An improved process-oriented hydro-biogeochemical model for 1 simulating dynamic fluxes of methane and nitrous oxide in alpine 2 ecosystems with seasonally frozen soils 3

The hydro-biogeochemical model Catchment Nutrient Management Model DeNitrification-DeComposition 16 (CNMM-DNDC) was established to simultaneously quantify ecosystem productivity and losses of nitrogen and carbon at the 17 site or catchment scale. As a process-oriented model, this model is expected to be universally applied to different climate 18 zones, soils, land uses and field management practices. This study is one of many efforts to fulfil such an expectation, which 19 was performed to improve the CNMM-DNDC by incorporating a physical-based soil thermal module to simulate the soil 20 thermal regime in the presence of freeze-thaw cycles. The modified model was validated with simultaneous field 21 observations in three typical alpine ecosystems (wetlands, meadows and forests) within a catchment located in the seasonally 22 frozen region of the eastern Tibetan Plateau, including soil profile temperature, topsoil moisture and fluxes of methane (CH4) 23 and nitrous oxide (N2O). The validation showed that the modified CNMM-DNDC was able to simulate the observed 24 seasonal dynamics and magnitudes of above variables in the three typical alpine ecosystems, with index of agreement values 25 of 0.91‒1.00, 0.49‒0.83, 0.57‒0.88 and 0.26‒0.47, respectively. Consistent with the emissions determined from the field 26 observations, the simulated aggregate emissions of CH4 and N2O were highest for the wetland among three alpine 27 ecosystems, which were dominated by the CH4 emissions. This study indicates the possibility for utilizing the process28 oriented model CNMM-DNDC to predict hydro-biogeochemical processes, as well as related gas emissions, in seasonally 29 frozen regions. As the original CNMM-DNDC was previously validated in some unfrozen regions, the modified CNMM30 DNDC could be potentially applied to estimate the emissions of CH4 and N2O from various ecosystems under different 31 climate zones at the site or catchment scale. 32

physical-based soil thermal module to simulate the soil thermal regime in the presence of freeze-thaw cycles. The modified 23 model was validated with simultaneous field observations in three typical alpine ecosystems (wetlands, meadows and forests) 24 within a catchment located in the seasonally frozen region of the eastern Tibetan Plateau. Then, the model was further 25 applied to evaluate its performance in simulating the effects of alpine wetland degradation on methane (CH 4 ) and nitrous 26 oxide (N 2 O) fluxes. The validation showed that the modified CNMM-DNDC was able to simulate the observed seasonal 27 dynamics of soil temperature, moisture, and fluxes of CH 4 and N 2 O in the three typical alpine ecosystems, with index of 28 agreement values of 0.91-1.00, 0.49-0.83, 0.57-0.88 and 0.26-0.47, respectively. Consistent with the emissions determined 29 from the field observations, the simulated aggregate emissions of CH 4 and N 2 O were significantly reduced due to wetland 30 degradation and were dominated by a reduction in CH 4 emissions. This study indicates the potential for utilizing the process-31 oriented model CNMM-DNDC to predict hydro-biogeochemical processes, as well as related gas emissions, in seasonally 32 frozen regions. As the original CNMM-DNDC was previously validated in some unfrozen regions, the modified CNMM-33 https://doi.org/10.5194/bg-2020-433 Preprint. Discussion started: 13 January 2021 c Author(s) 2021. CC BY 4.0 License.

Introduction 36
The elements of nitrogen and carbon are essential components of ecosystems (e.g., Breuer et al., 2010;Canfield et al., 37 2010). Climate changes due to warming and human anthropogenic activities derived from food production have significantly 38 altered the cycling of nitrogen and carbon and led to increased reactive nitrogen availability and carbon losses, which result 39 in a series of environmental problems at the catchment, regional and even global scales (e.g., Galloway et al., 2004;40 Galloway et al., 2008;Ju et al., 2009). Excessive reactive nitrogen in soils can be lost in the forms of nitrogen gases, such as 41 nitrous oxide (N 2 O), nitric oxide (NO) and ammonia (NH 3 ), and nitrogen pollution, such as nitrate (NO 3 -) and ammonium 42 (NH 4 + ), in water through leaching or surface runoff (e.g., Seitzinger, 2008;Collins et al., 2016). In the face of increased air 43 temperatures and intensive land use changes, especially in cold regions, the soil organic carbon stored during long periods 44 has been lost to the atmosphere via methane (CH 4 ) and carbon dioxide (CO 2 ) (e.g., Piao et al., 2009;Fenner and Freeman, 45 2011; Schuur et al., 2015). These nitrogen and carbon losses contribute to potential global warming (CO 2 , CH 4 and N 2 O), air 46 pollution (NO and NH 3 ) and surface/groundwater pollution (NO 3 and NH 4 + ). Therefore, sustainable ecosystems urgently 47 need to be established that not only focus on net primary productivity but also are friendly to the environment with the 48 minimal hazards, including greenhouse gases, air pollution and water pollutants (e.g., Cui et al., 2018;. 49 The cycling of nitrogen and carbon is closely related to soil water processes (e.g., Breuer et al., 2010;Vereecken et al., 50 2016; . Thus, interactions among soil waters and the cycling of nitrogen and carbon govern biological 51 productivity and environmental outcomes (e.g., Zhu et al., 2018). The interactions consist of the redox potential for different 52 transformation processes influenced by the spatiotemporal variation in soil water content and the lateral transport of water 53 and dissolved nitrogen or carbon controlled by surface and subsurface flow (e.g., McClain et al., 2003;Castellano et al., 54 2013;Bechmann, 2014). For example, the variation in soil water content can create hot spots or moments of nitrogen and 55 carbon losses by influencing plant nitrogen uptake, redox potential, and the transport of dissolved nitrogen and carbon (e.g., 56 Zhu et al., 2012;Keiluweit et al., 2017). Therefore, a complete understanding of biogeochemical processes will inevitably 57 involve interactions among soil water and the cycling of nitrogen and carbon (e.g., Breuer et al., 2010;Vereecken et al., 58 2016;Zhu et al., 2018). 59 Biogeochemical models, such DNDC, WNMM, CENTURY and DayCent, are effective tools for simulating the 60 cycling of nitrogen and carbon and quantifying the effects of climate change and human anthropogenic activities on 61 ecosystems (e.g., Foereid et al., 2007;Li, 2007;Li et al., 2007;Cheng et al., 2014). However, comprehensive hydrological 62 processes, especially for the lateral transport of water and nutrients, are generally simplified or ignored in these models due 63 to specific questions that must be addressed (e.g., Li, 2007;Li et al., 2007;Chen et al., 2008;Deng et al., 2014). On the 64 other hand, land surface or hydrological models at large scales, which are designed with explicit mechanisms of hydrology, 65 https://doi.org/10.5194/bg-2020-433 Preprint. Discussion started: 13 January 2021 c Author(s) 2021. CC BY 4.0 License. generally focus on vertical and lateral nutrient transport, such as nitrate loads into rivers (e.g., Liu et al., 2019). However, the 66 simulations of nitrogen and carbon processes are usually based on empirical functions even without predicting gas loss. Due 67 to the various purposes of different models, coupling soil hydrological models with biogeochemical models can be an 68 effective strategy for integrating soil water and cycling of nitrogen and carbon to improve model performance. Thus, the 69 coupled model with improved performance can be applied to evaluate the sustainability of natural or agricultural ecosystems, 70 simultaneously predicting productivity and potential negative environmental effects (e.g., Zhu et al., 71 2018 DeNitrification-DeComposition (CNMM-DNDC), which was established by incorporating the core biogeochemical 79 processes of DNDC into the hydrological framework of the CNMM, was validated at a catchment with complex landscapes 80 in the subtropical region and showed good performance for simultaneously simulating various variables, including 81 ecosystem productivity, hydrological nitrogen losses and nitrate discharge in streams, and emissions of gaseous carbon and 82 nitrogenous gases . Therefore, the CNMM-DNDC has the capacity to evaluate the sustainability of 83 ecosystems with simultaneous simulations of various variables closely related to both productivity and environmental 84 hazards. 85 However, as a process-oriented hydro-biogeochemical model designed to be applicable to different climate zones, soils, 86 land uses and field management practices, CNMM-DNDC testing is still lacking due to limited observations for model 87 validation. In this study, the model was applied to a catchment in seasonally frozen regions located on the eastern Tibetan 88 Plateau (TP) with the land use types of alpine wetlands, meadows and forests to test its ability to simulate hydro-89 biogeochemical processes. However, scientific descriptions of soil thermal dynamics due to freeze-thaw cycles are still 90 lacking for the CNMM-DNDC. This gap may hinder model application in seasonally frozen regions, which account for 56% 91 of the exposed land surface of the Northern Hemisphere (Jiang et al., 2020). In addition, the soil freeze-thaw cycles that 92 occur in these mid-high latitude regions exert important influences on soil thermal dynamics, as well as on related 93 hydrological processes, thus increasing the availability of substrates and stimulating the processes of CH 4 and N 2 O 94 production and emissions in soils (e.g., Song et al., 2019). Therefore, we hypothesize that adding the missing scientific 95 processes of soil thermal dynamics into the internal model program codes can improve the performance of the CNMM-96 DNDC in simulating the soil thermal dynamics, hydrological processes and CH 4 and N 2 O fluxes in seasonally frozen regions. 97 Filling this gap is especially necessary to broaden model applicability. The thermal dynamics of the soil and snow were calculated by the one-dimensional heat conduction equation (Eq. 1), 141 which was solved numerically using Eqs. 2-4. In the above equations, C (J m -3 °C -1 ), k (W m -1 °C -1 ), T (°C) and G (W m -2 ) 142 denote the soil heat capacity, thermal conductivity, soil temperature and heat fluxes between layers, respectively. Both Z and 143 D are the thicknesses of the soil layer (m), △t is the time step of the calculation, and l denotes the soil layer l. S is the internal 144 heat exchange due to freezing or thawing (W m -3 ) when the soil temperature is around 0 °C. The soil temperature changes 145 affected by freezing or thawing were determined on the basis of energy conservation, which indicated that the latent heat 146 released during freezing equalled the amount of heat required for the increased soil temperature and vice versa. The soil heat 147 capacity (C, J m -3 °C -1 ) is the weighted average of five constituents in the volumetric fraction (θ), including organic matter 148 (2.5×10 6 ), minerals (2.0×10 6 ), water (4.2×10 6 ), ice (2.1×10 6 ) and air (1.2×10 3 ) (Eq. 5). The thermal conductivity (k, W m -149 1 °C -1 ) is the geometric mean of the thermal conductivities of the above five constituents (Eq. 6), with values of 0.25, 2.9, 150 0.57, 2.2 and 0.025 W m -1 °C -1 for organic matter, minerals, water, ice and air, respectively. The upper and lower boundary 151 conditions of the thermal dynamics were determined by the surface energy balance and the defined geothermal heat flux at a 152 soil depth of 35 m. 153 (4) 2019). A value of IA (0-1) closer to 1 showed a better simulation ranging from 0 to 1. An NSI value (ranging from minus 199 infinity to 1) closer to 1 was better. Better model performance was indicated by a slope and an R 2 value that were both closer 200 to 1 in a significant ZIR. For more details on these criteria, refer to the online supplementary materials (Eqs. S1−4 in Table  201 S2). In addition, the SPSS Statistics Client 19.0 (SPSS Inc., Chicago, USA) and Origin 8.0 (OriginLab, Northampton, MA, 202 USA) software packages were applied for the statistical analysis and graphical comparison.  Table 1). However, the CH 4 uptake rates during the 232 dormant season were obviously underestimated by the modified model at both sites, especially at the alpine forest site, which 233 was responsible for the underestimation of cumulative CH 4 uptake. The observed cumulative CH 4 emissions ranged from -234 2.60 to 33.5 kg C ha −1 yr −1 and the modelled values ranged from -1.90 to 31.0 kg C ha −1 yr −1 (Fig. 6a). These results indicate 235 that the modified CNMM-DNDC successfully simulated the CH 4 fluxes of the three typical alpine ecosystems at the 236 catchment scale and showed the capacity to predict the effects of wetland degradation on CH 4 emissions. 237

Nitrous oxide fluxes 238
The daily observed N 2 O emissions from the alpine wetlands were higher than those from the alpine meadows but 239 lower than those from the alpine forests (Figs. 3b, 4b and 5b). for the alpine wetlands. In addition, compared with the original model, the modified model captured the peak emissions that 247 occurred during the freeze-thaw period from the alpine meadows due to the death of microbes, but the dynamics of the peak 248 emissions were not well simulated. The observed cumulative N 2 O emissions ranged from 0.14 to 0.58 kg N ha −1 yr −1 and the 249 modelled values ranged from 0.12 to 0.32 kg N ha −1 yr −1 (Fig. 6b). These results indicate that the modified CNMM-DNDC 250 showed the potential to estimate N 2 O emissions in seasonally frozen regions and thus was able to assess the influences of 251 wetland degradation on N 2 O emissions. The simulated aggregate emissions by the modified model were 1.5, 0.015, and 0.061 Mg CO 2 eq ha −1 yr −1 for the observed 257 alpine wetlands, meadows and forests, respectively, which were consistent with those from the field observations (1.6, 0.014, 258 and 0.15 Mg CO 2 eq ha −1 yr −1 for the alpine wetlands, meadows and forests, respectively) (Fig. 6c) caused the upper soil layer to act as an efficient oxidative methanotrophic barrier for the diffusion of CH 4 from the subsoil 296 and thus decreased CH 4 emissions (Kandel et al., 2018;Tan et al., 2020). In addition, the highly fluctuating CH 4 emissions 297 simulated by the modified model were also attributed to the high dependency of CH 4 production on soil moisture, which 298 controlled the size of the CH 4 balloon. Theoretically, the CH 4 emissions simulated by the original model should not have 299 been higher than those simulated by the modified model due to the lower predicted soil moisture level. The overestimated 300 CH 4 emissions simulated by the original model were mainly attributed to the overestimated soil temperature due to their 301 influences on mineralized substrates for CH 4 production, as well as the processes of CH 4 production. This result implies that 302 global warming may trigger intensive CH 4 emissions from degraded wetlands, which could partly serve as a trade-off for the 303 decreased CH 4 emissions due to the lower water table level in degraded wetlands. Both observations and simulations showed 304 that the CH 4 uptake in alpine forests was higher than that in alpine meadows, which was mainly attributed to the high SOC 305 content and low soil clay fraction of the alpine forests in the simulation. Methane uptake by upland soils is a biological 306 process governed by the availability of CH 4 and oxygen as well as the activity and quantity of methanotrophic bacteria in 307 soils (e.g., Liu et al., 2007;Zhang et al., 2014). In the model, the simulated CH 4 uptake was positively related to the SOC 308 content, which is closely related to the population size of methanotrophic bacteria. On the other hand, soil permeability 309 affects gas diffusion into the soil and thus CH 4 uptake (e.g., Liu et al., 2007). For the CNMM-DNDC, the clay fraction, which is regarded as a key factor regulating soil permeability, showed a negative relationship with CH 4 uptake in the model. 311 Thus, the SOC content, as well as the soil clay fraction, contributed to the differences in CH 4 uptake from alpine meadows 312 and forests. As the simulated dynamic characteristics of CH 4 uptake were primarily regulated by soil temperature and 313 moisture, the effects of low soil temperature (< 0.0 °C) on CH 4 uptake rates resulted in obvious underestimations in the 314 dormant season for both alpine meadows and forests. Therefore, an improved parameterization for simulating CH 4 uptake 315 under low soil temperatures is required for the model to better capture the dynamics of CH 4 uptake in the dormant season. threshold values of soil temperature to trigger the decomposition of microbes during the freezing period and stimulate the 337 production of NO, N 2 O and N 2 using substrates derived from microbial decomposition during the thawing period. However, 338 the dynamics of peak emissions due to freeze-thaw cycles were inconsistent with those from the field observations. Thus, 339 improvements are required to optimize the parameterization scheme to better capture the dynamic characteristics. In addition, 340 the peak emissions during the freeze-thaw period were not captured by the original model due to the significantly 341 overestimated soil temperature. The low evaluation statistics for the daily fluxes, especially for the alpine forests, were also 342 attributed to the underestimation of background emissions, which resulted from both measurement errors due to low fluxes 343 https://doi.org/10.5194/bg-2020-433 Preprint. Discussion started: 13 January 2021 c Author(s) 2021. CC BY 4.0 License.
around detection limits and model deficiencies in the simulation of tight nitrogen cycling in natural ecosystems. The model 344 performances of simulating various variables for three typical alpine ecosystems in the Rierlangshan catchment imply that 345 the modified CNMM-DNDC can be applied to predict the thermal dynamics, hydrology, nitrogen and carbon cycling and 346 related greenhouse gas emissions in seasonally frozen regions. 347

Implications for degraded alpine ecosystems 348
The typical natural wetland alpine ecosystems, which are annually inundated, act as greenhouse gas sinks or are 349 neutral (e.g., Cai., 2012;Tan et al., 2020). A previous study showed that more than 90% of the annually inundated wetlands 350 on the TP have been degraded and become seasonally inundated or wet alpine meadows due to intentional drainage for 351 grazing since the 1960s (Wei et al., 2015). Both the observations and simulations showed that in comparison to annually 352 inundated wetlands, wetland degradation induced by drainage stimulated N 2 O emissions to a small extent but reduced CH 4 353 emissions to a large extent. Thus, compared to that from natural wetlands, the aggregate emissions of CH 4 and N 2 O from 354 degraded wetlands were largely reduced but still higher than those of adjacent wet alpine meadows. These results were 355 consistent with the field observations of CH 4 and N 2 O emissions along different water table transects in the Zoige peatland, 356 which were primarily driven by soil water content and SOC . The decline in the water table induced by 357 intentional drainage resulted in recessive succession of the vegetation for the Zoige wetland with a typical mode of marsh, 358 marsh meadow and meadow (Xiang et al., 2009). Thus, one may deduce that the degradation of annually inundated wetlands 359 at a large-scale might have greatly reduced the aggregate emissions of CH 4 and N 2 O from the Zoige wetland, especially for 360 CH 4 . However, a recent meta-analysis showed that the reduction in the aggregate emissions of CH 4 and N 2 O due to draining 361 may be completely offset by the decreased net CO 2 uptake (Tan et al., 2020). For natural wetlands, anaerobic conditions 362 under high a water table inhibit litter decomposition, and thus, a large amount of organic matter is sequestered (Nahlik and 363 Fennessy, 2016). When the annually inundated wetlands degrade to seasonally inundated wetlands or meadows, the rate of 364 peat soil oxidation is enhanced, thus significantly increasing ecosystem respiration and resulting in a shift from net sinks of 365 greenhouse gas emissions to notable sources (Tan et al., 2020). Consistent with the results of the meta-analysis, the 366 simulation showed that the loss rate of SOC (consisting of microbes, humads and humus) was much higher for degraded 367 wetlands than for other typical alpine ecosystems. These results also indicate the large risk of soil carbon loss due to 368 intentional drainage, which has been sequestered for a long time. The simulation by the modified CNMM-DNDC showed 369 that the model has the capacity to simulate hydro-biogeochemical processes in seasonally frozen regions for various alpine 370 ecosystems. implies that hydro-biogeochemical model, such as the modified CNMM-DNDC, are able to predict soil thermal dynamics 387 and cycling of nitrogen and carbon in seasonally frozen regions with an improved physical-based soil thermal module. 388

Data availability 389
The model, input and output databases can be obtained from the first author and all the observed data sets used in this study 390 can be available from the co-authors. 391

Author contribution 392
Zheng, X. and Zhang, W. contributed to developing the idea and enhancing the science of this study. Zhang