Quantifying the effect of persistent dryer climates on forest productivity and implications for forest planning: a case study in northern Germany

A more and more aggravating climate change shakes one of the key features of forest planning, i.e. to assign species for cultivation according to site suitability (Brang et al. 2008). Formerly, species selection depended on static parameters, e.g. soil nutrients, site water budget and silviculture region as a substitute for regional climatic conditions. In the meantime, the rapidly changing climate affects the primary growing conditions well below a classical rotation period (e.g. Boisvenue and Running 2006; Pretzsch et al. 2014; Kohnle et al. 2014; Aertsen et al. 2014). Thus, forest planning has to account for dynamic site conditions and management decisions are based on expectations of future site developments. Besides projections of site conditions species selection for future forests depends on species resilience, its productivity and ecological conduciveness to name but a few. For sound decisions it is essential to accurately predict the expected values in future developments and to account for their inherent uncertainty, for example the impact of different climate change scenarios on forests.

Spittlehouse (2005) categorizes forest management actions to encounter climate change into societal adaptation, adaptation of the forest and adaptation to the forest. Societal adaptation covers adjusting people’s expectations on forest services and changes in forest policy on management and conservation. Adaptation of the forest includes, for example, species selection. And adaptation to the forest comprises amongst others applying the adequate silvicultural regime, harvesting and utilization of more salvage wood and modifications in wood processing. All these measures can be proactive approaches to reduce vulnerability and enhance recovery in advance or reactive responses to diminish the impact afterwards (Ogden and Innes 2007).

According to Spittlehouse and Stewart (2003) determination of vulnerabilities and critical thresholds is an integral part of forest management adaptation to climate change. One of the core issues for forest management in the north German lowlands is the vulnerability of forests to projected changes in precipitation pattern. The most likely reduced rainfall in the growing season and increasing evapotranspiration due to rising temperatures will result in persistent dryer conditions or even drought stress on many forest sites. This will most likely affect forest productivity and alter the risk profile of forests and forest enterprises (Lindner et al. 2010, 2014).

As drought stress reduces tree growth and vitality and might even trigger tree mortality (Hanson and Weltzin 2000; Allen et al. 2010), adaptation measures should account for the effects of persistent dryer climate conditions on forests (Hogg and Bernier 2005). Two main future drought types can be distinguished. First, intense and more frequent dry spells in the growing season will occur in central Europe (IPCC 2013; Lindner et al. 2014). Second, the decrease of mean precipitation in the growing season along with increasing mean temperatures will result in long-term persistent dryer conditions. Many studies quantify growth reactions after short-term extreme drought events (e.g. Orwig and Abrams 1997; Klos et al. 2009; Michelot et al. 2012; Weber et al. 2013). These exemplary studies confirm the obviously negative effect of extreme drought events on tree growth while evincing differences according to for example species, sites, forest layers or distribution range. While these growth declines after singular short-term drought events are mostly reversible on many sites (Bigler et al. 2006; Zingg and Bürgi 2008; Sutmöller et al. 2017), still with increasing frequency of drought events the short-term growth reductions might aggregate to considerable losses in the long run.

Other case studies investigate the long-term effect of persistent dryer conditions on forest productivity. Kint et al. (2012) observed a drought-induced growth decline for beech in Belgium since the 1960s. Piovesan et al. (2008) related the reduction in stand basal area increment observed for beech forests in the central Appennines, Italy, since the last three to four decades to dryer conditions. Furthermore, they found evidence for a stronger effect of temperature increase to drought than reduction of precipitation. They conclude that for beech persistent drought stress will overcompensate the positive influence of an earlier beginning of the growing season. Barber et al. (2000) assume drought stress as one possible explanation for the lessening of the positive effect of temperature increase on radial growth of white spruce in interior Alaska. Based on observed radial growth data from Scots pine, European beech and Pedunculate oak in northern Germany Bauwe et al. (2015) developed regression models and projected growth until 2100 using the emission scenario A1B for future climate conditions. Their simulation results show growth reductions for beech and oak with increasingly dryer conditions while Scots pine did not display a clear trend.

These case studies give evidence for long-term growth trends which correspond to persistent drought stress. But so far, it is not possible to satisfactorily quantify effects of persistent dryer climates on a level suitable for forest management (e.g. Adams et al. 2009; Choat et al. 2012; Anderegg et al. 2013). Until now, in Germany forest planning relies on drought vulnerability thresholds which are based on expert knowledge and/or singular observations on monitoring plots or experimental sites or current presence/absence observations in relation to various environmental factors (e.g. Spellmann et al. 2011, 2015; Kölling et al. 2009; Falk et al. 2013; Hanewinkel et al. 2014). To enhance the understanding of interactions between persistent dryer climates and forest productivity and to harness quantitative drought assessment information for forest planning we pursue three goals in the presented research: (1) We quantify the effect of persistent dryer climates on forest productivity in two north German forest regions until 2070. This is pivotal as we want to complement the currently qualitative drought vulnerability assessment by quantitative effects of climate change on forest growth. For this purpose we apply a single-tree growth simulator which already serves as a prediction tool for practical forest planning in northern Germany. For our analysis we present and employ an extended site-sensitive model version. (2) We quantify the uncertainty inherent in climate change by applying three different climate projections. (3) Finally, we point out how forest growth simulations of actual current forests in a specified region and for a defined projection period might add information to support forest planning decisions. This analysis serves as a first step towards a quantitative drought vulnerability assessment to be considered in forest management on a regional basis, for example for silvicultural management strategies, species selection decisions or long-term timber supply studies.

This study is based on projections of forest development applying the site-sensitive version of the single-tree growth simulator WaldPlaner 2.1 (Hansen and Nagel 2014). Site-sensitivity comprises, among others, effects of climatic conditions on tree growth, i.e. the impact of precipitation and temperature. In the simulation design we vary the climatic conditions according to three climate scenarios while all other factors are held constant. Thereby, differences in forest growth in the simulation results are traced back solely to variations in climate.

Study area

In this study we use forest inventory data from two regions Uelzen and Fläming in the north German lowlands (Fig. 1).

The two regions feature different current climatic and site conditions with favourable water supply in Uelzen and already today problematic precipitation and available water capacity in the lower parts of the Fläming region at least for less tolerant tree species (Table 1). Climate change is believed to increase water shortage in the future. Standard forest management surveys (cf. Böckmann et al. 1998) provide information on forest stands at 999 locations in Uelzen and 1008 locations in Fläming.

	Species area proportion (%)			Annual mean temp. (°C)	Annual prec. (mm)	cwbgs (mm)	Mean awc (mm)
	pine	beech	oak
Uelzen	61.3	4.8	6.1	9.2	727	-65	110
Fläming	73.5	12.8	5.9	9.8	572	-216	123

Model description

Figure 2 displays the individual components of the WaldPlaner 2.1 simulation framework to project forest development (for equations and parameters refer to Hansen and Nagel 2014). Single growth functions and a previous version of the model system have been extensively validated (Schmidt and Hansen 2007; Vospernik et al. 2010, 2015; Nagel 2013; Sprauer and Nagel 2015). One growth projection step commonly covers a 5-year period. The single tree data base stores individual tree parameters as described in the initialization data section below. An ingrowth routine adds trees to the data set and the density-dependent mortality model predicts single tree mortality. The competition index (c66) and the change in the competition index (c66c) due to tree removal and mortality are calculated afterwards. Basal area increment (bai) is a function of the tree parameters age, crown surface area (csa), the competition index and its change. The updated diameter at breast height (dbh) is then used to derive an updated tree height applying the site-sensitive longitudinal height-diameter model by Schmidt (2010) which utilizes climate and soil data. When applying climate scenarios in impact studies it is recommended to use average values of 30-year periods (IPCC 2013, p. 1450). Therefore, for each 5-year growth prediction the values of mean annual temperature sum in the growing season and aridity index are calculated for a 30-year period centered around the respective projection period and applied to the height-diameter model. Furthermore, development of crown surface area (csa) depends on crown width and crown length which, in turn, are functions of tree diameter and tree height. Thus, as tree height development is predicted site-sensitively soil and climate parameters also affect diameter increment. Finally, thinning and harvesting rules can be applied.

The height-diameter model by Schmidt (2010) constitutes the site-sensitive core of the single-tree growth model. It builds on a height-diameter model by Lappi (1997) which is a modified version of the Korf function (Eq. 1).

$$ \mathit{\ln}\left({H}_{kti}\right)=A-B{x}_{kti}+{\epsilon}_{kti} $$

(1)

with

$$ {x}_{kti}=\frac{{\left({dbh}_{kti}+\lambda \right)}^{-C}-{\left(30+\lambda \right)}^{-C}}{{\left(10+\lambda \right)}^{-C}-{\left(30+\lambda \right)}^{-C}} $$

(1.1)

and

H_kti: tree height value [m] for tree i in stand k at time t with ε_kti ~ N(0,σ²);

dbh_kti: dbh of tree i [cm] in stand k at time t;

A, B, C, λ: parameters of the height-diameter model.

The parameter values of C and λ in Eq. 1.1 are optimized to yield minimum quadratic deviation of the model. Further on, these values are fixed and Lappi’s model becomes linear. As convenient properties Lappi’s model formulation is biologically interpretable with A being the level and B the slope of the height-diameter curve on logarithmic scale and the two parameters show low collinearity.

The longitudinal height-diameter model (Eq. 2) by Schmidt (2010) is formulated as a generalized additive model.

$$ {\displaystyle \begin{array}{c}\mathit{\ln}\left\{E\left({H}_{kti}\right)\right\}=\widehat{p_{1a}}+{p}_{2a}{\left(1-{e}^{\widehat{-{p}_{3a}}{age}_{kti}}\right)}^{\widehat{p_{4a}}}+{f}_{2a}\left({rd}_{kti}\right)+{f}_{3a}\left({yg}_{ki}\right)+{f}_{4a}\left({tsum}_k\right)\\ {}+{f}_{5a}\left({ari}_k\right)+{f}_{6a}\left({lon}_k,{lat}_k\right)-{p}_{0b}{x}_{kti}-{p}_{1b}{age}_{kti}{x}_{kti}-{p}_{2b}{elev}_k{x}_{kti}\end{array}} $$

(2)

with

E(H_kti): expected value of tree height [m] for tree i in stand k at time t with H_kti ~ N (exp(η_kti), σ²);

age_kti: age [years] of tree i in stand k at time t;

rd_kti: relative diameter [cm] as the ratio of dbh of tree i in stand k at time t and the respective quadratic mean diameter;

yg_ki: year of germination of tree i in stand k;

tsum_k: average temperature sum in the growing season [°C] in stand k in the climatic reference period 1961 to 1990;

ari_k: average aridity index value by De Martonne (1926) in stand k in the climatic reference period 1961 to 1990;

lon_k, lat_k: Gauss-Krueger coordinates of stand k’s centroid;

elev_k: elevation [m] above sea level of stand k’s centroid;

\( \widehat{p_{xa}} \): parameter constants (with x = 1,3 and 4, respectively) of Chapman-Richards function used to approximate the age trend of the original parameter A (see formula 1);

p_2a: parameter of Chapman-Richards function used to approximate the age trend of the original parameter A (see formula  1);

f_xa: one-dimensional smoothing regression term (with x = 2, 3, 4 and 5, respectively) to describe non-linear effects of independent variables on the original parameter A (see formula 1);

f_6a: two-dimensional smoothing regression spline to account for large scale spatial autocorrelation of the original parameter A (see formula 1);

p_xb: parameter of linear effects (with x = 0, 1 and 2, respectively) on the original parameter B (see formula 1).

In a first step the expected tree height value is estimated using, among others, the climate variables temperature sum in the growing season (tsum) and De Martonne’s (1926) aridity index (ari) as described in the climate data section below. The model is parameterized based on data of the first two German National Forest Inventories, the forest enterprise inventory of Lower Saxony and experimental plots for Pedunculate and Sessile oak, European beech, Norway spruce, Douglas-fir and Scots pine in Germany (Eq. 2).

The second stage of the height-diameter model (Eq. 3) uses the predictors available water capacity (awc), soil nutrients (nut) and ground water class (gw) to further enhance site sensitivity (cf. Fleck et al. 2015, p. 47). A separate second stage is necessary as the soil parameters were not available throughout Germany during model development but only for Lower Saxony.

$$ \mathit{\ln}\left\{E\left({H}_{kti}\right)\right\}=\mathit{\ln}\left\{\widehat{E\left({H}_{kti}\right)}\right\}+{f}_{1a}\left({awc}_k\right){Z}_k+{gw}_k^T{\beta}_1+{nut}_k^T{\beta}_2+{f}_{2a}\left({lon}_k,{lat}_k\right) $$

(3)

with

E(H_kti): expected tree height value [m] for tree i in stand k at time t with H_kti ~ N (exp(η_kti), σ²);

\( \widehat{E\left({H}_{kti}\right)} \): estimated expected tree height value [m] for tree i in stand k at time t using Eq. 2;

awc_k: available water capacity in stand k [mm];

Z_k: indicator variable to define ground water impact in stand k with categories “ground water impact” and “no ground water impact”;

gw_k: vector of the categorical variable to define ground water impact with categories “no ground water impact (ground water table > 2.5 m)”, “low to medium ground water impact (ground water table 0.95 to 2.5 m)” and “medium to strong ground water impact (ground water table < 0.95 m)”;

nut_k: vector of the categorical variable to define soil nutrient supply in stand k (1 = very poor, 2 = poor, 3 = medium, 4 = good, 5 = rich; including further differentiation using + and – variants);

lon_k, lat_k: Gauss-Krueger coordinates of stand k’s centroid;

f_1a: one-dimensional smoothing regression spline to describe the non-linear effect of awc;

f_2a: two-dimensional smoothing regression spline to account for large scale spatial autocorrelation;

β₁, β₂: coefficient vectors.

The estimated expected tree height values calculated with Eq. 2 are used as an offset in the second model stage (Eq. 3).

Both stages of the site-sensitive height-diameter model are implemented in the forest simulation framework WaldPlaner. Various studies already proved the climate-sensitivity and biological plausibility of the model (Schmidt 2010; Albert et al. 2015, 2016; Fleck et al. 2015). In a comparative analysis bias and precision of the model proved satisfying (Albrecht et al. 2017).

Climate data

The 5^th IPCC report introduces a new generation of climate scenarios to depict the future concentration of global warming gases in the atmosphere (IPCC 2013). Four scenarios, denominated representative concentration pathways (RCP), are selected in the report. The scenarios differ in the underlying pathways, i.e. under RCP 8.5 rising CO₂-equivalents after 2100 (>1,370 ppm in 2100), two pathways stabilizing the CO₂-equivalents on different levels in 2100 (~850 ppm under RCP 6.0 and ~650 ppm under RCP 4.5), and a peak and decline pathway with a maximum of ~490 ppm before 2100 under RCP 2.6.

Based on the climate scenario RCP 8.5 (Moss et al. 2010; van Vuuren et al. 2011) three climate projections from 2011 to 2070 specify the range of potential climate change in this study. The three projections are calculated using the global circulation models INM-CM4 (Volodin et al. 2010), ECHAM6 (Stevens et al. 2013) and ACCESS1.0 (Bi et al. 2013).

The regional climate model STARS (Orlowsky et al. 2008) further downscales the realisations of the global circulation models to Germany. In the period 2051 to 2070 the projected temperature based on INM-CM4 shows a 1.1°C deviation from today’s (1991 to 2010) mean annual temperature for the north German lowlands. ECHAM6 exhibits a 1.6°C deviation and ACCESS1.0 2.7°C. Subsequently, these three climate scenarios are labelled minimum, median and maximum climate run. Furthermore, a retrospective climate scenario from 1951 to 2010 is projected using STARS based on daily weather records from 1218 climate stations run by the German Weather Service. Thus, a consistent time series from 1951 to 2070 is provided for each climate station.

The site-sensitive height-diameter model applies the climate parameters temperature sum in the growing season and aridity index as predictors. The beginning of the growing season is defined using the model LNVAR by Menzel (1997, p. 52 ff). The end is calculated according to suggestions by Walther and Linderholm (2006) and Frich et al. (2002) using either the temperature criterion or the short day criterion, whichever is met first. The temperature criterion applies, if the moving average of mean temperature in a 7-day period between July and October falls below 5°C. We follow von Wilpert (1990) and fix the short day criterion, defined as the day length needed for xylem growth, to October 5. De Martonne (1926) defines the aridity index as the ratio of annual precipitation sum [mm] and mean annual temperature [°C] +10. We calculate the aridity index based on annual values as the soil functions as a water storage. Finally, the temperature sum in the growing season and the aridity index is provided for all 2007 locations in Uelzen and Fläming in yearly resolution from 1951 to 2070 by interpolating the projected values of all climate stations within a 20 km distance from the respective location. The interpolation is done using WaSiM-ETH’s (Schulla 2015; Schulla and Jasper 2007) distance-weighted regression model.

Figure 3 illustrates how the temperature sum in the growing season and aridity index develop differently in Uelzen and Fläming over time. The linear trend lines clearly display an increase in temperature over time (Fig. 3a and b). For the aridity index, with lower values indicating dryer conditions, the findings show increasing dryer conditions with time (Fig. 3c and d). The temperature trend in Uelzen is on a slightly lower level than in Fläming. However, the aridity index states much dryer conditions in Fläming.

The baseline climate scenario is used as a reference in this study. It requires to avoid any temporal changes in temperature sum and aridity index looking at 20-year averages. Annual fluctuations, however, are necessary to display short-term climate variability. Therefore, the annual climate values of the reference period 1991 to 2010 are randomly mixed to create new sequences for each of the three projection periods 2011 to 2030, 2031 to 2050, and 2051 to 2070. This procedure returns identical means and variances for the three future periods.

It is the primary objective of this simulation study to quantify the effect of long-term persistent dryer climates on forest productivity and to derive tangible recommendations for regional forest planning. Thus, we first compare volume increment under climate change to volume increment under constant current climate. The analysis based on mean values from single trees provides deeper insights into individual growth reactions to climate change. The innate age trend on volume increment is eliminated by looking at deviations. However, the absolute value of volume increment depends on initial tree dimension (c.f. Berrill and O’Hara 2014). Consequently, the larger the initial volume the more likely a large deviation between volume increment under climate change and constant climate will occur.

Five general findings are apparent when analyzing the climatic effects on mean single-tree volume increment over initial volume classes (Figs. 4 and 5). (1) The expected trend of larger deviations in absolute terms in larger initital volume classes is obvious for all species in both regions except for Scots pine in Uelzen. Accompanying, the uncertainty range prominently increases over initial volume classes, i.e. the differences in volume increment deviation among the three climate scenarios are relatively minor for small trees but increase to a significant extent for larger trees. (2) Climate change has a positive effect throughout initial volume classes in most cases in the first period. This effect diminishes or even reverses over time. (3) While the uncertainty range includes only positive deviations in most cases in the first period, the uncertainty range indicates positive as well as negative climate impacts on volume increment throughout initial volume classes in most cases in the second and third period. This is especially apparent for Scots pine in the third period and for European beech as well as oak in the second and third period in both regions. (4) Scots pine is the least sensitive species regarding the climatic effect on volume increment followed by oak and beech being affected the strongest (please observe the dashed lines in Figs. 4 and 5 indicating different scales of the vertical axes). (5) Furthermore, the climatic effect is stronger in Fläming than in Uelzen.

Deviations in mean single-tree volume increment over initial volume classes help to analyze climate effects on tree growth on the scale of biological productivity. However, the climatic effect on volume increment at the forest stand level might differ. Thus, it is also necessary to report results of deviations in periodic volume increment. On the scale of area productivity using periodic volume increment more general tendencies of climatic effects on forest productivity can be identified (Fig. 6). As a reference we provide distributions on absolute values of periodic volume increment in the supplement (Additional file 2: Figures S1–S6).

The overall deviations in periodic volume increment displayed in Fig. 6 point out five general tendencies. (1) In accordance with the findings on mean single-tree volume increment periodic volume increment of all species benefits from climate change in the first projection period in both regions except for European beech and oak in Uelzen under the median and maximum scenario. (2) The uncertainty range widens over time and in most cases encompasses positive as well as negative deviations. Especially European beech strongly reacts on climate scenarios with positive deviations up to 2 m³·ha^–1·yr^–1 as well as negative values down to –3 m³·ha^–1·yr^–1 in the third period. (3) Basically, the positive effect of climate change in the first period diminishes over time and even turns negative in many cases. While the deviation in periodic volume increment of Scots pine shows only a few small negative values in the third period, periodic volume increment of oak is negatively affected by climate change in Uelzen entirely but not so much in Fläming. The climate effect on periodic volume increment of European beech shows a disparate picture over time with positive tendencies under minimum and median climate scenario in Uelzen and a clearly negative trend under median and maximum climate scenario in Fläming. (4) Scots pine is the least sensitive species, both positive and negative. Climate change has the strongest impact on European beech with the most positive deviations in Fläming in the first period and the most negative deviations in Uelzen and Fläming in the third period. Oak takes on an intermediate position looking at the deviations in periodic volume increment. (5) Comparing both regions a stronger climate effect is noticeable in Fläming looking at Scots pine and European beech. In Fläming the periodic volume increment of both species shows more positive reactions in the first period and under median and maximum scenario significant negative deviations in the third period. Periodic volume increment of pedunculate and sessile oak, however, is affected more heavily by climate change in Uelzen.

While the results displayed in Figs. 4, 5 and 6 emphasize the immediate short-term climatic effect on volume increment during the 20-year projection periods, the analysis of mean annual increment at biological rotation age (MAI_max) detects the cumulative long-term consequences of climate change on forest productivity. The differences in MAI_max indicate how the gradual divergences in volume growth between the baseline and the three climate scenarios accumulate to economically meaningful additional or fewer productivity, respectively (for absolute values on MAI_max please refer to Additional file 2: Figures S1–S6).

In Uelzen Scots pine and oak both benefit with additional MAI_max of up to 0.4 m³·ha^–1·yr^–1 depending on climate scenario and projection period (Fig. 7a). The projection only indicates a marginal loss for oak under maximum climate scenario in the third period. In Fläming the deviation in MAI_max shows similar patterns for pine and oak, however with slightly larger uncertainty ranges comprising larger positive values up to 0.6 m³·ha^–1·yr^–1 for pine (0.25 m³·ha^–1·yr^–1 for oak) and negative values down to –0.2 m³·ha^–1·yr^–1 for both species in the third period (Fig. 7b). While the uncertainty range increases significantly for both species in both regions over time, there is no obvious trend in productivity with increasing dyrer conditions until 2070. Contrary for European beech, the losses in MAI_max clearly increase over time. While beech productivity benefits from climate change in the first period in most cases with deviations in MAI_max up to 0.7 m³·ha^–1·yr^–1 in Uelzen and Fläming, there is a strong negative effect in Uelzen of down to –1 m³·ha^–1·yr^–1 in the second and even down to –3 m³·ha^–1·yr^–1 in the third projection period. In Fläming a similar trend for beech is observed with negative values down to –0.8 m³·ha^–1·yr^–1 in the second and down to –2 m³·ha^–1·yr^–1 in the third period. However, in Fläming under minimum climate scenario positive deviations in MAI_max of 0.4 m³·ha^–1·yr^–1 in the second period and even of 0.8 m³·ha^–1·yr^–1 in the third period are projected.

Generally, simulation studies serve to evaluate future developments in the light of differing framework conditions (Clark et al. 2001). The main objective of simulation studies is to provide information beyond measurements and observations for forest planning and to support decision making. In contrast, resource assessments, field experiments and observational studies collect data and record forest conditions in the first place (Zhao et al. 2014). The measured data and the results and conclusions derived thereof are used for policy making, forest planning and resource allocation, hypothesis testing and modelling. Therefore, simulation studies on possible future developments and measurement data from different sources with a retrospective view may complement each other especially for forest planning issues under uncertainty.

This study illustrates a method to assess the long-term effects of persistent dryer climate conditions on forest growth and productivity by comparing mean single-tree volume increment, periodic volume increment and mean annual increment at biological rotation age (MAI_max) under constant climate conditions and under three climate projections until 2070. The projected changes in forest growth within a specified uncertainty range serve as quantitative contributions to provide decision support in forest planning, e.g. adapting silvicultural management strategies, evaluating species future site suitability and assessing the impact on timber supply. The quantitative analysis of possible productivity changes under persistent dryer climate complements the expert-knowledge based drought vulnerability assessment which is applied in northwestern Germany today (cf. Table 2). The drought vulnerability assessment as a synoptic approach qualitatively aggregates the effects of drought on tree growth, vitality and mortality based on thresholds for site water budget. The reported projection of productivity changes quantifies effects of persistent dryer climates on forest growth separately. In addition, the effects of drought on vitality and mortality still needs to be quantified on a level suitable for forest management.

In particular, on the grounds of the specific simulation results following conclusions for forest planning in the two regions can be drawn: (1) Although the uncertainty range in the third period (2051 to 2070) encompasses positive as well as negative effects on volume increment in many cases for European beech, pedunculated and sessile oak and Scots pine in both regions decisions on species suitability have to be made. In this case, differences in site conditions need to be taken into account and furthermore the decision also is a question of risk attitude. The results presented are average values for all sites considered. Thus, favourable site conditions, e.g. sufficient available water capacity, might well compensate the negative effect of persistent dryer climates and different recommendations for these sites apply. (2) Scots pine is rated as the least sensitive species to dryer climate conditions. MAI_max of Scots pine even increases in the time period and regions under consideration. Thus, from the growth and yield point of view Scots pine will be well suited for forest management in the future. The same but not quite as pronounced applies to the management of oak in the Fläming region while being doubtful in Uelzen. (3) European beech faces the most severe growth reductions. In stands with beech as the leading species the negative impact of climate change might call for silvicultural adaptation measures and a change in species in the next forest generation. On sites, where a vulnerability analysis does not exclude beech from cultivation, forest management with beech as an admixture species is still advisable at the cost of lower productivity. (4) The clear temporal trend in magnitude and direction of the effect of persistent dryer climates on forest growth and productivity has to be considered.