Decision Trees for the prediction of environmental and agronomic effects of the use of Compost of Sewage Slugde (CSS)

Abstract The use of biosolids for soil improvement and for the reduction of inorganic fertilization costs has been a common practice in recent decades and is being used more and more often as inorganic fertilization cost increases. This practice is useful because it can be effective for the recovery of low fertility soils and to recycle urban and industrial waste, but it can also have negative effects. Some components of biosolids, like heavy metals, can have a potential hazard on human or animal health if they reach the edible part of the plant. In addition, there are no mathematical models able to predict both, the increases in yield and quality of the crops by the input of nutrients or the hazards due to the heavy metals incorporated into the soil. Data mining allows the creation of predictive models and in addition, some techniques, such as Decision trees, allow the generation of fast and interpretable models. The dataset used in this study come from an experimental field located in Burgos (Central-Northern Spain), in which, additions of Compost of Sewage Sludge (CSS) have been applied alternatively for 6 years. The main consequence of this application in terms of nutrients is the increase in available phosphorus mainly with the addition of the highest doses of CSS. There have been found correlations between the heavy metal contents in soil and in plant tissues described by Decision trees that are refuted by numerous publications using other classical statistical methodologies. Decision trees make it easy to interpret predictions, opening a new path for interpreting complex datasets commonly present in mid- or long-term agronomic experiences.