New Zealand has about four million hectares in medium to steep hill country pasture to which granular solid fertiliser is applied by airplane.  On most New Zealand hill country properties where cultivation is not possible the only means of influencing pasture production yield is through the addition of fertilizers and paddock subdivision to control grazing and pasture growth rates. Pasture response to fertilizer varies in production zones within the farm which can be modelled using decision mining trees.

New Zealand pastoral farmers target production of a blend of perennial rye grass (Lolium perenne_cv ) and white clover (Trifolium repens _cv ). Concentrations of these desired species reduce as pasture fertility reduces and topography becomes steeper, being replaced by wild grasses and weed species. As a result there are often significant differences in pasture quality, expressed as Crude Protein, (CP) Metabolisable Energy, (ME) and in-vitro Organic Matter Digestibility, (OMD). Traditionally these have not been measured as the process is extremely time consuming, expensive and there is often a long delay between sampling and the farmer receiving the results, which is inconvenient for decision making. This results in a lack of information around pasture productivity and differences in pasture quality not being fully recognised.  

These differences in quality have been measured in the field using a hyperspectral sensor  (ASD FieldSpec Pro), which allowed the full range of  pasture quality within one large station to be observed and measured in-situ. In one farm case study completed, the sites examined had a range of pasture quality results, ME (7.2 to 11 MJ kg^-1), CP (7.8 to 22%) and OMD (45 to 75%) which would significantly impact animal production. Results were compared to wet chemistry in order to calibrate and validate the methods. The results indicate that this method can give useful results with a high level of explanation of the data, providing a reliable means to determine pasture quality parameters from non-destructive field measurement.  

Pasture production is considerably effected by variations in slope, aspect, seasonal rainfall and soil type, factors which in some cases do not change and other variables which do change but are outside the control of the farmer. In this variable environment the traditional method of soil testing a monitor paddock, assuming it represents the whole farm and planning a blanket application of fertiliser makes little sense as hill country produces variable production from different zones reflected in the decision mining tree. Many of the factors affecting productivity can be represented in a GIS, which allows production zones and fertiliser response to be mapped once the farm plan is overlaid over an accurate digital terrain map.

In order to gain this benefit from such a decision mining tree the fertiliser application system must be accurate in terms of achieving the desired application rate from an aircraft. A computer controlled delivery system fitted to the topdressing aircraft has been developed which has significantly improved the coefficient of variation c.v. of spread. Trials have a reduction in the in-field coefficient of variation (CV) in spread from around 70% to about 40%, which is the level achieved on flat dairy pasture by spreading trucks using differential global positioning system (DGPS).  This has enabled more complex fertiliser application plans to be carried out.