Uncertainty in Great Barrier Reef Catchment soil nutrient data - implications for land use management

The reduction of particulate nutrient loads (nitrogen and phosphorus) has beenan important objective for managers of Great Barrier Reef Lagoon (GBRL) catchments.These loads are believed to be dominated by non-point sources, but end-of-catchment loadmeasurements provide little insight into the distribution and relative magnitude of pollutantsources within the catchment. This has led to a reliance on computer modelling to identifyprobable pollutant sources and to devise and compare management strategies to deal withthem. The most common modelling approach used in the GBRL region is to computespatially-distributed budgets of nutrient sources and sinks. The largest source is usuallyhillslope erosion which is calculated using a form of the revised universal soil loss equation(RUSLE) and then assigned nitrogen and phosophorus contents to the soil based on valuesheld in the Australian Soil Resource Information System (ASRIS). To this is added subsoilerosion from river banks and gullies where subsoil has traditionally been assumed to havespatially uniform phosphorus and nitrogen contents.We assessed the uncertainty in the most recently updated ASRIS GBRL soil nutrient dataand the implications of using these data for modelling and management of particulatenutrient loads at spatial scales ranging from individual farms to subcatchments with areas andgt;1000 sq. km. Bias in the assumed subsoil nutrient content suggests that particulate loads ofsubsoil phosphorus may be underestimated by 14% whereas subsoil nitrogen loads arelikely to be overestimated by a factor of three. Variability (expressed as std dev/mean) ofsurface soil nutrient content at a single sampling site was 34% for nitrogen and 21% forphosphorus. When considering variability within the smallest resolved spatial units (uniquemapping area, UMA) in ASRIS encompassing four or more sample sites the results were38% and 48% for N and P, respectively. This level of intrinsic variability in the observeddata in combination with the relatively large size of a UMA makes it very difficult todiscriminate differences in nutrient fluxes at scales much less than 25 sq. km withconfidence.