Precision agriculture - common sense with a dash of technology

Figure 1: a) elevation map, b) electromagnetic induction readings, c) zoning using three seasons of yield maps, and d) zoning using cluster analysis with three seasons of yield map plus elevation and EM map

Growers know which parts of their property give the best yields. Technology helps refine that knowledge, write Peter Fisher and Mohammad Abuzar.

Paddock zoning is a process that growers should approach without trepidation; after all, many are already practising it, though they might just call it common sense.

Farmers are constantly tailoring management operations according to the properties of different paddocks, what the current and future conditions might hold, and what risks might be involved. Most growers would already know that they need to manage particular areas of some paddocks differently. This is the basis of precision agriculture.

So what is new about PA? By integrating agronomy knowledge with recent advances in technology, PA provides a more scientific process for identifying and managing areas on the farm that might perform differently.

Managing individual paddocks differently might be considered the coarsest form of zone management. The other extreme is where constantly variable technology (CVT) is used to manage every location to match its conditions as they vary throughout the paddock.

This is also not a new concept to growers. Tractor draft control, which is probably the first example of CVT, was developed by harry Ferguson back in 1926. This mechanism senses the force on ground-engaging implements and adjusts the working depth "on-the-fly". Today, other CVTs are available such as variable-rate spray nozzles and fertiliser applicators.

In general, CVT is not limited by the application technology, but rather by our ability to accurately sense and locate the variability. In most cases, our agronomic knowledge of the crop responses to spatial variability is also not accurate enough to merit CVT More usually, a paddock is divided into a small number of sub-paddock zones in which different management can be confidently applied.

There are three main approaches to how sub-paddock zones can be delineated.

The first is based on a single factor of interest, such as soil sodicity or weed prevalence. For such factors there are clear remedial actions, such as applying gypsum or herbicide.

The second approach is to zone the paddock according to the yield produced - high-, medium- and low-yielding areas. Providing these zones are stable over time, growers know with confidence that each zone has a different return, so they can explore ways of maximising the gross margin for each zone. The criteria to identify each zone can be modified to account for the economic threshold of different management options, including making the area of each of these zones big enough for practical use.

The third approach is a combination of the first two, and uses a mixture of yield and other site properties, such as elevation and electromagnetic measurement (EM) maps. The approach uses a statistical methodology known as "cluster analysis" to identify areas within the paddock that are similar to each other in terms of a range of site properties included in the analysis.

However, the thresholds for different zones cannot be easily manipulated using cluster analysis techniques. Once the zones have been identified, further site investigation is usually required to understand why the zones are different and what changes in management practice might be required.

The impact of selecting these different zoning techniques is being studied by the Victorian Department of Primary industries (DPI) on a paddock in the southern Mallee region of the state, as part of the GRDC"s national PA Initiative. The first two images in Figure 1 show the individual site characteristics of (a) elevation and (b) EM induction readings, a measure of soil conductivity.

DPI has also been mapping the gammaradiometric emission profile of the paddock, because we are interested in whether this can provide a rapid estimate of the soil"s plant available water capacity (PAWC).

The third image is the average yield from three recent wheat crops (including the drought season of 2004), zoned into high, medium, and low yield. The fourth image is the paddock zoned using cluster analysis in which the same three seasons of yield data plus the elevation and EM (38-vertical mode) data have been used.

Figure 1: a) elevation map, b) electromagnetic induction readings, c) zoning using three seasons of yield maps, and d) zoning using cluster analysis with three seasons of yield map plus elevation and EM map.
[Note: Elevation in m; yield in t/ha; and EM shows electrical conductivity readings]

Although there are many similarities between the different zoning methods (after all, they both use the same three years of yield data), there are also some marked differences. A big difference in this example is that in the yield-only analysis each zone has a distinctly different mean yield, whereas in the cluster analysis it can be seen that the red and green zones have similar average yields.

The three zoning techniques described earlier require the grower to already have collected various amounts of information about their paddocks. This information can cost growers a significant amount of time and money in buying a yield monitor and for consultants to carry out site mapping, as well as in cleaning, storing and analysing the data sets.

It can also take several years for growers to build up a significant set of yield maps for each crop across different seasons.

An increasing number of growers are happy to make this investment, having confidence that the variability in their paddocks is sufficiently large that they need to be actively learning how to manage it. However, many other growers would like to know more about the yield variability of their paddocks before making any capital investment. DPI has been investigating the best way of providing this information to growers.

The methodology it has developed is based on estimating the variability in crop yield from historical satellite digital images. These are not a direct measure of yield but are used to estimate the crop biomass near anthesis (flowering), which in turn is used to estimate yield variability.

The advantage of this system is that these images are relatively cheap, especially if a group of growers in an area can share the cost. The prototype system being tested requires the grower to provide the rotation history for each paddock and the average paddock yields.

Analysis of the satellite data provides the grower with a map of the estimated differences in yield found within each paddock. However, paddock areas may perform differently in different seasons, so the analysis also tells the grower if the zones tend to be consistently high or low yielding, or tend to vary between high and low yields in different seasons, known as "flip-flopping".

The main advantage of the proposed system is that it not only provides growers with high, medium, and low-yielding zones, but enables these zones to be quantified in terms of tonnes per hectare.

Growers can use this information to compare whether the high or low-yielding areas in one paddock are higher or lower yielding than zones in another paddock. Not only can growers then prioritise areas for treatment within each paddock, but they can also develop a whole-farm planning approach to sub-paddock zoning.

For example, in a season with a late break growers can quickly prioritise which areas within specific paddocks are likely to make the greatest profit and concentrate their sowing in those areas. Growers can therefore estimate the economic benefits of making greater investment in precision agriculture by evaluating the yield variability across their complete farm.

GRDC Precision Agriculture Initiative (SIP09)
GRDC Research Code: DAV00030

For more information: Dr Peter Fisher, 03 5833 5222,