Realizing Mitigation Efficiency of European Commercial Forests by Climate Smart Forestry

European temperate and boreal forests sequester up to 12% of Europe’s annual carbon emissions. Forest carbon density can be manipulated through management to maximize its climate mitigation potential, and fast-growing tree species may contribute the most to Climate Smart Forestry (CSF) compared to slow-growing hardwoods. This type of CSF takes into account not only forest resource potentials in sequestering carbon, but also the economic impact of regional forest products and discounts both variables over time. We used the process-based forest model 4 C to simulate European commercial forests’ growth conditions and coupled it with an optimization algorithm to simulate the implementation of CSF for 18 European countries encompassing 68.3 million ha of forest (42.4% of total EU-28 forest area). We found a European CSF policy that could sequester 7.3–11.1 billion tons of carbon, projected to be worth 103 to 141 billion euros in the 21st century. An efficient CSF policy would allocate carbon sequestration to European countries with a lower wood price, lower labor costs, high harvest costs, or a mixture thereof to increase its economic efficiency. This policy prioritized the allocation of mitigation efforts to northern, eastern and central European countries and favored fast growing conifers Picea abies and Pinus sylvestris to broadleaves Fagus sylvatica and Quercus species.


Data
To analyze forest growth across Europe and management outcomes, we applied the stand-scale process-based forest model 4C (http://www.pik-potsdam.de/4c/). We examined outputs from 4C model simulations covering a network of 132 intensively monitored forest plots, distributed over 18 European countries and 10 environmental zones (S1- Figure 1, see Reyer et al. 1 for more details on the plot selection and environmental zones. The main central European tree species included in our study were: Fagus sylvatica, Picea abies, Pinus sylvestris, Quercus petraea and Quercus robur. The forest data for each plot (e.g. tree species, age, stem number, diameter at breast height and height) were used to initialize the 4C model, which was then modified by the climate scenarios. Soils were initialized from a horizontal description of soil physics and chemistry as described in Reyer et al. 1 .
The plot distribution applied in our study was not homogeneous, with lower sampling density in Mediterranean areas and South-Eastern Europe, representing majorly forests in higher elevations. Nevertheless, as we applied a country and species-specific approach, our analysis was performed exclusively for sampled species. Moreover, the simulated forest development of the plots in Mediterranean areas obtained through 4C is compatible with values reported in the literature 2,3,4 .
In order to calculate the economic outcomes of forest management for the various species and countries we derived the harvesting revenues and costs for each species in each country from the EFISCEN model database. We considered static wood prices during the simulation period, given the stability of real sawn wood prices in Europe during the past 30 years 5 . Forest areas were derived from the evaluation performed by Brus et al. 6 , the EFISCEN inventory database and national forest inventories of each country. Tables 1, 2, and 3 illustrate the allocation of plots, coverage area, and prices to different European countries and tree species, respectively. S1- Figure 1. Map of forest plots used in this study. The model 4C was initialized for each plot. Figure 1  Long term interest rates S1- Table 4 shows the long-term interest rates reported by the European Central Bank as an average of the long-term interest rates from November/2014 to November/2015. The data for Norway and Switzerland was retrieved from the OECD database.

Climate
The climate change scenarios were taken from the following combinations of regional climate models and general circulation models: CCLM/ECHAM5 (CCLM_A1B), HadRM3/HadCM3 (HAD_A1B) and HIRHAM3/Arpège (HIR_A1B) driven by the A1B emission scenario. This emission scenario assumes a medium-high emission level, with economic growth, rapid technology development, and a material-intensive lifestyle 9 so that a climate target of maximum 2°C global surface temperature warming at the end of century is exceeded. In addition, we used the CCLM/ECHAM5 model driven by the B1 emission scenario (CCLM_B1) whose GCM forcing corresponds nearly to the 2°C target. The B1 emission scenario considers changes in economic structures, with reduction in material intensity and introduction of resource-efficient technologies 9 . S1- Figure 2 shows the timeline of the average temperature anomalies of the four climate scenarios for the plots in each country and S1- Figure 3 illustrates the pathway of CO2 concentration increase for both A1B and B1 scenarios over time (2010-2090).

The 4C model
Process-based forest models describe forest dynamics in detail, based on controlling processes at tree, soil and atmosphere level 10,11 . Given the capacity of these models to capture carbon, nitrogen and water cycles, they are most suitable to assess the forest ecosystems responses to new climatic conditions 12 . Here, we applied the process-based model 4C (http://www.pik-potsdam.de/4c/), for analyzing forest responses in terms of wood production and carbon sequestration for different species in Europe and under different climate change scenarios. S1- Figure 4. Scheme of the model 4C with the main processes, fluxes and pools considered in this study. 4C is a process-based forest model that simulates forest responses to changing climatic conditions. The model is capable of simulating forest structure, LAI, carbon and water balance and various management interventions, including harvesting, thinning and regeneration 13 . It describes processes on tree-and stand-level based on eco-physiological experiments, long term observations and physiological modeling (see S1- Fig. 4). The trees of the forest stand are aggregated into cohorts and their establishment, growth and mortality are explicitly modeled for each cohort on the stand level assuming horizontal homogeneity whithin each cohort.
The annual course of net photosynthesis was simulated with a mechanistic formulation of net photosynthesis as a function of environmental influences (temperature, water and nitrogen availability, radiation and CO2). The physiological capacity (maximal carboxylation rate) was calculated on the basis of optimization theory (modified after Haxeltine and Prentice 14 ), plus calculation of total tree respiration following the concept of constant annual respiration fraction as proposed by Landsberg and Waring 11 . The allocation pattern of annual net primary production (NPP) to the tree organs and tree growth were modeled with a combination of pipe model theory 15 , the functional balance hypothesis 16 and several allometric relationships extended to respond dynamically to water and nutrient limitations. The start and end of the vegetation period were estimated as functions of air temperature and length of the day 17 . Water and nitrogen availability, which affect growth and mortality of trees, depend on the soil parameters, the climatic conditions, and the stand development. Mortality depends explicitly on the carbon balance of the tree cohorts and failure to reproduce foliage over several years increases mortality probability. Mortality due to disturbances has not yet been modeled. The water balance was calculated from potential evapotranspiration according to Turc/Ivanov 18 , interception, and percolation Transport of water in the multi-layered soil was calculated with a daily time step 19 by a simple percolation model and controlled by a modelspecific water conductivity parameter 20, 21 depending on the soil texture. Root uptake is limited by the transpiration demand of all trees and the plant available water. The tree cohorts compete for water and nitrogen and satisfy their demand layer by layer through their fine roots in proportion to the fine root mass of all cohorts in the respective layer. Currently, the model is parameterized for the five most abundant tree species of Central Europe (beech (Fagus sylvatica L.), Norway spruce (Picea abies L. Karst.), Scots pine (Pinus sylvestris L.), oaks (Quercus robur L., and Quercus petraea Liebl.), and birch (Betula pendula Roth)) as well as for other tree species. We applied 4C to examine the response of forest ecosystems to climate change, in terms of the total carbon budget in the ecosystem, wood productivity, and harvesting volume. The total carbon in the ecosystem represents a sum of the total carbon in biomass (above and belowground) and the total carbon in soil. Although the development of Quercus robur and Quercus petraea were modelled identically in 4C, the economic analysis for these species was performed separately due to the distinct wood prices and forest cover in each country. Thus, country-specific conditions for the implementation of mitigation strategies by sequestering carbon in forest biomass were defined.
4C has been evaluated across Europe using long-term forest growth data eddycovariance flux measurements, daily transpiration and soil water content as well as a longterm dataset on annual tree ring increments at different temporal scales 1,22,23,24,25 .

Response to climate change
The impacts of climate change on forest productivity forecasted by 4C are extremely positive in Europe (Reyer et al. 1 ). Changes in NPP forecasted by 4C, according to Reyer et al. 1 show that under increasing CO2 concentration (Persistent CO2 effects), forest productivity increases across Europe, especially for Boreal regions. At the species level, the model indicates a NPP change of 0 to 0.4 Mg of C/year for Picea abies, -0.2 to 3.9 Mg of C/year for Pinus sylvestris, -0.5 to 3.1 Mg of C/year for Fagus sylvatica and -0.9 to 3.7 Mg of C/year for Q. petraea and Q. robur 1 . With constant CO2 concentration (Acclimation CO2 effects) the climate change impacts are predominantly positive in Scandinavia, with mixed effects in central Europe, whereas in Mediterranean areas negative impacts were observed.