Field data for this study were collected on dryland maize over seven summer seasons from 2012 to 2019 (2018 missed due to weather conditions) in an experiment established in 2005 under permanent beds and rainfed conditions at the ‘Valle de Mexico’ experimental station of INIFAP, Mexico. The study site location is in the Eastern region of the trans-Mexican volcanic belt (19^o26.7’ N, 98^o53.2’ W, 2280 masl). According to the USDA taxonomy, the soil classification corresponds to a fine-loamy mixed thermic Cumulic Haplustoll (Govaerts et al., 2008). At the onset of the study, soil available P averaged 46 mg P/kg estimated as Bray P-1 extractable, organic matter 24 g/kg with the Walkley-Black method, pH 6.8 (soil:water ratio 1:2), total N 1.1 g/kg following Kjeldahl procedures, and exchangeable K 470 mg/kg extracted with ammonium acetate. The climate is temperate with dry winters; heavy rain events are frequent during the cropping seasons with significant dry spells. As management factors, leaving crop residues as stubble and reshaping beds during winters were the every-year practices.

This study was a single-factor experiment conducted as a randomized complete block design and six replications. Treatments consisted of a single application of 1.1 kg P/ha in liquid form to foliage at the V5 growth stage and 21 kg P/ha in granular form to the soil, banded and broadcast, applied by hand before planting. The application of liquid P fertilizer included a surfactant. Granular fertilizer was incorporated when reshaping beds and during the planting operation. The source of fertilizer was triple superphosphate (45% P₂O₅ per unit of fertilizer). Treatments included a control (0P) plot without fertilizer. Application of glyphosate (726 g a.i./ha) before planting and atrazine (1.8 kg a.i./ha), 30-40d after plant emergence, allowed appropriate weed control.

Plots were planted to a dent-kernel hybrid (H-40), 4-5 cm deep at about 60,000 seeds/ha at the onset of the rainy season when soil moisture was considered appropriate. The plot size was 10 m long by four beds wide (0.8 m each). Harvest area for grain yield estimation was from ears collected from the two central beds after physiological maturity; removing foreign material, impurities, and broken kernels allowed to draw clean grain samples to perform grain evaluations. Grain yield was adjusted to 140 g/kg moisture basis.

A kernel subsample was taken to determine grain P removal, bulk density (test weight), and thousand grain weight (TGW). Grain P removal was estimated from the P concentration using the vanadium-molybdenum method on digested samples with H₂SO₄-H₂O₂. The product of grain P concentration and grain yield estimated grain P removal (kg P/ha). Test weight was defined as the mass of grain in a volume of 1-L cup, including air space, using a filling hopper. Kernel weight was measured from a 200-kernel subsample and multiplied by five to obtain TGW. The agronomic efficiency (AE) was estimated as follows:

where Yf is the yield from the treatments with P fertilizer application, Y(0P) the yield from the control (0P) plot, and Fp the amount of P applied (kg/ha). AE indicates the change in yield per unit of P applied (kg/kg P). Stover P removal was not estimated; stover was left on the soil surface as mulch, assuming that most of the P recovery went into grain (Yu et al., 2021). Phosphorus RE was estimated by the balance method as:

where Gp is grain P removal (kg P/ha). This parameter shows the quantity of P removed by the grain for each P unit applied (kg P/kg P).

Composite soil samples to 30 cm depth were collected before planting to estimate soil P content following Bray P-1 procedures. As weather variables, accumulated precipitation during a 40-d period (40d PP) comprised within the bracketing-silking stage (~20 d before tasseling through ~20 d after) and heat units as accumulated growing degree-days estimated as:

where GDD(t-m) is the accumulated heat units from tasseling to physiological maturity, (Tmax-Tmin)/2 is the average temperature in 24-hr period, and Tb is the base temperature (10 ^oC). However, in days when Tmin<10 ^oC, then Tmin=Tb.

Data were subjected to ANOVA using SAS (SAS/STAT® Software, 2017) to test treatment effects as modeling correlation of times with Proc Mixed statistical model approach. Years and treatments were analyzed as fixed effects, while replications nested years were considered random effects. For the Year-P fertilizer interaction, regression analysis with 40d PP and GDD(t-m) as independent variable explained treatment variations over the years. As Eghball et al. (2003) applied, regression procedures allowed the estimation of soil available P decay over the years using the following non-linear equation:

where Pavai was the level of soil available P, x the time in years (1, 2…7 for 2012, 2013, 2014, 2015, 2016, 2017, 2019, respectively), and A and b non-linear regression parameters.

The average maize yield ranged from a low of 2624 kg/ha in 2019 to a high of 7533 kg/ha in 2016, when the lowest and highest 40d PP occurred, respectively (Table 2). Average yield as affected by P fertilizer application decreased in the order of foliar > broadcast > control (0P) > band (5856, 5507, 5456, and 4888 kg/ha, respectively) with some variations within crops seasons, mainly in the early seasons after the onset of the study (Fig. 1). According to a multiple regression analysis applied to each P treatment, 40d PP and GDD(t-m) combined effect explained 64-83% of the annual yield variations (Fig. 2). In agreement with the regression coefficients, the effect of 40d PP on yield was more significant than GDD(t-m). Similarly, the influence of 40d PP on grain yield variations was more significant for the P application to foliage and lowest for the control (0P) plot. For example, holding constant GDD(t-m) at 600 units, the yield for those two treatments increased 1540 and 960 kg/ha, respectively, for each 30 mm increase in 40d PP around tasseling.

Results showed two crop seasons (2015 and 2019) with negative AEs (Table 2). As defined for this work, negative AEs in these two years indicated that the control (0P) plot yield was higher than the yield from P treatments. Indeed, the annual average of AE showed considerable variation, from-193 kg/kg P in 2015 to 413 kg/kg P in 2014. The effect of P fertilizer treatments on the average AE was higher for the foliar P application (350 kg/kg P) than for the granular band-applied and broadcast (-27 kg/kg P and 6 kg/kg P, respectively) (Fig. 3). However, the variation of AE across 40d PP due to foliar P was more significant (Fig. 4a), while variations for both granular P applications were lower (Fig. 4b). Even though the AE of foliar P treatment showed a tendency to increase as 40d PP increased from 71 to 123 mm, regression procedures failed to identify statistical significance (p=0.13).

Phosphorus fertilizer treatments consisted of the spray of 1.1 kg P/ha to foliage and 21 kg P/ha applied to the soil. Results showed an average RE of 9.3 kg P/kg P; however, the RE of foliar P was consistently higher (Fig. 5a) than granular treatments (Fig. 5b). Recovery due to foliar P ranged between 16 and 38 kg P/kg P, while for granular application, between 0.7 and 2 kg P kg/P. As the Year-P fertilizer interaction was significant (Table 1), replacing the effect of Year in this interaction with 40d PP in regression procedures explained the RE variations. Results indicated a significant effect of 40d PP on the RE of foliar P and broadcast application (Fig. 5a and b, respectively). The effect of 40d PP explained about 70% of RE variations for the foliar and broadcast P application. The band-applied treatment was unaffected by 40d PP. According to the RE-40d PP relationship, as precipitation improved, the corresponding RE of each treatment decreased. Regression coefficients suggested that this reduction (slope) was larger for the foliar P application.

Conversely, RE increased as 40d PP approached dry conditions (intercept). This RE improvement was larger for the application to foliage than granular to soil. In any case, grain P removal (16-38 kg P/kg P) due to foliar P application was substantially higher than the P rate applied. On the other hand, regression analysis with GDD showed that heat units did not affect RE variations.

On average, estimating soil available P on each treatment from samples collected to 30 cm depth on the top of the beds showed differential tendencies over the years; broadcast remained about constant, the foliar and control (0P) decreased, and the band-applied treatment increased (Fig. 6). At the inception of the study, soil available P among treatments averaged about 46 mg/kg. After seven crop harvests, according to a non-linear model to estimate P decay over seasons (Eghball et al., 2003), the available P declined by 3%, 24%, and 34% for broadcast, foliar P, and control (0P) plot, respectively. On the contrary, the band-applied application treatment increased by 13%. Notably, even though the P rate for both granular applications was the same, available P content showed opposite results. Soil available P decay estimation, on the other hand, to reach a level of about 15 mg/kg was in the order of control (0P) plot > foliar (years 2031, 2042, respectively).

Results from the analysis of variance (Table 1) showed that Year’s main effect had a significant influence on grain quality measured as bulk density and TGW. Bulk density varied from a low of 703 kg/m³ in 2012 to a high of 770 kg/m³ in 2015, and TGW from 154 g in 2019 to 344 g in 2015 (Table 2). These two measurements averaged 734 kg/m³ and 282 g, respectively. Pooled data regressed on GDD(t-m) explained the effect of years on bulk density and TGW (Fig. 7); as GDD(t-m) increased from about 500 to 680 units, there was a corresponding increase in these two measurements. According to regression coefficients, TGW was more responsive (slope) to GDD(t-m) changes than bulk density.

Even though the soil tests high available P, maize grain yield showed a differential response to P fertilizer application. In general, grain yield was higher for the foliar P fertilizer. This increment agrees with Rafiullah et al. (2021), who reported biomass increments attributed to foliar P spray. Veneklaas et al. (2012), on the other hand, ascribed this effect to leaf senescence, which may be bound to an improved P distribution within the plant. However, a significant drawback in spraying foliage is that an over-application can lead to leaf burn (Bushong et al., 2016). Nevertheless, as this effect was not observed in this study, the foliar rate and growth stage of application is deemed appropriate.

The differential yield response to P treatments, on the other hand, was linked to precipitation (40d PP) and heat units [GDD(t-m)]. The association between yield response and these two weather factors varied among treatments. Even though the effect was remarkably more significant on foliar P spray, results suggest that weather is a principle that governs the optimum use of P fertilizer (Withers et al., 2005). This suggestion is consistent with Ye et al. (2017), who showed a positive response to post-silking precipitation and GDD. Furthermore, research in Mexico by Murray-Tortarolo et al. (2018) found an average correlation coefficient of r=0.45 between rainfed maize yield and precipitation, but this strength of dependency varies from a low of 0.35 to 0.91 suggesting that other factors such as planting date are also involved. The size of yield variation due to weather factors is consistent with the findings in this study. However, if the amount of precipitation continues decreasing and temperatures increase, the potential effect on maize yield and its relationship with fertilizer P is unknown.

The two methods of granular P application also showed differential results; the average maize yield due to broadcast was higher than the band-applied treatment. This result confirms that the broadcast P application is an acceptable practice for no-till systems (Freiling et al., 2022). Then, once crops remove soil P to an acceptable level, it is necessary to maintain it to prevent yield declination (Sonmez & Pierzynski, 2017). On the other hand, the control (0P) plot produced intermediate yield results, although the cessation of P fertilizer application to soil has been underway for over seven crop seasons. This result indicates that soil available P has not reached a sufficiency Bray P-1 level, despite the presumably decreased supply capacity over time (Tang et al., 2011).

Treatments showed positive and negative AE values; in the latter case, untreated-P plots produced more yield than treated-P plots. This result suggests that available P accumulated from previous fertilization is a valuable resource (Zhang et al., 2022) that needs specific management to maintain adequate yields. Contrastingly, AE for granular P applications was negative in most crop seasons, indicating that this source of P fertilizer warrants concerns in terms of yield. Moreover, although the AE of foliar P was strongly negative in two crop seasons, yield response was consistently larger than granular P applied to the soil. This enhancement indicates that foliar P can generally improve efficiency (Girma et al., 2007; Rafiullah et al., 2021). Nevertheless, the effect of P fertilizer sources and soil available P on AE varies due to multiple biophysical and management factors (Zhang et al., 2022). Accordingly, Girma et al. (2007) found that precipitation is a weather factor that affects AE. However, Yu et al. (2021) conjectured that temperature also affects AE but might depend on moisture conditions. Although the effect of these two weather variables was not statistically significant in this study, accumulated precipitation during the bracketing-silking period seemed to affect the AE of foliar P positively.

The average AE for the foliar P treatment was positive, band-applied negative, and broadcast almost nil. The negative AE for the band-applied treatment raises concerns as previous research indicates that this practice improves efficiency (Ridoutt et al., 2013; Dhillon et al., 2017). Then, as the benefit of the band-applied treatment in this study was not evident, the advantages of this application method should occur at sites with different soil P conditions (Freiling et al., 2022) and tillage practices. Thus, the granular broadcast application seems an appropriate practice to fertilize crops grown under permanent beds in soils testing low P.

Generally, grain P recovery was higher than the P dose applied in fertilizers. However, the RE of foliar P fertilizer was much higher than the granular application. This result suggests that foliar spray enhances the access to soil available P through changes in root architecture (Noack et al., 2010); unfortunately, there is limited information on what degree roots can explain variations in P efficiencies (Yu et al., 2021). Even though P recovery from granular treatments was much lower than foliar P, the RE was larger than one unity indicating that the soil has not reached the critical P level (Johnston & Poulton, 2019). Irrespective of the differential results between foliar and granular P application, the size of RE points out that significant amounts of soil available P have accumulated from past fertilizations and still is the most important source of P nutrition for maize (McLaughlin et al., 2011; Roberts & Johnston, 2015) after seven crop seasons.

The maize crop in this study behaved as plants over-fertilized as they appeared to have achieved a level of ‘‘luxury consumption’’, i.e., not all P taken up by the plants turned into grain yield or biomass production (Cadot et al., 2018). However, as the soil has not reached a critical Bray P-1 level, foliar P fertilization should continue as the crop management practice to maintain the high rate of grain P removal. Nevertheless, once the soil reaches a critical level, breeding programs should have developed maize hybrids that translocate minimum amounts of P to grains (Rose et al., 2013). Research has shown the possibility that reducing grain-P concentrations has no adverse effects on seedling vigor or grain yield (Pariasca-Tanaka et al., 2015).

Phosphorus application to foliage and the control (0P) plot reduces soil available P over the years. This reduction by foliar P suggests crop uptake exceeds the applied P rate. Soil P reduction due to foliar P, as the sole P fertilization source, indicates that the practice is appropriate to reach an adequate soil P level without harming maize yield. Although the control (0P) plot also mined available soil available P, the strategy may not be appropriate due to the effect on yield. According to the non-linear model applied to foliar P to estimate P decay over time (Eghball et al., 2003), the soil would be testing about 15 mg/kg by 2039 (28 years after the onset of the experiment). Assuming this soil’s available P content is the critical Bray P-1 level, the next step will be to maintain it by replacing the P removed by the harvested maize (Johnston & Poulton, 2019). However, once the soil reaches the critical level, crop P requirement might be minor as plant breeding programs advance in developing genotypes with that trait (Withers et al., 2014), but probably still dependent on in-season rainfall (McBeath et al., 2012).

Although the two granular P fertilizer treatments received the same P rate, the resulting soil available P content over the years differed, the band-applied increased, and the broadcast maintained the initial label. In the former case, the fertilizer supposedly concentrates on top of the bed, complicating the collection of a representative sample (Grant & Flaten, 2019), meanwhile the broadcast treatment deposits most fertilizer granules at the bottom of the furrows, while a tiny portion is on the top of the bed. These application protocols may explain the differential results. However, as the purpose is to diminish the high soil available P content, the granular P applications should be replaced by P spray to foliage. To revert the supposed stratified P due to granular P applications in this no-till system into an enhanced soil available P reduction, one-time tillage (Wortmann et al., 2010) seems an option that should be further studied.

On the other hand, irrespective of soil P decay as related to P fertilizer management, grain quality parameters depend on weather conditions, mainly temperature (Tamagno et al., 2016).