Evaluation of options for increasing yield and water productivity of wheat in Punjab, India using the DSSAT-CSM-CERES-Wheat model

Evaluation of options for increasing yield and water productivity of wheat in Punjab, India using the DSSAT-CSM-CERES-Wheat model

agricultural water management 95 (2008) 1099–1110 available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/agwat Evaluation of ...

697KB Sizes 54 Downloads 244 Views

agricultural water management 95 (2008) 1099–1110

available at www.sciencedirect.com

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

Evaluation of options for increasing yield and water productivity of wheat in Punjab, India using the DSSAT-CSM-CERES-Wheat model J. Timsina a,*, D. Godwin b, E. Humphreys a,1, Yadvinder-Singh c, Bijay-Singh c, S.S. Kukal c, D. Smith a a

Commonwealth Scientific and Industrial Research Organization (CSIRO) Land and Water, Griffith, NSW 2680, Australia Altin Park, Dubbo, NSW 2830, Australia c Punjab Agricultural University, Ludhiana 141004, Punjab, India b

article info

abstract

Article history:

The DSSAT-CSM-CERES-Wheat V4.0 model was calibrated for yield and irrigation schedul-

Received 16 November 2007

ing of wheat with 2004–2005 data and validated with 13 independent data sets from

Accepted 12 April 2008

experiments conducted during 2002–2006 at the Punjab Agricultural University (PAU) farm,

Published on line 2 June 2008

Ludhiana, and in a farmer’s field near PAU at Phillaur, Punjab, India. Subsequently, the validated model was used to estimate long-term mean and variability of potential yield (Yp),

Keywords:

drainage, runoff, evapo-transpiration (ET), crop water productivity (CWP), and irrigation

Triticum aestivum

water productivity (IWP) of wheat cv. PBW343 using 36 years (1970–1971 to 2005–2006) of

CERES-Wheat

historical weather data from Ludhiana. Seven sowing dates in fortnightly intervals, ranging

Model calibration

from early October to early January, and three irrigation scheduling methods [soil water

Model evaluation

deficit (SWD)-based, growth stage-based, and ET-based] were evaluated. For the SWD-based

Model application

scheduling, irrigation management depth was set to 75 cm with irrigation scheduled when

Crop and irrigation water

SWD reached 50% to replace 100% of the deficit. For growth stage-based scheduling,

productivity

irrigation was applied either only once at one of the key growth stages [crown root initiation

Irrigation scheduling

(CRI), booting, flowering, and grain filling], twice (two stages in various combinations), thrice (three stages in various combinations), or four times (all four stages). For ET-driven irrigation, irrigations were scheduled based on cumulative net ETo (ETo-rain) since the previous irrigation, for a range of net ETo (25, 75, 125, 150, and 175 mm). Five main irrigation schedules (SWD-based, ET-driven with irrigation applied after accumulation of either 75 or 125 mm of ETo, i.e., ET75 or ET125, and growth stage-based with irrigation applied at CRI plus booting, or at CRI plus booting plus flowering stage) were chosen for detailed analysis of yield, water balance, and CWP and IWP. Nitrogen was non-limiting in all the simulations. Mean Yp across 36 years ranged from 5.2 t ha1 (10 October sowing) to 6.4 t ha1 (10 November sowing), with yield variations due to seasonal weather greater than variations across sowing dates. Yields under different irrigation scheduling, CWP and IWP were highest for 10 November sowing. Yields and CWP were higher for SWD and ET75-based irrigations on both soils, but IWP was higher for ET75-based irrigation on sandy loam and for ET150-based irrigation on loam. Simulation results suggest that yields, CWP, and IWP of

* Corresponding author at: International Rice Research Institute (IRRI), DAPO Box 7777, Metro Manila, Philippines. Tel.: +63 2 580 5600x2631; fax: +63 2 580 5600x2631. E-mail address: [email protected] (J. Timsina). 1 International Rice Research Institute (IRRI), DAPO Box 7777, Metro Manila, Philippines. 0378-3774/$ – see front matter # 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.agwat.2008.04.009

1100

agricultural water management 95 (2008) 1099–1110

PBW343 would be highest for sowing between late October and mid-November in the Indian Punjab. It is recommended that sowing be done within this planting period and that irrigation be applied based on the atmospheric demand and soil water status and not on the growth stage. Despite the potential limitations recognised with simulation results, we can conclude that DSSAT-CSM-CERES-Wheat V4.0 is a useful decision support system to help farmers to optimally schedule and manage irrigation in wheat grown in coarse-textured soils under declining groundwater table situations of the Indian Punjab. Further, the validated model and the simulation results can also be extrapolated to other areas with similar climatic and soil environments in Asia where crop, soil, weather, and management data are available. # 2008 Elsevier B.V. All rights reserved.

1.

Introduction

Wheat in Punjab, NW India, is grown on about 3.5 Mha of which 2.6 Mha is after rice (GOP, 2006). Depending on variety and transplanting date the time for rice harvesting ranges from late September to November, and consequently the time for wheat sowing ranges from late October to late November. Sowing wheat later than early November leads to reduction in yield due to high temperatures during the grain filling stage (Ortiz-Monasterio et al., 1994). Average total and long-term (for e.g., for 1970/1971–2005/2006) rainfall during the wheat season (November–April) are 110 mm and 12–248 mm, respectively. The rainfall is usually poorly distributed relative to the needs of wheat throughout the season. Thus there is heavy dependence on irrigation for high yields. Farmers in the Indian Punjab usually irrigate wheat 4 or 5 times, in addition to a pre-sowing irrigation. Gajri and Prihar (1985) and Gajri et al. (1989, 1991, 1993, 1997) concluded that, on the deeply wetted coarse-textured soils commonly used for rice–wheat systems in Punjab, one pre-sowing irrigation followed by another irrigation of about 50 mm around the time of crown root initiation [CRI, approximately 25 days after sowing (DAS)] induces deeper and greater rooting by decreasing the penetration resistance of the sub-surface hard pan layer, improves exploitation of residual soil water, and increases dry matter and grain yield. The effects of withholding the first post-sowing irrigation depends on soil type, with more harmful effects on less moisture-retentive loamy sands than on more moisture-retentive sandy loams (Gajri and Prihar, 1985; Gajri et al., 1989). Several studies in Punjab have investigated the effects of sowing dates on phenology and grain yield of wheat (Saini et al., 1986; Randhawa et al., 1981; Ortiz-Monasterio et al., 1994). Few others have investigated the contribution of stored soil water and compared the irrigation water requirements based on evapo-transpiration (ET) (Gajri et al., 1991, 1993, 1997), irrigation water/pan evaporation (IW/PanE) ratio (Prihar et al., 1974, 1976), and growth stage (Prihar et al., 1976), but except for Prihar et al. (1978), none are based on soil water deficit (SWD). Prihar et al. (1978) reported that grain yield of wheat did not decline when it was irrigated after depletion of the available soil water from the 180-cm profile with 50– 110 mm on sandy loam and loamy sand. Yields, however, declined largely when the crop was irrigated after soil water depletion with 140–170 mm. Prihar et al. (1974) also reported that irrigations with IW/PanE ratio of 0.75–1.0 resulted in higher yields compared to irrigations at five growth stages.

Groundwater is the major source of irrigation water for wheat in much of Punjab, and groundwater levels are declining rapidly in many locations (Singh, 2000). Therefore, there is an urgent need to maximise the productive use of water for wheat grown after rice. Knowledge of the impact of sowing and irrigation management on crop yield, water balance, crop water use and requirement, and water productivity can help identify management options for maximising crop water productivity (CWP, mass of grain production per unit of ET, or g grain kg1 ET) and irrigation water productivity (IWP, mass of grain production per unit of irrigation water, or g grain kg1 irrigation), and tradeoffs between yield, water use, and CWP and IWP. Analysis of CWP and IWP is considered to be a useful measure for evaluating options for improving agricultural water management and water resource related policy development (Kinje et al., 2003). CWP has been determined for wheat using the water balance data from field experiments in a wide range of environments (Choudhary et al., 2006; Zwart and Bastiaanssen, 2004) and using simulation models (Bastiaanssen et al., 2003; Droogers et al., 2006; Timsina et al., 2005; Singh et al., 2006). Yield, CWP, and IWP vary with seasonal weather conditions. Crop simulation models can be used to evaluate management options to increase yield and water productivity, taking into account seasonal variability and weather related risks, and to extrapolate the experimental results both spatially and temporally. Many simulation models have been developed and used to predict yield and water balance components of wheat in NW India, including Punjab. Of these, CERES-Wheat (Ritchie and Otter, 1985), WTGROWS (Aggarwal et al., 1994), SUCROS-WBM (Arora and Gajri, 1998; Jalota and Arora, 2002), and CROPMAN (Jalota et al., 2006) can predict both yield and water balance components. Some others (Arora et al., 1987; Jalota et al., 2000) can predict water balance components but not yield. DSSAT-CSM-CERES-Wheat V4.0, derived from CERESWheat and CROPSIM-Wheat, can simulate the growth and development of dryland and irrigated wheat across a range of latitudes in both the northern and southern hemispheres (Jones et al., 2003; Hoogenboom et al., 2004). The model has been evaluated and applied to a range of tropical, sub-tropical, and temperate environments in Asia (Timsina and Humphreys, 2006a,b). In the Indian Punjab, an earlier version of CERES-Wheat was evaluated and applied to predict the longterm trend in potential yield (Pathak et al., 2003), and the effects of sowing date and climate change on crop growth and yield (Hundal and Kaur, 1996, 1997). However, there was no

agricultural water management 95 (2008) 1099–1110

evaluation and application of the newer version (V4.0) of the model, unless recently by Arora et al. (2007). This paper presents the results of model studies exploring the potential for increasing CWP and IWP of wheat by manipulating sowing date and irrigation management in the Indian Punjab. The specific objective is to evaluate the DSSAT-CSM-CERES-Wheat V4.0 and use it to examine the effects of sowing date (from early October to early January) and irrigation management (irrigation scheduling based on SWD, ET, and growth stage) on yield, water use, water balance components, and CWP and IWP of wheat.

2.

Materials and methods

2.1.

Model modification

2.1.1.

Automatic irrigation scheduling in DSSAT V4.0

DSSAT V4.0 includes an automatic irrigation option which schedules irrigation based on SWD, for which the user sets the soil management depth (DSOIL), the SWD at which irrigation is triggered, and the amount of water applied as a per cent of SWD. DSSAT V4.0 calculates the amount of irrigation water applied automatically based on SWD in the soil management depth (DSOIL) in the ‘Simulation Controls’ section in the ’Experimental Management Details’ input file (File X). SWD is the difference between water content at the drained upper limit (DUL) and the actual water content (THETAC) over the management depth. If irrigation does not completely refill the soil to DUL over this depth an endpoint moisture content (IEPT), is specified. To invoke these calculations in the model an irrigation simulation method code A (automatic) is specified which determines the dates and amounts of irrigation. For cropping situations where irrigation may not be required after the crop has passed a certain phenological stage, a variable IOFF is specified in the File X. Details of the structure and format of input files are given in Tsuji et al. (1994). Two new automatic irrigation method codes (P and W) were recently added in DSSAT V4.0 (Hoogenboom et al., 2004). Both methods apply irrigation in the same fashion as irrigation method code R (amount reported or applied) up until the last reported date, but thereafter the model recharges the profile using the same approach as in method code A but with a fixed amount AIRAMT in method code W. In both methods, when an irrigation event occurs, the amount applied (either supplied or calculated) is passed into the water balance calculations using the variable DEPIR (depth of irrigation) with irrigation events, cumulative amount of irrigation, and the timing of each event as the final outputs. Details of the water balance simulation in DSSAT have been described elsewhere (Ritchie, 1998). Few farmers, especially in NW India including the Indian Punjab, currently have the capability of scheduling irrigations based on SWD. Many farmers schedule irrigations based on growth stage, while the guidelines for irrigation in NW India are actually based on average cumulative pan evaporation. Therefore, we added the capability of automatically scheduling irrigations based on growth stage and potential evaporation.

1101

2.1.2. Modifications to irrigation routines 2.1.2.1. Growth stage-based irrigation scheduling. In wheat, a common way to report phenological stages is Zadok’s system of decimal codes between 0 and 100 (Zadocs et al., 1974). For example, code 14 is used to indicate the time of first tiller appearance (CRI), 31 is booting, 60 is anthesis, and 83 is the milking (mid-way grain filling) stage. The Zadok stage is calculated internally in the model from its own stage system which in turn is dependent on plant response to photoperiod, vernalization and temperature. To facilitate model application for irrigation management, a code was developed and added to DSSAT V4.0 to apply irrigation automatically at CRI, booting, flowering, and grain filling, either as a single irrigation at one growth stage only, or for various combinations of growth stages as specified by the user in the simulation control section of File X.

2.1.2.2. ET-based irrigation scheduling. To simulate ET-based irrigation, a variable CumNET [cumulative amount of potential ET minus any rain (net ETo) which has occurred since the last irrigation] and a corresponding method code E (operating in a similar fashion to method code A) were introduced. DSOIL (in mm) for method code E, specified in the automatic irrigation section in File X, refers to the amount of net ETo while THETAC refers to the proportion of CumNET, with irrigation triggered when CumNET reaches a value at least equal to DSOIL. The calculated irrigation amount is then passed to the water balance calculations using the variable DEPIR. CumNET is then reset to zero, with logging of amounts and timing of irrigations as for method code A. 2.1.2.3. Summary output files for seasonal analysis. Relevant routines in DSSAT-CSM-CERES-Wheat V4.0 were modified to generate additional output. These include: duration from sowing to anthesis and from anthesis to physiological maturity, intercepted photosynthetically active radiation during vegetative stage (IPARVeg) and reproductive stage (IPARRep), mean water stress indices for emergence to end of juvenile (WS1), end of juvenile to panicle initiation (PI) (WS2), PI to heading (WS3), heading to grain filling (WS4), and grain filling to end of maturity (WS5), components of the water balance, and CWP. 2.2.

Model calibration

The model was calibrated using observations from conventionally tilled treatments in a replicated field experiment with wheat (var. PBW343) grown in rotation with maize or soybean on a loamy sand (Typic Ustochrept) at the Punjab Agricultural University (PAU), Ludhiana (308540 N, 758480 E, 247 m above mean sea level). The wheat was sown on 5 November 2004 and harvested on 20 April 2005. Nitrogen, phosphorus, and potassium fertilizers at a rate of 60 kg N, 30 kg P, and 25 kg K ha1 were applied at sowing as urea, single super phosphate, and muriate of potash, and another 60 kg ha1 N as urea was broadcast after the first irrigation. Irrigation amounts of 75 mm each were applied on 30 November, 28 December, and 16 March. Weeds, pests and diseases were well controlled. Management details are given by Ram (2006).

1102

agricultural water management 95 (2008) 1099–1110

Table 1 – Physical and chemical properties of the different soils used in model evaluations and applications Depth (cm)

LL (cm3 cm3)

DUL (cm3 cm3)

SAT (cm3 cm3)

SRGF

BD (Mg m3)

SOC (%)

Clay (%)

Silt (%)

Sand (%)

pH (1:2)

Experiments 1 and 2: sandy loam (PAU) 0–15 0.07 0.26 15–30 0.07 0.27 30–60 0.06 0.23 60–90 0.06 0.21 90–120 0.07 0.21 120–150 0.05 0.21 150–180 0.05 0.20

0.36 0.31 0.36 0.39 0.39 0.39 0.39

1.00 0.80 0.15 0.07 0.02 0.01 0.01

1.61 1.76 1.61 1.53 1.53 1.52 1.52

0.29 0.24 0.22 0.20 0.18 – –

17.2 15.3 16.6 14.8 14.0 8.2 8.6

17.2 17.4 12.0 13.0 12.2 10.9 5.3

65.6 67.3 71.4 72.2 73.8 80.9 88.1

6.7 7.3 7.6 7.7 7.8 – –

Experiment 1: loam (Phillaur) 0–15 0.09 15–30 0.09 30–60 0.10 60–90 0.13 90–120 0.14 120–150 0.14 150–180 0.14

0.31 0.32 0.33 0.33 0.33 0.33 0.33

0.38 0.36 0.36 0.36 0.36 0.36 0.36

1.00 0.50 0.05 0.03 0.02 0.01 0.00

1.55 1.79 1.70 1.71 1.67 1.67 1.67

0.37 0.23 0.22 0.19 0.24 – –

17.4 23.2 25.0 26.0 29.9 30.8 29.3

42.6 48.7 44.0 48.2 47.5 48.4 49.5

40.0 28.1 31.0 25.8 22.6 20.8 21.2

8.3 8.3 8.3 8.2 8.2 – –

Experiment 3: loamy sand (PAU) 0–8 0.07 9–17 0.07 18–22 0.06 23–37 0.05 38–55 0.06 56–72 0.06 73–89 0.06

0.26 0.25 0.25 0.25 0.24 0.25 0.24

0.44 0.38 0.35 0.36 0.36 0.39 0.38

1.00 0.80 0.15 0.07 0.02 0.01 0.01

1.60 1.64 1.62 1.63 1.62 1.56 1.54

0.42 0.40 0.40 0.22 0.32 0.20 0.11

24 25 25 24 23 25 25

10 11 12 19 15 16 16

66 64 63 57 62 59 59

7.8 8.1 8.3 8.3 8.3 8.3 8.4

LL: lower limit; DUL: drained upper limit; SAT: saturation; SRGF: soil root growth factor.

Default values for various soil parameters from DSSAT V4.0 were initially chosen and then calibrated based on the criteria and principles reported by Ritchie and Crum (1989) and Hoogenboom et al. (2004). The finally chosen values after calibration were: soil albedo, SALB (0.20); soil evaporation limit, SLU1 (5.0); whole profile drainage coefficient, SLDR (0.60); runoff curve number, SLRO (60.0); N fertility factor, SLNF (1.0); and fertility factor other than N, SLPF (0.90). The same values were used for both soils due to their close similarity in texture of the topsoil. The genetic coefficients for PBW343 were determined by first choosing a standard default ecotype and a cultivar within DSSAT V4.0, and adjusting the development- and growth-related coefficients to achieve the best possible match between simulated and observed phenology, final biomass, yield and yield components. The standard procedure suggested by Hunt and Boote (1998), and later modified by Hunt (Hunt, L.A., Univ. of Guelph, personal communication, 2005), was adopted.

loam at PAU and on a loam at Phillaur, Experiment 2 on a sandy loam at PAU, and Experiment 3 on a loamy sand at PAU. Experiment 2 included four N rates ranging from 0 to 160 kg ha1. The physical and chemical characteristics of the soils at each experimental site, as used for model calibration, validation, and application are presented in Table 1. To evaluate the performance of the model, R2, absolute and normalised root mean squared error (RMSE), and index of agreement (D-index) (Willmott et al., 1985) were used. RMSE is the ‘‘best’’ measure as it summarises the mean difference in the units of observed and predicted values. A model reproduces experimental data perfectly when R2 is 1, RMSE is 0, and D-index is 1. These parameters were calculated as follows: "

n X absolute RMSE ¼ N ðPi  Oi Þ2 i¼1

normalised RMSE ð%Þ ¼

2.3.

Model evaluation

The model was evaluated using 13 independent data sets on phenology, biomass, yield and yield components from the conventionally tilled flat treatments of three series of replicated field experiments conducted over 4 years (2002– 2003 to 2005–2006) at PAU and on a farmer’s field at Phillaur (Ram, 2006; Kukal et al., 2007; Singh et al., 2007). Experiments 1 and 2 involved rice–wheat cropping systems, while Experiment 3 involved maize–wheat and soybean–wheat cropping systems. Experiment 1 was conducted on a sandy

#0:5

1





absolute RMSE mean of the observed

" P n

ðP  Oi Þ2 DINDEX ¼ 1  Pn i¼1 i 0 2 0 1¼1 ½jPi j þ jOi j

 100

#

jPi j ¼ a þ bOi : ¯ Pi ¼ P0i  O;

O0i ¼ Oi  O¯

where Pi and Oi are predicted and observed values, and O¯ is the mean observed value over several replicates.

1103

agricultural water management 95 (2008) 1099–1110

2.4.

Model application

The model was used to estimate yield, components of the water balance, and water productivity for PBW343 for a range of sowing dates and irrigation management on the sandy loam and loam soils of the PAU Farm and Phillaur in Punjab (Experiments 1 and 2 soils). Plant available water holding capacity (PAWC), i.e., water content between the drained upper limit (DUL) and the lower limit (LL), for the sandy loam and loam was 290 and 368 mm, respectively, in the top 180 cm (Humphreys et al., 2007), while the water holding capacity of the pore space between DUL and saturation was 280 and 62 mm, respectively. The loam had higher PAWC but substantially lower drainable porosity than the sandy loam. Initial conditions for each simulation were set 3 days before sowing, and simulations were carried out using 36 years (1970/ 1971–2005/2006) of historical weather data from Ludhiana for both soils. Initial total soil water to 180 cm depth was 309 and 493 mm in the sandy loam and loam soils, respectively, which reflect actual moisture conditions at the time of sowing wheat after rice, after a pre-sowing irrigation of about 70 mm (Humphreys et al., 2007). Initial ammonium-N of the profile was 25 and 27 kg ha1 and nitrate-N was 50 and 46 kg ha1, respectively. The soil root growth factor (SRGF), with values ranging from 0 to 1 for each layer (1 = no mechanical impediments or chemical stresses, and thus soil hospitable for root growth; 0 = high degree of mechanical or chemical stresses, and soil inhospitable for root growth) (Hoogenboom et al., 2004) was adjusted as the majority of the roots grew in the top 30 cm in both soils, with a maximum rooting depth of 150 cm, reflecting field observations on these soils (Humphreys et al., 2004). All simulations were carried out under N non-limiting conditions. All grain yields and water productivities are for dry grain.

2.4.1.

Potential yield

Potential yield, as affected by temperature, solar radiation and genotypic characteristics, was simulated for a range of sowing dates (at 15-day intervals from 10 October to 10 January) for 36 years of weather data at Ludhiana (1970–1971 to 2005–2006).

2.4.2.

Comparison of irrigation methods

The effect of irrigation management on yield and components of the water balance and water productivity was investigated for various sowing dates for the two soils. Three irrigation scheduling methods were used: SWD-based, growth stagebased, and ET-based. For the SWD-based scheduling, the

irrigation management depth was set to 75 cm with irrigation scheduled when SWD increased to at least 50%, to replace 100% of the deficit. For growth stage-based scheduling, irrigation was applied either only once at one of the key growth stages (CRI, booting, flowering, and grain filling), twice (two stages in various combinations), thrice (three stages in various combinations), or four times (all four stages). Irrigation at CRI was applied when SWD at 0–25 cm depth was 25% or more, while for all other stages irrigation was applied when SWD over 0–75 cm depth was 25% or more. The irrigation water applied was the amount required to replace 100% of the deficit in the specified soil depth. For ET-driven irrigation, the first irrigation was scheduled at 25 DAS (approximately CRI) when SWD at 0–25 cm depth was 50% or more. Further irrigations were scheduled based on cumulative net ETo (ETorain) since the previous irrigation, for a range of net ETo (25, 75, 125, 150, and 175 mm). The irrigation water added was the amount needed to replace 90% of net ETo, consistent with the recommended practice for wheat in Punjab (Prihar et al., 1974). Five main irrigation schedules, (1) SWD-based, (2 and 3) ETdriven with irrigation after accumulation of either 75 mm (ET75) or 125 mm of ET (ET125), and (4 and 5) growth stagebased with irrigations applied at CRI plus booting, or at CRI plus booting plus flowering stage for crops sown on 10 November were chosen for detailed analysis of yield, irrigation water requirement, and CWP.

3.

Results and discussion

3.1.

Model calibration

The simulated and observed values for phenology, grain and biomass yields, and grains/m2, after calibration, for the two treatments are presented in Table 2. There was a good agreement between measured and simulated grain yields (5.3– 5.4 t ha1). We suspect that there was incomplete drying of straw, thus resulting in high moisture content and higher straw yield. In the model, however, straw yield prediction is based on 0% moisture content. Thus, mean observed straw yields (7.0–7.1 t ha1), were 20% higher than simulated straw yields (5.5–5.6 t ha1). As a result, the total top weight at maturity was also underestimated by about 10%. Simulated seasonal total above-ground biomass accumulation during the growing season, however, matched well with the observed values (Fig. 1). Errors and uncertainties associated with any model calibration and evaluation processes are discussed later.

Table 2 – Simulated and observed phenological dates, grain yield, and grain number for PBW343 grown after maize and soybean in 2004–2005 (calibration) Treatments

SB–WH MZ–WH

Anthesis (DAS)

Maturity (DAS)

Grain yield (t ha1)

Grain number (m2)

Straw yield (t ha1)

Top yield (t ha1)

Sim.

Obs.

Sim.

Obs.

Sim.

Obs.

Sim.

Obs.

Sim.

Obs.

Sim.

Obs.

130 130

129 128

160 160

157 158

5.3 5.4

5.4 5.3

11750 11920

13098 13007

5.5 5.7

7.1 7.0

10.8 11.0

12.5 12.2

SB–WH: soybean–wheat; MZ-WH: maize–wheat; Sim.: simulated; Obs.: observed; DAS: days after sowing.

1104

agricultural water management 95 (2008) 1099–1110

simulated grain yield (Table 3). The model also predicted the yields for the four N rates ranging from 0 to 160 kg ha1 on the sandy loam at PAU over three seasons quite satisfactorily, suggesting that the model is able to capture the yield response to N.

3.2.2.

Fig. 1 – Simulated and observed seasonal above-ground biomass of PBW343 sown on 5 November 2004 on a conventionally tilled flat plot at PAU, Ludhiana under Experiment 3 (continuous line represents simulations and symbols observed data; vertical bars are S.D.s from the means of observed data).

3.2.

Model evaluation

The model was evaluated for days to anthesis and physiological maturity, grain, straw and biomass yields, and harvest index (HI) being the ratio of grain biomass to total aboveground biomass, using 13 independent data sets from four seasons of Experiments 1 and 2, and two seasons of Experiment 3 (Fig. 2; Table 3). Model predicted number of days to anthesis and maturity quite well as seen from low RMSE and reasonably good D-index (Table 3).

3.2.1.

Grain yield

There was generally a good agreement between the model predictions and observed data, within the bounds of experimental error (Fig. 2; Table 3). The model predicted grain yield on the loam at Phillaur and on the loamy sand at PAU quite well, but tended to over-predict on the sandy loam at PAU. Standard deviation (S.D.) was greater for observed than for the

Fig. 2 – Simulated and observed grain yield of PBW343 sown on conventionally tilled flat plots across 4 years in three experiments at two sites in Punjab.

Straw and total above-ground biomass

Across all data sets from all experiments examined together, prediction for grain yield was best (absolute RMSE = 617 kg ha1; normalised RMSE = 15%; D-index = 0.92) while that for HI was not quite satisfactory (absolute RMSE = 14; normalised RMSE = 35%; D-index = 0.34). S.D.s were greater for the observed than for the simulated straw and above-ground biomass yields (Table 3). Analysis of data for individual years also revealed satisfactory performance of the model for grain, straw and biomass yields.

3.2.3.

Time-course of top weight

Fig. 3 shows the reasonably good agreement between observed and simulated top weight of wheat during the 2005–2006 growing season on loam at Phillaur. In conclusion, based on all model evaluation criteria (S.D., RMSE, and D-index), the performance of the model was satisfactory in terms of its predictive ability for seasonal above-ground biomass yield as well as final grain, straw and biomass yields, and HI.

3.3.

Model application

3.3.1.

Effect of sowing date on potential yield

Potential yield varied across years and sowing dates (Fig. 4), with mean yield ranging from 5.2 (10 October sowing) to 6.4 t ha1 (10 November sowing). Yields were greatest for sowings between 25 October to 25 November (mean 6.3– 6.4 t ha1), and least for 10 October and 10 January sowings (mean 5.2–5.5 t ha1). Yield variation due to seasonal weather variation was greater than variation across sowing dates. For example, for the November 10 sowing, yields ranged from 4.3 to 7.9 t ha1. There was a decline in potential yield by 40.9 kg ha1 year1 for sowing on 15 November from 1970–1971 to 2005–2006, and also a decline by 15.7 kg ha1 (range 0–40 kg ha1) for each day delay in sowing after 10 November to 10 January over the same 36 years period. The duration from sowing to anthesis was greater for sowings on 25 October and 10 November, but duration from anthesis to physiological maturity was greater for October 10 and least for January 10 sowing. Leaf area index around the time of anthesis also varied considerably across years, and was usually highest for November 10 sowing followed by late October and late November-sown crops. Based on experimental field data, Ortiz-Monasterio et al. (1994) also reported that for Ludhiana, the optimum sowing dates were 5 November for PBW 34 (long-season) and 15 November for PBW 154 (medium-season) and PBW 226 (shortseason) cultivars. After the optimum sowing dates, yields were reduced by 0.8, 0.7, and 0.7% per day, or by 37, 34, and 34 kg ha1 d1, respectively. Randhawa et al. (1981) also reported that delaying sowing from 25 October to 15 December reduced yields of Kalyansona, WL711, HD2009, and WG357 by

1105

agricultural water management 95 (2008) 1099–1110

Table 3 – Validation of DSSAT-CSM-CERES-Wheat for phenology, grain, straw, biomass yields, and HI Data setsa

N

S.D.obs

S.D.sim

Abs. RMSE

Normal. RMSE

Anthesis (day) Maturity (day) Grain yield (kg ha1) Straw yield (kg ha1) Biom. yield (kg ha1) HI (%)

22 28 32 24 24 24

5.5 5.2 1166 1958 2915 4.3

4.1 3.7 972 1014 1920 2.6

5.3 4.5 617 2962 3091 13.5

7.0 3.4 15.1 39.9 23.1 35.1

D-Index 0.70 0.73 0.92 0.57 0.72 0.34

N: number of data pairs; SDobs: standard deviation of observed values; SDsim: standard deviation of simulated values; Abs. RMSE: absolute root mean square error; Normal. RMSE: normalised root mean square error (%); Biom. yield: biomass yield; HI: harvest index. a 13 data sets from 2002 to 2006 experiments.

1.2, 0.9, 1.2, and 1.0% per day, respectively. Our results based on model predictions are thus similar to the results obtained from the field experiments. Using CERES V3.5, Pathak et al. (2003) predicted the potential yield of wheat variety HD 2329, with long-term (1985–1999) mean yields ranging from 5.2 to 7.9 t ha1 for the Indo-Gangetic Plains of India. The highest potential yield (7.9 t ha1) was obtained for Ludhiana. With WTGROWS, Aggarwal et al. (2000) also reported that the potential yield varied between 2.6 and 8.3 t ha1 for 138 locations in India,

Fig. 3 – Simulated and observed seasonal above-ground biomass of PBW343 sown on 5 November 2004 on a conventionally tilled flat plot on a loam at Phillaur under Experiment 1 (continuous line represents simulations and symbols observed data; vertical bars are S.D.s from the means of observed data).

Fig. 4 – Effect of sowing date on simulated potential yield of PBW343 at Ludhiana, Punjab, India.

with yields above 7 t ha1 for NW India. In general, the yield decrease after 15 November was between 0.25 and 0.75% for yield potential of less than 4 t ha1, between 0.5 and 1.0% for yield potential between 4 and 6 t ha1, and between 0.8 and 1.0% for yield potential greater than 6 t/ha. Arora and Gajri (1998) reported that SUCROS-WBM predicted a potential yield of 3.6, 6.0 and 3.1 t ha1 for 15 October (early), 15 November (optimum), and 15 January (late) sowings. The potential yields using CERES V3.5 by Pathak et al. (2003) and using WTGROWS by Aggarwal et al. (1994) for Ludhiana were slightly greater (mean 7.9 and 7.0 t ha1, respectively) while those using SUCROS-WBM by Arora and Gajri (1998) were lower (mean 6.0 t ha1) than ours. The mean decline in wheat yield (40.9 kg ha1 year1) for 15 November sowing over 36 years was almost double the decline (20 kg ha1 year1) as reported by Pathak et al. (2003) during 1985–1999. Aggarwal et al. (2000), however, reported slightly greater yield decline (50 kg ha1 d1) beyond 15 November sowing in Delhi than by us (0–40 kg ha1 d1) for crops sown beyond 10 November in Ludhiana. The studies by Aggarwal et al. (2000) and Pathak et al. (2003) used different varieties, and hence cannot directly be comparable to ours.

3.3.2. Effect of irrigation scheduling method on yield, water use, and water productivity 3.3.2.1. SWD-based irrigation. On both loam and sandy loam soils, yields were greatest for 10 November sowing, albeit close to those for 25 October and 25 November sowings. Differences in yields across sowing dates were much greater than between soil types. The effect of sowing date on yield with irrigation based on SWD was similar to that of potential yield as the crops with SWD-based irrigation were never stressed. As for potential yield, yield with all irrigation scheduling methods varied greatly with seasonal conditions on a loam soil (Fig. 5). The components of the water balance and CWP were also strongly influenced by seasonal conditions (Table 4, Figs. 6–8). Sowing date only had a small effect on the components of the water balance. Crop water use (ET) was least for the earliest sowing, and similar for all sowings on or later than November 10. Consequently, irrigation requirement was least for October 10 sowing. On the loam, mean irrigation requirement was 257 mm for October 10 sowing, compared with means of 277– 289 mm for other sowing dates. Runoff and deep drainage were small and similar for all sowing dates (means 0.5– 3.9 mm). Crop water productivity was highest for October 25 and November 10 sowings (mean 1.7 g grain kg1 ET), while

1106

agricultural water management 95 (2008) 1099–1110

Crops for all growth stage-based irrigation treatments were always water-stressed before the start of CRI. The stress during the growing season, however, varied across treatments, with more stress when irrigated only once, more so when the irrigation was delayed to flowering or grain filling stage. Even with four irrigations, there was still water deficit stress, indicating that the irrigations were not applied timely and/or that the amount applied was not sufficient to meet crop demand. There were no significant differences between soils in response to growth stage-based irrigation scheduling. Crop water use and irrigation amount varied greatly with seasonal conditions and frequency of irrigation (Figs. 6 and 7). Mean ET ranged from 226 mm for a single irrigation to 342 mm for four irrigations. Mean irrigation amount ranged from 40 to 245 mm for the same treatments (Table 5). In comparison, irrigation amount with SWD-based irrigation was considerably higher (mean 361 mm) for the same sowing date (e.g., November 10) (Fig. 7). Mean CWP ranged from 1.0 g kg1 (for two irrigations at flowering and grain filling) to 1.2 g kg1 with three irrigations (CRI, booting, flowering) to 1.6 g kg1 for two irrigations at CRI and booting (Table 5). The latter was close to the mean CWP achieved under non-stressed conditions with SWD irrigation scheduling (1.7 g kg1). Irrigation water productivity with a single irrigation at CRI was extremely high (9.0 g kg1) due to the very low irrigation amount, while yield was only reduced by about 30% in comparison with the highest yielding treatment. Otherwise, IWP ranged from 2.0 g kg1 (two irrigations at flowering and grain filling) to 4.1 g kg1 with two irrigations at CRI and booting. Irrigation water productivity of the highest yielding treatment (three irrigations at CRI, booting, flowering) was much lower at 2.8 g kg1. Irrigation water productivity with the highest yielding growth stagebased irrigation was, however, still higher than with SWDbased irrigation scheduling for the November 10 sowing (2.4 g kg1), due to a greater reduction in irrigation amount than in yield.

Fig. 5 – Effect of irrigation management on simulated grain yield of PBW343 sown on 10 November on a loam at Ludhiana, Punjab, India.

irrigation water productivity was highest for 10 November sowing (mean 2.4 g grain kg1 irrigation water) (Table 4). The effects of sowing date on yield and on the components of the water balance and water productivity on the loam were similar to those on the sandy loam.

3.3.2.2. Growth stage-based irrigation. Yields of crops sown on 10 November on a loam with growth stage-based irrigation scheduling ranged from 0.9 to 7.2 t ha1 (Table 5). With one irrigation, yields were highest when the crop was irrigated at booting (mean 3.8 t ha1) and least when irrigated at grain filling (mean 2.8 t ha1). With two irrigations, yields were highest when irrigated at CRI and booting (mean 4.9 t ha1) and least when irrigated at flowering and grain filling (mean 2.8 t ha1). With three irrigations, yields were highest when irrigated at CRI, booting, and flowering, or at CRI, booting and grain filling (mean 5.0 t ha1). Irrigation at three stages gave higher yields than at two stages in about 60% of years (Fig. 5). Yields with three irrigations were similar to those with irrigations at all four stages (Table 5). Yields with three and four irrigations were always considerably less than that with SWD-based irrigation (Fig. 5).

3.3.2.3. ETo-based irrigation. Grain yields of PBW343 sown on 10 November with irrigation scheduled according to net ETo were much higher with more frequent irrigation (i.e., after accumulation of net ETo of 25 and 75 mm) compared to less frequent irrigation (i.e., after accumulation of 125, 150, and

Table 4 – Mean grain yield, water balance components, and water productivity of PBW343 with irrigation to refill the profile to 75 cm depth whenever plant available water declined to 50% on a loam soila Treatments/ scenarios

Grain yield (t ha1)

October 10 October 25 November 10 November 25 December 10 December 25 January 10

4.73 5.93 6.25 6.03 5.69 5.25 4.96

a b c d

(3.1–6.0) (3.6–7.3) (4.0–7.8) (4.4–7.4) (4.2–6.9) (4.0–6.4) (4.1–6.1)

ET (mm)b

319 348 361 363 360 357 359

(256–387) (275–403) (286–412) (300–429) (300–428) (303–413) (302–433)

Values in parentheses give the range. ET: evapo-transpiration. CWP: crop water productivity. IWP: irrigation water productivity.

Irrigation (mm) 257 284 279 289 277 281 289

(87–350) (173–437) (173–434) (172–352) (173–352) (174–351) (173–353)

Runoff (mm) 3.9 3.5 2.9 3.3 3.0 2.2 1.9

(0–40) (0–31) (0–34) (0–35) (0–33) (0–17) (0–20)

Deep Evaporation drainage (mm) (mm)

CWP (g grain kg1 ET)c

1.9 1.3 1.2 2.1 0.7 0.6 0.5

1.5 1.7 1.7 1.6 1.6 1.5 1.4

(0–49) (0–16) (0–39) (0–30) (0–23) (0–11) (0–16)

46 40 38 45 50 57 60

(31–79) (31–63) (29–67) (30–64) (35–68) (41–82) (41–89)

(1.2–1.8) (1.3–1.8) (1.4–1.9) (1.4–1.9) (1.3–1.8) (1.3–1.7) (1.2–1.6)

IWP (g grain kg1 irrigation)d 1.9 2.2 2.4 2.2 2.1 1.9 1.8

(1.4–4.2) (1.4–4.0) (1.5–3.6) (1.6–4.0) (1.5–3.5) (1.4–3.3) (1.2–3.4)

agricultural water management 95 (2008) 1099–1110

Fig. 6 – Effect of irrigation management on simulated ET of PBW343 sown on 10 November on a loam at Ludhiana, Punjab, India.

Fig. 7 – Effect of irrigation methods on simulated irrigation water applied to PBW343 sown on 10 November on a loam at Ludhiana, Punjab, India.

175 mm net ETo) (Table 6). With frequent irrigation the crops were never stressed. Yield with ET75 was similar to that with SWD-based irrigation, but considerably higher than yield with growth stage-based irrigation (Tables 4–6). Yield declined more rapidly on the sandy loam than on the loam as irrigation interval increased, probably reflecting the greater initial PAWC. ET and irrigation amount declined as irrigation interval increased, however the effect on soil evaporation, runoff and deep drainage was negligible, except on the loam for irrigation intervals of 125 mm and more. Increasing irrigation interval from 25–75 to 125 mm decreased irrigation amount by about 30 mm for each increment. Crop water productivity was higher (mean 1.7 g kg1 on both soils) for greater irrigation intervals up to 75 mm on the sandy loam, and 125 mm on the loam (Table 6), but was similar to SWD-based irrigation (Fig. 8). However IWP increased as irrigation interval increased, to a maximum value of 3.3 g kg1 at ET75 on the sandy loam, and 5.7 g kg1 at ET150 mm on the loam. Grain yields on all soils were greatest for 25 October, and 10 and 25 November sowings, suggesting a relatively narrow

1107

Fig. 8 – Effect of irrigation methods on simulated crop water productivity of PBW343 sown on 10 November on a loam at Ludhiana, Punjab, India.

sowing period for wheat after rice in the Indian Punjab. Crop water productivity was greater for optimum sowing date compared to earlier or later sowing dates. For crops sown on 10 November (optimum sowing), yields were highest when irrigation was applied after 25 or 75 mm of ETo-rain, followed by SWD-driven irrigation, and were least for growth stagebased irrigation. In ET-driven irrigation, the crops were not at all stressed, while in SWD-driven irrigation, they were little stressed. In stage-based irrigation, the crops were already stressed before the initiation of CRI. These results suggest that irrigation be scheduled based on the status of water in the soil (i.e. considering SWD) or atmospheric demand (i.e. considering ET) and not on the basis of growth stages. With predictions from an earlier version of CERES, Hundal and Kaur (1997) reported that irrigated wheat yields of HD2329 across years and sowing dates in Ludhiana ranged from 3.0 to 5.5 t ha1. Similar to our results, they concluded that the yields were highest for the early November sowing and decreased progressively with delay in sowing. Their yield estimations with irrigation were lower than ours, but they did not report how many irrigations were given and how they were managed. Our simulation results are similar to results of field experiments by Gajri et al. (1997) who reported that frequent small (25 mm) irrigations (5–8) for wheat on a sandy loam after 25 mm net pan evaporation, increased yield and CWP over infrequent large irrigations (2–4), each with 50–75 mm given after net PanE of 50–75 mm, respectively. Our results for fairly high yield with one irrigation at CRI also corroborate to Gajri and Prihar (1985) and Gajri et al. (1989, 1991, 1993) who reported that one early season irrigation at about 30 days after sowing can utilize the profile water and give yields as high as that with more frequent irrigations. Our results for higher yields with two irrigations are quite similar to Jalota et al. (1980) who also recommended for at least two irrigations, one at CRI and another at flowering, for wheat after rice. A recent recommendation by PAU is also to irrigate wheat at least twice, one at CRI and another at flowering, but depending on the amount of rainfall during the previous rice season and the stored soil moisture, a third irrigation at booting may also be required.

1108

agricultural water management 95 (2008) 1099–1110

Table 5 – Grain yield, water balance components, and water productivity of PBW343, sown on 10 November, as affected by irrigation scheduling at various stages on a loam soila Treatments/ scenarios

Grain yield (t ha1)

CRI Booting (BT) Flowering (FL) Grain filling (GF) CRI + BT CRI + FL CRI + GF BT + FL BT + GF FL + GF CRI + BT + FL CRI + BT + GF BT + FL + GF CRI + FL + GF CRI + BT + FL + GF

3.58 3.77 2.80 2.80 4.91 3.90 3.90 3.88 3.88 2.80 5.04 5.04 3.88 3.90 5.04

a b c d

(1.5–5.9) (1.9–5.9) (0.9–5.9) (0.9–5.9) (2.9–7.2) (1.5–6.1) (1.5–6.1) (1.9–6.0) (1.9–6.0) (0.9–5.9) (2.9–7.2) (2.9–7.2) (1.9–6.0) (1.5–6.1) (2.9–7.2)

ET (mm)b

226 257 250 226 297 300 273 287 289 257 331 334 298 307 342

(121–322) (181–334) (122–342) (106–328) (229–376) (174–378) (152–374) (206–365) (215–363) (122–355) (258–403) (274–405) (222–368) (189–392) (276–405)

Irrigation (mm) 40 101 110 119 139 151 161 151 194 151 191 236 201 197 243

(32–43) (0–119) (53–129) (72–131) (34–160) (93–171) (105–173) (80–209) (94–237) (80–189) (121–246) (105–276) (80–265) (121–234) (121–302)

Deep drainage (mm) 1 3 0 0 4 1 1 3 4 0 5 5 4 1 5

(0–22) (0–38) (0–2) (00) (0–69) (0–22) (0–22) (0–38) (0–38) (0–2) (0–69) (0–69) (0–38) (0–22) (0–69)

Evaporation (mm) 48 47 53 53 49 55 55 52 55 59 52 54 57 59 55

(34–75) (23–81) (25–89) (27–83) (34–77) (34–86) (37–84) (23–90) (25–91) (26–95) (34–77) (35–83) (24–98) (36–93) (35–83)

CWP (g grain kg1 ET)c 1.5 1.4 1.1 1.2 1.6 1.3 1.4 1.3 1.3 1.0 1.5 1.5 1.3 1.2 1.5

(1.1–1.9) (0.9–2.0) (0.7–1.8) (0.7–1.8) (1.1–2.1) (0.9–1.7) (0.9–1.8) (0.8–1.8) (0.9–1.8) (0.6–1.8) (1.1–1.8) (1.1–1.9) (0.8–1.8) (0.8–1.7) (1.1–1.8)

IWP (g grain kg1 irrigation)d 9.0 3.8 2.8 2.4 4.1 2.7 2.5 2.8 2.1 2.0 2.8 2.2 2.1 2.1 2.2

(3.8–17.4) (0–15.5) (0.7–7.3) (0.7–8.2) (1.9–17.4) (1.0–5.6) (0.9–5.6) (1.1–6.4) (1.0–6.3) (0.7–6.4) (1.4–4.6) (1.2–5.6) (0.9–6.4) (0.8–4.6) (1.1–4.6)

Values in parentheses give the range. ET: evapo-transpiration. CWP: crop water productivity. IWP: irrigation water productivity.

3.3.3.

Some warnings and limitations

The results of the model evaluation showed that the model satisfactorily predicted the grain yield of wheat sown with conventional tillage. There were, however, some cases where the model results did not match well with the experimental data. Several factors could be responsible for the discrepancies between simulated and observed results. First of all, all models are abstractions of reality, and real systems can never be modelled perfectly. Thus scientific processes embedded and coefficients used in any model, including DSSAT-CSM-CERESWheat, can never be complete. Inclusion of incomplete or imperfect knowledge in the model processes and our incomplete understanding of those processes could lead to uncertainties in model predictions.

Second, there are uncertainties associated with parameters and inputs used in any model. The crop, soil, and weather inputs have a degree of uncertainty associated with them due to random errors, bias in their measurement and calibration, and spatial and temporal variability. For example, development related genetic coefficients in DSSAT-CSM-CERESWheat are based on the data of the phenological (emergence, flowering, and physiological maturity) dates, which can also be affected by irrigation and N regimes. We admit that there were some management and methodological issues in observing and recording the dates of phenological events as well as some discrepancies between data of yield components and grain yields obtained from separate samples. Such uncertainties in data of phenological and yield components

Table 6 – Grain yield, water balance components and water productivity of PBW343, sown on 10 November, for ET-driven irrigation on two soilsa Treatments/ scenarios

Grain yield (t ha1)

Sandy loam (25 mm) Sandy loam (75 mm) Sandy loam (125 mm) Sandy loam (150 mm) Sandy loam (175 mm) Loam (25 mm) Loam (75 mm) Loam (125 mm) Loam (150 mm) Loam (175 mm)

6.30 6.21 5.03 4.39 3.91 6.25 6.09 5.88 5.38 4.91

a b c d e

(4.2–7.7) (4.2–7.6) (3.5–6.4) (2.7–5.9) (2.1–5.9) (4.2–7.8) (4.0–7.5) (4.2–7.2) (3.8–6.7) (2.9–6.7)

Values in parentheses give the range. ET: evapo-transpiration. Soil evap.: soil evaporation. CWP: crop water productivity. IWP: irrigation water productivity.

ET (mm)b

374 364 327 304 281 378 362 352 333 317

(296–423) (281–416) (256–379) (245–363) (216–346) (303–428) (283–410) (261–403) (261–393) (248–376)

Irrigation (mm) 262 235 206 202 190 261 231 190 184 176

(128–353) (100–316) (31–268) (32–312) (31–358) (97–356) (96–317) (15–256) (15–303) (15–339)

Soil evap. (mm)c 47 43 50 55 57 54 48 47 50 52

(41–66) (35–66) (37–70) (33–82) (34–94) (46–71) (39–68) (39–68) (36–68) (36–72)

Deep drainage (mm) 2 1 3 7 5 0 0 112 124 134

(0–51) (0–26) (0–17) (0–30) (0–69) (0–7) (0) (101–173) (101–174) (101–199)

CWP (g grain IWP (g grain kg1 ET)d kg1 irrigation)e 1.7 1.7 1.5 1.4 1.4 1.7 1.7 1.7 1.6 1.5

(1.4–1.9) (1.4–1.9) (1.3–1.9) (1.0–1.9) (0.9–1.9) (1.4–1.8) (1.4–1.9) (1.4–2.0) (1.3–2.0) (1.0–2.0)

2.5 2.9 3.3 3.3 3.1 2.5 2.9 5.5 5.7 5.7

(1.8–4.6) (1.9–5.9) (1.3–19.1) (1.1–18.5) (0.6–19.1) (1.8–4.5) (1.9–5.7) (1.9–39.5) (1.5–34.8) (1.2–39.5)

agricultural water management 95 (2008) 1099–1110

could lead to uncertainties in the determination of genetic coefficients, and hence in the model calibration and validation. Third, there were some unavoidable uncertainties in measured initial mineral N and soil moisture as well as in weather data, especially in sunshine hour data. Hence for calibration and validation of models, it is essential that good quality data sets to which researchers have full confidence be used. Uncertainties in model parameters and inputs could add further uncertainties and complexities when one tries to extrapolate the findings across time and space. Since the CWP is based on yield and ET, and IWP on yield and irrigation amount, there would be increased uncertainties in CWP and IWP data as compared to phenology, yield, and yield components data. Hence, when interpreting and extrapolating the model results, due consideration should be given to uncertainties that arise from model structure, model parameters and inputs, and in the experimental data used for model calibration, validation, and application.

4.

Conclusions

The DSSAT-CSM-CERES-Wheat was calibrated, evaluated, and used as a research tool to provide estimates of climatically driven potential yield, and yield, water balance components, and CWP and IWP of wheat grown after rice for different irrigation scheduling conditions for the Indian Punjab. The scenarios tested using 36 years of weather data and the seasonal analysis option of the DSSAT software showed how to better schedule irrigation to increase yield, optimize irrigation amount with selection of a proper irrigation method, and increase CWP and IWP. Simulation results suggest that yield of PBW343 in the Indian Punjab would be highest for sowing between late October and mid-November, and that irrigation be scheduled based on the atmospheric demand, or soil water status, and not on the growth stages. Results further suggest that, for wheat sown on the optimum date (i.e., November 10), yield, and CWP and IWP would be highest when irrigation is applied based on soil water deficit, or after accumulation of 75–125 mm of ETo-rain. However, the presented results are based on certain assumptions. The model predictions may be affected by a degree of uncertainty of the validity of these assumptions and the accuracy by which input parameters can be established. This may affect results and also bias conclusions in terms of yield estimations, water balance components, and CWP and IWP. Further, the interactions between weather, soil characteristics, plant growth dynamics, and management alternatives may affect simulation results. Currently the DSSATCSM-CERES-Wheat model does not include effects of weed and pests which can also affect predictions. In conclusion, despite the potential limitations, the DSSATCSM-CERES-Wheat model can be a useful decision support system to assist farmers for irrigation scheduling and applying optimum amount of irrigation water. This can eventually help increase CWP and IWP and make efficient use of declining water resources in the Indian Punjab. Further, the model can be used and the results of this study extrapolated to other areas with similar climatic and soil environments in Asia where crop, soil, weather, and management data are available.

1109

Acknowledgements We are grateful to the Australian Centre for International Agricultural Research (ACIAR) for financial support for this activity. We also acknowledge S.S. Dhillon, P.R. Gajri, Hari Ram and other research fellows, and all field coordinators and labourers involved in the project from PAU for assisting in crop management and monitoring of the field experiments.

references

Aggarwal, P.K., Kalra, N., Singh, A.K., Sinha, S.K., 1994. Analyzing the limitations set by climatic factors, genotype, and water and nitrogen availability on productivity of wheat. Part I. The model description, parameterization, and validation. Field Crops Res. 38, 73–91. Aggarwal, P.K., Bandyopadhyay, S.K., Pathak, H., Kalra, N., Chnader, S., Kumar, S., 2000. Analysis of yield trends of the rice–wheat system in north-western India. Outlook Agric. 29, 259–268. Arora, V.K., Gajri, P.R., 1998. Evaluation of a crop growth-water balance model for analysing wheat responses to climateand water-limited environments. Field Crops Res. 59, 213–224. Arora, V.K., Prihar, S.S., Gajri, P.R., 1987. Synthesis of a simplified water use simulation model for predicting wheat yields. Water Resour. Res. 23, 903–910. Arora, V.K., Singh, H., Singh, B., 2007. Analyzing wheat productivity responses to climatic, irrigation and fertilizernitrogen regimes in a semi-arid sub-tropical environment using the CERES-Wheat model. Agric. Water Manage. 94 (1–3), 22–30. 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 GIS. WATPRO Final report, including CS-ROM. ISBN: 90-6464864-6, pp. 85–100. Choudhary, B.U., Bouman, B.A.M., Singh, A.K., 2006. Yield and water productivity of rice–wheat on raised beds at New Delhi, India. Field Crops Res. 100, 229–239. Droogers, P., Kite, G.W., Murray-Rust, H., 2006. Use of simulation models to evaluate irrigation performances including water productivity, risk and system analyses. Irrigation Sci. 19, 139–145. Gajri, P.R., Prihar, S.S., 1985. Rooting, water use, and yield relations in wheat on loamy sand and sandy loam soils. Field Crops Res. 12, 115–132. Gajri, P.R., Prihar, S.S., Arora, V.K., 1989. Effects of nitrogen and early irrigation on root development and water use by wheat on two soils. Field Crops Res. 21, 103–114. Gajri, P.R., Prihar, S.S., Cheema, H.S., Kapoor, A., 1991. irrigation and tillage effects on root development, water use, and yield of wheat on coarse textured soils. Irrigation Sci. 12, 161–168. Gajri, P.R., Prihar, S.S., Arora, V.K., 1993. Interdependence of nitrogen and irrigation effects on growth and input-use efficiencies in wheat. Field Crops Res. 31, 71–86. Gajri, P.R., Singh, J., Arora, V.K., Gill, B.S., 1997. Tillage responses of wheat in relation to irrigation regimes and nitrogen rates on an alluvial sand in a semi-arid subtropical climate. Soil Tillage Res. 42, 33–46. GOP, 2006. Agriculture at a Glance. Dept. of Agriculture, GOP, Chandigarh, India.

1110

agricultural water management 95 (2008) 1099–1110

Hoogenboom, G., Jones, J.W., Porter, C.H., Wilkens, P.W., Boote, K.J., Batchelor, W.D., Hunt, L.A., Tsuji, G.Y., 2004. DSSAT 4. 0., Overview, vol. 1. ICASA, University of Hawaii, Honolulu, USA. Humphreys, E., Thaman, S., Prashar, A., Gajri, P.R., Dhillon, S.S., Singh, Y., Nayyar A., Timsina J., Singh, B. 2004. Productivity, water use efficiency and hydrology of wheat on beds and flats in Punjab, India. CSIRO Land and Water Technical Report 03/04. CSIRO, Griffith, NSW, Australia. . Humphreys, E., Kukal, S.S., Amanpreet-Kaur, Thaman, S., Yadav, S., Singh, Y., Singh, B., Timsina, J., Dhillon, S.S., Prashar, A., Smith, D.J., 2007. Permanent beds for rice– wheat in Punjab, India. Part 2. Water balance and soil water dynamics. In: Humphreys, E., Roth, C.H. (Eds.), Permanent Beds and Rice-residue Management for Rice–Wheat Systems in the Indo-Gangetic Plain. ACIAR Proceedings No. 127, ACIAR, Canberra, Australia. Hundal, S.S., Kaur, P., 1996. Climate change and its impact on crop productivity in Punjab, India. In: Abrol, Y.P. (Ed.), Climate Variability and Agriculture. Narosa Publishing House, Northeast Delhi, India, pp. 377–393. Hundal, S.S., Kaur, P., 1997. Application of the CERES-Wheat model to yield predictions in the irrigated plains of the Indian Punjab. J. Agric. Sci. 129, 13–18 Cambridge. Hunt, L.A., Boote, K.J., 1998. Data for model operation, calibration, and evaluation. In: Tsuji, G.Y., Hoogenboom, G., Thornton, P.K. (Eds.), Understanding Options for Agricultural Production. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 9–39. Jalota, S.K., Arora, V.K., 2002. Model-based assessment of water balance components under different cropping systems in north-west India. Agric. Water Manage. 57, 75–87. Jalota, S.K., Prihar, S.S., Sandhu, B.S., Khera, K.L., 1980. Yield, water use, and root distribution of wheat as affected by presowing and post-sowing irrigation. Agric. Water Manage. 2, 289–297. Jalota, S.K., Arora, V.K., Singh, O., 2000. Development and evaluation of a soil water evaporation model to assess the effects of soil texture, tillage and crop residue management under field conditions. Soil Use Manage. 16, 194–199. Jalota, S.K., Sood, A., Chahal, G.B.S., Choudhary, B.U., 2006. Crop water productivity of cotton–wheat system as influenced by deficit irrigation, soil texture and precipitation. Agric. Water Manage. 84, 137–146. Jones, J.W., Hoogenboom, G., Porter, C.H., Boote, K.J., Batchelor, W.D., Hunt, L.A., Wilkens, P.W., Singh, U., Gijsman, A.J., Ritchie, J.T., 2003. The DSSAT cropping system model. Eur. J. Agron. 18, 235–265. Kinje, J.W., Barker, R., Molden, D., 2003. Water Productivity in Agriculture: Limits and Opportunities for Improvement. CAB International, Wallingford, UK. Kukal, S.S., Humphreys, E., Singh, Y., Singh, B., Yadav, S., Amanpreet-Kaur, Thaman, S., Timsina, J., Dhillon, S.S., Brar, N.K., Prashar, A., Smith, D.J., 2007. Permanent beds for rice-wheat in Punjab, India. Part 1. Crop performance. In: Humphreys, E., Roth, C.H. (Eds.), Permanent Beds and Riceresidue Management for Rice–Wheat Systems in the IndoGangetic Plain. ACIAR Proceedings No. 127, ACIAR, Canberra, Australia. Ortiz-Monasterio, J.I., Dhillon, S.S., Fischer, R.A., 1994. Date of sowing effects on grain yield and yield components of irrigated spring wheat cultivars and relationships with radiation and temperature in Ludhiana, India. Field Crops Res. 37, 169–184. Pathak, H., Ladha, J.K., Aggarwal, P.K., Peng, S., Das, S., Singh, Y., Singh, B., Kamra, S.K., Mishra, B., Sastri, A.S.R.A.S., Aggarwal, H.P., Das, D.K., Gupta, R.K., 2003. Trends of

climatic potential and on-farm yields of rice and wheat in the Indo-Gangetic plains. Field Crops Res. 80, 223–234. Prihar, S.S., Gajri, P.R., Narang, R.S., 1974. Scheduling irrigation to wheat using pan evaporation. Ind. J. Agric. Sci. 44, 567–571. Prihar, S.S., Khera, K.L., Sandhu, K.S., Sandhu, B.S., 1976. Comparison of irrigation schedule based on pand evaporation and growth stages in winter wheat. Agron. J. 68, 650–653. Prihar, S.S., Sandhu, B.S., Khera, K.L., Jalota, S.K., 1978. Water use and yield of winter wheat as affected by timing and amount of last irrigation. Irrigation Sci. 1, 39–45. Ram, H., 2006. Micro-environment and productivity of maize– wheat and soybean–wheat sequences in relation to tillage and planting systems. Ph.D. Dissertation. PAU, Ludhiana, India, p. 218. Randhawa, A.S., Dhillon, S.S., Singh, W., 1981. Productivity of wheat varieties, as influenced by the time of sowing. J. Res. Punjab Agric. Univ. 18, 227–233. Ritchie, J.T., 1998. Soil water balance and plant stress. In: Tsuji, G.Y., Hoogenboom, G., Thornton, P.K. (Eds.), Understanding Options for Agricultural Production. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 41–54. Ritchie, J.T., Crum, J., 1989. Converting soil survey characterization data into IBSNAT crop model input. In: Bouma, J., Brget, A.K. (Eds.), Land Qualities in Space and Time. Wageningen, The Netherlands, pp. 155–167. Ritchie, J.T., Otter, S., 1985. Description and performance of CERES-Wheat: a user-oriented wheat yield model. In: Willis, W.O. (Ed.), ARS Wheat Yield Project. ARS-38. Natl. Technol. Info. Serv., Springfield, MI, USA. Saini, A.D., Dadhwal, V.K., Phadnawis, B.N., Nanda, R., 1986. Influence of sowing dates on pre-anthesis phenology in wheat. Ind. J. Agric. Sci. 56 (7), 503–511. Singh, R.B., 2000. Environmental consequences of agricultural development: a case study from the Green Revolution state of Haryana, India. Agric. Ecosys. Environ. 82, 97–103. Singh, R., van Dam, J.C., Feddes, R.A., 2006. Water productivity analysis of irrigated crops in Sirsa district, India. Agric. Water Manage. 82, 253–278. Singh, Y., Brar, N.K., Humphreys, E., Singh, B., Timsina, J., 2007. Yield and nitrogen use efficiency of permanent bed rice– wheat systems in northwest India: effect of N fertilization, mulching and crop establishment method. In: Humphreys, E., Roth, C.H. (Eds.), Permanent Beds and Rice-residue Management for Rice–Wheat Systems in the Indo-Gangetic Plain. ACIAR Proceedings No. 127, ACIAR, Canberra, Australia. Timsina, J., Humphreys, E., 2006a. Performance of CERES-Rice and CERES-Wheat models in rice–wheat systems: a review. Agric. Syst. 90, 5–31. Timsina, J., Humphreys, E., 2006b. Applications of CERES-Rice and CERES-Wheat models in research, policy, and climate change studies in Asia: a review. Int. J. Agric. Res. 1 (3), 202–225. Timsina, J., Humphreys, E., Godwin, D., Mathews, S., 2005. Evaluation of options for increasing water productivity of wheat using CSM-CERES Wheat model. In: Proceedings of the MODSIM05 Conference, December 12–15, Melbourne, Australia. Tsuji, G.Y., Uehara, G., Balas, S., 1994. DSSAT V3.0, vol. 2, University of Hawaii, Honolulu, Hawaii, p. 284. Willmott, C.J., Ackleson, S.G., Davis, R.E., Feddema, J.J., Klink, K.M., Legates, D.R., O’Connell, J., Rowe, C.M., 1985. Statistics for the evaluation and comparison of models. J. Geophys. Res. 90 (C5), 8995–9005. Zadocs, J.C., Chang, T.T., Konzak, C.F., 1974. A decimal code for the growth stage of cereals. Weed Res. 24, 415–421. Zwart, S.J., Bastiaanssen, W.G.M., 2004. Review of measured crop water productivity values for irrigated wheat, rice, cotton and maize. Agric. Water Manage. 69, 115–133.