INTRODUCTION
Rainfed agriculture plays an important role in contributing to world food security. In India, the land area under rainfed agriculture is about 85 million ha representing about 60% of net cultivated area and it supports 40% of the population of the country. In sub-Saharan Africa, more than 95% farm land is rainfed, while the corresponding figure for Latin America is almost 90%, for South Asia about 60%, for East Asia 65%, and for the Near East and North Africa 75% (Wani et al., Reference Wani, Sreedevi, Rockstrom, Ramakrishna, Wani, Rockstrom and Oweis2009). In addition to the climatic constraints such as erratic and uncertain rainfall patterns, soils in the rainfed areas are highly degraded physically, chemically and biologically (Maruthi Sankar et al., Reference Maruthi Sankar, Sharma, Dhanapal, Shankar, Mishra, Venkateswarlu and Kusuma Grace2010a; Sharma et al, Reference Sharma, Mandal, Srinivas, Vittal, Mandal, Kusuma Grace and Ramesh2005; Vittal et al., Reference Vittal, Maruthi Sankar, Singh, Balaguravaiah, Padamalatha and Yellamanda Reddy2003). Besides the above, intensive tillage practices using inversion implements such as the mould board plough result in the loss of surface crop residue and subsequent loss of soil organic carbon from soil aggregates. This, in combination with imbalanced fertilization and poor recycling of crop residues, has resulted in deterioration of soil quality leading to low crop productivity in rainfed regions (Campbell et al., Reference Campbell, Selles, Lafond and Zentner2001; Roldan et al., Reference Roldan, Caravaca, Hernande, Garcia, Sanchez-Brito, Velasque and Tiscareno2003; Sharma et al., Reference Sharma, Kusuma Grace, Mandal, Pravin Gajbhiye, Srinivas, Korwar, Ramesh, Ramachandran and Yadav2008b).
Practices such as zero or reduced tillage, green manuring, recycling of crop residues, have proved effective in improving soil fertility and soil quality in irrigated and temperate regions (Unger, Reference Unger1990). No-tillage (NT) farming, practiced in combination with growing a cover crop in the rotation cycle, is widely recognized as a viable alternative to plough tillage as a way to improve the environment and sustain natural resources. Benefits of no tillage /zero farming (e.g. erosion control, water conservation, soil fertility enhancement, C sequestration) are directly attributed to the amount of crop residue mulch and application of dung/manure as soil amendments (Lal, Reference Lal2007). However, in tropical countries like India, after harvesting of the crops, residue is removed from the soil surface for feeding livestock and / or to use it as fuel for domestic cooking. Beside this, owing to moisture scarcity in rainfed areas, there is very little scope to grow green manure and biomass generating crops in rainfed regions without losing the regular cropping season. Hence, there are very limited amounts of crop residue / biomass available for surface field application.
Sharma et al. (Reference Sharma, Mandal, Srinivas, Vittal, Mandal, Kusuma Grace and Ramesh2005) reported that minimum tillage, when practiced in combination with 90 kg N ha−1 in a castor-sorghum system, maintained a desirable soil quality index of 1.10 in rainfed Alfisols. Further, they reported that to maintain higher crop yield as well as soil quality, primary tillage along with organic residues and nitrogen application are crucial. It has been reported that elimination of summer fallowing in arid and semi-arid regions and adopting no-till with residue mulching improves soil structure, lowers bulk density, increases infiltration capacity (Lal, Reference Lal2004; Shaver et al., Reference Shaver, Peterson, Ahuja, Westfall, Sherrod and Dunn2002) and ultimately enhances crop productivity. Minimum tillage maintains lower temperature, water, oxygen and thereby induces suitable environments for the growth and activity of microflora and microfauna (Blevins and Frye, Reference Blevins and Frye1993; Follet, Reference Follet1990). Thus, optimum tillage operations combined with weed and fertilizer management would be essential not only to enhance the productivity of crops but also to maintain soil health and sustainability over a long period (Maruthi Sankar et al., Reference Maruthi Sankar, Vittal, Ravindra Chary, Ramakrishna and Girija2006; Nema et al., Reference Nema, Maruthi Sankar and Chauhan2008).
Though much effort has gone in to such studies in temperate regions, systematic studies in rainfed semi-arid tropical regions are rare, especially in developing countries because of difficulties in controlling weeds, less water infiltration in compacted soil and the non-availability of appropriate seeding implements (Sharma et al., Reference Sharma, Kusuma Grace, Mandal, Pravin Gajbhiye, Srinivas, Korwar, Ramesh, Ramachandran and Yadav2008b).
Millets in general, are the important crops of tropical and subtropical countries. In Asia, India and China are the two major countries where millets contribute significantly towards the food basket. Among the millets, foxtail millet is the most important in China, whereas in India, it is pearl millet, which is grown on 9.1 m ha land with total production of 7.3 m t and very low productivity of 780 kg ha−1. About 30.7% of the area under pearl millet is in Maharashtra, Uttar Pradesh and Haryana. Since the pearl millet crop is grown in rainfed areas of these states, the present study was focused in these states only. In this study, the long-term effects of tillage and fertilizer practices on productivity and profitability of pearl millet were assessed in Inceptisols of Agra, Vertisols of Solapur and Aridisols of Hisar. The study was planned with the following specific objectives: i) to identify the best tillage and nutrient management treatments in terms of crop yields, yield sustainability and rainwater-use efficiency (RWUE) in rainfed pearl millet under different climatic and edaphic conditions and ii) to develop predictive functions/ models explaining the relationship between crop yield and monthly rainfall.
MATERIALS AND METHODS
Field experiments were conducted in the rainy season (June–September) every year from 2000 to 2008 to study the effects of tillage and fertilizer on the productivity of pearl millet (Pennisetum americanum) at three centres of the All India Coordinated Research Project for Dryland Agriculture (AICRPDA). These centres are located at Agra (Lat. 27.17°N, Long. 78.83°E), Solapur (Lat. 17.68°N, Long. 75.93°E) and Hisar (Lat. 29.17°N, Long. 75.73°E). Soils are Aridisol at Hisar, Inceptisol at Agra and Vertisol at Solapur with their respective climates as arid, arid and semiarid.
The field experiments were conducted in a spilt-plot design with three replications using pearl millet as the test crop. The treatments were randomized and superimposed to the plots in the first year and were continued in subsequent years. The recommended pearl millet varieties suitable for rainfed condition used in the study were MBH–163 at Agra, HHB–67 at Hisar and Shradha at Solapur. These varieties were chosen because of their high yield potential and the same variety was used every year at its respective location. The main plot treatments were three tillage practices: (i) conventional tillage (CT), (ii) low tillage + interculture (LT1) and (iii) low tillage + herbicide (LT2); the subplot treatments were three fertilizer treatments: (i) 100% N from organic source (F1), (ii) 50% organic N + 50% N inorganic N (F2) and (iii) 100% N inorganic N (F3). Hence, the nine treatments tested in the study were: CTF1, CTF2, CTF3, LT1F1, LT1F2, LT1F3, LT2F1, LT2F2 and LT2F3. The tillage operations performed are described in Table 1. The recommended N rate using inorganic fertilizer was 60 kg ha−1 at Agra, 50 kg ha−1 at Solapur and 40 kg ha−1 at Hisar. Farmyard manure (FYM) was used as the organic fertilizer source at all locations. Each year the recommended P2O5 was applied to all plots at a rate of 40 kg ha−1 at Agra, 25 kg ha−1 at Solapur and 20 kg ha−1 at Hisar. Due to variation in the soil and climatic conditions, the recommended rates of fertilizer were different for different sites. Standard recommended crop management practices were used at each location from sowing to harvest (Vittal et al., Reference Vittal, Maruthi Sankar, Singh and Samra2002).
Table 1. Tillage practices adopted for pearl millet at different locations.
DAS: days after sowing.
Soil and agronomic details of the experiments
Soil samples were collected from the experimental sites at Agra, Solapur and Hisar before the start and at the end of the study and were analysed for various physical and chemical parameters using the standard procedures. The same plots were used for the same treatments every year. Soil pH and electrical conductivity were measured in 1:2 soil water suspension (Rhoades, Reference Rhoades, Page, Miller and Keeney1982), organic carbon by the Walkley Black method (Walkley and Black, Reference Walkley and Black1934), available N by alkaline-KMnO4 oxidizable N method (Subbaiah and Asija, Reference Subbaiah and Asija1956), available P by the 0.5 M NaHCO3 method (Olsen et al., Reference Olsen, Cole, Watanabe and Dean1954), available K by the neutral normal ammonium acetate method (Hanway and Heidel, Reference Hanway and Heidel1952), and bulk density using soil cores (Blake and Hartge, Reference Blake, Hartge and Klute1986). Soil water retention at permanent wilting point (PWP) and field capacity (FC) were measured using a pressure plate apparatus at ˗1.5 MPa and ˗0.033 MPa (Cassel and Nielsen, Reference Cassel, Nielsen and Klute1986). The details of these soil parameters and some of the agronomic parameters, i.e. net plot size, spacing and seed rate, are given in Table 2.
Table 2. Details of agronomic parameters and initial soil characteristics of different experimental locations.
AWC: available water capacity; BD: bulk density; EC electrical conductivity; OC: organic carbon.
The values in parentheses indicate the status of nutrients at the end of the study.
Rainfall and its distribution at different locations
The data on monthly rainfall from June to September and the cumulative rainfall of the four months received during 2000 to 2008 were considered for assessing the performance of tillage and fertilizer treatments at different locations. Daily rainfall events of ≥ 2.5 mm only were used for computing the cumulative rainfall of a month for further analysis. Total seasonal rainfall over the study period at Agra ranged from 290 to 766 mm with a mean of 493 mm and coefficient of variation (CV) of 28.9% (Table 3). At this site, a maximum mean rainfall of 208 mm (CV 69.4%) was received in July, followed by 144 mm (CV 66.6%) in August while the lowest mean rainfall of 54 mm (CV 103.0%) was received during June. At Solapur, the total seasonal rainfall ranged from 280 to 573 mm with a mean of 460 mm (CV 19.7%). At this site, maximum mean rainfall of 150 mm (CV 49.5%) was received in September, followed by 144 mm (CV 58.7%) in August while the lowest mean rainfall of 71 mm (CV 56.4%) was received in June. At Hisar, the total seasonal rainfall ranged from 85 to 684 mm with a mean of 345 mm (CV 61.1%). At this site, maximum mean rainfall of 128 mm (CV 97.6%) was received in July, followed by 82 mm (CV 96.3%) in June while the lowest mean rainfall of 56 mm (CV 103.9%) was received in September. A maximum mean rainfall of 493 mm was received at Agra, followed by 460 mm at Solapur and 345 mm at Hisar over the years of the study.
Table 3. Distribution of rainfall during June to September at different study locations
CV: coefficient of variation (%); CRF: crop seasonal rainfall (mm); DOS: date of sowing, DOH: date of harvesting; CGP: crop growth period; F: experiment failed due to severe moisture stress.
Statistical analysis
Analysis of variance of the effects of tillage and fertilizer and their interactions. The analysis of variance (ANOVA) for all the experiments was performed using SPSS version 16 in a split-plot design and the differences between tillage and fertilizer practices over years were compared by least significant difference test (l.s.d.) to a significance level of p < 0.05 (Gomez and Gomez, Reference Gomez and Gomez1984; Kempthorne, Reference Kempthorne1954).
Methodology for developing prediction models between rainfall and crop yields. In order to assess the effect of rainfall on pearl millet yields, treatment-specific linear regression models were developed using grain yields and total monthly rainfall (RF) received from June to September (Draper and Smith, Reference Draper and Smith1998). The expression of the linear regression model is given as:
where, α is intercept and βs are the slopes or regression coefficients measuring the change in yield for a unit change in the rainfall.
For both this and the regression model described in the next paragraph, the coefficient of determination or the predictability of a regression model indicates how much variation in yield could be explained by a set of variables. It can range from 0 to 1. A model with higher R 2 would be preferable to a model with a lower R 2, since the former would capture maximum variability in the data. Further, the predicted yields would be closer to the observed yields based on the model. The prediction error would indicate an estimate of unexplained error based on the model.
Methodology for computing sustainability yield index, rainwater use efficiency and profitability of treatments. The term ‘sustainable’ implies a time dimension and the capacity of a farming system (in this case soil and nutrient management treatment) to endure indefinitely (Lockeretz, Reference Lockeretz1988) in order to identify the best (sustainable) combination of tillage and fertilizer practices for each location, sustainability yield indices (SYI) were computed (Behera et al., Reference Behera, Maruthi Sankar, Mohanty, Pal, Ravindra Chary, Subba Reddy and Ramakrishna2007; Maruthi Sankar et al., Reference Maruthi Sankar, Vittal, Ravindra Chary, Ramakrishna and Girija2006; Nema et al., Reference Nema, Maruthi Sankar and Chauhan2008; Vittal et al., Reference Vittal, Maruthi Sankar, Singh, Balaguravaiah, Padamalatha and Yellamanda Reddy2003). An efficient tillage and fertilizer treatment could be identified based on SYI derived as a ‘ratio of the difference of main yield and prediction error based on regression model and maximum yield attained by any treatment over years’. The SYI ‘Ak’ of treatment ‘k’ could be given as
where Y k is the mean yield of kth treatment and E k is the prediction error based on the regression model of kth treatment.
Rainwater use efficiency. The rainwater use efficiency (RWUE) was determined for each treatment every year using total rainfall and grain production The RWUE (kg ha−1 mm−1) is computed as a ratio of yield and crop seasonal rainfall (Rockstrom et al., Reference Rockstrom, Barron, Fox, Kijne, Barker and Molden2003). Based on ANOVA, the treatment differences for RWUE could be tested and superior treatments could be identified.
Economics factors. In order to compute the economics and profitability of the tillage and fertilizer treatments over years, the gross monetary returns, net monetary returns and benefit-cost ratios were calculated (Nema et al., Reference Nema, Maruthi Sankar and Chauhan2008). The gross monetary returns (Rs ha−1) were computed as a product of the mean yield of each treatment over years and value of the crop at each location. The monetary value of pearl millet was Rs 5.8 to Rs 7.2 kg−1 (mean Rs 6.5 kg−1) at Agra, Rs 8.2 to Rs 9.8 kg−1 (Rs 9.0 kg−1) at Solapur and Rs 6.4 to Rs 7.6 kg−1 (Rs 7.0 kg−1) at Hisar. The net monetary returns (Rs ha−1) were computed as a difference of gross monetary returns and cost of cultivation (Rs ha−1) for each treatment. The benefit-cost ratios were derived as a ratio of gross monetary returns and cost of cultivation for each tillage and fertilizer treatments tested at each location.
RESULTS
Effect of tillage and fertilizer treatments on crop yield
The mean yields attained under different tillage and fertilizer treatments in each year along with the l.s.d. values at p < 0.05 level are given in Table 4. At Agra, among the tillage practices, CT gave maximum mean yield of 1744 kg ha−1 (CV 44.7%) while LT1 gave minimum yield of 1590 kg ha−1 (CV 48.0%) over years. Among the fertilizer treatments, F3 gave maximum mean yield of 1697 kg ha−1 (CV 44.7%) while F1 gave minimum yield of 1601 kg ha−1 (CV 45.0%). Based on l.s.d. criteria, CT was significantly superior over LT1 and LT2 treatments in 2000, 2004, 2005, 2006 and 2007, while LT2 was significantly superior over LT1 in 2003. Among the fertilizer treatments, F3 was found significantly superior to F1 and F2 treatments in 2000, 2003, 2004, 2005 and 2007.
Table 4. Effect of tillage and fertilizer treatments on pearl millet yield (kg ha−1) at different locations during 2000 to 2008.
l.s.d.: least significant difference; F: experiment failed due to severe moisture stress; T: Tillage; Ft: fertilizer; CV: coefficient of variation (%).
† Experiment was conducted only upto 2007.
At Solapur, among tillage practices, CT gave maximum mean yield of 1728 kg ha−1 (CV 37.2%) while LT2 gave minimum yield of 1459 kg ha−1 (variation of 43.0%) over years). Among fertilizer treatments, F3 gave maximum mean yield of 1786 kg ha−1 (variation of 36.2%) while F1 gave minimum yield of 1424 kg ha−1 (variation of 43.2%). Based on l.s.d. criteria, CT was significantly superior over LT2 in all the years except 2001 and LT1 in 2006. The LT1 was superior over LT2 in 2003, 2004, 2005 and 2007, and CT in 2004 and 2005. Among the fertilizer treatments, F3 was significantly superior over F1 in all years, and F1 and F3 in 2000, 2001, 2002 and 2007, while F2 remained significantly superior over F1 during the years 2001 and 2003 to 2007.
At Hisar, the crop failed in 2002 and 2004 due to severe moisture stress and yield data was not available. Among tillage practices, CT gave maximum mean yield of 1741 kg ha−1 (CV 18.9%) while LT1 gave minimum yield of 1704 kg ha−1 (CV 21.9%) over years. Among fertilizer treatments, F3 gave maximum mean yield of 1766 kg ha−1 (CV 20.9%) while F1 gave minimum yield of 1675 kg ha−1 (CV 21.3%). Based on l.s.d. criteria, CT was found superior to LT2 in 2003, 2005, and 2006 while LT2 was superior to LT1 in 2007. The treatment F3 was superior to F1 in 2003, 2005, 2006, 2007 and 2008.
Rainwater use efficiency of tillage and fertilizer treatments at different locations
Because of the dependence of rainfed agriculture totally on rainfall, rainwater is very precious, hence, the slogan given to the farming community in rainfed areas is that ‘water is precious like gold – value it and use it most efficiently’. Thus, enhancing water use efficiency by all means becomes the cardinal principle in rainfed agriculture. During 2000–2008, the RWUE ranged from 1.39 to 5.57 kg ha−1 mm−1 at Agra, 1.38 to 6.01 kg ha−1mm−1 at Solapur and 2.56 to 8.72 kg ha−1mm−1 at Hisar (Table 5). The ANOVA indicated significant differences among tillage and fertilizer treatments for RWUE in different years and also when the data were pooled over years. Considering the l.s.d. criteria at p< 0.05 level, CT was superior with significantly higher mean RWUE of 3.59 kg ha−1 mm−1 at Agra, 3.83 kg ha−1 mm−1 at Solapur and 5.02 kg ha−1 mm−1 at Hisar. Among fertilizer treatments, F3 was superior with significantly higher mean RWUE of 5.04 kg ha−1 mm−1 at Hisar and 3.96 kg ha−1 mm−1 at Solapur, while F2 was superior with significantly higher mean RWUE of 3.46 kg ha−1 mm−1 at Agra.
Table 5. Effect of tillage and fertilizer treatments on Rain water use efficiency (kg ha−1 mm−1) in pearl millet at different locations during 2000 to 2008.
l.s.d.: Least significant difference; F: Experiment failed due to severe moisture stress; T: Tillage; Ft: Fertilizer; CV: Coefficient of variation (%).
† Experiment was conducted only upto 2007.
Profitability of tillage and fertilizer treatments at different locations
The details of cost of cultivation incurred (Rs ha−1), gross and net monetary returns (Rs ha−1), benefit-cost ratio from tillage and fertilizer treatments at different locations are given in Table 6. There was marginal variation in the cost of cultivation and value of the pearl millet grain in different years at all the three locations. At Agra, the mean cost of cultivation ranged from Rs 8332 ha−1 for LT1F2 to Rs 9756 ha−1 for LT2F1 (CV 5.4%). The mean gross returns ranged from Rs 18 227 ha−1 under LT1F1 to Rs 20 240 ha−1 under CTF2 (CV 3.5%). The mean net returns were in a range of Rs 8821 ha−1 under LT1F1 to Rs 11 439 ha−1 under CTF3 (CV 9.9%). The mean benefit-cost ratio was in a range of 1.93% under LT2F1 to 2.33 under CTF3 (CV 7.7%). The analysis indicated that CTF3 was superior for attaining maximum net returns and benefit-cost ratio at Agra.
Table 6. Monetary returns of tillage and fertilizer treatments in pearl millet at different locations during 2000 to 2008.
† Data of 2000 to 2007.
At Solapur, the mean cost of cultivation ranged from Rs 4630 ha−1 for LT2F3 to Rs 7725 ha−1 for CTF1 (CV 21.7%). The mean gross returns ranged from Rs 11 624 ha−1 under LT2F1 to Rs 17 898 ha−1 under LT1F3 (CV 13.2%). The mean net returns ranged from Rs 6654 ha−1 under LT2F1to Rs 12 818 ha−1 LT1F3 (CV 21.1%). The mean benefit-cost ratio ranged from 1.89 under CTF1 to 3.52 under LT1F3 (CV 20.5%). At Solapur, LT1F3 would be a better combination for attaining maximum net returns and benefit-cost ratio.
At Hisar, the mean cost of cultivation ranged from Rs 14 897 ha−1 for LT1F3to Rs 15 987 ha−1 for LTF1 (CV 2.6%). The mean gross returns ranged from Rs 17 638 ha−1 under LT1F1 to Rs 19 013 ha−1 under LT2F3 (CV 2.7%). The mean net returns were in the range of Rs 1995 ha−1 under CTF1 to Rs 3866 ha−1 under LT1F3 (CV 21.2%). The mean benefit-cost ratio ranged from 1.12 under CTF1 to 1.26 under LTF3 (CV 1.3%). At Hisar, LT1F3 would be the best combination for attaining maximum net returns and benefit-cost ratio.
Prediction of yield using monthly rainfall at different locations
Information on the regression models developed between pearl millet yield and monthly rainfall, including the estimates of regression coefficients of monthly rainfall, coefficient of determination (R 2) and prediction error under the model of each treatment are given in Table 7. The regression model gave a significantly higher yield predictability and lower prediction error for different treatments over years at all three locations. At Agra, the R 2 value ranged from 0.81 for CTF3 and LT2F3 to 0.89 for LT1F2 treatment. The predictability was significant only for four treatments: 0.89 for LT1F2, followed by 0.87 for LT1F1 and LT1F3, and 0.85 for CTF2. The prediction error was in the range of 367 kg ha−1 for LT1F1 to 521 kg ha−1 for LT2F3 treatment. The rainfall received in June had a significant positive effect on yield attained by all treatments except LT2F1, while July rainfall had a significant positive effect on yield attained by all treatments except CTF3 and LT2F3. August rainfall had a significant positive effect on yield attained by LT1 in combination with all three fertilizer sources. However, September rainfall had a non-significant negative effect on yield attained by all the treatments.
Table 7. Effect of monthly rainfall on pearl millet yield attained by tillage and fertilizer treatments at different locations during 2000 to 2008.
*, ** significant at p < 0.05 and p < 0.01 level, respectively.
CT: Conventional tillage; LT1: Low tillage + interculture; LT2: Low tillage + herbicide; F1: 100% N (organic); F2: 50% N (organic) + 50% N (inorganic); F3: 100% N (inorganic); R 2: Coefficient of determination.
† Data of 2000 to 2007.
At Solapur, the R 2 value ranged from 0.63 for LT2F3 to 0.92 for LT1F2 treatment. Significant predictability was observed only for three treatments: 0.92 for LT1F2, followed by 0.89 for LT1F1 and 0.86 for CTF2. Based on the models, the prediction error ranged from 312 kg ha−1 for LT1F1 to 613 kg ha−1 for LT2F3. The rainfall received in June had a non-significant positive effect on the yield attained by all treatments, while the rainfall received in July had a positive and significant effect on the yield attained by all treatments except LT2F1, and LT2F3. August and September rainfall had a negative effect on yield attained by all treatments. However, August rainfall was beneficial with a significant effect on the yield attained by four treatments only (CTF2, LT1F1, LT1F2 and LT2F2) while September rainfall had significant effect on the yield attained by three treatments (CTF2, LT1F1 and LTF2).
At Hisar, CTF3 had a minimum R 2 of 0.79, while LT2F1 had a maximum of 0.90 for predicting yield. The predictability of yield was significant for six treatments: 0.90 for LT2F1, followed by 0.87 for CTF1, LT1F1 and LT2F2, and 0.85 for LT1F3 and LT2F3. The model of CTF1 gave minimum prediction error of 143 kg ha−1, while the model of CTF3 gave a maximum of 194 kg ha−1. June and August rainfall had a non-significant positive effect, while July and September rainfall had a non-significant negative effect on the pearl millet yield attained by different treatments.
The regression models between yield and monthly rainfall have clearly indicated that June rainfall had a positive effect, while September rainfall had a negative effect on the yield attained at Agra, Hisar and Solapur. July rainfall had a positive effect on yield at Agra and Solapur while it had a negative effect at Hisar. August rainfall had a positive effect on yield attained at Agra and Hisar while it had a negative effect at Solapur.
Sustainability of tillage and fertilizer treatments over years
Higher estimates of yield predictability with lower prediction errors were observed based on the regression model (1) of yield calibrated through monthly rainfall received in different years. Accordingly, the superiority of tillage and fertilizer treatments was assessed based on the SYI derived using the regression model calibrated for each location. SYI values ranged from 19.9% to 68.3% across the three locations (Table 7). These values ranged from 35.4% for LT2F1 to 42.2% for CTF2 at Agra, 19.9% for LT2F1 to 45.6% for LT1F2 at Solapur and 64.1% for LT1F1 to 68.3% for CTF3 at Hisar. The results indicated a wide range in the SYI values of tillage and fertilizer treatments at different locations due to variation in crop seasonal rainfall, apart from erratic distribution of monthly rainfall received during June to September in each year. Among tillage practices, CT was superior at Agra and Hisar, while LT1 was superior at Solapur. Among fertilizer practices, F2 was superior at Agra and Solapur, while F3 was superior at Hisar. The tillage and fertilizer treatments had a better sustainability of yield of pearl millet at Hisar, followed by Solapur and Agra.
DISCUSSION
Effect of tillage and fertilizer treatments on yield, RWUE, sustainability and monetary returns
In terms of crop yields, at Agra, CT was significantly superior compared to LT1 and LT2 treatments in 2000, 2004, 2005, 2006 and 2007; while LT2 was significantly superior over LT1 in 2003. When mean yields of nine years were considered, CT recorded 9.7% higher yields compared to LT1 and 4.8% yield higher compared to LT2. Among fertilizer treatments, F3 performed significantly superior to F1 and F2 in 2000, 2003, 2004, 2005 and 2007. When RWUE was used as criterion, CT was again found superior with significantly higher mean RWUE of 3.59 kg ha−1 mm−1 at Agra. The F2 treatment gave maximum RWUE of 3.46 kg ha−1 mm−1, while F3 treatment gave RWUE of 3.44 kg ha−1 mm−1. According to the third criterion of monetary returns, CTF3 recorded maximum net returns of Rs 11 439 ha−1 with benefit-cost ratio of 2.33. Thus, based on the nine-year study, CT together with F3 could be considered as superior for attaining maximum productivity, profitability and RWUE in pearl millet in the arid light Inceptisols of Agra.
In case of Vertisols at Solapur when crop yields were considered, CT was significantly superior to LT2 in all years except 2001; and significantly superior to LT1 in 2006. LT1 was significantly superior to LT2 in 2003, 2004, 2005 and 2007; and CT in 2004 and 2005. Among fertilizer treatments, F3 was significantly superior over F1 in all years; and significantly superior to F2 in 2000, 2001, 2002 and 2007. Application of F2 was significantly superior over F1 in 2001, 2003 to 2007. When evaluated in terms of RWUE, among tillages, CT was significantly superior with mean RWUE of 3.83 kg ha−1 mm−1 while among fertilizer nutrient treatments, F3 was significantly superior with mean RWUE of 3.96 kg ha−1 mm−1 over years. LT1F3 resulted in maximum net returns of Rs 12 818 ha−1 with benefit-cost ratio of 3.52. When the pooled values over the period of nine years were considered, though CT and LT1 treatments were at par in their yield levels (1728 and 1726 kg ha−1 respectively) and with RWUE of 3.83 and 3.80 respectively, LT1 in combination with F3 was observed to be profitable with maximum net returns and benefit-cost ratio based on the study. Thus, for pearl millet grown under semi-arid Vertisols of Solapur, LT1F3 could be considered as superior for attaining maximum productivity, profitability and RWUE.
In case of Aridisols at Hisar, when crop yields were considered, among the tillage practices, CT was significantly superior compared to LT2 in 2003, 2005 and 2006, while LT2 was significantly superior over LT1 in 2007. Among the fertilizer nutrient treatments, F3 was significantly superior compared to F1 in 2003, and 2005 to 2008. In terms of RWUE, CT proved superior with maximum mean RWUE of 5.02 kg ha−1 mm−1, while F3 was superior with maximum mean RWUE of 5.04 kg ha−1 mm−1. However, the mean yield (1741 kg ha−1) and RWUE (5.02 kg ha−1 mm−1) attained by CT were marginally higher compared to the yield (1708 kg ha−1) and RWUE (4.86 kg ha−1 mm−1) attained by LT2 over years. However, the study indicated that LT1 in combination with F3 resulted in a maximum net returns of Rs 3866ha−1 with benefit-cost ratio of 1.26. Thus, LT1F3 application could be considered as superior for attaining maximum productivity, profitability and RWUE for pearl millet in the arid Aridisols at Hisar.
When another criterion, the SYI (Table 7) was considered for evaluating the treatments, it was found that the SYI of the treatments ranged from 35.4% for LT2F1 to 42.2% for CTF2 at Agra, 19.9% for LT2F1 to 45.6% for LT1F2 at Solapur and 64.1% for LT1F1 to 68.3% for CTF3 at Hisar. The wide variation of SYI between the locations observed in the present study was due to variation in the climate and soil type, while the wide variation of SYI within a location was due to variation in the amount of rainfall received in each month during June–September, apart from the efficiency of tillage and fertilizer treatments in utilizing the rainwater as measured by the RWUE in different years.
At Agra, the CTF3 which gave maximum mean yield, RWUE, net profit and benefit-cost ratio was also found to be the second best in terms of SYI (39.9%) compared to CTF2 with maximum SYI of 42.2%. At Solapur, LT1F3 which gave maximum mean yield, RWUE, net profit and benefit-cost ratio was found to be the fourth best with SYI of 42.6% compared to LT1F2 with maximum SYI of 45.6%. At Hisar, LT1F3 which was superior for yield, RWUE, net profit and benefit-cost ratio was found to be the fourth best with SYI of 67.8% compared to CTF3 with maximum SYI of 68.3%. The present study has thus indicated that although CTF3 at Agra, LT1F3 at Solapur and Hisar had a marginally lower SYI from sustainability of yield point of view, they can still be considered superior at the respective locations to attain maximum and sustainable yield, RWUE, net profit and benefit-cost ratio by growing pearl millet in different years.
The present study has clearly revealed that CT in combination with 100% inorganic fertilizer (F3) has performed well in terms of crop yield, RWUE, net profit and benefit-cost ratio compared to other treatments in rainfed Inceptisol soils at Agra. This observation on the superior performance of conventional tillage over reduced or minimum tillage is in confirmation with the earlier studies conducted in rainfed Alfisol conditions in a sorghum castor system (Sharma et al., Reference Sharma, Mandal, Srinivas, Vittal, Mandal, Kusuma Grace and Ramesh2005). It is well established that in rainfed agriculture, the two cardinal principles which help in growing a weed-free good crop are: practicing summer tillage (conventional method) to kill the weed seeds by exposing them to hot weather and to capture pre-monsoon and monsoon rainwater in profile by way of loosening the soil surface and enhancing the infiltration rate. Further, conventional tillage also helps in loosening the seed bed for good soil aeration, better root growth and ultimately bumper crop growth. The importance of tillage in weed control has been highlighted by Richey et al. (Reference Richey, Griffith and Parsons1977) and Hatfield (Reference Hatfield1990). They reported that controlling weeds is one of the major reasons of performing tillage and tillage aids in weed control by killing emerging seedlings, burying seeds, delaying growth of perennials, providing a rough surface, which hinders seed germination, providing loose surface soil for efficient action of herbicide and incorporating herbicide when necessary. Perhaps because of some of these above benefits, CT at Agra has proved superior to low tillage (LT). It has been reported that the beneficial effects of low tillage or zero tillage can be accrued more effectively, if adequate crop residue is retained on the soil surface on a long-term basis (Lal, Reference Lal1989, Sharma et al., Reference Sharma, Mandal, Srinivas, Vittal, Mandal, Kusuma Grace and Ramesh2005, Reference Sharma, Kusuma Grace, Mandal, Pravin Gajbhiye, Srinivas, Korwar, Ramesh, Ramachandran and Yadav2008b). Unger (Reference Unger1990) has very clearly emphasized the importance of surface residue maintenance and reported that surface residue, as with conservation tillage (low / zero tillage) systems, reduced runoff and increased infiltration by dissipating the energy of falling raindrops, thereby reducing soil aggregate dispersion that results in surface sealing, and retarding the flow rate of water across the surface, thus providing more time for water infiltration. In the present study, we could not maintain crop residue on the surface as crop residue is generally used for feeding the livestock. Further, while highlighting the importance of more tillage (especially deep tillage) in dry land crops such as sorghum and pearl millet earlier, Vittal et al. (Reference Vittal, Vijayalakshmi and Rao1983) reported that deep tillage upto 23.3 cm helped in improving grain yield by way of better recharge of soil profile and by enhancing the scope for better rooting depth in Alfisols. Further, in general, the rainfed soils are marginally low in fertility, especially nitrogen, the response and performance of inorganic fertilizers remain superior compared to organic sources of nutrients in the short run because of quick release and availability of the nutrients in case of former. Probably, this could be the reason that in the present study 100% inorganic fertilizer proved superior to other fertilizer levels. No doubt, the superior role of long-term use of organic sources of nutrients alone and in combination with inorganic sources in 1 : 1 ratio (N basis) in improving crop yields and physical, chemical and biological soil parameters and soil functions associated with them is well understood and documented (Nambiar, Reference Nambiar2002; Sharma et al., Reference Sharma, Neelaveni, Katyal, Srinivasa Raju, Srinivas, Kusuma Grace and Madhavi2008a). Although, it has been established that in the long-term, tilling the soil more and more or using inversion tillage with implements like the mould board plough may be deleterious to soil quality, in the short run, yield gains remain higher than under low or zero tillage owing to factors mentioned above (Venkateswarlu et al., Reference Venkarteswarlu, Sharma and Prasad2010; Sharma et al., Reference Sharma, Kusuma Grace, Mandal, Pravin Gajbhiye, Srinivas, Korwar, Ramesh, Ramachandran and Yadav2008b). Interestingly, at other two centres, Solapur and Hissar, low tillage in combination with inorganic fertilizer (LT1F3) proved very effective in giving maximum mean yield, RWUE, net profit and benefit-cost ratio. At Solapur, soils are rich in clay and more than 1 m in depth and consequently have higher water retention capacity, and could support the crop well under rainfed conditions. In these soils, the LT1 level of tillage (two harrowings + one hoeing + one hand weeding at 30 days after sowing) probably was adequate to capture the rainwater during pre and post monsoon showers to recharge the soil profile and control weeds, and could perform better in combination with 100% inorganic fertilizer (F3). Similarly, at Hisar where soils are sandy, use of one cultivation with cultivator + one harrowing + one planking + one interculture (LT1) was probably adequate to conserve rainwater, control weeds and to support good crop growth with 100% inorganic fertilizer. Hence, at these two locations low tillage (LT1) in combination with 100% inorganic fertilizer gave superior performance.
Relationship between crop yields and monthly rainfall
The predictability of yield through monthly rainfall was found to be significantly higher at all three locations. It was a maximum of 0.89 with prediction error of 374 kg ha−1 and mean RWUE of 3.24 kg ha−1mm−1 for LT1F2 treatment at Agra. The lowest prediction error of 367 kg ha−1 with yield predictability of 0.87 was observed for LT1F1 which had a RWUE of 3.10 kg ha−1mm−1. The pearl millet yield was significantly influenced by the June rainfall with positive effect in case of all treatments except LT2F1, while July rainfall had a significant positive effect for all treatments except CTF3 and LT2F3. August rainfall significantly increased the yield attained by LT1F1, LT1F2 and LT1F3, while September rainfall did not show any significant influence on crop yields.
At Solapur, the predictability of yield was maximum (0.92) for LT1F2 with prediction error of 337 kg ha−1 and mean RWUE of 3.94 kg ha−1mm−1. The lowest prediction error of 312 kg ha−1 with yield predictability of 0.89 was observed for LT1F1 which had a RWUE of 3.35 kg ha−1mm−1. The July rainfall was important in significantly increasing pearl millet yield under all treatments except LT2F1 and LT2F3, while the distribution of August and September rainfall had a negative effect. However, the rainfall received in June despite its non significant impact, positively influenced the crop yields over years.
At Hisar, the predictability of yield was maximum (0.90) for LT2F1 with prediction error of 147 kg ha−1 and mean RWUE of 4.87 kg ha−1mm−1. The lowest prediction error of 143 kg ha−1 was observed for CTF1 which had a predictability of 0.87 and RWUE of 5.03 kg ha−1mm−1. Although June and August rainfall had a positive effect, and July and September rainfall had a negative effect on the yield attained by all treatments, these were found to be statistically non-significant.
The regression models developed between long-term pearl millet yields and monthly rainfall of June to September, indicated that rainfall received in June is important and plays a vital role in positively influencing the crop yields under arid Inceptisols at Agra, arid Aridisols at Hisar and semi-arid Vertisols at Solapur. Rainfall in June is important because that is the optimum period for preparatory tillage and moisture conservation in soil profile. July rainfall had a positive effect on yield at Agra and Solapur while it had a negative effect at Hisar. July rainfall is also important as this period is optimum for seeding pearl millet crop at most of the study locations. Good rainfall during this month ensures timely sowing, effective germination and good crop stand. The reason for the negative influence at Hisar could be attributed to uneven distribution, intensity and duration of rainfall events. August rainfall is also equally important for uninterrupted plant growth in rainfed regions especially in low water retention soils. Therefore, the positive effect of August rainfall was obvious on yield at Agra and Hisar where soils were mostly low water retentive. The negative impact of August rainfall at Solapur could be attributed to the water stagnation in clayey Vertisols owing to occasional high rainfall events. When the crop heads towards grain formation and maturity, high intensity rains may not be desirable as was obvious with the negative effect of September rainfall at all the three study locations.
The regression models thus developed between pearl millet grain yield and monthly rainfall for the months of growing season for each management treatment (tillage and nutrient) for each study location would be of interest to predict the yield at a given level of rainfall with the likely fluctuation (as error). Further, the error output of these regression models has been instrumental in computing the sustainability of each of the treatments. Such models have been successfully developed and used earlier for soyabean, groundnut and many other crops. (Maruthi Sankar et al., Reference Maruthi Sankar, Mishra, Srinivasa Rao, Padama Latha, Sahadev Reddy, Babu, Ravindrnath Reddy, Veerabhadra Rao, Bhargavi, Ravindra Chary, Osman, Shalander, Vasundhara, Devasena and Girija2010b; Sharma et al., Reference Sharma, Maruthi Sankar, Thakurand and Sharma2009).
CONCLUSION
Based on the study conducted under (i) arid Inceptisols at Agra, (ii) semi-arid Vertisols at Solapur and (iii) arid Aridisols at Hisar during 2000 to 2008, suitable tillage and fertilizer practices have been identified for attaining maximum sustainable productivity and rainwater use efficiency in pearl millet in different years. The mean productivity was found to be sustainable in the range of 1590 to 1744 kg ha−1 at Agra, 1424 to 1786 kg ha−1 at Solapur and 1675 to 1766 kg ha−1 at Hisar. The regression models developed between crop yields and monthly rainfall indicated that the rainfall of June, July and August at Agra, June and July at Solapur and June and August at Hisar was beneficial to pearl millet with a positive effect. Rainfall of September at Agra, August and September at Solapur and July and September at Hisar had a negative influence on the yield attained in different years. The linear regression models gave significant yield predictability in the range of 0.64 to 0.81 at Agra, 0.63 to 0.92 at Solapur and 0.75 to 0.89 at Hisar for different tillage and fertilizer treatments. The SYI achieved through different treatments ranged from 35.4 to 42.2% at Agra, 19.9 to 45.6% at Solapur and 64.1 to 68.3% at Hisar. The CTF3 gave maximum net returns of Rs 11 439 ha−1 with benefit-cost ratio of 2.23, RWUE of 3.52 kgha−1mm−1 and second best SYI of 39.9% at Agra, while LT1F3 gave maximum net returns of Rs 12 818 ha−1 with benefit-cost ratio of 3.52, RWUE of 3.89kgha−1mm−1 and fourth best SYI of 42.6% at Solapur, and LT1F3 gave maximum net returns of Rs 3866 ha−1 with benefit-cost ratio of 1.26, RWUE of 5.05kgha−1mm−1 and fourth best SYI of 67.8% at Hisar. The study indicated that although CTF3 at Agra, LT1F3 at Solapur and LT2F3 at Hisar had a marginally lower SYI; they can still be considered superior for rainfed pearl millet to attain maximum productivity, profitability and RWUE under different climatic and edaphic conditions in different years. Further, the long-term data generated through this study could be used to establish useful yield predictive functions/ models explaining the behaviour of relationships between pearl millet yields and monthly rainfall. The results of the present study are not only useful to the given location but can also work by analogy for developing similar relationship for other crops in various part of the rainfed tropics across the world. The statistical methodology adopted in this study for interpreting the long-term yield data and for better understanding of the relationship between tillage and nutrient management treatments, crop yields and sustainability vis-à-vis crop growing season and monthly rainfall would also be useful to the readers in general and researchers in particular especially in rainfed tropics.
In the Indian sub-continent, which represents mostly subtropical and tropical environment, where lands are mostly at the verge of degradation and soil quality has deteriorated, such studies which warrant the shift of conventional reckless tillage practices to reduced tillage practices or no tillage should be very relevant for the future. However, further such studies need to be conducted in the very long term using appropriate reduced or low tillage, maintaining crop residue on the surface, in strict sense adopting selective weed control practices or herbicides and following proper cropping systems. The authors believe that the information generated in the present study using long-term data-base on crop yield as influenced by tillage and nutrient management treatments, rainwater use efficiency, net returns, cost-benefit ratio and sustainability for different climatic and edaphic locations in rainfed agro-ecology under a well-structured net work programme would be of much interest to the researchers, students, NGOs, land / farm managers and others in tropical and sub tropical rainfed regions, and elsewhere.
