#• \
\ d?
PRO*^
2020 National Emissions Inventory Technical
Support Document: Agriculture - Fertilizer
Application
-------
-------
EP A-454/R-23 -001 i
March 2023
2020 National Emissions Inventory Technical Support Document: Agriculture - Fertilizer
Application
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Research Triangle Park, NC
-------
Contents
List of Tables i
List of Figures i
9 Agriculture - Fertilizer Application 9-1
9.1 Sector Descriptions and Overview 9-1
9.2 Sources of data 9-1
9.3 EPA-developed estimates 9-1
9.3.1 Methodology overview 9-1
9.3.2 Activity data 9-3
9.3.3 Emission factors 9-5
9.3.4 Improvements/Changes in the 2020 NEI 9-6
9.4 References for agricultural fertilizer application 9-9
List of Tables
Table 9-1: SCCs in the Agricultural Fertilizer Application sector 9-1
Table 9-2: Agencies that submitted fertilizer application NH3 emissions in the 2020 NEI 9-1
Table 9-3: Environmental variables needed for an EPIC simulation 9-4
List of Figures
Figure 9-1: "Bidi" modeling system used to compute 2020 Fertilizer Application emissions 9-3
Figure 9-2: USDAfarm production regions used in FEST-C simulations 9-4
Figure 9-3: Simplified FEST-C system flow of operations in estimating NH3 emissions 9-6
Figure 9-4: Distribution of Ag Fertilizer NH3 emissions in the 2020 NEI 9-7
Figure 9-5: NH3 emissions difference between 2020 NEI and 2017 NEI, by county 9-8
Figure 9-6: Model evaluation of air quality based on updated ag fertilizer emissions 9-9
l
-------
9 Agriculture - Fertilizer Application
9.1 Sector Descriptions and Overview
Fertilizer in this category refers to any nitrogen-based compound, or mixture containing such a compound, that
is applied to land to improve plant fitness. The SCCs that compose this sector in the 2017 NEI are provided in
Table 9-1. The SCC level 1, 2 and 3 description is "Miscellaneous Area Sources; Agriculture Production - Crops;
Fertilizer Application" for both SCCs. EPA-estimated emissions are solely for SCC 2801700099, which comprise
the majority of the emissions for this sector and for which EPA methods are discussed further below. The first
SCC in Table 9-1 is included for completeness only, as EPA does not provide any estimates that are housed in
that SCC nor did SLT submit any data for that SCC in the 2020 NEI cycle. The only pollutant estimated by EPA for
this sector and expected to be reported by any SLT is ammonia (NH3).
Table 9-1: SCCs in the Agricultural Fertilizer Application sector
SCC
SCC Level 4 Description
EPA
S/L/T
2801700000
Total Fertilizers
2801700099
Miscellaneous Fertilizers
X
X
9.2 Sources of data
The agricultural fertilizer application sector includes data from the S/L/T agencies and the default EPA-generated
agricultural fertilizer emissions. The agencies listed in Table 9-2 submitted emissions for this sector; agencies not
listed used EPA estimates for the entire sector.
Table 9-2: Agencies that submitted fertilizer application NH3 emissions in the 2020 NEI
Region
Agency
S/L/T
3
Delaware Department of Natural Resources and Environmental Control
State
9
Maricopa County Air Quality Department
Local
10
Coeur d'Alene Tribe
Tribe
10
Kootenai Tribe of Idaho
Tribe
10
Nez Perce Tribe
Tribe
10
Shoshone-Bannock Tribes of the Fort Hall Reservation of Idaho
Tribe
9.3 EPA-developed estimates
9.3.1 Methodology overview
Direct flux measurements of ammonia (NH3) over agricultural fields and natural vegetation over the past few
decades have demonstrated that vegetation and soil can either be a source or a sink of atmospheric NH3. The
direction and magnitude of the exchange depends on the concentration gradient between the canopy and the
atmosphere. The bidirectional approach taken here accounts, in the most comprehensive way possible, for
estimated NH3 emissions from this complex process. The NH3 emissions estimated here are for fertilizer that has
been applied to the soil. Emissions from the application processes are also estimated in the manure
management portion of livestock emissions. Based on the methods used by EPA for estimating both sets of
these emissions, it is believed that there is no significant double counting across the two ag sectors. The
9-1
-------
approach to calculating emissions from this sector in 2020 is very consistent with the way it was done in the
2017 NEI, with some noted embellishments to the methods for the 2020 NEl, which have resulted in about a
60% increase in NH3 emissions nationwide compared to 2017 estimates. The bidirectional version of CMAQ
(v5.4) [ref 1] and the Fertilizer Emissions Scenario Tool for CMAQ FEST-C (vl.4) [ref 2, ref 3, ref 4] were used to
estimate ammonia (NH3) emissions from agricultural soils. These estimates were then loaded into EIS for use in
the 2020 NEI. The approach to estimate 2020 fertilizer emissions consists of these aggregate steps:
• Run FEST-C to produce nitrate (N03), Ammonium (NH4+, including Urea), and organic (manure) nitrogen
(N) fertilizer usage estimates
• CMAQ model with bidirectional ("bidi") NH3 exchange to generate gaseous ammonia NH3 emission
estimates.
• Calculate county-level emission factors as the ratio of bidirectional CMAQ NH3 fertilizer emissions to
FEST-C total N fertilizer application.
• Assign the NH3 emissions to one SCC: "...Miscellaneous Fertilizers" (2801700099).
We first estimate fertilizer application by crop type for 2020 using FEST-C modeled data. 2020 USDA Economic
Research Service crop specific fertilization data is not yet available, and we also did not receive state estimated
data. However, upgraded use of FEST-C vl.4 simulations resulted in better agreement with USDA ERS (Economic
Research Service) and U.N Food and Agriculture national fertilizer application estimates [ref 4] with an average
domain wide bias of -10.5% for 2002 to 2019 model simulations were observational data were available. Then
we ran the CMAQ v5.4 model with the Surface Tiled Aerosol and Gaseous Exchange (STAGE) deposition option
with bidirectional exchange to estimate fertilizer and biogenic NH3 emissions for 2020. We use this approach for
three reasons: (1) FEST-C estimates fertilizer applications based on crop nutrient needs which is typically lower
than real world fertilization rates; (2) FEST-C fertilizer timing and application methods are assumed to be
correct; and (3) If available, this CMAQ model option allows us to incorporate state-submitted and USDA
reported data into the final fertilization emission estimates.
FEST-C is the software program that processes land use and agricultural activity data to develop inputs for the
CMAQ model when run with bidirectional exchange. FEST-C reads land use data from the Biogenic Emissions
Landuse Dataset (BELD), meteorological variables from the Weather Research and Forecasting model [ref 5], and
nitrogen deposition data from a previous or historical average CMAQ simulation. FEST-C, then uses the USDA's
Environmental Policy Integrated Climate (EPIC) modeling system [ref 4] to simulate the agricultural practices and
soil biogeochemistry and provides information regarding fertilizer timing, composition, application method and
amount. Figure 9-1 provides a comprehensive flowchart of the complete EPIC/FEST-C/WRF "bidi" modeling
system.
9-2
-------
Figure 9-1: "Bidi" modeling system used to compute 2020 Fertilizer Application emissions
The Fertilizer Emission Scenario Tool for CMAQ
(FEST-C)
9.3.2 Activity data
The following activity parameters were input into the EPIC model:
• Grid cell meteorological variables from WRF
• Initial soil profiles/soil selection
• Presence of 21 major crops: irrigated and rain fed hay, alfalfa, grass, barley, beans, grain corn,
silage corn, cotton, oats, peanuts, potatoes, rice, rye, grain sorghum, silage sorghum, soybeans,
spring wheat, winter wheat, canola, and other crops (e.g., lettuce, tomatoes, etc.)
• Fertilizer sales to establish the type/composition of nutrients applied
• Management scenarios for the 10 USDA production regions shown in Figure 9-2 [ref 3, ref 4],
These include irrigation, tile drainage, intervals between forage harvest, fertilizer application
method (injected versus surface applied), and equipment commonly used in these production
regions.
9-3
-------
Figure 9-2: USDA farm production regions used in FEST-C simulations
We used the WRF meteorological model to provide grid cell meteorological parameters for 2020 using a national
12-km rectangular grid covering the continental U.S. The meteorological parameters in Table 9-3 below were
used as EPIC model inputs.
Table 9-3: Environmental variables needed for an EPIC simulation
EPIC input variable
Variable Source
Daily Total Radiation (MJ m2)
WRF
Daily Maximum 2~m Temperature (C)
WRF
Daily minimum 2-m temperature (C)
WRF
Daily Total Precipitation (mm)
WRF
Daily Average Relative Humidity (unitless)
WRF
Daily Average 10-m Wind Speed (m s"1)
WRF
Daily Total Wet Deposition Oxidized N (g/ha)
CMAQ
Daily Total Wet Deposition Reduced N (g/ha)
CMAQ
Daily Total Dry Deposition Oxidized N (g/ha)
CMAQ
Daily Total Dry Deposition Reduced N (g/ha)
CMAQ
Daily Total Wet Deposition Organic N (g/ha)
CMAQ
Initial soil nutrient and pH conditions in EPIC are based on the 1992 USDA Soil Conservation Service (CSC) Soils-5
survey. While this survey may seem outdated, it would not be expected that much change would occur over
time for these parameters as they have a typically fairly narrow range of pH requirements and soil pH that is
managed by the farmer. The EPIC model then is run for 25 years using current fertilization and agricultural
cropping techniques to estimate soil nutrient content and pH for the 2020 EPIC/WRF/CMAQ simulation.
9-4
-------
The presence of crops in each model grid cell was determined using USDA Census of Agriculture data (2006) and
USGS National Land Cover data (2011). These two data sources were used to compute the fraction of
agricultural land in a model grid cell and the mix of crops grown on that land.
Fertilizer sales data and the 6-month period in which they were sold were extracted from the 2014 Association
of American Plant Food Control Officials (AAPFCO). AAPFCO data are used to identify the composition (e.g.,
urea, nitrate, organic) of the fertilizer used, and the amount applied is estimated using the modeled crop
demand. These data are useful in making a reasonable assignment of what kind of fertilizer is being applied to
which crops.
Management activity data refers to data used to estimate representative crop management schemes. We used
the USDA Agricultural Resource Management Survey (ARMS) to provide management activity data. These data
cover 10 USDA production regions shown in Figure 9-2 and provide management schemes for irrigated and rain
fed hay, alfalfa, grass, barley, beans, grain corn, silage corn, cotton, oats, peanuts, potatoes, rice, rye, grain
sorghum, silage sorghum, soybeans, spring wheat, winter wheat, canola, and other crops (e.g., lettuce,
tomatoes, etc.).
The following variables were provided to stakeholders to enable review, comment, and potential improved
localized inputs on these data in the 2020 NEI process:
• Fertilizer application timing
• Plant/harvest dates
• Fertilizer application rates by crop and county
• Area planted
• Crop yields
9.3.3 Emission factors
The emission factors were derived from the 2020 CMAQ FEST-C outputs. Total fertilizer emission factors for each
month and county were computed by taking the ratio of total fertilizer NH3 emissions (short tons) to total
nitrogen fertilizer application (short tons).
12 km by 12 km gridded NH3 emissions were mapped to a county shape file polygon. The cell was assigned to a
county if the grid centroid fell within the county boundary. County-level fertilizer emissions (NH3) for 2020 are
subsequently derived from the diagnostic emission output from a 2017 CMAQ FEST-C model simulation [ref 9],
With this modeling system, it would be difficult to perform a sample calculation as part of this documentation.
These emissions are computed via the full chemical transport model, as illustrated in Figure 9-3.
9-5
-------
Figure 9-3: Simplified FEST-C system flow of operations in estimating NH3 emissions
Modeled
EPIC
Data
Ammonium
pH,and
application method
Modeled
WRF
Data
Soil Moisture
& Temperature
CMAQ
Modeled
Atmospheric
NH,
Deposition
9.3.4 Improvements/Changes in the 2020 NEI
The 2020 fertilizer emission estimates are based on the CMAQ FEST-C "bidirectional" approach outlined in
Figure 9-1 and Figure 9-3 that couples meteorological inputs, CMAQ, and the EPIC modeling system through the
FEST-C interface. The approach used for estimating ammonia emissions for the 2020 NEI is substantially the
same approach as that used for the 2017 NEI efforts, which is documented in the 2017 NEI TSD. However, the
2020 NEI used the latest model versions of CMAQv5.4 and FEST-Cvl.4. FEST-Cvl.4 largely corrected bugs in
FEST-Cvl.3 correcting the indexing of modeled beans and canola crop types. The estimates used FEST-Cvl.4
simulations with CMAQv5.4 using the land use specific deposition option, Surface Tiled Aerosol and Gaseous
Exchange (STAGE), and bidirectional NH3 exchange. In addition, the following substantial changes were made in
estimating NH3 emissions using this process in the 2020 NEI:
• Revised/improved treatment of the biogenic non-agricultural emissions based on recent observations
from ORD measurement programs [ref 6, ref 7],
• The CMAQ look up table of plant functional type, e.g., evergreen needleleaf forests, emission factors
were updated in CMAQv5.4 based on soil and vegetation surveys at some AMoN sites and soil and
vegetation and soil NH4+ observations in the global TRY plant trait database (https://www.try-db.org)
• These updates resulted in significantly higher emission factors for soils for all land cover types except
evergreen needle leaf forest and increases in the vegetation emission factors for grasslands. These data
were used to update the emission factors for no-agriculture land use in CMAQ that were previously
populated based soil measurements from a North Carolina pine forest and vegetation NH4+ estimated
from annual deposition fields from CMAQv5.0 simulations and the empirical relationship in Massad et
al. [ref 8] as documented in Bash et al. [ref 9, ref 10],
• Moving from MODIS to 2011 NLCD datasets for national coverage of agricultural lands resulted too in
contributing to an overall increase in to 2020 emissions
• The national coverage of Agriculture between NLCD and MODIS is similar, but MODIS estimates much
larger coverage for the upper midwest, where we see the decrease in NH3 emissions and NLCD has
much higher agriculture coverage KS, OK, North TX, and the southeast.
9-6
-------
Due to all these changes/ improvements made to the emissions estimation process in the 2020 cycle,
nationwide EPA NH3 estimates for this sector are about 60% higher than in 2017. In the 2017 NEI, about 1.1
million tons of NH3 were estimated in total from this source, in the 2020 NEI that number increases to about 1.8
million tons. Please note that in the 2017 NEI for this sector, emissions were reported incorrectly as "tons of N"
instead of NH3, making the correction for converting "tons of N" to NH3 results in the 1.1 million tons number
referenced above. Figure 9-4 shows how these emissions are distributed by county (note that AK and HI are
blank as EPA methods do not cover that domain and we did not receive any submissions from those states), and
Figure 9-5 shows the difference county by county of emissions in 2020 vs emissions in 2017. It can be seen from
Figure 9-5 that most counties showed an increased level of emissions in 2020 compared to 2017. By state, the
biggest increases in moving to 2020 for states with higher absolute amounts of NH3 emissions include FL, GA,
MO, SC, and AL. States that decreased in emissions in going to 2017 from 2020 were few, but included WA, MN,
ID, and AZ.
Figure 9-4: Distribution of Ag Fertilizer NH3 emissions in the 2020 NEI
Legend
I I State Boundaries
Sum NH3
Emissions (TON)
1 Min: 0.835663 - ISO
I I 150 • 1,000
1 1,000 - 5,000
5,000 - 10,000
10,000-Max:
11,874.11 -ot
I 10 or No Value
Alaska
Hawaii
9-7
-------
Figure 9-5: NH3 emissions difference between 2020 NEI and 2017 NEl, by county
Hawaii
Alaska
l l State Boundaries
2020-2017 NEI NH3
Emissions Difference
(TON)
¦1 -H,850 - -3.000
{¦ -2,999 - -500
r 1-499-0
i 1 1 - 500
¦¦ 501 - 3,000
¦¦ 3,001 • 7,962
0 or No Value
Legend
To check the validity of these increases in ammonia emissions going from 2017 to 2020, a CMAQ. model
simulation for the year 2018 was used to evaluate these natural ammonia emission potential increases (via new
measured emissions factors as discussed previously) and these simulations resulted in a lower model bias at 84%
of AmON NH3 monitoring sites (see Figure 9-6 which shows modeled concentrations of NH3 compared to
monitored values of NH3) and reduced the summertime model NH3 bias from -29% to -17% (still an
underestimate but better). In Figure 9-6, the cooler colors indicate a reduction in model bias at those locations.
9-8
-------
Figure 9-6: Model evaluation of air quality based on updated ag fertilizer emissions
CMAQv54_12US1_2C18_Base_STAGE_EM - CMAQv533J2llS1_2Q18_Base_STAGE NH3 bias difference lor 20180101 to 20181231
There is no point source subtraction that is needed for this sector. Additional Information regarding the general
methods employed here can be found in the Air Emissions Inventory Training site, search for "Key Ammonia
sectors" training materials.
9.4 References for agricultural fertilizer application
1. Community Multiscale Air Quality (CMAQ v5.4) model.
2. Benish, S.E., Bash, J.O., Foley, K.M., Appel, K.W., Hogrefe, C., Gilliam, R., Pouliot, G., Long-term regional
trends of nitrogen and sulfur deposition in the United States from 2002 to 2017, Atmospheric Chemistry
and Physics, v22, I 19, 12479-12767, 2002, https://doi.org/10.5194/acp-22-12749-2022
3. Foley, K.M., Pouliot, G.A., Eyth, A., Aldridge, M.F., Allen, C., Appel, K.W., Bash, J.O., Beardsley, M.,
Beidler, J., Choi, D., Farkas, C., Gilliam, R., Godfrey, J., Henderson, B.H., Hogrefe, C., Koplitz, S.N., Mason,
R., Mathur, R., Misenis, C., Possiel, N., Adams, E.: 2002-2017 anthropogenic data for air quality modeling
over the United States, ScienceDirect, v47, April 2023, 109022,
https://doi.Org/10.1016/i.dib.2023.109022
4. Fertilizer Emission Scenario Tool for CMAQ (FEST-C) system.
5. Weather Research Forecast (WRF) model.
6. Rumsey, I.C., Walker, J.T.: Application of an online ion-chromatography-based instrument for gradient
flux measurements of speciated nitrogen and sulfur, Atmos. Meas. Tech. 9, 2581-2592,
doi:10.5194/amt-9-2581-2016, 2016
7. Walker, J. T., Chen, X., Wu, Z., Schwede, D., Daly, R., Djurkovic, A., Oishi, A. C., Edgerton, E., Bash, J.,
Knoepp, J., Puchalski, M., liames, J., and Miniat, C. F.: Atmospheric Deposition of Reactive Nitrogen to a
Deciduous Forest in the Southern Appalachian Mountains, Biogeosciences, https://doi.org/10.5194/bg-
2022-133, in press, 2023
8. R.-S. Massad, E. Nemitz, and M. A. Sutton, Review and parameterisation of bi-directional ammonia
exchange between vegetation and the atmosphere Atmos. Chem. Phys., 10, 10359-10386, 2010.
9-9
-------
9. J. O. Bash, E. J. Cooter, R. L. Dennis, J. T. Walker, and J. E. Pleim, Evaluation of a regional air-quality
model with bidirectional NH3 exchange coupled to an agroecosystem model, Biogeosciences, 10, 1635-
1645, 2013
10. Cooter, E.J., Bash, J.O., Benson V., Ran, L.-M.; Linking agricultural crop management and air-quality
models for regional to national-scale nitrogen deposition assessments, Biogeosciences, 9, 4023-4035,
2012.
9-10
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-454/R-23-001i
Environmental Protection Air Quality Assessment Division March 2023
Agency Research Triangle Park, NC
------- |