg a YFM b Potential Yg a Potential YFMb

SWAP-WOFOST model

S1F1

S2F5

S3F11

S4F16

S5F20

1.94 1.56 0.90 7.0 16.1

1.80 1.38 0.94 5.9 15.4

1.82 1.31 1.07 4.4 11.3

2.05 1.48 1.45 5.2 11.4

1.77 1.22 0.83 4.3 9.2

1.78 0.94 0.83 8.1 18.0

1.67 0.93 0.85 9.0 19.0

0.14 0.09 0.07 0.4 1.3

0.40 0.31 0.30 2.3 12.9

0.35 0.29 0.25 2.3 14.1

Average

S1F1

S2F5

S3F11

S4F16

S5F20

Average

1.88 1.39 1.04 5.4 12.7

2.90 2.23 1.22 9.0 19.7 9.1 19.8

2.47 1.81 1.21 7.2 18.0 7.6 19.2

3.13 2.15 1.69 6.7 14.1 7.7 15.4

2.81 1.90 1.86 6.3 14.2 7.1 15.3

2.83 1.97 1.35 6.9 14.6 7.2 15.3

2.83 2.01 1.47 7.2 16.1 7.7 17.0

1.73 0.94 0.84 8.5 18.5

1.54 0.84 0.73 7.3 17.7 7.5 19.1

1.53 0.88 0.80 8.4 17.5 9.1 18.9

0.29 0.23 0.21 1.6 9.4

1.54 0.86 0.77 7.8 17.6 8.3 19.0 0.32 0.22 0.18 1.0 5.8 3.3 20.2

0.45 0.36 0.34 2.6 13.9 3.6 22.7

0.42 0.35 0.29 2.7 12.8 4.0 23.5

0.40 0.31 0.27 2.1 10.8 3.6 22.1

Water productivity WP (kg m
3) is calculated in different forms, viz. Yg/T (transpiration) or ET (evapotranspiration) or ETQ (evapotranspiration + percolation) (Table 3). a Yg denotes the grain (or seed) yield. In case of SWAP, the Yg is the actual (measured) yield, while in case of SWAP-WOFOST, the Yg is the simulated water and salt limited yield considering 80% grain (or seed) in simulated storage organs for wheat, 81% for rice and 44% for cotton. The Yg includes 14, 16 and 15% moisture in grain (or seed) for wheat, rice and cotton, respectively. b YFM denotes the total fresh matter yield. In case of SWAP, the YFM is the actual (measured) yield, while in case of SWAP-WOFOST, the YFM is the simulated water and salt limited yield considering 12% moisture in air-dry fresh matter for wheat and rice, and 18% for cotton.

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Wheat WPT (Yg/T) WPET (Yg/ET) WPETQ (Yg/ETQ) Yg a YFM b Potential Yg a Potential YFMb

SWAP model

272

Table 9 Water productivity of wheat, rice and cotton at farmer fields in Sirsa district during the agricultural year 2001–02

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

273

The percolation Qbot further reduces the WPET to WPETQ (Table 3). The average WPETQ, expressed as Yg/ETQ (kg m
3), at the selected farmer fields was 1.04 for wheat, 0.84 for rice and 0.21 for cotton. Note the high reduction from WPET to WPETQ for wheat at wheat–rice fields (S1F1 and S2F5) and the wheat–cotton field S5F20 (Table 9). Usually in irrigated areas, Qbot contributes to the groundwater recharge, which is recycled through groundwater pumping in good quality groundwater areas. The groundwater pumping is not possible in the poor quality groundwater areas. Therefore, the reduction in Qbot will be beneficial for improving the low WPETQ values in the poor quality groundwater areas as in the northern parts of Sirsa district. Even in good quality groundwater areas, recycling of groundwater has energy cost that can be reduced if less water is lost through percolation. Optimal irrigation scheduling and precise land levelling can be promoted to reduce the Qbot, and hence to improve the WPETQ. Additionally, the WPT, WPET and WPETQ values were calculated using the SWAPWOFOST simulated water balance and crop grain (or seed) yields (Table 9). Accurate simulation of crop yields is necessary for the calculation of reliable water productivity values. In this analysis, the simulated potential Yg for wheat varied from 7.1 (S4F16) to 9.1 t ha
1 (S1F1) with an average value of 7.7 t ha
1. These potential productions of wheat in Sirsa district were in agreement with the reported potential Yg of 7.3 t ha
1 by Aggarwal et al., 2000. They also reported the potential Yg of 10.8 t ha
1 for rice in Sirsa district. In another study, Aggarwal et al. (2001) mentioned the potential Yg for cotton from 4.2 to 5.7 t ha
1 in Haryana conditions. In this study, the simulated potential Yg ranged from 7.5 to 9.1 t ha
1 for rice, and from 3.3 to 4.6 t ha
1 for cotton (Table 9). These values were slightly lower than the above mentioned values by Aggarwal et al. (2000, 2001). This might be due to high temperatures during the kharif season of the agricultural year 2001–02 (Fig. 4). Higher temperatures accelerate the crop development rate, which results into a shorter growing period and a lower crop production. Taking into accounting the water and salt stress, SWAP-WOFOST reduces the potential Yg to water and salt limited Yg. In addition to the water and salt stress, the actual Yg under field conditions depends on other factors such as nutrient supply, pest, disease and weed control (Fig. 2). The lowest actual (measured) Yg of 4.3 t ha
1 at wheat field S5F20 (Table 9) with the high irrigation of 568 mm (Table 7) suggests the presence of nutrition, pest, disease or weed stress at farmer fields in Sirsa district. Also, note the high actual Yg of wheat at field S1F1 as compared to other wheat fields. This is attributed to improved crop management and early sowing (i.e. Oct 25, 2001) of wheat at field S1F1. As expected, SWAP-WOFOST simulated water and salt limited Yg for wheat were 20– 60% higher than actual Yg (Table 9). Furthermore, the simulated water and salt limited Yg for wheat were almost equal to the simulated potential Yg at the corresponding field. The relative transpiration (T/Tp) at the selected wheat fields ranged from 0.85 to 1.00. This presents almost negligible water and salt stress on wheat in Sirsa district, in contrast to substantial nutrition, pest, disease or weed stress. The differences in WPT, WPET and WPETQ values for wheat obtained from SWAP and SWAP-WOFOST are mainly due to the differences in actual (measured) and simulated Yg at the corresponding fields. The average WPET for wheat obtained from SWAP-WOFOST was 45% higher than that obtained from SWAP (Table 9). Improved crop management in terms of timely sowing and optimal

274

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

fertilizer, and better pest, disease and weed control is expected to achieve this significant increase in the WPET for wheat in Sirsa district. For rice, the average water and salt limited Yg was 7.8 t ha
1, which was slightly lower than the average actual Yg of 8.5 t ha
1. Apparently it seems that rice growth at the selected rice fields was hardly limited by the water and salt stress, or reduced by the nutrient deficiency, pests, diseases or weeds: the actual Yg were almost equal to the potential Yg (Table 9). The relative transpiration (T/Tp) at the selected rice fields was about 0.95. However, the T/Tp at the selected cotton fields ranged from 0.50 to 0.65, and shows serious water and salt stress on cotton crop. Moreover, the low actual Yg of 0.4 t ha
1 at the cotton field S3F11 confirms crop failure, mainly due to water stress: 301 mm irrigation only (Table 8). The low rainfall of 177 mm only will have contributed to the water stress on kharif crops, especially cotton. Therefore, the water and salt limited Yg at the selected cotton fields were 1.4–3.3 times lower than the potential Yg (Table 9). Also, the WPT, WPET and WPETQ values for cotton obtained from SWAP-WOFOST were slightly higher than those obtained from SWAP. This suggests that ensuring the irrigation supplies, especially during the dry years, and improved crop management at cotton fields will increase cotton yields, and subsequently its water productivity in Sirsa district.

4. Conclusions The ecohydrological model SWAP in combination with field experiments can be used to quantify hydrological variables such as transpiration, evapotranspiration and percolation, and biophysical variables such as dry matter or grain yields, which are required for water productivity analysis of irrigated crops. Inverse modelling is efficient in the calibration of model input parameters. The use of observed soil moisture u and salinity profiles EC1:2 as system response in inverse modelling is successful to determine the soil hydraulic parameters. The good agreement between the observed and simulated u and EC1:2 values (Table 6) provided confidence to use the calibrated and validated SWAP to quantify the water and salt balances, and subsequently water productivity at farmer fields. Water productivity of main crops (wheat, rice and cotton) in Sirsa district corresponds to the average value for the climatic and growing conditions in Northwest India. Further, water productivity of summer crops (cotton and rice) is lower than that of winter crop (wheat). This is attributed to high temperatures and large vapour pressure deficit during the kharif (summer) season. Considerable spatial variation in WPT, WPET and WPETQ values was observed not only for different crops but also for the same crop. For instance, the WPET varied from 1.22 to 1.56 kg m
3 for wheat among different farmer fields (Table 9). The corresponding value for cotton varied from 0.09 to 0.31 kg m
3. This indicates a considerable variation and scope for improvement in water productivity. The significant (17–47%) share of evaporation into evapotranspiration, especially for rice, presents a major non-beneficial loss of water. The average WPET, expressed as Yg/ET (kg m
3), was 1.39 for wheat, 0.94 for rice and 0.23 for cotton. In this study, the WPT is expressed as Yg/T (kg m
3), and sets the lower limit of water used by crop: the crop transpiration T only. Using the actual (measured) crop yields and simulated water balance (SWAP), the calculated average WPET was significantly lower than the average WPT: 46%

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

275

for rice, 26% for wheat and 22% for cotton. Improving agronomic practices, especially aerobic rice cultivation, could reduce this non-beneficial loss of water through evaporation, and subsequently will improve the WPET. The simulated water and salt limited yields for wheat were 20–60% higher than the actual yields at the selected wheat fields. This is attributed to almost negligible water and salt stress at these fields: the relative transpiration (T/Tp) ranged from 0.85 to 1.00. This water productivity analysis suggests that improved crop management in terms of timely sowing, and optimal nutrition, pest, disease and weed control for wheat will multiply its WPET by a factor of 1.5! Moreover, severe water stress was observed on cotton (relative T/Tp < 0.65) during the kharif (summer) season. The simulated water and salt limited yields for cotton were 1.4 to 3.3 times (in case of actual yields 1.6–8.3 times) lower than the simulated potential yields at the selected cotton fields. Therefore, benefits in terms of increased cotton yields and improved water productivity will be gained by ensuring irrigation supply at cotton fields, especially during the dry years. Water productivity quantifies the ‘kg crop produced per m3 water used in different hydrological processes such as transpiration, evapotranspiration and percolation’, and provides the opportunity to identify ‘where water can be saved’ in the irrigated agriculture. Such a detailed water use analysis is very useful for many irrigated areas which deal with water scarcity. As this study shows, ordinary data on climate, soil and crop, in combination with ecohydrological models such as SWAP, can be used to produce the required hydrological and biophysical information in these regions.

Acknowledgements This research in Sirsa district (Haryana), India was financed by the Dutch Ministry of Agriculture through the Water Productivity (WATPRO) project from Januray 2001 to November 2003. The authors thank the anonymous reviewers for their constructive comments. References Aggarwal, M.C., Roest, C.W.J., 1996. Towards improved water management in Haryana state. Final report of the Indo-Dutch operational research project on hydrological studies. Chaudhary Charan Singh Haryana Agricultural University, Hisar. International Institute for Land Reclamation and Improvement, Wageningen, DLO Winand Staring Centre for Integrated Land, Soil and Water Research, now Alterra Green World Research, Wageningen, The Netherlands, 80 pp. Aggarwal, P.K., Talukdar, K.K., Mall, R.K., 2000. Potential yields of rice–wheat system in the Indo-Gangetic plains of India. Rice–Wheat Consortium Paper Series 10, Rice–Wheat Consortium for Indo-Gangetic plains, New Delhi, India, 16 pp. Aggarwal, P.K., Roeter, R.P., Kalra, N., Van Keulen, H., Hoanh, C.T., Van Laar, H.H. (Eds.), 2001. Land use analysis and planning for sustainable food security: with an illustration for the state of Haryana, India. Indian Research Institute, New Delhi, International Rice Research Institute, Los Banos and Wageningen University and Research Centre, Wageningen, The Netherlands, 167 pp. Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration, Guidelines for computing crop water requirements. Irrigation and Drainage Paper 56, FAO, Rome, Italy, 300 pp.

276

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Angstrom, A., 1924. Solar and atmospheric radiation. Q. J. R. Meteorol. Soc. 50, 122–125. Bastiaanssen, W.G.M., Zwart, S.J., Pelgrum, H., 2003. Remote sensing analysis. In: Van Dam, J.C., Malik, R.S. (Eds.), Water productivity of irrigated crops in Sirsa district, India. Integration of remote sensing, crop and soil models and geographical information systems. WATPRO final report, including CD-ROM. ISBN 90-6464864-6, pp. 85–100. Bessembinder, J.J.E, Dhindwal, A.S., Leffelaar, P.A., Ponsioen, T., Singh, S., 2003. Analysis of crop growth. In: Van Dam, J.C., Malik, R.S. (Eds.), Water productivity of irrigated crops in Sirsa district, India. Integration of remote sensing, crop and soil models and geographical information systems. WATPRO final report, including CD-ROM. ISBN 90-6464-864-6, pp. 59–82. http://library.wur.nl/way/catalogue/documents/ WATPRO_final_report.pdf. Black, T.A., Gardner, W.R., Thurtell, G.W., 1969. The prediction of evaporation, drainage, and soil water storage for a bare soil. Soil Sci. Soc. Am. J. 33, 655–660. Boesten, J.J.T.I., Van der Linden, A.M.A., 1991. Modelling the influence of sorption and transformation on pesticide leaching and persistence. J. Environ. Qual. 20, 425–435. Bouman, B.A.M., Tuong, T.P., 2001. Field water management to save water and increase its productivity in irrigated lowland rice. Agric. Water Manag. 49 (1), 11–30. Doherty, J., Brebber, L., Whyte, P., 1995. PEST: Model independent parameter estimation. Australian Centre for Tropical Freshwater Research James Cooke University, Townsville, Australia, 140 pp. Doorenbos, J., Kassam, A.H., 1979. Yield response to water. FAO Irrigation and Drainage Paper 33, FAO, Rome, Italy. Droogers, P., Kite, G., 2001. Estimating productivity of water at different spatial scales using simulation modelling. Research Report 53, International Water Management Institute, Colombo, Sri Lanka, 16 pp. Droogers, P., Bastiaanssen, W.G.M., 2002. Irrigation performance using hydrological and remote sensing modelling. J. Irrig. Drain. Eng. 128, 11–18. Gribb, M.M., 1996. Parameter estimation for determining hydraulic properties of fine sand from transient flow measurements. Water Resour. Res. 32, 1965–1974. Hussain, I., Sakthivadivel, R., Amarasinghe, U., Mudassar, M., Molden, D., 2003. Land and water productivity of wheat in the western Indo-Gangatic plains of India and Pakistan: a comparative analysis. Research Report 65, International Water Management Institute, Colombo, Sri Lanka, 50 pp. Jackson, M.L., 1973. Soil Chemical Analysis. Prentice Hall of India Pvt. Ltd, New Delhi, India. Jhorar, R.K., 2002. Estimation of effective soil hydraulic parameters for water management studies in semi-arid zones. PhD Thesis, ISBN 90-5808-644-5. Wageningen University and Research Centre, Wageningen, The Netherlands, 157 pp. Kijne, J., Barker, R., Molden, D. (Eds.), 2003. Water productivity in agriculture: limits and opportunities for improvement. Comprehensive assessment of Water Management in Agriculture, Series No. 1, CABI press, Wallingford, UK, 352 pp. Klute, A., Dirksen, C., 1986. Hydraulic conductivity and diffusivity: Laboratory methods. In: Klute, A. (Ed.), Methods of Soil Analysis. Part F. Second ed. Agron. Monogr. 9. ASA and ASSS, Madison, Wisconsin, USA, pp. 687–734. Kool, J.B., Parker, J.C., 1988. Analysis of the inverse problem for transient unsaturated flow. Water Resour. Res. 24, 817–830. Kroes, J.G., Van Dam, J.C. (Eds.), 2003. Reference Manual SWAP version 3.03. Alterra Green World Research, Alterra report 773, 1566–7197. Wageningen University and Research Centre, Wageningen, The Netherlands, 211 pp. Lo¨venstein H.M., Lantinga, E.A., Rabbinge, R., Van Keulen, H., 1995. Principles of production ecology: text for course F 300-001, pp. 8, Fig. 8. Department of Theoretical production ecology, Wageningen University and Research Centre, Wageningen, The Netherlands, 121 pp. Malik, R.S., Kumar, R., Dabas, D.S., Dhindwal, A.S., Singh, S., Singh, U., Singh, D., Mal, J., Singh, R., Bessembinder, J.J.E., 2003. Measurement program and description database. In: Van Dam, J.C., Malik, R.S. (Eds.), Water productivity of irrigated crops in Sirsa district, India. Integration of remote sensing, crop and soil models and geographical information systems. WATPRO final report, including CD-ROM. ISBN 90-6464864-6, pp. 29–39.

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

277

Molden, D., 1997. Accounting for water use and productivity. SWIM Paper 1. International Irrigation Management Institute, Colombo, Sri Lanka. 16 pp. Molden, D., Sakthivadivel, R., 1999. Water accounting to assesses and productivity of water. J. Water Resour. Dev. 15 (1/2), 55–72. Molden, D., Murray-Rust, H., Sakthivadivel, R., Makin, I., 2001. A water productivity framework for understanding and action. Workshop on Water productivity. Wadduwa, Sri Lanka, November 12 and 13, 2001. Monteith, J.L., 1965. Evaporation and the environment. In: Fogg, G.E. (Ed.), The State and Movement of Water in Living Organisms. Cambridge University Press, pp. 205–234. Monteith, J.L., 1981. Evaporation and surface temperature. Quart. J. Royal Soc. 107, 1–27. Mualem, Y., 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resour. Res. 12, 513–522. Nielsen, D.R., Van Genuchten, M.Th., Biggar, J.W., 1986. Water flow and solute transport in the unsaturated zone. Water Resour. Res. 22 (Suppl.), 89S–108S. PARC, 1982. Consumptive use of water for crops in Pakistan, Pakistan Agricultural Research Council, Final Technical report: PK-ARS-69/FG Pa 251, Islmabad, Pakistan, 193 pp. Piper, C.S., 1966. Soil and Plant analysis. Hans Publishers, Bombay, India. Richards, L.A., 1931. Capillary conduction of liquids through porous mediums. Physics 1, 318–333. Ritter, A., Hupet, F., Munoz-Carpena, R., Lambot, S., Vanclooster, M., 2003. Using inverse methods for estimating soil hydraulic properties from field data as alternative to direct methods. Agric. Water Manag. 59, 77–96. Roelevink, A., 2003. Stratification and parameterization for regional water productivity analysis of Sirsa District, Haryana (India). MSc Thesis. Wageningen University and Research Centre, Wageningen, The Netherlands, 42 pp. Russo, D., 1988. Determining soil hydraulic properties by parameter estimation: on the selection of a model for the hydraulic properties. Water Resour. Res. 24, 453–459. Smith, M., 1992. CROPWAT, a computer program for irrigation planning and management. Irrigation and Drainage Paper 46, FAO, Rome, Italy. Singh, K.B., Gajri, P.R., Arora, V.K., 2001. Modelling the effect of soil and water management practices on the water balance and performance of rice. Agric. Water Manag. 49, 77–95. Spitters, C.J.T., Van Keulen, H., Van Kraalingen, D.W.G., 1989. A simple and universal crop growth simulator: SUCROS87. In: Rabbinge, R., Ward, S.A., Van Laar, H.H. (Eds.), Simulation and systems management in crop protection, Simulation Monographs, Pudoc, Wageningen, The Netherlands, pp. 147–181. Supit, I., Hooyer, A.A., Van Diepen, C.A. (Eds.), 1994. System description of the WOFOST 6.0 crop simulation model implemented in CGMS. Vol. 1: Theory and algorithms. EUR publication 15956, Agricultural series, Luxembourg, 146 pp. Tuong, T.P., Bouman, B.A.M., 2003. Rice production in water-scarce environments. In: Kijne, J., Barker, R., Molden, D. (Eds.), Water productivity in agriculture: limits and opportunities for improvement. Comprehensive assessment of Water Management in Agriculture, Series No. 1, CABI Press, Wallingford, UK, pp. 53– 67. Van Dam, J.C., Huygen, J., Wesseling, J.G., Feddes, R.A., Kabat, P., Van Walsum, P.E.V., Groenendijk, P., Van Diepen, C.A., 1997. Theory of SWAP version 2.0. Simulation of water flow, solute transport and plant growth in the Soil–Water–Atmosphere–Plant environment. Report 71, Sub department of Water Resources, Wageningen University, Technical document 45, Alterra Green World Research, Wageningen, The Netherlands, 167 pp. Van Dam, J.C., Feddes, R.A., 2000. Numerical simulation of infilteration, evaporation and shallow groundwater levels with Richards’s equation. Simulation of field-scale water flow and bromide transport in a cracked clay soil. J. Hydrol. 233, 72–85. Van Dam, J.C., 2000. Simulation of field-scale water flow and bromide transport in a cracked clay soil. Hydrol. Process. 14, 1101–1117. Van Genuchten, M.Th., Cleary, R.W., 1979. Movement of solutes in soil: computer simulated and laboratory results. In: Bolt, G.H. (Ed.), Soil Chemistry B, Physico-Chemical Models. Elsevier, Amsterdam, The Netherlands, pp. 349–386. Van Genuchten, M.Th., 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, 892–898.

278

R. Singh et al. / Agricultural Water Management 82 (2006) 253–278

Willmott, C.J., 1982. Some comments on the evaluation of model performance. Bull. Am. Meteor. Soc. 63, 1309– 1313. Wo¨sten, J.H.M., Lilly, A., Nemes, A., Bas, C. Le., 1998. Using existing soil data to derive hydraulic parameters for simulation models in environmental studies and in land use planning. Report 156, Alterra Green World Research, Wageningen, The Netherlands, 106 pp. Zwart, S.J., Bastiaanssen, W.G.M., 2003. Review of measured crop water productivity values for irrigated wheat, rice, cotton and maize. Agric. Water Manag. 69 (2), 115–133.