In this dissertation, I explore the role that agricultural fire management plays in global patterns of vegetation fire and thus carbon cycling between ecosystems and the atmosphere. No estimates previously existed of the amount of fire associated with pasture and rangelands at the global scale, and so my first chapter details the development of a dataset separating the influences of cropland, pasture, and non-agricultural land on fire activity. Pasture turns out to be responsible for nearly half of all the area that burns every year, and often differs in frequency and seasonal timing from adjacent non-agricultural ecosystems.
These new observational data allowed me to design and parameterize a global fire model that, for the first time, explicitly simulates the way that modern land managers use fire on cropland and pasture. The information about non-agricultural fire also allowed me to construct the first model for burning on non-agricultural land based on observations with the influence of pasture burning – which is governed by wholly different processes – removed. Chapter 2 details the structure and performance of this new fire and vegetation model. After calibration against observations using a new, automated method, the model is shown to successfully reproduce the general global patterns of fire activity.
In my third chapter, I use the model to test whether pasture burning practices have any effect on terrestrial carbon cycling. Significantly more carbon emissions are associated with pasture fire than would be expected if left to burn according to the same mechanisms that govern non-agricultural fire. The mean difference is even larger than the mean annual modeled net land carbon flux. Although there is a fair amount of interannual variation, and there are some areas for improvement in the model’s simulation of both vegetation dynamics and fire activity, these results highlight the importance to Earth system modeling of including realistic representations of how people manage fire on agricultural lands.