Uncertainty assessment of soil organic carbon content spatial distribution using geostatistical stochastic simulation

Citation metadata

Date: Feb. 2010
From: Australian Journal of Soil Research(Vol. 48, Issue 1)
Publisher: CSIRO Publishing
Document Type: Report
Length: 4,947 words

Document controls

Main content

Abstract :

Soil organic carbon (SOC) affects many processes in soil. The main objective of this study was the prediction and uncertainty assessment of the spatial patterns of SOC through stochastic simulation using 2 simulation algorithms, sequential Gaussian simulation (sGs) and sequential indicator simulation (sis). The dataset consisted of 158 point measurements of surface SOC taken from an 18-ha field in Lower Austria. Conditional stochastic simulation algorithms were used to generate 100 maps of equiprobable spatial distribution for SOC. In general the simulated maps represented spatial distribution of SOC more realistically than the kriged map, i.e. overcoming the smoothing effect of kriging. Unlike sGs, sis was able to preserve the connectivity of extreme values in generated maps. The SOC simulated maps generated through sGs reproduced the sample statistics well. The reproduction of class-specific patterns of spatial continuity of SOC for the simulated model produced through sis was also reasonably good. The results highlight that when the class-specific patterns of spatial continuity of the attribute must be preserved, sis is preferred to sGs. For local uncertainty, standard deviations obtained using kriging varied much less across the study area than those obtained using simulations. This shows that the conditional standard deviations achieved through simulations depend on data values in addition to data configuration for greater reliability in reporting the estimation precision. Further, according to accuracy plots and goodness statistic, G, sis performs the modelling uncertainty better than sGs. The simulated models can provide useful information in risk assessment of SOC management in Lower Austria. Additional keywords: prediction, kriging, sequential Gaussian simulation, sequential indicator simulation.

Source Citation

Source Citation   

Gale Document Number: GALE|A222313804