Simulated effects of nitrogen saturation on the global carbon budget using the IBIS model

Over the past 100 years, human activity has greatly changed the rate of atmospheric N (nitrogen) deposition in terrestrial ecosystems, resulting in N saturation in some regions of the world. The contribution of N saturation to the global carbon budget remains uncertain due to the complicated nature of C-N (carbon-nitrogen) interactions and diverse geography. Although N deposition is included in most terrestrial ecosystem models, the effect of N saturation is frequently overlooked. In this study, the IBIS (Integrated BIosphere Simulator) was used to simulate the global-scale effects of N saturation during the period 1961–2009. The results of this model indicate that N saturation reduced global NPP (Net Primary Productivity) and NEP (Net Ecosystem Productivity) by 0.26 and 0.03 Pg C yr−1, respectively. The negative effects of N saturation on carbon sequestration occurred primarily in temperate forests and grasslands. In response to elevated CO2 levels, global N turnover slowed due to increased biomass growth, resulting in a decline in soil mineral N. These changes in N cycling reduced the impact of N saturation on the global carbon budget. However, elevated N deposition in certain regions may further alter N saturation and C-N coupling.


SI 1 N module in the IBIS model
The original IBIS (Integrated BIosphere Simulator) 1 has a very simple N (nitrogen) control on the NPP (Net Primary Productivity) calculation through a constant leaf N level. Liu et al. 2

incorporated
a largely complete N cycle module into IBIS that includes the dynamic leaf N level and N controls on C (carbon) assimilation, C allocation and the C-N (Carbon-Nitrogen) cycle in soil 2 .
Vegetation photosynthesis is characterized by the Farquhar equation in the IBIS model, and the maximum photosynthetic velocity is controlled by leaf-available N (equation (1)).
is the optimal C:N ratio for foliage, i.e., the ratio at which the maximum photosynthetic rate ( max V ) occurs. L B is the actual foliar C:N ratio, and m V is the actual photosynthetic rate. If an N shortage exists, the actual foliar C:N ratio ( L B ) will increase and max V will decline. In the IBIS N is the available N in the soil, and Mmax N is the maximum available N in the soil, which was set to 2 g m -2 in the study of Liu et al.
NPP is the daily NPP, and g R is the growth respiration ratio. p K is used to adjust the effect of respiration on NPP. When the soil N level is low, even leaf photosynthesis permits a high NPP, and the actual NPP will be low due to modification by p K . p K is also used to regulate the carbon allocation of vegetation (equation (4)). In N-limited ecosystems, is the maximum available mineral N in the soil (2 g N m -2 ) that allows N limitation to occur. M N is the actual available mineral N in the soil. A comparison of the simulated annual GPP from IBIS and MTE is presented in Figure S1. The multiple-year average IBIS GPP is similar to the MTE GPP, although the trends in the IBIS GPP and MTE GPP differ after 1995. It is generally agreed that the MTE should not be used as a benchmark for GPP trends 6 . One reason is that the CO 2 fertilization effect was not considered in the MTE. Another reason is that the flux tower sites used in the MTE are mainly distributed in northern temperate regions, whereas tropical ecosystems largely drive the inter-annual variability in the C cycle 7 . Other modelled GPP trends have been summarized by Anav et al. 6 The inter-annual trend in GPP ranged from 0.28 to 0.62 Pg C yr -2 between 1990 and 2010, which is comparable to the trend in the IBIS GPP (i.e., 0.39 Pg C yr -2 ) during the same period (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)). Thus, the IBIS GPP is similar to previous findings in terms of both magnitude and trend.

SI. 2.1 Validation of GPP
The spatial differences between the IBIS GPP and MTE GPP are shown in Figure S2.   We also compared our GPP with the GPP values obtained by Beer et al. 4 and Saugier et al. 9 in different biomes (Table S1). The distribution of global biomes is based on the work of Roy et al. 10 and Griggs and Noguer 11 . In addition, cropland has been incorporated into global biomes, which is extracted from the MODIS (MODerate resolution Imaging Spectroradiometer) land cover product.
The IBIS-simulated forest GPP value is higher than Beer's value but similar to Saugier's value.
Compared with Beer's GPP, the IBIS GPP is higher in tropical and temperate forests. However, in tropical savanna and grasslands, the IBIS GPP is lower than in previous studies. In other biomes, the IBIS GPP results are similar to those of other studies.

SI. 2.2 Validation of the NPP
We compiled recent global terrestrial NPP results covering the past 30 years ( In addition to using data from published papers to validate the global total NPP, we used the MODIS NPP product to examine the performance of the IBIS N-saturation module. Without considering N saturation, IBIS overestimates the NPP compared with the MODIS NPP. However, with the Nsaturation module, the IBIS NPP is more consistent with the MODIS NPP ( Figure S3).

SI. 2.3 Validation of the NEP
In this study, we collected global NEP (Net Ecosystem Productivity) results evaluated by different methods to validate our simulated NEP (Table S3). The IBIS-simulated NEP average was found to be 2.5 Pg C yr -1 from 1980 to 2000, which is within a reasonable range. A previous study also showed that the NEP result of a C-N coupling model is similar to that of a C model, although the spatial distribution of NEP and the sensitivity of the terrestrial C balance to its driving factors are substantially altered by N dynamics 21 .

SI. 2.4 Validation of the N-cycling module
To validate the N control on C cycling, we used the observations of aboveground NPP responses to N addition in forests and grasslands. For forests, Thomas et al. 19 reviewed 40 N-addition experiments in which the N-addition rates ranged from 0.9 to 15.0 g N m -2 yr -1 over 2-30 years 19 .  19 The modelled aboveground NPP (ANPP) was calculated for the same time frame as the field study, and the N addition began after 1985.
The N control of C cycling has been studied for forests but not for grassland areas according to a comparison with observation results. The rate of forest ANPP increased by 23±9% on average in response to N fertilization in the field experiments. IBIS-simulated ANPP increased by 18±13%, which is similar to the empirical results ( Figure S4b). However, in grassland areas, the rate of ANPP increase observed at field sites averaged 62±15%, whereas the IBIS-simulated ANPP increase was 35±25% ( Figure S4c).  Figure S5. Historical NPP and NEP changes.