Biogeophysical impacts of peatland forestation on regional climate changes in Finland

Biogeosciences, 11, 7251–7267, 2014 www.biogeosciences.net/11/7251/2014/ doi:10.5194/bg-11-7251-2014 © Author(s) 2014. CC Attribution 3.0 License. Biogeophysical impacts of peatland forestation on regional climate changes in Finland Y. Gao1,2, T. Markkanen1, L. Backman1, H. M. Henttonen3, J.-P. Pietikäinen1, H. M. Mäkelä1, and A. Laaksonen1,4 1Finnish Meteorological Institute, P.O. Box 503, 00101 Helsinki, Finland 2University of Helsinki, Department of Physics, P.O. Box 64, 00014 Helsinki, Finland 3Finnish Forest Research Institute, P.O. Box 18, 01301 Vantaa, Finland 4University of Eastern Finland, Department of Applied Physics, P.O. Box 1627, 70211 Kuopio, Finland Correspondence to: Y. Gao ([email protected]) Received: 11 June 2014 – Published in Biogeosciences Discuss.: 22 July 2014 Revised: 10 November 2014 – Accepted: 12 November 2014 – Published: 17 December 2014 Abstract. Land cover changes can impact the climate by influencing the surface energy and water balance. Naturally treeless or sparsely treed peatlands were extensively drained to stimulate forest growth in Finland over the second half of 20th century. The aim of this study is to investigate the biogeophysical effects of peatland forestation on regional climate in Finland. Two sets of 18-year climate simulations were done with the regional climate model REMO by using land cover data based on pre-drainage (1920s) and postdrainage (2000s) Finnish national forest inventories. In the most intensive peatland forestation area, located in the middle west of Finland, the results show a warming in April of up to 0.43K in monthly-averaged daily mean 2m air temperature, whereas a slight cooling from May to October of less than 0.1K in general is found. Consequently, snow clearance days over that area are advanced up to 5 days in the mean of 15years. No clear signal is found for precipitation. Through analysing the simulated temperature and energy balance terms, as well as snow depth over five selected subregions, a positive feedback induced by peatland forestation is found between decreased surface albedo and increased surfaceair temperatureinthe snow-melting period. Ourmodelled results show good qualitative agreements with the observational data. In general, decreased surface albedo in the snow-melting period and increased evapotranspiration in the growing period are the most important biogeophysical aspects induced by peatland forestation that cause changes in climate. The results from this study can be further integrally analysed with biogeochemical effects of peatland forestation to provide background information for adapting future forest management to mitigate climate warming effects. Moreover, they provide insights about the impacts of projected forestation of tundra at high latitudes due to climate change. 1 Introduction Climate response to anthropogenic land cover change happens more locally and occurs on a much shorter time scale compared to global warming due to increased greenhouse gases (GHG) (IPCC, 2013). The influences on the climate from the biogeophysical effects caused by land cover changes can enhance or reduce the projected climate change (Bathiany et al., 2010; Bonan, 2008; Feddema et al., 2005; Gálos et al., 2011; Göttel et al., 2008; Ge and Zou, 2013; Pielke et al., 2011, 1998; Pitman, 2003). Especially for the climate impacts of past large-scale afforestation, studies show that the most obvious effects of the increase of forests in boreal areas are warming during snow-cover periods due to decreased surface albedo and cooling in summertime from increased evapotranspiration (ET) in tropical areas with sufficient soil moisture (Bala et al., 2007; Betts, 2000; Betts et al., 2007). Vast areas of naturally treeless or sparsely treed peatlands have been drained to grow forests for timber production in northern European countries (Päivänen and Hånell, 2012). In Finland, it is the dominant land cover change over the last half century due to the high fraction of pristine peatland and the need for timber production. The total peatland area of Finland was estimated to Published by Copernicus Publications on behalf of the European Geosciences Union. 7252 Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland be 9.7millionha in the 1950s (Ilvessalo, 1956). In the beginning of the 2000s, the area of drained peatland for forestry was estimated to be 5.7millionha by Minkkinen et al. (2002) and 5.5millionha by Tomppo et al. (2011). The area of drained peatlands is unlikely to increase further because no more public subsidisation is given for the first-time drainage of peatlands and the increased awareness of natural conservation (Metsätalouden kehittämiskeskus Tapio, 1997). The area of restored mires was 15000ha between 1990 and 2008 (http://www.biodiversity. fi/en/indicators/mires/mi17-mire-restoration) (Kaakinen and Salminen, 2006). However, land cover change is not only a result of human land-use activities but can also be a consequence of climate change. Global warming in the future is also considered to be a factor that affects boreal peatland through water-level drawdown due to increased ET (Laiho et al., 2003; Laine et al., 1995). Attention has been paid to the climate effects of peatland forestation. A decrease in the local night-time minimum temperature during the growing season was observed roughly for the first 15years after drainage (Solantie, 1994). The reason for this nocturnal cooling phenomenon is the insulation of lower soil layers from the atmosphere by dry peat. Therefore, the heat flux from drained peat soil can not compensate for the radiative cooling at the surface, which leads to a drop in daily minimum temperature (Venäläinen et al., 1999). On alongertime scale,thegrowingforest on formerlyopen peatlands leads to a decrease in surface albedo. The reasons for this are the darker tree cover in comparison to the lighter moss/grass cover in the snow-free period and the partial snow cover in forest areas compared to the full snow cover in open areas in the snow-cover period. This increases the daily maximum temperature due to an increase in the absorption of short-wave radiation (Solantie, 1994). Consistent results on the seasonal cycles of surface albedo and net surface solar radiation due to peatland forestation were found by Lohila et al. (2010), based on measurement data at two pairs of drained and undrained peatland sites located in the south and north of Finland. The results showed a notably decreased surface albedo and corresponding increased net surface solar radiation in springtime. Furthermore, Lohila et al. (2010) indicated the local climate impacts of peatland forestation by investigating long-term (1961–2008) spring surface temperature trends over southern (<65◦N) and northern (>65◦N) Finland. The largest positive daytime maximum temperature trend of 0.64Kdecade−1happened in April in southern Finland, where a total of 2.7millionha of peatlands were drained (Hökkä et al., 2002). The night-time minimum temperature trend through the same period was 0.37Kdecade−1. Lohila et al. (2010) attributed the substantially larger increase in the daytime maximum temperature than in the night-time minimum temperature to the change in surface radiative properties after drainage. However, these studies about the effects of peatland forestation on climate are based on site-level data or observation-based regional data, which can not attribute the climate impacts to different influencing factors. Specifically, they can not distinguish the local biogeophysical effects from the global climate change due to the increase in GHG concentrations. The climate effects of peatland forestation have not been quantified on a regional scale/country level to investigate the biogeophysical effects in particular. Also, the magnitude and pattern of land-use change effects on climate depend on regional conditions such as soil property, topography, etc. Information from regional studies is essential for the development of future strategies for climate mitigation or forest management. Thus, it is necessary to investigate the effects regionally and systematically. In recent years, regional climate models have become suitable for simulating regional climate in a fine resolution to resolve small-scale atmospheric circulation (Déqué et al., 2005; Jacob et al., 2001, 2007; McGregor, 1997). For this, a regional climate model with a realistic land scheme to interpret more detailed land surface information needs to be applied. In this study, the long-term climate effects caused by peatland forestation are assessed from two sets of 15-year simulation results with the regional climate model REMO, by using the historical (1920s) and present-day (2000s) land cover conditions. The intention of this study is to understand how peatland forestation in Finland influences regional climate conditions through biogeophysical processes. 2 Model description and methodology 2.1 REMO climate model The regional climate model REMO is a three-dimensional hydrostatic atmospheric circulation model developed at the Max Planck Institute for Meteorology in Germany (Jacob et al., 2001, 2007; Jacob and Podzun, 1997). Its dynamical core is based on the “Europa-Modell”, the former numerical weather prediction model of the German Weather Service (Majewski, 1991). The land surface scheme (LSS) of REMO mainly follows that of the global atmosphere circulation model ECHAM4 (Roeckner et al., 1996) with several physical package updates (details are shown below). The prognostic variables are pressure, temperature, horizontal wind components, specific humidity, cloud liquid water and ice. REMO is driven by large-scale forcing data according to the relaxation scheme (Davies, 1976). The eight outermost grid boxes at each lateral boundary are the sponge zone. Because land cover is central to this study, a brief introduction of the LSS in REMO is given below. In REMO LSS, the total area of each model grid box is composed of fractions of land (vegetation cover and bare soil), water (ocean surface and inland lake) and sea ice (Semmler et al., 2004). The biogeophysical characteristics of major land cover classes (Olson, 1994a, b) are described by the following surface Biogeosciences, 11, 7251–7267, 2014 www.biogeosciences.net/11/7251/2014/ Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland 7253 parameters: background surface albedo (albedo over snowfree land areas), roughness length, fractional green vegetation cover, leaf area index (LAI; one-sided green leaf area per unit ground area), forest ratio (fr; fractional coverage of trees regardless of their photosynthetic activity), soil waterholding capacity (maximum amount of water that plants may extract from the soil before wilting begins) and volumetric wilting point (percentage of moisture in a soil column below which plants start to wilt) (Hagemann, 2002; Hagemann et al., 1999). The land surface parameters are averaged linearly according to fractional coverage of land cover types within a model grid box, except for the roughness length that isaveragedlogarithmically (Claussenet al., 1994; Hagemann et al., 1999). As LAI, fractional green vegetation cover and background surface albedo strongly depend on the vegetation phenology, they are prescribed with intra-annual cycles by using a monthly varying growth factor that determines the seasonal growth characteristics of the vegetation (Hagemann, 2002; Rechid and Jacob, 2006). The growth factor for latitudes higher than 40◦north or south is derived from a 2m temperature climatology (Legates and Willmott, 1990); in other latitudes, the fraction of photosynthetically active radiation is used. The simple bucket scheme (Manabe, 1969) is used for soil hydrology where the partitioning of surface runoff and infiltration follows the Arno scheme (Dümenil and Todini, 1992). The soil temperature profile from the ground surface to around 10m deep is described by five soil layers with increasing thickness. The heat conductivity and heat capacity, required in the heat conduction equation for calculating the soil temperature, depend on the soil types (Kotlarski, 2007). The distribution of soil types is from the FAO/UNESCO soil map of the world (FAO/UNESCO, 1971–1981; Kotlarski, 2007). The Arno scheme used for the soil hydrology was further improved by considering the high resolution subgridscale heterogeneity of the field capacities within a climate model grid box (Hagemann and Gates, 2003). The resolution of subgrid-scale heterogeneity is set to be 10 times higher than the model resolution when using the default REMO land cover map-Global Land Cover Characteristics Database (GLCCD) (Loveland et al., 2000; US Geological Survey, 2001). The three parameters in the improved Arno scheme account for the shape of the subgrid distribution of soil water capacities (Beta), subgrid minimum (Wmin)and maximum (Wmax)soil water capacities. Also, the original annual background albedo cycle was modified by using MODIS satellite data between 2001 and 2004 in order to derive more realistic global distributions of pure soil albedo and pure vegetation albedo, which are then used to compute the annual background albedo cycle with monthly varying LAI (Rechid, 2008; Rechid et al., 2009). Figure 1. Orography of the model domain and the five selected subregions (subregion1 – blue; subregion2 – red; subregion3 – purple; subregion4 – green; subregion5 – orange). The inner black frame shows the extent of the relaxation zone from the outer boundary, i.e. the eight outer-most grid boxes in each direction of the model domain. 2.2 The model domain and land cover data sets Our model domain covers Fennoscandia, a part of Russia and the northern part of central Europe, and it is centred on Finland (Fig. 1). Typical features influencing the climate of this domain include the North Atlantic Ocean and the Baltic Sea that surround the Fennoscandian countries, many inland lakes located in Sweden and Finland and the relatively high Scandinavian mountain range; the rest of the area has a topography lower than 300m above sea level. The default land cover map in REMO is the GLCCD. However, its description of the land cover in Finland is unrealistic. For instance, there is no peatland in Finland in the GLCCD, whereas 7.4% (22377km2) of the land is covered by naturally treeless or sparsely treed peatlands according to the 10th Finnish national forest inventory (FNFI10) (Korhonen et al., 2013). The GLCCD was therefore substituted by the more realistic and up-to-date CORINE land cover map (CLC; 2006) for the same model domain in Gao et al. (2014), except for the Russian part where the CLC (2006) is not available. Unfortunately, land cover maps describing the land cover conditions of Finland before the most intensive period www.biogeosciences.net/11/7251/2014/ Biogeosciences, 11, 7251–7267, 2014 7254 Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland Figure 2. Changes of fractional coverage of the 10 land cover classes in Finland from the 1920s to the 2000s (FNFI10–FNFI1). of peatland drainage in the 1960s are quite limited. Nevertheless, the data collected in the 1st Finnish national forest inventory (FNFI1) provide the possibility for tracing back the land cover condition of Finland in the 1920s (Ilvessalo, 1927; Tomppo et al., 2010). Also, the FNFI10, rather than the CLC (2006),is adopted todescribe the landcovercondition of Finland in the 2000s, with the aim to avoid the uncertainties in comparing land cover maps with different land cover classification methods and different spatial resolutions. The FNFI1 and FNFI10 land cover maps are post-products that were specially prepared for this study from the respective FNFI field measurementdata. The detailed descriptionof the procedures for deriving the FNFI1 and FNFI10 land cover maps is shown in Appendix A. The two FNFI land cover maps are in 3km resolution and include 10 land cover classes following CLC nomenclature. The fractional coverage for the 10 land cover classes over the land area of Finland in the 1920s and the changes from the 1920s to the 2000s based on the two FNFI land cover maps are as follows (fractional coverage in the 1920s; changes from the 1920s to the 2000s): coniferous forest (33.0%; 5.2%); mixed forest (13.5%; −5.7%); broadleaved forest (4.7%; −0.8%); artificial areas (0.7%; 4.1%); natural grasslands (3.4%; −3.4%); peat bogs (14.3%; −5.2%); open spaces (1.5%; −0.1%); transitional woodland/shrub (18.9%; 4.3%); moors and heathland (2.1%; 0.7%); and agricultural areas (8.0%; 0.9%). Regional differences of those land cover classes can be seen in Fig. 2. In the FNFI maps, the land cover class ”peat bogs” is defined as naturally treeless peatland and pine mires where the stocking level is low or the mean height of trees is below 5m at maturity. Therefore, the shifting from peat bogs to forests represents a major land cover change due to peatland forestation. In addition to regional inspections, five subregions were selected to represent different land cover change conditions between FNFI1 and FNFI10 (Fig. 1), and the changes of fractional coverage of the 10 land cover classes in those five subregions are given in Table 1. This was done to specifically assess the local climate effects of different intensities of peatland forestation. From subregion1 to subregion4 there is a decrease in the reduction of peat bogs. Subregion1 and subregion2 are two peatland forestation areas located in the middle and south of Finland respectively. In subregion1 and subregion2 there were decreases in the fractional coverage of peat bogs of more than 20%, and the decreases were mainly compensated by coniferous forest. The decrease in the fractional coverage of peat bogs was 2% less in subregion2 than that in subregion1, but the increase in the fractional coverage of coniferous forest was 5% higher in subregion2 than that in subregion1. The total increase in the fractional coverage of forest types was about 16% in both subregion1 and subregion2. Subregion3 is located in the east of subregion1. There was a 12% decrease in the fractional coverage of peat bogs, but instead of an increase of forests, the fractional coverage of transitional woodland/shrub increased by 14.3%. SubBiogeosciences, 11, 7251–7267, 2014 www.biogeosciences.net/11/7251/2014/ Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland 7255 Table 1. Changes of fractional coverage (%) of the 10 land cover classes from the 1920s to the 2000s (FNFI10–FNFI1) in the five subregions. Class Legend Subregion1 Subregion2 Subregion3 Subregion4 Subregion5 1 coniferous forest 13.40 18.03 −2.24 −11.74 −10.13 2 mixed forest 1.23 −3.46 −2.30 −1.86 −2.10 3 broad-leaved forest 1.24 0.98 1.68 −0.52 −4.11 4 artificial areas 4.44 4.95 2.44 5.69 2.52 5 natural grasslands −4.41 −2.10 −1.71 −2.82 −1.60 6 peat bogs −22.92 −20.82 −12.60 −3.80 8.64 7 open spaces 0.06 −0.12 −0.11 −0.31 −1.14 8 transitional woodland/shrub 3.64 −0.72 14.26 4.84 9.12 9 moors and heathland 0.00 0.00 0.00 0.00 −1.37 10 agricultural areas 3.31 3.26 0.57 10.52 0.17 region4 is an area where the most intensive anthropogenic activities have occurred in the five subregions. There was a 14% decrease in the fractional coverage of forest types and a 3.8% decrease in that of peat bogs, with a 5.7% increase in the fractional coverage of artificial areas and a 10.5% increase in that of agriculture areas. Subregion5 is an area with an 8.64% increase in the fractional coverage of peat bogs and a 16.3% decrease in the fractional coverage of forest types. Herein one should notice that some uncertainties may arise from sampling in the FNFI1 and FNFI10 data. This applies especially to FNFI1, where the distance between inventory lines was as high as 26km. Therefore, subregions that are smaller than 100km×100km may not be sufficient to represent the actual land cover changes spatially. However, the dynamics of the local effects of land cover changes on climate can not be detected when averaging climate signals over large areas with diverse land cover changes. Therefore small subregions, which cover a range of land cover change intensities, are chosen to reflect local climate impacts due to different land cover changes. Moreover, the FNFI data only cover the land surface in Finland without considering inland lakes. Therefore, the land–sea mask in the model domain is adopted from the CLC (2006). In addition, the land cover conditions of the area outside Finland in the model domain are the same as those in Gao et al. (2014), i.e. based on the CLC (2006) and the GLCCD, and thus identical in both simulations. In order to make the land surface parameters more suitable for this study, several modifications in REMO LSS were done. Details of those modifications are documented in Appendix B. 3 Experiment design Two simulations were conducted with the FNFI1 and FNFI10 land cover maps, representing the land cover conditions before and after peatland forestation activities in Finland respectively. The simulations were driven with 6-hourly lateral boundary conditions from ECWMF ERA-Interim reanalysis data (Simmons et al., 2007) from 1 January 1979 to 31 December 1996. The 18-year forward runs were preceded by 10-year (1 August 1979–1 January 1990) simulations in order to stabilise the deep soil temperatures and soil moistures. The last 15 years (1 December 1981– 30 November 1996) out of the 18-year forward simulations were adopted for further analysis. The analysed period starts from 1 December in order to keep all 3 winter months continuous. The simulated first 1.5 years were excluded in order to minimise the influences of the initial boundary conditions on simulated climate conditions, which have a much quicker adaptation speed than deep soil temperature. The model grid is in an 18km resolution horizontally and extends over 27 vertical levels (up to 25km). The model time step was set to 90s and the time steps of output variables are 6-hourly for 3D variables and hourly for 2-D variables. Daily data covering 24h are processed from 18:00UTC on the previous day to 17:00UTC on the current day. For 6-hourly data, 18:00UTC on the previous day and 00:00UTC, 06:00 and 12:00UTC on the current day were used for daily values. For this study domain, the growing season and the dormancy season cover the period from May to October and from November to April respectively. 4 Results The land cover change effects on regional climate conditions in Finland are analysed based on the differences in climate variables between the post-drainage and pre-drainage simulations (FNFI10–FNFI1). This “delta change approach” is adopted to eliminate the uncertainties related to model bias (Gálos et al., 2011; Jacob et al., 2008). 4.1 Effects on climate over Finland The differences in monthly-averaged daily mean 2m air temperature (T2m) are quite heterogeneous temporally and spatially. T2m differences are most prominent in springtime and summertime (Fig. 3). The most noticeable difference in T2m, up to 0.43K, takes place in the most intensive peatland www.biogeosciences.net/11/7251/2014/ Biogeosciences, 11, 7251–7267, 2014 7256 Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland Figure 3. The 15-year averaged differences (FNFI10–FNFI1) in monthly-averaged daily mean 2m air temperature in spring and summer months. forestation area in the middle west of Finland in April. The warming is also evident in February and March, with differences of 0.2K in this area. However, T2m turns to show a slight cooling, generally less than 0.1K, in a few parts of this area from May to October. There are also two regions in northern Finland that show opposite changes compared to the peatland forestation area in the middle west of Finland with cooling in the spring and warming in the growing season. This is because of decreased forest cover and increased fraction of peat bogs in those two areas from FNFI1to FNFI10-based land cover maps. An increase of less than 0.2K is seen in T2m in the southeast of Finland in July and August as well as in the very south of Finland throughout the growing season, which is mainly due to the change from mixed forest to coniferous forest and the increased artificial areas respectively. The 15-year averaged monthly precipitation shows only small differences, less than 10mmmonth−1, in varied patterns in the model domain from April to August (not shown). The snow clearance day is also an important indicator of springtime climate change at high latitudes (Peng et al., 2013). Therefore, the snow clearance day for each grid box in Finland is determined for the 15years. The snow clearance day is defined here as the first day after which the total number of snow-covered days does not exceed the total number Figure 4. The 15-year averaged differences (FNFI10–FNFI1) in the snow clearance days over model grid boxes in Finland. of snow-free days, and the selection of this day ends before midsummer in a year. The differences between the 15-year averaged snow clearance days of the two simulations (Fig. 4) show almost the same pattern as the differences in T2m in April (Fig. 3). In the peatland forestation area in the middle west of Finland, the snow clearance days are mostly advanced by 0.5 to 3 days and, in a few grid boxes, advanced by up to 5 days in the 15-year mean. The two small areas in the north of Finland with reverse land cover changes in comparison to peatland forestation show up to 2-day delays in general. In the very south of Finland, the snow clearance days are also generally advanced in accordance with the warming seen in T2m, but delayed in several scattered grid boxes due to increased fraction of artificial areas at the expense of forests. 4.2 Effects on climate over five subregions T2m, precipitation and several closely related climate variables (surface albedo, net surface solar radiation, snow depth, ET) for the five subregions were processed into 11-day running means to reduce the influence of day-to-day variations. The differences between the simulations in each of the regionally averaged climate variables were further averaged over the 15years (Fig. 5). The date information herein (day of year, DOY) represents the middle contributing day of the 11-day averaging period. T2m of subregion1 shows a warming of 0.1 to 0.2K from February until the end of March and an evident peak of increase from early April to early May (from DOY 95 to DOY 125) that reaches a maximum of 0.5K in late April. Biogeosciences, 11, 7251–7267, 2014 www.biogeosciences.net/11/7251/2014/ Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland 7257 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 0 60 120 180 240 300 360 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 (e) DOY ET (mm) 0 60 120 180 240 300 360 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 (a) DOY T2m (K) 0 60 120 180 240 300 360 -0.02 -0.02 -0.01 -0.01 0 0.01 0.01 0.02 (b) DOY snow depth (m) Subregion1 Subregion2 Subregion3 Subregion4 Subregion5 0 60 120 180 240 300 360 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 (c) DOY albedo 0 60 120 180 240 300 360 -6 -4 -2 0 2 4 6 8 (d) DOY net surface solar radiation(W/m2) 0 60 120 180 240 300 360 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 (f) DOY precipitation (mm) Figure 5. The 15year averaged regional mean differences (FNFI10 – FNFI1) in 11 day running mean of daily mean (a) two-metre air temperature, (b) snow depth (presented as equivalent water), (c) surface albedo, (d) net surface solar radiation, (e) ET and (f) precipitation of the five subregions. 42 Figure 5. The 15-year averaged regional mean differences (FNFI10 – FNFI1) in 11-day running mean of daily mean (a) 2m air temperature, (b) snow depth (presented as equivalent water), (c) surface albedo, (d) net surface solar radiation, (e) ET and (f) precipitation of the five subregions. T2m of subregion2 has the same development as subregion1 throughout the whole year, but the warming is much smaller and the biggest difference, only 0.12K, occurs in the beginning of April. This is consistent with the differences in snow depth. The snow-cover period in subregion2 is shorter along with an earlier maximum difference in snow depth. Moreover, those characteristics of the differences in snow depths are in qualitative agreement with the differences in surface albedo because snow is the key factor that controls the surface albedo in the snow-cover period. From the beginning of May to the beginning of October, T2m shows a cooling of less than 0.1K in subregion1 and subregion2 because the cooling caused by ET exceeds the warming caused by the slightly lower albedo. The variability of the differences in net surface solar radiation in the growing season is induced by the variability of cloud cover rather than surface albedo. In November, December and January, the differences in T2m vary in both directions. At high latitudes, incoming solar radiation is quite small and cloud cover fraction is high in late autumn and winter. Therefore, the differences in surface albedo are not able to induce differences in net surface solar radiation in this period. Instead, the surface air temperature is sensitive to changes in the long-wave radiation balance that may lead to atmospheric air temperature inversion under a clear sky, manifesting itself as extreme cold surface air temperature. Thus, the variability of the differences in cloud cover caused by short-term variations in the climate contributes to the varied differences in T2m in this period. The differences in T2m for subregion3 show a warming of less than 0.1K from DOY 91 to DOY 120 but also a warming in an even smaller magnitude throughout the growing season. The difference in surface albedo in subregion3 is close to 0, although the difference in snow depth is similar to that of subregion2 but with a time lag of around 15 days in the most intensive point. In subregion4, the snow depth shows a quite small increase from the beginning of January until the end of March, which is consistent with the increase in surface albedo and explains the slight decrease of up to 0.1K in T2m from the middle of February until the end of March. Subregion5 displays the opposite characteristics compared to subregion1 and subregion2 for all the investigated variables. The absolute differences in snow depth of subregion5 are smaller than those of subregion1 but larger than those of subregion2. Because subregion5 is located in the north of Finland, the biggest difference in snow depth occurs later than that of subregion1. The magnitude of the maximum differences in T2m in the snow-cover period of subregion5 also lies between that of subregion1 and subregion2 and happens later than that of subregion1. The differences in T2m in the growing season depend on the surplus of energy balance terms where ET manifests itself as latent heat flux. In general, the increase of ET in subregion2 is slightly higher than that in subregion1. As a consequence, the decrease of T2m in subregion2 is slightly larger than that in subregion1 during the growing season when the albedo difference is quite small. The decreased ET and the slightly decreased surface albedo together result in a slight warming during the growing season in the other subregions. The extents of warming in the other subregions follow the magnitudes of the decreased ET because the differences in surface albedo are almost the same in the growing season. Precipitation has higher variability than ET throughout the year in the five subregions. In general, the differences in precipitation are much larger in the growing season than in the dormancy season, when they are close to 0mmday−1. In the growing season, the increase in precipitation of subregion1 occurs during a longer period and has a larger magnitude than that of subregion2. There are slight increases in the precipitation in subregion3 and subregion4, whereas the precipitation of subregion5 shows a decreasing tendency in the growing season, with the biggest differences less than 0.2mmday−1. Furthermore, the maximum and minimum differences of grid pointwise and regionally averaged 11-day running mean of T2m over 15years for subregion1 were investigated as complements to the regionally averaged 15-year mean differences (Fig. 6). T2m shows a maximum difference in grid pointwise of nearly 2K in the snow-melting period over the 15years, which is 1K higher than the maximum difference www.biogeosciences.net/11/7251/2014/ Biogeosciences, 11, 7251–7267, 2014 7258 Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 0 60 120 180 240 300 360 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 DOY T2m (K) Mean Min_Grid Max_ Grid Max_Subregion Min _Subregion ∇ Figure 6. Maximum, minimum and mean differences of gridpoint-wise and regionally averaged 11 day running mean of daily mean two-metre air temperature over 15 years in subregion1. 43 Figure 6. Maximum, minimum and mean differences of grid pointwise and regionally averaged 11-day running mean of daily mean 2m air temperature over 15years in subregion1. in regionally averaged T2m over the 15years and 4 times that of the 15-year mean of regionally averaged T2m. The timings of the three kinds of maximum differences in spring deviate from each other by 3 to 10 days. The minimum differences show only a small deviation between the grid pointwise and regional mean values over the 15years. During the snowmelting period, the minimum differences of regionally averaged T2m is above 0, but not that of the grid pointwise T2m. The springtime differences between regional mean and grid pointwise extremes elucidate that, even within one subregion with homogenous characteristics related to peatland forestation, the spring warming of T2m is temporally and spatially heterogeneous. This implies that local effects are more pronounced than the regional and temporal statistics can reveal. For the rest of the year, the differences between the maximum (minimum) of the grid pointwise and regionally averaged T2m are small and of a more regional nature. In the period between November and January, the large variations of maximum (minimum) T2m are contributed by the inversion effects due to short-term variations in the climate. Additionally, for a more thorough understanding of the relationships between spring warming and albedo changes in the snow-cover period due to peatland forestation, two correlation relationships were investigated over the 15years for subregion1 (Fig. 7). One is between the maximum temperature difference day (DOY) and the maximum surface albedo difference day (DOY). The other is between the inflection day of total albedo (the day when surface albedo just finishes a fast decrease from its wintertime level; DOY) and the snow clearance day (DOY). The maximum temperature difference days match to maximum albedo difference days in 6years, and the rest of the years generally show a delayed maximum temperature difference day compared to the maximum albedo difference day, with a maximum deviation of 14 days. In general, the snow clearance day correlates well Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 100 110 120 130 140 150 100 110 120 130 140 150 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 (b) Inflection point of total albedo (DOY) Selected snow clearance day (DOY) 70 80 90 100 110 120 130 70 80 90 100 110 120 130 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 19951996 (a) Maximum temperature change (DOY) Maximum total albedo change (DOY) Figure 7. (a) Correlation between maximum temperature change day (DOY) and maximum total albedo change day (DOY); (b) correlation between inflection day of total albedo (the day when surface albedo just finishes a fast decrease from its wintertime level; DOY) and the snow clearance day (DOY). The plots show regional means over subregion1 for all 15 years. 44 Figure 7. (a) Correlation between maximum temperature change day (DOY) and maximum total albedo change day (DOY); (b) correlation between inflection day of total albedo (the day when surface albedo just finishes a fast decrease from its wintertime level; DOY) and the snow clearance day (DOY). The plots show regional means over subregion1 for all 15years. with the inflection point of surface albedo. For most years, the differences are less than 6 days, but 3 years show differences up to around 20 days. In those years, sporadic snowfall with a small accumulated snow depth cannot really introduce differences in total surface albedo over the subregion but influences the determination of the snow clearance day. 4.3 Relationships between the changes in biogeophysical aspects and the impacts on climate To assess the generality of the causal relationships between land cover changes and climate variables, the spatial correlations between changes in the two surface energy balance relevant variables, surface albedo and ET, and T2m are investigated. Consequently, the spatial correlations between changes in surface albedo and ET and changes in the surface parameter values are also explored. The correlations with fractional green vegetation cover is not shown Biogeosciences, 11, 7251–7267, 2014 www.biogeosciences.net/11/7251/2014/ Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland 7259 in Fig. 8 because LAI and green vegetation ratio are both modulated with the monthly varying growth factor by the same scheme, and they are highly correlated (Pearson correlation coefficient, r2=0.984 for March, r2=0.674 for June). Monthly means of 15-year averaged changes in March and June are selected to represent springtime and summertime respectively. The changes in T2m are in accordance with the changes in surface albedo in March (Fig. 8a), which is almost linearly correlated with the changes in LAI (Fig. 8c) and forest ratio (Fig. 8e). The changes in T2m are linearly correlated with the changes in ET over most of the area in June (Fig. 8b). In general, the changes in ET are also correlated with the changes in LAI (Fig. 8d), roughness length (Fig. 8f) and forest ratio (yearly constant, not shown), despite theinfluences from droughtsthatmay happen inlate summer. Overall, the changes in surface albedo and ET are closely dependent on the changes in land surface parameters, which are induced by the changes in fractional coverages of land cover types in the five subregions (Table 1). The changes in T2m are mainly modulated by the changes in surface albedo and ET in spring and summer respectively. Some grid boxes located in the southeast of Finland, where mixed forest was substituted by mainly coniferous forest, show deviations in the correlations with LAI (marked by yellow circles in Fig. 8b, c, d). In this area, LAI increased with almost no change in forest ratio, which led to a relatively smaller decrease in surface albedo compared to other areas with the same magnitude of changes in LAI in March; the ET-induced cooling is outweighed by the albedo-induced warming, which causes a slight warming in June. In the following summer months, July and August, the ET-induced cooling typically gets smaller because of surface water limitation and consequent warming. 5 Discussion 5.1 Biogeophysical impacts of peatland forestation on regional climate Surface albedo shows a notable decrease in peatland forestation areas during the snow-cover period and a slight decrease in the growing season, whereas LAI, roughness length, fractional green vegetation cover and forest ratio increase throughout the year after peatland forestation. Those changes lead to an increase in springtime T2m, which occurs locally in accordance with the decrease in surface albedo. In the growing season, an increase in ET related to the increased LAI and fractional green vegetation cover leads to more energy consumed by latent heat flux than gained by slightly lower albedo. Additionally, higher roughness length can play a role by increasing turbulent mixing and consequently the magnitudes of turbulent fluxes. Thus, the scattered differences in precipitation in summer are contributed to more convective structures, while for the rest of the year the precipitation is basically controlled by large-scale meteorology. From the Figure 8. Spatial correlations between (a) changes in monthlyaveraged daily mean 2m air temperature (T2m) and changes in albedo for March, (b) changes in T2m and changes in ET for June and relationships between changes in land surface parameters in REMO LSS following land cover changes and changes in albedo (c, e) (changes in ET, d,f) in the corresponding month. The changes in the grid boxes in selected subregions are shown with coloured dots (subregion1 – blue; subregion2 – red; subregion3 – purple; subregion4 – green; subregion5 – orange). The grid boxes in yellow circles show the changes in the southeast area of Finland. analysis of the results in the five subregions, the differences in the climate variables show that their magnitudes depend on the extent of land cover changes, while the timings of the extremesmostly depend ongeographical locations (latitudes) that define the radiation balance through the seasonal cycle. Results also illustrate a positive feedback induced by peatland forestation between lower surface albedo and warmer T2m in the snow-melting period. The warming caused by lower surface albedo in the snow-cover period due to more forest leads to a quicker and earlier snow melting; meanwhile, the surface albedo is reduced and consequently the surface air temperature is increased. Additionally, the maxiwww.biogeosciences.net/11/7251/2014/ Biogeosciences, 11, 7251–7267, 2014 7266 Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland forests of Finland. Results of the general survey of the forests of the country carried out during the years 1921–1924), Communicationes ex Instituto Quaestionum Forestalium Finlandiae 11, Valtioneuvoston kirjapaino, 1927. Ilvessalo, Y.: Suomen metsät vuosina 1921–24 vuosiin 1951–53: kolmeen valtakunnan metsien inventointiin perustuva tutkimus (the forests of Finland from 1921–24 to 1951–53. A survey based on three national forest inventories), Communicationes Instituti Forestalis Fenniae, Finnish Forest Research Institute, Helsinki, Finland, 47, 277 pp., 1956. IPCC: Climate Change 2013: The Physical Science Basis. Working Group 1 Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge, UK and New York, NY, USA, 1535 pp., 2013. Jacob, D. and Podzun, R.: Sensitivity studies with the regional climate model REMO, Meteorol. Atmos. Phys., 63, 119–129, 1997. Jacob, D., Van den Hurk, B. J. J. M., Andrae, U., Elgered, G., Fortelius, C., Graham, L. P., Jackson, S. D., Karstens, U., Köpken, Chr., Lindau, R., Podzun, R., Rockel, B., Rubel, F., Sass, B. H., Smith, R. N. B., and Yang, X.: A comprehensive model inter-comparison study investigating the water budget during the BALTEX-PIDCAP period, Meteorol. Atmos. Phys., 77, 19–43, 2001. Jacob, D., Bärring, L., Christensen, O. B., Christensen, J. H., De Castro, M., Déqué, M., Giorgi, F., Hagemann, S., Hirschi, M., Jones, R., Kjellström, E., Lenderink, G., Rockel, B., Sánchez, E., Schär, C., Seneviratne, S. I., Somot, S., Van Ulden, A., and Van den Hurk, B.: An inter-comparison of regional climate models for Europe: model performance in present-day climate, Climatic Change, 81, 31–52, doi:10.1007/s10584-006-9213-4, 2007. Jacob, D., Kotova, L., Lorenz, P., Moseley, C., and Pfeifer, S.: Regional climate modeling activities in relation to the CLAVIER project, Idöjárás, 112, 141–153, 2008. Kaakinen, E. and Salminen, P.: Mire conservation and its short history in Finland, in: Finland – land of mires, The Finnish Environment, 23, edited by: Lindholm, T. and Heikkilä, R., Finnish Environment Institute, Helsinki, Finland, 229–238, 2006. Køltzow, M.: The effect of a new snow and sea ice albedo scheme on regional climate model simulations, J. Geophys. Res.-Atmos., 112, D07110, doi:10.1029/2006JD007693, 2007. Korhonen, K. T., Ihalainen, A., Viiri, H., Heikkinen, J., Henttonen, H. M., Hotanen, J. P., Mäkelä, H., Nevalainen, S., and Pitkänen, J.: Suomen metsät 2004–2008 ja niiden kehitys 1921–2008 (the forests of Finland in 2004–2008 and their development from 1921 to 2008), Metsätieteen aikakauskirja, 3, Finnish Forest Research Institute, Helsinki, Finland, 269–608, 2013. Kotlarski, S.: A subgrid glacier parameterisation for use in regional climate modelling, PhD thesis, University of Hamburg, Max Planck Institute for Meterology, Hamburg, Germany, 2007. Kuusinen, N., Tomppo, E., and Berninger, F.: Linear unmixing of MODIS albedo composites to infer subpixel land cover type albedos, Int. J. Appl. Earth Obs., 23, 324–333, 2013. Laiho, R., Vasander, H., Penttilä, T., and Laine, J.: Dynamics of plant-mediated organic matter and nutrient cycling following water-level drawdown in boreal peatlands, Global Biogeochem. Cy., 17, 1053, doi:10.1029/2002GB002015, 2003. Laine, J., Vasander, H., and Laiho, R.: Long-term effects of water level drawdown on the vegetation of drained pine mires in southern Finland, J. Appl. Ecol., 32, 785–802, 1995. Legates, D. R. and Willmott, C. J.: Mean seasonal and spatial variability in global surface air temperature, Theor. Appl. Climatol., 41, 11–21, 1990. Lohila, A., Minkkinen, K., Laine, J., Savolainen, I., Tuovinen, J.-P., Korhonen, L., Laurila, T., Tietäväinen, H., and Laaksonen, A.: Forestation of boreal peatlands: impacts of changing albedo and greenhouse gas fluxes on radiative forcing, J. Geophys. Res.- Biogeo., 115, G04011, doi:10.1029/2010JG001327, 2010. Loveland, T. R., Reed, B. C., Brown, J. F., Ohlen, D. O., Zhu, Z., Yang, L., and Merchant, J. W.: Development of a global land cover characteristics database and IGBP DISCover from 1km AVHRR data, Int. J. Remote Sens., 21, 1303–1330, 2000. Majewski, D.: The Europa-Modell of the Deutscher Wetterdienst, in: ECMWF Seminar on numerical methods in atmospheric models, 2 (Vol.), Reading, UK, 147–191, 1991. Manabe, S.: Climate and the ocean circulation 1: I. The atmospheric circulation and the hydrology of earth’s surface, Mon. Weather Rev., 97, 739–774, 1969. McGregor, J.: Regional climate modelling, Meteorol. Atmos. Phys., 63, 105–117, 1997. Metsätalouden kehittämiskeskus Tapio: Metsätalouden säädökset (Forestry regulations), Tapio, Finland, 111 pp., 1997. Melton, J. R., Wania, R., Hodson, E. L., Poulter, B., Ringeval, B., Spahni, R., Bohn, T., Avis, C. A., Beerling, D. J., Chen, G., Eliseev, A. V., Denisov, S. N., Hopcroft, P. O., Lettenmaier, D. P., Riley, W. J., Singarayer, J. S., Subin, Z. M., Tian, H., Zürcher, S., Brovkin, V., van Bodegom, P. M., Kleinen, T., Yu, Z. C., and Kaplan, J. O.: Present state of global wetland extent and wetland methane modelling: conclusions from a model intercomparison project (WETCHIMP), Biogeosciences, 10, 753– 788, doi:10.5194/bg-10-753-2013, 2013. Minkkinen, K. and Laine, J.: Vegetation heterogeneity and ditches create spatial variability in methane fluxes from peatlands drained for forestry, Plant Soil, 285, 289–304, doi:10.1007/s11104-006-9016-4, 2006. Minkkinen, K., Korhonen, R., Savolainen, I., and Laine, J.: Carbon balance and radiative forcing of Finnish peatlands 1900–2100 – the impact of forestry drainage, Glob. Change Biol., 8, 785–799, 2002. Olson, J. S.: Global ecosystem framework-definitions, USGS EROS Data Center Internal Report, Sioux Falls, SD, 37 pp., 1994a. Olson, J. S.: Global ecosystem framework-translation strategy, USGS EROS Data Center Internal Report, Sioux Falls, SD, 39 pp., 1994b. Päivänen, J. and Hånell, B.: Peatland Ecology and Forestry: A Sound Approach, University of Helsinki Department of Forest Sciences Publication 3, Department of Forest Ecology, University of Helsinki, Helsinki, Finland, 2012. Pebesma, E. J.: Multivariable geostatistics in S: the gstat package, Comput. Geosci., 30, 683–691, 2004. Peng, S., Piao, S., Ciais, P., Friedlingstein, P., Zhou, L., and Wang, T.: Change in snow phenology and its potential feedback to temperature in the Northern Hemisphere over the last three decades, Environ. Res. Lett., 8, 014008, doi:10.1088/17489326/8/1/014008, 2013. Pielke, R. A., Avissar, R., Raupach, M., Dolman, A. J., Zeng, X., and Denning, A. S.: Interactions between the atmosphere and terrestrial ecosystems: influence on weather and climate, Glob. Change Biol., 4, 461–475, 1998. Biogeosciences, 11, 7251–7267, 2014 www.biogeosciences.net/11/7251/2014/ Y. Gao et al.: Biogeophysical impacts of peatland forestation on regional climate changes in Finland 7267 Pielke, R. A., Pitman, A., Niyogi, D., Mahmood, R., McAlpine, C., Hossain, F., Goldewijk, K. K., Nair, U., Betts, R., Fall, S., Reichstein, M., Kabat, P., and de Noblet, N.: Land use/land cover changes and climate: modeling analysis and observational evidence, Wiley Interdisciplinary Reviews: Climate Change, 2, 828–850, 2011. Pitman, A. J.: The evolution of, and revolution in, land surface schemes designed for climate models, Int. J. Climatol., 23, 479– 510, 2003. Pitman, A. J., de Noblet-Ducoudré, N., Cruz, F. T., Davin, E. L., Bonan, G. B., Brovkin, V., Claussen, M., Delire, C., Ganzeveld, L., Gayler, V., van den Hurk, B. J. J. M., Lawrence, P. J., van der Molen, M. K., Müller, C., Reick, C. H., Seneviratne, S. I., Strengers, B. J., and Voldoire, A.: Uncertainties in climate responses to past land cover change: First results from the LUCID intercomparison study, Geophys. Res. Lett., 36, L14814, doi:10.1029/2009GL039076, 2009. Preuschmann, S.: Regional surface albedo characteristics – analysis of albedo data and application to land-cover changes for a regional climate model, PhD thesis, University of Hamburg, Max Planck Institute for Meterology, Hamburg, 2012. R Development Core Team, R: A Language and Environment for Statistical Computing, The R Foundation for Statistical Computing, Vienna, Austria, 2011. Räisänen, P., Luomaranta, A., Järvinen, H., Takala, M., Jylhä, K., Bulygina, O. N., Riihelä, A., Laaksonen, A., Koskinen, J., and Pulliainen, J.: Evaluation of North Eurasian snow-off dates in the ECHAM5.4 atmospheric GCM, Geosci. Model Dev. Discuss., 7, 3671–3715, doi:10.5194/gmdd-7-3671-2014, 2014. Rechid, D.: On biogeophysical interactions between vegetation phenology and climate simulated over Europe, PhD thesis, University of Hamburg, Max Planck Institute for Meterology, Hamburg, 2008. Rechid, D. and Jacob, D.: Influence of monthly varying vegetation on the simulated climate in Europe, Meteorol. Z., 15, 99–116, 2006. Rechid, D., Raddatz, T. J., and Jacob, D.: Parameterization of snowfree land surface albedo as a function of vegetation phenology based on MODIS data and applied in climate modelling, Theor. Appl. Climatol., 95, 245–255, 2009. Roeckner, E., Arpe, K., Bengtsson, L., Christoph, M., Claussen, M., Dümenil, L., Esch, M., Giogetta, M., Schlese, U., and Schultz-Weida, U.: The Atmospheric General Circulation Model ECHAM4: Model Description and Simulation of the PresentDay Climate, MPI Report No. 218, Max Planck Institute for Meterology, Hamburg, Germany, 90 pp., 1996. Roesch, A., Wild, M., Gilgen, H., and Ohmura, A.: A new snow cover fraction parametrization for the ECHAM4 GCM, Clim. Dynam., 17, 933–946, 2001. Semmler, T., Jacob, D., Schlünzen, K. H., and Podzun, R.: Influence of sea ice treatment in a regional climate model on boundary layer values in the Fram Strait Region, Mon. Weather Rev., 132, 985–999, 2004. Simmons, A., Uppala, S., Dee, D., and Kobayashi, S.: ERA-Interim: new ECMWF reanalysis products from 1989 onwards, ECMWF newsletter, 110, 25–35, 2007. Solantie, R.: Albedo in Finland on the Basis of Observations on Aircraft, Meteorological publications, 12, Finnish Meterological Institute, Helsinki, Finland, 106 pp., 1988. Solantie, R.: Suurten suo-ojitusten vaikutus ilman lämpötilaan erityisesti Alajärven Möksyn havaintojen perusteella (the impact of large scale wetland drainage on air temperature based on observations in Möksy in Alajärvi), Meteorological publications, 29, Finnish Meterological Institute, Helsinki, 40 pp., 1994. Tomppo, E., Gschwantner, M., Lawrence, M., and McRoberts, R. E.: National Forest Inventories, Pathways for Common Reporting, Springer, The Netherlands, 2010. Tomppo, E., Heikkinen, J., Henttonen, H. M., Ihalainen, A., Katila, M., Mäkelä, H., Tuomainen, H., and Vainikainen, N.: Designing and Conducting a Forest Inventory-case: 9th National Forest Inventory of Finland, Springer, The Netherlands, 2011. Turetsky, M. R., Kotowska, A., Bubier, J., Dise, N. B., Crill, P., Hornibrook, E. R. C., Minkkinen, K., Moore, T. R., MyersSmith, I. H., Nykänen, H., Olefeldt, D., Rinne, J., Saarnio, S., Shurpali, N., Tuittila, E.-S., Waddington, J. M., White, J. R., Wickland, K. P., and Wilmking, M.: A synthesis of methane emissions from 71 northern, temperate, and subtropical wetlands, Global Change Biol., 20, 2183–2197, doi:10.1111/gcb.12580, 2014. US Geological Survey: Global land cover characteristics data base version 2.0, available at: http://edc2.usgs.gov/glcc/globdoc2_0. php, 2001. Venäläinen, A., Rontu, L., and Solantie, R.: On the influence of peatland draining on local climate, Boreal Environ. Res., 4, 89– 100, 1999. Wiscombe, W. J. and Warren, S. G.: A model for the spectral albedo of snow. I: Pure snow, J. Atmos. Sci., 37, 2712–2733, 1980. Wramneby, A., Smith, B., and Samuelsson, P.: Hot spots of vegetation climate feedbacks under future greenhouse forcing in Europe, J. Geophys. Res.-Atmos., 115, D21119, doi:10.1029/2010JD014307, 2010. Zhang, W., Jansson, C., Miller, P. A., Smith, B., and Samuelsson, P.: Biogeophysical feedbacks enhance the Arctic terrestrial carbon sink in regional Earth system dynamics, Biogeosciences, 11, 5503-5519, doi:10.5194/bg-11-5503-2014, 2014. www.biogeosciences.net/11/7251/2014/ Biogeosciences, 11, 7251–7267, 2014