Introduction
European forest use for fuel, timber and food dates back to pre-Roman times . Diverse forest uses have been documented for western Europe (UK, Belgium, Denmark, Netherlands, Germany, Switzerland), southern Europe (Spain, Italy, Greece), eastern Europe (Hungary) and northern Europe (Sweden, Finland) . Common uses fell into two broad categories: timber removal (harvesting, thinning and coppicing) and non-timber removal (branches, leaves and undergrowth removed for livestock usage through litter raking, pollarding and grazing). One of the most visible legacies of centuries-long intensive forest use is a shift in species distribution across central Europe, where forest managers replaced a substantial share of the deciduous vegetation by coniferous tree species . This anecdotal evidence in combination with references to increasing restrictions on common forest use supports the hypothesis that European forests have been intensively used for centuries, with forest cover reaching lows around the year 1850, followed by a turnaround in forest area .
Summary of various data sources used to reconstruct forest management between 1600 and 2010. The years listed are the years of the data set which were used in this work, not necessarily all the years available. The following abbreviations were used: GDP for gross domestic product, pers for person, live for head of livestock, S for southern, C for central and N for northern.
| Component | Data | Comments |
|---|---|---|
| General | ||
| Historical population maps | 1600–2000 | |
| Historical GDP data | 1600–2000 | |
| Wood density | 400 | |
| Carbon content | 50 % | |
| Charcoal calorific value | 30 | |
| Wood calorific value | 15 | |
| Preindustrial wood to charcoal mass loss ration | 4 | |
| Fuelwood – cooking | ||
| Energy demand | 8 | |
| Upscaled by population. People used charcoal for cooking instead of wood because of transport and air quality, although charcoal is less efficient. Energy demand is taken to be similar all over Europe despite differences in diet and climate. It was assumed that only wood was used as an energy source. | ||
| Fuelwood – heating | ||
| Temperature threshold | 15.5 C | |
| Heating degree days S Europe | 1000 | |
| Heating degree days N Europe | 4000 | |
| Personal space | 30 | Assumed |
| Heating efficiency | 10 % | Assumed |
| Upscaled by population. The energy released while cooking was used for heating. It was assumed people would use 50 % charcoal and 50 % wood heating because of transport and air quality, although charcoal is less efficient. | ||
| Salt production | ||
| Wood demand for brine of 250 | 1 | |
| Wood demand for brine of 50 | 10 | |
| Production | Table | |
| Upscaled by production. Steady production of the mines through time. Brine was transported no more than 25 through pipeworks. It was assumed that only wood was used as an energy source. | ||
| Beer and distillates | ||
| Consumption | 300 | |
| Wood demand | 4.8 | |
| Upscaled by population. It was assumed that only wood was used as an energy source. | ||
| Ceramics and brick | ||
| Consumption | 45 | |
| Wood demand | 2 | |
| Upscaled by population. It was assumed that only wood was used as an energy source. | ||
| Metal works | ||
| Iron consumption | 1.50 | |
| Lead consumption | 1.45 | |
| Copper consumption | 0.27 | |
| Silver consumption | 0.004 | |
| Gold consumption | 0.0002 | |
| Tin consumption | 0.022 | Estimate |
| Zinc consumption | 0.07 | Estimate |
| Wood demand | 81 | |
| Adapted from production estimates from the Roman Empire. Tin and zinc consumption is scaled to copper based on typical bronze and brass alloys. Upscaled by population. Iron and lead included military as well as domestic uses. These estimate includes both metal melting and working. Wood estimates are prior to charcoal conversion. Charcoal is only mandatory for iron smelting. However, the assumption was made that the preferred fuel for kilns and furnaces was charcoal because charcoal burns at higher temperatures and because of the ease of transport to production sites. |
Continued.
| Component | Data | Comments |
|---|---|---|
| Glass production | ||
| Total wood demand for glass | Similar to metal works | |
| Upscaled by metal production. Wood estimates are prior to charcoal conversion and it was assumed only charcoal was used. | ||
| Timber | ||
| Construction spending Spain | ||
| Construction spending England | ||
| Upscaled by gross domestic product. The share of different building materials was assumed constant and a relationship between construction and gross domestic product was assumed. | ||
| Shipbuilding | ||
| Share in wood demand | 1 % | |
| Not accounted for | ||
| Litter raking | ||
| Livestock densityN Europe | 0.6 | |
| Livestock densityC Europe | 0.5 | |
| Livestock densityS Europe | 0.3 | |
| Litter demand | 200–480 | |
| Correction factor1600–1850 | Tuned to | |
| Correction factor1851–2010 | Tuned to | |
| Upscaled by population. The correction factor was fitted such that agricultural litter use peaked in 1850 and was abandoned completely by 1950. Northern Europe was defined as the land mass above 55 N, central Europe was located between 45 and 55 N and southern Europe the land below 45 N. Forage collection was not considered. | ||
| Land cover changes | ||
| Historical land use maps | 1600–2000 | |
| Land cover map | 2010 | |
| Tree species map | 2010 | |
| Natural vegetation | 1600 | |
| No upscaling required | ||
| Annual wood increment | ||
| Annual wood increment | 3.8 | Country level productivity from Table S3 in |
| Reduction factor | 2 (–) | |
| Upscaled by land cover maps. Contemporary estimates of annual wood increment are restricted to stem wood increment. | ||
| GEN-MAPS (1600–1828) | ||
| Annual productivity of coppice | Country level productivity % | Assumed based on |
| Annual productivity of high stand | Country level productivity % | Assumed based on |
| No upscaling required. Coppicing was used to meet the fuelwood demand (i.e., cooking and heating), while high-stand management was used for all other wood uses. | ||
| GEN-MAPS (1829–1949) | ||
| Age class reconstruction | 1950–2010 | |
| Rotation high stand | 50–180 | Table |
| Rotation coppice | 35 | |
| Relaxation for transitions | 20 | Assumed |
| No upscaling required. Missing pixels in the age class reconstruction were linearly interpolated assuming age class structure between neighboring pixels is similar. It was assumed that 50 % of forests between 1 and 40 years old in 1950 were under coppice management throughout their life. The backcasting algorithm formalized the following assumptions: (1) permanent transition to high-stand management; (2) revival of coppice management during World Wars and interbellum; and (3) unmanaged forest being taken into production. |
Continued.
| Component | Data | Comments |
|---|---|---|
| GEN-MAPS (1950–2010) | ||
| Coppice fraction (1990) | 10–50 % | |
| Coppice fraction Fenno-Scandinavia (1990) | 1 % | Assumed based on |
| No upscaling required. All pixels were assumed to contain the same percentage of coppiced forest area as the country value. The 1990 coppiced areas were held constant until 2010. | ||
| ORC-MAPS (1600–2010) | ||
| General maps | 1600–1828 | This study, see Sect. |
| General maps | 1829–1949 | This study, see Sect. |
| General maps | 1950–2010 | This study, see Sect. |
| Land cover reconstruction | 1600–2010 | This study, see Sect. |
| Management threshold | 15 % | Assumed |
| Unmanaged threshold | 100 % in 1910 | Assumed |
| 15 % by 1950 | Assumed based on | |
| No upscaling required. The demand for wood products from coppice management was satisfied before the other wood demands. Dominant PFTs were managed first. |
In the last decades forest management has been reported to have important
impacts on the carbon balance , resulting in its inclusion in
the Kyoto Protocol as a tool to mitigate climate change and to buy time while
emissions from land cover change and fossil fuel burning could
be reduced. More recently, evidence is accumulating that forest management
(defined as forest remaining forest but undergoing tree or other biomass
removal, clear-cuts, fire suppression and/or species change) may also have
substantial effects on the atmospheric concentration
in addition to affecting the surface climate
through changes in albedo, evapotranspiration and surface roughness
.
Hence, to fully understand the impact of the large-scale application of forest
management, taking into consideration only the carbon balance is not
sufficient. Models need to be developed to examine the effects of forest
management on the surface climate
About a decade ago the land cover change community faced similar issues and made substantial
progress through reconstructing land cover changes. At present these reconstructions are mainly
derived from combining reconstructed population densities with agricultural models, guided or
constrained by historical research, archaeological findings and pollen records
Similar to the historical land cover studies, several groups have studied the impact of forest management on the contemporary (1950–2010) forest carbon sink over Europe . These groups reduced forest management to “age class management”, which consists of regular thinning followed by a clear-cut. In this way, the model can be driven by an age class reconstruction rather than a management reconstruction. Several methods have been proposed and used for age class reconstructions of European forests . As a result, forest age reconstruction exists for the period 1950 to 2010 over Europe. A first attempt to create a forest management map for the year 2000 for Europe was completed by . However, while this map originally started off as an effort to indicate the current management strategies across Europe, it ended as a map which depicts optimal management strategies based purely on the rational knowledge of forest managers. Although this effort could be a valuable starting point for developing future forest management scenarios, it is of limited interest for historical reconstructions.
In this manuscript, we reconstructed historical forest management for four contrasting but widely used management strategies (see Sect. ) with the purpose of spinning up soil carbon and biomass pools for the year 1600 for European-wide simulations. Management reconstructions from 1600 to 2010 are needed to drive transient model experiments simulating the impact of forest management on the carbon budget as well as the surface and atmospheric climate. In this paper, we present the data sources and algorithms that were used to construct the forest management maps in as much detail as possible. The application of these maps in model experiments will be the subject of future studies.
Material and methods
General approach
Annual spatially explicit wood demand and supply was estimated for the period from 1600 to 1829, based on various sources (Table ). Wood supply was based on estimates of historical annual wood increment and land cover reconstructions (see Sect. ). Demand accounted for the fuelwood needed in households, wood used in food processing, charcoal used in metal smelting and salt production, and timber for construction (see Sect. ). The estimate was based on reconstructed individual needs and population estimates. Coppice management was assumed to satisfy the fuelwood demand, whereas charcoal and construction wood were assumed to come from deforestation and high-stand management. Comparing estimated demand and supply resulted in spatially explicit reconstructions of the share of forests under coppice and high-stand management and forest left unmanaged (see Sect. ). For the reconstruction between 1829 and 2010, a supply-driven backcasting method was used. The method used a 1950 to 2010 age reconstruction as its starting point. The general idea of such an approach is straightforward: a high stand 90 years old at the end of 1950 must have been cut and replanted at the end of 1860, before which it was either a high stand or a coppice. If, based on the demand and supply approach, it was decided that the stand was already under high-stand management in 1859 and the rotation length of the species is 120 years, then the stand was assumed to be 120 years old in 1859.
The above approaches for 1600 to 1828 and 1829 to 2010 resulted in resolution maps containing the amount of forest area (in ) required to be managed as high stands and coppice to satisfy estimated wood demand (from 1600 to 1828) and age class reconstructions (1829–2010). Usage with ORCHIDEE-CAN (ORganizing Carbon and Hydrology In Dynamic Ecosystems Environment, CANopy version) , the land surface model of our choice, requires that the previously assigned management types and surface areas are assigned to specific tree species or species groups. The final product is thus a map which describes the forest area, the area of each species and the management strategy used for that species for each pixel. All variables may change from one year to the next.
Finding comprehensive data to parametrize and validate these historical activities is not currently possible. Hence, a variety of proxy information was used. These proxies, their usage and underlying assumptions are discussed in more detail below. We deliberately separated the carbon removal activities into various forms of wood harvesting and litter raking, which is expected to allow easier incorporation of future advances into the proposed procedure. As such, the resulting maps should be considered as a starting point rather than a final data product.
Reconstructing wood and litter demand for 1600–1828
Historical wood demand primarily comes from the following sectors: fuelwood, industrial charcoal production (e.g., iron smelting, salt and glass production) and timber (e.g., shipbuilding and construction). Although each of these sectors is treated in a different way in this manuscript, most estimates rely on historical population maps available in the HYDE 3.1 database . These maps are only available for certain years, e.g., 1600, 1700, 1710, 1720. Instead of using the map for 1600 to cover the whole range of 1601 to 1699, we used linear interpolation to fill in the missing years. The same is true for the decadal data, i.e., 1701–1709, 1711–1719, etc., up until 2005, which is the latest year in the database.
Fuelwood demand (1600–1828)
For the period before electricity became available, household energy requirements can be divided into two categories: cooking and heating. The energy required for cooking depends on the local diet because different foodstuffs and food quality require different cooking times. The amount of wood required for cooking in preindustrial Europe ranged from 533 to 1067 per person per year. The lower number comes from using wood itself as the fuel, while the higher number comes from converting wood to charcoal first. Despite the lower efficiency, charcoal is often preferred as a fuel, in particular in urban areas, due to lower transport costs and lower production of particulate matter. To simplify matters these general estimates were assumed valid across Europe.
The required energy for heating depends on temperature, housing insulation, the number of people in the household, furnace efficiency and other heat sources. As a first approximation we accounted for heat as a by-product of cooking, and the additional heating requirement was calculated solely based on temperature, more specifically on heating degree days. Heating degree days quantify the total energy that is needed to keep the minimum indoor temperature above a threshold. It is calculated by multiplying the number of days by the number of degrees that the mean temperature is below a temperature threshold, generally set at 15.5 C in Europe . Heating degree days range from around 1000 to 4000 from the Mediterranean to Scandinavia, respectively. The heat released from cooking was estimated to already heat 50 to 400 of air per year up to this threshold in Scandinavia and the Mediterranean, respectively. An additional 30 of air were assumed to have been heated to compensate for the fact that heat from cooking in summer did not contribute to the heat requirements during winter. This required an additional 24 to 97 of wood per person per year from the Mediterranean to Scandinavia, assuming a heating efficiency of 10 %. Wood demand for heating was calculated as the product of air volume, air density, heat capacity of air, heating degree days and heating efficiency divided by the wood caloric values (Table ).
The above numbers for cooking and heating were combined to give four fuelwood estimates, using different combinations of (1) fuelwood demand calculated based on 100 % charcoal usage or 100 % wood usage and (2) European heating demand based on Scandinavian or Mediterranean per capita heating wood demand. Given the inherent uncertainty in these numbers, a simple mean was calculated between all four scenarios to give the final fuelwood demand maps.
Industrial wood and charcoal demand (1600–1828)
Between 1600 and 1828, wood and charcoal were used in the industrial production of salt, beer, metal and glass. Certain of these uses require a higher temperature than what can be achieved using wood directly, e.g., iron smelting, although other large-scale uses, e.g., brewing, could have made use of either wood or charcoal. For simplicity, we assumed that salt, beer and distillates were produced with wood as the fuel source, while ceramics, glass and metal works used charcoal as a fuel source.
By 1600 much of the salt production from brine springs had stopped in favor of sea salt production and rock salt mines , which consume much less wood than the evaporation of brine. We incorporated only production from sites which are known to have operated after 1600 (Table ). For these sites, the period of operation, the average annual production over different periods and the salt content of the brine are known. The amount of wood required depended on the production volume and the salt content of the brine . For brine with a concentration of 250 , 400 of wood produced 400 of salt, while 10 times as much wood was needed for the same amount of salt at a brine concentration of 50 . We used a linear function to interpolate between these concentrations. Given that the spatial resolution of our maps is on the order of 50 and that pipeworks rarely extended more than 40 , all of the wood demand from salt production was placed in the pixel where the brine spring was located.
Summary of European salt mine characteristics important for wood consumption during 1600–1900. The final two columns are estimates based on information found in .
| Mine | Location | Brine | Years of | Production |
|---|---|---|---|---|
| () | production | () | ||
| Halle an der Saale | 51.48 N, 11.97 E | 250 | 1600–1699 | 6000 |
| 1700–1799 | 3500 | |||
| Lüneburg | 53.25 N, 10.42 E | 250 | 1600–1649 | 20 000 |
| 1650–1699 | 15 000 | |||
| 1700–1749 | 10 000 | |||
| 1750–1799 | 5000 | |||
| Hallein (Salzburg) | 47.68 N, 13.10 E | 50 | 1600–1649 | 20 000 |
| 1650–1699 | 15 000 | |||
| 1700–1749 | 10 000 | |||
| 1750–1799 | 5000 | |||
| Hallstatt (Bad Ischl) | 47.72 N, 13.63 E | 250 | 1600–1649 | 50 000 |
| 1650–1699 | 40 000 | |||
| 1700–1799 | 10 000 | |||
| Saulnois (Lorraine, France) | 49.12 N, 6.18 E | 150 | 1600–1699 | 8000 |
| 1700–1799 | 25 000 | |||
| 1800–1899 | 10 000 |
In northern and eastern Europe, where grape cultivation was difficult or impossible, beer was consumed daily by all social classes. Alcoholic beverages were an important source of the much needed caloric intake and a popular medicine . We assumed an average annual consumption of 300 of beer and distillates per person per year, which required 4.8 of wood per liter of beverage . Drinks and food were prepared, served, consumed, processed, stored and transported in ceramics; in order to produce 45 ceramics per person per year, 360 wood per person per year were required . Spatial population estimates were scaled by these per capita demands to provide spatially explicit wood demand maps for 1600–1828.
Historical annual per capita usage of metals has been reported to be 1.5 for iron and lead and to range between 0.2 and 0.0002 for copper, silver, gold, tin and zinc (Table ). Total wood requirements for metal works were calculated assuming that 81 of wood, prior to charcoal conversion, was needed to smelt 1 of metal and by making use of the HYDE population estimates . The glass industry also consumed charcoal, mainly for potash production and less for the melting process itself . As there were many local production sites and production could be easily moved to regions with accessible wood resources , the historical production has been poorly documented. During the Napoleonic era , however, wood demand for glass and iron production has been estimated at up to 3 for each process in central Europe. In the absence of other data, we assumed the wood demand for glass and iron to remain the same between 1600 and 1828.
Timber production (1600–1828)
The use of wood as a building material varied both spatially and temporally, and it was often substituted by loam, stone, straw and brick. By restricting our estimate to between the years 1600 and 1828, the use of concrete in homes or other structures can mostly be ignored, since its use was largely abandoned after Roman times until after 1850 . We assumed that building construction better relates to gross domestic product (GDP) than to population, as more productive societies were able to build larger houses and buildings. Furthermore, it was assumed that the share of other building materials remained constant between 1600 and 1828 in Europe, which ignores the possibility that, during that period, affluence may have shifted building preferences from wood to stone.
Spatially explicit estimates for gross domestic product per capita in Europe for the time period from 1600 to 1828 were created using the gross domestic product per capita for years and countries for which it has been previously estimated , expressed in the hypothetical international currency Geary–Khamis dollars (). When needed, missing values were supplied through spatial or temporal interpolation. Next, estimates of wood used for construction in Spain and England in 1660 and 1700 (Table ) were combined with spatially interpolated gross domestic product estimates to calculate a construction wood usage of and for Spain and England, respectively. For simplicity, the result was rounded to and used for all of Europe. Subsequently, this value was scaled to the spatially explicit gross domestic product estimate between 1600 and 1828.
Timber shortage was early on reported as a critical issue, especially for naval shipyards . However, this shortage was due to very specific requirements, e.g., long sailing masts or specially shaped oak wood for the body, and not because of a general lack of timber . Shipbuilding demand was estimated to be around 1 % of the total wood demand by the end of the 1700s, with a lower demand in earlier periods . Despite its political, military and economic importance, we neglected the impact of shipbuilding activities on the historical wood demand in Europe due to its comparatively minor significance regarding total wood consumption.
Litter demand (1600–1828)
Towards the end of the Middle Ages, farmers began to keep their cattle inside during winter, which led to a demand in forest litter to absorb animal waste . Initially forest litter was used as it was easily available, cheap and, contrary to straw, had no other uses such as construction . In spring, the waste-soaked litter would then be spread on the fields as a form of fertilizer. From 1750 throughout the 1800s, litter demand increased significantly in the course of the agrarian modernization, which promoted indoor feeding of cattle year round . The expanding railroad network, however, made straw more easily available for areas without grain production, and forest litter collection was abandoned towards the end of the 1800s and the beginning of the 1900s. From this, it was assumed that litter raking started at a modest level during the first years of our simulations (1600), peaked in the mid-1800s and faded out afterwards.
European maps of litter demand were based on historical livestock estimates , taken to be equal to 0.6, 0.5 and 0.3 for northern, central and southern Europe, respectively. For this application, the dividing parallels between northern, central and southern Europe were taken to be 55 and 45 N latitude. It has been reported that 200 to 480 of dry litter were collected per livestock unit per year . We further assumed that 480 litter per livestock unit per year corresponded to the peak demand for the study period (see Sect. ).
First, litter maps were generated from the livestock density maps using peak demand for all years from 1600 to 1828 and 1829 to 2010. Next, these initial litter maps were multiplied by a correction factor (Table ) to account for the temporal evolution in litter demand. The correction factor was tuned to give the desired behavior, based on historical information from Switzerland .
Reconstructing wood supply
Wood supply was calculated from annual forest area and wood increment estimates. Wood increment was assumed to differ for coppice and high-stand management and thus partly depends on the wood demand.
Forest area and species composition (1600–2010)
Historical forest area was reconstructed by making use of three different data sources: (1) the tree species maps of ; (2) the land cover map of (hereafter abbreviated as P&M), and (3) the historical land use maps of (hereafter abbreviated as K&K). Maps (1) and (2) were combined by taking the fractions of bare soil, grasses and crops in Europe from P&M. The forests inside Europe from the P&M map were replaced by the more detailed species composition fractions in the map, scaling the sum of the fractions to be equal to the total forested fractions present in the P&M map in Europe. Details about the correspondence of the forest species in and the plant functional types (PFTs) used in the ORCHIDEE land surface model can be found in .
The European tree species map is only available for the year 2000, whereas the P&M
map is available for 2000, 2005 and 2010. As we are interested in historical forest management back
to 1600, this combined species–land-cover map had to be extended through historical land use
maps. Although the same procedure could be done with any historical land use maps
Our reconstruction contained 28 PFTs, whereas the historical maps of K&K, which were used to reconstruct the historical forest area, only contain three classes, i.e., crop, pasture and forest. The procedure used to convert the three land cover classes into the 28 classes was based on that of . In brief, for a given year , per pixel, the crop fractions in the P&M maps were scaled to match the total crop fraction from the K&K map for year ; the ratio of C3 to C4 crops on the historical K&K map was kept identical throughout the period, as the K&K maps do not distinguish between C3 and C4 vegetation. If there were no crops on a pixel in P&M but there were crops in the same pixel on the historical K&K map, the C3 C4 ratio of the neighboring pixels on the P&M map was used to determine the C3 C4 ratio in our reconstruction. Throughout this study, neighboring pixels were defined as pixels touching either an edge or a vertex of the current pixel, i.e., there are eight possible neighbors.
Next, the amount of grassland on the pixel was computed using the pasture fraction on the K&K map and following a procedure similar to the crop fraction above. The K&K maps have a pasture fraction, while the contemporary P&M map only has C3 and C4 grass fractions; C3 and C4 refer to the photosynthetic pathway in the grass and give no indication of management. The extent of unmanaged grassland is therefore not known. If the pasture fraction on the K&K map was lower than what was present on the P&M map, the reconstruction took the P&M grassland fraction, under the assumption that managed and non-managed grasslands can coexist on the same pixel and that present-day non-managed grasslands fraction is at its historical low. If the K&K pasture fraction was greater than the grassland fraction present on the P&M map, our reconstruction made use of the K&K fraction, and the difference in surface area was made up by leaving crops untouched and reducing the forest fraction. The C3 C4 ratio was treated in the same way as for crops, i.e., it was kept constant for each pixel throughout the period. If the fractions of crop or pasture usage were greater than the total vegetative fraction present on P&M, we did not increase the vegetative fraction. Instead, all forests were replaced. In essence, we assumed that the non-vegetative fraction of the pixel has not changed in the past 400 years. Furthermore, all sub-grid-level water was classified as bare soil.
Over the course of the past several hundred years, the species composition of European forests has changed as much as the extent of the forests. The largest difference is the large-scale planting of coniferous species, resulting in modern forests, which are significantly more conifer-rich than historical forests . While forests in Europe have been heavily managed for centuries, and while human activities had certainly changed the species composition on certain scales already before 1600 (e.g., through the introduction of chestnut Castanea sativa Mill.), it is likely that the composition of growth forms (e.g., coniferous vs. non-coniferous forests) in 1600 closely resembled that of natural vegetation on scales of tens of kilometers, which is what we are interested in here. Therefore, we have scaled the conifer fraction of the forests on our pixels in 1600 to be equal to that found in the natural vegetation map of .
The question remains how to make the transition from the conifer fraction in 1600 towards the conifer fraction found in our reconstruction for the year 2010. For this purpose, the European countries were divided into “early” and “late” conversion countries. Early countries, i.e., Germany, the Netherlands, Belgium and Switzerland, began experimenting with large-scale conifer plantations sooner than the rest of Europe . In these countries, the conifer fraction is kept constant from 1600 to 1800, then linearly increased to the 2010 conifer fraction. For the rest of Europe, the same linear transformation starts in 1850. This represents one possible transformation method, simulating a slow establishment of conifer plantations.
Wood production (1600–1829)
The contemporary mean annual wood increment for 16 European countries calculated from forest inventory
data ranges from 0.6 in Greece to 9.5 in Germany, with
a value of 3.8 for the combined productivity of all countries listed
We used the increment estimates of for each of the countries listed, filling in countries for which no estimates were available by taking an average of the neighboring pixels. To correct for the increase in productivity, we differentiated between countries in western Europe, where nitrogen deposition is the greatest, i.e, UK, France, Germany, Switzerland, Netherlands, Ireland, Denmark, Belgium and Austria, and the rest of Europe. Annual increments for forest in western Europe were scaled down by a factor of 2 prior to 1920, in accordance with what is reported by . After 1920 we linearly increase the productivity of every country to its current value by 1990. This gives estimates for forest productivity in Germany, France and England around 1700–1800 which are within the 2.8 to 3.8 of timber independently reported for the same time period . Despite the simplifications underlying this approach, central European countries have the highest productivity throughout the reconstruction, which we believe correctly reflects an earlier implementation of more advanced silvicultural practices in central Europe.
The annual wood increment gives a maximum annual wood supply for sustainable harvests. To simplify the calculations it was assumed that the annual increment was harvested annually. Hence, this approach does not require explicitly accounting for clear-cuts and other management interventions.
Wood production maps from 1600 to 1828 were made by using the historical forest maps above (see Sect. ) and the productivity estimates. For each year between 1600 and 1828 (the final year of the wood demand maps), the total forest area in a grid cell was estimated. This area is compared to the forest area for the same grid cell in the following year. If the forest area in the current year is greater than the forest area in the subsequent year, this difference is multiplied by the annual wood increment for that pixel and the resulting wood amount stored as “land cover change”. If the current forest area is less than the subsequent year's forest area, the newly planted area will be taken into account only in the following year. In any case, the rest of the forest area for the current year is multiplied by the increment and stored under “harvest”. For the final year of the calculation (1828), land cover change is still available, as the historical PFT maps have been created up until the present day, and therefore the map from 1829 exists and can be used even if the demand map is not calculated for 1829.
Summary of independent information, i.e., information not used to create our estimates, for European wood demand from around 1600 to 1900.
| Summary | Source | Reconstructed |
|---|---|---|
| Fuelwood | ||
| 2–2.5 in northwest and central Europe from 1500 to 1830 | 0.9 | |
| 1.4 in Madrid in 1680 | 0.9 | |
| 4.7 in Sweden in 1800 | 0.8 | |
| 1–2.3 in Italy before 1861 | 0.7 | |
| Energy demand was 9 . Fuelwood pressure on forests in Switzerland declined by the end of the 1800s since coal was used significantly from 1850 onwards. | 5 | |
| In Paris, per capita fuelwood consumption amounted to1.8 in 1815. Fuelwood consumption started falling in 1830s in France due to the rise of coal. | 0.9 | |
| From 1800, the wood demand was 4 for urban populations in Finland. Wood demand started to decline around 1840. | 0.4 | |
| From 1800, the wood demand was 9.5 for rural populations in Finland. Wood demand started to decline around 1840. | 0.9 | |
| In 1700, 2.3 of firewood was produced in England. Wood had mostly been replaced by coal after 1810 in England. | 1.9 | |
| In 1700, firewood production required (740 000 ) of coppiced land. | ||
| Timber | ||
| 53 % of total wood consumption as sawn timber in Europe in 1920, varying from 15 to 96 % of the total wood production based on the country | N.A. | |
| In Austria the total wood production consisted of 66 % fuelwood in 1877 | N.A. | |
| In Austria the total wood production consisted of 34 % construction wood in 1877 | N.A. | |
| Charcoal production | ||
| (0.866 ) for iron production in England in 1700 | 3.8 | |
| Iron production in England in 1700 required 330 000 (133 000 ) of woodland | 950 000 | |
| Total wood demand | ||
| In 1800, each Finn consumed 20 | 5 |
Rotation length by tree species and country collected from . If an explicit year is not given, the year 1990 (the date of the latest data set used) is assumed. Region names are listed in bold font and are defined in , from which the quotations are also taken.
| Region/Country | Species | Years | Rotation lengths (years) |
|---|---|---|---|
| Northern | Pine | 1990 | 80–120 |
| Spruce | 1990 | 70–90 | |
| Pine | 1990 | 110–180 | |
| Spruce | 1990 | 90–140 | |
| Finland | – | – | – |
| Norway | – | – | – |
| Sweden | – | – | – |
| Central | – | – | – |
| Denmark | Pine | 1990 | 50–150 |
| Germany West | Conifers | 1920s–1930s | 80 |
| Fir | 1920s–1930s | 90 | |
| Spruce and fir | 1980s | 80–100 | |
| Pine and larch | 1980s | 100–120 | |
| Oak | 1980s | 140–180 | |
| Beech | 1980s | 120–140 | |
| Germany East | – | 1990 | “the rotations are somewhat shorter [than those in the rest of the region]” |
| Poland | – | 1990 | “the rotations are somewhat shorter [than those in the rest of the region]” |
| Czechoslovakia | – | 1990 | “the rotations are somewhat shorter [than those in the rest of the region]” |
| Atlantic | Pine | 1990 | 50–60 |
| Other conifers | 1990 | 50 | |
| Oak and/or beech | 1990 | 120–150 | |
| Ireland | – | – | – |
| United Kingdom | Broadleaf | 1990 | 120 |
| Sub-Atlantic | – | – | – |
| Netherlands | – | 1990 | “rotation ages are relatively short compared with those in Central Europe” |
| Belgium | – | – | – |
| Luxembourg | Conifers | 1990 | 80 |
| Broadleaf | 1990 | 130 | |
| France | Broadleaf | 1990 | 140 |
| Spruce and fir | 1990 | “long rotation” | |
| Alpic | – | – | – |
| Austria | – | 1990 | “relatively long” |
| Switzerland | Conifers | 1990 | 80–250 |
| Broadleafs | 1990 | 80–180 | |
| Pannonic | – | – | – |
| Hungary | Pine | 1990 | 60–70 |
| Other conifers | 1990 | 60–70 | |
| Oak | 1990 | 90–110, 70–100 | |
| Beech | 1990 | 100–120 | |
| Poplar | 1990 | 15–30, 15–40 | |
| False acacia | 1990 | 15–30, 30–35 (up to 75) | |
| Coppice or coppice with standards | 1990 | 30–75 | |
| Romania | Conifers | 1990 | 100–110, 100–140 |
| Broadleafs | 1990 | 100–120, 100–140 |
Continued.
| Region/Country | Species | Years | Rotation lengths [years] |
|---|---|---|---|
| Mediterranean west | Maritime pine | 1990 | 50 |
| Other pines | 1990 | 60–100 | |
| Cork and/or oaks | 1990 | 150–200 | |
| Eucalyptus | 1990 | 8–15 | |
| Poplars | 1990 | 8–20 | |
| Coppice | 1990 | 8–20 | |
| Portugal | – | – | – |
| Spain | – | – | – |
| Mediterranean Middle | – | 1990 | “regimes used in the coniferous forestresemble those applied in Central European regions although rotation ages are shorter” |
| Italy | – | 1990 | “Part of the stands are grown to high rotation ages” |
| Yugoslavia | – | – | – |
| Albania | – | – | – |
| Mediterranean east | – | – | – |
| Bulgaria | – | – | – |
| Greece | – | – | – |
Southern part of the region. Northern part of the region. Actual rotation ages are often 10–20 years longer than prescribed. Differing values given at different places in the text.
Reconstructing forest management
Defining forest management strategies
Although forest management has developed a wide range of locally appropriate and species-specific strategies, the nature of this study requires a limited number of contrasting strategies that are expected to be relevant at the spatial resolution (e.g., ) of global and regional modeling studies. To this end four management strategies were distinguished based on their expected impact on the biogeochemical and biophysical processes:
- 1.
FM1: no human intervention. Expected to result in tall, vertically and horizontally complex canopies with a closed nutrient cycle.
- 2.
FM2: high stands with thinning and harvesting based on stem diameter and stand density. Expected to result in tall, homogeneous canopies with open nutrient cycles.
- 3.
FM3: coppicing with harvesting of aboveground biomass, based on stem diameter. Expected to result in medium-high, homogeneous canopies with open nutrient cycles. Note that at present it is not possible to simulate coppicing-with-standards in ORCHIDEE-CAN.
- 4.
FM4: short-rotation coppicing with harvesting of all aboveground biomass based on age, using rotations of less than 6 years. Expected to result in short, homogeneous canopies with intensive nutrient exports and possibly requiring fertilization to maintain productivity.
We have chosen not to divide our management strategies into biogeographic regions since we considered that, regardless of region and species composition, the net biogeochemical result from the different management types is to transfer woody biomass (and carbon) from the forest into an agricultural PFT or to remove it completely from the ecosystem (as harvest). The net biophysical result will be driven by changes in canopy structure and vegetation height. We believe that the first-order effect of forest management can be quantified without accounting for management differences between biogeographic regions. The lack of available data for parametrization and validation discourages introducing too much detail at this time.
Two sets of forest management maps
Using the demand and supply estimate, a first set of general maps (GEN-MAP) was made, which was, however, independent of the historical land cover maps (see Sect. ). These maps list the forest area under each management type required to meet the wood demand (see Sect. ; from 1600 to 1828) and match the reconstructed age class structure (see Sects. and ; from 1829 to 2010) for each pixel. In the GEN-MAPS, we report only the land areas of high-stand management and coppicing required. Short-rotation coppice are not included in these maps, as short-rotation coppices were not significantly used historically. Unmanaged forest is not included in the maps since that fraction is estimated as a residual of the total amount of forest area on a given pixel, and land cover data are not used in this step. If the total forest area required in GEN-MAP is greater than the forest area found in a given pixel on a historical land cover map, the user will have to decide how to prioritize and manage this conflict.
Rather than simply assigning forest areas to the different management types as is the case for the GEN-MAPS, we went one step further and assigned forest area and tree species to the different management types. The resulting maps (ORC-MAP) can be directly read into the ORCHIDEE-CAN model. These maps take into account the historical land cover maps as well as a few model-specific constraints (see Sect. ), resulting in maps which list, for every pixel, the amount of forest required under coppice, high-stand and short-rotation coppice management to meet demand and age class information. The ORC-MAPS, as opposed to the GEN-MAPs, do contain fractions of unmanaged forest, although there is no short-rotation coppice area.
GEN-MAP (1600–1828)
The general maps before 1828 simply represent how much forest is required to satisfy the wood demand. Initially, only high-stand management and coppicing were considered. The reason for this is that our pre-1829 maps are based entirely on wood demand. The approach assumes that coppicing is done to meet fuelwood demand (see Sect. ), while industrial wood demand (see Sect. ) was met using high stands. A simplified approach to calculating historical forest productivity is used here, as outlined in Sect. . In order to distinguish between high-stand and coppice systems, the productivity for a pixel estimated as explained in Sect. was decreased by 10 % for high-stand forests and increased by 10 % for coppice. This differential productivity reflects the fact that the historical productivity estimates are averages over all forest ages, while the productivity of coppice systems will be higher because they do not have to grow a completely new root system after harvest .
These productivity estimates were then used to calculate the area of managed forests that needs to exist on this pixel to meet the wood demand. All fuelwood for cooking and heating was assumed to come from coppicing, whereas all other wood was assumed to come from high-stand forests. For example, if 120 of fuelwood is required in a pixel where coppice produces 3.0 , 40 of forest is coppiced for this year. Note that in the GEN-MAP approach the required surface area under management may be greater than the total area of the pixel if the demand is very high.
GEN-MAP (1829–1949)
The period after 1829 was characterized by the onset of so-called scientific forest management , resulting in the abandonment of coppicing, pollarding and litter raking in favor of a production-oriented forestry, which is still practiced today, i.e., high-stand management. After 1829, our forest management reconstruction therefore gradually replaced coppicing by high-stand management such that it reproduces the reconstructed modern European age class structure . Although non-timber forest usage (e.g., litter raking) persists at present on local scales , it is no longer considered significant for large-scale modeling applications over Europe.
Age class reconstructions are available in 5-year intervals from 1950 until 2010 and include the following age classes: 1–20, 21–40, 41–60, 61–80, 81–100, 101–120 and . To our knowledge, this age reconstruction is the only European age reconstruction that has been validated , and the year 1950 was therefore used as the target age structure for our management reconstruction. The age class reconstruction does not cover all European countries. The most notable missing countries in terms of surface area are Belarus and Ukraine. We therefore used an iterative filling procedure under the assumption that the age class structure of neighboring pixels is likely to be correlated. For this, any European pixels which were not found on the age class map but were touching pixels that were on the age class map were given a target distribution equal to the average of neighboring pixels. This procedure was repeated until all pixels were given a value. After the iterations had finished, a few pixels, mostly small islands, remained untouched. For these pixels we used the value of the nearest neighboring pixel. In addition, the areas for each age class were interpolated between the years of the reconstruction to give maps for every year between 1950 and 2010. Note that the European age class reconstruction contains no direct information on the applied forest management.
We applied a so-called backcasting age reconstruction that
is mainly based on an assumed rotation length
Despite its straightforward implementation, the backcasting algorithm needs to deal with several specific cases. For the 1829 map, we checked the fraction of forests under high-stand management in 1828 which was estimated by the demand-based approach (see Sect. ). If the surface area of forests under high-stand management in 1829 is greater than the area we would expect based on A121 in 1950, the FM distribution for 1828 was also used for 1829 on the pixel (Fig. a). If the fraction of high-stand forest in 1828 is less than expected based on A121, forests under coppice in 1828 were converted into high-stand management in 1829 until a forest area of at least A121 was obtained in 1829 (Fig. b). If we had no coppice in 1828, the amount of high-stand forests was increased while keeping the coppice fraction at zero (Fig. c) so that the total amount of managed forest required in the pixel increases. This procedure was repeated for all age classes, and to obtain smoother maps, it was assumed that one twentieth of the forest in, for example, age class A101 was planted every year between 1830 and 1849.
Conceptual presentation of the common forest management transitions at the pixel level and how these transitions were dealt with in the reconstruction. (a) In 1828 the demand–supply approach resulted in more high stands than expected from the backcasting method. (b) The surface area of high-stand forest in 1828 is less than expected based on backcasting. (c) No coppice management in 1828 based on demand–supply, but there is coppicing in 1829 based on backcasting. Dark grey shows coppice management; light grey shows high-stand management.
[Figure omitted. See PDF]
Coppicing is known to have been important and it still exists in significant quantities especially in southern Europe and France ; however, it is uncertain how much coppice forest existed historically. Since coppiced forests were seldom on rotations longer than 35 years , we assumed that the fraction of coppice forests on a pixel is smaller than the amount of forests in the 1–20 and 21–40 year age classes. We further assumed that 50 % of forests between 1 and 40 years old in 1950 were under coppice management throughout their life, thus, in our example, dating back to 1910 when the 40-year-old forest was planted (Fig. a–c).
The following two cases provided an upper bound on the area of high-stand forest or a lower bound on the area of coppice in a pixel between 1828 and 1949. The algorithm checked and accounted for these cases only after the procedure described in the previous paragraphs was performed. (1) The area of high-stand forest in 1828 may have been larger than required by the age class maps for 1950. In this case, we created a linear transition between the areas in 1828 and 1950. This represented a situation of early modern forestry, with the conversion of coppice to high stands occurring earlier than what can be observed in the age class data (Fig. a). (2) There is a possibility that the amount of coppice in 1828 was less than what was required in 1950. In this case, the amount of coppice was held constant until 1910. Between 1910 and 1950, the amount of coppice was increased linearly to the required amount in 1950 (Fig. c). Increasing coppice accounted for increased forest usage during the First (1914–1918) and Second (1939–1945) World War.
Furthermore, the amount of coppice area between 1829 and 1950 could face one of two situations. Either the amount of coppice forest in 1828 was greater than what was required in 1950, or it is less. In the former case, we convert coppice area to high stands as required to reproduce the age reconstruction. For example, assume that in 1850 there are 10 000 of coppice and no high stands, and in 1950 there are 1000 of coppice and 4000 of 81–100--old forest and no other age classes populated (Fig. a). From 1850–1869, the amount of coppice forest needs to be reduced by 200 per year to match the amount of high-stand forest by 1950 (the total 4000 is converted evenly over the 1850–1869 time period). In addition, 1000 of high stand is introduced between 1910–1950, as our method takes 50 % of the youngest two age classes in 1950 as coppices, implying that the other 50 % are managed as high stands. This method will not bring new forest under management until there are no further coppice forests to convert, which does not happen in this case (Fig. b). As the GEN-MAPS are not constrained by actual forest area on the pixel, these “new forests” are not taken from any other management type.
Conceptual presentation of the common forest management transitions at the pixel level and how these transitions were dealt with in the reconstruction. (a) The amount of coppice area in 1850 was higher than what was required in 1950. (b) The amount of coppice area in 1850 was lower than what was required in 1950. Dark grey shows coppice management; light grey shows high stand management.
[Figure omitted. See PDF]
Following the algorithm outlined above we created general management maps from 1829 to 1949 which include the forest area needed per pixel to match the wood demand in 1828 and reconstructions of in 1950, ignoring any historical maps detailing forest area. This means that values of forest area given during this time period may be substantially larger than the total area of the pixel.
GEN-MAP (1950–2010)
The percentage of coppiced forest in many European countries, including in Belgium, Luxembourg, France, Hungary, Italy, the former Yugoslavia, Albania, Bulgaria, Spain, Portugal and Greece, varied from 10 to 50 % in 1990 . The area under coppice was linearly decreased from the 1950 value to its 1990 value reported in . Missing data were assigned coppiced areas based on nearest neighboring pixels. For Finland, Sweden and Norway, a target fraction of 0.01 was set as these regions do not have an extensive history of coppicing . All pixels were assumed to contain the same percentage of coppiced forest area as the country value in 1990, and the 1990 coppiced areas were held constant until 2010. The maps from 1950 to 2010 have the same total forest area as .
ORC-MAPS (1600–2010)
The GEN-MAPS described above are 0.5 resolution maps containing the forest area (in ) required to be managed under high-stand and coppice management to satisfy the estimated wood demand from 1600 to 1828 and the age reconstruction maps of from 1829 to 2010. For usage by the ORCHIDEE-CAN model, however, the assigned management types need to be (1) allocated to a specific tree species or species groups and (2) combined with historical land cover maps to take into account the total forest area within a pixel.
An ORCHIDEE-CAN-specific map lists a single forest management strategy for each PFT on a pixel. The ORCHIDEE-CAN model targeted in this study consists of 28 PFTs, including bare soil, 3 grassland types, 3 agricultural types and 21 forest types. Some of these forest PFTs are parametrized to represent individual tree species, while others are more general metaclasses . The first step is to decide which PFTs are allowed to be managed in which style? Obviously none of the non-forest PFTs can be managed as forests. Additionally, conifers cannot be coppiced due to phylogenetic constraints (with the exception of Taxus baccata L.). This leaves 12 deciduous PFTs which can be left unmanaged or be managed as a high stand or a coppice. The nine coniferous PFTs can only be managed as a high stand or left unmanaged.
For each year between 1600 and 2010, our land cover reconstruction (see Sect. ) and the GEN-MAPS map (see Sects. , and ) were consulted. For each pixel, the forest PFTs were sorted by size from largest to smallest, in terms of surface area. Priority was given to coppice management because under our assumptions, coppice provided wood for primary needs such as cooking and heating. Hence, deciduous PFTs, starting from the dominant species, were put under coppice management if the demand required that more than 15 % of the pixel was put under coppice management. If so, PFTs were managed as coppice until the total coppice forest area in GEN-MAP was exceeded or there were no more deciduous PFTs. At that point, whether solely conifers or both conifers and deciduous PFTs were available depended on the pixel. Starting from the largest stand, these were put under high-stand management only if the demand required that more than 15 % of the pixel was put under high-stand management. If so, PFTs were managed as high stand until the total high-stand area in GEN-MAP was exceeded or there were no more forest PFTs. The 15 % threshold was introduced to avoid artifacts from the fact that one PFT is assigned a single management strategy. For example, a pixel dominated by few PFTs, say 90 % of PFT1 and 10 % of PFT2, and with a low fuelwood requirement, say 5 %, would nevertheless be under coppice management to 90 % because the algorithm dictates starting with the dominant PFT. The 15 % threshold protected against this artifact. Any remaining forests were left unmanaged.
Furthermore, a cap was set on the maximal share of unmanaged forest within a pixel during the 20th century. The cap starts at 100 % in 1910 (i.e., a whole pixel can be left unmanaged) and reduces to 15 % by 1950.
Results and discussion
Land cover reconstruction
As a base layer for our management reconstructions, we used the reconstruction of land cover change by . Preindustrial agricultural land use in Europe is substantially higher than in other common reconstructions . Following , our reconstruction mirrors their central assumption that European countries experienced forest transitions during the mid 1800s . In Europe, forest area reached its all-time low around 1850, after which encroachment and afforestation made the forest area increase again to surpass the forest area of 1600 only in the late 1900s (Fig. ). In this study, forest area was further detailed by filling in tree species fractions and their spatial and temporal dynamics. The 539 000 increase in conifer area outnumbers the decrease in deciduous forest by 55 000 . While conifers made up 29 % of the forest area in 1600, their present-day share has increased to over 56 %. Since 1600, the net increase in conifer forests is thus the single most important net land dynamic and exceeds the net change in agricultural land by a factor of 10 (Fig. ).
Land surface area in covered by forest (blue), grassland (orange), cropland (brown) and bare soil (grey) between 1600 and 2010. The forest area is further separated into coniferous (dark blue) and deciduous (light blue) forest.
[Figure omitted. See PDF]
It is well documented that Europe's present-day species distribution does not reflect its natural vegetation . It has been estimated that between 1500 and 2000, about 1300 different vascular plants species were introduced into Germany alone . Most of these neophytes were introduced for agricultural uses, but at least 11 tree species were introduced into forestry . In addition to introducing new species, humans are thought to have affected the distribution and share of tree species by favoring one species over another. Beech (Fagus sylvatica L.), for example, is thought to be responsible for the decrease of fir (Abies alba Mill.) . For the period under study, planting conifer species outside their natural growing area (e.g., Pinus sylvestris L. and Picea abies H. Karst.), the introduction of new species (e.g., Pseudotsuga menziesii Franco and Picea sitchensis Carriére) , the conversion of mixed boreal forest into homogeneous conifer forests and the onset of a production-oriented forestry all contributed to a dramatic change in European species distribution. Recent efforts in converting conifer forest back to mixed or deciduous forest were not taken into account and, therefore, not reflected in our reconstruction (Fig. ). Although such efforts have received substantial attention in forestry literature, their large-scale spatial extent remains to be quantified .
In this study, the reconstruction of changes in species distribution is based on anecdotal evidence as found in historical archives and forced to match present-day observations . Although our reconstruction uses national estimates for trends in conifer coverage, the neighborhood function that was used to decide which conifer species was planted during conversion indirectly takes climatic preferences of tree species into account. By design, however, our reconstruction cannot be expected to reproduce small-scale patterns such as the preferential use of spruce at elevated cool sites. This limitation is partly compensated for by the rather coarse resolution (i.e., ) of the reconstruction.
Demand and supply (1600–1828)
The 410-year time span of our reconstruction by far exceeds the 100- to 150-year life expectancy of a European forest. Living archives for forest management are thus absent for most of our reconstruction period. Given that forest management aims at satisfying diverse resource demands, we believe that in the absence of these archives, forest management can be reconstructed by making use of wood demand and supply estimates. Between 1600 and 1828, our estimates indicate that more than half of the wood demand came from the industrial production of ceramics, glass, beer, distillates and metal works (Fig. ). The remaining demand was mainly for construction and fuelwood, with wood for salt production being negligible. These estimates assume unlimited access to forest resources. However, as early as 1717, the estimated demand exceeds the estimated supply in our reconstruction. Given that Sylva, or a Discourse of Forest-Trees and the Propagation of Timber in His Majesty's Dominions by the English writer John Evelyn (1664) and Colbert's ordinance from 1669, known as l'aménagement forestier, were likely inspired by locally decreasing forest resources, a European-wide crossover 5 decades later seems reasonable. If our estimates reflect history, the crossover may have reshaped society by triggering a search for alternative energy sources and by playing a role in the colonial wars in search of wood resources to close the demand–supply gap ; two events the consequences of which still affect today's society.
Total estimated wood demand (stacked surfaces) and wood supply (dashed line) in between 1600 and 1829. The wood demand distinguishes between fuelwood for households (grey), fuelwood for salt production (not visible), fuelwood for industrial processes (blue) and timber used for construction (brown). Expansion factors and assumptions of this reconstruction are detailed in Table .
[Figure omitted. See PDF]
The demand–supply gap suggests unresolved issues with our estimates on the demand side, the supply side, or both the demand and supply sides. On the European scale our historical supply and demand estimates are of the same magnitude as present-day wood production and consumption, which provides empirical evidence that our historical estimates are at least feasible. Furthermore, independent estimates for fuelwood demand indicate that our estimates are on the low end (Table ). A cross-check of the charcoal demand based on our methodology against a single independent estimate indicates a fivefold overestimate by our method for England in 1700 (Table ). The mismatch in surface area was further amplified because we used a productivity of 4 , whereas the independent estimate used 6.5 . The latter seems closer to present forest production than to historical production expected in 1700. For Finland, however, total wood demand obtained with our methodology appears to be on the low side (Table ). Validation against independent estimates thus suggest that our demand estimate has a realistic order of magnitude and leans towards the low end of available estimates.
Similar to our approach for construction wood, the demand for glass and iron could have been scaled to gross domestic product. Given the negligible share of the wood demand for glass production, its scaling method is likely not relevant for our reconstruction. Such a scaling method could, however, more easily account for the decreasing marginal consumption of wood with increasing gross domestic product per capita. Such an approach would increase realism by partly decoupling wood demand from population growth, but it requires more data to establish the relationship between gross domestic product per capita and the use of glass, metal and construction wood.
As both the total forested area and the production per unit of land area increased between 1600 and 2010, the estimate for historical wood supply should be lower than the present-day supply. The increase in the area of forest land since 1600 is well established and is reflected in numerous historical documents as well as regional maps dating back to the late 1700s. One of the most compelling examples of such maps is represented by the Ferraris maps, which archived land cover of the Austrian Netherlands (now largely Belgium) in great detail . When combined with more recent topographic surveys, a clear and objective picture emerges of the land cover changes. Similar maps exist across Europe, e.g., Atlas de Trudaine (France, 1740–1780), Atlas of Napoleonic Cartography in Italy (1795–1815) and the Gyllenhielms atlas (Småland, Sweden, 1633–1655). It is this type of information that forms the scientific basis of land cover reconstructions, including that of , which was used in this study.
It is well established that productivity has increased over the past 6 decades, and this increase has been attributed to (1) the use of more productive, often exotic, species (for example Pseudotsuga menziesii Franco and Picea sitchensis Carriére), (2) elevated atmospheric nitrogen deposition , (3) fertilization and climate change , (4) an almost complete abandonment of practices such as forest grazing and litter raking , and (5) a more scientific approach to forest management . Combining historical archives with bookkeeping methods suggests that productivity has increased by 100 % since 1830 . Resampling Fagus sylvatica L. and Picea abies H. Karst. showed increased tree growth of 32 to 77 % and increased stand volume growth of 10 to 30 % since 1870 . An analysis covering the period 1970 to 1990 documents a general increase in productivity all over Europe by 25 % in 20 time . The 100 % increase in productivity applied in this study should thus be considered a high rather than a low estimate. Consequently, our estimate of historical wood supply should be considered low rather than high.
If site-level productivity by 1600 was only reduced by 30 % and thus agreed with relatively short-term observations , the demand would not have exceeded the supply until after 1815. Applying present-day definitions to forest and harvest for our reconstruction could also have contributed to an underestimation of the wood supply. When pressure on forest resources was high, it cannot indeed be excluded that branches, twigs and litter were collected and burned. Although it is labor intensive, stumps may have been uprooted and used for charcoal production , which could increase the supply by 15 to 25 % in the boreal and temperate zone and even 25 to 50 % for Mediterranean species . Furthermore, the strict separation between forest and agricultural lands emerged only at the onset of scientific forestry and modern nation states . The aforementioned early maps confirm that hedges, tree rows and solitary trees between and within fields and grazing lands were abundant in agricultural landscapes, in particular in the extensive heathlands used for rough grazing. A tree cover of 15 % across agricultural lands, suggesting common use of agroforestry, would delay the crossover by 15 years. Finally, importing 25 of wood C, which is a quarter of today's import , would have delayed the crossover by 40 . It seems unlikely, however, that fewer people with a much smaller logistical capacity than today could have imported such a volume even from Europe's forest-rich colonies.
Our reconstruction of wood demand uses a relationship between population density, economic development and consumption. However, wars and collapses of political systems may have decoupled population growth from wood demand for years to decades. Our reconstructions, for example, show no drop in demand during the Thirty Years' War (1618–1648), as is expected based on building dates of roughly 200 000 wooden structures across Europe (U. Büntgen, personal communication, 2014). During wars and crises, forest resources may even have regenerated from earlier (over)use . Regions with resource scarcity may have imported commodities such as glass and iron rather than importing the wood to produce these products themselves. Furthermore, we may have overestimated the share of charcoal, which, in turn, would lead to a substantial overestimation of the wood demand. A unit of energy released by charcoal required twice as much wood in its production compared to burning wood directly (see Table ). Recycling may have been more common than thought, e.g., decommissioned construction wood may have been used as fuelwood. Recycling all construction wood as fuelwood would reduce the wood demand by 25 % and delay the crossover by 60 .
During the centuries under study, industrial procedures evolved, e.g., slag
from Roman iron works was reused thus requiring less energy for smelting
, and a procedure was discovered to use coked coal rather than
charcoal for the extraction of iron , and as a consequence wood
as a construction material was replaced by iron . It is not
clear whether these increases in efficiency resulted in higher production at
the same amount of wood consumption or effectively reduced the wood demand.
Present-day studies show that in a market economy, effective substitution is
only likely under resource scarcity . Finally, alternative
energy sources such as solar and wind energy (for example, for pumping water
into evaporation pans for salt production; ), the burning of peat,
manure and agricultural residuals , and the initially local
rise in the use of coal may all have contributed to a lower
wood demand than estimated in this study. We interpret these data as a clue
that the replacement of wood as an everyday fuel source in the home likely
began to occur in the early 1800s
Although the crossover of demand and supply suggests that more work is needed and that most likely both estimates need to be better constrained, the key question for this study is whether these estimates are expected to substantially affect our forest management reconstructions. The assumption that wood did not travel outside the pixels is the key to answering this question. This appears to be a fair assumption, in particular for salt production, where channels to guide the brine rarely exceeded 40 . Transport of fuelwood was limited, especially in areas with no convenient river transportation . However, other evidence does indicate that long-range timber transport was common. Transport of fir in Roman times, for example, covered hundreds of kilometers .
Following the assumption of limited wood transport, the management maps are largely affected when wood supply exceeds wood demand at the pixel level because, in this case, a fraction of the forest will remain unmanaged. If however, wood demand exceeds its supply, as is common from our reconstruction, all forest in the pixel will be managed, but the mismatch between demand and supply is not propagated into neighboring pixels or to subsequent time steps. Given the priority rules that were set to allocate the fractions to different management practices, an unsatisfied demand is likely to result in a higher share of coppice management compared to high-stand management. Despite the mismatch between wood supply and demand, our management maps will correctly reflect that forests were under pressure but may result in a biased allocation of the management strategies towards coppice management. In 1600, our reconstruction suggests that at the pixel level, the percentage of unaccounted-for demand across Europe is around 40 % and increases to around 65 % by 1820. This reflects the increased forest pressure due to increasing population and provides support for an increased incentive to transition to alternative fuel sources.
Litter demand
Finding comprehensive data to parameterize and validate historical litter-based activities is not currently possible. To our knowledge, there exist two quantitative studies on traditional non-timber forest uses . The quantitative studies were localized in Swiss valleys and showed that the model results were sensitive to data such as litter demand for livestock and straw production rate, data which are impossible to find for all of Europe for different times between 1600 and 2010. Present-day soil carbon stocks that experienced historical litter raking were estimated to be 2 % lower than they would have been in the absence of litter raking . The effects of litter raking on the nutrient balance of soils is expected to be more important than its effect on the carbon balance .
When litter raking is accounted for in ORCHIDEE-CAN simulations, the litter
maps give an estimate of the amount of litter to be removed from each grid
cell at the end of every year. If a grid cell does not have enough litter to
cover the demand, we remove all, but we do not attempt to take any litter from
surrounding grid cells. Litter raking was used for wintering livestock, with
the manure collected and spread on the fields in the spring. Hence,
ORCHIDEE-CAN adds the carbon contained in the raked litter to the litter
pools of any agricultural PFTs on the grid square. Owing to the tight link
between forest use and agriculture in this epoch, the main result of litter
raking is thus clear: carbon which would enter the forest soil if the litter
were not raked was diverted to litter pools outside the forest. At
its peak, litter raking was estimated to collect 50 of dry litter
and thus to redistribute 25 of carbon and 0.5 of nitrogen
Forest management reconstruction
Trends in forest management
Our forest management reconstruction shows a clear and continuous decrease in unmanaged forest between 1600 and 1950 (Fig. ). A spatial representation shows that between 1600 and 2010, unmanaged forest was mainly lost in Scandinavia and eastern Europe (Fig. a, b). Although few will argue against the reconstructed decline, a share of 30 % of unmanaged forests in 1600 cannot be confirmed (or rejected) on the European scale. The gain in unmanaged forest since 1950, which is scattered over Europe, reflects a modest but increasing effort in forest conservation and the recent trend towards abandoning management by an increasingly urban population. Our reconstruction suggests that 17 % of the forest is unmanaged. At present the share of untouched forest or forest with minimal human intervention is, however, estimated to be just below 5 % . However, that does not include a notable fraction of public and private forest, which is only lightly managed and with harvest rates well below the annual increment . Overestimation of unmanaged forests may also be partly caused by classification issues. The ORC-MAPS were prepared to be used with ORCHIDEE-CAN and based on a remote-sensing product. Given that ORCHIDEE-CAN currently does not distinguish between forest and shrubland, dense shrubland was classified as forest and sparse shrubland as grasslands. After 1950, our forest management relies on forest inventory data, which, by their nature, exclude shrubland. Our reconstruction algorithm thus classified dense shrubland as unmanaged forest. Although classifying shrubland as forest results in overestimating basal area and vegetation height over much of the Mediterranean region , the consequences of considering shrubland as unmanaged ecosystems are expected to be minor.
Total area ( ) of unmanaged (grey), high-stand (light blue) and coppice (dark blue) forests between 1600 and 2010. The full lines show the best available estimates, whereas the dashed lines show the sensitivity of the management reconstruction to the “minimal importance” threshold (see Sect. ).
[Figure omitted. See PDF]
The abrupt decreases in 1829 and 1950 in unmanaged forests in Figure are due to Finnish and by extension Scandinavian forestry. In 1950, Finland had a lot of forests that, according to the inventory data, were more than 121 years old. Given that these forests were in the inventory data and thus under commercial use in 1950, our algorithm has put all these forests under management in 1829. In 1829, the backcasting algorithm thus required much more managed forests than the demand–supply algorithm that was used prior to 1829. Furthermore, the algorithm caps the unmanaged forests in the 20th century (see Sect. ). Recall that the cap starts at 100 % in 1910 (i.e., a whole pixel can be left unmanaged) and reduces to 15 % by 1950. The cap was found to be redundant until 1932 but became substantial after 1942. By then, the maximum amount of unmanaged forest on a pixel was capped at 30 %, which is the case for many pixels. Hence, the algorithm placed a lot of previously unmanaged forests under management between 1940 and 1950. Although the origin of the abrupt decrease in unmanaged forests in 1950 is an artifact of the algorithm, it happens to relate to a decade in which a lot of unmanaged forest were harvested to support post-war reconstruction efforts .
Coppice management groups diverse forest management strategies that leave the root system intact but may differ in rotation length, tree species and wood use depending on region and climate. Although the wood from coppice stands was often used as fuelwood in households, its use is also documented for construction, agriculture, viticulture and tannin production . Part of its popularity was due to its ecological stability, flexibility in management and common rights. Throughout history, coppicing has been the basis for silvo-pastoral systems with temporary cropping, husbandry, hunting, gathering and beehives . Between 1600 and 1828, population growth drove the reconstructed wood demand, resulting in an increasing share in coppice management albeit accompanied by an overall decrease in forest area (Fig. ). After 1828, an increasing substitution of fuelwood with coal and of construction wood with metals made coppice lose its importance (Fig. e, f). Coppice stands were converted or neglected. At present, independent estimates suggest that 15 % of the European forest is currently under coppice management. Outside the coppice strongholds in southeast Europe (Fig. f), high-stand forests in Europe often show relics of the typical features of coppice management. Although there is little doubt that coppice management has substantially decreased since 1600, the reconstructed share of 36 % of coppice in 1600 cannot be confirmed (or rejected) on the European scale.
Distribution of management strategies over Europe. (a) Unmanaged forests in 1600 and (b) in 2010. (c) High-stand forests in 1600 and (d) in 2010. (e) Coppice management in 1600 and (f) in 2010. The color bar shows the fraction of the forest managed by a given strategy. Pixels along the coastlines often contain few forests, frequently resulting in the entire pixel being under a single management strategy.
[Figure omitted. See PDF]
During the World Wars and the interbellum, our reconstructions suggest a revival of coppice management (Fig. ). Whether this reflects reality is uncertain, but during the wars, rotations were considerably shortened to meet the increased demand. For example, in the New Forest, UK, all conifers of 20 to 35 years were cut between 1940 and 1946 . Although the reconstruction suggests that coppice management in 2010 is at its lowest in over 400 (Fig. ), it has been rediscovered because of its contribution to biodiversity and as a source of renewable bioenergy . In the latter case, fewer species and shorter rotations are used compared to traditional coppice management, leading to the term short-rotation coppice (SRC). At present short-rotation coppice is limited to about 500 and was therefore not considered substantial enough to be taken into account in our reconstruction.
In the late 1700s it was realized that when used in a controlled manner, forest could produce wood without being destroyed. This changing attitude towards natural resource management was reflected by the establishment of German, Austrian and French forestry schools and is considered to be the start of the so-called scientific forest management which is based on the idea that the productivity of forests is stable over time. The focus on productivity resulted in the widespread use of conifers at sites that were previously only occupied by deciduous tree species. Since the late 1700s a variety of management strategies that differ in objectives in the and intensity and frequency of the interventions has emerged . This variety is grouped as high-stand management in our reconstruction. Following centuries of forest destruction this then-new approach to land management may have contributed to a steady increase in forest area over Europe (Fig. ), along with broader economic, technological, cultural and political changes .
Our management reconstruction also allocates forest to high-stand management prior to the birth of scientific forest management. Given the specific objective of our reconstruction, i.e., studying large-scale effects of forest management, it was considered acceptable to classify both scientific forest management in the late 1700s and unplanned forest use prior to the late 1700s as high-stand management. In both approaches, some wood is removed prior to the final harvest, and tall vegetation is maintained. Although the outcome of scientific high-stand management is expected to be sustained production, the outcome of unplanned management could be sustained production, degradation or even forest destruction. Given the success of scientific high-stand management in sustaining wood production and providing other ecosystem services, its increasing use and dominance between 1600 and 2010 (Fig. ) are to be expected.
Quality of the reconstructions
Although 80 % of the data and expansion factors used in this reconstruction (see Table ) are justified by previous research, its spatial and temporal representativeness remains unknown as the data are most often based on local case studies. Estimates of several components are based on sparse data, and for these components, it was not possible to separate the data into parameterization and validation sets. Where sufficient data were available (Table ), the reconstruction was compared against validation data. Although our reconstruction performs reasonably well across space and time, the quality and representativeness of the validation data are unknown. Given the lack of quality-checked data for validating our reconstructions, we aimed for the highest transparency by listing all assumptions made to reconstruct forest management (Table ). Wood removal activities were deliberately separated by various forms of wood harvesting and litter raking, which is expected to allow easier incorporation of future advances. The resulting maps should thus be considered as a starting point for future developments on the topic rather than a final data product.
The uncertainty of the forest management reconstruction must be considerable given that it relies heavily on population and land cover reconstructions, which are also uncertain . When creating the ORC-MAPs it was assumed that within a pixel, a single tree species was managed with a single strategy. Large artifacts from this assumption were avoided by only putting a tree species under a specific management strategy if the demand required that more than 15 % of the pixel was put under that specific management (see Sect. ). Many of the previous assumptions are believed to be of minor importance compared to this 15 % threshold. Therefore, the effect of the 15 % threshold was tested by recreating the ORC-MAP for a threshold of 5 and 25 %. The trends in forest management between 1600 and 2010 were found to be insensitive to the specific threshold used (Fig. ). Although different settings for this threshold could result in a 5 to 25 % increase or decrease in the surface area for a given management strategy, these changes are not random and tend to favor one management strategy over another. Hence, using different threshold values rarely changed the dominance of a certain strategy within a given time period.
In this study, demand and supply maps were used to reconstruct forest management between 1600 and 1828, and a supply-based backcasting method was applied to reconstruct forest management between 1829 and 2010. This implies that except for the year 1829, the management reconstruction for a single year was based on a single approach. Continuing the reconstruction of demand and supply maps would enable us to constrain the management reconstruction between 1829 and 2010 by an independent data set. It remains to be tested whether this would really result in more reliable management maps. An extended and improved estimate of the supply should depend on the reconstructed forest management because different management strategies will supply different amounts and dimensions of wood. Also, a demand map for 1829–2010 would have to deal with a very dynamic period in terms of energy consumption and wood use. In the end, the uncertainty of the 1829 to 2010 demand reconstruction may be so large that it would only marginally constrain the backcasting method. By contrast, the backcasting method relies on a more restricted data set with fewer assumptions.
Ways forward
Future efforts to reconstruct forest management could move this field forward by considering the following things:
Our reconstruction builds on land cover reconstructions but does not interact with this land cover reconstruction. Simultaneous consideration of both wood demand and the food demand that currently drives these land use reconstructions during their creation would increase the internal consistency of the land use and thus the forest management reconstructions.
An interdisciplinary team with access to physical land science, historical archives in local languages, archaeological archives and expertise in economic history could better constrain the physical, sociological and economic aspects that underlie the forest management reconstruction. A broader expertise would not only further internal consistency of the reconstruction but is likely to increase the number of data sources that could be used for parametrization and validation.
It is well known that forest management strategies other than coppicing and high-stand management were used historically in Europe . Hence, using regional information could help to refine and expand the management strategies currently distinguished. Such a regional approach would likely result in the addition of continuous cover management (central Europe) , coppice with standards (central, southern Europe and the UK) and swidden (northern Europe) .
The historical species distribution is one of the few components of the reconstructions which could be improved based on comprehensive evidence by making use of pollen, charcoal and genetic records. Recent efforts in bringing together pollen records from across Europe and studies on genetic lineages are expected to refine and better constrain the spatial and temporal resolution of future reconstructions of species distribution. Charcoal dating and analyses could add further detail on species occurrence.
The limited sensitivity tests we conducted suggested that better knowledge of the supply side will be essential as will a better understanding of wood trade and both of production techniques and the demand of energy-intensive products such as metals.
Data and map availability
ORCHIDEE-CAN is a version of the land surface model ORCHIDEE that was developed and parametrized to simulate the biogeochemical and biophysical effects of forest management . Applications making use of the full potential of ORCHIDEE-CAN require a spatially and temporally explicit description of forest management as boundary condition and, hence, the need for a reconstruction of forest management as presented in this study. Contrary to the forest management reconstruction, the demand and supply maps are intermediate products and not essential for driving ORCHIDEE-CAN. Wood supply, for example, is at present calculated by ORCHIDEE-CAN. In the future, modeled wood supply, rather than our supply map, could thus be used to derive the most likely management strategy. Nevertheless, we decided to make all maps available to other research groups. The demand, supply and management maps as well as the coded algorithms (scripts in the R language) can be requested from one of the authors. Every use of (part) of these maps and their underlying data other than the original products (Table ) should cite this paper. It is strongly advised to contact the authors to verify the correct use and interpretation of the maps.
Conclusions
By making use of land use reconstructions, present-day vegetation maps, tree species maps, expansion factors and forest age reconstructions, we created general (GEN-MAP) and model-specific (ORC-MAP) historical forest management maps. Where the GEN-MAPS can serve as a starting point for other modeling groups to create their own maps, the ORC-MAPS were created to be used with the ORCHIDEE-CAN land surface model. The ORC-MAPS assign a management strategy (unmanaged, high-stand management or coppice management) to each forest PFT in Europe. Our reconstruction reproduces the most important changes in forest management between 1600 and 2010: (1) an increase of 593 000 in conifers at the expense of deciduous forest (decreasing by 538 000 ); (2) a 612 000 decrease in unmanaged forest; (3) a 152 000 decrease in coppice management; (4) a 818 000 increase in high-stand management, and (5) the rise and fall of litter raking, which at its peak in 1853 removed 50 dry litter per year.
M. J. McGrath and S. Luyssaert wrote the manuscript. M. J. McGrath wrote all scripts and created the maps. J. Kaplan, U. Gimmi, M. Bürgi, M.-J. Schelhaas, K. Erb, P. Meyfroidt and D. McInerney provided the majority of the expansion factors for estimating wood demand and wood supply. M. J. McGrath, S. Luyssaert, J. Ryder, Y. Chen, K. Naudts, J. Otto and A. Valade developed ORCHIDEE-CAN such that it can make use of the forest management and litter maps. All authors contributed to discussing the approach and its caveats.
Acknowledgement
M. J. McGrath, J. Ryder, Y. Chen, K. Naudts, J. Otto and S. Luyssaert were funded through ERC starting grant 242564 (DOFOCO) and A. Valade was funded through ADEME (BiCaFF). The study benefited from early discussion with Gert-Jan Nabuurs and Geerten Hengeveld and additional funding through ECV CCI Landcover and an STSM (COST, TERRABITES ES805) offered to M. J. McGrath. S. Luyssaert, A. Valade, K. Erb and P. Meyfroidt benefited from the ISSI International Team on “Integrating Earth Observation data and the description of land management practices into global carbon cycle models”. Edited by: E. Veldkamp
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© 2015. This work is published under http://creativecommons.org/licenses/by/3.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
Because of the slow accumulation and long residence time of carbon in biomass and soils, the present state and future dynamics of temperate forests are influenced by management that took place centuries to millennia ago. Humans have exploited the forests of Europe for fuel, construction materials and fodder for the entire Holocene. In recent centuries, economic and demographic trends led to increases in both forest area and management intensity across much of Europe. In order to quantify the effects of these changes in forests and to provide a baseline for studies on future land-cover–climate interactions and biogeochemical cycling, we created a temporally and spatially resolved reconstruction of European forest management from 1600 to 2010. For the period 1600–1828, we took a supply–demand approach, in which supply was estimated on the basis of historical annual wood increment and land cover reconstructions. We made demand estimates by multiplying population with consumption factors for construction materials, household fuelwood, industrial food processing and brewing, metallurgy, and salt production. For the period 1829–2010, we used a supply-driven backcasting method based on national and regional statistics of forest age structure from the second half of the 20th century. Our reconstruction reproduces the most important changes in forest management between 1600 and 2010: (1) an increase of 593 000 km
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
Details
; Meyfroidt, P 2 ; Kaplan, J O 3
; Bürgi, M 4 ; Chen, Y 1
; Erb, K 5 ; Gimmi, U 4 ; McInerney, D 6 ; Naudts, K 1
; Otto, J 7 ; Pasztor, F 1
; Ryder, J 1 ; M-J Schelhaas 8 ; Valade, A 9
1 Laboratoire des Sciences du Climat et l'Environnement, IPSL, CEA-CNRS-UVSQ, 91191 Gif-sur-Yvette, France
2 F.R.S.-FNRS & Université catholique de Louvain, Earth and Life Institute, Georges Lemaître Centre for Earth and Climate Research, 3 Place Pasteur, 1348 Louvain-la-Neuve, Belgium
3 Institute of Earth Surface Dynamics, Université de Lausanne, Géopolis, 1015 Lausanne, Switzerland
4 Swiss Federal Research Institute, Zürcherstrasse 111, 8903 Birmensdorf, Switzerland
5 Institute of Social Ecology, Alpen-Adria Universität, Klagenfurt-Wien-Graz, Schottenfeldgasse 29, 1070 Vienna, Austria
6 Laboratoire des Sciences du Climat et l'Environnement, IPSL, CEA-CNRS-UVSQ, 91191 Gif-sur-Yvette, France; now at: Coillte Teoranta, Irish State Forestry Board, Newtownmountkennedy, Co. Wicklow, Ireland
7 Laboratoire des Sciences du Climat et l'Environnement, IPSL, CEA-CNRS-UVSQ, 91191 Gif-sur-Yvette, France; now at: Helmholtz-Zentrum Geesthacht, Climate Service Center 2.0, Hamburg, Germany
8 Alterra, Droevendaalsesteeg 3, 6708 PB Wageningen, the Netherlands
9 Institut Pierre Simon Laplace, CNRS, Université Pierre et Marie Curie, 4 Place Jussieu, BP 99, 75252 Paris, France