• nitrous oxide;
  • nitric oxide;
  • outliers;
  • uncertainty;
  • watershed


To develop an advanced method for estimating nitrous oxide (N2O) emission from an agricultural watershed, we used a closed-chamber technique to measure seasonal N2O and nitric oxide (NO) fluxes in cornfields, grassland, pastures and forests at the Shizunai Experimental Livestock Farm (467 ha) in southern Hokkaido, Japan. From 2000 to 2004, N2O and NO fluxes ranged from –137 to 8,920 µg N m−2 h−1 and from –12.1 to 185 µg N m−2 h−1, respectively. Most N2O/NO ratios calculated on the basis of these N2O and NO fluxes ranged between 1 and 100, and the log-normal N2O/NO ratio was positively correlated with the log-normal N2O fluxes (r2 = 0.346, P < 0.01). These high N2O fluxes, therefore, resulted from increased denitrification activity. Annual N2O emission rates ranged from –1.0 to 81 kg N ha−1 year−1 (average = 6.6 kg N ha−1). As these emission values varied greatly and included extremely high values, we divided them into two groups: normal values (i.e. values lower than the overall average) and high values (i.e. values higher than average). The normal data were significantly positively correlated with N input (r2 = 0.61, P < 0.01) and the “higher” data from ungrazed fields were significantly positively correlated with N surplus (r2 = 0.96, P < 0.05). The calculated probability that a high N2O flux would occur was weakly and positively correlated with precipitation from May to August. This probability can be used to represent annual variation in N2O emission rates and to reduce the uncertainty in N2O estimation.