Process-based models are increasingly used to study mass and energy fluxes from agro-ecosystems, including nitrous oxide (N2O) emissions from agricultural fields. This data set is the output of three process-based models – DayCent, DNDC, and EPIC – which were used to simulate fluxes of N2O from dairy farm soils. The individual models’ output and the ensemble mean output were evaluated against field observations from two agricultural research stations in Arlington, WI and Marshfield, WI. These sites utilize cropping systems and nitrogen fertilizer management strategies common to Midwest dairy farms.
The models were calibrated and validated using data collected at Arlington and Marshfield over five years (nine years for crop yield). Calibration and validation used observations of soil temperature (n = 887), volumetric soil water content (VSWC, n = 880), crop yield (n = 67), and soil N2O flux (n = 896). The observed data are presented here with the model output to document model calibration and validation; most of these observed data are also held by Ag Data Commons in separate data sets from field experiments at Arlington and Marshfield (http://dx.doi.org/10.15482/USDA.ADC/1361194, http://dx.doi.org/10.15482/USDA.ADC/1401975, http://dx.doi.org/10.15482/USDA.ADC/1399470). The remaining observed data is described in Osterholz et al. 2014.
Model simulations were run from 2010-2015 for the Arlington site and 2013-2015 for the Marshfield site. The three models were parameterized (i.e. calibrated) for each site using the same climate, initial soil physical and chemical conditions, hydraulic properties, initial soil carbon, and management schedules. Weather data for each site (daily minimum and maximum temperature, precipitation, relative humidity, wind speed, and solar radiation) was reconstructed using the NOAA online climate database (NOAA, 2016). Initial soil physical and chemical properties were constructed from available on-site measurements and supplemented using the Web Soil Survey (Soil Survey Staff, 2016). Soil carbon data was available for each site, and to prioritize model agreement initial soil carbon for the 0-20cm layer was set at 55.7 Mg C ha-1 for Arlington (Sanford et al., 2012), and at 52.6 Mg C ha-1 for Marshfield. Following parameterization of soil C, a 17 year spin-up period (1993-2009) at each site was simulated prior to the years during which data was collected (2010-2015). While DayCent developers typically recommend a spin-up of at least 1000 years, DNDC has been run with spin-up periods as low as 2 years (Zhang et al., 2015). Given that observations of soil C were available, a 17 year spin-up was chosen to reflect the duration between initial soil C sampling (Sanford et al., 2012) and the first measurement of N2O in our data set (Osterholz et al., 2014). Management and input schedules were constructed from on-site data and record-keeping; these are available in the supplementary online data of the primary journal paper. All other initial parameters, such as crop-specific productivity or soil carbon turnover rate, were independently established by each model in calibration.
This work was part of “Climate Change Mitigation and Adaptation in Dairy Production Systems of the Great Lakes Region,” also known as the Dairy Coordinated Agricultural Project (Dairy CAP), funded by the United States Department of Agriculture – National Institute of Food and Agriculture (award number 2013-68002-20525). The main goal of the Dairy CAP was to improve understanding of the magnitudes and controlling factors over greenhouse gas (GHG) emissions from dairy production in the Great Lakes region. Using this knowledge, the Dairy CAP has improved life cycle analysis (LCA) of GHG production by Great Lakes dairy farms, developing farm management tools, and conducting extension, education and outreach activities.