Abstract for presentation at The 13th Australasian Remote Sensing and Photogrammetry Conference

Relative purchase costs, more certainty of supply and greater opportunity for cloud-free acquisitions from satellites such as MODIS, have necessitated reviewing existing routine programs based on Landsat data. An operational monitoring program for environmental impacts, spanning 8 years, has mapped annual distribution of summer cropping in the northern inland catchments of New South Wales. This has resulted in a substantial digital baseline of cropping paddocks, built into a geodatabase, While there are a few new cropping paddocks each year, most of the arable land has now been delineated from the medium resolution Landsat imagery over the monitoring period. Crop rotation practices, climatic variation, and economic factors result in only a percentage of paddocks being cropped in any one year. The feasibility of substituting MODIS (or similar) data, with distinct cost and certainty of supply advantages was investigated. Using an a priori approach, the established paddock GIS database was utilized to analyse the MODIS data. Centroids of the largely rectilinear paddock polygons were used to sample and subset corresponding pixels in the MODIS data. These centroid pixels were considered to have the highest probability of reflecting land cover within the paddock. The method sought to eliminate the mixels resulting from the poor spatial scale of the imagery and minimise the confusion contributed from neighbouring land covers. An NDVI was derived for the single “pure” paddock pixels and these were subsequently classified into crop or no crop consequent to field calibration and statistical analysis. The paper presents the results and supports the substitution or concessions afforded by cheaper lower resolution imagery where a priori knowledge environment exists.