Kruskal–Wallis ANOVAs showed significant differences between ranked estimations obtained from sampling schedules for all individual years (2006–2007: P < 0.001; 2007/2008: P < 0.001; 2008/2009: P < 0.001) and over three years (P < 0.001). The accuracy of estimation of annual cumulative N2O emissions was generally improved with increasing number of samplings per year ( Fig. 5). Over BYL-719 three years, outcomes from the weekly and fortnightly sampling schedules were significantly poorer (P < 0.05) than all other sampling schedules. Increasing the sampling frequency to biweekly or triweekly after rainfall substantially reduced the errors in the estimated annual cumulative N2O emissions. For example, increasing sampling in a axons weekly schedule to two or three times per week following > 20 mm daily rainfall increased the estimation accuracy over the three years by 8% and 9%, respectively. Estimations, over the three years, based on fortnightly samplings were improved by 18% and 20%, when biweekly and triweekly samplings, respectively, were applied after daily rainfall of > 20 mm. Of the rainfall-based sampling schedules, weekly + triweekly and fortnightly + triweekly consistently delivered estimates with errors <± 7% in all years, while the weekly + biweekly and fortnightly + biweekly schedules appeared less satisfactory in 2007–2008 with the errors reaching ± 10%.