Browsing by Author "Grafton, MCE"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    The classification of hill country vegetation from hyperspectral imagery
    (Fertilizer & Lime Research Centre, 10/04/2017) Cushnahan, T; Yule, IJ; Grafton, MCE; Pullanagari, R; White, M; Currie, L; Hedley, M
    Remotely sensed hyperspectral data provides the possibility to categorise and quantify the farm landscape in great detail, supplementing local expert knowledge and adding confidence to decisions. This paper examines the novel use of hyperspectral aerial imagery to classify various components of the hill country farming landscape. As part of the Ravensdown / MPI PGP project, “Pioneering to Precision”, eight diverse farms, five in the North and three in the South Island were sampled using the AisaFENIX hyperspectral imager. The resulting images had a 1m spatial resolution (approx.) with 448 spectral bands from 380 – 2500 nm. The primary aim of the PGP project is to develop soil fertility maps from spectral information. Images were collected in tandem with ground sampling and timed to coincide with spring and autumn seasons. Additional classification of the pasture components of two farms are demonstrated using various data pre-processing and classification techniques to ascertain which combination would provide the best accuracy. Classification of pasture with Support Vector Machines (SVM) achieved 99.59% accuracy. Classification of additional landscape components on the same two farms is demonstrated. Components classified as non-pasture ground cover included; water, tracks/soil, Manuka, scrub, gum, poplar and other tree species. The techniques were successfully used to classify the components with high levels of accuracy. The ability to classify and quantify landscape components has numerous applications including; fertiliser and farm operational management, rural valuation, strategic farm management and planning.
  • Thumbnail Image
    Item
    Improving aerial topdressing in New Zealand through particle ballistics modelling and accuracy trials
    (Fertilizer & Lime Research Centre Massey University, 1/06/2016) Chok, SE; Grafton, MCE; Yule, IJ; Currie, L; Singh, R
    Fixed wing aircraft are utilised in New Zealand to apply dry bulk fertiliser on hill country farms. The fertiliser is most often applied manually as a blanket rate over the entire farm. Previous study indicates that this yields a field coefficient of variation (CV), which is the standard deviation over the mean application rate, of 63 – 70%. The CV decreased to 44% when the hopper door was automatically controlled using aircraft installed global positioning system (GPS) in lieu of manual intervention by the pilot. This is comparable to fertiliser application by fully GPS enabled truck spreaders. Spreadmark® specifies that the transverse overlap CV should be 15% for nitrogen-based fertilisers and 25% for all other products; however transverse overlap tested CV is considerably different to field CV. Variation in aerial topdressing is a barrier to achieving these CV. These variables include wind conditions, topography, aircraft speed and fertiliser properties. Ravensdown Limited is upgrading their topdressing aircraft fleet with differential rate application technology (DRAT), which uses the automated hopper door and GPS to apply various application rates over specified target areas within a farm. The advantage of this system is that fertiliser can be applied to these areas with the largest potential benefit in terms of increase pasture productivity and reduced environmental impact. Two trials utilising cone shaped collectors were carried out at coastal sheep and beef farms to determine the DRAT system’s accuracy when applying two application rates. Proof of release maps, which is deduced from aircraft recorded data, showed the system was able to vary rate. The CV ranged between 34% and 56%. The CV can be further improved by using a granular fertiliser ballistics model that predicts the transverse and longitudinal spread patterns based on wind conditions, fertiliser properties and aircraft operation. Validation data for this model was collected in validation trials for superphosphate, urea and di-ammonium phosphate. A validated model can provide guidelines on the optimum conditions and settings for aerial topdressing.
  • Thumbnail Image
    Item
    Integration of precision farming data and spatial statistical modelling to interpret field-scale maize grain yield variability in New Zealand
    (1/12/2019) Jiang, G; Grafton, MCE; Pearson, D; Bretherton, M; Holmes, A
    Spatial variability in soil, crop, and topographic features, combined with temporal variability in weather can result in variable annual yield patterns within a paddock. The complexity of interactions between these yield-limiting factors requires specialist statistical processing to be able to quantify spatial and temporal variability, and thus inform crop management practices. This paper evaluates the role of multivariate linear regression and a Cubist regression model to predict spatial variability of maize-grain yield at two sites in the Waikato Region, New Zealand. The variables considered were: crop reflectance data from satellite imagery (Sentinel 2 and Landsat 8), soil electrical conductivity (EC), soil organic matter (OM), elevation, rainfall, temperature, solar radiation, and seeding density. The datasets were split into training and validation sets, proportionally 75% and 25% respectively. Both models learn using 10-fold cross-validation. Statistical performance was evaluated by leaving out one year of yield data as the validation set for each iteration, with all remaining years included in the training set for building the prediction models. In the multiple-year analysis, the Cubist model (RMSE=1.47 and R2=0.82 for site 1; RMSE=2.13 and R2=0.72 for site 2) produced a better statistical prediction than the MLR model (RMSE=2.41 and R2=0.51 for site 1; RMSE=3.37 and R2=0.30 for site 2) for the prediction of the validation set. However, for the leave-one-year-out analyses, the MLR model provided better statistical predictions (RMSE=1.57 to 4.93; R2 = 0.15 to 0.31) than the Cubist model (RMSE = 2.62 to 5.9; R2 = 0.05 to 0.14) for Site 1. For Site 2, both models produced poor results. Yield data for additional years and inclusion of more independent variables (e.g. soil fertility and texture) may improve the models. This analysis demonstrates that there is potential to use statistical modelling of spatial and temporal data to assist farm management decisions (e.g. variable rate application, precision land levelling, irrigation, and drainage). Once the functional relationship between within-paddock yield potential and complementary variables is established, it should be possible to provide an accurate management prescription, enabling variable rates of an input (e.g. plant density, fertiliser) to be applied automatically across the paddock based on the “yield-input” response curve.