DISCUSSION

The repeated fertilizer applications did not produce an important increase in available P, moreover the soils that received the highest rate were just attaining the recommended critical level for L. corniculatus in Uruguay (12 mg/kg) after 5 years. This reflects the high P retention capacity of the soils developed from basaltic rocks (Hernández & Meurer Reference Hernández and Meurer1998). Although these results indicate that P was a limiting factor for pasture production, they should be taken with caution, because there is some evidence that the Bray-1 method underestimates the real P availability in these soils. The causes of this poor behaviour of the method are high organic P mineralization, which is difficult to detect, and strong P retention of iron oxide compared with the Bray-1 solution extraction capacity (Hernández & Meurer Reference Hernández and Meurer1998). In spite of these difficulties, no other soil P test has been extensively used and calibrated for the soils of Uruguay, so consequently the Bray-1 method was used in the current study.

While lotus was successfully established and produced abundant herbage in all the oversown plots, white clover establishment was patchy, dominating in some plots and almost absent in others (data not shown). This heterogeneity could not be related to fertilizer rates, or a given area of the experiment. The persistence of the two species was also different, with a longer productive period for lotus and only 2 years for white clover. This may also reflect that white clover has a higher P requirement than Lotus.

Drawing conclusions on the benefits of oversowing and fertilizer P based solely on total herbage yield would be very misleading without looking at the changes that occurred in grass and legume growth. For both species there was a trend towards lower production as time passed, which has also been reported by other researchers for this type of improved pasture in Uruguay (Carámbula Reference Carámbula1996) and other countries (McIvor et al. Reference McIvor, Guppy and Probert2011). The causes of this have been related mainly to the competence of the native grasses, which have more developed root systems and are able to absorb water and nutrients more efficiently than legumes. In the current study, the legume decline also reflects the cessation of P fertilizer inputs in 2000 and 2001.

The P fertilizer used in most of the applications was single superphosphate, which is also a sulphur (S) source. In a plot experiment, testing different P sources (single and triple superphosphate, phosphate rock and partially acidulated phosphate rock) near to the grazing experiment, the effect of S addition on total forage and L. corniculatus yield was examined under mowing. Phosphate rock plus elemental S, at comparable P rates as single superphosphate, and phosphate rock without S addition were tested. The responses in total forage and legume production to phosphate rock plus elemental S application were not significantly different from those to phosphate rock (data not presented). These results were taken into consideration to interpret the processes that occurred in the grazing experiment. Although it is possible that the benefits of single superphosphate application could be related to the improvement in S nutrition of the pasture, and the decline of legume production after fertilization ceased partially attributed to the decrease in S input, the results of the plot experiment suggest that the main limitation to legume production was P availability.

In spite of the declining trend in legume production throughout the study, they produced a marked increase in N availability in the whole system. Mallarino et al. (Reference Mallarino, Wedin, Perdomo, Goyenola and West1990) reported significant amounts of N transferred from legumes to grasses by decaying nodules, roots and shoots in pastures of Uruguay. The other means of N transfer would be through dung and urine deposition of the grazing animals (Ledgard Reference Ledgard2001). This transference was probably relevant to foster the N content of grasses in P2, because this treatment supported higher stocking rates than P1 in most of the rotation cycles. The rate of N cycling was, therefore, probably faster in P2, which increased the N concentration of the herbage.

An examination of the legume component of the sward shows a marked increase in legume production in the first 3 years of the study, declining in the last year of measurement. The higher legume production was in general counteracted by lower grass production, explaining the lack of differences in total herbage yield in the second to fourth years. A similar trend of decreasing grass production following legume introduction and P application was reported in the south of Brazil by Bandinelli et al. (Reference Bandinelli, Gatiboni, Trindade, de Quadros, Kaminsky, Flores and Saggin2005). Although the current study does not show differences between fertilizer rates overall, in many cuts the P2 produced significantly higher amounts of legume herbage than P1 (data not presented). Interestingly the last 2 years, when P was not added, the oversown pastures produced higher grass yield than NG, maybe reflecting the increasing amounts of N being cycled in these swards, as a result of the legume symbiotic N fixation in the previous 4 years. This higher grass production is therefore the residual response to the applied P fertilizer, but without further P inputs the benefits from the residual P fertilizer will be short-lived.

The seasonal trends of pasture N and P contents, with the lowest contents in late summer, have been reported by other authors (Li et al. Reference Li, Hao, Willms, Zhao and Han2009) and can be attributed to the physiology of the different species. Although native pastures are composites of many species, most of the grasses present the highest nutrient accumulation for the spring growth and the lowest at maturity in the seeding period at the end of the summer (Berretta Reference Berretta and Berreta1998; Ferreira et al. Reference Ferreira, Nabinger, Elejalde, de Freitas, Carassai and Schmitt2011). Coates (Reference Coates1994), in a long-term study of pasture fertilization in Australia, also reported seasonal variations of grass P content, which were attributed not only to plant maturity but also to low soil moisture, conditions frequently seen in late summer. Regarding the decrease in lotus nutrient content at the end of the summer, the low leaf to stem proportion could explain the decline in quality of the forage (Hill et al. Reference Hill, Coates, Whitebread, Clem, Robertson and Pengelly2009). Another possible explanation for the declining P content of the oversown pastures is related to fertilizer application at the end of autumn, which is likely to have lost availability after 1 year, given the high P retention of the soils. As mentioned previously, the soil showed rather low available P values, even in the fertilized plots, throughout the experiment.

Similar P concentrations and seasonal trends were reported by Wunsch et al. (Reference Wunsch, Barcellos, Prates, da Costa, Montanholi and Brandao2006) for NG in the south of Brazil, with an average of 0·7 g/kg of P in the forage. The concentration of P in the herbage biomass of NG was insufficient for the nutrition of the calves, according to NRC (1996) and the critical levels reported by Karn (Reference Karn2001), but are within the range reported by other authors in pastures of this region (Berretta Reference Berretta and Berreta1998; Mazza et al. Reference Mazza, Motta, de Moraes, Vezzani, Adami and Rabel2012). On the other hand, given the relatively low stocking rates, the calves were probably able to select the younger leaves, which are likely to present higher nutrient contents than trimmed herbage (Suttle Reference Suttle2010; Avondo et al. Reference Avondo, Secchiari, Battaglini, Bonanno and Pulina2013).

It was expected, given the important increase in P absorption produced by fertilizer application, that the effect of P rate on the amount of P of the mixed pastures would be apparent. However, significantly higher P amounts in the herbage of P2 than in P1 were found only in the third and fourth years, after four and five fertilizer applications, respectively. Haan et al. (Reference Haan, Russell, Kovar, Powers and Benning2007) attributed increases in pasture P uptake when higher cattle stocking rates were used to manure return. In the current experiment, higher P contents were measured in dung collected in the oversown pastures compared with the native grasses (Del Pino & Hernández Reference Del Pino and Hernández2002), which indicates that P cycling was enhanced by the improvement in diet quality.

The weaned calves gained weight throughout the experimental period of each year. There were exceptions, including calves on the NG treatment in the winter of 1997 and in the summer of 1999 and 2001 that lost weight, as did the calves on the P1 treatment in summer 2001. This was unexpected because in those periods the herbage allowance was similar to the other groups. It is possible that the poor performance could be related to extreme weather conditions affecting animal welfare, or to the poor quality of herbage (Birkelo et al. Reference Birkelo, Johnson and Phetteplace1991; Ferreira et al. Reference Ferreira, Nabinger, Elejalde, de Freitas, Carassai and Schmitt2011). In this region of Uruguay the first cause is more likely to occur in winter, because of frost, and the latter in summer, as observed in the current experiment.

The long-term average annual production in Uruguayan extensive grazing systems is 90–100 kg/ha/year, depending on the technology, which is in line with the result obtained in the NG treatment in the current study (Mila & Tambler Reference Mila and Tambler2011). The increased production from oversowing and P fertilizer addition translated into a 1·8- and 2·5-fold increase in animal production for the P1 and P2 treatments, respectively. If this comparison is limited to the first 4 years when fertilizer was applied each year, this translated to a 2·12- and 2·92-fold increase, respectively, highlighting the importance of P fertilizer inputs. The contrasting results from herbage production cuts and animal production data demonstrate the importance of obtaining animal production data when assessing the merits of oversowing and fertilizer.

Considering that the herbage allowance was similar for different groups, the differences in DLWG could be attributed to herbage quality, which is closely related to the nutrient content of consumed herbage (Coates Reference Coates1994). There was a higher response in DLWG and SR to P1 than P2 rate, indicating a higher efficiency of the P1 rate. However, comparing the effect of the treatments in LW throughout the experiment, the production per kg of applied P were 5·102 and 5·193 kg of LWG for the P1 and P2, respectively, indicating a linear response to the P rate. McIvor et al. (Reference McIvor, Guppy and Probert2011), in a review of many experiments that measured response to P application, reported that the response was in general <4 kg LWG/kg P. This linear response contrasts with the flat pasture production response, but is consistent with the response observed in some periods in the herbage P. It is important to note that the superiority of P1 and P2, compared with the native pasture, is not only attributable to fertilizer P, but also to legume introduction, which should be taken into consideration in the technological evaluation of the different options. The interaction of N (crude protein) and P concentration in the pasture quality have been widely recognized (Karn Reference Karn2001).

The effects of each treatment on LWG of the grazing animals in the last two periods, when no fertilizer was added, suggest that there was a cumulative effect of pasture improvement, as was also observed in herbage yield. This residual effect of pasture improvement also has practical implications, because farmers frequently stop fertilizer applications to improved pastures when the legume population declines.

The declining trend in pasture quality, especially regarding lower legume production, throughout the study produced a lessening in the LWG response to pasture improvement. The outline of this trend is the poor performance of all treatments in the second evaluation period, but this could be mainly due to the extreme weather conditions, with drought during winter–early spring and excessive rainfall in late spring–summer. The negative effects of withholding P application from pastures were also observed in New Zealand by Mackay & Lambert (Reference Mackay and Lambert2011), who reported a decline in quality of the pastures after withholding fertilizer inputs, which was reflected in animal production, and this effect was aggravated with time.

The establishment of relationships between LWG and pasture characteristics was aimed at providing tools for pasture management, especially regarding stocking rate. Simple regression relationships were adjusted: a multiple regression was not tested because of the high correlation among the parameters. Other parameters tested (N and P concentration in herbage, grass yield) did not reach statistical significance. Coates et al. (Reference Coates, Kerridge, Miller and Winter1990) reported close relationships between LWG per head and pasture characteristics, but in the current study per head measurement underestimates the whole effect of the treatment; therefore per hectare beef production was considered a better representation of animal performance. The amount of nutrients in herbage is dependent partly on pasture yield, but also herbage quality; hence it is better related to animal behaviour than herbage production. In the separated set of oversown pastures herbage quality was not likely to represent an important limitation; therefore the main parameter determining animal performance was pasture yield. In contrast, in the native pasture the most important limitation was the low nutrient content and this was reflected in the significant effect of the amount of P in herbage. Hill et al. (Reference Hill, Coates, Whitebread, Clem, Robertson and Pengelly2009) in northern Australia developed relationships between mixed pasture characteristics and LW changes. The pastures varied in species composition, with legume/grass and grass-only swards. They obtained positive correlations of per head LW gains with green leaf herbage mass and proportion, diet crude protein (CP), dietary DM digestibility, diet non-grass % and total green herbage mass, but the relationships found were best fitted by an asymptotic model. The LW gains were relatively poorly predicted by dietary CP, compared with green leaf proportions. This was attributed by Hill et al. (Reference Hill, Coates, Whitebread, Clem, Robertson and Pengelly2009) to the good quality of the herbage. It is important to note that in their experiment, the diet CP was measured from faecal samples and represented the consumed rather than the offered herbage N content. In contrast, Ferreira et al. (Reference Ferreira, Nabinger, Elejalde, de Freitas, Carassai and Schmitt2011) comparing native and improved pastures in Brazil reported close relationships of LWG with total forage mass, but in their experiment the main effect of the treatments were on pasture yield, with similar per head gains of the animals grazing the different treatments. Jones (Reference Jones2003) in Australia also reported linear relationships between per head annual LWG and pasture yield until 4000 kg/ha, with an asymptotic trend thereafter.

These relationships have practical implications, providing a useful indicator for assessing performance where animal production data are not available. For low-quality pastures the amount of available nutrients, especially P, in herbage was a better predictor of beef production than herbage yield. For oversown fertilized pastures, in contrast, herbage availability was a good indicator of animal performance. In a given situation the nutrient content of the available herbage could allow the comparison of pastures with different species composition.

This finding emphasizes the importance of including animal performance and not limiting assessments to herbage production when conducting cost-benefit analysis of oversowing and fertilizer investments. Phosphorus availability in the herbage was a better predictor of animal performance than the pasture yield. The current study highlights that the benefits of oversowing and fertilizer inputs are short-lived, with the withholding of fertilizer in the last 2 years resulting in a loss of sown legumes and decline in animal production.