Ammonia (NH3) volatilization is the major source of nitrogen (N) loss resulting from the application of synthetic and organic N fertilizers to croplands. It is well known that in Mediterranean cropping systems, there is a relationship between the intrinsic characteristics of the climate and nitrous oxide (N2O) emissions, but whether the same relation exists for NH3 emissions remains uncertain. Here, we estimated the impact of edaphoclimatic conditions (including meteorological conditions after N fertilization), crop management factors, and the measurement technique on both the cumulative emissions and the NH3 emission factor (EF) in Mediterranean climate zones, drawing on a database of 234 field treatments. We used a machine learning method, random forest (RF), to predict volatilization and ranked variables based on their importance in the prediction. Random forest had a good predictive power for the NH3 EF and cumulative emissions, with an R2 of 0.69 and 0.76, respectively. Nitrogen fertilization rate (N rate) was the top-ranked predictor variable, increasing NH3 emissions substantially when N rate was higher than 170 kg N ha-1. Soil pH was the most important edaphoclimatic variable, showing greater emissions (36.7 kg NH3 ha-1, EF = 19.3%) when pH was above 8.2. Crop type, fertilizer type, and N application method also affected NH3 emission patterns, while water management, mean precipitation, and soil texture were ranked low by the model. Our results show that intrinsic Mediterranean characteristics had only an indirect effect on NH3 emissions. For instance, relatively low N fertilization rates result in small NH3 emissions in rainfed areas, which occupy a very significant surface of Mediterranean agricultural land. Overall, N fertilization management is a key driver in reducing NH3 emissions, but additional field factors should be studied in future research to establish more robust abatement strategies.
Keywords: Ammonia emissions; Crop management; Mediterranean cropping systems; Meteorological conditions; Nitrogen use efficiency; Random forest.
Copyright © 2023. Published by Elsevier Ltd.