NMIP

NMIP-2: global Nitrogen/N₂O Model Inter-comparison Project phase 2 Protocol

NMIP (The global Nitrogen/N₂O Model Inter-comparison Project) is an international partnership of model-model and data-model intercomparison under the umbrella of the Global Carbon Project (GCP) and International Nitrogen Initiative (INI).

1. Goals and Objectives

The first phase of the Global N₂O model Inter-comparison Project (NMIP-1) aimed at understanding and quantifying the historical budgets of global and regional terrestrial N₂O fluxes, environmental controls and uncertainties associated with model structure and parameters (Tian et al., 2018). To support the biannual assessment of the global N₂O budget and other associated activities such as RECCAP2, we plan to implement the second phase of NMIP simulations (NMIP-2).

• The Primary Goal of the NMIP-2 is to focus on the quantification and attribution of N₂O emissions, which includes the objectives as follows: 1) to provide more reliable estimates of regional (and/or national) and global soil N₂O emissions with an update to the year 2019/2020, covering a period of 1850-2020, 2) to attribute soil N₂O emissions to anthropogenic nitrogen addition and other factors, and 3) to provide an estimate of indirect N₂O emission from N leaching to inland water systems (River, Lakes and Reservoir).

• The Secondary Goal of NMIP-2 is to better understand the terrestrial nitrogen cycle and its interactions with other factors, which includes the objective as follows: 1) to provide an assessment of the global terrestrial nitrogen budget, 2) to understand and quantify N-Nexus: how do anthropogenic N inputs affect carbon (C sink), water (N loading), air (NH3 and NOx) and crop yield?

2. Timelines

• Input data available for modeling groups:  September 20, 2021;
• Deadline for model output submission: December 30, 2021 (for RECCAP2); January 30, 2022 (for ESSD paper – the Global N2O budget 2020)
• Data Analysis and Synthesis of all components of the global N2O budget: February-March 2022
• ESSD paper submission: The Global N₂O Budget 2020 (in Summer, 2022)
• The global terrestrial nitrogen budget (planned)
• Other papers ideas (to be developed)

3. Input data for NMIP-2

(Nitrogen Data Development Team: Hanqin Tian, Zihao Bian, Hao Shi, Xiaoyu Qin, Chaoqun Lu, Rongting Xu and Bowen Zhang, Shufen Pan, Francesco N. Tubiello, Jinfeng, Chang, Naiqing Pan)

To support NMIP-2, we have developed a comprehensive dataset of historical anthropogenic nitrogen input, with a spatial resolution of 0.25⁰ x 0.25⁰ latitude/longitude, annual time step covering a period from 1850 to 2020. This Harmonized Anthropogenic Nitrogen Input (HaNi) data product has integrated different sources of data including fertilizer N application in cropland (Lu and Tian 2017) and pasture (Xu et al. 2019), manure N production and application to cropland and pasture (Zhang et al. 2018, Xu et al. 2019), partition of ammonium N and nitrate (Nishina et al. 2017), and latest fertilizer and manure N data from FAO and IFA. The HaNi data includes 6 components: 1) fertilizer N application to cropland, ammonium N (NH₄⁺ -N) and nitrate N (NO₃⁻ -N), separately, 2) fertilizer N application to pasture, ammonium N (NH₄⁺ -N) and nitrate N (NO₃⁻ -N), separately, 3) Manure N application to cropland, 4) manure N application to pasture, 5) manure N deposition to pasture and rangeland, 6) atmospheric N deposition, NH_x and NO_y, separately. A detailed description of methods and data sources to generate the HaNi data is in preparation. Here we provide a brief description of all input data with a consistent spatial resolution (0.5⁰ x 0.5⁰ lat/lon) we used in NMIP-2:

1) Fertilizer N use in cropland

Annual synthetic N fertilizer use in cropland during 1910-2019 at 0.5º spatial resolution. This fertilizer dataset is consistent with LUH2 cropland distribution, and was developed by combining IFA country-level inventory, IFA crop-specific N fertilizer use rate, M3-crop type distribution map, and the LUH2 cropland dataset (Lu and Tian, 2017). Fertilizer N is further broken down into ammonium N (NH₄⁺ -N) and nitrate N (NO₃⁻ -N), according to the ratio developed by Nishina et al. (2017). N fertilization rate is assumed to be 0 before 1910.

2) Fertilizer N use in pasture

We provide annual synthetic N fertilizer use in pasture during 1901-2019. N fertilization rate is assumed to be 0 between 1901 and 1960. To develop spatially explicit datasets on fertilizer applied to pasture, we used country level proportion of total fertilizer allocated to grasslands from Lassaletta et al. (2014). The total N fertilizer applied to pasture was then divided by total pasture area based on LUH2 pasture dataset to develop spatially explicit datasets at 0.5^o × 0.5^o resolution. Fertilizer N is further broken down into ammonium N (NH₄⁺ -N) and nitrate N (NO₃⁻ -N), according to the ratio developed by Nishina et al. (2017).

3) Manure N application in cropland

Annual manure N application in cropland during 1860-2019 were developed on the basis of N production from manure in six livestock and poultry groups (duck, chicken, goat, sheep, pig, and cattle). And a portion of manure N applying to cropland is model-dependent (Zhang et al., 2017). Manure N application rate before 1860 is assumed to be the same with that in 1860.

4) Manure N application in pasture

To develop annual manure N application in pasture during 1860-2019, we used country-level “manure application to soils” dataset during 1961-2019 from FAOSTAT and separated total manure application amount in pasture and cropland based on their land area. The annual spatial patterns of manure production in Zhang et al. (2017) were used to spatialize manure N application in pasture. LUH2 pasture dataset was used to calculate the N application rate in pasture. Manure N application rate before 1860 is assumed to be the same with that in 1860.

5) Manure N deposition in pasture & rangeland

We also provide annual manure N deposition in grazing land (pasture + rangeland) during 1850-2019 at 0.5º spatial resolution. This dataset is developed on the basis of country-level manure deposition dataset during 1961-2019 from FAOSTAT. The annual spatial patterns of manure production in Zhang et al. (2017) were used to spatialize manure N deposition in grassland. LUH2 rangeland & pasture datasets were used to calculate the N deposition rate in grassland. For models which don’t simulate grazing in pasture and rangeland, we don’t suggest using manure N deposition as input to avoid unrealistic accumulation and losses of N because of net manure N input into the soil.

6) Atmospheric N deposition

Gridded NH_x and NO_y deposition during 1850-2020 were from the International Global Atmospheric Chemistry (IGAC)/Stratospheric Processes and Their Role in Climate (SPARC) Chemistry–Climate Model Initiative (CCMI) N deposition fields, which explicitly considered N emissions from natural biogenic sources, lightning, anthropogenic and biofuel sources, and biomass burning.

7) Climate data:

The climate data used to run historical simulations is the half degree CRU-JRA55 6-hourly forcing over 1901- 2020. Ian Harris (UEA) merged the “new generation” reanalysis from JRA-55 (Japanese 55-year Reanalysis) with the monthly CRU TS (v4.03) dataset. All JRA-55 data were regridded to the CRU 0.5° grid using appropriate NCL routines based on the Spherepack package, and masked to give a land-only (excluding Antarctica) dataset. For years between 1958 and 2020, JRA-55 was used, and alignment to CRU TS occurred where appropriate. For years before 1958, random (but fixed) years from JRA-55 for 1958-1967 were used to fill. Alignment to CRU TS applied separately to each instance, as appropriate (ie, using the appropriate CRU TS year).

8) Atmospheric CO₂ concentration

Annual CO₂ concentration during 1850-2020 were derived from ice core CO₂ data and NOAA annual observations. Data from March 1958 are monthly average from MLO and SPO provided by NOAA’s Earth System Research Laboratory. Data prior to March 1958 were estimated with a cubic spline fit to ice core data from Joos and Spahni (2008). Annual mean concentration was generated from these monthly data.

9) Land cover change (LCC)

For potential and present vegetation map, we used a mix of SYNMAP (one time, present indicates the year of 2000, and potential is the year of 1900 or before that, biome type and biome fraction, without wetland distribution), C3/C4 grass fraction, and wetland map (12-year (1993-2004) average of annual maximum inundated land surface percentage in each pixel). Historical distribution of cropland, pasture and rangeland during 1850-2020 were from LUH2 dataset (Hurtt et al., 2020), the original dataset is at 0.25° resolution, and we aggregated them to 0.5° to keep consistent with other datasets.

10) Irrigation Change

Annual irrigation data during 1850-2020 are from LUH2 (Hurtt et al., 2020).

11) Tillage pattern

This  one time gridded tillage  dataset  is  the  result  of  a  study  on  global  classification  of  tillage  practices  and  the spatially explicit mapping of crop-specific tillage systems for around the year 2005 (Porwollik et al., 2019). To simulate tillage impacts is optional

12) Historical burning area data

A gridded historical burned area data set will be provided if some participating model groups are capable of simulating fire effect. To simulate fire impacts is optional.

4. NMIP model simulation methods and experimental design

Model initialization: In NMIP2, the model simulations are divided into two stages: 1) spin-up and 2) transient runs. During the spin-up run, models were driven by the repeated climate data during 1901-1920 (earliest period that CRU-JRA climate data is available) and by other driving forces in 1850 (atmospheric CO₂ concentration, N deposition, N fertilizer use, manure N application, irrigation and land cover change). Each model group could determine the spin-up running years according to the model’s specific requirements for carbon, water, and nitrogen status. One standard for reference is that model reaches the equilibrium status when the differences of grid-level C, N, and water stocks were less than 0.5 g C m–2, 0.5 g N m–2, and 0.5 mm in two consecutive 50 years. Once these requirements were met, the spin-up run stopped, and the model reached an equilibrium state. Each model group is suggested to report their specific requirements in the readme file.

Simulation experiments: For the historical transient runs, there are 10 simulation experiments (SH0-SH10, see Table1). All historical runs start with the equilibrium carbon, water, and nitrogen status in 1850 obtained from previous spin-up run, and transiently run through the period 1850–2020. For the period 1850–1900 when CRU–JRA climate data are not available, the 20-yr average climate data between 1901 and 1920 will be used. SH0 is the reference run and is designed to track the model internal fluctuation and model drift. In SH0, we will use the 20-yr average climate data between 1901 and 1920, and all other drivers (CO₂, land cover, irrigation, Nfer, ManureN and Ndep) keep the level of 1850 throughout the simulation period. The purpose of SH1 run is to estimate the overall contribution of multiple drivers on nitrogen and carbon fluxes (SH1-SH0). SH2-SH8 are factorial experiments, which are used to quantify the effects of each driver on nitrogen and carbon fluxes (ManureN, Nfer, Ndep, irrigation, land cover, CO₂ and climate, respectively), and the effect of one specific factor is equal to the difference between S1 and cur­rent experiment (SH1-SH2, SH1-SH3, SH1-SH4, SH1-SH5, SH1-SH6, SH1-SH7, and SH1-SH8, respectively). SH9-SH0 is used to determine direct effect of Nfer and ManureN excluding their interaction with other factors. SH10-SH0 is used to determine direct effect of climate excluding its interaction with other factors. In the two experiments (SH11 and SH12), we keep the management practice constant at the level of 1850 to distinguish

the contributions of land cover change without the impacts of farming practices (irrigation, fertilizer N and Manure N applications).

Table 1. Simulation design

Historical	Climate	CO2	Land cover	Irrigation	Ndep	Nfer	ManureN
SH0	1901-1920	1850	1850	1850	1850	1850	1850
SH1	•	•	•	•	•	•	•
SH2	•	•	•	•	•	•	1850
SH3	•	•	•	•	•	1850	•
SH4	•	•	•	•	1850	•	•
SH5	•	•	•	1850	•	•	•
SH6	•	•	1850	•	•	•	•
SH7	•	1850	•	•	•	•	•
SH8	1901-1920	•	•	•	•	•	•
SH9	1901-1920	1850	1850	1850	1850	•	•
SH10	•	1850	1850	1850	1850	1850	1850
SH11	•	•	1850	1850	•	1850	1850
SH12	•	•	•	1850	•	1850	1850

Note: For historical simulations, “•” indicates the forcing during 1850-2020 is included in the simulation, “1901-1920” indicates the 20-year mean climate condition during 1901-1920 will be used over the entire simulation period, and “1850” indicates the forcing will be fixed in 1850 over the entire period. Climate data is only available from 1901, we assume the 20-yr average value between 1901 and 1920 for years 1850-1900. N deposition is available only from 1850. Manure N is available only from 1860, we assume manure N at the 1860 value for years 1850-1860. N fertilizer before 1910 was zero. In addition, we assume the N input data for 2020 is the same as 2019.  A gridded historical data set will be provided if participating model group is capable of simulating fire effect.

4. Required output variables

The output variables from each model should include nitrogen fluxes and pools, carbon fluxes and pools (see the attached Excel file). They are divided into two categories: the first priority variables and the second priority variables. The first priority variables are necessary (Table 3 and 4), and the second priority variables are optional (Please see appendix Excel File). We suggest that model teams provide annual/monthly outputs at the spatial resolution of 0.5º x 0.5 º latitude and longitude in the format of NetCDF.

Please define PFTs in the header of Vegtype level netcdf files, e.g. PFT 1 = broadleaf tree, PFT2 = … Please supply Fractional Land Cover [0-1] of PFT for each simulation as requested (1=total land). If Dynamic Vegetation is not enabled in your DGVM (i.e. changing natural PFT fraction in response to climate) please indicate (e.g. include information in an associated README file). Note the ocean fraction of any given gridcell may not be zero (e.g. at coastal gridcells). Please provide your gridbox fluxes in units per m2 of land fraction, PFT fluxes should be per m2 of PFT, and the PFT land cover fraction should be provided. Please upload the land-sea mask that you are applying.

5. Output file name convention

One file per variable for entire time-series. The name of output file should be Model_Simulation_variable.nc (e.g. DLEM_SH1_n2o.nc) for grid-level outputs and Model_Simulation_variablepft.nc (e.g. DLEM_SH1_n2opft.nc) for biome-level outputs. Please see the following requirements for an example netcdf header for variable nomenclature

The aim is to be more consistent with TRENDY, CMIP, LUMIP, LS3MIP in our format/variable requests to aid analysis:

1. Please follow the protocol (or explicitly state why not).
2. All modelling teams provide a methodology (in a README file) of how to calculate global annual N₂O emissions from gridded monthly files (grid and pft level). This will avoid confusion of whether to use landmasks/landcover /gridareas/etc.
3. Order of dimensions should be consistent. Eg [lon,lat,PFT,time]. When using ncdump this reads as [time,PFT,lat,lon].
4. Please provide a list of variables that are not applicable for your model. E.g. Nsoilpft might not exist. This gives us an idea of what variables we can request/expect.
5. Using cf-complient units. Remove “N” for nitrogen and “C” for carbon from the units and don’t measure time in years or months, e.g. N₂O emissions were previously requested in units kgN m^-2 s^-1, respectively, please remove the letter N to be cf-complient in the netcdf files.
6. Gridbox fluxes should be per m² of land
7. PFT fluxes should be per m² of PFT
8. Pools and coverages should be per m² of land
9. All models to provide a land fraction file if using regular lat-lon grids, or a land fraction and grid area if using non regular grids.
10. All models should use a consistent file naming (e.g. DLEM_SH1_n2o.nc/ DLEM_SH1_n2opft.nc ). Eg. do not include annual/monthly/perpft tag.
11. Following this, PFT labels are different among DGVMs (pft, PFT, vegtype…). Please all use nomenclature, PFT.
12. Consistent latitude/longitude use (e.g. do not use lat/lon)
13. Consistent fill value of -99999 to be used (e.g. not -9999)
14. All data from -180 -> 180 and -90 -> 90.
15. All annual models output over the same time period, 1850-2020, and monthly data cover 1980-2020.

To ensure accessibility by broad users, avoid to format netcdf files with netcdf library 4.4.0 or earlier, combined with libhdf5 1.10.0 or greater. There is a known issue with netcdf formatted by this set of libraries.