INTRODUCTION
In the semiarid regions from Argentina the expansion of wheat/barley monocultures increased productivity at the expense of environmental stability and sustainability (Viglizzo, 1986; Viglizzo et al., 1991). In this region, the variability in the occurrence of rainfall produces random crop yields. At the same time, the effective depth of the soils (SD) is affected by the presence of a petrocalcic horizon locally known as “tosca”, similar to the “caliche”, ”calcrete” or the “crôutes calcaires” in French (Giai et al., 2002), which has a great variability in CaCO3 content, depth, structure and degree of induration (Pazos & Mestelan, 2002). In the Buenos Aires province only, near 2,866,000 hectares with this limitation exists, (INTA, 1995) being the main limiting factor in the productivity of crops in this region.
The limited depth of soil restrains the radical exploration of the soil, water accumulation (Puricelli et al., 1997, Krüger et al., 2018), stubble accumulation, water use efficiency (Krüger et al., 2014), availability of nutrients (Salih et al., 1989; Volmer & Buffa, 2005) and response to nitrogen fertilization (Frolla et al., 2016), having a negative impact on biomass production and crop yield (Calviño & Sadras, 1999; Sadras & Calviño, 2001; Calviño et al., 2003). At the same time, greater emissions of greenhouse gases have been proven in these soils (Vazquez-Amabile et al., 2013).
The incorporation of no-tillage system in semiarid region over the petrocalcic soils allowed yields to rise as water is used more efficiently (Hansen et al., 1994; Cutforth & McConkey, 1997; Cutforth et al., 2002; Buschiazzo et al., 2007; Schuller et al., 2007) organic matter is preserved (Bossuyt et al., 2002; Balesdent et al., 2000) and wind erosion is reduced (Buschiazzo et al., 2007; Hevia et al., 2007; Hansen et al., 2012). However, the incorporation of no-till management in shallow depth soils with a wheat/barley monoculture in semi-arid regions may not be enough to achieve acceptable yields.
Interaction of the weather and soil depth determines different yields across the spatial variability of soils. Considering the extended presence of the petrocalcic horizon in the region, this relationship generates appropriate conditions for site specific management based on SD. It is of interest, in this case, the determination of critical values for the production of wheat or other winter cereals. These critical values will be useful in the definition of management units. It is postulated that even the use of deep soils would not make crop production independent of quantity and distribution of rainfall, but could improve the response to nitrogen fertilization, the probability of economically viable wheat yields, and generate a criterion for site-specific management. The objectives of this work were i) to determine and rank the effect of variables such as SD, precipitation and soil water availability at different times on wheat yields under zero tillage and ii) to establish a SD critical depth value for less random economic returns.
MATERIALS AND METHODS
The experiment was carried out in a commercial field of 66 ha in which wheat (Triticum aestivum (L.) Thell) is grown every year since 2010 under no-till. The field, located in the extreme western of the Austral Pampa region of Argentina (Figure 1), is representative of the semi-arid southwest of Buenos Aires province. Climate is temperate, continental and semiarid with a mean annual rainfall of 660 mm and mean annual temperatures of 15.2 °C. Rainfall is concentrated in fall (wheat- fallow) and spring (crop cycle) but during grain filling period available water could be lower than crop requirements (Galantini et al., 2014).
The soils are an association of Argiustolls and Haplustolls (SAG y P-INTA, 1989). Later modifications of the taxonomic system place them as Petrocalcic paleustolls (Amiotti, pers. comm.). Soil texture was loamy on the surface and clay loam in the subsoil, with sequence of horizons of type: A-B-C-2Ckm. Determinations made at different points of the study site did not detect significant textural differences. Characteristics of the surface horizon include: 15.9 g kg-1 organic carbon content (Walkley & Black, 1934), 8.6 mg kg-1 extractable phosphorus (Bray & Kurtz, 1945) and pH = 7.4 (soil-water dilution 1:2.5). SD in the studied field ranged between 0.1 and 1 m with a modal value of 0.5 m (Frolla et al., 2015).
During the 2011-2017 period, available soil water content at wheat sowing (AWS) and at crop flowering in November (AWN) and yield (YLD) were recorded in 15 geo-referenced points randomly selected annually. AWS and AWN were measured by gravimetrically (Hillel, 1998) in 0.2m layers to the petrocalcic horizon. At the same points wheat was hand harvested (0.84 m2 plots) and mechanically threshed. Relative yield (RYLD) was calculated as the rate of each point to maximum yield of the year. Simple regression procedures (Cate & Nelson, 1971) and classification- regression tree (Johannes & Hoddinott, 1999) were applied to analyze the relationship between variables by using Infostat® software (Di Rienzo et al., 2012) and R (R Core Team, 2017). Significance levels used were α=0.05.
RESULTS AND DISCUSSION
Table 1 shows monthly rainfall distribution for each year, and average values for available historical records (2003-2017). The amount and distribution of rainfall varied between years and in relation to historical averages. In 2011, precipitation during fallow (January-May) exceeded the average while it was reduced during the crop cycle (June-November). In 2012 it was close to the average during fallow and crop cycle while in 2013 it was scarce during both periods. In 2014, rainfall was normal during fallow and excessive during the crop cycle. In 2015 it was excessive during fallow and relatively low during the cycle. In 2016 it was higher during fallow and somewhat lower than the average during crop cycle. In 2017 it was excessive during fallow and normal during the cycle. Although there were no two similar years, there was a general trend towards more precipitation during fallow than in the crop cycle.
Figure 2 presents YLD according to SD for each year. Average yield per season ranged between 1800 and 3600 kg ha-1, with an absolute maximum in 2016 (5898 kg ha-1) in a deep soil, and a minimum in 2012 (871 kg ha-1) in a shallow soil.
Significant relationships were observed between SD and YLD in six over seven years. Quiroga et al. (2012) identify different factors that, together with soil thickness, contribute to water availability and yield (climate, soil texture, porosity, consumptive use, ancestor crop fallow management). Damiano & Taboada (2000) related the variation in the available water capacity of soils in the Pampean Region with texture and depth of rooting, the latter determined by mechanical limitations in the profile or the characteristics of the crop. Díaz Zorita et al. (1999) observed that the dependence of wheat yields on soil water retention and total organic C contents in years with low moisture availability appears to be related to the positive influence of these soil proper- ties on available water-holding capacity. In this study, since no textural differences were detected, and other factors than climate remained constant, it was reasonable to expect a close relationship between SD and YLD. However, the relationship was different each year explaining SD variation between 36 and 63% of the YLD variation.
According to Scian & Bouza (2005), several authors highlighted the dependence of yields on precipitation during fallow and the wheat cycle in the region. From Table 1 and Figure 2 it appears that the relative distribution of precipitation during fallow and crop cycle was an important factor in the differentiation of yields between deep and shallow soils. The greater water retention capacity of the former can increased yield in some years with low rainfall during wheat cycle (2011, 2012, 2013 and 2017). But for this to be achieved, precipitation during fallow had to be enough to complete a high proportion of this retention capacity like 2017. This did not happen in 2012 but the reloading of the profile occurred in the early stages of the wheat cycle. In 2014, excessive rainfall during crop cycle, nitrogen leaching and high incidence of fungal diseases reduced yields in deep soils and matched them with those of shallow soils. Consequently, the relationship with SD was low and not significant. In 2015 and 2016, sufficient or well-distributed rains may have increased yields in shallow soils reducing the differences with deeper ones.
Figure 3 shows a Cate & Nelson (1971) diagram representing the variation of wheat RYLD according to SD. Observations included in quadrants II and IV make up the model that determined a critical value of 0.52 m SD to obtain RYLD values greater than 0.68. Strictly speaking, this SD value sets a limit for wheat production in the studied area. It also sets a reference value to separate shallow and deep soils for site-specific management in the case of fields with complex patterns. According to the maximum yields observed in each season (Figure 2), a RYLD value of 0.68 corresponds to a range between 1500 and 4000 kg grain ha-1. In a more humid environment and clayey soils, Puricelli et al. (1997) estimated the critical depth for wheat crops at 0.4 m. Bravo et al. (2004), in the same region observed a 54% variation in yield due to the variation of SD, the lowest yields being located in soils with a depth of less than 0.4 m. In sandy soils of the Semi-arid Pampean Region, Quiroga et al. (2012) related soil thicknesses less than 0.8 m with lower wheat yield and response to nitrogen fertilization than deeper soils.
Table 2 shows the results of simple linear regression analysis between the variables SD, AWS, AWN and YLD. All years showed relationship between SD and AWS (R2> 0.31, p<0.06), and six of them with AWN (R2> 0.34, p<0.02). In turn, soil water content influenced YLD: AWS in five years (R2> 0.41, p<0.01), and AWN in six (R2> 0.23, p<0. 07). Relationships between AWS and YLD of wheat and other crops has been described in similar regions (Fontana et al., 2006; Quiroga & Bono, 2007). However, the initial water supply often does not explain acceptably the variation in yields. If rainfall during the cycle is low enough, AWS must be combined with rainfall during the critical period for the definition of yield (Quiroga et al., 2007; Venanzi et al., 2008), even if SD is greater than 0.5 m. As previously mentioned, in years with sufficient rainfall during the crop cycle, AWS has less influence on yield
Within a certain period the available soil water content (AWS or AWN) depends on the precipitation previously received, assuming that there is no significant runoff and evaporation. Verón et al. (2002) estimated that precipitation is the main determinant not only of the net primary productivity of crops of wheat in Argentina but also of its interannual variability, especially in soils where the water storage capacity is low and the system becomes more dependent on precipitation in the crop cycle, as those studied in this case. Different authors in the region explained the variation of wheat yields based on the rainfall received: Loewy (1987) with annual rainfall, Miranda & Junquera (1994) with september-november, Calviño & Sadras (2002) with the rain that occurred 60 days before flowering and 10 days after, Zilio et al. (2014) with august-november and, to a lesser extent, with october-november. The great variability in the amount and distribution of rainfall between years, together with other climatic and biotic factors, determines that these relationships are not always met or that the adjustment of the models is not acceptable. In this study, the amount of precipitation during different periods: fallow, crop cycle (June-November), vegetative stage (August and September), critical period (October and November) and annual rainfall, did not explain significantly the variation of wheat yields (data not shown). There was also no relationship between these variables and the degree of response of YLD to SD, represented by the slope of the regression lines in Figure 2. The limited number of years analyzed may have made it difficult to establish simple linear relationships between these variables.
The existence of collinearity prevents the use of multiple regression models to explain the relationship and the relative importance of the variables involved in the crop’s water dynamics and yield. The classification-regression tree (Figure 4) allows the ranking of variables and the determination of critical values despite this restriction (Johannes & Hoddinott, 1999). In this case, considering SD, AWS, AWN, rainfall during the fallow (RNFf), during the cycle (RNFc) and during the month of November (RNFn) as determining variables and RYLD as a dependent variable, the model selected SD as the first variable determinant of RYLD. With a limit value of 0.52 m, SD separated two groups with average RYLD of 0.56 and 0.80 respectively. The coincidence with Cate and Nelson result is due to the fact that both procedures have the same calculation procedure.
At the second level of discrimination, SD continued to be the most important variable for deeper soils while for soils with <0.52 m SD the model selected was RNFc. At the third level, RNFf was the discriminating variable for deeper soils (>0.68 m) and RNFc for 0.52-0.68 m SD range. On the other hand, the shallower soils that received RNFc> 191 mm were, again, separated for the SD.
Regardless of the validity of critical values and average RYLD, which should be checked in a greater number of years, the hierarchy of variables seems logical on the basis of the experience gathered throughout the study. In the shallower soils, the RNFc is important to achieve acceptable yields. An adequate rainfall distribution, with a certain volume in November, combines the relatively low water retention capacity with its frequent recharge to meet the water requirement of the crop. For this reason, in rainy cycles, the SD vs. YLD relationship usually decreases. In soils with intermediate values of SD and water retention capacity, RNFc is still important to achieve higher yields than in shallower soils. In deeper soils, RNFf allows soil water storage at levels close to field capacity, combined with low RNFc a high SD vs. YLD relationship can be achieved.
Combination of these situations determines the greater or lesser influence of SD on YLD (the slope of the regression line). Although the relationship is significant most of the years, in some of them the yield differences between shallow and deep soils are narrowed, determining a certain level of economic risk if it is intended to make a high use of inputs.
CONCLUSIONS
Soil depth, limited by the petrocalcic horizon, was significantly related to wheat yields explaining between 36 and 63% of yield variation throughout six of seven growing seasons. The relationship is based on its influence on the available water content of the soil at sowing and during the critical period for the crop (November). Due to the poor water retention capacity of the soil, both variables are combined to determine the crop yield. Although there was a relationship between rainfall and available soil water content at these times, with the available data this was not significant.
A reference depth of 0.5 m was obtained, above which an acceptable production of wheat grain is expected. On the basis of the yields observed during the study, the relative yield related to this depth (0.68) corresponds to a range between 1500 and 4000 kg grain ha-1 depending on climate conditions.
The classification-regression tree produced a logical ranking among the variables involved in the definition of the relative yield of wheat. Soil depth was established as first discrimination level, followed by rainfall during crop cycle and in November for the shallowest soils, and rainfall during crop cycle and fallow for the deepest ones.
The results indicate that site specific management based on soil depth is possible and advisable in this environment, provided that there are no important variations in soil texture and relief. The differences in yield observed in most years allow for the formulation of differential strategies for inputs use, mainly application levels of fertilizers and herbicides, which will contribute to the sustainability of the involved production systems.