Yearly mean flux values were then calculated for only hot moments, the entire flux dataset, and the flux dataset with not moments removed to determine the impact of very high flux events on annual GHG emissions. The term “outlier” is often used to connote values requiring removal or transformation within a dataset to maintain statistical power and limit overinflated estimates from high leverage observations . However, systematic elimination or data transformation ignore or underweight important processes such as hot moments of GHG flux . Given our large and continuous dataset, we could also compare mean fluxes with and without hot moments to better quantify the importance of hot moments. We further explored the importance of capturing hot moments by also recalculating mean N2O and CH4 flux after excluding fluxes greater than one, two, and three standard deviations from the mean. A modified statistical jackknifing technique was used to explore the response of mean N2O and CH4 flux estimates to changes in sampling interval by repeatedly sampling the dataset at 1-, 2-, 7- , 14-, and 28-day intervals . Two sets of soil sensors were installed from September 2018-July 2020 at depths of 10 cm, 30 cm, and 50 cm. Combination SO-110 Oxygen and thermistor temperature sensors and CS616 moisture sensors were connected to CR1000 dataloggers that stored data at 15 min intervals. A period of sensor removal occurred in May and June 2019 as multiple agricultural events, including tillage, planting, and discing prevented continuous installation. Sensors were also removed for 3 weeks in September-October 2019 for crop harvest and discing and for 2 weeks the following spring before planting, April to May 2020. Erroneous data corresponding to sensor malfunction were removed from the dataset,raspberry cultivation pot which include 0.6% of soil moisture measurements and 0.05% of soil O2 and temperature measurements.

Power loss also contributed to data loss, with a total of 58 days of missing data from agricultural activity or power loss during the sensor measurement period . To explore the potential distribution of GHG production across the soil profile, two replicate soil gas samples for CO2, CH4, and N2O were also taken in parallel with the soil sensors above at 10 cm, 30 cm, and 50 cm depths weekly during unflooded periods from September 2018 through November 2018, and April through December 2019. Instrument grade stainless steel 1/8” tubing was installed in parallel to the soil sensors above, with approximately 15 cm of tubing installed with multiple sampling holes parallel to the soil surface. Sampling septa were installed in 1/8” Swagelok unions permanently connected to the stainless-steel tubing. Septa were changed monthly. Two gas samples were collected with 30 ml BD syringes, discarding the first sample to clear the dead volume in the sampling line. Sampling lines were removed from the field in May and June 2019 for tillage, planting, and discing. The 30 ml gas samples were stored in over-pressurized 20 mL glass vials with thick septa until manual sample injection analysis on a Shimadzu GC-34 . Differences in soil gas concentrations, O2, moisture, mineral N, and pH across time periods were tested with one-way analysis of variance . Growing season time periods were classified as planting date to harvest date, preceded and followed by fallow periods. Unflooded periods were defined as soil moisture less than 50% at 10 cm depth. For CH4 fluxes, anaerobic periods were defined as any period of time where daily 10 cm O2 concentrations were equal to 0. Linear regressions were used to explore relationships between soil atmosphere GHG concentrations and net soil GHG fluxes . We conducted a hypothetical upscaling exercise to estimate the potential impact of agricultural maize peat land emissions in the region.

We multiplied our annual GWP values with areal values of 40,000 ha for agricultural maize with similar management practices on peat land soils in the Rindge soil series within the Sacramento-San Joaquin Delta, California, USA . Similar management throughout the region includes conventional maize agricultural practices and winter flooding of fallow maize fields to limit weed growth and provide habitat for migrating waterfowl .Significant CH4 fluxes were only observed 60 days into an extended period of anoxic conditions lasting a total of 124 days. This period of anoxic conditions was associated with complete soil saturation following winter flooding and corresponded to decreased soil O2 concentrations across depths . Short periods of elevated NH4 + concentrations observed during flooding were also associated with decreases in CH4 production . Shorter periods of sustained anoxic conditions in 2019-2020 did not produce hot moments of CH4 fluxes. Wavelet coherence analysis of CH4 fluxes suggested that soil moisture, soil temperature, and bulk soil O2 concentrations drove patterns in net CH4 fluxes at a daily time scale . Only soil O2 concentrations across soil depths had significant coherence with CH4 fluxes on a weekly timescale , with no significant coherence at longer timescales. Seasonality explained the high intra-annual variation observed in CO2 fluxes. Higher soil respiration rates occurred during the growing season and following harvest . Fluxes were significantly lower when soils were saturated . There was significant coherence with moisture, temperature, and O2 concentrations across depths at the daily scale . At weekly and seasonal scales, temperature and O2 concentrations displayed significant coherence with soil CO2 fluxes. We compared chamber fluxes with ecosystem respiration measurements conducted via eddy covariance in parallel at this field site . Similar values were observed for soil CO2 chamber fluxes and Reco eddy-covariance measurements across the study period . Soil CH4 chamber fluxes were lower than the eddy-covariance CH4 fluxes , although eddy covariance also captured similar hot moments of CH4 emission .We further explored the importance of missing potential hot moment fluxes by calculating the change in mean N2O and CH4 fluxes after removing observations greater than one, two, or three standard deviations from the overall mean flux. Removing all observations more than one standard deviation of the mean underestimated annual N2O fluxes by 56.6%, while removing observations greater than two and three standard deviations underestimated N2O fluxes by 42.7% and 34.5%, respectively.

Missing N2O fluxes greater than three standard deviations corresponded to an underestimation of annual N2O emissions up to 14.3 ± 0.6 kg N-N2O ha-1 yr-1. For CH4, removing observations greater than one standard deviation underestimated annual CH4 fluxes by 79%, while removing observations greater than two and three standard deviations underestimated CH4 fluxes by 69% and 63%, respectively. Finally, we calculated the minimum number of randomized flux measurements needed to calculate annual and total flux values for N2O and CH4 with a 95% confidence interval and margins of error of 10%, 25%, and 50% when the occurrence of hot moments are unknown . For N2O, an average of 8,342 individual flux measurements were needed to accurately calculate the annual mean flux within a 10% margin of error. This represents up to 35% of the dataset. Increasing the margin of error to 25% and 50% reduced the number of measurements needed, with a range of 475 to 1,904 and 121 to 507 individual randomized measurements per year, respectively. When analyzing the total N2O dataset, the minimum number of flux measurements needed was 6,401 with a 10% of margin error,low round pots decreasing to 1,108 and 281 for margins of error of 25% and 50%, respectively. The minimum sample size needed for calculating annual and total mean CH4 fluxes were greater than N2O . For annual CH4 fluxes, the minimum sample size needed to recalculate the mean flux within a 10% margin of error was at least 17,133 . Increasing the margin of error to 25% and 50% reduced the minimum annual sample sizes needed to at least 7,562 and 2,525 , respectively. The minimum number of flux measurements needed for the total CH4 dataset was also higher than N2O with 68,137, 54,419, and 31,656 for margins of error of 10%, 25%, and 50%, respectively. We conducted an upscaling exercise to provide a first approximation of the potential impact of peat land maize agriculture on regional GHG emissions. Using the three years of field data, we upscaled these flux measurements using the total regional land area with similar soil series and management practices. We calculated a mean annual GWP of 1.86 Tg CO2e y-1 for agricultural peat lands in the region, with N2O emissions representing 0.48 Tg CO2e y-1 . Assuming the field estimates measured here are representative of local management practices, N2O fluxes alone could represent 26% of agricultural maize peat land CO2e emissions in this region, a significantly higher percentage than previous estimates . Soil types with similar organic matter content represent over 40,000 ha of agricultural peat lands in the Sacramento-San Joaquin Delta region and these soils are dominated by maize production. They are often flooded in the winter for waterfowl habitat . The agricultural peat land soils in this study were extreme N2O emitters, with mean rates that were 4-27 times greater than other non-peat cropland N2O emissions .

It is notable that these values for both peat land and non-peat land ecosystems were largely derived from non-continuous data that may not capture all N2O emission hot moments. The three year average N2O emissions were greater than the highest IPCC estimates for temperate organic cropland soils, and the peak annual N2O emissions from this study were five times greater than the average values of 8 kg N2O-N ha-1 y-1 . Estimated mean annual N2O emissions of 16.8 ± 14.8 kg N2O-N ha-1 y-1 have been reported for other drained peat lands with data derived from bulk densities . Average N2O emissions observed in this study were similar to or higher than studies of N2O emissions from agricultural peat lands in the Sacramento-Delta, which ranged from 6.6 ± 3.8 using model estimates to 24 ± 13 kg N2O-N ha-1 y-1 using shorter-term periodic manual static chamber measurements . Surprisingly winter flooding, not fertilization, was the dominant driver of N2O emissions.The high NO3 – measured shortly after flooding likely accumulated under oxic, well-drained soil conditions as a result of N mineralization following crop harvest , and may have been supplemented by iron coupled anaerobic ammonium oxidation in these iron and C-rich soils . Urea-ammonium-nitrate fertilizer was applied once per year during planting. This inorganic N fertilizer application also contributed to a short-term increase in N2O emissions, although this was not the dominant source of annual N2O emissions. Denitrification was likely the main pathway of N2O during hot moments of N2O flux given elevated NO3 – concentrations observed immediately prior to peak emissions, as well as the observed increases in soil moisture and decreases in soil O2 and NO3 – concentrations during the N2O hot moments. The NO3 – was likely consumed during denitrification, with significant amounts of N2O released as a byproduct of incomplete denitrification in these N-rich soils . The strong correlations observed between daily mean N2O fluxes and soil atmosphere N2O concentrations also suggest that significant N2O production was occurring at depth and thus production throughout the profile likely contributed to the large fluxes observed.The large continuous data set allowed us to explore the importance of hot moments of N2O and CH4 emission in total ecosystem GHG budgets. While hot moments represented only 0.63-1.50% and 0.06-0.76% of annual N2O and CH4 flux measurements, respectively, they contributed up to 76% of total N2O emissions and 486% of total CH4 emissions. This corresponded to N2O hot moment emissions alone contributing up to 18% of the annual GWP of these agricultural peat lands. This highlights that missing hot moments may lead to significant underestimates of total ecosystem GHG budgets. We also explored the effects of sampling interval on N2O and CH4 flux quantification. Our results further highlighted the necessity of continuous measurements to accurately estimate total ecosystem N2O and CH4 fluxes. Even weekly sampling intervals may underestimate annual N2O fluxes by up to 20%, a significant fraction of total GWP, even from these high emitting agricultural peat lands. While continuous automated chamber or eddy covariance measurements are ideal to capture hot moments of emissions, long-term continuous measurements are still cost prohibitive in many locations and ecosystems. If hot moments are predictable and well defined, daily flux measurements are likely effective in appropriately quantifying hot moments of N2O emissions .