Filling the yield gap – Optimising yield and economic potential of high input cropping systems in the HRZ
Author: Malcolm McCaskill (Agriculture Victoria Research, Department of Economic Development Jobs, Transport and Resources; Hamilton Research Centre, Hamilton) Kerry Stott (Agriculture Victoria Research, Department of Economic Development Jobs, Transport and Resources, Centre for AgriBioscience, La Trobe University, Bundoora) Brendan Christy (Agriculture Victoria Research, Department of Economic Development Jobs, Transport and Resources, Rutherglen Research Centre, Rutherglen) Penny Riffkin (Agriculture Victoria Research, Department of Economic Development Jobs, Transport and Resources; Hamilton Research Centre, Hamilton) Rob Norton (Norton Agronomic, Horsham, (previously International Plant Nutrition Institute (IPNI), Horsham) Amanda Pearce (SARDI, Struan Research Centre, Naracoorte), Claudia Gebert (Southern Farming Systems Inverleigh) Elly Polonowita, Frank Henry, Debra Partington (Agriculture Victoria Research, Department of Economic Development Jobs, Transport and Resources; Hamilton Research Centre, Hamilton) | Date: 27 Feb 2018
Take home messages
- Under-fertilising appears to be a major cause of yield gaps in cropping systems in the high rainfall zone (HRZ).
- Yield gaps need to take into account seasonal risk and relative crop and fertiliser prices.
- Soil test critical values should be higher than commonly used because of the higher yield potential of the HRZ.
- Return on investment in nitrogen (N) fertiliser is maximised if phosphorus (P), potassium (K) and sulphur (S) are non-limiting.
- The project has produced three Excel-based decision support tools to determine the economic optimum application rate of N, P, K and S under a range of conditions.
In the HRZ of southern Australia, commercial wheat and canola yields are well below their water-limited potential (Yield Gap Australia 2018). The yield gap in this case was defined as the difference between actual yields reported by growers to the Australian Bureau of Statistics (ABS), and a potential yield calculated for each region and cropping year using the Agricultural Production Systems sIMulator (APSIM) model supplied with non-limiting nutrients.
Since nutrient limitations are one of the most common causes of yield gaps, a plant nutrition component was incorporated into the DAV00141 project. One of the questions posed was whether the soil test interpretation guidelines developed in the low and medium rainfall areas were appropriate to the HRZ with its higher yield potential. The nutrition component comprised field experiments, crop modelling, economics, and the development of three Excel-based decision support tools to assist decision makers choose the most economic application rate of various nutrients for a given season.
To determine which nutrients were responsible for crop responses, a series of nutrient omission experiments were conducted in the 2015 and 2016 growing seasons in the HRZ between Bool Lagoon in South Australia (SA) and Rutherglen in Victoria (VIC). At each site, one treatment was supplied with non-limiting rates of all the nutrients to which responses could be expected (P, K, S, copper (Cu), zinc (Zn)), while in other treatments, one or all of these nutrients was omitted. Nitrogen was applied at a minimal rate — 60% of estimated requirements or 100% of requirements. The experiments were conducted with either wheat (cv. Beaufort) or canola (cv. Archer) (Table 1). Soil samples were collected prior to sowing to develop yield relationships appropriate to the HRZ. These included soil N and available K to a depth of 1.4m, and Colwell, DGT-P and KCl-40 available S to 10cm. Further details are given by McCaskill et al. (2016) and Pearce et al. (2017).
In the 2017 season, the experimental program was modified to examine a range of application rates for nutrients to determine the economic optimum nutrient application rate. Results are presented here for a canola P response experiment conducted on the Hamilton Long-Term Phosphate Experiment at five starting fertility levels, and sufficient N applied for it to be non-limiting. Background fertility ranged from a Colwell P of 14mg/kg where virtually no P fertiliser had been applied over the previous 40 years, and to a Colwell P of 143mg/kg where the annual application rate had averaged 27kg/ha.
Data presented here have been analysed in Genstat (18th Edition) using the restricted maximum likelihood (REML) and standard curve procedures, and are reported at the 5% significance level. However, as some of the data are from incomplete data sets, the findings must be considered preliminary.
Utilising the experimental findings of this and previous projects, a series of Excel-based decision support aids were developed. Firstly, we utilised grain yield response relationships to soil tests for P, K and S from this project and the database of Better Fertiliser Decisions for Cropping in Australia (BFDC). Secondly, these were embedded in the Catchment Analysis Toolkit (CAT) model (Christy et al. 2013) to derive a series of predicted yields for wheat and canola in response to a range of fertiliser application strategies across multiple sites and years. CAT is a biophysical model that operates on a daily time-step, and has a dynamic N model.
Scenarios of starting soil conditions and fertiliser application were developed through discussion with commercial agronomists in south-western VIC and southeast SA. Starting soil conditions were based on soil samples collected at the nutrient omission experimental sites. Thirdly, these scenarios were summarised into a series of coefficients for response functions showing diminishing marginal returns and incorporated into Excel look-up tables within the decision support tools.
The spreadsheet tools use conventional marginal investment and return economics to calculate the economic optimum application rate of N, P, K and S for a given set of input conditions, grain and fertiliser prices and the user’s required benefit/cost ratio or rate of return on the marginal dollar invested in fertiliser. The key risk factor is seasonal outcomes and production functions were determined for four season types — ‘very poor’, ‘poor’, ‘good’ and ‘very good’. Three spreadsheet tools were developed from a common base and these address different questions — (i) an awareness tool showing likely response to in-crop N based on the initial P, K and S fertility, (ii) a planning tool to assist with pre-sowing applications of N, P, K and S and in-crop decisions based on climate forecasts, and (iii) an evaluation tool, to check whether the crop was under fertilised or over fertilised, post crop.
Results and discussion
Could full nutrient application close the yield gap?
Grain yields for the ’all’ treatments were close to or exceeded the water-limited yield potential in six of the twelve experiments (Table 1). In four experiments, yields below potential were associated with prolonged waterlogging (Bool Lagoon in 2016 and 2017, and Rutherglen in 2016). For example, wheat at Bool Lagoon in 2017 was inundated continuously mid July until mid November, and yielded 2.6t/ha compared with a region-wide yield potential calculated by APSIM of 6.0t/ha for a rainfall decile of 10. In two experiments, yields below potential were associated with an exceptionally dry finish (canola at Francis and Inverleigh in 2015).
Which nutrients were required and what are the critical soil test values?
Statistically significant grain yield responses were found to N, P, K and S, but not to the micronutrients Cu and Zn (Table 1). The magnitude of the P response was related to the Colwell soil test. The data set from this project was supplemented by four previous trials in the HRZ in the BFDC database. An exponential curve described 64% of the variation, with the 90% critical value at a Colwell P of 30mg/kg (±SE 23 to 44mg/kg) (Figure 1). There was no significant difference between wheat and canola (for comparison, 90% critical values from the BFDC database from all trials in Australia are 24mg/kg for wheat and 20mg/kg for canola). Unlike most relationships in the BFDC database, which plateau at 100% of maximal yield, this relationship plateaued at 88% of maximal yield. This is the ‘starter P’ effect, whereby P banded just below the seed assists early crop establishment.
There were insufficient responses to K and S to derive similar relationships from this project alone. However, from the information collected to date from trials and the experience of crop agronomists, we suggest that the K response relationship for pastures be used for HRZ cropping. The pasture relationship has a 90% critical level at a Colwell K of between 96mg/kg and 109mg/kg Colwell K depending on soil texture (Gourley et al. 2007) (this is much higher than 90% critical values from the BFDC database of 57mg/kg for wheat and 47mg/kg for canola based on trials in drier parts of Australia). For S using the KCl-40 extractant, a preliminary value of 8mg/kg appears to be more appropriate for both than the current BFDC values of 4.5mg/kg for wheat and 6.7mg/kg for canola.
A budgeting approach was used for N to determine application rates for the treatment where we aimed to provide 100% of N requirements. This approach involved calculating plant demand less soil N to a depth of 1m, less an allowance for mineralisation. The approach worked well for wheat but for canola it appeared much of the soil N was unavailable to the crop, despite the crop being highly responsive to fertiliser N. A parallel study (DAV00151 - Understanding how waterlogging affects water and nitrogen use by wheat) has shown that under waterlogged conditions, soil layers below approximately 5cm, become anaerobic. This would limit the capacity of roots to actively take up N and other nutrients, except where the roots have aerenchyma that allow oxygen diffusion. Wheat has aerenchyma in its adventitious roots, whereas canola lacks adventitious roots. This may explain why canola is much more dependent on fertiliser N application under waterlogged conditions than wheat.
Table 1. Summary of nutrient omission and response experiments conducted under the project, including the decile of growing season rainfall (April to November inclusive), measured grain yield of the all-nutrients treatment, the yield potential estimated by APSIM for seasons of the same rainfall decile from the Yield Gap Australia website, and the relative yield (%) where particular nutrients are omitted (only reported where responses were statistically significant).
Yield of ‘all’ (t/ha)
Yield potential (t/ha)
Relative yield if a nutrient is omitted
P (6%), N (24%)
K (83%), N (17%)
P (78%), N (33%), S (68%)
P (76%), S (78%), N (41%)
P (62%), N (59%), S (70%)
P (83%), N (80%)
Figure 1. Relative grain yield response to Colwell P in wheat and canola for experiments in the HRZ in this project, and four previous trials in the BFDC database. Vertical line shows where fitted yield is 90% of the maximal value at a Colwell P of 30mg/kg. Note that because the relationship plateaued at 88% of the yield achievable when P is applied at sowing, the critical value is at 90% x 88% = 79%.
How much nutrient was required?
While soil test response relationships describe the magnitude of response to a non-limiting amount of particular nutrient, they do not indicate the economic optimum amount to apply. This needs a fertiliser rate experiment such as that in Figure 2 (or equivalent model output such as from CAT). Here, seven rates of P were applied to fields with starting P fertility ranging from 14mg/kg to 143 mg/kg Colwell P. Canola grain yield followed a common relationship once adjustment was made for the starting fertility. For example, at a background P of 53mg/kg Colwell, yield of the nil P treatment was equivalent to a treatment receiving 58kg P/ha at a starting fertility of 14mg/kg Colwell.
Figure 2. Canola grain yield response to applied P for a starting fertility of 14mg/kg Colwell, on the Hamilton Long-term Phosphate Experiment in 2017. Starting fertility ranged from 14mg/kg to 143mg/kg Colwell P, and P rates are adjusted so they are equivalent to the lowest starting fertility.
Table 2. Background Colwell P of the response experiments on the Ham ilton Long-term Phosphate Experiment, the long-term (40 year) annual P application that has produced the fertility level, the equivalent P application rate of the background P using the combined relationship in Figure 1, and the economic optimum P application rate at a 2:1 benefit cost ratio for canola at each background level.
Starting soil fertility
Economic optimum P application rate
Agricultural economists calculate the optimum fertiliser application rate as where $1 of extra grain is produced from $1 of extra fertiliser (Figure 3a), which is a 1:1 benefit cost ratio. A 1:1 benefit cost ratio is suitable if there is a high level of confidence in the response relationship, and no cost of capital. However, if there is some doubt whether a fertiliser investment will return sufficient additional yield despite seasonal variation and other possible crop growth constraints, a benefit cost ratio of 1.25:1 or 2:1 may be preferred, but the overall profits will be lower in the long term. In the example of P application to canola at Hamilton, the optimum P application rate at a 2:1 benefit cost ratio was 88kg P/ha less the allowance for background fertility (Table 2). Key factors that favour either high or low optimum application rates are:
Higher optimum fertiliser application rates
Lower optimum fertiliser application rates
High crop prices
Low crop prices
Low fertiliser prices
High fertiliser prices
1:1 benefit cost ratio optimum
2:1 benefit cost ratio (or wider)
The yield factor is illustrated in Figure 3(b) by using the same curve as in Figure 2 scaled down to represent lower yield potentials in the Wimmera and Mallee. The 2:1 economic optimum occurs at 92% of yield potential in the HRZ, compared with 83% in the Wimmera and 66% in the Mallee. Soil tests are often interpreted in relation to a critical level at which 90% of maximum yield is achieved, whereas a higher threshold should be used in areas of greater yield potential.
The crop price factor is illustrated in Figure 3(c) by using the wheat price of $224/t in the canola yield response relationship, rather than the $495/t canola price. The economic optimum at a 2:1 benefit cost ratio declines to 55kg P/ha (from 88kg P/ha), less the allowance for background fertility.
Figure 3. (a) Economic optimum nutrient application for a 1:1 and 2:1 benefit cost ratio; (b) economic optimum P application (circles) for a 2:1 benefit cost ratio for yield potentials representative of the HRZ, Wimmera and Mallee using the same curve as in Figure 2, and the fertility required for 90% of yield potential in all three environments; (c) economic optimum P application at a 2:1 benefit cost ratio for canola using the same curve as in Figure 2 and current prices, and for wheat if the yield response relationship also applied to wheat.
It should be noted that this P response relationship was for a soil with a Phosphate Buffering Index (PBI) of 200, whereas the average PBI of commercial samples submitted in 2015 to the Nutrient Advantage laboratory from south-west VIC was only 108 (McCaskill et al. 2016 and unpublished). While a similar relationship would apply to all soils in the HRZ, the economic optimum application rate is likely to be lower than shown here.
What if two or more nutrients are limiting?
In the P rate experiment given above, non-limiting rates of N were applied, and N was not considered in the economic optimisation. In practice, most sites have an interaction between two or more limiting nutrients. This is illustrated from the 2016 wheat omission experiment at Bool Lagoon (Figure 4). There was a strong response to additional N where all the required other nutrients were applied, but there was a weaker response if P or S were omitted. Where both P and S were applied, each additional kilogram of N fertiliser between the mid and high rate of N produced 11.7kg of extra grain, compared with 6.8kg if P was omitted, 4.5kg if S was omitted and no additional yield if both were omitted. Correction of other nutrient limitations is the first step in obtaining a good response to applied N. Conversely, the P and S responses were only statistically significant at the high, but not the mid-rate of N. Similar findings were made from the other omission experiment sites. As cropping in the HRZ adopts varieties with higher potential yields and higher N rates are applied, we can expect more responses to P, K and S unless soil conditions are closely monitored.
Figure 4. Wheat grain yield response to applied N at Bool Lagoon in 2016 as affected by the omission of all other nutrients at sowing, and the omission of P or S, at N application rates of 30kg, 68kg and 187kg N/ha. Error bars show the 5% least significant difference. Redrawn from Pearce et al. (2017).
Putting it together — decision support
Since the economic optimum changes with input costs and product prices, economic information is better conveyed by calculation tools than static information. The tools combine well established production economics principles with relatively poorly developed (to-date) nutrient response relationships from the HRZ and are available on the eXtensionAUS website. The spreadsheet tools allow users to adjust prices for crops and inputs and reveal optimum nutrient ratios and fertilisation levels for the range of seasonal conditions. For limited capital and/or high risk situations, users are also able to specify their required benefit/cost ratio or rate of return on the marginal dollar invested in fertiliser. Simple graphs and tables were used to illustrate expected outcomes. A screen grab from the awareness tool (Figure 5) shows how limitations of P, K or S affect the optimal application rate of N.
Figure 5.a Screen shot of the awareness tool, showing some of the input data required.
Figure 5.b A dynamic calculation of the economic optimum N application under conditions of limited P, K or S, and if these nutrients are fully supplied.
The effect of season variability on the optimal fertiliser strategy is accommodated by a drop-down box of yield quartiles. At sowing, these yield outcomes have equal probability, and possible N, P, K and S fertiliser strategies can be tested under both good and poor seasonal conditions. As the season progresses, the probability of achieving a particular yield outcome becomes more certain because of rainfall received after sowing, and drought influences become apparent such as El Niño or a positive Indian Ocean Dipole (IOD). Much of this information is available in late August and can influence decisions on split N application in late winter and early spring. The planning tool allows users to test how these factors affect the probability of achieving a low or high final yield, and the economic optimum N application rate. We expect to conduct training and feedback sessions with the tools over the next year, leading to improved versions. Eventually the tools may be made available in other forms, through incorporation into existing decision support tools and possibly smartphone apps, but the current Excel form provides a way of prototyping in parallel with gathering more information on nutrient response relationships.
Through a series of nutrient response experiments, we have established that by providing sufficient nutrients, the yield of wheat and canola crops can be equal to or exceed the water-limited potential, except in cases of severe waterlogging or drought. The strongest responses were to P followed by N, S and K. The magnitude of these responses was related to soil tests, but with critical values at which 90% of maximal yield was achieved slightly higher than from previous trials in other parts of Australia. Economic analysis showed that the 90% critical value underestimated the economic optimum because of the higher yield potential in the HRZ. Since the economic optimum fertiliser application rate is also dependent on input prices, product price and seasonal outlook, we have prepared three spreadsheets to calculate the optimum under a wide range of conditions. The spreadsheets are populated with yield and nutrient response data from a biophysical model, but allow modification to suit individual circumstances.
BFDC (Better Fertiliser Decisions for Cropping in Australia) (accessed 19 Jan 2018)
Christy B, O’Leary G, Riffkin P, Acuna T, Potter T and Clough A (2013). Long-season canola (Brassica napus L.) cultivars offer potential to substantially increase grain yield production in south-eastern Australia compared with current spring cultivars. Crop and Pasture Science 64, 901–913.
Gourley CJP, Melland AR, Waller RA, Awty IM, Smith AP, Peverill KI, Hannah MC (2007). Making Better Fertilizer Decisions for Grazed Pastures in Australia. Department of Primary Industries, Melbourne.
Pearce A, McCaskill M, Ludwig I, Partington D, Riffkin P (2017). What are the limiting nutrients for crops of high yield potential in the South East of South Australia? Edited by Garry O’Leary. Proceedings of the 18th Australian Agronomy Conference 2017, 24-28 September 2017, Ballarat, Victoria
McCaskill M, Stott K, Christy B, Clough A, Norton R, Pearce A, Crozier C, Killoran J, Henry F, Francis M, Vague A, Farlow C, Henson C, McLean T, Partington D, Edwards J, Seven J, Riffkin P (2016). Crop Nutrient Decisions in the High Rainfall Zone: Technical Report. Department of Economic Development, Jobs, Transport and Resources, Melbourne.
Yield Gap Australia (2018) (accessed 17 January, 2018)
The research undertaken as part of this project is made possible by the significant contributions of growers through both trial cooperation and the support of the GRDC — the author would like to thank them for their continued support. We thank the Victorian Department of Economic Development, Jobs Transport and Resources, and the International Plant Nutrition Institute (IPNI) for co-investing in this project. We also thank the landholders upon whose properties the experiments were conducted, and the technical staff who have contributed to the study.
915 Mt Napier Rd, Hamilton, Vic 3300
03 5573 0957
GRDC Project code: DAV00141
Was this page helpful?