Some interesting observations have emerged from research at Trangie to assess the viability of precision agriculture (PA) for the Western Plains of NSW.

Low rainfall in the last two years has limited yield and variability in yield at the project’s sites, making it impossible to draw many general conclusions; however, one issue of interest relates to the use of remotely sensed spectral indices.

Many PA practitioners use the Normalised Difference Vegetation Index (NDVI), obtained from air-borne or satellite imagery at about anthesis, to predict yield.

NDVI indicates the amount of biomass present and the vigour of the plants.

Researchers in Western Australia associated with the project have found NDVI to be a reliable indicator of eventual yield variability but there has been some debate about its reliability in the eastern States where rainfall and temperatures att he end of the growing season are more variable.

Trangie-based research by a team from NSW DPI, AGnVET Services, Sustainable Soils Management and Charles Sturt University in 2004 justifies caution in using image-based spectral indices as substitutes for a yield map.

Figure 1 shows a map of a Red-Blue spectral index from an image of a canola crop acquired on September 21, 2004; Figure 2 shows the eventual yield map of the same crop.

Figure 1: Flowering intensity of 2004 canola crop indicated by Red-Blue spectral index.

Flowering intensity is usually a good indicator of yield of canola: areas of high flowering intensity generally yield well. The Red-Blue index appears to be a better indicator of flowering intensity in canola than NDVI.

Two things to note are: the obvious difference between the index in the north-east part of the paddock and in the rest of the paddock in Figure 1, and the fact that some areas of high flowering intensity - blue in Figure 1, have subsequently yielded poorly - red in Figure 2.

The difference in flowering intensity between the north-east corner and the rest of the paddock at the time of image capture resulted from a three-day interruption to sowing caused by rain.

Figure 2: Canola yield in 2004.

The failure of flowering intensity to translate into yield is particularly noticeable in the small block at the north-west corner which went from highest flowering intensity to lowest yield.

This happened because some areas where the plants had grown most strongly before flowering ranout of moisture and the plants were unable to turn their potential into yield; while other areas where the plants had grown less - there-fore using less moisture up to flowering - still had sufficient moisture to allow the plants to produce grain and higher yields.

This is an extreme case but the evidence of the sowing interruption’s effect on flowering shows that imagery can be a powerful and useful tool. Work on the links between spectral indices and yield is ongoing with the aim of determining the circumstances in which imagery can be used to indicate yield.

Examples such as this, where an explanation of the outcome from ground observation is available, will lead to greater skill in image interpretation.

The project is based at Trangie Agricultural Research Centre and is part of GRDC’s Strategic Initiative Program on Precision Agriculture (SIP09). To date, the project has monitored winter cropping on two sites during 2004 and 2005. One site is located between Trangie and Tottenham on predominately grey clay soil and the other is just west of Nyngan on redsoil.