Understanding and analysing spatial variability of nitrous oxide emissions from a grazed pasture

Nitrous oxide (N2O) emissions exhibit a high degree of spatial variability. Within a grazed pasture the uneven deposition of urine patches is one of the major sources of spatial variability in soil attributes and environmental conditions. Understanding the spatial variability of N2O emissions is necessary to estimate the size of the sampling errors using a given number of static chambers. In this study we measured N20 emissions for three weeks following a grazing event using 100 chambers. These chambers were divided into 4 blocks of 25 chambers arranged in a5m x 5m grid so the within-block and between block variability could be compared. A known amount of urine was applied to 20% of the chambers. The behaviour of the sample mean using different numbers of chambers was investigated by randomly sub-sampling the chamber measurements from the 80 chambers to which urine was not applied. A cluster analysis based on final soil NH4(+), NO3- and cumulative N2O emissions correctly identified 90% of the known urine patches. As expected the presence or absence of urine patches was one source of differences in emissions. However, not all plots responded equally to urine application and the difference in response seemed to be regulated by soil moisture. The low emitting plots corresponded to very high moisture contents resulting in low nitrification rates supported by low measured soil NO3-, but high NH4+ in these chambers. However, a high degree of variability within a treatment and plot has not been accounted for in this study. The distribution of the sample mean remained skewed even up to 40 chambers per sample. However, the standard deviation and the 95% confidence interval reduced with increasing numbers of chambers. For random sampling with 16 chambers there was a 95% chance that the sample mean would be within -62% to 109% of the true mean. Stratifying the sampling, either by plot, urine patch/non-urine patch, or both factors combined reduced the skewness and narrowed the 95% confidence interval of the mean. The greatest improvement was seen with the combined plot/urine stratification. Stratifying by urine/nonurine patch produced a greater improvement than stratifying by plot. Therefore if it is possible to identify either urine patches or regions of high N2O production potential accurately then sampling errors could be reduced using a stratified sampling scheme. However, care would be needed to ensure that systematic errors were not introduced. (C) 2014 Elsevier B.V. All rights reserved.