Combining satellite data with a simulation model to describe spatial variability in pasture growth at a farm scale

Practical application of simulation modelling as a decision aid for grazing system management usually involves an assumption of uniformity of model in puts over a farm paddock or property. In reality, paddocks and farms displa y high spatial variability in model inputs. There is considerable interest in assessing the significance of this spatial variablity for anmal producti on and enterprise profitability. This study seeks to demonstrate the use of spatial data with the GRAZPLAN pasture model to provide estimates of annua l net primary production from pastures at a farm scale on the Northern Tabl elands of New South Wales, Australia. The GRAZPLAN pasture model was valida ted against data from 2 separate field experiments for a typical improved p asture based on Phalaris aquatica from 1968 to 1972. A spatial coverage, cl assifying paddocks into 9 pasture types based on a botanical survey, was us ed to define the pasture parameter sets used in simulations. A Landsat TM s atellite image classified to give 3 pasture growth status classes was used to define within-paddock levels of a fertility index used in the simulation model. Simulations over 1975-94 were conducted for all combinations of pas ture types and fertility scalar values using climate data for the CSIRO Pas toral Research Laboratory near Armidale. Simulation output was written to a lookup table and imported into a PC-based geographic information system. T he spatial data layers were combined to form a display template representin g spatial variation in pasture type, pasture condition and fertility. The s patial template was reclassified using the lookup tables to create maps of annual net primary production from pastures. Spatial variability in simulat ed annual net primary production was greater for the paddocks with diverse mixtures of sown and native species than for the more uniform highly improv ed or pure native pastures. The difference in response to rainfall of simul ated net primary production was greater between different pastures types th an between different levels of the fertility index. The resulting maps prov ide a demonstration of the way in which satellite imagery and other data ca n be interfaced with a decision support system to provide information for u se in precision management of grazing systems. Implementation of such metho ds as a management tool will depend on development of quantitative spatial data layers which provide accurate and repeatable initial conditions and pa rameter values for simulation models.