Modelling long-term impacts of changes in climate, nitrogen deposition and ozone exposure on carbon sequestration of European forest ecosystems

Science of the Total Environment 605–606 (2017) 1097–1116

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Modelling long-term impacts of changes in climate, nitrogen deposition and ozone exposure on carbon sequestration of European forest ecosystems Wim de Vries a,b,⁎, Maximilian Posch c, David Simpson d,e, Gert Jan Reinds a a

Wageningen University and Research, Environmental Research (Alterra), PO Box 47, NL-6700 AA Wageningen, The Netherlands Wageningen University and Research, Environmental Systems Analysis Group, PO Box 47, NL-6700 AA Wageningen, The Netherlands c Coordination Centre for Effects (CCE), RIVM, PO Box 1, NL-3720 BA Bilthoven, The Netherlands d EMEP/MSC-W, Norwegian Meteorological Institute, PO Box 43-Blindern, N-0313 Oslo, Norway e Dept. Space, Earth & Environment, Chalmers University of Technology, SE-412 96 Gothenburg, Sweden b

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• We modelled effects of changes in climate and air quality on carbon sequestration. • The impacts of drivers on forest carbon sequestration are in line with literature data. • Large past forest growth changes were mainly due to N deposition and CO2 increase. • Limited future forest growth changes were mainly caused by climate and CO2 change. • The adverse growth impacts of ozone were b50% of the stimulating N effects

Contribution of drivers to relative change in carbon sequestration in trees and soil over Europe in the periods 1950-2000 the 2000-2050 as compared to 1900 using an interactive (Int) and multiplicative (Mul) model

a r t i c l e

i n f o

Article history: Received 21 March 2017 Received in revised form 13 June 2017 Accepted 16 June 2017 Available online xxxx Keywords: Carbon sequestration Climate change Forest Nitrogen deposition Ozone CO2

⁎ Corresponding author. E-mail address: [email protected] (W. de Vries).

http://dx.doi.org/10.1016/j.scitotenv.2017.06.132 0048-9697/© 2017 Elsevier B.V. All rights reserved.

a b s t r a c t We modelled the effects of past and expected future changes in climate (temperature, precipitation), CO2 concentration, nitrogen deposition (N) and ozone (O3) exposure (phytotoxic ozone dose, POD) on carbon (C) sequestration by European forest ecosystems for the period 1900–2050. Tree C sequestration was assessed by using empirical response functions, while soil C sequestration was simulated with the process-based model VSD, combined with the RothC model. We evaluated two empirical growth responses to N deposition (linear and non-linear) and two O3 exposure relationships (linear function with total biomass or net annual increment). We further investigated an ‘interactive model’ with interactions between drivers and a ‘multiplicative model’, in which the combined effect is the product of individual drivers. A single deposition and climate scenario was used for the period 1900–2050. Contrary to expectations, growth observations at European level for the period 1950–2010 compared better with predictions by the multiplicative model than with the interactive model. This coincides with the fact that carbon responses in kg C ha−1 yr− 1 per unit change in drivers, i.e. per °C, ppm CO2, kg N ha− 1 yr−1 and mmol m− 2 yr−1 POD,

1098

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

are more in line with literature data when using the multiplicative model. Compared to 1900, the estimated European average total C sequestration in both forests and forest soils between 1950 and 2000 increased by 21% in the interactive model and by 41% in the multiplicative model, but observed changes were even higher. This growth increase is expected to decline between 2000 and 2050. The simulated changes between 1950 and 2000 were mainly due to the increase in both N deposition and CO2, while the predicted increases between 2000 and 2050 were mainly caused by the increase in CO2 and temperature, and to lesser extent a decrease in POD, counteracted by reduced N deposition. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Forest ecosystems provide various services that are vital to human health and livelihood, including wood production (provisioning service) and carbon (C) sequestration reducing climate change (regulating service) (Millennium Ecosystem Assessment, 2005). With their considerable potential for C sequestration, forests constitute one of the most important elements of the global C cycle, and C sequestration both in trees and soils is thus a very important regulating ecosystem service. Biomass productions of forests and the related above-ground and below-ground C sequestration are influenced by climate, air quality and soil quality. Impacts are due to effects on photosynthesis or gross primary production (GPP), on autotrophic plant respiration (Ra), and thus on net primary production (NPP, being equal to GPP minus Ra), and on heterotrophic soil respiration (Rh). The most relevant drivers include (i) climatic parameters, i.e. temperature (Beedlow et al., 2004; Heimann and Reichstein, 2008; Wamelink et al., 2009) and precipitation (Toledo et al., 2011; Wamelink et al., 2009), (ii) air quality, i.e. changes in N deposition (De Vries et al., 2014; De Vries et al., 2009b; Liu and Greaver, 2009) in CO2 concentrations (Kets et al., 2010; McCarthy et al., 2006; Norby et al., 2005), tropospheric ozone (O3) (Beedlow et al., 2004; Felzer et al., 2005; Sitch et al., 2007b) and SO2 (Bytnerowicz et al., 2007; Guardans, 2002), and (iii) soil quality, especially the availability of non-nitrogen soil nutrients (P, Ca, Mg, K etc.) (Körner, 2006; Körner et al., 2005; McCarthy et al., 2006). N inputs, however, are not always promoting C sequestration; at very high inputs, N may cause adverse effects due to elevated risks of pests and diseases, drought and frost (e.g. Bobbink et al., 2003; Flückiger et al., 2002) and due to Al toxicity caused by N induced soil acidification (summary reviews in De Vries et al., 2015; and Sverdrup and Warfvinge, 1993). Many experiments have been conducted across the world to quantify forest ecosystem responses, mostly focusing on net primary production (NPP) and heterotrophic soil respiration (Rh) in response to single and sometimes multiple environmental drivers. Examples of single driver experiments include CO2 fertilization (Gielen et al., 2005; Janssens et al., 2005; Norby et al., 2005), N fertilization (Allison et al., 2008; Bowden et al., 2004; Haynes and Gower, 1995), ozone exposure (King et al., 2005; Pregitzer et al., 2006), warming (Allison and Treseder, 2008; Bronson et al., 2008) and irrigation (Borken et al., 1999; Trichet et al., 2008). Examples of multiple driver experiments, assessing potential interactions, include CO2 and N fertilization (McCarthy et al., 2010; Spinnler et al., 2002; Vose et al., 1995), CO2 fertilization and O3 exposure (King et al., 2005; Tingey et al., 2006) and N fertilization and irrigation (Kasurinen et al., 2004; Maier and Kress, 2000; Samuelson et al., 2009; Trichet et al., 2008). There is ample evidence for interactions between the drivers of C sequestration. Interactions can be either synergistic or antagonistic, i.e. effects of multiple drivers are either larger or smaller compared to the sum of the effects of the individual drivers (Zavaleta et al., 2003). For example, elevated CO2 effects are limited for N-limited systems (Reich et al., 2006), but they increase N uptake (Finzi et al., 2007) and water use efficiency through reductions in stomatal conductance (Drake and Gonzàlez-Meler, 1997) or by an increase in fine root biomass and

increased ectomycorrhizal activity (Luo et al., 2004). Increased water stress, O3 levels and CO2 levels can all lead to stomatal closure (Mauzerall and Wang, 2001) that reduces the exchange of gases, and thus limits the damaging effect of O3 and the fertilization effect of CO2 on photosynthesis (Karnosky et al., 2003). Conversely, increased light and elevated temperature lead to stomatal opening, thus amplifying the effects of CO2 and O3 (Mauzerall and Wang, 2001). Future C sequestration of forest ecosystems depends on the response to all the above environmental drivers, including both their interactions and their expected changes in the coming decades. In this context, many global biogeochemical carbon-nitrogen cycle models have been developed and used to assess the potential impacts of N deposition in interaction with climate change, CO2 fertilization and/or ozone stress on photosynthesis/plant productivity and the sequestration of carbon in terrestrial ecosystems across the globe. These models have originally been developed to assess changes in ecosystem structure and C sequestration in response to shifts in climate and CO2 concentration on a global scale (e.g. Cramer et al., 2004; Cramer et al., 2001). Recently, one or more of those models also include impacts of ozone exposure (e.g. Sitch et al., 2007a), effects of N availability (Churkina et al., 2009; Thornton et al., 2007; Zaehle et al., 2011; Zaehle et al., 2010a; Zaehle and Friend, 2010; Zaehle et al., 2010b), and in a very limited number of cases, both N and P availability (Goll et al., 2012; Wang et al., 2010; Zhang et al., 2011). Most existing global C–N models, focus on interacting impacts of N deposition, climate change and CO 2 fertilization on C and N cycling (Churkina et al., 2009; Thornton et al., 2007), thus lacking inclusion of ozone stress, impacts of non-N nutrient availability and the effects of soil acidity on the availability of those nutrients. This is a clear drawback since the availability of key plant nutrients, not only N, but also phosphorus (P), calcium (Ca), magnesium (Mg) and potassium (K), might be a key determinant of the C sequestration in forests at the global scale (see also De Vries, 2014; Fernández-Martínez et al., 2014). Another drawback in these studies is the use of a limited number of so-called plant functional types. In contrast, De Vries and Posch (2011) developed a simple empirical model that included interacting effects of climate change and N deposition on C sequestration in trees (above-ground woody biomass), while accounting for non-N nutrient availability. In this study, we further elaborated that empirical tree growth model, called EUgrow, by improving the response functions for climate change and N deposition and adding empirical response functions for CO2 and O3. We furthermore linked the extended EUgrow model to a process-based soil model VSD+ (Bonten et al., 2016; Posch and Reinds, 2009), denoted as EUgrow-VSD +, to evaluate the combined effects of past and expected future changes in climatic variables (precipitation and temperature), CO2 concentrations, N deposition and ozone exposure on C sequestration in trees and soils of European forest ecosystems during the period 1900–2050. We also accounted for the possible limitation of forests growth by other major nutrients, i.e. P, Ca, Mg and K, and for the impacts of both climate change and soil acidification on soil C sequestration. We did not consider the impacts of changes in land-use and forest age-structure on growth and C sequestration, but assumed that the area and

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

age-structure of European forest equals the 2005 distribution for the whole simulation period. We evaluated the use of two different empirical growth responses to N deposition (linear and non-linear) and O3 exposure (linear relationships with total biomass and with net annual increment). We further investigated two models: (i) an ‘interactive model’ including interactions between drivers; and (ii) a ‘multiplicative model’, in which the combined effect is the product of the individual drivers. 2. Methods and data 2.1. The EUgrow model to simulate tree carbon sequestration

G ¼ Gref 

f ðdriversÞ
f driversref

ð1Þ

where G stands for the growth rate of above-ground woody biomass expressed in C (kg C ha−1 yr−1), being part of the net primary production (NPP) at a given location and in a given year, f(drivers) stands for the location- and time-dependent impact factors affecting NPP, both through photosynthesis and plant respiration, and f(driversref) is the value of f in a given reference year at the same location (implying G = Gref in the reference year). The factors included are the availability of CO2, water and nutrients, distinguishing N from other nutrients (P, Ca, Mg, K), temperature and exposure to O3. This approach is comparable to Felzer et al. (2004), who derived gross primary production (GPP) as a multiplication of the maximum rate of C assimilation with factors related to photosynthetically-active radiation (PAR), leaf area, monthly air temperature, atmospheric CO2 concentration, relative canopy conductance, ozone concentration and the feedback of N availability on C assimilation. In this study the reference growth, Gref, was based on growth data for 20 tree species in 250 regions covering Europe in the EFISCEN model data base (Schelhaas et al., 2007) for the reference year, tref, 2005. In assessing the impacts on the various tree species, a distinction was made between Norway spruce, Scots pine, oak, beech, birch, other conifers and other broadleaves. To gain insight into the interacting impact of drivers, we investigated two models: (i) an ‘interactive model’ with a plausible formulation of interactions; and (ii) a ‘multiplicative model’, in which the combined effect is the product of the individual drivers. In the interactive model, a distinction is made between interacting factors affecting NPP, both through photosynthesis and plant respiration (CO2, water and nutrients) and non-interacting factors due to temperature and O3 exposure, according to:  f ðdriversÞ ¼ min f CO2  f water ; f N ; f nutlim ðCa; Mg; K; PÞ  f temp  f O3

line with results by Feng et al. (2011), who found that N supply up to 60 kg ha−1 yr−1 did not significantly change the sensitivity to ozone of Cinnamomum camphora (Camphor tree) seedlings, a dominant evergreen broadleaf tree species in sub-tropical regions. Impacts of climate on ozone effects are not included by an interaction between the response factors, but through an interaction in the calculation of the phytotoxic ozone dose (POD), being the used ozone exposure metric (see below). In the multiplicative model the combined effect of all drivers on tree C sequestration is simulated by multiplying the reference growth, Gref, with the product of the individual drivers, i.e.: f ðdriversÞ ¼ f CO2  f water  f N  f nutlim ðCa; Mg; K; PÞ  f temp  f O3

2.1.1. Overall approach The empirical model EUgrow simulates the net increase in C in the above-ground woody biomass (stems and branches) of forests. To assess the impact of all drivers, we used a relatively simple approach, modifying a known reference growth rate, Gref, with the impact of all major drivers and scaling them to the reference value of those drivers according to:

ð2Þ

The model assumes that net photosynthesis (NPP) is affected by changes in CO2 concentration, water, availability of N and other nutrients (P, Ca, Mg, K), in which the minimum of the product of CO2 and water versus nutrients is used. The effects of elevated CO2 and water availability can thus be limited by nutrient availability (see also De Vries, 2014; Fernández-Martínez et al., 2014), whereas – conversely – effects of elevated nutrient availability may be limited by CO2 and water availability. Net photosynthesis is also influenced by the exposure to O3, which is assumed to be multiplicative (no interactive effects). The absence of an interaction between O3 and N deposition on growth is in

1099

ð3Þ

Note that even the multiplicative model includes implicit interactions as a no interaction would require a fully additive model. All factors are dimensionless, but depend on location (forest plot) and time (simulation year), and some can be both higher and lower than one. The response is scaled by dividing the overall response function in a certain year to its value in the reference year 2005, being the most recent year for which growth data were available (see Eq. (1)). 2.1.2. Modelling drivers of tree growth Here we describe how the drivers of impacts on tree growth (aboveground woody biomass) and related tree C sequestration are modelled, i.e. temperature, water availability, CO2 exposure, N availability, other nutrient limitations and O3 exposure. Only for N the response function was explicitly derived in view of impacts on above-ground woody biomass growth. For all other drivers, the response function refers to NPP (being the woody and non-woody biomass growth), but we assumed that the impact of the considered driver on above-ground woody biomass growth is proportional to NPP. 2.1.2.1. Temperature impact. The impact of temperature on NPP was included by two factors, accounting for the impact on photosynthesis (GPP) and on autotrophic respiration (Ra). This was done by multiplying the temperature impact factor on GPP, fGPP, with a temperature impact factor for the forest carbon use efficiency (CUE), fCUE: f temp ðT Þ ¼ f GPP ðT Þ  f CUE ðT Þ

ð4Þ

The forest carbon use efficiency, which stands for the fraction of assimilated photosynthates retained in forest biomass, is defined as 1–Ra/ GPP, where Ra/GPP is the fraction of assimilated photosynthates consumed by autotrophic respiration. The influence of temperature on GPP or CO2 exchange by photosynthesis, fGPP, was taken from the C-Fix model (Veroustraete et al., 2002):   ΔH a exp C 1 − RT   f GPP ðT Þ ¼ ΔST−ΔH d 1 þ exp RT

ð5Þ

where T the annual mean temperature (K), ΔHa is the activation energy (52,750 J mol−1), ΔHd the deactivation energy (211,000 J mol−1), ΔS the entropy of the denaturation equilibrium of CO2 (704.98 J K−1 mol−1), C1 = 21.77, and R = 8.314 J K−1 mol−1 is the universal gas constant. The influence of temperature on CUE was modelled by a non-linear function of temperature according to Piao et al. (2010), converted to absolute temperature T (K):   f CUE ðT Þ ¼ 1− 0:0012  T 2 −0:68186  T þ 97:381

ð6Þ

In Fig. 1a, fGPP, fCUE and their product ftemp are displayed as a function of temperature.

1100

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

Fig. 1. Dimensionless response functions for temperature impacts on photosynthesis, fGPP, carbon use efficiency, fCUE and their product ftemp (see Eq. (4)) (a); for CO2, fCO2, for spruce and beech (b); for N deposition plus N fixation, fN, using a linear approach (c) and a non-linear approach (d).

2.1.2.2. Water availability impact. Following Prentice et al. (1993), the impact of water availability (drought stress) on forest growth was modelled as the ratio of the actual evapotranspiration, AET, and potential evapotranspiration, PET, at a site: f water ¼

AET PET

ð7Þ

Prentice et al. (1993) used this approach to assess impacts of drought stress on net assimilation, being equal to NPP, and we assumed that it holds as well for the woody part of the NPP (see above). PET was calculated as a function of net radiation, which in turn was calculated from temperature and cloudiness. AET was calculated as the lesser of a supply function S, which was related to the soil moisture content, and a demand function D, being equal to PET. Changes in the daily soil moisture content were simulated by adding the difference between actual daily precipitation and evapotranspiration in a single soil layer, from which runoff occurs when the store is full (tipping bucket principle). More details on the approach are given in Prentice et al. (1993). AET and PET were estimated on a daily basis, using monthly climatological data on precipitation rate, temperature and cloudiness that were interpolated to daily values. Data on the average monthly temperature, precipitation and cloudiness were derived from a high resolution European database (Mitchell et al., 2004; New et al., 1999) that contained monthly values for the years 1901–2100 for land-based grid-cells of 10′ × 10′ (approx. 15 km × 18 km in central Europe). 2.1.2.3. CO2 impact. The fertilizing impact of an elevated CO2 concentration on forest growth is included in EUgrow by a logarithmic scaling factor according to Gifford (1980):  f CO2 ¼ 1 þ β  log10

CO2a CO2ref

 ð8Þ

where β is a dimensionless tree species dependent parameter, and CO2a and CO2ref are the CO2 concentrations at ambient level and at reference level, the latter being the CO2 concentration in 2005 (383 ppm). This

function – also used by, e.g., Goudriaan and Ketner (1984) and Wamelink et al. (2009) – applies to NPP, but we assumed that it also holds for above-ground biomass growth rates. The average values for β were taken from Wamelink et al. (2009), who estimated β for eleven tree species from literature data on experiments with information on NPP rates at elevated and reference CO2 levels, using the same relation as Gifford (1980) between growth rate and CO2 concentration, according to:    NPP CO2a −1 =log10 CO2ref NPP0

 β¼

ð9Þ

where NPP and NPP0 are the net primary production rates at ambient (elevated) and reference levels of carbon dioxide, the latter being the concentration during the experiment, taken at 350 ppb. Results are given in Table 1, and an allocation of values to the tree species modelled is given in Table 2. In Fig. 1b fCO2 as function of CO2a is shown for two tree species with the lowest response (spruce) and the highest response (beech) to CO2.

Table 1 Average values and standard deviations of the coefficient β (see Eqs. (8) and (9)) per species describing the effect of CO2 concentration on forest growth (N = number of data points), based on the supporting material in Wamelink et al. (2009). Tree species (Latin)

Tree species (English)

β average

β standard deviation

N

Pinus sylvestris Picea abies Picea sitchensis Abies alba Larix deciduas Pseudotsuga menziesii Quercus petrea Quercus robur Fagus sylvatica Betula pendula Fraxinus excelsior Average

Scots pine Norway spruce Sitka spruce Silver fir Larch Douglas-fir Sessile oak Pedunculate oak Beech Birch Ash

0.30 0.27 0.32 0.45 0.18 0.04 1.08 0.51 1.01 – 0.41 0.59

– 0.28 0.30 0.55 – – – 0.18 0.82 – – –

1 4 3 4 1 1 1 4 5 – 1 39

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116 Table 2 Values of β for the lumped tree species used in this study describing the effect of CO2 concentration on forest growth (Eqs. (8) and (9)).

j

β

Based on original data for

Norway spruce Scots pine Other conifers

0.27 0.30 0.33

Oak

0.62

Beech Birch

1.01 0.78

Other broadleaves

0.41

Norway spruce Scots pine Weighted (on number of experiments) average of Sitka spruce, silver fir, larch and Douglas-fir Weighted (on number of experiments) average of Sessile oak and Pedunculate oak European beech Weighted (on number of experiments) average of all deciduous trees European ash

2.1.2.4. Nitrogen availability impact. The response function for N deposition has been derived from direct empirical relationships between N deposition and above-ground biomass growth, primarily based on field data, including Solberg et al. (2009), and further substantiated by a literature review presented in De Vries et al. (2014), further elaborating on the approach used by De Vries and Posch (2011). De Vries and Posch (2011) defined a hormetic response function of tree growth to N deposition approximated by a trapezoidal shape, with the following considerations: (a) a non-zero growth at zero N deposition (e.g., due to N fixation), (b) a maximum growth between 25 and 35 kg N ha−1 yr−1 based on Solberg et al. (2009), and (c) a decrease again above 35 kg N ha−1 yr−1 with a complete collapse of the system (no growth) at 150 kg N ha−1 yr−1, due to adverse impacts caused by pests/diseases and acidification, inspired by the results of Magill et al. (2004). In a recent overview of N deposition impacts on C sequestration by forest ecosystems, De Vries et al. (2014) substantiated a maximum growth rate between 15 and 35 kg N ha−1 yr−1, based on non-linear growth responses in different long-term field studies (Etzold et al., 2014; Kint et al., 2012; Nellemann and Thomsen, 2001). In this study we used that linear approach again (Fig. 1c), but alternatively used those field responses to fit the following non-linear (relative) growth curve: a

f N ðxÞ ¼ ð1 þ xÞ expð−axÞ

N −Nmax with x ¼ in v

modelled in the following way: instead of using the value of a function f at time step (year) j, fj, one uses: g j ¼ ∑k¼ j−Lþ1 w j−kþ1 f k

Lumped tree species

ð10Þ

where Nin stands for the N input (kg N ha−1 yr−1), Nmax is the N input at maximal growth (kg N ha−1 yr−1), ν is a scale parameter (in kg N ha−1yr−1) and a is a dimensionless shape parameter. The function reaches its maximum of 1 for Nin = Nmax, and for x → ∞ it approaches zero. The parameter Nmax was set at 25 kg N ha−1 yr−1. The other two parameters, ν and a, were determined by: (i) setting fN = 0.5 for Nin = 0, i.e. the relative growth at zero N input; and (ii) setting fN = 0.1 (10% of the maximum growth) at Nin = 125 kg N ha− 1 yr−1. The value of fN = 0.5 was based on data from Solberg et al. (2009), as elaborated further in De Vries and Posch (2011) and Etzold et al. (2016), whereas the value of 0.1 was based on results of Magill et al. (2004). Determining the parameters by least-square optimization yielded (rounded) a = 1.8 and ν = 39 kg N ha−1 yr−1. The resulting curve is shown in Fig. 1d. The results show that at an N input of 60 kg N ha−1 yr−1, the growth is approximately 70% of the maximum growth for deciduous trees, which is in line with results for oak found by Kint et al. (2012). The upward part of the curve relates to N limitation and the downward part to adverse N effects, caused by enhanced sensitivities to frost/drought, pests/diseases and acidification. The drawback of using an empirical N response function is that it does not take into account the time lag between N deposition and growth response caused by large N reservoirs in soils. Therefore, it is the accumulated N deposition that leads to changes in soil N availability (e.g. Borken et al., 2002) and thus in growth. Such a time lag has been

1101

L

with ∑k¼1 wk ¼ 1

ð11Þ

where the wk are weights, which indicate the (relative) strength of memory for year k, (w1 for year j, w2 for year j–1, etc.) and L is the length of the memory (in years). Constant memory of length L is obtained by setting wk = 1/L, k = 1,…,L, i.e. the value in year j is the arithmetic mean of the values of the last L years. A linear declining memory into the past is obtained by setting wk = 2(L − k + 1)/[L(L + 1)], k = 1,…,L; and here we use a linear declining memory for the N response with L = 40 years. Note that the x-axis in Fig. 1c and d (the N input) includes not only N deposition, but also N fixation. This refers to the lichen-induced N fixation in boreal forests that decreases with increasing N deposition. This was modelled by using a hyperbolic curve: N fix ¼ Nfix;0 

1 1 þ 0:2  Ndep

ð12Þ

with both Nfix and Ndep in kg N ha−1 yr−1 and Nfix,0 being the N fixation at negligible N deposition. The value of Nfix,0 was set at 2 kg N ha−1 yr−1 based on DeLuca et al. (2002), who found that the feather moss (Pleurozium schreberi) in symbiosis with a cyanobacterium (Nostoc sp.) fixes between 1.5 and 2.0 kg N ha−1 yr−1 in boreal forests of northern Scandinavia. The reduction function was based on stand-scale N-enrichment experiments at five levels (0, 3, 6, 12, and 50 kg N ha−1 yr−1) in the boreal zone of northern Sweden by Gundale et al. (2011), as shown in Fig. 2a. The relationship between N deposition and the sum of N deposition and N fixation is given in Fig. 2b. In principle, the growth response will also differ between fertile and less fertile sites, especially in view of N limitation (De Vries, 2014; Fernández-Martínez et al., 2014). Solberg et al. (2009) thus found a distinction in response in sites with a C/N b 25 in the organic layer as compared to sites with a C/N N 25 in the organic layer. This aspect has, however, not been included in our empirical modelling approach. 2.1.2.5. Nutrient limitation impact. The growth of forests is also limited by the availability of other nutrients than N, i.e. base cations (Ca, Mg, K), and P. This was included according to the approach used by De Vries and Posch (2011). In the model, these nutrients are supplied by deposition and weathering. The demand of nutrient k is derived from the reference growth, Gref, and the average nutrient content, ck, in the tree. Allowing for a dilution of the average nutrient content by a fraction zk (related to the uncertainty in optimal nutrient content in stem wood for a given growth), the minimum demand for nutrient k is given by Dk = zk·ck·Gref. Let Sk be the annual supply of nutrient k, then the (potential) reduction of growth is computed as:  f nutlim ¼ min 1; SCa =DCa ; SMg =DMg ; SK =DK ; SP =DP

ð13Þ

This formula says that the most limiting nutrient determines growth, and the ‘1’ ensures that fnutlim does not exceed one. 2.1.2.6. Ozone exposure impact. Ozone affects vegetation by direct cellular damage once it enters the leaf through the stomata, i.e. O3 uptake is a function of both ambient O3 levels and stomatal conductance (Mauzerall and Wang, 2001). The cellular damage is probably the result of changes in membrane permeability (Krupa and Manning, 1988) and a reduction in stomatal conductance, as the stomata close in response to increased internal O3 (Reich, 1987). The effect of O3 exposure on forest growth in the period 1900–2050 is included by a scaling factor for O3 exposure relative to the reference ambient O3 exposure in 2005, fO3, using the phytotoxic ozone dose (POD, in mmol m−2) as the measure for O3 exposure. This stomatal flux-based POD is calculated by accounting for

1102

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

Fig. 2. The reduction in N-fixation in response to N deposition (a) based on data by Gundale et al. (2011) and the relationship between N deposition and the sum of N deposition and N fixation (b) for an N fixation at zero N deposition, Nfix,0, of 2 kg N ha−1 yr−1 (see Eq. (12)).

the effects of climate (temperature, humidity, light), ambient ozone concentration, soil (moisture availability) and plant development (growth stage) on the extent of opening of the stomatal pores on leaf surfaces through which ozone enters the plant (e.g. Ashmore et al., 2004). In this approach, the hourly mean stomatal flux of O3 is accumulated over a stomatal flux threshold of Y nmol m−2 s−1 (Büker et al., 2015; Mills et al., 2011). The flux is based on the projected leaf area. The stomatal flux-based POD for forests is based on a stomatal flux threshold of 1 nmol m−2 s−1, called POD1. A summary of the calculation approach of POD (including the POD1 values for forests) is given in Simpson et al. (2007, 2012). The function fO3 used in EUgrow is: f O3 ¼ 1−a2  POD1

ð14Þ

with a2 being based on empirical relationships between POD1 and relative changes in tree growth. Relationships were initially derived between POD1 and changes in total biomass, being the net primary production (NPP) as reported in Harmens et al. (2010) and Calatayud et al. (2011). More recently, Emberson (2015) derived dose-response relationships for net annual increment (NAI) and POD1 for beech/ birch, Norway spruce/Scots pine, oak and Aleppo pine. Results for the relationship between POD1 and both total biomass and NAI are given in Table 3. Relationships are highly significant, except for Aleppo pine, which is hardly sensitive to ozone and thus the relationship is poor. An allocation of a2 values to the tree species according to Table 3 is given in Table 4. The allocation of coefficients to trees which are missing is based on the sensitivity of tree species to ozone exposure presented in Paoletti (2006). 2.2. The VSD+ model for simulating soil carbon sequestration The dynamic model VSD + is used to calculate soil C and N mineralisation and C and N pool changes, based on the difference Table 3 Coefficients in the relationship (see Eq. (14)) between POD1 (phytotoxic ozone dose, in mmol m−2, above the stomatal flux threshold of 1 nmol m−2 s−1) and: (i) relative total biomass (left; based on Harmens et al. (2010) for all tree species except oak and on Calatayud et al. (2011) (for oak) and (ii) relative net annual increment NAI (right; data based on Emberson, 2015) (R2 is explained variation). Tree species

Norway spruce/scots pine Oak Beech/birch Holm oak/aleppo pine a

Refers to POD1.6.

Total biomass

NAI

between litterfall and mineralisation. VSD + predicts soil C and N mineralisation, considering the impacts of temperature, water availability, pH and N concentration of the litter.

2.2.1. Modelling approach The VSD+ model has been developed from the VSD model (Posch and Reinds, 2009) by including an explicit description of organic C and N turnover (Bonten et al., 2016). The changes in the soil organic C contents are calculated using the RothC-26.3 model formulation (Coleman and Jenkinson, 2005), which is a five-compartment soil organic C model including (i) Decomposable Plant Material (DPM), (ii) Resistant Plant Material (RPM), (iii) Microbial Biomass (BIO), (iv) Humified Organic Matter (HUM) and (v) Inert Organic Matter (IOM). Turnover of the first four pools follows a first-order process, which is modified by temperature, soil moisture, clay content and pH. The IOM pool does not decompose. The mineralisation and immobilisation of N are dependent on the turnover of the C pools and the C/N ratio of the respective C pools. The C/N ratios are (i) calculated from the N content of the plant material input for the DPM pool, (ii) set at fixed values of 100, 8.5 and 10, respectively, for the RPM, BIO and IOM pools and (iii) again calculated for the HUM pool by assuming that N turnover from the DPM, RPM and BIO pools is transferred to this pool (Bonten et al., 2016).

Table 4 Coefficients in the linear relationship between POD1 (in mmol m−2) and (i) relative total biomass (TB) and (ii) relative net annual increment (NAI) for different trees species (see Eq. (14)). Tree species (groups)

a2 (TB)

a2 (NAI)

Based on (see Table 3)

Norway spruce Scots pine Other conifers Northern Europe Southern Europe Oak Beech Birch Other broadleaves Northern Europe Southern Europe

0.0024 0.0024

0.0078 0.0078

Norway spruce/Scots pine Norway spruce/Scots pine

0.0024 0.00063 0.0034a 0.011 0.011

0.0078 0.0051 0.0057 0.0101 0.0101

Norway spruce/Scots pineb Aleppo pineb Oak Beech/birch Beech/birch

0.0034a 0.00063

0.0057 0.0051

Oakc Aleppo pinec

a

a2

R2

a2

R2

0.0024 0.0034a 0.011 0.00063

0.55 0.67 0.64 0.20

0.0078 0.0057 0.0101 0.0051

0.77 0.76 0.63 0.66

Refers to POD1.6 but used in this study as if it was POD1. The allocation of other conifers is set equal to Norway spruce/Scots pine in Northern Europe and to Aleppo pine in Southern Europe. c The allocation of other broadleaves is set equal to Oak in Northern Europe and to Aleppo pine in Southern Europe, based on the assumption is that other broadleaves are mainly evergreen broadleaf species in the Mediterranean area, which are tolerant to ozone due to their sclerophyllous leaves and lower gas exchange rate (Calatayud et al., 2011; Paoletti, 2006). b

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

2.2.2. Litterfall input 2.2.2.1. Calculation of litterfall. Net C and N accumulation (sequestration) or release is determined by litter inputs to the soil and mineralisation of organic C and N in the soil. Changes in litterfall rates with time are based on a scaling with growth rates, calculated by EUgrow according to: Rlf ¼ Rlf ;ref  G=Gref

ð15Þ

where Rlf stands for litterfall rate, G for growth rate and the subscript ref refers to the litterfall rate and growth rate in the reference year 2005. As with reference growth rates, litterfall inputs for the reference year 2005 were based on standing biomass data for stem wood (Nabuurs et al., 2000; Schelhaas et al., 2007), multiplied by biomass expansion factors (BEFs), thus giving information on the standing biomass of foliage/fine roots and branches/coarse roots, which were then multiplied by component-specific biomass turnover rates (BTRs) to estimate litterfall and root turnover as input to the VSD + model. The whole tree biomass was calculated per biomass component (foliage/ fine roots, branches/coarse roots and stems) on the basis of tree species and tree component specific BEFs (the ratio of a given biomass component, i.e. foliage, fine roots or branches/coarse roots to stem wood biomass in kg C) that depend on age class and region (country) in Europe, as derived in the project ‘Carbo-invent’ SCEN (http://cordis. europa.eu/result/rcn/82855_en.html) and incorporated in EFISCEN. By assuming a constant age structure during the 150 year simulation period, use was made of an age class area weighted average BEF per biomass component per tree species per region/country. As with the BEFs, BTRs also depend on age class, and use was made of age class area weighted average BTRs per biomass component per tree species per region/ country. Inputs of woody material (stems, branches, coarse roots) to the soil were not taken into account, implying that all of it is removed from the system, with the nutrients in it. The model calculations were based on the assumption of a constant standing biomass, implying that additional growth is harvested each year. A fraction of the harvest is returned in the form of foliage/fine roots and this fraction is given by EFISCEN. The C input to the soil was calculated by multiplying the litterfall rate with a C concentration in litterfall of 50%. 2.2.2.2. Calculation of N concentrations in litterfall. For the N contents in leaves we assumed that they depend on N deposition, according to Bonten et al. (2016): ctN leaves ¼ ctN  leaves; min þ ctN leaves;  max −ctNleaves; min expNlfdepNdepÞ  1−e




1103

model also predicts soil pH and soil solution chemistry, and initial values were derived by assuming equilibrium with the N and S deposition in 1900. The plausibility of the results was evaluated by comparing the predicted mean and standard deviations of C pools below major tree species (Scots pine, Norway spruce, oak and beech) on clustered soil types, i.e. peat soils, sandy soils (texture class 1, being b 8% clay), loamy soils/light clay soils (texture class 2 with clay content between 8 and 18%) and clay soils (texture class 3, 4 and 5 with clay content above 18%) in 2000 with measurements for those combinations at a subset of approximately 6000 ICP Forest plots (De Vos et al., 2015). 2.3. Input data 2.3.1. Geographical data bases The input data for the calculations on the European scale, which consist of spatial information describing land cover, soils, and forest growth, were derived by overlaying and combining maps and data bases of: (i) land cover, using a harmonised land cover map for Europe (Slootweg et al., 2005), (ii) soils, using the European Soil Database v2 map at scale 1:1 M (JRC, 2006) and (iii) about 250 forest regions with growth data for a variety of species and age classes from the European Forest Institute (Schelhaas et al., 2007). Overlaying these maps and merging polygons with common soil, vegetation and regional characteristics within blocks of 0.10° × 0.05° (about 5 km × 5 km at 60°N) resulted in about 362,000 records (‘sites’) (after neglecting those with size b1 km2) covering about 1.1 million km2 of forests west of 32°E. More details on the data and procedures can be found in Reinds et al. (2008). 2.3.2. Forest growth data Growth data were taken from the database of the European Forest Institute (Schelhaas et al., 2007) for the year 2005 (reference growth; see above). Values used are area-weighted annual averages over the seven considered (lumped) tree species, i.e. Norway spruce, Scots pine, oak, beech, birch, other conifers and other broadleaves. Nutrient contents in stems and branches were taken from Jacobsen et al. (2003). The nutrient contents for ‘other conifers’ were obtained by averaging spruce and pine and for ‘other broadleaves’ by averaging oak, beech and birch, using a branch-to-stem ratio of 0.15 and 0.20, respectively. Growth volume was converted to mass, using average densities of 450 and 650 kg m−3 for conifers and deciduous trees, respectively. For mixed forests, coniferous and deciduous tree values were averaged. The C sequestered in trees was obtained by multiplying the net biomass growth with a C content of 50%.

ð16Þ

where ctNleaves is N content in leaves (%), ctNleaves,min is minimum N content in leaves (%), ctNleaves,max is maximum N content in leaves (%), expNlfdep is the exponent for the relation between N in litter fall and N deposition and Ndep is the N deposition (eq m−2 yr−1). Default values of ctNleaves,min, ctNleaves,max and expNlfdep are included in GrowUp for different tree species. These default values have been derived from a European database of N leaf contents and N deposition (De Vries et al., 2003). 2.2.3. C and N mineralisation rates Data for the initial soil C pools in 1900 were derived by assuming equilibrium with C inputs by litterfall in 1900, calculated as described above, using standard RothC turnover rate constants of the DPM, RPM, BIO and HUM pools, modified by temperature, soil moisture, clay content and pH. The initial distribution of soil organic matter content over the five C pools was calculated by: (i) estimating the IOM pool from the initial total organic C content, (ii) assuming a steady state for the DPM, RPM and BIO pools with the C inputs at the starting year, and (iii) from the total C balance for the HUM pool. Details of all processes and the initialisation are given in Bonten et al. (2016). The VSD +

2.3.3. Climate data Climate data for the period 1961–2050 were bias-corrected data from the ECHAM5 A1B-r3 RCA3 simulation (Kjellström et al., 2011) described in more detail by Engardt et al. (2017). Bias correction was available for daily temperature and precipitation. To run the EUgrow-VSD+ model from 1900 onwards, climate data for the period 1900–1960 were generated by random draws out of the climate data for 1961–1970, comparable to other studies (e.g. Engardt et al., 2017). In the simulations 10-year averages were taken centred around 1900, 1910,…, 2050 to smooth the climate pattern. 2.3.4. CO2 concentration Trends in annual average atmospheric CO2 concentration between 1900 and 2005 were taken from Etheridge et al. (1996) for the period 1900–1960, based on measured CO2 concentrations from air in Antarctic ice and firn, and from Keeling and Whorf (2006) for the period 1960– 2005, based on measurements at Mauna Loa. For the year 2050 a concentration of 573 ppm was assumed (based on the IPCC SRES A1B scenario), and between 2005 and 2050 values were obtained by linear interpolation.

1104

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

2.3.5. Air quality: N and S deposition and POD The depositions of S as well as oxidised and reduced N were calculated with the EMEP model (Simpson et al., 2012) using the same setup (with RCA3 meteorology) as described in Simpson et al. (2014) and Engardt et al. (2017). Also the phytotoxic ozone dose (POD) was calculated by the EMEP model, incorporating the DO3SE ozone deposition module, which parameterises ozone uptake as functions of tree phenology, light, temperature, humidity, and soil moisture (Emberson et al., 2001; Simpson et al., 2012; Simpson et al., 2003). Historic S, NOx, NH3 and VOC emissions were based upon GAINS 2005, EMEP1990 and scaling factors for earlier years from Lamarque et al. (2010). Further details are given in Simpson et al. (2014) and Engardt et al. (2017). Predictions for the period 2005–2050 were based on GAINS emission scenarios, (www.iiasa.ac.at/web/home/research/researchPrograms/Overview2. en.html) as improvement over the Representative Concentration Pathways (RCPs), being greenhouse gas concentration trajectories adopted by the IPCC for its Fifth Assessment Report (AR5) in 2014 (Van Vuuren et al., 2011). Compared to the RCPs, the GAINS emissions have a stronger focus on air pollutant emissions and abatement technologies, not just limited to CO 2 (Amann et al., 2013). Base cation deposition data over Europe were taken from simulations with an atmospheric dispersion model for base cations (Van Loon et al., 2005) and kept constant throughout the simulation period.

2.4. Temporal trends and geographic variation in drivers Spatially-averaged trends in annual mean temperature, CO2 concentration, NOx, NHy and S deposition as well as POD1 exposure over European forests for the period 1900–2050 are shown in Fig. 3 for the whole of Europe and three sub-regions: North, Central and South. The southern region corresponds to that defined by Carnicer et al. (2011), and the boundary of the northern region was chosen to include all of Scandinavia. Under the A1B climate scenario, the CO2 concentration rises from approximately 300 ppm in 1900 to N500 ppm in 2050 (Fig. 3A). Temperatures at the ~ 362,000 forest sites increase on average by about 2 °C between 1900 and 2050 (Fig. 3B), whereas median precipitation hardly changes, with precipitation at southern sites decreasing and at northern sites increasing compared to 1900 (not shown; see also De Vries and Posch, 2011). The deposition of both NOx and NH3 shows a steep rise during 1950–1980, followed by a steady decline in Central Europe. In Southern and Northern Europe, the deposition stayed relative constant between 1980 and 1990 and declined afterwards (Fig. 3C and D). Note, however, that the predicted NOx deposition further declines towards 2050, whereas the predicted NH3 deposition stays nearly constant or even slightly increases. The European average total N deposition in 1900 is estimated near 3 kg N ha−1 yr−1, while the highest European average deposition, occurring in 1980, is nearly 13 kg N ha−1 yr−1, decreasing to about 7 kg N ha−1 yr−1 in 2050.

Fig. 3. Temporal development of annual mean CO2 (A), temperature (B), oxidised N deposition (C), reduced N deposition (D), total sulphur deposition (E) and POD1 (F) over European forests in the period 1900–2050. Overall average over European forested area (red line) and averages over three European regions: North (N55°N), Central (46°N–55°N) and South (b46°N) are shown. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

The deposition of S (influencing soil C via soil pH) follows a similar trend as NOx and NH3 deposition, with a peak near 1980 in Central Europe and near 1990 in Southern and Northern Europe (Fig. 3E). The ozone exposure in terms of POD1 (Fig. 3F) shows a similar trend as the reduced N deposition.

2.5. Scenarios: impact of individual drivers and their interactions The scenarios used to study the impact of each driver and their interactions on the sequestration of C in forests and forest soils are listed in Table 5. The impacts of the four drivers (climate, CO2 concentration, N deposition and O3) were investigated (a) by running all combinations of the 4 drivers (both their past trend and future scenario values) switched on or kept constant at their 1900-value (resulting in 16 socalled scenarios) and (b) by subtracting scenario results that differ only in a certain driver being on or off. A ‘scenario’ is any combination of the 4 drivers – past and future climate, CO2 concentration, N deposition and O3 concentration – switched on or kept at their 1900-value; it is denoted Sijkl, where i, j, k, l each take the value 1 (driver scenario on) or 0 (driver constant at 1900 value), with i, j, k, l referring to Climate, CO2, Ndep and O3, resp. Obviously, S0000 (scenario 1) refers to constant (1900) model input for all 4 drivers (the ‘zero-line’) and in the reference scenario (scenario 16; S1111), all drivers are assumed to change according to past estimates (period 1900–2005) and future predictions (period 2005–2050), using the IPCC A1B scenario. In the other scenarios, three drivers (scenarios 2–5), two drivers (scenarios 6–11) or one driver (scenario 12–15) is assumed to stay constant at the 1900 level, while the other drivers are assumed to vary. The impacts of the individual drivers (Climate, CO2 concentration, N deposition or O3 exposure) is not additive in the EUgrow model and thus depends on the status of the other divers. The impact of a particular driver was calculated by comparing two scenarios that differ only in one driver being on or off. For every driver there are 8 such possibilities, as listed in Table 6. The impact of each driver was assessed using the linear N and total biomass O3 model.

Table 5 Scenarios used to study the impact of each driver, alone or in combination with others. ‘1900’ (denoted as 0 in scenario code) means that the driver was kept constant at the 1900 level. ‘1900–2050’ (denoted as 1 in scenario code) means that the driver changed during this period, based on historical (1900–2005) and predicted data (2005–2050). Scenario

Climate

All drivers constant 1 S0000 1900

CO2

N deposition

Ozone

1900

1900

1900

Three drivers constant (one driver varying) 2 S1000 1900–2050 1900 3 S0100 1900 1900–2050 4 S0010 1900 1900 5 S0001 1900 1900

1900 1900 1900–2050 1900

1900 1900 1900 1900–2050

Two drivers constant (two drivers varying) 6 S1100 1900–2050 1900–2050 7 S1010 1900–2050 1900 8 S1001 1900–2050 1900 9 S0110 1900 1900–2050 10 S0101 1900 1900–2050 11 S0011 1900 1900

1900 1900–2050 1900 1900–2050 1900 1900–2050

1900 1900 1900–2050 1900 1900–2050 1900–2050

One driver constant (three drivers varying) 12 S1110 1900–2050 1900–2050 13 S1101 1900–2050 1900–2050 14 S1011 1900–2050 1900 15 S0111 1900 1900–2050

1900–2050 1900 1900–2050 1900–2050

1900 1900–2050 1900–2050 1900–2050

All drivers varying 16 S1111 1900–2050

1900–2050

1900–2050

1900–2050

1105

Table 6 The eight possible subtractions of scenarios used to study the impact of each driver and their interaction with the other drivers (see Table 5 for an explanation of the codes). Nr

1 2 3 4 5 6 7 8

Influence of single effect Climate

CO2 conc.

N deposition

O3 exposure

S1000–S0000 S1100–S0100 S1010–S0010 S1001–S0001 S1110–S0110 S1101–S0101 S1011–S0011 S1111–S0111

S0100–S0000 S1100–S1000 S0110–S0010 S0101–S0001 S1110–S1010 S1101–S1001 S0111–S0011 S1111–S1011

S0010–S0000 S1010–S1000 S0110–S0100 S0011–S0001 S1110–S1100 S1011–S1001 S0111–S0101 S1111–S1101

S0001–S0000 S1001–S1000 S0101–S0100 S0011–S0010 S1101–S1100 S1011–S1010 S0111–S0110 S1111–S1110

3. Results 3.1. Trends in tree and soil carbon sequestration Simulated temporal changes in the European average C sequestration in trees and in soil in response to changes in climate, CO2 concentration, N deposition and O3 exposure are shown in Fig. 4. The results include those predicted with the interactive model (including driver interactions; top) and the multiplicative model (bottom), excluding the impact of non-nitrogen nutrient (base cations and phosphorus) limitations, but distinguishing the various used N and ozone response functions. Both the calculated tree and soil C sequestration is averaged over 10 year periods. Results show substantial differences in results between the interactive and the multiplicative model, the latter simulating a much stronger impact of the combined drivers. This implies that the interactions between the drivers reduce their combined effect on tree and soil C sequestration. The results also show a rather strong impact of the different growth responses to N deposition (both linear and non-linear) for both the multiplicative and interactive model, both in the past and future. In the past, there is large impact of the two different O3 exposure functions (total biomass or net annual increment impacts), but for the future the impact of the O3 function used is limited, due to the relative small predicted change in POD (see Fig. 3F) as compared to the change in N deposition (Fig. 3C and D). The calculated annual C sequestered in trees increases almost continually over time in the period 1900–2005 when using the linear N deposition response and the O3 exposure function for total biomass, and the same holds for the future period up to 2050 for both the multiplicative and interactive model approach. Simulations with the multiplicative and interactive model show that the estimated C sequestration in trees and soils in the year 2005 was 36% and 14% higher, respectively, as compared to 1900. These changes are only 16% and 12%, respectively, in 2050 as compared to 2005. The relative large impact in the past is mainly due to the increase in both N deposition and CO2, while the relative small increase in the future is mainly due to the fact that the stimulating effect of climate and CO2 change is counteracted by the ‘adverse’ effect of reduced N deposition. In the interactive model, tree growth is not much affected up to 1970 but it starts to increase from that period onwards (in response to the increased N deposition), and even more after 2005, which is mainly due to climate and CO2 changes, since N deposition decreases in that period. Apparently, the decrease in N deposition has a smaller decreasing effect on forest growth, compared to the increasing impacts of climate and CO2 changes. When using the O3 exposure function for net annual increment, the calculated tree C sequestration stays either nearly constant in the last 100 years or even decreases over time, depending on the N response function used. Unlike N, the impact of O3 response is much smaller in future predictions. This is due to the estimated smaller change in POD as compared to N (especially NOx) deposition in the future (see Fig. 3).

1106

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

Fig. 4. Temporal development of European average C sequestered in trees (left) and in soil (right) in response to changes in climate, CO2 concentration, N deposition and O3 exposure for the four combinations of two growth responses to N deposition (linear and non-linear) and O3 exposure (total biomass and net annual increment, NAI), using the interactive model (Int, top) and the multiplicative model (Mul, bottom), excluding non-N nutrient limitations.

The soil C pool changes reflect on average the changes in tree C pools, as this affects the C input by litterfall. However, unlike tree C sequestration, the changes can be negative since soil respiration can be higher than litter C input. The limited increase or even decrease in the period after 2020, to even negative values in 2050 for the interactive model using a linear growth response to N deposition, is mainly due to climate change that increases soil respiration through increased temperature. Compared to tree C sequestration, the changes in soil C pools are on average very limited (between − 25 and 150 kg C ha−1 yr−1), being generally 10 or more times lower than the changes in tree C pools (between1100 and 1500 kg C ha−1 yr−1). This result is comparable with previous studies (see Section 4.1).

3.2. Impacts of individual drivers on tree carbon sequestration The impact of individual drivers (climate, CO2 concentration, N deposition or O3 exposure) on tree C sequestration depends on the status of the other divers. This is most prominent in the interactive model, of which the results are presented in Fig. S1. Results for the multiplicative model are given in Fig. S2. A more detailed insight in the impacts of individual drivers is shown in Fig. 6, using the multiplicative model with the linear N and total biomass O3 approaches, which seem to produce the most plausible results (see Discussion). Results obtained for the interactive model are given in the supplementary material (Fig. S3). The impact of each particular driver was calculated by subtracting any two scenarios that differ only in that driver being on or off, resulting in the eight possibilities shown in Fig. 5 (see also Table 6). As expected, effects of adding a driver (left figures) or removing a driver (right figures) are similar in a multiplicative model, whereas there are clear differences when using an interactive model (Fig. S3). The impact of each driver, i.e. climate, CO2, N and O3 on the relative change in tree C sequestration for the 150 year simulation period are about 5, 9, 12.5 and −4%, respectively, when applying the multiplicative model and using the linear N and total biomass O3 response (Table 7). The impact of all individual drivers on tree C sequestration obtained with other response functions and for the interactive model can be

found in the Supplementary material (Table S1). When applying the non-linear N and NAI O3 response multiplicative model, the percentages for N and O3 change to about 13.5 and −8.5%, resp. (Table S1b). When applying the interactive model, these percentages are about 5, 2, 9 and − 4%, respectively, using the linear N and total biomass O3 response (Table S1c), with the N and O3 effect changing to about 6.5 and − 8.5%, resp., when using the non-linear N and NAI O3 responses (Table S1d). Note that the percentages are the average changes over the whole 150 year period. The effect of CO2 is substantially lower when using the interactive model, the effect of N is slightly lower, whereas those of climate and O3 are quite comparable. For both models, the relative impacts of N on tree C sequestration remain relatively similar when using the non-linear N formulation (slightly lower for the interactive model, i.e. near 7% and slightly higher for the multiplicative model, i.e. near 13%). However, for both models, the relative impacts of O3 on tree C sequestration double when applying the NAI biomass O3 response model, i.e. on average to about −8.5%.

3.3. Spatial patterns in tree and soil carbon sequestration Spatial patterns for the time-averaged tree and soil C sequestration for the period 1900–1950, 1950–2000 and 2000–2050 are shown in Fig. 6 for the interactive model and multiplicative model with the linear N response and the total biomass response to ozone. The maps present the annual average C sequestered in trees and soil in a grid cell. Results show that the 50 year average additional tree and soil C sequestration in the recent past (1950–2000) is higher in Central Europe, as compared to Northern and Southern Europe, for both the interactive model (top graphs) and multiplicative model (bottom graphs). However, in the future (2000–2050) large changes in tree C sequestration occur both in Central and Northern Europe, specifically with the multiplicative model. This difference in pattern is due to the predicted large impact of N deposition in Central Europe in the past and of climate change in both Central and Northern Europe in the future. Apparently, water availability limitations mainly offset the effects of CO2 and temperature increase in Southern Europe.

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

1107

Fig. 5. The influence of the single drivers on tree carbon sequestration, derived from eight different possibilities for subtracting two relevant scenarios (see Table 6), using the multiplicative model (Mul) with the linear N response and O3 total biomass response.

The role of N deposition in the period 1950–2000 is clearly illustrated for the Netherlands where tree C sequestration is predicted to stay nearly constant for both the interactive and multiplicative model, due to the highly elevated N deposition levels, leading to an initial growth stimulation followed by a growth decline (see the response function for N deposition and N fixation in Fig. 1). The role of driver interactions is nicely illustrated by the difference in tree C sequestration in Northern Europe for the period 2000–2050. When using the interactive model, limitations due to N availability and ozone exposure mainly offset effects of climate and CO2 changes in this region. The calculated changes in C sequestration in soils largely follow the additional tree C sequestration pattern in the period 1950–2000 as a result of increased litterfall, where N deposition effects dominate, but

much less in the period 2000–2050, when climate change effects dominate. In Northern Europe, additional soil C sequestration becomes even negative when using the interactive model, whereas additional tree C sequestration is highly positive. This is because soil C sequestration is influenced by both litterfall C input, which increases with an increase in tree growth, and soil C mineralisation rates, which specifically increases in Northern Europe due to increasing temperature. The opposite effect can be seen in the south (specifically in Italy) where the calculated additional soil C sequestration is positive, when using both the interactive and multiplicative model, while additional tree C sequestration is negative in both cases (Fig. 7). This is because water limitation limits tree C sequestration and thereby litterfall C input, but even more the soil C mineralisation rates, leading ultimately to a net soil C sequestration.

1108

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

Table 7 The contribution of each driver to the average change in tree carbon sequestration during the 150 year simulation period relative to 1900 in interaction with the other drivers, using the multiplicative model and interactive model (values in brackets) applying the linear N and total biomass O3 response. Scenario

S0000 S1000 S0100 S0010 S0001 S1100 S1010 S1001 S0110 S0101 S0011 S1110 S1101 S1011 S0111

Drivers switched off

None Climate CO2 N O3 Climate + CO2 Climate + N Climate + O3 CO2 + N CO2 + O3 N + O3 Climate + CO2 + N Climate + CO2 + O3 Climate + N + O3 CO2 + N + O3

Added effect (%)a Climate

CO2 concentration

N deposition

O3 exposure

4.6 (4.9) – 5.1 (5.4) 4.6 (4.0) 4.5 (4.8) – – – 5.1 (4.6) 5.0 (5.3) 4.6 (3.9) – – – 5.0 (4.6)

8.6 (1.1) 9.0 (1.5) – 8.9 (2.7) 8.3 (1.0) – 9.3 (3.4) 8.7 (1.5) – – 8.6 (2.6) – – 9.0 (3.3) –

12.7 (8.5) 12.7 (7.4) 13.0 (10.2) – 12.4 (8.4) 13.1 (9.4) – 12.5 (7.3) – 12.8 (10.0) – – 12.8 (9.2) – –

−3.9 (−3.9) −3.9 (−4.0) −4.1 (−4.0) −4.1 (−4.0) – −4.2 (−4.0) −4.1 (−4.1) – −4.3 (−4.1) – – −4.4 (−4.2) – – –

a The impact in % is calculated as (scenario + added effect − scenario) / scenario × 100. E.g., the first entry for the added effect of O3 is computed as (S0001 − S0000) / S0000, and the last entry for the added effect of climate as (S1111 − S0111) / S0111.

3.4. Contribution of drivers to tree and soil carbon sequestration in past and future The contribution of individual drivers and their interactions to the relative change in total C sequestration by trees and soil over Europe in the period 1950–2000 and 2000–2050 as compared to 1900, for both the interactive and multiplicative model, is shown in Fig. 7. Compared to 1900, the impacts of climate and CO2 change on elevating growth and tree carbon sequestration are lower in the period 1950–2000 compared to 2000–2050, due to larger changes in CO 2 and climate in the latter period. The average growthincreasing impact of N deposition compared to 1900 is comparable in both periods. Even though the annual changes are different in sign, due to increasing N deposition to 1985 and a decline afterwards, the average effect is about equal. Similarly, the average growth-decreasing impacts of ozone, compared to 1900, are comparable in both periods. The positive impacts of N deposition and adverse impacts of ozone exposure on soil carbon sequestration, respectively, are however, larger in 1950–2000 as compared to 2000–2050. As indicated before for the whole 150 year period (Table 7), the effect of CO2 and N deposition is substantially lower for the interactive model than for the multiplicative model while the effects of climate and O3 are quite comparable in both models. The difference is especially pronounced for the impact of N on soil C sequestration in the period 1950–2000, The European average increase in total forest carbon sequestration compared to 1900 is 41% in the period 1950–2000 and 58% in the period 2000–2050, when using the multiplicative model, while it is 16% and 20% when using the interactive model, again illustrating the slower increase in growth rates the future (see also Fig. 3). The results presented in Fig. 7 are based on the approach in which one driver has been added (scenarios 2–5 and scenario 16 minus scenario 1; see Table 6). The results are slightly different when removing a driver (see Table S2). The impacts of drivers per unit change, i.e. the responses in kg C ha−1 yr−1 per °C, ppm CO2, kg N ha−1 yr−1 and mmol m−2 yr−1 POD for the interactive and multiplicative model with the linear N response model and total biomass response to POD are given in Table 8. The results presented in Table 8 are again based on the approach in which one driver has been added. Results show that per unit change, the effects of temperature and POD are relatively constant, varying between 53 and 86 kg C ha− 1 yr− 1 per °C and between − 037 and − 0.54 kg C per mmol POD (converting mmol m − 2 yr− 1 POD to mol ha− 1 yr− 1 POD) for all periods and models. For CO2, the effects

are, however, five to ten times as low for the interactive model (0.09–0.31 kg C ha− 1 yr− 1 per ppm CO2) as compared to the interactive model (1.02–1.41 kg C ha− 1 yr− 1 per ppm CO2). For N deposition the effects are nearly twice as low substantially lower for the interactive model (12.1–17.1 kg C kg N− 1) as compared to the interactive model (14.7–28.6 kg C kg N− 1). In the period 1900–1950, with very small changes in drivers, the N deposition impact was comparable for both models. 4. Discussion 4.1. Plausibility of tree and soil carbon sequestration 4.1.1. Tree carbon sequestration The plausibility of the simulated trends in tree growth and related C sequestration was evaluated by comparing observations on growth (net annual increment, NAI) at European level for the period 1950–2010, based on Nabuurs et al. (2013), with our model results for that period. However, apart from changes in air quality and climate, observed growth changes were also affected by a gradual change in the species composition and age structure of forests as well as forest management (e.g. Kauppi et al., 2014). To gain insight into the effects of changing drivers (air quality and climate) on changes in past growth (NAI) rates, the changes in NAI were corrected for changes in age structure, which also affect NAI changes. At European scale, the observed changes in the age-class distribution in European forests between 1950 and 2010 show an increase in the percentage of forest in the age classes between 21 and 80 years and a decrease above 81 years, while the percentage of forest in the age class below 20 years remained almost the same (Vilén et al., 2012). This indicates that part of the NAI increment might be due to a changed age structure, since trees in the age classes between 21 and 80 years grow faster that those above 81 years. By using average data on normalized relative growth rates per age class (Schelhaas pers. comm.) and multiplying them with the differences in fraction of forests 1950, 1960… up to 2010 in eight 20 year classes (i.e. 1–20, 21–40, 41–60, 61–80, 81–100, 101–120, 121–140 and N141 years), the relative impact of age structure change between 1950 and 2010 on NAI increment was estimated at only 2% (Fig. 8), implying that the change in growth rate was predominantly due to changes in environmental drivers. Observations show a strongly increasing growth (NAI) rate and related tree C sink up to 2005, especially between 1980 and 1990, followed by signs of C sink saturation after 2005, as reported in detail in Nabuurs et al. (2003). The observed increase in that period was as

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

1109

Fig. 6. Spatial variation in tree and soil C sequestration over Europe in the period 1900–1950, 1950–2000 and 2000–2050 (annual average C sequestered after the first year of the respective period) using the interactive model (Int, top 6 maps) and multiplicative model (Mul, bottom 6 maps) with the linear N response and the total biomass response to POD1.

1110

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

Fig. 7. Contribution of individual drivers and their interactions to the relative change in carbon sequestration in trees and soil over Europe in the past (period 1950–2000, left) and the future (2000–2050, right) compared to 1900 using the interactive (Int, top bars) and the multiplicative model (Mul, bottom bars) with the linear N response and the total biomass response to POD1. The results are based on the approach in which one driver has been added.

much as 70%, with a European average NAI in 1960 being only about 60% of the NAI in 2005 (Fig. 8). The simulated growth trend with the model using the multiplicative model (with a linear N and total biomass O3 response) compared better with observations than the results of the interactive model, but in both cases the growth trends were underpredicted. The simulated changes were (much) lower than the observed changes, with a growth (NAI) rate in 1960 near 90% and 80% of those in 2005, respectively for the interactive and multiplicative model, while observed values are near 60% (see Fig. 8). This discrepancy can partly be explained by changes in management that also affect forest growth and which are not included in our model approach. This includes the drainage of peat soils and the choice of better growing provenances that have occurred from 1960 onwards. Results for Finland by Kauppi et al. (2014) indicate that effects of improved forest management could be substantial. The difference may also be affected by the impact of light on CO2 exchange, which may have changed over time, amongst others, due to the effect of a change in diffuse radiation induced by aerosols. The formation of aerosols (fine particulates) can cause an increase in diffuse radiation, which may cause an increase in ecosystem production, since photosynthesis seems more efficient under diffuse light conditions (Mercado et al., 2009). These authors estimate that variations in diffuse fraction enhanced the land C sink by approximately one quarter between 1960 and 1999. Under a climate mitigation scenario for the twenty-first century in which sulphate aerosols decline, the ‘diffuse-radiation’

fertilization effect declined rapidly to near zero by the end of the twenty-first century. These effects, of which the uncertainties are large however, are not included in our study. It is interesting to mention that, as with the observations, both our interactive and multiplicative model also indicate signs of hardly any change in tree C sink between 1990 and 2010. Unexpectedly, however, use of the multiplicative model, assuming no interactions between drivers, leads to the most reliable predictions for past change in growth. The enormous observed increase in NAI during 1980–1990 seems to indicate that the positive effect of N impacts has been underestimated and/or the POD effects overestimated in our model. Using non-linear N response or NAI O3 responses lead, however to even more limited effects in this period (see Fig. 4) and thus an even lower comparability with observations (not shown in Fig. 8). The spatial pattern in simulated tree C sequestration seems plausible. A similar regional pattern was found by Papale and Valentini (2003), presenting spatial (1 km × 1 km) estimates of C fluxes of European forests based on the net CO2 exchange fluxes collected at sixteen

Table 8 The impacts of drivers on tree carbon sequestration in kg C ha−1yr−1 per unit, i.e. per °C, ppm CO2, kg N ha−1yr−1 deposition and mmol m−2 yr−1 POD1 for the interactive and multiplicative model with the linear N response model and total biomass response to POD1. Model

Interactive

Period

1900–1950 1950–2000 2000–2050 Multiplicative 1900–1950 1950–2000 2000–2050

Relative driver effect in kg C ha−1 yr−1 per unit changea °C

ppm CO2

kg N ha−1 yr−1

mmol m−2 yr−1 POD1

– 53.0 77.0 – 74.7 86.2

0.31 0.21 0.09 1.42 1.31 1.02

14.5 17.2 12.1 14.7 27.7 28.6

−4.7 −4.8 −5.4 −3.7 −3.8 −4.1

a For every period, the impact is calculated as (scenario + added effect − S0000) divided by the change in driver during that period.

Fig. 8. Temporal development of observed relative changes in European net annual increment (NAI) and corrected NAI changes for changes in age structure, indicating the response to changes in environmental factors (climate, CO2 concentration, N deposition and O3 exposure) as compared to model predictions with the interactive and multiplicative model using a linear N and total biomass O3 response.

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

sites in the EUROFLUX network, using neural networks for the spatial extrapolation. 4.1.2. Soil carbon sequestration The plausibility of the simulated soil C sequestration is hard to assess on observed data since systematic surveys on the change in soil C pools are not available and also need a long period before significant temporal changes can be found in view of the spatial variability of C pools (De Vries et al., 2009b). The only available approach is the use of measured C pools in so-called chronosequences, i.e. a series of forest stands planted in different years on similar soils in the same area (e.g. Vesterdal et al., 2007). Data from two chronosequences in Denmark and one in Sweden showed soil C accumulation rates varying between approximately 50–500 kg C ha−1 yr−1 (Mol Dijkstra et al., 2009). This is in line with the results of this study, which vary on average between 150 and 300 kg C ha−1 yr−1 around 2000, depending on the model (interactive or multiplicative) (see Fig. 4). However, the observations are limited to three locations, and thus not representative for large regions. An indication of the plausibility can furthermore be obtained by comparing the model results with large-scale results obtained from other model approaches, such as those applying: (i) the limit value concept, in which the soil C accumulation is estimated by multiplying the annual litterfall with the recalcitrant fraction of the decomposing plant litter (Berg et al., 2001), (ii) the N-balance method, where C sequestration is calculated from the N retention in the soil multiplied with the present soil C/N ratio in organic layer and mineral topsoil (De Vries et al., 2006; Gundersen et al., 2006) or (iii) process-based descriptions of C cycling (e.g. Liski et al., 2005; Mol Dijkstra et al., 2009; Nabuurs and Schelhaas, 2002). Application of these methods on 192 intensive monitoring plots in the northern and western part of Europe resulted in median soil C sequestration rates of 446 kg C ha−1 yr−1 with the limit-value concept, 184 kg C ha−1 yr−1 with the N-balance method and 64 kg C ha−1 yr−1 with the process-based model SMART2, a model comparable to VSD+. Validation of those models on the three chronosequences indicated that the N-balance method gave best results, while the other approaches either overestimated or underestimated soil C sequestration rates. Other model-based large-scale applications across Europe around the year 2000 gave results in a similar range, i.e. 190 kg C ha−1 yr−1 (Liski et al.,

1111

2002) and 110 kg C ha−1 yr−1 (Nabuurs and Schelhaas, 2002). These results are all in the same order of magnitude as the calculated soil C sequestration rates in this study. It should be acknowledged, however, that the effects of environmental drivers on turnover rate constants in the RothC model are limited to effects of temperature, soil moisture and soil pH. This may affect future predictions, since many literature studies show that elevated CO2 concentrations and N additions also affect (increase) soil respiration rates, with the combined effects of multiple factors potentially being larger than the sum of the single factors, as summarized in a recent meta-analysis (Zhou et al., 2016). 4.2. Inclusion of non-nitrogen nutrient limitations Our simulations indicate that increased N deposition was the main driver for increased carbon sequestration in the past (1950–2000) while CO2 is the main driver for increased carbon sequestration in the future (2000 and 2050). This is in line with other model simulations (e.g. Bala et al., 2013). Impacts of limitations by non-nitrogen nutrients were, however, not taken into account in the above results. This was done to focus the study on the effects of drivers that changed most in the past decades. Results including the limitation of those nutrients (Fig. 9) indicated a much lower impact of the changes of the combined drivers, both for the interactive and multiplicative (compare with Fig. 4), due to the additional limiting effect of nutrients on growth at various sites. Considering the very limited comparability with observed growth changes in the past, however, these results do not seem plausible. Nevertheless, results indicate that limitations by non-nitrogen nutrients such as calcium, magnesium, potassium and phosphorus could play a role in the future and lack of inclusion may cause too optimistic growth predictions. 4.3. Plausibility of the impacts of drivers The (relative) impact of individual drivers mainly depended on the use of either the interactive model or the multiplicative model, and much less on the interaction with the other drivers (see Table 7). The plausibility of the results can be assessed in view of literature data on

Fig. 9. Temporal development of European average C sequestration in trees (left) and in soil (right) in response to changes in climate, CO2 concentration, N deposition and O3 exposure for two growth responses to N deposition (linear and non-linear) and O3 exposure (total biomass and net annual increment, NAI), using the interactive model (Int, top) and the multiplicative model (Mul, bottom), and including non-nitrogen nutrient limitations.

1112

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

both absolute % change effects (see Table 7) and on data about responses in kg C ha− 1 yr− 1 per °C, ppm CO2, kg N and mol POD (see Table 8), as discussed below. 4.3.1. Climate change The impact of climate change on tree C sequestration is determined by both the increase in temperature, generally leading to an increase in C sequestration, and a change in precipitation. Taking both aspects into account and relating the change in C sequestration to temperature change alone over the 150 year period shows an overall relative change between 2.5 and 2.9% depending on the model (Table 7), implying an average increase varying from 53 to 86 kg C ha−1 yr−1 °C−1 (Table 8), i.e. an increase near 200 kg C ha−1 yr−1 (see Fig. 5) for an average temperature change near 2.3 °C. This is comparable to results by Wamelink et al. (2009) who found an average response in Europe for trees between 60 and 200 kg C ha−1 yr−1 °C−1, depending on latitude. Especially in the Northern latitudes, impacts of climate warming tend to be large (e.g. Kauppi et al., 2014). Unfortunately, there are no meta-analysis results on the effects of temperature on NPP per unit change in temperature. The results for soil carbon sequestration are quite different from those for trees, especially in the Northern countries (Fig. 6). The simulations suggest a clear increase in tree C pools while soil C pools either increase slowly (multiplicative model) or decline (interactive model), driven by the increase in soil respiration in response to climate change. As mentioned by Davidson (2016), there are few direct observations of the effects of warming on stocks of soil carbon, and future predictions of Earth-system models (ESMs) thus neither agree in the sign nor in the magnitude of change (Todd-Brown et al., 2014). As with most ESMs, the soil model RothC used in our study simulates the impact of temperature rate on soil-carbon decomposition by a simple Arrhenius function, with the role of microbes implicitly included in the decay constants (Kirschbaum, 2004; Knorr et al., 2005). Recently Crowther et al. (2016) synthesized the results from 49 soil-warming experiments conducted across six biomes, ranging from arctic permafrost to dry Mediterranean forests. They reported that the effects of warming on changes in soil C pool in a layer of 0–10 cm depth were highly variable, with positive, negative and neutral impacts, up to carbon pools near 50 tons C ha−1, but that considerable losses occurred at higher carbon pools. The authors mention that especially the large C pools in carbonrich Arctic soils may cause a global-scale soil-carbon loss in response to climate change, in support of other publications that indicate declining soil carbon stocks in the Arctic due to enhanced decomposition in a warmer world (Abbott et al., 2016; Billings, 1987). Such high C pools, however, hardly occur in the topsoil of European forests. De Vries et al. (2009a) reported a median soil C pool of 21 tons C ha−1 and a 95-percentile of 55 tons C ha−1 in a layer of 0–10 cm depth for 522 forested plots located in various European regions. The fact that large soil C losses not occur in our model simulations is thus in line with the results from soil-warming experiments (Crowther et al., 2016). 4.3.2. CO2 fertilization In the multiplicative model modelled CO2 effects showed an impact between 1900 and 2050 of approximately 250–300 kg C ha−1 yr− 1 (Fig. 5) for a CO2 increase between 1900 and 2050 of 250 ppm (Fig. 3), in line with the estimated increase of 1–1.4 kg C ha−1 yr−1 ppm CO−1 2 (Table 8). This is lower than the results by Wamelink et al. (2009), who calculated an average response for trees in Europe near 2 kg C ha−1yr−1 ppm CO−1 2 . A meta-analysis of 15 years of FACE studies, however, showed that elevated CO2 only slightly increased crop yields, i.e. 5–7% in rice and 8% in wheat for a change in CO2 of 200 ppm (Ainsworth and Long, 2005),which is lower than our simulated average response near 25% over 250 ppm, being equal to 0.1% per ppm CO2. Ainsworth and Long (2005) did not report yield changes of trees, but reported that trees in general were more responsive to CO2 than crops. The responses reported in this meta-analysis were lower than those from previous studies that mostly measured the effect of young tree seedlings in labs

or enclosures, which seem to overestimate CO2 effects. The simulated when using an interacresponses of 0.1–0.3 kg C ha−1 yr−1 ppm CO−1 2 tive model (Table 8) seem, however, clearly too low. This corresponds with the fact that predictions with the multiplicative model gave a better comparison with observations than the interactive model. It also corresponds to a FACE CO2-enrichment experiment for forests indicating a response in net primary production, NPP, of 10% for a step increase of 100 ppm (Norby et al., 2010; Smith et al., 2016), being equal to 0.1% per ppm CO2. Another aspect, not included in the model analysis, that may affect the results is the role of fungi in N uptake and related CO2 fertilization effect. A recent synthesis of 83 elevated CO2 experiments (Terrer et al., 2016) showed that plants associated with ectomycorrhizal (ECM) fungi exhibited an overall CO2-driven ~28% enhancement in biomass, while plants associated with arbuscular mycorrhizal (AM) fungi were unresponsive to elevated CO2 under low N availability, except when associated with N2-fixers. The coincidence of trees with ECM fungi, may have caused a possible underestimation of the CO2 effect, which could (partly) explain the difference between predicted and observed forest growth over the past 50 years. 4.3.3. Nitrogen deposition The EU average carbon response to N deposition in the period of N deposition rise between 1900 and 1985 was about 200–300 kg C ha−1yr−1, depending on the type of model (see Figs. 5 and S3) for an average N deposition change near 10 kg N ha−1 yr−1 (Fig. 3), in line with the reported responses of 12–17 kg C kg N−1 for the interactive model and 15–29 kg C kg N− 1 for the interactive model (Table 8). These results are in line with reported literature data. Overall, long-term N addition experiments in boreal forests indicate biomass C–N responses of 20– 40 kg C per kg N (Gundale et al., 2014; Högberg et al., 2006; Hyvönen et al., 2008; Pregitzer et al., 2008). This range compares best with stoichiometric scaling estimates assuming an overall average retention in forests of 25% (De Vries et al., 2014). In addition, N deposition shows a positive effect on soil C sequestration, generally increasing C uptake by 10 kg C per kg N in boreal forest (Maaroufi et al., 2015). 4.3.4. Ozone exposure The EU average carbon response to ozone exposure (POD1) in the period of POD increase (between 1900 and 1985) was about − 80 kg C ha−1 yr− 1, independent of the type of model (see Figs. 5 and S3), indicating a 7–8% decrease in growth over this period (see also Fig. 4). This is line with an extensive meta-analysis by Wittig et al. (2009), who reported a 7% decrease in total biomass of trees for a change from background O3 concentrations (near 20 ppb) and the current concentration (near 40 ppb). However, since the POD response in our model is directly related to empirical data on forest growth response, it is of course in line with those data. 5. Conclusions We modelled impacts of various drivers on tree growth empirically, not accounting for all the impacts of drivers on tree physiological processes such as interacting changes of climate, N, CO2 and POD on stomatal closure, water and N use efficiency. Such effects are included in process based earth system models, and there is a need to further investigate and model the role of CO2, water and nutrients on above- and below-ground C sequestration in such models. However, the applied empirical approach does not include such details, which may have affected the results, although especially the multiplicative model gave most plausible results in view of literature information. Compared to 1900, the simulated European average total C sequestration in both forests and forest soils increased by 41% between 1950 and 2000, when using the multiplicative model. This growth increase is expected to decline between 2000 and 2050, but still an additional 17% growth (from 41 to 58%) is expected in this period. When using the interactive model, the simulated changes are 16% and 4% for the

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

two periods. As expected, the inclusion of interactions between the drivers reduces their combined effect on tree and soil C sequestration. Limitations by non-nitrogen nutrients (Ca, Mg, K, P) could also play a role in the future and lack of inclusion may cause too optimistic growth predictions. Rather unexpectedly, simulated changes in tree C pools by a multiplicative approach compared better with observed past forest growth changes than results obtained with the interactive model. The simulations using linear N and total biomass O3 response functions in a multiplicative model gave the best comparison with growth observations at European level for the period 1950–2010. Even in this case, the simulated changes were lower than the observed changes, which can partly be explained by changes in management, such as the drainage of peat soils and the choice of better growing provenances which are not included in our model approach. The impacts of changes in forest age-structure are likely small. Compared to tree C sequestration rates (between 900 and 1450 kg C ha−1 yr−1), the changes in soil C pools are much lower (between −25–300 kg C ha−1 yr−1) and can be negative due to enhanced soil respiration in response to climate change. The relatively large soil C pool changes in the past are mainly due to increased N deposition and to a lesser extent by CO2 increase, both increasing litterfall rates. The relative small increase in the future is caused by an increase in CO2 and temperature and to lesser extent a decrease in POD, counteracted by a decrease in N deposition. The impacts of climate and O3 on the relative change in tree C sequestration for the 150 year simulation period are about 2.5–3.0% and − 4.5–5.0%, respectively, in both the interactive and multiplicative model. The impacts of CO2 and N deposition in the same period are, however (much) higher in the multiplicative model (9% and 12.5%, respectively) than in interactive model (2% and 8%, respectively), since CO2 impacts are reduced by N limitations and vice versa in the latter model. The impacts of drivers per unit change, i.e. the responses in kg C ha−1 yr−1 per °C, ppm CO2, kg N ha−1 yr−1 and mmol m−2 yr−1 POD are in line with published literature data with the multiplicative model, especially for CO2 that was most affected by using an interactive model. This is in line with the observed larger discrepancy with observations when using the interactive model. The simulated 50 year average additional tree and soil C sequestration in the recent past (1950–2000) is highest in Central Europe, due to the predicted large impact of N deposition in that period, but in the future (2000–2050) large changes in tree C sequestration occur both in Central and Northern Europe in response to the expected climate change. Water availability limitations mainly offset the effects of CO2 and temperature increase in Southern Europe. Acknowledgements The authors thank Magnuz Engardt for providing bias-corrected climate data from the ECHAM5 A1B-r3 RCA3 simulation for the period 1961–2050, Gert Jan Nabuurs and Mart Jan Schelhaas for providing data on the changes in net annual increment and age structure per country and Jan Cees Voogd for evaluation of those data. This research was funded by the European Commission under DG Research 7th Framework Program for the project “Effects of Climate Change on Air Pollution Impacts and Response Strategies for European Ecosystems” (ECLAIRE), Grant agreement no: 282910, and has been co-financed by the strategic research program “Sustainable spatial development of ecosystems, landscapes, seas and regions (KB-14)" funded by the Dutch Ministry of Economic Affairs and carried out by Wageningen University and Research Centre. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2017.06.132.

1113

References Abbott, B.W., Jones, J.B., Schuur, E.A.G., Chapin I, F.S., Bowden, W.B., Bret-Harte, M.S., et al., 2016. Biomass offsets little or none of permafrost carbon release from soils, streams, and wildfire: an expert assessment. Environ. Res. Lett. 11, 034014. http://dx.doi.org/ 10.1088/1748-9326/11/3/034014. Ainsworth, E.A., Long, S.P., 2005. What have we learned from 15 years of free-air CO2 enrichment (FACE)? A meta-analytic review of the responses of photosynthesis, canopy properties and plant production to rising CO2. New Phytol. 165:351–371. http://dx. doi.org/10.1111/j.1469-8137.2004.01224.x. Allison, S.D., Treseder, K.K., 2008. Warming and drying suppress microbial activity and carbon cycling in boreal forest soils. Glob. Chang. Biol. 14:2898–2909. http://dx.doi. org/10.1111/j.1365-2486.2008.01716.x. Allison, S.D., Czimczik, C.I., Treseder, K.K., 2008. Microbial activity and soil respiration under nitrogen addition in Alaskan boreal forest. Glob. Chang. Biol. 14:1156–1168. http://dx.doi.org/10.1111/j.1365-2486.2008.01549.x. Amann, M., Klimont, Z., Wagner, F., 2013. Regional and global emissions of air pollutants: recent trends and future scenarios. Annu. Rev. Environ. Resour. 38:31–55. http://dx. doi.org/10.1146/annurev-environ-052912-173303. Ashmore, M.R., Emberson, L., Karlsson, P.E., Pleijel, H., 2004. New directions: a new generation of ozone critical levels for the protection of vegetation in Europe. Atmos. Environ. 38:2213–2214. http://dx.doi.org/10.1016/j.atmosenv.2004.02.029. Bala, G., Devaraju, N., Chaturvedi, R.K., Caldeira, K., Nemani, R., 2013. Nitrogen deposition: how important is it for global terrestrial carbon uptake? Biogeosciences 10: 7147–7160. http://dx.doi.org/10.5194/bg-10-7147-2013. Beedlow, P.A., Tingey, D.T., Phillips, D.L., Hogsett, W.E., Olszyk, D.M., 2004. Rising atmospheric CO2 and carbon sequestration in forests. Front. Ecol. Environ. 2:315–322. http://dx.doi.org/10.1890/1540-9295(2004)002[0315:RACACS]2.0.CO;2. Berg, B., McClaugherty, C., Virzo De Santo, A., Johnson, D., 2001. Humus buildup in boreal forests: effects of litter fall and its N concentration. Can. J. For. Res. 31:988–998. http://dx.doi.org/10.1139/cjfr-31-6-988. Billings, W.D., 1987. Carbon balance of Alaskan tundra and taiga ecoystems. Past, present and future. Quat. Sci. Rev. 6:165–177. http://dx.doi.org/10.1016/0277-3791(87)90032-1. Bobbink, R., Ashmore, M., Braun, S., Flückiger, W., van den Wyngaert, I.J.J., 2003. Empirical nitrogen critical loads for natural and semi-natural ecosystems: 2002 update. In: Achermann, B., Bobbink, R. (Eds.), Empirical Critical Loads for Nitrogen. Swiss Agency for Environment, Forest and Landscape SAEFL, Berne, pp. 43–170. Bonten, L.T.C., Reinds, G.J., Posch, M., 2016. A model to calculate effects of atmospheric deposition on soil acidification, eutrophication and carbon sequestration. Environ. Model. Softw. 79:75–84. http://dx.doi.org/10.1016/j.envsoft.2016.01.009. Borken, W., Xu, Y.-J., Brumme, R., Lamersdorf, N., 1999. A climate change scenario for carbon dioxide and dissolved organic carbon fluxes from a temperate forest soil drought and rewetting effects. Soil Sci. Soc. Am. J. 63:1848–1855. http://dx.doi.org/10.2136/ sssaj1999.6361848x. Borken, W., Beese, F., Brumme, R., Lamersdorf, N., 2002. Long-term reduction in nitrogen and proton inputs did not affect atmospheric methane uptake and nitrous oxide emission from a German spruce forest soil. Soil Biol. Biochem. 34:1815–1819. http://dx.doi.org/10.1016/S0038-0717(02)00171-2. Bowden, R.D., Davidson, E., Savage, K., Arabia, C., Steudler, P., 2004. Chronic nitrogen additions reduce total soil respiration and microbial respiration in temperate forest soils at the Harvard Forest. For. Ecol. Manag. 196:43–56. http://dx.doi.org/10.1016/j. foreco.2004.03.011. Bronson, D.R., Gower, S.T., Tanner, M., Linder, S., Van Herk, I., 2008. Response of soil surface CO2 flux in a boreal forest to ecosystem warming. Glob. Chang. Biol. 14:856–867. http://dx.doi.org/10.1111/j.1365-2486.2007.01508.x. Büker, P., Feng, Z., Uddling, J., Briolat, A., Alonso, R., Braun, S., et al., 2015. New flux based dose-response relationships for ozone for European forest tree species. Environ. Pollut. 206:163–174. http://dx.doi.org/10.1016/j.envpol.2015.06.033. Bytnerowicz, A., Omasa, K., Paoletti, E., 2007. Integrated effects of air pollution and climate change on forests: a northern hemisphere perspective. Environ. Pollut. 147:438–445. http://dx.doi.org/10.1016/j.envpol.2006.08.028. Calatayud, V., Cerveró, J., Calvo, E., Garcia-Breijo, F.-J., Reig-Armiñana, J., Sanz, M.J., 2011. Responses of evergreen and deciduous Quercus species to enhanced ozone levels. Environ. Pollut. 159:55–63. http://dx.doi.org/10.1016/j.envpol.2010.09.024. Carnicer, J., Coll, M., Ninyerola, M., Pons, X., Sanchez, G., Penuelas, J., 2011. Widespread crown condition decline, food web disruption, and amplified tree mortality with increased climate change-type drought. Proc. Natl. Acad. Sci. U. S. A. 108:1474–1478. http://dx.doi.org/10.1073/pnas.1010070108. Churkina, G., Brovkin, V., von Bloh, W., Trusilova, K., Jung, M., Dentener, F., 2009. Synergy of rising nitrogen depositions and atmospheric CO2 on land carbon uptake moderately offsets global warming. Glob. Biogeochem. Cycles 23, GB4027. http://dx.doi.org/10. 1029/2008GB003291. Coleman, K., Jenkinson, D.S., 2005. RothC - a model for the turnover of carbon in soil. Model Description and Users Guide. Rothamsted Research, Harpenden, UK Available at. http://rothamsted.ac.uk/sites/default/files/users/kcoleman/RothC_guide_DOS.pdf:. Cramer, W., Bondeau, A., Woodward, F.I., Prentice, I.C., Betts, R.A., Brovkin, V., et al., 2001. Global response of terrestrial ecosystem structure and function to CO2 and climate change: results from six dynamic global vegetation models. Glob. Chang. Biol. 7: 357–373. http://dx.doi.org/10.1046/j.1365-2486.2001.00383.x. Cramer, W., Bondeau, A., Schaphoff, S., Lucht, W., Smith, B., Sitch, S., 2004. Tropical forests and the global carbon cycle: impacts of atmospheric carbon dioxide, climate change and rate of deforestation. Philos. Trans. R. Soc. Lond. 359:331–343. http://dx.doi. org/10.1098/rstb.2003.1428. Crowther, T.W., Todd-Brown, K.E.O., Rowe, C.W., Wieder, W.R., Carey, J.C., Machmuller, M.B., et al., 2016. Quantifying global soil carbon losses in response to warming. Nature 540:104–108. http://dx.doi.org/10.1038/nature20150.

1114

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

Davidson, E.A., 2016. Biogeochemistry: projections of the soil-carbon deficit. Nature 540: 47–48. http://dx.doi.org/10.1038/540047a. De Vos, B., Cools, N., Ilvesniemi, H., Vesterdal, L., Vanguelova, E., Carnicelli, S., 2015. Benchmark values for forest soil carbon stocks in Europe: results from a large scale forest soil survey. Geoderma 251–252:33–46. http://dx.doi.org/10.1016/j.geoderma.2015. 03.008. De Vries, W., 2014. Nutrients trigger carbon storage. Nat. Clim. Chang. 4:425–426. http:// dx.doi.org/10.1038/nclimate2255. De Vries, W., Posch, M., 2011. Modelling the impact of nitrogen deposition, climate change and nutrient limitations on tree carbon sequestration in Europe for the period 1900–2050. Environ. Pollut. 159:2289–2299. http://dx.doi.org/10.1016/j.envpol. 2010.11.023. De Vries, W., Vel, E., Reinds, G.J., Deelstra, H., Klap, J.M., Leeters, E.E.J.M., et al., 2003. Intensive monitoring of forest ecosystems in Europe: 1. Objectives, set-up and evaluation strategy. For. Ecol. Manag. 174:77–95. http://dx.doi.org/10.1016/S03781127(02)00029-4. De Vries, W., Reinds, G.J., Gundersen, P., Sterba, H., 2006. The impact of nitrogen deposition on carbon sequestration in European forests and forest soils. Glob. Chang. Biol. 12:1151–1173. http://dx.doi.org/10.1111/j.1365-2486.2006.01151.x. De Vries, W., Reinds, G.J., Bierkens, M.F.P., 2009a. Assessment of the optimal time interval for repeated soil surveys at intensively monitored forest plots. J. Environ. Monit. 11: 2009–2021. http://dx.doi.org/10.1039/B909480F. De Vries, W., Solberg, S., Dobbertin, M., Sterba, H., Laubhann, D., van Oijen, M., et al., 2009b. The impact of nitrogen deposition on carbon sequestration by European forests and heathlands. For. Ecol. Manag. 258:1814–1823. http://dx.doi.org/10.1016/j. foreco.2009.02.034. De Vries, W., Du, E., Butterbach-Bahl, K., 2014. Short and long-term impacts of nitrogen deposition on carbon sequestration by forest ecosystems. Curr. Opin. Environ. Sustain. 9–10:90–104. http://dx.doi.org/10.1016/j.cosust.2014.09.001. De Vries, W., Posch, M., Sverdrup, H.U., Larssen, T., de Wit, H.A., Bobbink, R., et al., 2015. Geochemical indicators for use in the computation of critical loads and dynamic risk assessments. In: De Vries, W., Hettelingh, J.-P., Posch, M. (Eds.), Critical Loads and Dynamic Risk Assessments. 25. Springer Netherlands, pp. 15–58. DeLuca, T.H., Zackrisson, O., Nilsson, M.C., Sellstedt, A., 2002. Quantifying nitrogen-fixation in feather moss carpets of boreal forests. Nature 419:917–920. http://dx.doi. org/10.1038/nature01051. Drake, B.G., Gonzàlez-Meler, M.A., 1997. More efficient plants: a consequence of rising atmospheric CO2? Annu. Rev. Plant Physiol. Plant Mol. Biol. 48:609–639. http://dx.doi. org/10.1146/annurev.arplant.48.1.609. Emberson, L., 2015. Effects of climate change on air pollution impacts and response strategies for European ecosystems. Deliverable 12.3 – Delivery of Novel Thresholds for Key Dose-response Relationships of the EU Project Eclaire. Emberson, L., Ashmore, M., Simpson, D., Tuovinen, J.-P., Cambridge, H., 2001. Modelling and mapping ozone deposition in Europe. Water Air Soil Pollut. 130:577–582. http://dx.doi.org/10.1023/A:1013851116524. Engardt, M., Simpson, D., Schwikowski, M., Granat, L., 2017. Deposition of sulfur and nitrogen in Europe 1900–2050. Model calculations and comparison to historical observations. Tellus Ser. B Chem. Phys. Meteorol. 69 (1):1328945. http://dx.doi.org/10. 1080/16000889.2017.1328945. Etheridge, D.M., Steele, L.P., Langenfelds, R.L., Francey, R.J., Barnola, J.-M., Morgan, V.I., 1996. Natural and anthropogenic changes in atmospheric CO2 over the last 1000 years from air in Antarctic ice and firn. J. Geophys. Res. 101:4115–4128. http://dx.doi.org/10.1029/95JD03410. Etzold, S., Waldner, P., Thimonier, A., Schmitt, M., Dobbertin, M., 2014. Tree growth in Swiss forests between 1995 and 2010 in relation to climate and stand conditions: recent disturbances matter. For. Ecol. Manag. 311:41–55. http://dx.doi.org/10.1016/j. foreco.2013.05.040. Etzold, S., Reinds, G.J., Solberg, S., Waldner, P., Schaub, M., Simpson, D., et al., 2016. Impacts of Site Quality, Air Quality and Climate on Growth and Carbon Sequestration of European Forests (submitted manuscript). Felzer, B., Kicklighter, D., Melillo, J., Wang, C., Zhuang, Q., Prinn, R., 2004. Effects of ozone on net primary production and carbon sequestration in the conterminous United States using a biogeochemistry model. Tellus 56B:230–248. http://dx.doi.org/10. 1111/j.1600-0889.2004.00097.x. Felzer, B., Reilly, J., Melillo, J., Kicklighter, D., Sarofim, M., Wang, C., et al., 2005. Future effects of ozone on carbon sequestration and climate change policy using a global biogeochemical model. Clim. Chang. 73:345–373. http://dx.doi.org/10.1007/s10584-005-6776-4. Feng, Z., Niu, J., Zhang, W., Wang, X., Yao, F., Tian, Y., 2011. Effects of ozone exposure on sub-tropical evergreen Cinnamomum camphora seedlings grown in different nitrogen loads. Trees 25:617–625. http://dx.doi.org/10.1007/s00468-011-0538-x. Fernández-Martínez, M., Vicca, S., Janssens, I., Sardans, J., Luyssaert, S., Campioli, M., et al., 2014. Nutrient availability as the key regulator of global forest carbon balance. Nat. Clim. Chang. 4:471–476. http://dx.doi.org/10.1038/nclimate2177. Finzi, A.C., Norby, R.J., Calfapietra, C., Gallet-Budynek, A., Gielen, B., Holmes, W.E., et al., 2007. Increases in nitrogen uptake rather than nitrogen-use efficiency support higher rates of temperate forest productivity under elevated CO2. Proc. Natl. Acad. Sci. 104: 14014–14019. http://dx.doi.org/10.1073/pnas.0706518104. Flückiger, W., Braun, S., Hiltbrunner, E., 2002. Effects of air pollutants on biotic stress. In: Bell, J.N.B., Treshow, M. (Eds.), Air Pollution and Plant Life, 2nd edition John Wiley and Sons, Chichester, pp. 379–406. Gielen, B., Calfapietra, C., Lukac, M., Wittig, V.E., De Angelis, P., Janssens, I.A., et al., 2005. Net carbon storage in a poplar plantation (POPFACE) after three years of free-air CO2 enrichment. Tree Physiol. 25:1399–1408. http://dx.doi.org/10.1093/treephys/25.11.1399. Gifford, R.M., 1980. Carbon storage by the biosphere. In: Pearman, G.I. (Ed.), Carbon Dioxide and Climate: Australian Research. Australian Academy of Science, Canberra, pp. 167–181.

Goll, D.S., Brovkin, V., Parida, B.R., Reick, C.H., Kattge, J., Reich, P.B., et al., 2012. Nutrient limitation reduces land carbon uptake in simulations with a model of combined carbon, nitrogen and phosphorus cycling. Biogeosciences 9:3547–3569. http://dx.doi. org/10.5194/bg-9-3547-2012. Goudriaan, J., Ketner, P., 1984. A simulation study for the global carbon cycle, including man's impact on the biosphere. Clim. Chang. 6:167–192. http://dx.doi.org/10.1007/ BF00144611. Guardans, R., 2002. Estimation of climate change influence on the sensitivity of trees in Europe to air pollution concentrations. Environ. Sci. Pol. 5:319–333. http://dx.doi. org/10.1016/S1462-9011(02)00033-3. Gundale, M.J., DeLuca, T.H., Nordin, A., 2011. Bryophytes attenuate anthropogenic nitrogen inputs in boreal forests. Glob. Chang. Biol. 17:2743–2753. http://dx.doi.org/10. 1111/j.1365-2486.2011.02407.x. Gundale, M.J., From, M., Bach, L.H., Nordin, A., 2014. Anthropogenic nitrogen deposition in boreal forests has a minor impact on the global carbon cycle. Glob. Chang. Biol. 20: 276–286. http://dx.doi.org/10.1111/gcb.12422. Gundersen, P., Berg, B., Currie, W.S., Dise, N.B., Emmett, B.A., Gauci, V., et al., 2006. CarbonNitrogen Interactions in Forest Ecosystems - Final Report. Danish Centre for Forest, Landscape and Planning, KVL, p. 62. Harmens, H., Mills, G., Hayes, F., Norris, D., et al., 2010. Air Pollution and Vegetation. Centre for Ecology & Hydrology, Bangor, UK, p. 40. Haynes, B.E., Gower, S.T., 1995. Belowground carbon allocation in unfertilized and fertilized red pine plantations in northern Wisconsin. Tree Physiol. 15:317–325. http:// dx.doi.org/10.1093/treephys/15.5.317. Heimann, M., Reichstein, M., 2008. Terrestrial ecosystem carbon dynamics and climate feedbacks. Nature 451:289–292. http://dx.doi.org/10.1038/nature06591. Högberg, P., Fan, H., Quist, M., Binkley, D., Tamm, C.O., 2006. Tree growth and soil acidification in response to 30 years of experimental nitrogen loading on boreal forest. Glob. Chang. Biol. 12:489–499. http://dx.doi.org/10.1111/j.1365-2486.2006.01102.x. Hyvönen, R., Persson, T., Andersson, S., Olsson, B., Ågren, G.I., Linder, S., 2008. Impact of long-term nitrogen addition on carbon stocks in trees and soils in northern Europe. Biogeochemistry 89:121–137. http://dx.doi.org/10.1007/s10533-007-9121-3. Jacobsen, C., Rademacher, P., Meesenburg, H., Meiwes, K.J., 2003. Element Contents in Tree Compartments - Literature Study and Data Collection [in German]. Niedersächsische Forstliche Versuchsanstalt, Göttingen, Germany, p. 80. Janssens, I.A., Medlyn, B., Gielen, B., Laureysens, I., Jach, M.E., Van Hove, D., et al., 2005. Carbon budget of Pinus sylvestris saplings after four years of exposure to elevated atmospheric carbon dioxide concentration. Tree Physiol. 25:325–337. http://dx.doi.org/ 10.1093/treephys/25.3.325. JRC, 2006. The European Soil Data Base, Distribution Version v2.0. Joint Research Center, Ispra (Italy). Karnosky, D.F., Zak, D.R., Pregitzer, K.S., Awmack, C.S., Bockheim, J.G., Dickson, R.E., et al., 2003. Tropospheric O3 moderates responses of temperate hardwood forests to elevated CO2: a synthesis of molecular to ecosystem results from the Aspen FACE project. Funct. Ecol. 17:289–304. http://dx.doi.org/10.1046/j.1365-2435.2003.00733.x. Kasurinen, A., Kokko-Gonzales, P., Riikonen, J., Vapaavuori, E., Holopainen, T., 2004. Soil CO2 efflux of two silver birch clones exposed to elevated CO2 and O3 levels during three growing seasons. Glob. Chang. Biol. 10:1654–1665. http://dx.doi.org/10.1111/ j.1365-2486.2004.00841.x. Kauppi, P.K., Posch, M., Pirinen, P., 2014. Large impacts of climatic warming on growth of boreal forests since 1960. PLoS One 9, e111340. http://dx.doi.org/10.1371/journal. pone.0111340. Keeling, C.D., Whorf, T.P., 2006. Atmospheric CO2 records from sites in the SIO air sampling network. Trends: A Compendium of Data on Global Change. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, Tenn., U.S.A. Kets, K., Darbah, J.N., Sober, A., Riikonen, J., Sober, J., Karnosky, D.F., 2010. Diurnal changes in photosynthetic parameters of Populus tremuloides, modulated by elevated concentrations of CO2 and/or O3 and daily climatic variation. Environ. Pollut. 158: 1000–1007. http://dx.doi.org/10.1016/j.envpol.2009.09.001. King, J.S., Kubiske, M.E., Pregitzer, K.S., Hendrey, G.R., McDonald, E.P., Giardina, C.P., et al., 2005. Tropospheric O3 compromises net primary production in young stands of trembling aspen, paper birch and sugar maple in response to elevated atmospheric CO2. New Phytol. 168:623–636. http://dx.doi.org/10.1111/j.1469-8137.2005.01557.x. Kint, V., Aertsen, W., Campioli, M., Vansteenkiste, D., Delcloo, A., Muys, B., 2012. Radial growth change of temperate tree species in response to altered regional climate and air quality in the period 1901–2008. Clim. Chang. 115:343–363. http://dx.doi. org/10.1007/s10584-012-0465-x. Kirschbaum, M.U.F., 2004. Soil respiration under prolonged soil warming: are rate reductions caused by acclimation or substrate loss? Glob. Chang. Biol. 10:1870–1877. http://dx.doi.org/10.1111/j.1365-2486.2004.00852.x. Kjellström, E., Nikulin, G., Hansson, U., Strandberg, G., Ullerstig, A., 2011. 21st century changes in the European climate: uncertainties derived from an ensemble of regional climate model simulations. Tellus 63A:24–40. http://dx.doi.org/10.1111/j.1600-0870. 2010.00475.x. Knorr, W., Prentice, I.C., House, J.I., Holland, E.A., 2005. Long-term sensitivity of soil carbon turnover to warming. Nature 433:298–301. http://dx.doi.org/10.1038/ nature03226. Körner, C., 2006. Plant CO 2 responses: an issue of definition, time and resource supply. New Phytol. 172:393–411. http://dx.doi.org/10.1111/j.1469-8137. 2006.01886.x. Körner, C., Asshoff, R., Bignucolo, O., Hättenschwiler, S., Keel, S.G., Peláez-Riedl, S., et al., 2005. Carbon flux and growth in mature deciduous forest trees exposed to elevated CO2. Science 309:1360–1362. http://dx.doi.org/10.1126/science.1113977. Krupa, S.V., Manning, W.J., 1988. Atmospheric ozone: formation and effects on vegetation. Environ. Pollut. 50:101–137. http://dx.doi.org/10.1016/0269-7491(88)90187-X.

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116 Lamarque, J.-F., Bond, T.C., Eyring, V., Granier, C., Heil, A., Klimont, Z., et al., 2010. Historical (1850–2000) gridded anthropogenic and biomass burning emissions of reactive gases and aerosols: methodology and application. Atmos. Chem. Phys. 10: 7017–7039. http://dx.doi.org/10.5194/acp-10-7017-2010. Liski, J., Perruchoud, D., Karjalainen, T., 2002. Increasing carbon stocks in the forest soils of western Europe. For. Ecol. Manag. 169:159–175. http://dx.doi.org/10.1016/S03781127(02)00306-7. Liski, J., Palosuo, T., Peltoniemi, M., Sievänen, R., 2005. Carbon and decomposition model Yasso for forest soils. Ecol. Model. 189:168–182. http://dx.doi.org/10.1016/j. ecolmodel.2005.03.005. Liu, L., Greaver, T.L., 2009. A review of nitrogen enrichment effects on three biogenic GHGs: the CO2 sink may be largely offset by stimulated N2O and CH4 emission. Ecol. Lett. 12:1103–1117. http://dx.doi.org/10.1111/j.1461-0248.2009.01351.x. Luo, Y., Su, B., Currie, W.S., Dukes, J.S., Finzi, A., Hartwig, U., et al., 2004. Progressive nitrogen limitation of ecosystem responses to rising atmospheric carbon dioxide. Bioscience 54:731–739. http://dx.doi.org/10.1641/0006-3568(2004)054[0731:pnloer]2.0. co;2. Maaroufi, N.I., Nordin, A., Hasselquist, N.J., Bach, L.H., Palmqvist, K., Gundale, M.J., 2015. Anthropogenic nitrogen deposition enhances carbon sequestration in boreal soils. Glob. Chang. Biol. 21:3169–3180. http://dx.doi.org/10.1111/gcb.12904. Magill, A.H., Aber, J.D., Currie, W., Nadelhoffer, K., Martin, M., McDowell, W.H., et al., 2004. Ecosystem response to 15 years of chronic nitrogen additions at the Harvard Forest LTER, Massachusetts, USA. For. Ecol. Manag. 196:7–28. http://dx.doi.org/10.1016/j. foreco.2004.03.033. Maier, C.A., Kress, L., 2000. Soil CO2 evolution and root respiration in 11 year-old loblolly pine (Pinus taeda) plantations as affected by moisture and nutrient availability. Can. J. For. Res. 30, 347–359. Mauzerall, D.L., Wang, X., 2001. Protecting agricultural crops from the effects of tropospheric ozone exposure: reconciling science and standard setting in the United States, Europe, and Asia. Annu. Rev. Energy Environ. 26:237–268. http://dx.doi.org/ 10.1146/annurev.energy.26.1.237. McCarthy, H.R., Oren, R., Finzi, A.C., Johnsen, K.H., 2006. Canopy leaf area constrains CO2induced enhancement of productivity and partitioning among aboveground carbon pools. Proc. Natl. Acad. Sci. U. S. A. 103:19356–19361. http://dx.doi.org/10.1073/ pnas.0609448103. McCarthy, H.R., Oren, R., Johnsen, K.H., Gallet-Budynek, A., Pritchard, S.G., Cook, C.W., et al., 2010. Re-assessment of plant carbon dynamics at the Duke free-air CO2 enrichment site: interactions of atmospheric [CO2] with nitrogen and water availability over stand development. New Phytol. 185:514–528. http://dx.doi.org/10.1111/j. 1469-8137.2009.03078.x. Mercado, L.M., Bellouin, N., Sitch, S., Boucher, O., Huntingford, C., Wild, M., et al., 2009. Impact of changes in diffuse radiation on the global land carbon sink. Nature:1014–1017 http://dx.doi.org/10.1038/nature07949. Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-being: Synthesis. Island Press, Washington, DC http://www.millenniumassessment.org/documents/ document.356.aspx.pdfN. Mills, G., Pleijel, H., Braun, S., Büker, P., Bermejo, V., Calvo, E., et al., 2011. New stomatal flux-based critical levels for ozone effects on vegetation. Atmos. Environ. 45: 5064–5068. http://dx.doi.org/10.1016/j.atmosenv.2011.06.009. Mitchell, T., Carter, T.R., Jones, P.D., Hulme, M., New, M., 2004. A Comprehensive set of High-resolution Grids of Monthly Climate for Europe and the Globe: The Observed Record (1901–2000) and 16 Scenarios (2001 − 2100). Working Paper. Tyndall Centre. Mol Dijkstra, J.P., Reinds, G.J., Kros, H., Berg, B., de Vries, W., 2009. Modelling soil carbon sequestration of intensively monitored forest plots in Europe by three different approaches. For. Ecol. Manag. 258:1780–1793. http://dx.doi.org/10.1016/j.foreco.2008. 09.011. Nabuurs, G.J., Schelhaas, M.J., 2002. Carbon profiles of typical forest types across Europe assessed with CO2FIX. Ecol. Indic. 1:213–223. http://dx.doi.org/10.1016/S1470160X(02)00007-9. Nabuurs, G.J., Schelhaas, M.J., Pussinen, A., 2000. Validation of the European Forest Information Scenario Model (EFISCEN) and a projection of Finnish forests. Silva Fennica 34, 167–179. Nabuurs, G.J., Schelhaas, M.J., Mohren, G.M.J., Field, C.B., 2003. Temporal evolution of the European forest sector carbon sink from 1950 to 1999. Glob. Chang. Biol. 9: 152–160. http://dx.doi.org/10.1046/j.1365-2486.2003.00570.x. Nabuurs, G.J., Lindner, M., Verkerk, P.J., Gunia, K., Deda, P., Michalak, R., et al., 2013. First signs of carbon sink saturation in European forest biomass. Nat. Clim. Chang. 3: 792–796. http://dx.doi.org/10.1038/nclimate1853. Nellemann, C., Thomsen, M.G., 2001. Long-term changes in forest growth: potential effects of nitrogen deposition and acidification. Water Air Soil Pollut. 128:197–205. http://dx.doi.org/10.1023/A:1010318800180. New, M., Hulme, M., Jones, P.D., 1999. Representing twentieth century space-time climate variability. Part 1: development of a 1961–90 mean monthly terrestrial climatology. J. Clim. 12, 829–856. Norby, R.J., DeLucia, E.H., Gielen, B., Calfapietra, C., Giardina, C.P., King, J.S., et al., 2005. Forest response to elevated CO2 is conserved across a broad range of productivity. Proc. Natl. Acad. Sci. U. S. A. 102:18052–18056. http://dx.doi.org/10.1073/pnas.0509478102. Norby, R.J., Warren, J.M., Iversen, C.M., Medlyn, B.E., McMurtrie, R.E., 2010. CO2 enhancement of forest productivity constrained by limited nitrogen availability. Proc. Natl. Acad. Sci. U. S. A. 107:19368–19373. http://dx.doi.org/10.1073/pnas.1006463107. Paoletti, E., 2006. Impact of ozone on Mediterranean forests: a review. Environ. Pollut. 144:463–474. http://dx.doi.org/10.1016/j.envpol.2005.12.051. Papale, D., Valentini, R., 2003. A new assessment of European forests carbon exchanges by eddy fluxes and artificial neural network spatialization. Glob. Chang. Biol. 9:525–535. http://dx.doi.org/10.1046/j.1365-2486.2003.00609.x.

1115

Piao, S., Luyssaert, S., Ciais, P., Janssens, I.A., Chen, A., Cao, C., et al., 2010. Forest annual carbon cost: a global-scale analysis of autotrophic respiration. Ecology 91:652–661. http://dx.doi.org/10.1890/08-2176.1. Posch, M., Reinds, G.J., 2009. A very simple dynamic soil acidification model for scenario analyses and target load calculations. Environ. Model. Softw. 24:329–340. http://dx. doi.org/10.1016/j.envsoft.2008.09.007. Pregitzer, K., Loya, W., Kubiske, M., Zak, D., 2006. Soil respiration in northern forests exposed to elevated atmospheric carbon dioxide and ozone. Oecologia 148:503–516. http://dx.doi.org/10.1007/s00442-006-0381-8. Pregitzer, K.S., Burton, A.J., Zak, D.R., Talhelm, A.F., 2008. Simulated chronic nitrogen deposition increases carbon storage in northern temperate forests. Glob. Chang. Biol. 14: 142–153. http://dx.doi.org/10.1111/j.1365-2486.2007.01465.x. Prentice, I.C., Sykes, M.T., Cramer, W., 1993. A simulation model for the transient effects of climate change on forest landscapes. Ecol. Model. 65, 51–70. Reich, P.B., 1987. Quantifying plant response to ozone: a unifying theory. Tree Physiol. 3: 63–91. http://dx.doi.org/10.1093/treephys/3.1.63. Reich, P.B., Hobbie, S.E., Lee, T., Ellsworth, D.S., West, J.B., Tilman, D., et al., 2006. Nitrogen limitation constrains sustainability of ecosystem response to CO2. Nature 440: 922–925. http://dx.doi.org/10.1038/nature04486. Reinds, G.J., Posch, M., de Vries, W., Slootweg, J., Hettelingh, J.P., 2008. Critical loads of sulphur and nitrogen for terrestrial ecosystems in Europe and northern Asia using different soil chemical criteria. Water Air Soil Pollut. 193:269–287. http://dx.doi.org/10. 1007/s11270-008-9688-x. Samuelson, L., Mathew, R., Stokes, T., Feng, Y., Aubrey, D., Coleman, M., 2009. Soil and microbial respiration in a loblolly pine plantation in response to seven years of irrigation and fertilization. For. Ecol. Manag. 258:2431–2438. http://dx.doi.org/10.1016/j.foreco. 2009.08.020. Schelhaas, M.J., Eggers, J., Lindner, M., Nabuurs, G.J., Pussinen, A., Päivinen, R., et al., 2007. Model Documentation for the European Forest Information Scenario Model (EFISCEN 3.1.3). Alterra, Wageningen, p. 118. Simpson, D., Tuovinen, J.-P., Emberson, L., Ashmore, M., 2003. Characteristics of an ozone deposition module II. Sensitivity analysis. Water Air Soil Pollut. 143:123–137. http:// dx.doi.org/10.1023/A:1022890603066. Simpson, D., Ashmore, M.R., Emberson, L., Tuovinen, J.-P., 2007. A comparison of two different approaches for mapping potential ozone damage to vegetation. A model study. Environ. Pollut. 146:715–725. http://dx.doi.org/10.1016/j.envpol.2006.04.013. Simpson, D., Benedictow, A., Berge, H., Bergström, R., Emberson, L.D., Fagerli, H., et al., 2012. The EMEP MSC-W chemical transport model — technical description. Atmos. Chem. Phys. 12:7825–7865. http://dx.doi.org/10.5194/acp-12-7825-2012. Simpson, D., Christensen, J., Engardt, M., Geels, C., Nyiri, A., Soares, J., et al., 2014. Impacts of climate and emission changes on nitrogen deposition in Europe: a multi-model study. Atmos. Chem. Phys. 14:6995–7017. http://dx.doi.org/10.5194/acp-14-69952014. Sitch, S., Cox, P.M., Collins, W.J., Huntingford, C., 2007a. Indirect radiative forcing of climate change through ozone effects on the land-carbon sink. Nature 448 (7155): 791–794. http://dx.doi.org/10.1038/nature06059. Sitch, S., Cox, P.M., Collins, W.J., Huntingford, C., 2007b. Indirect radiative forcing of climate change through ozone effects on the land-carbon sink. Nature 448:791–794. http://dx.doi.org/10.1038/nature06059. Slootweg, J., Hettelingh, J.-P., Tamis, W., van't Zelfde, M., 2005. Harmonizing European land cover maps. In: Posch, M., Slootweg, J., Hettelingh, J.-P. (Eds.), European Critical Loads and Dynamic Modelling: CCE Report 2005, Bilthoven, the Netherlands, pp. 47–62. Smith, W.K., Reed, S.C., Cleveland, C.C., Ballantyne, A.P., Anderegg, W.R.L., Wieder, W.R., et al., 2016. Large divergence of satellite and Earth system model estimates of global terrestrial CO2 fertilization. Nat. Clim. Chang. 6:306–310. http://dx.doi.org/10.1038/ nclimate2879. Solberg, S., Dobbertin, M., Reinds, G.J., Lange, H., Andreassen, K., Fernandez, P.G., et al., 2009. Analyses of the impact of changes in atmospheric deposition and climate on forest growth in European monitoring plots: a stand growth approach. For. Ecol. Manag. 258:1735–1750. http://dx.doi.org/10.1016/j.foreco.2008.09.057. Spinnler, D.S., Egli, P.E., Körner, C.K., 2002. Four-year growth dynamics of beech-spruce model ecosystems under CO2; enrichment on two different forest soils. Trees - Structure and Function 16:423–436. http://dx.doi.org/10.1007/s00468-002-0179-1. Sverdrup, H., Warfvinge, P., 1993. The effect of soil acidification on the growth of trees, grass and herbs as expressed by the (Ca + Mg + K)/Al ratio. Reports in Ecology and Environmental Engineering 1993. Lund University, Department of Chemical Engineering II, p. 108. Terrer, C., Vicca, S., Hungate, B.A., Phillips, R.P., Prentice, I.C., 2016. Mycorrhizal association as a primary control of the CO2 fertilization effect. Science 353:72–74. http://dx.doi. org/10.1126/science.aaf4610. Thornton, P.E., Lamarque, J.F., Rosenbloom, N.A., Mahowald, N.M., 2007. Influence of carbon-nitrogen cycle coupling on land model response to CO2 fertilization and climate variability. Glob. Biogeochem. Cycles 21, GB4018. http://dx.doi.org/10.1029/ 2006GB002868. Tingey, D.T., Johnson, M.G., Lee, E.H., Wise, C., Waschmann, R., Olszyk, D.M., et al., 2006. Effects of elevated CO 2 and O 3 on soil respiration under ponderosa pine. Soil Biol. Biochem. 38:1764–1778. http://dx.doi.org/10.1016/j.soilbio.2005.12. 003. Todd-Brown, K.E.O., Randerson, J.T., Hopkins, F., Arora, V., Hajima, T., Jones, C., et al., 2014. Changes in soil organic carbon storage predicted by Earth system models during the 21st century. Biogeosciences 11:2341–2356. http://dx.doi.org/10.5194/bg-11-23412014. Toledo, M., Poorter, L., Peña-Claros, M., Alarcón, A., Balcázar, J., Leaño, C., et al., 2011. Climate is a stronger driver of tree and forest growth rates than soil and disturbance. J. Ecol. 99:254–264. http://dx.doi.org/10.1111/j.1365-2745.2010.01741.x.

1116

W. de Vries et al. / Science of the Total Environment 605–606 (2017) 1097–1116

Trichet, P., Loustau, D., Lambrot, C., Linder, S., 2008. Manipulating nutrient and water availability in a maritime pine plantation: effects on growth, production, and biomass allocation at canopy closure. Ann. For. Sci. 65:814. http://dx.doi.org/10.1051/forest: 2008060. Van Loon, M., Tarrasón, L., Posch, M., 2005. Modelling Base Cations in Europe, Oslo, Norway. p. 47. Van Vuuren, D.P., Edmonds, J., Kainuma, M., Riahi, K., Thomson, A., Hibbard, K., et al., 2011. The representative concentration pathways: an overview. Clim. Chang. 109:5–31. http://dx.doi.org/10.1007/s10584-011-0148-z. Veroustraete, F., Sabbe, H., Eerens, H., 2002. Estimation of carbon mass fluxes over Europe using the C-Fix model and Euroflux data. Remote Sens. Environ. 83:376–399. http:// dx.doi.org/10.1016/S0034-4257(02)00043-3. Vesterdal, L., Rosenqvist, L., van der Salm, C., Hansen, K., Groenenberg, B.J., Johansson, M.B., 2007. Carbon sequestration in soil and biomass following afforestation: experiences from oak and Norway spruce chronosequences in Denmark, Sweden and the Netherlands. In: Heil, G.W., Muys, B., Hansen, K. (Eds.), Environmental Effects of Afforestation in North-Western Europe. From Field Observations to Decision Support. Springer Series: Plant and Vegetation Vol. 1, pp. 19–51. Vilén, T., Gunia, K., Verkerk, P.J., Seidl, R., Schelhaas, M.J., Lindner, M., et al., 2012. Reconstructed forest age structure in Europe 1950–2010. For. Ecol. Manag. 286:203–218. http://dx.doi.org/10.1016/j.foreco.2012.08.048. Vose, J.M., Elliott, K.J., Johnson, D.W., Walker, R.F., Johnson, M.G., Tingey, D.T., 1995. Effects of elevated CO2 and N fertilization on soil respiration from ponderosa pine (Pinusponderosa) in open-top chambers. Can. J. For. Res. 25:1243–1251. http://dx. doi.org/10.1139/x95-137. Wamelink, G.W.W., Wieggers, H.J.J., Reinds, G.J., Kros, J., Mol-Dijkstra, J.P., van Oijen, M., et al., 2009. Modelling impacts of changes in carbon dioxide concentration, climate and nitrogen deposition on carbon sequestration by European forests and forest soils. For. Ecol. Manag. 258:1794–1805. http://dx.doi.org/10.1016/j. foreco.2009.05.018.

Wang, Y.P., Law, R.M., Pak, B., 2010. A global model of carbon, nitrogen and phosphorus cycles for the terrestrial biosphere. Biogeosciences 7:2261–2282. http://dx.doi.org/ 10.5194/bg-7-2261-2010. Wittig, V.E., Ainsworth, E.A., Naidu, S.L., Karnosky, D.F., Long, S.P., 2009. Quantifying the impact of current and future tropospheric ozone on tree biomass, growth, physiology and biochemistry: a quantitative meta-analysis. Glob. Chang. Biol. 15:396–424. http://dx.doi.org/10.1111/j.1365-2486.2008.01774.x. Zaehle, S., Friend, A.D., 2010. Carbon and nitrogen cycle dynamics in the O-CN land surface model: 1. Model description, site-scale evaluation, and sensitivity to parameter estimates. Glob. Biogeochem. Cycles 24, GB1005. http://dx.doi.org/10.1029/2009GB003521. Zaehle, S., Friedlingstein, P., Friend, A.D., 2010a. Terrestrial nitrogen feedbacks may accelerate future climate change. Geophys. Res. Lett. 37, L01401. http://dx.doi.org/10. 1029/2009GL041345. Zaehle, S., Friend, A.D., Dentener, F., Friedlingstein, P., Peylin, P., Schulz, M., 2010b. Carbon and nitrogen cycle dynamics in the O-CN land surface model: 2. Role of the nitrogen cycle in the historical terrestrial carbon balance. Glob. Biogeochem. Cycles 24, GB1006. http://dx.doi.org/10.1029/2009GB003522. Zaehle, S., Ciais, P., Friend, A.D., Prieur, V., 2011. Carbon benefits of anthropogenic reactive nitrogen offset by nitrous oxide emissions. Nat. Geosci. 4:601–605. http://dx.doi.org/ 10.1038/ngeo1207. Zavaleta, E.S., Shaw, M.R., Chiariello, N.R., Mooney, H.A., Field, C.B., 2003. Additive effects of simulated climate changes, elevated CO2, and nitrogen deposition on grassland diversity. Proc. Natl. Acad. Sci. U. S. A. 100:7650–7654. http://dx.doi.org/10.1073/pnas. 0932734100. Zhang, Q., Pitman, A., Dai, Y., 2011. Limitations of nitrogen and phosphorous on the terrestrial carbon uptake in the 20th century. Geophys. Res. Lett. 38, L22701. http://dx.doi. org/10.1029/2011GL049244. Zhou, L., Zhou, X., Shao, J., Nie, Y., He, Y., Jiang, L., et al., 2016. Interactive effects of global change factors on soil respiration and its components: a meta-analysis. Glob. Chang. Biol. 22:3157–3169. http://dx.doi.org/10.1111/gcb.13253.