Nitrous Oxide Prediction from Farmland Using Fuzzy Logic

Abstract

Undoubtedly, agriculture is vital for sustaining humanity, but how often do we concern ourselves with the negative consequences of arable farming, particularly the amount of greenhouse gas emissions that result from it? Nitrous oxide (N₂O) is a greenhouse gas (GHG) emitted from farmland as a result of natural soil processes, including nitrification and denitrification. The amount of this potent GHG that is emitted into the atmosphere is exacerbated by the widespread application of nitrogen-rich fertilizers to crops using the conventional broadcast method.

It is alarming to know that from 2005 to 2019, fertilizer use in Canada increased by 71%, and this caused a 54% increase in N₂O emissions. The Government of Canada has decided to take action against climate change and set a goal to reduce N₂O levels by 2030. Note that this target does not mean there is a ban or mandatory reduction in fertilizer use. Their goal is to reduce fertilizer-related emissions without jeopardizing the success of maximizing food production, while improving the sustainability of farming. To accomplish it, there must be ways to monitor N₂O from farmland.

The work presented in this article is indirectly related to this goal, as I will explore the use of a fuzzy logic (FL) model to predict N₂O from relatively easily obtained inputs: daily average soil temperatures and moisture readings. This is not intended to replace traditional measurement methods, such as trace gas chambers, but to provide an alternative means of obtaining a rough estimate of N₂O emissions, since traditional methods are very costly. If this yields sufficiently accurate results, it can be deployed in real-world settings.

Between the years of 2021 and 2023, data was collected during the growing season on many different parameters, including soil temperature and moisture. These two variables are known to impact the flux of N₂O. In this research project, a FL model was developed, and daily average soil temperature and moisture data for each year were obtained and input into the model. The output was the predicted amount of N₂O emitted from the farmland. These predicted values were compared statistically with real, daily-averaged N₂O readings. The results indicate that this model is promising as a cost-effective technology for predicting N₂O emissions from farmland.