Maryem-Arshad
Maryem Arshad
Role: 
PhD Candidate
Contact details:
Office: 

Room 570
Biological Sciences North (D26)
UNSW, Kensington 2052

Mapping Soil Clay Using Bayesian Modelling and Proximal Sensors at the Field

Owing to its substantial role in maintaining soil quality and soil health, knowledge of the distribution of the clay particle size fraction at the field level is important. Contrarily, acquiring clay content data is time-consuming and expensive at the field level. Digital soil mapping (DSM) offers robust and efficient approach to value-add to limited topsoil (0-0.3 m), subsurface (0.3-0.6 m) and subsoil (0.9-1.2 m) clay data. To this end, proximally sensed ancillary data was collected from three sources, including; a digital elevation model, a gamma-ray (g-ray) spectrometer (RS700) which provided radioelement (e.g. Th – Thorium and TC – Total Count) data and an electromagnetic (DUALEM-421) instrument which provided soil electrical conductivity (e.g. 1mPcon and 4mPcon) data. To understand the uncertainty in the DSM, a Bayesian inference approach called Integrated Nested Laplace Approximation with Stochastic Partial Differential Equation (INLA-SPDE) was used. Model accuracy (RMSE), bias (ME) and concordance (Lin’s) was also compared for different combinations of ancillary data sources. We concluded that the INLA-SPDE approach could provide estimations of the posterior marginal distributions of the model parameters as well as the model responses. Using more than one ancillary data, in various combinations, proved better at predicting clay.

      

Spatial distributions of predicted clay (%) by INLA-SPDE for a) topsoil (0–0.3 m) b) subsurface (0.3-0.6 m) and c) subsoil (0.9-1.2 m)

Spatial distributions of credibility interval of predicted clay (%) by INLA-SPDE for a) topsoil (0–0.3 m) b) subsurface (0.3-0.6 m) and c) subsoil (0.9-1.2 m)