1Farming Systems Institute, QDPI, Toowoomba
2School of Mathematical Sciences, QUT, Brisbane
3CIAT, Cali, Columbia
4Wesfarmers Landmark, Dalby
Phone 07 4688 1524; Fax 07 4688 1192
Although precision agriculture tools employed by farm managers have the capability to provide data-rich information, agronomic interpretation of this information lags well behind. Protein mapping, using remote or on-board sensor technology, is set to improve this interpretation, but a framework is needed that can extract the maximum value from these data. We have collected coincidental yield and protein data sets across cereal crops in northern Australia for 3 years, and a number of useful strategies have been demonstrated that maximise the extraction of agronomic information for advisers and farm managers. Nitrogen (N) management, using N removal, N supply and N deficit maps, and probability maps, that estimate the likelihood of response to applied N, can be obtained with coincidental yield/protein maps. Water use efficiency and starting-soil moisture supplies, using reverse-cycle modelling, can also be verified from these layers. Economic interpretations, using gross margins and sensitivity analyses, can similarly be provided. Issues still to be resolved in mapping coincidental maps are discussed.
Information technology, in synchrony with global positioning systems, has created opportunities for site-specific land management through the collection and synthesis of resource and production data. However, collection of information over space and time has outperformed our ability to interpret and apply the data. Consequently, the major goal of site-specific management, which is to see a continuous improvement of management decisions, has lagged as a result of poor agronomic interpretation and information delivery (Cook and Bramley 2001). Engineers, agronomists, and geo-spatial analysts must together develop frameworks to maximise the flow of information back to farm managers to improve land management.
Protein mapping is set to provide a unique dimension in site-specific management (Engel et al. 1997; Williams 2000). It is timely, then, to devise a framework for analysis and interpretation that ensures management decisions, derived from coincidental yield and protein maps, are fully informed.
We have collected coincidental yield and protein maps from grain crops over 7 seasons throughout the northern cereal belt. We have used agronomic principles to derive recommendations for the farm manager in relation to supplies, where they are yield-limiting, of soil water and nitrogen. A number of specific outputs are presented that will improve the usage of coincidental yield and protein data, and allow farm managers to properly evaluate the cost-effectiveness of collecting these data.
Methods of data collection and mapping
Yield and protein data collection
Site-specific yield was collected using an AgLeader mass flow yield monitor with positions provided by an OmniSTAR differentially-corrected global positioning system (GPS). The monitors were calibrated at the commencement of each season, and accuracy errors were found to be less than 1% over the period. The monitor also collected grain moisture.
Grain samples were collected during harvest using a customised sampling device capable of siphoning off a 50–100 g sample every 30–50 m. A serially linked palm-top captured GPS location and GPS time, and recorded this against the sample numbers. Samples were then analysed for protein and moisture using a Bran-Luebbe Infra-alyzer 500, and matched to the original site.
Data filtering and matching
Both yield and protein data sets were corrected to moisture contents of traded grains (i.e. wheat 12.5%, malting-grade barley 0%, and sorghum 13.5%). The data sets were combined using two alternative methods based either on time alignment or spatial alignment. However, the clock times recorded by the yield monitor relate to the UNIX definition of time (based on coordinated universal time or UTC), while those recorded from the GPS relate to GPS time. Alignment was possible – but complicated due to the inclusion of several “leap seconds” since the 1980’s. The second option was chosen to avoid the risk that the times were misaligned. Once the projected points were transformed into mapping grid units, each yield and protein set was interpolated to a common 10-m grid using a local variogram approach by kriging using Vesper (Minasny et al. 1999). The interpolated data were combined for coincidental interrogation.
Further comparisons were made in an attempt to reduce the residual errors associated with each coincidental estimate. For example, a “moving window” approach to obtain regression relationships between yield and protein with either global or neighbouring estimates was attempted, but the reduced residual error was associated with lower precision (Norng et al. 2001).
Outputs of coincidental yield and protein mapping
The results of a suite of N trials of wheat, barley, and sorghum throughout the northern grain belt can be employed to make inferences about the state of site-specific management of N supplies to cereal crops.
Nitrogen budgeting techniques can be used to estimate the amounts (kg/ha) of N removed from the field as the crop is harvested (e.g. Dalgliesh and Foale 1998). Nitrogen removal (kg N/ha) can be estimated by:
N removal = yield (t/ha) . protein (%) . a
where a = 1.75 for wheat and 1.6 for barley and sorghum. An understanding of the spatial variation in N removal can indicate zones within the field that are in an N-deficit. Under tactical N management, managers will then attempt to replace what was removed as grain, and “balance the budget” as much as possible.
In addition to N being removed from the field in the form of grain, various amounts of N will remain in stubble and root material. This N reserve will slowly become available to successive crops at rates dependent on the management of crop and soil and the carbon/N ratio of crop residues. Strategic application of N requires one to understand what supplies of N have been utilised to produce a crop or series of crops and to offset the removal rate by N additions. Reserves of N that have been temporarily “tied up” through immobilisation in stubble/root material are then accounted for.
As a rule-of-thumb, the transfer efficiency of N to cereal grains is around 50% (Dalgliesh and Foale 1998). However, this efficiency of transfer from N supply to N removal does vary according to the supply of plant-available N. For example, wheat grain at 11.5% requires 1.7 times the available-N as that removed by the grain, whereas wheat grain at 13.5% requires 2.2 times the available-N as that removed by the grain (W. Strong, unpublished data). For wheat, this inverse-linear transfer function can be expressed as:
N transfer (%) = a – b . protein (%)
where a = 135 and b = 6.67. Site-specific estimates of N supply, using this approach, should recur under uniform N application since the major drivers of N supply, which includes organic matter, are relatively stable.
Retrospective likelihood of N limitation
In the northern grain region, certain combinations of yield, expressed on a relative scale, and protein can infer retrospectively whether the supplies of N were sufficient for maximum crop production. These inferences, observed in a large suite of N-related trials over several decades and regions, can apply to wheat (Table 1), barley, and sorghum.
Using the same trial data, the sum of trials where significant responses occurred, relative to the unfertilised control, to applied N can be plotted against the protein content of the grain in these controls. This provides a robust relationship to describe the likelihood of grain yield being limited given a specific protein content.
This analysis can act as an evaluation tool of the success of past N application strategies, and can provide ancillary information on where N-related on-farm trials should take place.
Table 1. Agronomic inferences for wheat in northern Australia for N and moisture supplies based on relative yield coincidental with protein (Strong and Holford 1997).
low supplies of one or more factors
Coincidental yield and protein data, in combination with in-crop rainfall data, can provide some information on the levels of available moisture present at sowing. Crop models, such as Wheatman (Woodruff 1992), estimate yield and protein for wheat and barley crops given a number of input parameters. These parameters include location, soil type, sowing date, crop/cultivar, moisture profile at sowing, and in-crop rain.
This model can be used to derive the initial levels of moisture available to the crop can be estimated, when the other parameters are known, for any given combination of yield and protein (D. Woodruff, pers. comm.). Site-specific information on the variation in starting soil moisture, particularly over a number of seasons, could prove useful for landscape assessment or soil capability studies. In addition, seasonal assessments of water use efficiency could be made that allow managers to evaluate their management practices.
Gross margin analysis
Payment of grain, for wheat and barley, is dependent on both yield and protein. The incentive for farm managers to produce grain of a certain protein quality is dependent on the additional payment secured by attaining that premium grade. Protein content is a useful indicator of whether the crop has achieved maximum yield potential, and a trade-off exists between payment for additional yield in comparison to payment for higher quality grain.
Site-specific variation in gross margin, calculated from the premiums paid for yield and high-quality grain, can indicate which areas in the field contribute most to farm profit. Options could include segregation of harvest loads to maximise the tonnage of premium grain, fencing off or restoring areas that consistently lead to losses, or undertaking on-farm studies to establish optimum fertiliser input rates.
Assessment of yield and protein allows for the proper interpretation of on-farm fertiliser strips. By contrasting yield/protein responses to a number of fertiliser rates, applications can be tailored for the field. With the unreliable in-crop rainfall in northern Australia, there will always be errors in targeting fertiliser requirements. Nevertheless, consistent use of spatial yield and protein collected during grain harvest should enable identification of areas where crops have exhausted all available N supplies, as well as areas where substantial available soil N may be carried over to the next crop.
By capturing coincidental yield and protein data on a site-specific basis, significant opportunity is provided to the farm manager to investigate and review their approach to N management, moisture usage, and economic crop inputs. Further analysis is required to construct statistical frameworks by which site-specific data, which may vary in sampling intensity, can be combined and interrogated. Frameworks could encompass a number of levels such as basic nutrient budgeting, probabilistic outputs that specific factors lead to yield limitations, or complex modelling that assists in our understanding of systematic changes in the system.
Methodologies that allow both data sets to be collected at grain harvest have the potential to be rapidly adopted. However, this will only occur as farm managers recognise the usefulness of these data, and are prepared to purchase suitable technology. Producers of mapping software and farm consultants should recognise the need to combine coincidental data sets, such as yield and protein, and follow a framework that allows farm managers to extract maximum value from cropping records.
We wish to thank Jamie Grant, Rob Taylor, Mike Smith, and Richard Prior for access to their fields. This work was made possible through the provision of funds by GRDC and QDPI, and in-kind support from Wesfarmers Landmark and RDS Technologies.
Cook, S.E., and Bramley, R.G.V. (2001). Is agronomy being left behind by precision agriculture? In ‘10th Australian Agronomy Conference, Hobart, 28 January – 1 February.’ (ASA: Hobart.)
Dalgliesh, N., and Foale, M. (1998). ‘Soil Matters’. (APSRU/CSIRO: Toowoomba.)
Engel, R.E., Long, D.S., and Carlson, G.R. (1997). On-the-go protein sensing is near – Does it have a future in precision nitrogen management for wheat? Better Crops with Plant Food 81(4), 20–3.
Minasny, B., McBratney, A.B., and Whelan, B.M. (1999). ‘VESPER version 1.0’. (Australian Centre for Precision Agriculture, The University of Sydney, NSW). URL: www.usyd.edu.au/su/agric/acpa.
Norng, S., Pettitt, A.N., Strong, W.M., and Butler, D. (2001). Strategies to interpret yield maps: Predicting grain protein using yield. In ‘10th Australian Agronomy Conference, Hobart, 28 January – 1 February.’ (ASA: Hobart.)
Strong, W.M., and Holford, I.C. (1997). Impact of land use practices on sustainability, fertilisers, and manures. In ‘Sustainable Crop Production in the Sub-Tropics: an Australian Perspective.’ (Eds. A.L. Clarke and P.J. Wylie) pp. 214-34 (QDPI: Brisbane.)
Williams, P. (2000). Case IH/Textron take sample presentation seriously! NIR News 11(3), 3–4.
Woodruff, D.R. (1992). ‘WHEATMAN’, a decision support system for wheat management in subtropical Australia. Australian Journal of Agricultural Research 43, 1483–99.