The impact of nitrogen and phosphorous limitation on the estimated terrestrial carbon balance and warming of land use change over the last 156 yr

We examine the impact of land use and land cover change (LULCC) over the period from 1850 to 2005 using an Earth system model that incorporates nitrogen and phosphorous limitation on the terrestrial carbon cycle. We compare the estimated CO 2 emissions and warming from land use change in a carbon-only version of the model with those from simulations, including nitrogen and phosphorous limitation. If we omit nutrients, our results suggest LULCC cools on the global average by about 0.1 C. Including nutrients reduces this cooling to∼ 0.05C. Our results also suggest LULCC has a major impact on total land carbon over the period 1850–2005. In carbon-only simulations, the inclusion of LULCC decreases the total additional land carbon stored in 2005 from around 210 Pg C to 85 Pg C. Including nitrogen and phosphorous limitation also decreases the scale of the terrestrial carbon sink to 80 Pg C. Shown as corresponding fluxes, adding LULCC on top of the nutrient-limited simulations changes the sign of the terrestrial carbon flux from a sink to a source (12 Pg C). The CO 2 emission from LULCC from 1850 to 2005 is estimated to be 130 Pg C for carbon only simulation, or 97 Pg C if nutrient limitation is accounted for in our model. The difference between these two estimates of CO2 emissions from LULCC largely results from the weaker response of photosynthesis to increased CO 2 and smaller carbon pool sizes, and therefore lower carbon loss from plant and wood product carbon pools under nutrient limitation. We suggest that nutrient limitation should be accounted for in simulating the effects of LULCC on the past climate and on the past and future carbon budget.


Geoscientific Instrumentation
Methods and Data Systems The Cryosphere Discussions This discussion paper is/has been under review for the journal Earth System Dynamics (ESD). Please refer to the corresponding final paper in ESD if available.

The impact of nitrogen and phosphorous limitation on the estimated terrestrial carbon balance and warming of land use change over the last 156 yr
in 2005 from around 210 Pg C to 85 Pg C. Including nitrogen and phosphorous limitation also decreases the scale of the terrestrial carbon sink to 80 Pg C. In particular, adding LULCC on top of the nutrient limited simulations changes the sign of the terrestrial carbon flux from a sink to a source (12 Pg C). The CO 2 emission from LULCC from 1850 to 2005 is estimated to be 130 Pg C for carbon only simulation, or 97 Pg C

Introduction
Human activity has modified 42-68 % of the terrestrial surface via deforestation, reforestation, clearing for crops, pasture and urban settlements (Hurtt et al., 2006). Land use and land cover change (LULCC) is concentrated in regions including eastern North tent heat fluxes (Bonan, 2008;de Noblet-Ducoudré et al., 2012;Boisier et al., 2012), which in turn can affect air temperature and the larger scale climate (Feddema et al., 2005;Findell et al., 2007Findell et al., , 2009Pitman et al., 2009;de Noblet-Ducoudré et al., 2012). In addition to the biogeophysical impacts of LULCC, changing the nature of the surface also has a major impact on terrestrial biogeochemical cycles (Arneth et al., 2010;15 Levis, 2010; Houghton et al., 2012). If forests are replaced by crops or pasture, the soil carbon is reduced by 25-30 % as a result of cultivation (Houghton and Goodale, 2004). The effect of ecosystem carbon balance will depend on the total ecosystem carbon before the land use change occurs, net primary productivity (NPP) of the crop or pasture and the rate of ecosystem carbon change after land use change. In addition, increases 20 in atmospheric CO 2 likely stimulates photosynthesis (Field et al., 1995) although nutrient limitation by nitrogen (N) and phosphorous (P) moderate this fertilization effect (Vitousek et al., 2010). The interactions between CO 2 -induced climate change and the terrestrial carbon balance, and the feedbacks associated with the response by the surface via CO 2 emissions to climate change is extremely complex and uncertain 25 (Friedlingstein et al., 2006) and LULCC is superimposed onto these interactions. As noted by Arneth et al. (2010), examining how LULCC interacts with biogeochemical cycling is a research priority. The biogeochemical effects of LULCC in terms of land use emissions have been investigated previously within several climate models (e. g. Pongratz et al., 2009Pongratz et al., , 2011Shevliakova et al., 2009). Carbon emissions from LULCC dampen biogeophysical cooling in some studies (Brovkin et al., 2004;Bala et al., 2007). In other studies, LULCC induced cooling can be changed to warming once the terrestrial carbon feedback is 5 included (Sitch et al., 2005;Pongratz et al., 2010). Recent studies in Global Carbon Project suggest that LULCC nearly offsets the entire land sink from reforestation and CO 2 fertilization since the pre-industrial period (Canadell et al., 2007;le Quéré et al., 2009). There are, however, large uncertainties in the magnitude of carbon loss linked to LULCC (Denman et al., 2007). Estimates of the scale of CO 2 emission from LULCC 10 between 1850 and 2000 vary from 44 to 150 Pg C (Houghton, 2008;Arora and Boer, 2010). A recent inter-comparison study reported carbon emissions due to LULCC for the 1990s had a range of 0.75-1.50 Pg C yr −1 , with a median value of 1.1 Pg C yr −1 based on 13 model estimates (Houghton et al., 2012). One weakness of existing studies of the impact of LULCC on biogeochemical cycles 15 is the lack of the inclusion of nutrients. N limitation reduces the net carbon uptake by the global land biosphere by 37 % to 74 % from the preindustrial through to 2100 in some modeling studies (Thornton et al., 2007;Sokolov et al., 2008;Zaehle et al., 2010

Model description
We used the CSIRO Mk3L  coupled with a land surface model including carbon, nitrogen and phosphorous cycles, CABLE (Wang et al., 2010. Mk3L is a relatively low-resolution but computationally efficient general circulation 5 model developed for studies of climate on centennial to millennial time scales . The atmospheric component has a horizontal resolution of 5.6 • by 3.2 • and 18 vertical levels. CABLE performs well in comparison to other land surface models (LSMs) in simulating latent and sensible heat as well as CO 2 fluxes at the site scale (Abramowitz et al., 2007(Abramowitz et al., , 2008Wang et al., 2011). An earlier version was used 10 in the Land Use Change IDentification of robust impacts (LUCID) project (Pitman et al., 2009;de Noblet-Ducoudre et al., 2012). Mao et al. (2011) documents the performance of Mk3L coupled to CABLE, which provides strong evidence that the coupled model produces a reasonable large-scale climatology. The version of CABLE used here includes the biogeochemical model CASA-CNP (Wang et al., 2010). CASA-CNP 15 simulates dynamics of carbon, nitrogen and phosphorus in plant and soil. The coupled Earth System Model has recently been used to explore the dependence of terrestrial carbon uptake due to N and P limitation through the 20th century without land cover change (Zhang et al., 2011). 20 The interpretation of the Coupled Model Intercomparison Project (CMIP-5, Taylor et al., 2012) land cover trajectories by Lawrence et al. (2012) for the period 1850-2005 is used to provide the change in area fractions of different PFT within a land cell as a function of time compatible for CABLE and CASA-CNP. Figure 1 shows the pattern of changes in (a) forests; (b) grass and (c) crops. In general, a pattern of forest reduction is Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | also been reduced in similar regions; though note an increase in grasslands in eastern South America coincident with decrease in forests. In most regions of the Northern Hemisphere, the forest removal has resulted in increased croplands (Fig. 1c). The scale of these changes, according to the CMIP-5 experimental protocols are a reduction in total forest area from about 54 × 10 6 km 2 in 1850 to about 47 × 10 6 km 2 in 2005.
In this study, the atmosphere model was forced by CO 2 from CMIP-5 database for 1850-2005 (Meinshausen et al., 2011). The ocean is prescribed using monthly sea 10 surface temperatures (SSTs) simulated by CSIRO-Mk3.6 (Rotstayn et al., 2010(Rotstayn et al., , 2012 for the CMIP-5 experiments associated with the same CMIP-5 CO 2 for the same period. To evaluate the effects of land use change on terrestrial carbon balance, we under-20 took two sets of experiments started from the same initial states. In the "LUC" experiment, the model was run using the land cover change and CO 2 data. The land model was set up to run carbon cycle only (LUC-C) and carbon, nitrogen and phosphorous cycle (LUC-CNP) cases for each LULCC ensemble simulations. In the "CTL" experiment, the same simulations were run except the vegetation distribution was kept constant as

Analysis of CO 2 emission from land use change
Total carbon on land (c L ) comprises three carbon pools in vegetation (c V ), litter (c L ) and soil (c S ). For land points that never experience LULCC over the simulation period [0, t], the budget equations of carbon in each of these three pools are where f GPP is gross primary production in g C m −2 yr −1 , r V , r L and r S are the respired 10 CO 2 from vegetation, litter and soil carbon pools in g C m −2 yr −1 , respectively; f L and f S are carbon fluxes from vegetation to litter, and from the litter to the soil pool in g C m −2 yr −1 , respectively. Land use change can affect total carbon pool sizes directly and indirectly. Land use change can affect climate through the biogeophysical effects and biogeochemical ef-15 fects, and the changed climate will then impact all the fluxes on the right-hand sizes of Eqs. (1) and (3). Since atmospheric CO 2 concentration from 1850 to 2005 is prescribed as an input to our model in this study, only the biogeophysical effect is taken into account here. The dynamic equations for carbon in vegetation, litter and soil of a land point under land use change over the study period are given by The flux terms with star as superscript in Eqs. (4) to (6) Similar to Shevliakova et al. (2009) we partitioned c * P equally into three anthropogenic pools characterized by their turnover rates: fuel wood (1 yr −1 ), paper and paper prod-15 ucts (0.1 yr −1 ) and wood products (0.01 yr −1 ).
is the associated CO 2 flux released from consumption of anthropogenic pools. CO 2 emission from land use change, F LUC is calculated as Net ecosystem exchange (NEE) of a land point is calculated as 10 3 Results

Impacts of nutrient limitation on terrestrial carbon
At the global scale, simulations using the C-only mode and omitting LULCC show a strong terrestrial sink of CO 2 . The magnitude of this sink exceeds 200 Pg C between 1850 and 2005 (Fig. 3). Adding N and P limitation reduces this sink to 85 Pg C, a result 15 consistent with other studies that demonstrate N-limitation strongly reduces terrestrial carbon sinks (Thornton et al., 2007;Sokolov et al., 2008;Zaehle et al., 2010). Adding LULCC has a major impact on terrestrial carbon stores. Simulations using the C-only 515 Introduction mode, but including LULCC, also simulate a net carbon sink (80 Pg C between 1850 and 2005). The sink is negligible from 1850 through to about 1960, and increases rapidly under accelerating atmospheric CO 2 concentrations through to 2005. The N and P limitation reduces the capacity of the terrestrial biosphere to take up CO 2 such that in the LUC-CNP simulations the land is a weak source of CO 2 because emis-5 sions from land cover change are not fully offset by land carbon uptake in response to increased atmospheric CO 2 (Fig. 3). Further, the acceleration in the terrestrial sink shown in the C-only LULCC simulation is largely suppressed in the CNP simulation with LULCC. Over the period 1850 and 2005, the LUC-CNP simulation therefore remains a source for CO 2 with a magnitude of 12 Pg C. That is, N and P limitation changes the 10 terrestrial surface from a sink of CO 2 to a net source over the period 1850 to 2005. The CO 2 emissions from LULCC can be estimated as the difference in pool size changes from 1850 to 2005 between the simulations with and without LULCC. With C-only, the plant biomass carbon was reduced by 104.6 Pg C, litter and soil carbon by 45.5 Pg C but wood product pool was increased by 20.7 Pg C between the sim-15 ulations with and without LULCC. The total CO 2 emission from LULCC was therefore 129.6 Pg C (Fig. 4). Most of the CO 2 emitted from LULCC was from the increased heterotrophic respiration (0.56 Pg C yr −1 ) and consumption from wood products (0.48 Pg C yr −1 ) offset slightly by an increase in gross primary production (GPP) of 0.15 Pg C yr −1 for 1850-2005. 20 For the CNP simulation, LULCC resulted in a decrease in the plant, litter and soil carbon pools, and a small increase in wood product carbon pool (Fig. 5). As a result, the total land carbon pool including wood product carbon decreased by 12 Pg C from 1850 to 2005. In simulations without LULCC, plant biomass, litter and soil carbon pools increased from 1850 to 2005. Therefore the total CO 2 emission from LULCC Imposing the N and P limitation in our model reduced the estimated CO 2 emission from LULCC by 32.9 Pg C from 1850 to 2005. Most of this difference can be accounted by the effect of nutrient limitation on the contribution of vegetation biomass change.

5
LULCC increased plant biomass slightly by increased NPP for both C-only and CNP simulations, and this increase was reduced by N and P limitation. On the other hand LULCC also increased the amount of carbon transferred to litter, soil and wood product pools, and thereby reduced the plant biomass carbon. The reduction in plant carbon under N and P limitation was 25 Pg C less than under C-only simulation. N and P 10 limitation also reduced the CO 2 emission from litter, soil and wood product pools due to LULCC because all simulated pool sizes and fluxes under N and P limitation were much smaller than those under C-only simulations. However, the magnitude of changes in these pools was much less than the change in vegetation biomass pool.
The geographical pattern of changes in NEE with and without LULCC are shown in 15 Fig. 6. These patterns include a climate signal associated with the increase in CO 2 between 1850 and 2005, and a CO 2 fertilization effect, as well as any impact from LULCC. The changes shown in Fig. 6 should therefore not be interpreted as simply a LULCC signal. The combined impact in the C-only simulations (Fig. 6a) includes decreases in NEE of ∼ 20-30 g C m −2 yr −1 over Europe, parts of SE Asia, eastern North 20 America, isolated parts of South America and Africa coincident with LULCC. Increases in NEE occur over North America, Eurasia, parts of South America and Africa of ∼ 10-30 g C m −2 yr −1 . Figure 6c shows the results from simulations excluding LULCC but with the same CO 2 forcing and any associated fertilization effect as used in Fig. 6a.
Here, in the C-only simulations, NEE increases over most of the vegetated surfaces by 25 10-30 g C m −2 yr −1 (Fig. 6c).
In both LUC (Fig. 6b) and CTL (Fig. 6d) simulations, the addition of N and P limitation moderates the impact of climate and elevated CO 2 on NEE. In the LUC experiments including N and P limitation (Fig. 6b)  the areas of decreased NEE become more clearly associated with LULCC particularly over Europe and S.E. Asia. There are still areas of increased NEE over South America and central Africa, but the magnitude has decreased from ∼ 20-30 g C m −2 yr −1 (Fig. 6a) to ∼ 10 g C m −2 yr −1 (Fig. 6c). Similarly, in the CTL simulations, the magnitude of the increase in NEE decreases from 10-30 g C m −2 yr −1 to ∼ 10 g C m −2 yr −1 when 5 N and P limitation is included (Fig. 6d). It is interesting to compare Fig. 6b and Fig. 6d.
In the CTL (but nutrient limited) simulations, NEE increase over Europe, eastern North America, and SE. Asia by ∼ 10 g C m −2 yr −1 . These same regions show large reductions in NEE once LULCC is included (Fig. 6c). That is, omitting LULCC leads to a misleading conclusion on the sign of the change in NEE over the period 1850-2005. 10 The decreases of global carbon uptakes by including N and P limitation mainly occur at the tropics and northern hemisphere high latitudes. The difference between the LUC and CTL simulations can be seen in Fig. 7 where the averaged annual emissions of CO 2 from LULCC for the period 1850-2005 are shown. The impact of LULCC can be clearly seen in both C-only and the CNP simula-15 tions. However, there is a general reduction in the area affected when N and P limitation is included and the larger changes become more geographically constrained to areas of intensive LULCC. This is most clear in Fig. 7c, which shows the difference between the land use emissions in the C-only simulation (Fig. 7a) and those from the CNP simulations (Fig. 7b). First, the pattern of LULCC can be clearly seen in Fig. 7c as we would 20 expect if the impact of LULCC on emissions is substantially constrained to the regions of LULCC and remote changes are limited. Also noteworthy is that Fig. 7c highlights a general tendency to positive values pointing to higher LULCC emissions in the C-only simulation. Thus, the addition of nutrient limitation tends to offset the impact of LULCC on carbon loss over the historical period.

Impact of nutrient limitation on climate
Warming between 1850 and 2005 due to the increase in atmospheric CO 2 is shown in Fig. 8 show a very similar overall changes in temperature, which is to be expected given all models are forced using the same CO 2 and aerosols, and SSTs. Clearly, LULCC and N and P limitation are small effects on climate at the global scale in comparison to human emissions of CO 2 . That said, LULCC does lead in our simulations to a small reduction in the amount of CO 2 induced warming. In the C-only simulations, warming is reduced 5 from 0.87 to 0.76 • C. This is consistent with earlier experiments suggesting LULCC cools the planet on the global average. In the NP-limited simulations, the warming of 0.78 • C is reduced to 0.72 • C. Overall, this suggests that LULCC offsets global warming (although by a very small amount on the global average) but the inclusion of N and P limitation reduces the impact of LULCC. While LULCC reduces warming by 0.11 • C, 10 with N and P limitation included this is reduced to 0.06 • C (Table 1). This is clear in the regional impacts of LULCC on temperature that is strongly regionalized (Pitman et al., 2009). In our simulations, LULCC cools primarily over North America and Eurasia by ∼ 0.5 • C for the C-only simulation (Fig. 9a). Impacts are not statistically significant at a 90 % confidence level elsewhere. Adding N and P limitation affects the spatial extent of 15 significant cooling over both North America and Eurasia (Fig. 9b) but not its magnitude.
Further study with much larger ensemble of simulations is required to ascertain this with confidence. The biogeophysical impacts of LULCC are also shown in Table 1 for different regions. Including LULCC reduced warming, as a result of the increase in surface albedo 20 by 0.003 to 0.005, and decreases in net surface radiation absorptions by ∼ 1 W m −2 and sensible heat flux by about 1 W m −2 . This cooling impact of LULCC is stronger in northern hemisphere mid and high latitudes for both the C-only and CNP cases. Compared to the C-only case, inclusion of N and P limitation reduced the effect of LULCC on surface climate globally except in northern hemisphere mid latitudes from ESDD 4,2013 Nitrogen and phosphorous limitations on past LULCC

Discussion and conclusions
In a range of earlier studies, LULCC has been shown to be an important driver of regional temperature change, at least over those regions where changes have been significant (Pielke et al., 2011, and references therein). Most of these studies have focused on the biogeophysical impacts of LULCC and have shown, most commonly, 5 cooling in the higher latitudes (Lawrence and Chase, 2010). This is associated with the dominance of the albedo impacts of LULCC and the associated snow-albedo feedback in high latitudes, which tends to cool on the annual average. In this study, our results suggest LULCC cools on the global average by about 0.1 • C without nutrient limitation ( Fig. 8). This cooling grows through the period 1850-1920 but from around 1940 re-10 mains similar. If nutrients are included, LULCC still cools the global mean temperature, but only by around 0.05 • C with a similar temporal pattern shown for the non-limited simulations. In all our simulations, the statistically significant impact of LULCC on climate remains limited to regions of intensive change (Fig. 9). Our results therefore provide support for including LULCC when examining regional-scale impacts, particularly in 15 regions of intense LULCC (de Noblet-Ducoudré et al., 2012). Our results suggest that the impact of LULCC on regional-scale temperature may be overestimated if N and P limitation are not incorporated (Table 1). Focusing on the impact of LULCC on the terrestrial carbon balance, our results suggest LULCC has a major impact on changes of total land carbon over the period 1850-20 2005. In carbon only simulations, the inclusion of LULCC decreases the additional land carbon stored in 2005 from around 210 Pg C to 80 Pg C (Fig. 3). As anticipated based on earlier simulations using our modeling system (Zhang et al., 2011) adding N and P limitation significantly decreases the scale of the terrestrial carbon sink from 210 Pg C to 85 Pg C (Fig. 3). Adding LULCC on top of this system changes the sign of the ter-Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the CO 2 -fertilization effect and an overly efficient uptake of CO 2 by the biosphere. This masks the impacts of LULCC. Once nutrients are included, CO 2 fertilization becomes less efficient and the lower uptake of CO 2 by the land can be more clearly affected by LULCC allowing for a more accurate account of the net biogeophysical and biogeochemical impacts of LULCC to be determined. The significance of this result depends 5 on how generalizable our results are to other modelling systems and highlights the conclusion by Arneth et al. (2010) that examining how LULCC interacts with biogeochemical cycling is a research priority. Our estimate of CO 2 emissions from LULCC is 130 Pg C for C-only or 97 Pg C for CNP simulation from 1850 to 2005. These estimates are lower than the estimate of 10 155 Pg C using the book-keeping method by Houghton (2008) over the same period and much higher than the estimate of 40-77 Pg C by Arora and Boer (2010) from 1850 to 2000. The 80 Pg C net sink as estimated for LUC-C approaches the higher end of the estimated land sink of 23-90 Pg C by Arora and Boer (2010), whereas the 12 Pg C land source for LUC-CNP compares well with the source of ∼ 10 Pg C inversely calcu-15 lated from other better-constrained fluxes (Denman et al., 2007). Based on the same interpretation of land cover trajectories from Hurtt et al. (2006), the LULCC emission (119 Pg C) simulated by CLM4 with N limitation (Lawrence et al., 2012) is consistent but lower than our C-only estimate (130 Pg C) and higher than the estimate for CNP (97 Pg C). CLM4 also produced a land carbon source of 68 Pg C for the 1850-2005 pe-20 riod, which is higher than the estimated source of 12 Pg C in our LUC-CNP simulation.
Our study shows that nutrient limitation significantly reduced CO 2 emission from LULCC from 1850 to 2005, and this has significant implication on the global carbon budget. From 1850 to 2005, the total CO 2 emission from fossil fuel burning was estimated to be 314 Pg C (Andres et al., 2011), about 200 Pg C was accumulated in the 25 atmosphere, 135 Pg C was taken up by the ocean (Khatiwala et al., 2009). If the CO 2 emission from LULCC was 97 Pg C over the same period, the accumulated land carbon uptake is calculated as 76 Pg C for the nutrient-limiting simulation, which is  (2007). As stated by Houghton et al. (2012), the high uncertainty in estimating carbon fluxes linked to LULCC not only because of uncertainties in rates of changes in land surface, but also because of the incomplete processes adopted by different models (e.g. wood 5 harvest and shifting cultivation). We note that there are inevitably some caveats to our study. It is dependent on one Earth System Model, one representation of the biogeochemical cycles and one implementation of LULCC. We have also used prescribed SSTs from earlier simulations with our modeling system, which has the potential to suppress impacts from LULCC (Davin and de Noblet-Ducoudré, 2010). Clearly, we would advocate experiments such as ours being repeated with other Earth System Models that include N and P limitation. That said, we suspect that our core conclusion that the inclusion of N and P limitation reduces the impact of LULCC on both temperature and on the terrestrial carbon balance will be supported by other modeling results in the future. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Andres, R. J., Gregg, J. S., Losey, L., Marland, G., and Boden, T. A.: Monthly, global emissions of carbon dioxide from fossil fuel consumption, Tellus B, 63, 309-327, 2011. Arneth, A., Harrison, S. P., Zaehle, S., Tsigaridis, K., Menon, S., Bartlein, P. J., Feichter, J., Korhola, A., Kulmala, M., O'Donnell, D., Schurgers, G., Sorvari, S., and Vesala T.: Terrestrial biogeochemical feedbacks in the climate system, Nat. Geosci., 3, 525-532, 5 doi:10.1038/ngeo905, 2010. Arora, V. and Boer, G. J.: Uncertainties in the 20th century carbon budget associated with land use change, Glob. Change Biol., 16, 3327-3348, doi:10.1111/j.1365-2486.2010.02202.x, 2010, and Abramowitz G.: Climate model simu- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | cycle feedback analysis: Results from the C4MIP model intercomparison, J. Climate, 19, 3337-3353, 2006. Gallo, K. P., Owen, T. W., Easterling, D. R., and Jamason, P. F.: Temperature trends of the U.S. historical climatology network based on satellite designated land use/land cover, J. Climate, 12, 1344Climate, 12, -1348Climate, 12, , 1999.   (