Applying geostatistical simulation to compare measurement strategies for estimation of nitrous oxide emissions

by Katherine Sherril

Abstract
Emissions of nitrous oxide (N₂O), a greenhouse gas, have been increasing over the last two decades, and emissions from soils is a lead contributor. N₂O is produced during microbial processes of nitrification and denitrification, which can be greatly stimulated in agriculturally managed soils as a result of fertilization and other management decisions. Assessing emissions to better link them to processes, can inform adaptive management. Nonetheless, sources of N₂O emissions from soils are complex to measure given the high spatial and temporal variability of the process. Fluxes of N₂O have been observed to have neither normal distribution in space nor in time, but rather occur in episodic which can last sometimes less than a day or in small spatial hotspots. Assessing the stochasticity of the process can be an important step towards improving accuracy in sampling designs and models. In this research project, a geostatistical model was applied to simulate such temporal and spatial variability of emissions in order to inform sampling design decisions. The model includes a spatio-temporal residual, a trend for seasonal effects, and a trend to simulate patterns that are a result of topography and continuous soil properties. The model imposed a Space-Time semivariogram to simulate a residual with spatio-temporal correlation. Simulating the spatio-temporal residual component, a number of possible realities were generated for nitrous oxide emissions over a 1 hectare study area and 6-month growing season. From this simulated data, common sampling designs of varying spatial and temporal sampling patterns were compared to assess their accuracy. Appropriately capturing the spatio-temporal variability is challenging when upscaling measurements given the variability of the process as well as budget and time constraints of an experiment. Results indicate that increasing the number of points in space versus time results in higher accuracy.