Take home messages
- ‘Uniform’ paddocks in the Southern Mallee still demonstrated economically significant variation in response to fertiliser.
- These paddocks might be worth another look for growers considering variable rate application.
Over years of management, growers may observe consistent patterns of crop performance in different areas of a paddock. This can be due to underlying constraints that, if managed with tailored input application rates, could return production or cost benefits. This is known as Variable Rate Application (VRA).
VRA of fertiliser involves varying inputs to target paddock zones that will yield higher if more inputs are applied, or maintain yield with fewer inputs applied. Savings of up to $57/ha have been reported from VRA of fertiliser within a paddock (Robertson et al. 2007).
However, there are many potential sources of variability in a paddock and the benefits of VRA tend to be paddock specific. It is important to understand what is likely to work prior to widespread adoption of VRA. If there is no production gain or cost saving to be made in a paddock, VRA of fertiliser is unlikely to have a fit. At worst, apparent fertiliser savings may be a net loss if they cause a greater loss in yield.
There is a common perception that many paddocks in the southern Mallee are too ‘uniform’ to justify VRA. However, the challenges involved in identifying paddock zones and separating variability in nutrient response from other spatial characteristics suggest room for uncertainty in this assumption, and the value in demonstrating more focused methods of assessing paddock variability.
To ascertain whether a production or financial benefit can be gained through the use of VRA of fertiliser on four ‘uniform’ and four ‘variable’ paddocks, and demonstrate a practical approach to this kind of assessment.
Four growers across the southern Mallee each selected two cereal paddocks. These consisted of a ‘uniform’ paddock and a ‘variable’ paddock, identified on the basis of previous experience and perception. For paddock details see Table 1.
Table 1. Paddock details of eight paddocks selected by growers. Rainfall, crop type and variety and dates of relevant operations.
Trial details and inputs
Test strips were established in each paddock with varying fertiliser inputs at sowing and topdressing based on each growers standard practice (Table 1.). Total inputs are outlined in Table 2. Pests, weeds and diseases were managed by individual grower practice.
Table 2. Strip treatment outlines for each paddock paired by grower and identified as ‘uniform’ or ‘variable’. Grower practice outlines the surrounding paddock treatment/normal management. Zero phosphorus (P) treatments are used to measure P responsiveness. Low nitrogen (N) and high N treatments are used to measure levels of N responsiveness. Reasonable efforts were made to balance for non-target nutrients but this was not always practicable with grower operations at the paddock scale.
Historical paddock performance, based on satellite NDVI data, gamma radiometrics, and elevation, was analysed using the ‘PA Stack’ online tool. Based on this analysis, soil sampling was conducted across three transects within each paddock, targeting areas of differing average performance, or higher variability in performance. Yield Prophet® subscriptions were established for each transect within Grower 1’s paddocks to provide an indication of variation in yield potential.
Demonstration strips were established as outlined in Table 2. Nutrient response within the season was assessed using NDVI data collected by UAV at approximately GS30 and using yield maps collected by growers. A kriging process was applied to reduce the level of noise in the raw yield maps, and numerical data was extracted using the QGIS platform.
Nutrient response was calculated as the difference between the treatment strip and closely paired sampling areas in neighbouring control strips. For example, a much higher average yield in sampled points within a high N treatment strip compared to paired samples within the neighbouring low N control treatment strips would indicate a strong N response at that point in the paddock (see Figure 1).
Results and interpretation
Variability was evident in the results from initial soil sampling. A greater spread in values generally occurred in ‘variable’ paddocks. For example, in Grower 1’s paddocks:
- PBI varied between 17-91 in the ‘variable’ paddock, compared to a consistent 120 for the ‘uniform’;
- Organic carbon varied between 0.25-1.1% in the ‘variable’ paddock, compared to a narrower
0.99-1.4% spread in the ‘uniform’.
Based on these soil test results, Yield Prophet modelling on the 15 August suggested that the 50 per cent probability water limited/nitrogen unlimited yield potential varied by 0.5t/ha between transects in the ‘uniform’ paddock compared to 1.9t/ha in the ‘‘variable’’.
Most paddocks received above average growing season rainfall, although the pattern of rainfall was highly variable. Due to technical difficulties, data from Grower 3 was not received in time to be included in the analysis for this report. In the case of Grower 2’s paddocks, patches of low plant density caused by crabholes and waterlogging contributed to a very high level of variability in results. Yield data from Grower 2’s ‘variable’ paddock was also affected by spray operations that resulted in several large bare patches.
Yield variation was observed across control strip(s) in every paddock. Paddocks that were identified as ‘variable’ had consistently higher variability as measured by the standard deviation of yield (Table 3).
Table 3. Yield data summary statistics.
Variability in nutrient response, as measured by the standard deviation, was generally higher for ‘variable’ compared to ‘uniform’ paddocks (see Table 3). However, responsiveness was poorly correlated with yield in the majority of cases. This suggests that yield variability was not a sufficient predictor of the location of response variability.
Grower 1’s paddocks both demonstrated high degree of small scale variation (Figure 2). The ‘uniform’ paddock displayed only minor responses to the P treatment, without any large contiguous areas clearly outperforming others. N response was much more clearly differentiated between a highly responsive area in the first third of the trial and lower responses elsewhere. The ‘variable’ paddock displayed similar but opposite patterns. There were two areas of higher P response, separated by a low responding area. Apart from small-scale variation, there was a uniform N response across the paddock.
Grower 2’s ‘variable’ paddock had an extreme degree of variability in response to both treatments. The ‘uniform’ paddock also displayed a very ‘noisy’ pattern of small scale variation. In both cases, this included multiple areas of apparently negative response. However, it is likely that these results were heavily affected by the in-season factors outlined above. Small scale bare patches that fall into only one treatment strip can result in extreme values in either direction that are not truly indicative of nutrient response.
Grower 4’s ‘uniform’ paddock displayed distinct zones of high P response in the middle of the trial and N response at the far end. The ‘variable’ paddock also demonstrated a high N response over substantial areas. P response was characterised by a high degree of small scale variability. Apart from a handful of outlying data points, only the first half of the trial appeared P responsive.
Variability needs to occur in a targetable pattern for effective VRA management (GRDC, 2006). As with previous BCG research, variability in the response to nutrients has been used as the main indicator of where a benefit to VRA may exist. This is because the optimal nutrient rate that maximises profit for each management zone is affected not by yield itself, but the yield response to fertiliser (Abadi and Farre, 2015). Applying more nutrients to higher responding areas, or less to non-responsive areas, makes intuitive financial sense.
The results of this trial support growers’ ability to differentiate between relatively ‘variable’ and ‘uniform’ paddocks on the basis of past experience, both in terms of yield and apparent response to nutrition. However, variability in response was still evident in paddocks considered ‘uniform’.
The treatments in this trial were chosen to demonstrate whether responses were consistent across a paddock. How economically significant they are can be gauged by comparing the response to the breakeven response required to pay for the treatment.
The results of these calculations for Grower 1 and 4 are shown in Figure 3. On this basis, ‘uniform’ paddocks still have economically significant variation in response. 61 per cent of Grower 1’s ‘uniform’ paddock achieved less than the breakeven yield response for the amount of nitrogen applied, and 70 per cent was below the breakeven yield for P. Similarly, Grower 4’s ‘uniform’ paddock had 18 per cent below the breakeven for N applied, and 29 per cent for P. If they were predictable, both growers could have increased profit by applying less nutrients to these areas.
Would it be practical to target these areas of different performance? Grower 1’s ‘uniform’ paddock is a promising candidate for VRA of nitrogen, since it is clearly divided between a large responsive and non-responsive area. In the context of relatively uniform yield potential, this could indicate a deficiency in the responsive area. If the pattern was identified using test strips and in-season imagery (e.g. satellite, drone), and if appropriate to the season, the grower could target additional applications to this area.
The potential benefits of VRA of phosphorus in the same paddock is not as clearly indicated, and would have been more appropriate in the ‘variable’ paddock. Variation that occurs over a very short distance is hard to differentiate from random ‘noise’ and impractical to target using a zone based VRA approach, although there may still be large-scale trends under the ‘noise’ that are worth targeting. Ultimately a more detailed, longer term understanding may be required to answer this effectively and enable accurate zoning.
The results from Grower 2’s paddocks demonstrate the importance of validating test strip results against a broader knowledge of the paddock and the season. It is important to be able to differentiate between relevant underlying differences, and season-specific factors that can affect the reliability of this approach.
Ultimately, both in-depth grower knowledge and targeted analysis are required to determine whether VRA of fertiliser is an appropriate response to observed variability. Nevertheless, this trial suggests that the ‘uniform’ paddocks of the Southern Mallee might be worth a second look for growers who are considering VRA.
Abadi, A., and Farre, I., 2015, ‘A simple framework for profitable fertiliser use under risk and soil constraints’, Building Productive, Diverse and Sustainable Landscapes – Proceedings of the 17th ASA Conference.
This research was funded as part of the National Landcare Programme – Sustainable Agriculture Small Grants Round 2015-16. BCG acknowledges the extensive assistance of the participating growers in performing this trial.
CHAIRMAN’S WELCOME I THE BIG DOZEN I BOARD, STAFF AND COMMITTEES I THE YEAR THAT WAS 2017 I 2017 RESEARCH SITES I SITE DESCRIPTIONS I BCG RESEARCH METHODOLOGY I GUIDELINES FOR INTERPRETING SOIL TEST RESULTS I GRAIN PRICES I PRODUCTION COSTS I ACRONYMS AND ABBREVIATIONS I WHEAT VARIETY SPECIFICATIONS I NVT WHEAT YIELD RESULTS I BARLEY VARIETY SPECIFICATIONS I NVT BARLEY YIELD RESULTS I CANOLA VARIETY SPECIFICATIONS I NVT CANOLA YIELD RESULTS I NVT PULSE YIELD RESULTS I CEREAL GROWTH STAGE CHART