Abstract
Preliminary spatial predictive modelling of 0-30 cm Soil Organic Carbon (SOC) stocks has been undertaken using legacy site data, across a 1 million hectare study area (>10 million cells) in the Mid North region of South Australia. SOC stock estimation required the development of a bulk density pedotransfer function, incorporation of key soil parameters (available at the horizon level), and application of depth-smoothing splines to evaluate stocks over a consistent depth range. Data analysis and spatial modelling has been performed within the R software environment, including use of the raster package to apply predictive models block by block (in order to overcome inherent memory limitations). A preliminary combined linear and regression tree model has performed well at training sites but lacks predictive ability at independently collected validation sites. Key limitations in the legacy site data have been identified which will guide future potential improvements in mapping of SOC stocks.
Original language | English |
---|---|
Title of host publication | GlobalSoilMap |
Subtitle of host publication | Basis of the Global Spatial Soil Information System |
Editors | Dominique Arrouays, Neil McKenzie, John Hempel, Anne C. Richer de Forges, Alex McBratney |
Place of Publication | Boca Raton, FL |
Publisher | CRC Press |
Pages | 253-259 |
Number of pages | 7 |
ISBN (Electronic) | 9781315775586, 9780429227417 |
ISBN (Print) | 9781138001190 |
DOIs | |
Publication status | Published - 2014 |
Externally published | Yes |
Event | 1st GlobalSoilMap Conference: Basis of the Global Spatial Soil Information System - Orleans, France Duration: 7 Oct 2013 → 9 Oct 2013 |
Conference
Conference | 1st GlobalSoilMap Conference: Basis of the Global Spatial Soil Information System |
---|---|
Country/Territory | France |
City | Orleans |
Period | 7/10/13 → 9/10/13 |
Keywords
- GIS
- Soil Science
- Soil Sciences
- Remote sensing and cartography
- Geology - Earth scinences
- Digital soil mapping