The integration of nitrogen dynamics into a land surface model.Part 1:model description and site-scale validation

YANG Xiujing,DAN Li,YANG Fuqing,PENG Jing,LI Yuyu,GAO Dongdong,d,JI Jinjun nd HUANG Mi

aKey Laboratory of Regional Climate-Environment for Temperate East Asia,Institute of Atmospheric Physics,Chinese Academy of Sciences,Beijing,China;bCollege of Earth and Planetary Sciences,University of Chinese Academy of Sciences,Beijing,China;cKey Laboratory of Water Cycle and Related Land Surface Processes,Institute of Geographical Sciences and Natural Resources Research,Chinese Academy of Sciences,Beijing,China;dSchool of Atmospheric Sciences,Chengdu University of Information and Technology,Chengdu,China;eKey Laboratory of Ecosystem Network Observation and Modeling,Institute of Geographic Sciences and Natural Resources Research,Chinese Academy of Sciences,Beijing,China

ABSTRACT Nitrogen cycling has profound effects on carbon uptake in the terrestrial ecosystem and the response of the biosphere to climate changes.However,nutrient cycling is not taken into account in most land surface models for climate change.In this study,a nitrogen model,based on nitrogen transformation processes and nitrogen fluxes exchange between the atmosphere and terrestrial ecosystem,was incorporated into the Atmosphere-Vegetation Interaction Model(AVIM)to simulate the carbon cycle under nitrogen limitation.This new model,AVIM-CN,was evaluated against site-scale eddy covariance-based measurements of an alpine meadow located at Damxung station from the FLUXNET 2015 dataset.Results showed that the annual mean gross primary production simulated by AVIM-CN(0.7073 gC m-2d-1)was in better agreement with the corresponding flux data(0.5407 gC m-2d-1)than the original AVIM(1.1403 gC m-2d-1)at Damxung station.Similarly,ecosystem respiration was also down-regulated,from 1.7695 gC m-2 d-1to 1.0572 gC m-2d-1,after the nitrogen processes were introduced,and the latter was closer to the observed vales(0.8034 gC m-2d-1).Overall,the new results were more consistent with the daily time series of carbon and energy fluxes of observations compared to the former version without nitrogen dynamics.A model that does not incorporate the limitation effects of nitrogen nutrient availability will probably overestimate carbon fluxes by about 40%.

KEYWORDS Coupled carbon and nitrogen dynamics;nitrogen limitation;land surface model;carbon-nitrogen-water cycles

1.Introduction

The terrestrial carbon cycle in land surface models is primarily used to predict and quantify the variations of carbon fluxes under the in fluences of rising CO2,rising nitrogen deposition,and changing climate and landuse type(Peng and Dan 2015;Tian et al.2016).Unfortunately,the-magnitude-ofprojected-plant responses to the processes mentioned above remains highly uncertain(Ahlstr?m et al.2012;Ballantyne et al.2012;Anav et al.2013;Arora et al.2013;Walker et al.2014;Luo,Keenan,and Smith 2015;Shao et al.2016).According to previous investigation,approximately 28%of CO2emissions from fossil fuel combustion and land use is assimilated by terrestrial ecosystems annually(Le Quere et al.2013).On the other hand,nitrogen limitation counteracts CO2fertilization effects remarkably(Shi et al.2016;Sokolov et al.2008;Thornton et al.2007;Wang,Law,and Pak 2010;Zaehle,Friedlingstein,and Friend 2010;William et al.2015).Based on decade-long field experiments and the simulations of the fifth phase of the Coupled Model Intercomparison Project(CMIP5),Peng and Dan(2015)and Reich and Hobbie(2013)found that,without nitrogen constraints,earth system models perhaps overestimate terrestrial photosynthesis and productivity under the condition of CO2fertilization.In addition,nitrogen deposition,which has dramatically increased owing to the rapid expansion of industrialization and the heavy use of fertilizer in agriculture(Galloway et al.2004;Lu and Tian 2014;Cai et al.2016),boosts CO2fixation by plants(Lu et al.2012;Thomas,Bonan,and Goodale 2013;Chen et al.2015;Zhang et al.2015).However,such enhancement is not applicable in all regions across the globe(Nadelhoffer et al.1999).

The terrestrial carbon stock and carbon flux is subject to the restrictions imposed by the amount of available-nitrogen.Nitrogen-strongly-restricts-plant photosynthesis,organic matter decomposition,carbon allocation,and the response of ecosystems to the increased atmospheric CO2concentration.However,due to the complexity of the nitrogen cycle,previous studies on the global changes of ecosystem processes have focused mainly on the impacts of temperature,moisture,and the concentration of CO2.Little research has been carried out on the mechanisms of carbonnitrogen cycle interactions and the responses of the carbon-nitrogen cycle to other factors,including surface carbon-water fluxes.Recent studies have shown that primary production in the tropical and subtropical forests of China is limited by the nitrogen supply(Fang et al.2011),and plants can adapt to long-term elevated nitrogen deposition by increasing transpiration to maintain a nutrient balance(Lu et al.2018).In addition,nitrogen limitations also reduce the land carbon sink(Exbrayat et al.2013).Presently,land models-including-theAtmosphere-Vegetation-Interaction-Model(AVIM),developed over the past several decades in China-are mainly static models,with soil fertility and leaf nitrogen content fixed as constants.A model of evapotranspiration controlled by nitrogen cycling was established by Dickinson et al.(2002),and it showed that the carbon and nitrogen stores respond to interannual changes of climate.Jia,Wang,and Xie(2018)suggest that the incorporation of a prognostic nitrogen cycle into CLM4.5 reduces the predicted gross primary production(GPP)and its interannual variability.Other studies have also proven the necessity to incorporate nitrogen processes in land process models(Kirschbaum 1999;Lin et al.2000;Kirschbaum and Paul 2002;Yang et al.2009;Cai et al.2016).Also,few coupled models consider the nitrogen cycle;for example,among all the models that participated in the phase5oftheCoupled-ModelIntercomparison Project(CMIP5),only CESM,NorESM,and BNU-ESM have nitrogen modules in their land surface models,and-are-relatively-simple.Furthermore,very-few coupled models have considered nitrogen dynamics(Yu and Piao 2014).

To enable better projection of the responses of the terrestrial carbon cycle to climate changes,it is necessary to introduce the in fluences of nutrients on the terrestrial carbon-cycle,especially-carbon-nitrogen-coupled dynamics.In this study,we incorporated a nitrogen model into the AVIM land surface model,which has been coupled with the FGOALS global model(Peng and Dan2014)andaregionalmodel(Dan,Cao,andGao2015),and validate its performance using site-scale observed data at Damxung station during 2004-05.And the further region validated tests of the new nitrogen model will be elaborated in the following second part paper.

2.Model,data,and methods

2.1 Model description

The AVIM land surface model incorporates physical processes(Ji and Hu 1989),ecophysiological processes(Ji 1995)and soil carbon dynamic processes(Huang et al.2007;Ji,Huang,and Li 2008).The model participated in the-IGBP’sEcosystem-Model-Data-Intercomparison Project,and has shown outstanding performance in terms of carbon flux(Dan and Ji 2007).It has a unique capability in simulating the soil carbon for earth system models(Todd-Brown et al.2013).

AVIM has been validated for various ecosystems,including forests,shrubland,grassland,and cropland(Dan,Ji,and Li 2002;Dan,Ji,and He 2007;Huang et al.2007;Ji,Huang,and Liu 2005;Ji,Huang,and Li 2008),indicating AVIM can simulate the physical,biological,and biogeochemical processes of different plant types at local,regional,continental,and global scales.Compared with the former version with the carbon cycle,a soil nitrogen dynamic model combined with the nitrogen cycle was incorporated to form a new version,AVIM-CN.In AVIM-CN,the maximum carboxylation is constrained by the available nitrogen in soil:Here,f(T),f(Ws),and f(N)are the effects of canopy temperature,soil water content,and available nitrogen in soil on photosynthesis(Dan,Ji,and He 2007).f(N)varies with the amount of inorganic nitrogen uptake by roots relative to the requirement of plant photosynthesis(Ndemand).To compare the performance of the new version of the model with coupled carbon and nitrogen processes,f(N)is set as a constant value,1,in the original model as a control run.A detailed description of the nitrogen model can be found in the supplementary material.

2.2 Data and modeling protocol

The driving data were a half-hourly dataset during 2004-05,including air temperature,specific humidity,wind velocity,air pressure,precipitation,and short-and longwave radiation data from the FLUXNET2015 dataset observed at Damxung(site ID:CN-Dan)in an alpine grassland ecosystem(30.4978°N,91.0664°E),approximately 3 km from Damxung County in Tibet,China.The measurement site is part of the Chinese Terrestrial Ecosystem Flux Observational Network(ChinaFlux).The annual mean temperature was 2.18°C during the study period of 2004-05,due to its high elevation of 4295.7 m.The mean precipitation was 520.15 mm during 2004-05,with 85.5%concentrated in the growing season from June to September.The plant community,classified as meadow-steppe,is dominated by Kobresia pygmaea,Stipa capillacea Keng,and Carex montiseverestii Kükenth at the flux measurement site.

The simulations were conducted by AVIM and AVIMCN independently,to validate the new model’s performance.Themodelswerealsocomparedagainst observed daily data obtained by the flux tower.The validation data used here included soil surface temperature,soil surface water content,sensible heat flux,latent heat flux,GPP,net ecosystem exchange(NEE),and ecosystem respiration.

3.Model results and analysis

Figure 1.Daily variation of observed(black spots)and modeled(red line for AVIM;blue line for AVIM-CN)(a)GPP,(b)NPP,(c)aboveground biomass,(d)LAI,(e)belowground biomass,and(f)soil carbon storage,at Damxung station during 2004-05.

The introduction of carbon-nitrogen coupling had a significant effect on the simulation of GPP,as shown in Figure 1(a).The annual mean GPP modeled by the new model,AVIM-CN,was 0.7073 gC m-2d-1,and the simulation with the carbon cycle only was 1.1403 gC m-2d-1,compared to the site observation of 0.5409 gC m-2d-1.The fraction of GPP reduction due to nitrogen limitation was 37.97%.The above results indicate that nitrogen is a key factor for the carbon flux in steppe meadow on the Tibetan Plateau,especially during the growing season,due to the great carbohydrate consuming process to obtain nitrogen in the soil poor in nutrients.The root-mean-square error(RMSE)of AVIM-CN was 0.3387 gC m-2d-1,against 0.9147 gC m-2d-1for AVIM.The correlation coefficient(r)of AVIM-CN with observation(r=0.9293)was also slightly larger than for AVIM(r=0.9282).NPP,leaf area index(LAI),and above-and belowground biomass,in Figure 1(b-e),were all reduced significantly,due to the insuffi-cient availability of nitrogen nutrients according to the biogeochemical processes,with down-regulation proportions of 36.52%,14.35%,27.52%,and 51.38%,respectively.According to field experiments conducted by Shi et al.(2006),at the same location in 2004,the observed annual peak value was 67.9 gC m-2for aboveground biomass.The modeled leaf biomass was 94.97 gC m-2for AVIM and 72.47 gC m-2for AVIM-CN,indicating the latter was more reasonable.The effects of nitrogen limitation were more obvious in the root than that in the leaf,and the former was mainly affected during the growing season while the latter was affected throughout the year.With the effects of deficient inorganic nitrogen,the simulated soil carbon pool reduced by 26.51%,from 10.7120 kgC m-2to 7.8725 kgC m-2(Figure 1(f)).

The simulation with a coupled carbon-nitrogen cycle improved the lower soil moisture of the original model(Figure 2(a)).The RMSE value was 0.0541 m3m-3,with a correlation coefficient of 0.8587 at the 95%con fidence level,compared with a higher RMSE(0.0544)and lower correlation coefficient(0.8442)without nitrogen cycling.The smaller LAI in AVIM-CN when nitrogen limitation was taken into consideration led to less canopy interception of rainfall,and therefore more throughfall of rain entering into the soil and increasing the soil moisture.Subsequently,less water would be evaporated to the overlying air via transpiration effects with less vegetation morphology under nitrogen limitation,which would lead to more water remaining in the soil(Figure 2(b)).Transpiration was 19.2571 W m-2in AVIM and 18.4933 W m-2in AVIM-CN.Figure 2(c)shows that the simulation of sensible heat flux improved a little,with the RMSE decreasing by 0.6182 W m-2and the correlation coefficient increasing by 0.024.The modeled peaks of latent heat flux shown in Figure 2(d)(86.793 W m-2for AVIM and 78.972 W m-2for AVIM-CN)were-much-lowerthan-the-observed-peak-of 219.2344 W m-2.However,the observed annual peak of latent heat flux at Damxung station during 2004-05 was no more than 100 W m-2(Li et al.2008;Liu et al.2010),which showed that the observed data of FLUXNET 2015 were overestimated and our simulations captured the actual latent heat flux.

Figure 2.Daily variation of observed(black spots)and modeled(red line for AVIM;blue line for AVIM-CN)(a)soil water content,(b)transpiration,(c)sensible heat flux,and(d)latent heat flux,at Damxung station during 2004-05.

Nitrogen limitation produced negative changes in respiration,as shown in Figure 3(a).Compared with the-observed-respiration,-the-simulated-RMSE decreased significantly from 1.3468 gC m-2d-1(carbononly)to 0.5842 gC m-2d-1(coupled carbon-nitrogen),and was greatly improved by incorporating the nitrogen dynamics in AVIM-CN.The average respiration modelled by AVIM-CN(1.0572 gC m-2d-1)agreed more with the measurements(0.8034 gC m-2d-1),as compared to 1.7695 gC m-2d-1from AVIM.We also analyzed the three main components of ecosystem respiration,i.e.soil respiration,plant growth,and maintenance respiration(Figure 3(b-d)).The simulations of the three components were regulated down by 36.61%,36.51%,and 50.80%,respectively,due to carbon-nitrogen interaction.The magnitude of maintenance respiration was the largest,which may be explained by the reduction in root biomass maintained throughout the whole year.For the other two types of respiration,temperature might be the key control factor during the non-growing season,and the constraints of nitrogen on respiration mainly focused on summer.However,-incorporating-carbon-nitrogen-coupled dynamics did not improve the net ecosystem changes at Damxung station(Figure 3(e)).The RMSE increased slightly from 0.4942 to 0.5113,while the correlation coefficient decreased from 0.5387 to 0.4286.This might be attributable to the nonlinear interaction of new-parameters-related-to-the-nitrogen-cycle.Compared to NPP,the NEE showed large uncertainty due to the limited observations and the complexity of soil respiration in the model.Nitrogen dynamics had no strong effect on the soil temperature.The annual mean soil temperature was 5.8071°C for the carbon-nitrogen coupled cycle,verses 5.8017°C for the carbon-only cycle,which was more consistent with observation(6.0325°C).In terms of RMSE and correlation coefficient,AVIM-CN also performed slightly better than AVIM,as shown in Figure 3(f).

Figure 3.Daily variation of observed(black spots)and modeled(red line for AVIM;blue line for AVIM-CN)(a)ecosystem respiration,(b)soil respiration,(c)growth respiration,(d)maintenance respiration,(e)net ecosystem exchange,and(f)soil surface temperature,at Damxung station during 2004-05.

4.Conclusions and discussion

The new model,AVIM-CN,not only simulated the GPP,ecosystem respiration and NEE closer to observation,but also improved the simulation of soil surface temperature,soil water content,and sensible heat fluxes.Results showed that carbon fluxes and storage simulated by AVIM-CN were reduced by approximately 50%under nitrogen limitation.The nitrogen cycle in other land surface models,e.g.CLM4.5,has also down-regulated NPP(Peng and Dan 2015;Jia,Wang,and Xie 2018),which shows consistency with our study and the importance of nitrogen limitation for carbon cycles.

According to the site-scale observations and previous research,it is clear that the incorporation of key nitrogen processes within AVIM-CN was successful.The performance of the land surface model in simulating the interaction of carbon and nitrogen within the terrestrial ecosystem was improved overall.GPP,LAI,aboveground biomass,soil temperature,and water content were simulated more reasonably by AVIM-CN.The coupling of the terrestrial carbon and nitrogen cycles in AVIM-CN improved the carbon fluxes under the limitation of nitrogen nutrients.It turned out that nitrogen dynamics just down-regulated the magnitude of the diurnal variations of the carbon fluxes,instead of altering their phase.This might be of great importance for the investigation of carbon flux in the Tibetan Plateau with global warming.

It is important to note that nitrogen dynamics might cause some uncertainty,by introducing carbon-nitrogen interaction processes and bringing nonlinear feedback into the model.The largest uncertainty lies in the simulation of soil respiration and NEE.Carbon sequestration is likely to be overestimated by land surface models that do not consider nitrogen constraints on the carbon cycle.This would impact the terrestrial carbon balance driven by the carbon-nitrogen dynamics,which again shows the importance of the coupling of nitrogen for land surface models.

Acknowledgments

We acknowledge the data providers for the FLUXNET2015 dataset at Damxung station(https:// fluxnet. fluxdata.org/data/ fluxnet2015-dataset).

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This study was supported by a project of the National Key Research and Development Program of China[grant number 2016YFA0602501]and a project of the National Natural Science Foundation of China[grant numbers 41630532 and 41575093].