Nitrification and denitrification in the Community Land Model compared to observations at Hubbard Brook Forest

Models of terrestrial system dynamics often include nitrogen (N) cycles to better represent N limitation of terrestrial carbon (C) uptake but simulating the fate of N in ecosystems has proven challenging. Here, key soil N fluxes and flux ratios from the Community Land Model version 5.0 (CLM5.0) are compared to an extensive set of observations from the Hubbard Brook Forest Long-Term Ecological Research (LTER) site in New Hampshire. Simulated fluxes include microbial immobilization and plant uptake, which compete with nitrification and denitrification, respectively, for available soil ammonium (NH4+) and nitrate (NO3-). In its default configuration, CLM5.0 predicts that both plant uptake and immobilization are strongly dominated by NH4+ over NO3-, and that the model ratio of nitrification:denitrification is approximately 1:1. In contrast, Hubbard Brook observations suggest that NO3- plays a more significant role in plant uptake and that nitrification could exceed denitrification by an order of magnitude. Modifications to the standard CLM5.0 at Hubbard Brook indicate that a simultaneous increase in the competitiveness of nitrifying microbes for NH4+ and reduction in the competitiveness of denitrifying bacteria for NO3- are needed to bring soil N flux ratios into better agreement with observations. Such adjustments, combined with evaluation against observations, may help improve confidence in present and future simulations of N limitation on the C cycle, although C fluxes such as gross primary productivity (GPP) and net primary productivity (NPP) are less sensitive to the model modifications than soil N fluxes.
Methods
Modifications to CLM5.0 at Hubbard Brook LTER

The Community Land Model version 5.0 (CLM5.0) was modified at a single grid cell corresponding to the Hubbard Brook Experimental Forest, a northern hardwood forest site in the White Mountain National Forest in New Hampshire USA (43°56´N, 71°45´W).  The purpose was to test alternative parameterizations for nitrification and denitrification in CLM5.0.  The CLM5.0 simulations use site-level present day meteorological (from GSWP3 v. 1) and N deposition inputs, created by extracting the single grid cell values from the global gridded forcing data for CLM5.0 [Lawrence et al., 2019].  Atmospheric CO2 concentration (= 367 ppm), land use and N deposition (= 0.7 gN/m2/yr) were fixed at year 2000 conditions throughout the simulations, while meteorological forcings were cycled over 1991-2010.  The plant functional type of the grid cell was prescribed as 100% broadleaf deciduous temperate forest.  Spin-up for each simulation was run in accelerated decomposition mode for 400 years, followed by a final spin-up for 200 years, of which the last 20 years were sampled for the results archived here.  The N fluxes varied interannually but displayed no obvious drift or trends over these 20 years.  

We tested a variety of new parameterizations, described in detail in the text, in which model nitrification and/or denitrification was revised based on observed empirical relationships.

1a) Increased nitrification (Parton)

We added an NH4+ mineralization-based term to the CLM5.0 formula for potential nitrification in accord with the Parton et al. [2001] equation, from which the formula is derived. 

2a) Increased nitrification (Zhang)

In an alternative approach to boosting nitrifier competitiveness, we implemented a parameterization in which we parameterized potential nitrification as a direct linear function of gross mineralization multiplied by a scalar computed as (pH-4)6, reflecting the empirical linear relationship found by Zhang et al. [2018].  Since CLM5.0 has a uniform default pH of 6.5, this scalar was effectively 0.42

1b and 2b) Reduced denitrification (Reduced Denit)

We reduced the [NO3-]-limited and CO2 respiration-limited equations for potential denitrification by a factor of 100 and 10, respectively.  We ran two reduced denitrification modifications: 1b) Reduced Denitrification with Parton nitrification scheme (from modification 1a) and 2b) Reduced Denitrification with Zhang nitrification scheme (from modification 2a).

1c) Denitrification scaled to Nitrification (Denit=Nitrif/10)

We tested an alternative parameterization, building off modification 1a), to reduce the rate of denitrification.  In this alternative approach, we bypassed the Del Grosso et al. [2000] algorithm altogether and instead set potential denitrification equal to potential nitrification divided by 10. 

1bx) No N2 fixation

We turned off N2 fixation (beginning from year 1 in the spin-up phase) due to concern that CLM adds an excessive amount of N to northern temperate ecosystems such as Hubbard Brook that lack symbiotic N2 fixers and where heterotrophic N fixation rates are low.   N deposition was left turned on in this experiment.  Modification 1bx was performed with the 1b) modifications (Parton increased nitrification and Reduced Denitrification adjustments) also turned on.

3) Swap NO3-  

The order of competition for mineral N between plants and soil microbes was switched such that they competed first for NO3- and second for NH4+.  This was a swap in the sense that the default CLM5.0 competition occurs in the opposite order, i.e., first for NH4+ then for NO3-.  Aside from reversing that order, we made no other adjustments to the algorithms for potential nitrification and denitrification in modification 3.