Soil carbon and nitrogen stocks in sugarcane systems by Bayesian conditional autoregressive model – an unbiased prediction strategy
Abbruzzini, T.F.
Soil carbon and nitrogen stocks in sugarcane systems by Bayesian conditional autoregressive model – an unbiased prediction strategy - United Kingdom : Taylor and Francis 2017.
Peer review
Spatially dependent data are predominant in soil science and prone to biased inferences from standard statistical analysis. Thus, the aims of this study were: to model the spatial dependency among soil sampling points using a Bayesian conditional autoregressive (CAR) prior; and to determine the effects of different sugarcane management systems on soil C and N stocks. Four sugarcane sites were evaluated: conventional burned (BSC); unburned (USC); and organic sugarcane for 4 years (O04) and 12 years (O12). A native vegetation forest (NVF) site was used as a reference. The CAR model prediction agreed with the observed results of both soil C and N stocks. The highest predicted soil C and N stocks at 0-30 cm depth were observed for O12 (57.3 and 4.8 Mg ha-1), and the lowest were for BSC (37.6 and 3.0 Mg ha-1). The Bayesian CAR model captured the spatial dependence among soil sampling points and allowed to compare soil C and N stocks of different sugarcane managements. Thus, Bayesian spatial modeling is a novel approach to evaluate soil management practices when performing ad hoc monitoring of soil carbon within contiguous areal units.
Text in English
https://doi.org/10.1080/17583004.2017.1309204
Bayesian theory
Soil
Carbon
Soil carbon and nitrogen stocks in sugarcane systems by Bayesian conditional autoregressive model – an unbiased prediction strategy - United Kingdom : Taylor and Francis 2017.
Peer review
Spatially dependent data are predominant in soil science and prone to biased inferences from standard statistical analysis. Thus, the aims of this study were: to model the spatial dependency among soil sampling points using a Bayesian conditional autoregressive (CAR) prior; and to determine the effects of different sugarcane management systems on soil C and N stocks. Four sugarcane sites were evaluated: conventional burned (BSC); unburned (USC); and organic sugarcane for 4 years (O04) and 12 years (O12). A native vegetation forest (NVF) site was used as a reference. The CAR model prediction agreed with the observed results of both soil C and N stocks. The highest predicted soil C and N stocks at 0-30 cm depth were observed for O12 (57.3 and 4.8 Mg ha-1), and the lowest were for BSC (37.6 and 3.0 Mg ha-1). The Bayesian CAR model captured the spatial dependence among soil sampling points and allowed to compare soil C and N stocks of different sugarcane managements. Thus, Bayesian spatial modeling is a novel approach to evaluate soil management practices when performing ad hoc monitoring of soil carbon within contiguous areal units.
Text in English
https://doi.org/10.1080/17583004.2017.1309204
Bayesian theory
Soil
Carbon