Auditing on-farm soil carbon stocks using downscaled national mapping products: Examples from Australia and New Zealand

A method has been tested to audit soil carbon stocks at the farm-scale. This is needed for emission trading schemes and for tracking sustainability of the soil resource through time. The method begins with baseline maps derived from national models which are then disaggregated with fine scaled environmental data via statistical modeling; before optimally stratifying to guide soil sampling positions. Field sampling provides a statistically valid estimate of soil carbon stocks, and the method is repeated through time to monitor any changes in stocks. Case studies are presented from Australia and New Zealand. While spatial downscaling is useful for generating soil maps relevant to the farm scale, the optimal stratification of these maps for guiding soil sampling for baseline soil carbon auditing purposes should not be recommended if the national scale mapping is thought to be unreliable. Because of unresolvable differences between spatial scales associated with bias and incorrect specification of uncertainties, results from Australia revealed stratified simple random sampling was not as efficient when compared to the less costly simple random sampling. Conversely, results from New Zealand do show that stratified simple random sampling to be more efficient than simple random sampling. From soil sampling, clear differences in carbon stocks to 30 cm were observed when comparing the stocks measured at sites from both countries. In New Zealand, soil carbon stocks were estimated to be as high as 101 t/ha for the top 30 cm of soil. For the Australian sites which were all situated in the Hunter Valley region of NSW, the highest measured soil carbon stocks were 25 t/ha for the top 30 cm. The largest soil carbon variance was observed at the New Zealand hill country farm, where the landscape consists of alluvial terraces, complex broken valley sides and a summit plateau mantled with volcanic ash. In Australia, the presence of subsoil pedogenic carbonate (marl) contributed to high variance estimates relative to the other sites.