B7: Improved Nitrogen Cycling in Sea Ice Design Document
The Design Document page provides a description of the algorithms, implementation and planned testing including unit, verification, validation and performance testing. Please read Step 1.3 Performance Expectations that explains feature documentation requirements from the performance group point of view.
Design Document
In the table below 4.Equ means Equations and Algorithms, 5.Ver means Verification, 6.Perf - Performance, 7. Val - Validation, 
- competed, 
- in progress, 
- not done
Overview table for the owner and an approver of this feature
1.Description | Corrects inadequacies in sea ice nitrogen dynamics |
| 2.Owner | Nicole Jeffery |
| 3.Created | 08/26/2019 |
| 4.Equ | |
| 5.Ver | |
| 6.Perf | |
| 7.Val | |
| 8.Approver | Katherine Calvin (Unlicensed) |
| 9.Approved Date |
Title: Improved nitrogen cycling in sea ice
Requirements and Design
E3SM Sea Ice Group
Date:Aug 26, 2019
Summary
The E3SMv1.1 CBGC simulations of sea ice biogeochemistry (BGC) drastically underestimate ice algal primary production in the Arctic. This is in large part due to biases in the upper ocean nitrate pool. However, sea ice algal primary production is also underestimated in the Southern Ocean, albeit not as drastically, where surface nutrients do not play a role. This suggests insufficiencies in our modelled biochemical interactions. Several studies (Fripiat et al. 2013, New insights into sea ice nitrogen biogeochemical dynamics from the nitrogen isotopes papers; Baer et al. 2015, Nitrogen uptake dynamics in landfast sea ice of the Chukchi Sea) point to dissolved organic nitrogen (DON) remineralization and nitrate production from nitrification as important nutrient sources for algal production. These sources are important for two reasons. First, DON is found in high concentrations in the polar oceans and, unlike nitrate, DON adsorbs to the ice crystals during sea ice growth and so can accumulate to high concentrations in the sea ice. Remineralization of DON produces ammonium which is readily taken up by ice algae, in fact, preferentially to nitrate. Thus DON remineralization to ammonium provides a new and abundant source of nitrogen for ice algae that will almost surely mitigate our production biases. Second, nitrate produced from nitrification is distinct from the ocean source of nitrate because it is trapped in a biofilm which adsorbs to ice crystals (Deman et al. 2019, High production going along with respiration, Impact of bio-film formation for sea ice biogeochemistry, IGS Sea Ice Symposium). This process allows for build-up of nitrate in the sea ice brine, a feature observed in sea ice cores (Baer et al., 2015; Duarte et al., 2017 Sea ice thermohaline dynamics and biogeochemistry in the Arctic Ocean: Empirical and model results) but that is virtually impossible in our current model design.
The current model has a DON pool that follows the transport physics of adsorbing BGC tracers. Fluxes between ocean and sea ice have already been implemented. In terms of biochemistry, DON is produced in situ during algal mortality and has a loss due to bacterial reminerialization, however that loss is tracked but not included as a source of ammonium. This source will be added and tuned based on observations. Nitrification is already implemented in the v1 model but with a default rate constant of zero. Observations indicate a rate constant of 0.046 per day. The impacts of the bio-film, however, still need to be added. This requires some minor coding to allow for adsorption of nitrate during the biogeochemical reaction process but not, as occurs for DON, during brine transport.
Alterations of the nitrogen cycle should also conserve nitrogen and, because we use fixed carbon to nitrogen ratios, carbon as well. Remineralization of DON, production of DON from zooplankton grazing and uptake of inorganic sources of nitrogen by algae can be made closed by requiring all zooplankton losses to entire the DON pool. For carbon, this requires enabling the DIC tracer (already in MPAS-seaice and coupled to MPAS-ocean) and adding a loss term for algal uptake and a remineralization proportional to DON and DOC loss terms.
The majority of code changes will involve a the MPAS-SI column package and the ice_algae.F90 subroutine. Enabling of DIC will require and additional parameter in Registry, a new namelist parameter and some small changes to the mpassi interface. Implementation and tuning will require updates to the default values in namelist mpassi_in. It is expected that these changes will conserve carbon, at least in the reaction equations, and increase ice algal primary production (reduce biases) in both polar regions, though we still expect underestimations in the Arctic.
Requirements
Requirement: Add remineralization and nitrification fluxes to sea ice biogeochemical reactions.
Date last modified: Aug 27, 2019
Contributors: Nicole Jeffery
Currently, dissolved organic nitrogen (DON), ammonium (NH4), nitrate (NO3), and algal nitrogen (N) are the four fundamental pools in the sea ice nitrogen cycle. They satisfy coupled non-linear advection-diffusion-reaction equations. As modelled, this cycle is not closed, though losses are tracked in a diagnostic zooplankton pool (ZOO). With conservation in mind, we will modify the DON equation source terms so that together with ammonium sources they balance nitrogen losses from algae during grazing and mortality. We will then add a source term to the NH4 equation that can be made equivalent (via a parameter, fdon) to the remineralization loss term already present in the DON equation. Finally, we will change namelist parameters to turn on nitrification at observed rates.
This development will be first tested algebraically to show conservation of nitrogen. In addition, the ZOO diagnostic will be used to print an error message if nitrogen is not conserved in the reaction terms. We can test this in short stand alone MPAS-SI runs similar to the standard-bgc testsuite.
Requirement: Add adsorption of nitrate produced from nitrification
Date last modified: Aug 27, 2019
Contributors: Nicole Jeffery
Nitrate tracers are currently purely mobile. This means that they never adhere to the ice crystals and cannot build up in the brine in significant excess of the ocean concentration. Tracers that adhere, such as DON, move between the mobile and stationary phases via prescribed timescales with the condition that the mobile to stationary transformation occurs only during sea ice growth and the stationary to mobile transformation occurs during sea ice melt. In the case of nitrate, we first must define a stationary fraction of nitrate equal to the nitrification source term. Secondly, we need to implement the latter transformation, stationary to mobile, while preventing mobile to stationary transformations.
With the nitrification parameter set to 0, the code should be identical up to round off error in the biogeochemical tracers. With nitrification at observed values, we expect to see higher nitrate concentrations in some polar sea ice, but no unphysical accumulations. This can be tested by activating a warning flag already implemented that will indicate higher than expected concentrations of BGC tracers.
Requirement: Enable DIC (dissolved inorganic carbon)
Date last modified: Sept 13, 2019
Contributors: Nicole Jeffery
The dissolved inorganic carbon tracer was partially implemented in v1. In particularly, DIC is already coupled to MPAS-O, the tracer transport is enabled with a config_use_carbon namelist option, default dimensions already exist in Registry and are set to appropriate values. However, the parameter config_mobility_type_inorganic_carbon needs to be added in Registry and the namelist to ensure that the correct transport physics treatment is used. As an initial test, we will treat DIC in the same manner as DOC, an initially mobile tracer that strongly adsorbs to the ice matrix (config_mobility_type_inorganic_carbon = 0.0) .
Reaction source and sink terms for DIC are straight forward, following the nitrogen cycle.
Algorithmic Formulations
Design solution: modify the reaction terms in the NH4 and DON equations
Date last modified:Aug 27, 2019
Contributors: Nicole Jeffery
- Requirement: Add remineralization and nitrification fluxes to sea ice biogeochemical reactions.
The improved formulation of the reaction equations for NH4, DON, algal nitrogen (N) and NO3 are described in NitrogenCycleMPAS_SI.pdf
The v1 code does not include DONremin in equation (3) and the fraction of grazing contributing to DON in equation (4) has been modified from fdon fgs to (1-fgs + fng fgs) to close the grazing pathway. In addition, Algal mortality is now a source of DON.
The carbon only DOC source terms from grazing and mortality should be the additional carbon after removing the source to DON (converted to carbon units using R_C2N_DON).
2. Requirement: Add adsorption of nitrate produced from nitrification
The sea ice vertical bgc subroutine follows the algorithm: 1) compute the new mobile and stationary fractions based on the sea ice area tracer, f_mobile, and evolution timescales; 2) solve the vertical transport equation for the mobile fraction; 3) evolve the stationary fraction based on sea ice boundary changes (melt and growth); 4) solve the reaction equation for the total brine concentration of the tracer; and 5) combine to evaluate the new bulk tracer concentration.
In order to add adsorption of nitrate from nitrification, we need to first modify (1) to allow transformation from stationary to mobile for nitrate but not mobile to transformation. The change in the mobile fraction for tracer C, dmobile(m), is currently:
dmobile = mobileC* ([C]mobile * (exp(-dt * tauret)-1) + [C]stationary* (1-exp(-dt * taurel))
where mobile_C is 1 for tracers that move between phases and 0 for purely mobile tracers like nitrate. If we now use
dmobile = mobileC* ([C]mobile * (exp(-dt * tauret)-1) + [C]stationary* (1-exp(-dt * taurel)) + (1-mobileC )* [C]stationary* (1-exp(-dt * taurel))
then tracers with mobile_C = 0 (nitrate and silicate) will have transformations in one direction only, stationary to mobile. Although we are not adding silicate remineralization, it would now be a simple matter to include this improvement if later warranted.
3. Requirement: Enable DIC
Reaction source (DIC_s) and sink (DIC_r) terms for DIC are as follows:
DIC_r = 0
DIC_s = 0
do n = 1, n_algal ! loop over algal groups
DIC_r = DIC_r + (1-fr_resp) * grow_N(n) * R_C2N(n) * dt ! fr_resp is the fraction of respiration losses,
! grow_N(n) is the uptake of nitrogen by algal n and R_C2N(n) is the carbon to nitrogen ratio of algal n
end do
do n = 1, n_don ! loop of DON groups
DIC_s = DIC_s + DON_r(n) * R_C2N_DON(n) ! DON_r is the remineralization term for DON
! R_C2N_DON(n) is the carbon to nitrogen ratio of DON n
end do
do n = 1,n_doc ! loop over the DOC groups
DIC_s = DIC_s + DOC_r(n).
end do
Design and Implementation
Implementation: Improvements to nitrogen cycling reaction terms
Date last modified: Sept 10, 2019
Contributors: Nicole Jeffery
In the subroutine algal_dyn (column/ice_zbgc.F90) add source terms from zooplankton grazing:
The current source term to DON (DON_s) is
DON_s(n) = graze_N*f_don(n)*fr_graze_s * dt
Change to
DON_s(n) = graze_N*dt - Am_s_e + mort_N*dt - Am_s_mo
Here we've subtracted the ammonium source terms for grazing and mortality from the total nitrogen sources (graze_N+mort_N)*dt.
And allow remineralization of DON to ammonium (AM)
Am_s = Am_s + DON_r(n)*f_don_Am(n)
Save the nitrification source of Nitrate for adsorption to ice crystals:
nitrification = Nit_s_n
Implementation: Adsorption of nitrification source of nitrate to ice crystals
Date last modified: Sept 10, 2019
Contributors: Nicole Jeffery
Redefine the stationary (adsorbed) concentration of a tracer to allow for nitrified nitrate:
initcons_stationary(k) = mobile(mm)*(in_init_cons(k,mm)-initcons_mobile(k))
+ initcons_stationary(k) = max(c0,in_init_cons(k,mm)-initcons_mobile(k))
Allow release of nitrate but not adsorption:
dmobile(k) = mobile(mm)*(initcons_mobile(k)*(exp(-dt*rtau_ret( mm))-c1) + &
initcons_stationary(k)*(c1-exp(-dt*rtau_rel(mm))))
Becomes...
dmobile(k) = mobile(mm)*(initcons_mobile(k)*(exp(-dt*rtau_ret( mm))-c1) + &
initcons_stationary(k)*(c1-exp(-dt*rtau_rel(mm)))) + &
(1-mobile(mm))*initcons_stationary(k)*(c1-exp(-dt*rtau_rel(mm)))
Change
if (sum_tot > c0 .and. mobile(mm) > c0) trcrn(nt_zbgc_frac+mm-1) = sum_new/sum_tot
To allow for adsorption when mobile(mm) = c0 (i.e. for nitrate):
if (sum_tot > c0) trcrn(nt_zbgc_frac+mm-1) = sum_new/sum_tot
Then using "nitrification" defined in algal_dyn and passed to z_biogeochemistry:
if (m .eq. nlt_bgc_Nit) then ! for nitrate
initcons_mobile(k) = max(c0,(biomat_brine(k,m)-nitrification(k) + &
react(k,m))*iphin_N(k)*trcrn(nt_zbgc_frac+m-1))
initcons_stationary(k) = max(c0,((c1-trcrn(nt_zbgc_frac+m-1))*(biomat_brine(k,m)- &
nitrification(k) + react(k,m)) + nitrification(k))*iphin_N(k))
sum_new = sum_new + initcons_mobile(k)*dz(k)
sum_tot = sum_tot + (initcons_mobile(k) + initcons_stationary(k))*dz(k)
end if
And at the end of the subroutine, redefine the mobile fraction for nitrate
if (m .eq. nlt_bgc_Nit .and. MAXVAL(nitrification) > c0) then
trcrn(nt_zbgc_frac+m-1) = zbgc_frac_init(m)
if (sum_tot > c0) trcrn(nt_zbgc_frac+m-1) = sum_new/sum_tot
end if
enddo
Implementation: Enable DIC
Date last modified: Sept 13, 2019
Contributors: Nicole Jeffery
The most challenging aspect here is to conserve both carbon and nitrogen. There are additional organic carbon pools that can track material that is not in the DON (protein) pool. Here we redefine the DOC source terms as differences between the total carbon grazed or from algal mortality and subtract the contribution that fluxed to the DON pool (in carbon units). For n = 1,2
DOC_s(n) = f_doc(n)*(graze_C*dt + mort_C*dt - DON_s(1)*R_C2N_DON(1))
For conservation, f_doc(1) + f_doc(2) = 1. We'll take f_doc(1) = f_doc(2)= 0.5 for now since we do not yet model the impact of organic macromolecules.
DIC reaction equations are straight forward, but enabling DIC requires an addition (config_mobility_type_inorganic_carbon) to the default namelist, namelist definitions and build-namelist.
We will also add two variables (dC and dN) to sum the total changes in carbon and nitrogen in each grid level and timestep. These values should sum to zero up to round off error.
Planned Verification and Unit Testing
Verification and Unit Testing: Run standard tests (bgc and physics)
Date last modified: Sept 10, 2019
Contributors: (Nicole JefferyAdrian Turner)
seaice/njeffery_zbgcNMods will be tested to check for BFB in standard_physics test suite and the snow_tracer_physics test suite which include regression tests with seaice/develop master, parallelism and restartability tests.
Will also test standard_bgc and the aerosol_shortwave test suite. These will not be BFB in the regression tests, but both parallelism and restartability should pass. Changes in BGC fields will be significant for standard_bgc. All the tests should run to completion. That is, both nitrogen and carbon conservation tests should pass. For the aerosol_shortwave regression test, we should only see round off level differences in the verticalAerosolConc and mobileFraction tracers.
Update: These tests were completed on Sept 17, 2019. All tests were BFB except the regression standard_bgc and regression aerosol_shortwave. In the latter, verticalAerosolConc and mobileFraction tracers differed only to round off. Conservation tests for nitrogen and carbon reaction kinetics passed. PR to seaice/develop is submitted https://github.com/MPAS-Dev/MPAS-Model/pull/369. These tests will be performed by Adrian Turner before the PR is accepted.
Planned Validation Testing
Validation Testing: GCASE tests with ocean-ice bgc
Date last modified: Sept 10, 2019
Contributors: (Nicole Jeffery)
We wil also run three tests:
- Run 5 year GCASE (CORE 2) with restart files from bgc run
- Run 5 year GCASE with new code but default namelist parameters except those critical for conservation (this will reduce the impact of our nitrogen cycling improvements)
- Run 5 year GCASE with new code and updated parameters.
We are looking for increases in total sea ice primary production (PP) in both the Arctic and Southern ocean regions particularly in test 1 compared with test 3. We will not tune further at this stage until all the physics and bgc code improvements are in. Current model versions underestimate total PP in both poles. We should also see increases in maximum and average sea ice nitrate concentrations but values should not build up unphysically. There is some data which shows a 1-10x increase in sea ice nitrate relative to the ocean values in some locations.
Planned Performance Testing
Performance Testing: Compare performance of GCASE runs
Date last modified: Sept 10, 2019
Contributors: (Nicole Jeffery)
Compare years per day of 1 and 3 (on anvil). We don't expect to see any degradation of performance.