Concepts in modelling N2O emissions from land use

Abstract

Modelling nitrous oxide (N2O) emissions from soil is challenging because multiple biological processes are involved that each respond differently to various environmental and soil factors. Soil water content, organic carbon, temperature and pH are often used in models that predict N2O emissions, yet for each of these factors there are concepts that are not fully understood. Though a ubiquitous measure of soil water for models, the application of functions based on water filled pore space across soils that vary in bulk density is not ideal. Diffusion of gases and solutes in soil are controlled by the volume fractions of air and water present. Across soils with different bulk densities, both of these terms vary at constant water filled pore space. Soil organic carbon influences N2O emissions in two ways: as a source of energy for denitrifiers and also by driving biological oxygen demand and the creation of anaerobic zones in the soil. Soil temperature influences N2O emissions through its effect on the activity of microorganisms and enzymes. A variety of temperature response functions have been proposed. The preferred response function should contain a temperature optimum that can be varied in response to climatic conditions to account for microbial adaptation. Soil pH can have direct and indirect influences on rates and product ratios of nitrification and denitrification. The concepts of pH optima and microbial adaptation need to be considered in modelling. Methodological issues such as microsite versus bulk soil measurements and apportioning N2O fluxes to the various N transformation processes remain an impediment to characterising the influence of pH and other factors on N2O emissions. Quantifying the response of N2O emissions to individual factors using regression analysis requires all other factors to be controlled experimentally. Boundary line analysis provides a way of defining the response to a single input variable where other influencing variables are not controlled. Such analyses can aid in the definition of the shape and magnitude of response functions to be incorporated into process simulation models. Process/mechanistic simulation models offer a greater transferability than empirical models but careful consideration of temporal and spatial scale and the availability of data to run these models is critical in developing model structure.