University of California, Berkeley Berkeley, CA, United States
Soil nitrous oxide (N2O) fluxes are often characterized by hot spots and hot moments that contribute disproportionately to annual N2O budgets but are difficult to predict without continuous, high-frequency measurements. It is understood that the production and consumption of N2O are likely controlled by availability of carbon (C), nitrogen (N), or oxygen (O2) substrates and additional interactions with temperature, moisture, and plant activity. However, poor data resolution leads to an incomplete understanding of how these drivers alter the magnitude and occurrence of hot moments of soil N2O flux. Our experimental approach using long-term continuous, automated chamber measurements in agricultural peatland soils found significant annual differences in the magnitude of N2O emissions associated with scale-emergent relationships with moisture, temperature, and O2 variability.
Here we defined hot moments as flux greater than four standard deviations from the mean, as statistically 99.9% of the population should fall within four standard deviations. We found that these short-term N2O hot moments accounted for up to 76% (range: 31%-76%) of annual emissions across agricultural ecosystems. This also highlighted that non-hot moment background emissions can also represent a significant portion of total N2O budgets. To determine drivers of N2O hot moment and background emissions, we used wavelet coherence to determine that seasonal and diel trends. We found that rainfall and irrigation were the primary drivers of hot moments of N2O emissions (p < 0.05). With satellite-derived photosynthetic activity, we also found that plant activity and associated C or N substrate availability was an important driver of background emissions (p < 0.05). Importantly, high-resolution datasets can be used to train machine learning models. Using only continuous moisture, O2, and temperature sensor data from three soil depths (10, 30, and 50 cm), our machine learning model predicted over 80% of the variability in N2O emissions in multiple agroecosystems. These combined approaches help better quantify ecosystem budgets of N2O fluxes and have inspired our current research developing new environmental sensing technologies alongside machine learning models to quantify both the spatiotemporal distribution of N2O hot moments and better predict the magnitude of emissions.
