Application of Bayesian Networks for agricultural land suitability classification: a case study of biosolids amendment

Land suitability classification tools have been largely applied by environmental managers to support decision making. These tools are needed in the case of the management of biosolids on agricultural soil, to minimise environmental contamination and human exposure. To define the suitability of the agricultural areas is a complex task that involves integrating parameters that are incomparable and sometimes incommensurate, such as soil classification, crop type, etc. In addition, the lack of knowledge on the representation of natural systems is of concern and high uncertainty is associated to model development. In this study, we propose the application of Bayesian networks (BNs), a recognised modelling solution to deal with uncertain and complex problems, to classify the suitability of agricultural land to receive biosolids as an organic amendment. A case study of sewage sludge application in agricultural land in North-Eastern part of Spain (Lleida Catalonia) was used to describe model’s application. The developed Bayesian network represents causal relationships between the terrain characteristics and the identified issues related to the amending practice (environmental contamination and human exposure). The causal relationships were defined by local stakeholders and environmental experts in 3 workshops, where they were able to identify the main problems related to this practice and investigate the impacts according to different soil and landscape characteristics. As a final step, model outputs were represented in GIS, given the general trends for each agricultural area. In addition, uncertainty was represented in maps that give the probability of reaching each suitability class. These suitability maps represent an innovative way of evaluating the results of the land suitability classification.