Ecological Modelling 316 (2015) 14–27 Contents lists available at ScienceDirect Ecological Modelling j ourna l h omepa ge: www.elsevier.com/locate/ecolmodel Original article Hierarchical Bayesian calibration of nitrous oxide (N 2 O) and nitrogen monoxide (NO) flux module of an agro-ecosystem model: ECOSSE Xi Li a,∗ , Jagadeesh Yeluripati b , Edward O. Jones c , Yoshitaka Uchida d , Ryusuke Hatano a a Laboratory of Soil Science, Graduate school of Agriculture, Hokkaido University, Kita 9 Nishi 9, Kita-ku, Sapporo 060-8589, Japan b The James Hutton Institute, Craigiebuckler, Aberdeen AB15 8QH, Scotland, UK c Institute of Biological and Environmental Sciences, School of Biological Sciences, University of Aberdeen, 23St Machar Drive, Aberdeen, UK d Laboratory of Environmental Biogeochemistry, Research Faculty of Agriculture, Hokkaido University, Kita 9 Nishi 9, Kita-ku, Sapporo 060-8589, Japan a r t i c l e i n f o Article history: Received 11 April 2015 Received in revised form 16 July 2015 Accepted 21 July 2015 Keywords: Hierarchical Bayesian Uncertainty Variability Nitrous oxide Agro-ecosystem a b s t r a c t With the position of nitrous oxide (N 2 O) being the greenhouse gas with the highest global warming potential and its long atmospheric lifetime, the anthropogenic production of N 2 O is of major concern. The process-based model, ECOSSE, was partly developed to quantify emissions of greenhouse gases with an input data requirement that is readily available at regional scale. Hierarchical Bayesian (HB) meth- ods are potentially used to reduce the uncertainty and to explain the spatial variability of estimated parameters. Here, we used a Hierarchical Bayesian method to calibrate the parameters of the N 2 O and nitrogen monoxide (NO) sub-model of ECOSSE and to quantify the uncertainty of model simulations and to investigate the model extrapolation using soil information. The sub model simulated N 2 O emis- sion from nitrification and denitrification, while the simulated NO from nitrification. The HB calibration reduced the uncertainty in the N 2 O and NO simulations. The model’s root mean square error (RMSE) was decreased by 18% and 29% for N 2 O and NO across field sites compared to an uncalibrated model. Param- eters for nitrification could be considered universal, while parameters for denitrification challenged the assumption that these parameters may be considered universal constant values across sites. Parameters of the NO module could be considered constant for model extrapolation to regional scale. The calibrated parameters derived from soil-specific calibration could be served as default values for the N 2 O module extrapolation for similar soil types. Otherwise, the mean value of posterior distribution of calibration parameters in multi-dataset could be served as the parameter for model up scaling at regional scale. © 2015 Elsevier B.V. All rights reserved. 1. Introduction Emission of nitrous oxide (N 2 O), which has a global warming potential 298 times greater than carbon dioxide (CO 2 ) over a 100- year period and an atmospheric lifetime of approximately 120 years (Abdalla et al., 2010), is a major contributor to climate change. Soils are the main source of nitrous oxide (N 2 O) emission via the micro- bial processes of nitrification and denitrification, and they are also the source of nitric monoxide (NO) emissions via the microbial pro- cesses of nitrification (Wrage et al., 2001). Approximately 55–65% total global N 2 O emissions originate from agriculture (Mosier et al., 1998; Olivier et al., 1998), with emissions currently increasing at a rate of 0.2–0.3% per year (Kroeze, 1994), due to greater use of nitrogen (N) fertilizers and agricultural intensification which have ∗ Corresponding author. Tel.: +81 08032361320; fax: +81 117062494. E-mail addresses: icy124@hotmail.com, lixi124@chem.agr.hokudai.ac.jp (X. Li). enhanced these two processes (Abdalla et al., 2010; Babu et al., 2006). There are two general approaches to estimating direct N 2 O and NO fluxes from soils in agro-ecosystems. In an empirical model, N 2 O and NO fluxes are proportional to some easily quantifiable factors, and the model ignores complex processes and environmental fac- tors in N cycle (IPCC, 2006; Lokupitiya and Paustian, 2006). The process-oriented ecosystem models attempt to simulate many or all of the complex components of N cycle; therefore, process-based models have a unique potential to predict N 2 O and NO emissions from arable soils (Butterbach-Bahl et al., 2004; Gabrielle et al., 2006a). However, model testing is often limited by a lack of field data to which the simulations can be compared (Desjardins et al., 2010). The input data requirement for process-based models is also often very high, which leads to the limited use in regional or national predictions from agricultural land (Beheydt et al., 2007; Bell et al., 2012). The ECOSSE model is designed and produced to simulate soil carbon (C) and N stocks and gaseous emissions from soils using only data which is available at national or regional scale http://dx.doi.org/10.1016/j.ecolmodel.2015.07.020 0304-3800/© 2015 Elsevier B.V. All rights reserved.