Irrigation water fitness assessment based on Bayesian network and FAO guidelines

The use of inappropriate irrigation water may lead to unwanted consequences for the soil, plants, irrigation system equipment and more. Ideally, the assessment of water fitness should include all influential factors and their interdependence. Because that is rarely the case, as an attempt to satisfy such requirements we propose a new approach based on the FAO guidelines and Bayesian network (BN) modelling. Fifteen factors recommended by the FAO were used to create a network to assess irrigation water fitness. The usability of the BN was tested on two water samples from two measuring points in Vojvodina Province, Serbia. The results obtained from the same sample are diverse and depend on the plants' tolerance to salt and ionic toxicity, and on soil texture. For the different conditions at the measuring point Hetin, the probability of no restriction on use varied from 50 to 84%. In the case of the Bačko Gradište measuring point, the probability varied from 66 to 91.5%, and the strongest fitness depends on the soil texture. Comparison of the obtained results with the FAO classification shows high correlation and points to flexibility and adaptability of the created network to change data or working with missing data.