Managing pasture nutrient requirements on large hill country sheep and beef properties based on information from soil sampling is expensive because of the time and labor involved. High levels of error are also expected as these properties are often greatly variable and it is therefore extremely difficult to sample intensively enough to capture this variation. Extensive sampling was also not considered viable as there was no effective means of spreading fertilizer with a variable rate capability over this terrain. A large project was commissioned which looked to replace soil sampling in areas of mixed pasture with remotely sensed information from a hyperspectral imaging system to inform fertilizer application. One of the major objectives of the project was to establish whether hyperspectral imaging could map pasture nutrient concentration levels from aerial surveys. Canopy reflectance data was measured using a high resolution airborne visible-to-shortwave infrared (Vis-SWIR) imaging spectrometer measuring in the wavelength region 380 to 2500 nm to predict nutrient concentrations. The main nutrients of interest were nitrogen (N) phosphorus (P), potassium (K), sulfur (S).

Nutrient prediction models were developed using a number of regression, such as Support Vector Regression, methods which utilized calibrating ground based surveys. The level of explanation for the various nutrients is stated. High levels of explanation were achieved and the best training models were used to extrapolate the models to the whole farm. The methodology has been used over 8 geographical areas of New Zealand and in different seasons to maximize the levels of variation observed and avoid model over fitting. The project is working closely with Ravensdown Fertiliser Co-operative Limited who has funded the project in partnership with the New Zealand Ministry for Primary Industry. The project will allow the company to develop a much more precise and efficient nutrient planning and application service. The nutrient application will be completed through a fleet of topdressing aircraft and vehicles capable of variable rate application. The spatial maps demonstrate large variations in pasture nutrient content and other pasture quality parameters which are not normally measured and therefore not taken into account by farmers in their pasture management. Information from the project presents a significant opportunity to improve pasture management and resulting animal production.