1

A GEOGRAPHICALLY EXPLICIT APPROACH FOR PRICE

DETERMINATION OF FOREST FEEDSTOCK UNDER DIFFERENT

NEXT-GENERATION BIOFUEL PRODUCTION SCENARIOS – THE

CASE OF SWEDEN

Ismail Ouraich, Luleå University of Technology, +46920493519, ismail.ouraich@ltu.se

Robert Lundmark, Luleå University of Technology, +46920492346, robert.lundmark@ltu.se

Abstract The objective of this paper is to develop a geographically-explicit model of price determination of forest biomass

for Sweden. The model uses a simple demand-supply framework to determine new equilibrium prices at the

gridcell level. Data on forest biomass supply and harvest cost is available for Sweden for 334 0.5x0.5 degree

gridcells. We use the data to construct regional supply curves, and investigate the impacts on equilibrium prices

of increased demand for forest biomass for bioenergy production under a number of scenarios generated via the

BeWhere-Sweden model. We run simulations for two energy production scenarios from forest biomass: 10 TWh

and 20 TWh. The scenarios represent biofuel targets aimed at in Sweden for the decoupling of the transportation

sector from fossil fuels by 2030. Overall, the results indicate that the price of forest resources would only

moderately increase. The average price increase does not exceed one percent. The simulation results show that

the spatial distribution of price changes does not track spatial distribution of demand pressure, which holds for

the 10 TWh and 20 TWh scenarios. The largest impacts are observed in the southern and middle parts of

Sweden, despite large endowments of forest, owing to high demand clustering.

Introduction In recent decades, the paradigm of transitioning from a fossil fuel-based economic system to a sustainable bio-

based one has gained much traction in policy circles, motivated by a number of interlinked issues such as

reduction of greenhouse gas emissions, energy security and independence, climate change, etc. To this end, a

number of countries have devoted substantial resources in developing technologies for transportation biofuels,

based on feedstock from either the agricultural sector or the forestry sector, broadly categorized as first and

second generation biofuel technologies, respectively. However, due to growing concerns over negative spill-over

effects (e.g., food security) from using agricultural feed-stocks, the development of second generation

technologies has become more important. The expansion of second generation production facilities will also

affect the feed-stock markets. Thus, an important question is how large-scale developments of second generation

transportation biofuels will affect the market price for forest resources? This will not only resonate back to the

feasibility of the investment decision itself but also affect the competitive situation for the other areas the forest

resources are used. Moreover, the local and regional price effects might be more severe than what might be

suggested by using national averages. Spatial considerations and dimensions are therefore important to

incorporate in order to obtain reliable results and implications.

For second generation technologies, the main industrial framework covered in the literature centres around the

concept of ‘biorefineries’. A biorefinery can be described as sustainable processing of biomass into a wide range

of marketable products and energy carriers [1]. This can be either in the form of stand-alone production facilities

or as add-ons to existing industrial processes, i.e., the production of chemical pulp. Initial research efforts

focused on the technological dimension of producing transportation biofuels and energy from forest resources. In

more recent years, and as new bioenergy conversion technologies mature, the focus shifted to studies on the

techno-economic feasibility of large-scale implementation of biorefineries. To this effect, the empirical literature

has focused primarily on model development, either by system analyses or by analysing new and improved value

chains of biorefineries. The focus of such models spans a number of themes that covers issues related to e.g., the

procurement costs of forest resources, logistics, and optimal localization of biofuel conversion capacity. An

important strand of literature empirically analyse biofuel supply chains, integrated into existing industrial

processes [2-5]. Another development in the literature is the explicit treatment of the spatial dimension. Much of

the modelling effort uses geographical information system (GIS)-based models that explicitly account for the

spatial dimension [6-11], and/or a hybrid approach which uses a techno-economic routine of cost-minimization

of the whole value-chain, all the while explicitly incorporating the spatial dimension [12-20].

The objective of this paper is to develop a spatial price determination model of forest resources that accounts for

the potential impacts of increased demand pressure from an expanding second generation biofuel industry. Price

determination is based on a bottom-up supply-demand framework, where local supply schedules are constructed

2

and used in conjunction with estimates of potential demand, at a gridcell level. As such, the model adopts a

spatial disaggregation of determinants. The developed framework will be applied on Sweden. Forest resources

are the major feedstock in the production of transport biofuels in Sweden. This is can be explained by (i) the

large endowment of forests in Sweden and (ii) a mature, large and export-orientated forest industry with

considerable flows of by-products suitable for biofuel value chains and technical adaptation. The increasing

production of transport biofuels using forest resources is in direct competition with the more traditional uses of

the resources from the forest industries. In addition, conservation, recreational use and ecosystem services are

also competing for the same resources. Many studies have pointed to the potential that exists in Sweden with

respect to the usage of forest resources for energy production [21-27]. In connection to that, a number of studies

have developed spatially-explicit harvest cost model and/or hybrid models [4, 28-31]. Lundmark [28] argue that

increased bioenergy production will not cause a major disruption in the supply of forest resources to the

traditional forest industries. Other studies in the Swedish context support this finding. For instance, empirical

studies on the factor substitution in the forest industries and the heating industry indicate that there exists a

potential for negative impacts on the heating industry from an increased feed-stock competition. However, the

parameter estimates are in general low, which suggests that the level of competition would remain relatively

limited [24, 30, 32]. Other studies, nonetheless, show that the feed-stock competition will on a global scale

increase significantly. Many factors will have an impact on this process. For instance, demand for forest

resources is projected to increase, driven primarily by increasing demand from developing economies. As global

market integration continues, coupled with reduced transport cost, the trade in forest resources is also expected

to increase [26]. This is also true for the trade within EU, which is expected to be driven by the promotion of

bioenergy targets, in general and by transportation biofuels targets in particular [33]. Countries with large forest

endowments will most likely have a major role in meeting the targets. Sweden has been shown to have a

moderate potential to contribute to EU production of bioenergy by trading forest resources to EU [34].

Swedish forest sector In 2015, gross felling reached 92.6 Mm

3sk

based on the latest estimates from the Swedish Forest Agency’s

statistical database.1 The net felling volume was 74.4 million cubic metres solid volume excluding bark (Mm

3f

ub), of which 35.9 Mm3f ub of sawlogs (48.3%), 30.9 of pulpwood (41.5%), 7.1 of fuelwood (9.5%), and other

roundwood accounted for 0.5 Mm3f ub (0.7%). Figure 1 depicts the spatial distribution of gross felling volumes

in Sweden for 2013. Noticeable is that the northern and central counties have the highest gross felling. This is

connected to the fact that those counties also have the largest share of forest endowments.

The harvested volumes were predominantly used by the pulp and paper industry and the sawmill industry. They

accounted for 41 and 57 percent of the total net felling, respectively. Between 1995 and 2012, sawmills and the

pulp and paper mills increased their consumption of forest resources by 5 and 14 percent, respectively. The

overall industrial consumption of forest resources increased steadily from 59.7 in 1975 to its peak of 86.1 Mm3f

ub in 2005. This corresponds to an increase by 44 percent over the period. The consumption volume has declined

somewhat after 2005 due to various reasons. In 2012, the total consumption reached 81 Mm3f ub (incl.by-

products), which is a 6 percent decline compared to 2005 [35]. The declining trends can partly be explained by a

decline in the production capacity. Nonetheless, if all areas of usage are included, the total utilisation of forest

resources exhibits an increasing trend during the last decades. The decline in the industrial usage has been more

than off-set by an increasing use in the energy sector. Indeed, an increasing share of forest resources is being

converted to energy. Approximately 129 terawatt-hours (TWh) out of a total energy supply of 565 TWh comes

from biomass, which represents 23 percent of total energy supply. The increase in bio-based energy supply is

projected to continue owing to policy reforms that aim to decouple the Swedish economy from fossil fuels.

In terms of the spatial distribution of productive forest land and gross felling volumes, we notice that the

northern and middle counties in Sweden represent the largest share in terms of forest land endowment, and hence

in terms of supply of wood material. However, when investigating the county-level spatial distribution of gross

felling yield2 for productive forest land, we notice that the southern counties exhibit higher yields compared to

the northern counties (Figure A1, Appendix).

1 See

http://pxweb.skogsstyrelsen.se/Default.aspx?px_language=sv&px_db=Skogsstyrelsens%20statistikdatabas&rxid

=57840e53-1f66-45c7-bee8-a2a7dc29fabf [Last accessed 10/3/2016] 2 We compute gross yield of productive forest land in Sweden by taking the ratio of gross felling volumes (in

1000 m3sk) and productive forest land area (in 1000 ha).

3

Harvesting volumes and costs The supply-side dynamics is captured through the construction of regional supply curves using data on harvested

volumes and harvesting cost for different wood assortment, harvesting operation and geographical area

(gridcell)[31]. The supply covers four types of wood assortment: sawlogs, pulpwood, branches & tops, and

stumps. The harvesting operations are separated by final felling and thinning. Table 1 summarizes the total

volume harvested and average harvesting cost, by wood assortment and harvesting operation.

Table 1: Aggregate harvested volume and average harvesting cost in Sweden

Total supply (TWh yr-1) Average harvest cost (MEUR TWh-1)

Final felling

Sawlogs 75.3 20.7

Pulpwood 32.3 13.7

Branches & Tops 20.2 14.6

Stumps 16.0 22.1

Thinning

Sawlogs 14.1 25.1

Pulpwood 37.0 16.7

Branches & Tops 10.8 16.0

Source: Authors’ adaptation (Datasource: Lundmark et al. (2015) [31])

The spatial distribution of harvested volumes indicates that the harvesting is mainly concentrated in the southern

and central regions of Sweden. This finding holds true for all the wood assortments included in the analysis.

However, the spatial distribution for harvested volumes in thinning operations displays a stronger concentration

in the identified regions compared to final felling. Nonetheless, this finding is not surprising especially since the

spatial distribution of the gross felling yield of productive forest land in Sweden.

The harvesting cost is calculated per geographical area and is based on the results from Lundmark et al. [31].

They are constructed using a bottom-up approach using productivity measures together with unique

topographical and geographical information for the specific are. With respect to the spatial distribution of

harvesting cost, a fairly homogeneous cost structure across Sweden can be observed. The highest harvesting

costs are observed in the northern region of Sweden. However, there exists a substantial difference between the

spatial distribution of harvesting cost for final felling and thinning operations. For final felling, the harvesting

costs are characterized by a uniform spatial distribution, whereby higher costs occur in a limited scope in the

northern regions. This finding holds for all type of wood raw material. But for thinning operations, a different

spatial pattern can be observed. Indeed, it seems that the southern, and to a limited extent, the northern regions of

Sweden exhibit a strong concentration of higher harvesting cost in thinning operations compared to the central

regions of the country.

Demand scenarios The demand-side of the model consists of two parts. The first part describes the demand changes derived from an

expanding production of transportation biofuels to satisfy different policy targets. These demand changes are

results obtained from BeWhere-Sweden model simulations [29]. BeWhere-Sweden is a techno-economic,

geographically explicit optimization model to determine optimal localization and conversion technology of

integrated biorefineries. The strength of the model consists in its explicit treatment of spatial aspects of supply

and demand of forest resources from different sectors. The optimization routine of BeWhere-Sweden consists of

the cost minimization of the entire system cost, which includes harvest cost, transportation cost, and investment

costs of different technology alternatives, and various policy instrument constraints (e.g. taxes, subsidies). The

model is expressed as linear programming problem. The spatial dimension is captured via 334, 0.5x0.5 degrees,

gridcells (50km x 50km) representation of Sweden.

With respect to the demand side, we obtain input data under different scenarios of biofuel production targets in

Sweden from the BeWhere-Sweden model [29]. The data is provided at a spatial resolution identical to the data

on supply and harvest cost. The biofuel scenarios characterizing the demand for wood raw material pertain to

projected increase in biofuel production to supply transportation fuel in Sweden, which aims at achieving a

complete decoupling of the transport sector from fossil fuels by 2030. To this effect, the analysis includes three

scenarios (their effects are summarized in Table 2):

Business-as-usual (BAU) scenario, which captures current demand for wood raw material from its

different uses (incl. energy generation);

10TWh scenario, which represents the BAU scenario + a 10 terawatt hour biofuel production target for

use in the transportation sector; and

4

20TWh scenario, which represents the BAU scenario + a 20 terawatt hour biofuel production target for

use in the transportation sector.

Table 2: Demand scenarios in level and in percent change from BAU included in the analysis

Demand scenarios (in TWh yr-1)

BAU 10TWh 20TWh

Level Change (%) Level Change (%) Level Change (%)

Final felling Sawlogs 69.5 n.a. 71.2 3% 73.6 6%

Pulpwood 32.1 n.a. 32.2 0.2% 32.2 0.2%

Branches & Tops 14.6 n.a. 18.7 28% 19.9 36%

Stumps 1.1 n.a. 4.8 322% 10.7 835%

Thinning Sawlogs 5.7 n.a. 8.1 41% 12.4 117%

Pulpwood 36.4 n.a. 36.8 1% 36.9 1%

Branches & Tops 6.6 n.a. 9.9 50% 10.5 60%

Source: Authors’ adaptation (Datasource: Lundmark et al. (2015) [31], Wetterlund et al. (2013) [29])

The spatial distribution of demand indicates that for most of the wood assortments, the demand seems to be

concentrated primarily in the southern and, to a lesser extent, the central regions of Sweden under the BAU

scenario. This comes as no surprise owing to the fact that population and industrial density is higher in the latter.

However, it can be observed that demand peaks in particular locations in the northern regions. Comparing the

spatial distribution of demand for the scenarios suggest only small differences. For most of the wood

assortments, the demand intensity seems to be located within the same regions regardless of production target

(Figures A2-A8, Appendix).

Method This section will outline a bottom-up, spatially-explicit, supply-demand model of price determination for

geographically dispersed resources and fixed demand nodes. The geographical partition is based on

approximately 50x50 km gridcells. In each gridcell, a potential harvesting volume and cost is calculated per

wood assortment and harvesting operation. These volumes and costs constitute the building blocks for the

‘merit-order’ regional supply curve. In each gridcell a regional supply curve can be constructed for each wood

assortment by ranking the enclosed gridcells’ harvesting cost – from lowest to highest – and where the width of

the cost-order is the harvesting volume. The enclosed gridcells is determined by assuming a maximum

transportation distance (or by adding a constant transportation cost per km). This approach allows the

construction of a regional supply curve in any arbitrary gridcell.

The demand changes are exogenously given and reflect the change in demand in a specific gridcell in which new

or changed biofuel production capacity has been added. In order to account for the potential demand competition

between different locations (and uses), an adjusted-demand is estimated using a distance-decay framework.

Historically, distance-decay functions represent the workhorse tool that evolved in two separate fields of

research: spatial interaction studies in the social sciences and gene dispersal in genetics [36]. Within the

economic field, the most widely used model of spatial interaction is the gravity model, which is based on

Newtonian physics. The ‘gravity model’ has been used extensively within the international trade literature in

recent years. The objective of such models is to analyse the interaction between two ‘masses’ or ʻflowsʼ (e.g.

exports and/or imports flows, etc.) and how distance impacts competition among different locations [36-42]. For

the purpose of this study, a distance-decay adjusted demand is estimated in order to capture the potential

competition dynamics of demand nodes for the forest resources. This is achieved through the following:

𝑋𝐷𝑖,𝑚𝑎𝑑𝑗

= 𝑋𝐷𝑖,𝑚 + ∑ 𝑋𝐷𝑖,𝑚,𝑛𝑑𝑒𝑐𝑎𝑦

𝑛≠𝑚 (𝐸𝑞. 1)

and

𝑋𝐷𝑖,𝑚,𝑛𝑑𝑒𝑐𝑎𝑦

= 𝛼 × 𝑋𝐷𝑖,𝑚 × (1 − (𝑑𝑚,𝑛

𝑑𝑚𝑎𝑥)𝛾

) (𝐸𝑞. 2)

where 𝑋𝐷𝑖,𝑚𝑎𝑑𝑗

is the distance-decay adjusted demand for wood assortment 𝑖 in gridcell 𝑚; 𝑋𝐷𝑖,𝑚 is the demand

for wood assortment; 𝑋𝐷𝑖,𝑚,𝑛𝑑𝑒𝑐𝑎𝑦

is the distance-decay demand flow from gricell 𝑚 to gridcell 𝑛; 𝑑𝑚,𝑛 is the

distance matrix; 𝑑𝑚𝑎𝑥 is the maximum distance range and ; 𝛼 and 𝛾 are calibration parameters.

5

The equations suggest that an increasing demand in a specific gridcell will radiate outwards to neighbouring

gridcells with a diminishing effect. The effect will accumulate in gridcell that are affected by more than one

demand change. As a consequence, the most intense demand effect might be observed in gridcells within the

vicinity of several expected capacity expansions, but that does not have a capacity expansion itself. The main

objective is to capture how the expected increase in the demand for forest resources in a particular location (i.e.,

gridcell) competes with the demand from other locations. Figures A9-A15 (Appendix) display the results of the

distance-decay demand adjustment. The spatial price determination model is implemented using GAMS.

Results From results, we observe that the spatial distribution of the adjusted demand displays substantial changes when

comparing the BAU vs. 10TWh and the BAU vs. 20TWh scenarios. This is not a surprise, owing to the fact that

increased production of biofuels from wood material will put increased pressure on their demand. For sawlogs

from final felling and thinning, we notice that the locus of high demand locations is concentrated in the southern,

and to a lesser extent, middle regions of Sweden (Figures A9 and A13, Appendix). The situation seems to be

reversed for pulpwood, where we observe that demand seems to be clustered in the in the middle regions of

Sweden, primarily eastward along the coast where urban agglomerations are largest (Figures A10 and A14,

Appendix). For both sawlogs and pulpwood, the spatial distribution does not exhibit differential across scenarios,

i.e. BAU vs. 10TWh vs. 20TWh. The only difference corresponds to the degree of diffusion and level of the

adjusted demand.

A different situation emerges for branches & tops and stumps. Whereas there does not seem to be stark

differences between the 10TWh and 20TWh scenarios in terms of spatial distribution; there exists, however, a

significant shift in the adjusted demand between the BAU scenario on one hand and the 10TWh and 20TWh

scenarios on the other hand. Indeed, we notice that under the BAU scenario, the southern regions in Sweden

concentrate the locus of locations with the highest level of demand. Whereas these regions contain still an

important locus of demand point, we observe that the locus of the highest demand points shifted to the mid-

northern regions under the 10TWh and 20TWh scenarios (Figures A11-A12 and A15, Appendix).

As argued earlier, we use regionally-constructed supply curves and the distance-decay adjusted demand to gauge

the potential impact of increased production of biofuels on market prices of wood raw materials. Table 4

summarizes the aggregated results in terms of percent impacts.

In terms of magnitude of impacts, we notice that prices are projected to increase the most for pulpwood and

branches & tops. In addition, we notice that the impacts are slightly larger for final felling compared to thinning.

For pulpwood, the price increase ranges from 0.004% to 18.95% for final felling and 0.01% to 14.19% for

thinning; whereas for branches & tops, it ranges from 0.001% to 7.17% for final felling and from 0.01% to

1.91% for thinning. However, we notice that the average percent change in prices is not high where it reaches

0.93% and 0.99% for pulpwood from final felling and thinning respectively. Similarly, average impact for

branches & tops reaches 0.45% and 0.38% for final felling and thinning respectively. This indicates that the loci

of locations that exhibit the maximum impacts in terms of price change are few and limited in numbers. For

stumps from final felling and sawlogs from thinning, the results did not yield any impact on prices. Hence, there

are no results reported (Table 3).

In terms of the spatial distribution of the price changes, we notice that the latter behaves as expected given that it

matches the spatial distribution of demand. However, the magnitude of the price does not always correspond to

areas where the demand pressure is higher. This is so given that we must take into consideration the spatial

distribution of supply. Indeed, some of the highest impacts in terms of price change are located in locations with

relatively low-to-moderate demand pressure. However, these locations are also characterized by a relative supply

scarcity of wood material. This is typically the case in northern regions in Sweden. Another finding is that there

exists little change with respect to impact on prices between the two biofuel scenarios analysed. The impact

seems to be more pronounced, as previously argued on aggregate, for final felling vs. thinning. Also, the level of

dispersion of price impacts is larger for final felling compared to thinning operations (Figures A16-A20,

Appendix).

6

Table 3: Summary of impacts on prices of wood raw material by biofuel production scenarios in level and

in percent change

Scenario 1: 10 TWh Scenario 2: 20 TWh

Level (MEUR/TWh) Percent change Level (MEUR/TWh) Percent change

Avg Max Min Avg Max Min Avg Max Min Avg Max Min

Fin

al f

elli

ng

Sawlogs 19.8 20.7 19.1 0.23 1.13 0.002 19.8 20.7 19.1 0.23 1.13 0.002

Pulpwood 13.3 16.7 12.7 0.93 18.95 0.004 13.3 16.7 12.7 0.93 18.95 0.004

Branches

& Tops 13.9 15.7 13.6 0.45 7.17 0.003 13.9 15.7 13.6 0.45 7.17 0.001

Stumps 21.2 22.1 20.8 n.a. n.a. n.a. 21.2 22.1 20.8 n.a. n.a. n.a.

Th

inn

ing Sawlogs 24.1 25.7 22.5 n.a. n.a. n.a. 24.1 25.7 22.5 n.a. n.a. n.a.

Pulpwood 16.4 20.5 15.0 0.99 14.19 0.01 16.4 20.5 15.0 0.99 14.19 0.01

Branches

& Tops 15.6 16.8 14.9 0.38 1.91 0.01 15.6 16.8 14.9 0.38 1.91 0.01

Conclusions In this paper, an original, spatially-explicit price determination model is developed. The model is used to analyse

the price impacts of increased transportation biofuel production in Sweden and its spatial characteristics. The

model is applied using a scenario-analysis of potential increased transportation biofuel production in Sweden.

Two scenarios are developed for 10 and 20 TWh production of transportation biofuels using second generation

technologies per year.

Overall, to achieve the scenario goals using the location, input mix and technology choice suggested by

BeWhere-Sweden, the price of forest resources would only moderately increase. The average price increase does

not exceed one percent. The simulation results show that the spatial distribution of price changes does not track

spatial distribution of demand pressure, which holds for the 10 TWh and 20 TWh scenarios. The largest impacts

are observed in the southern and middle parts of Sweden, despite large endowments of forest, and this is due to

high demand clustering owing to population density, industrial cluster, etc. However, relatively large impacts

can be observed in the northern regions as well, especially for biomass obtained via the thinning operation. The

spatial distribution of the price changes differ depending on type of harvest operation (final felling vs. thinning).

Biofuel production targets might affect the spatial distribution, but relatively minor.

References 1. IEA, IEA Bioenergy Task 42 on Biorefineries - Co-production of Fuels, Chemicals, Power and Materials

from Biomass. 2007, IEA: In: Minutes of the third Task meeting. Copenhagen, Denmark, 25–26 March 2007.

2. Leduc, S., et al., Optimal location of lignocellulosic ethanol refineries with polygeneration in Sweden. Energy, 2010. 35(6): p. 2709-2716.

3. Leduc, S., Development of an Optimization Model for the Location of Biofule Production Plants. 2009, Luleå University of Technology.

4. Pettersson, K., et al., Integration of next-generation biofuel production in the Swedish forest industry – A geographically explicit approach. Applied Energy, 2015. 154: p. 317-332.

5. Wetterlund, E., et al., Optimal localisation of next generation biofuel production in Sweden - Report from an f3 R&D project. 2013, f3 Centre.

6. Ranta, T., Logging residues from regeneration fellings for biofuel production–a GIS-based availability analysis in Finland. Biomass and Bioenergy, 2005. 28(2): p. 171-182.

7. Melo, M.T., S. Nickel, and F. Saldanha-da-Gama, Facility location and supply chain management - A review. European Journal of Operational Research, 2009. 196(2): p. 401-412.

8. Yoshioka, T., et al., A GIS-based analysis on the relationship between the annual available amount and the procurement cost of forest biomass in a mountainous region in Japan. Biomass and Bioenergy, 2011. 35(11): p. 4530-4537.

9. Gold, S. and S. Seuring, Supply chain and logistics issues of bio-energy production. Journal of Cleaner Production, 2011. 19(1): p. 32-42.

7

10. Höhn, J., et al., A Geographical Information System (GIS) based methodology for determination of potential biomasses and sites for biogas plants in southern Finland. Applied Energy, 2014. 113: p. 1-10.

11. Zhang, F., et al., Decision support system integrating GIS with simulation and optimisation for a biofuel supply chain. Renewable Energy, 2016. 85: p. 740-748.

12. Parker, N., et al., Development of a biorefinery optimized biofuel supply curve for the Western United States. Biomass & Bioenergy, 2010. 34(11): p. 1597-1607.

13. Murphy, G., et al., Management tools for optimal allocation of wood fibre to conventional log and bio-energy markets in Ireland: a case study. European Journal of Forest Research, 2010. 129(6): p. 1057-1067.

14. Natarajan, K., et al., Optimal Locations for Methanol and CHP Production in Eastern Finland. BioEnergy Research, 2011. 5(2): p. 412-423.

15. Wetterlund, E., et al., Optimal localisation of biofuel production on a European scale. Energy, 2012. 41(1): p. 462-472.

16. Jenkins, T.L. and J.W. Sutherland, A cost model for forest-based biofuel production and its application to optimal facility size determination. Forest Policy and Economics, 2014. 38: p. 32-39.

17. Natarajan, K., et al., Optimal locations for second generation Fischer Tropsch biodiesel production in Finland. Renewable Energy, 2014. 62: p. 319-330.

18. Santibañez-Aguilar, J.E., et al., Optimal planning and site selection for distributed multiproduct biorefineries involving economic, environmental and social objectives. Journal of Cleaner Production, 2014. 65: p. 270-294.

19. Santibañez-Aguilar, J.E., et al., A mixed-integer dynamic optimization approach for the optimal planning of distributed biorefineries. Computers & Chemical Engineering, 2015. 80: p. 37-62.

20. Cambero, C., et al., Strategic optimization of forest residues to bioenergy and biofuel supply chain. International Journal of Energy Research, 2015. 39(4): p. 439-452.

21. Johansson, J. and U. Lundqvist, Estimating Swedish biomass energy supply. Biomass & Bioenergy, 1999. 17(2): p. 85-93.

22. Ericsson, K.H., S. Nilsson, L. J. Svenningsson, P., Bioenergy policy and market development in Finland and Sweden. Energy Policy, 2004. 3: p. 1707-1721.

23. Geijer, E., G. Bostedt, and R. Brannlund, Damned if you do, damned if you do not-Reduced Climate Impact vs. Sustainable Forests in Sweden. Resource and Energy Economics, 2011. 33(1): p. 94-106.

24. Olsson, A., Examining the Competition for Forest Resources in Sweden Using Factor Substitution Analysis and Partial Equilibrium Modelling. 2011, Luleå University of Technology.

25. Poudel, B.C., et al., Potential effects of intensive forestry on biomass production and total carbon balance in north-central Sweden. Environmental Science & Policy, 2012. 15(1): p. 106-124.

26. Jonsson, R., How to cope with changing demand conditions - The Swedish forest sector as a case study: an analysis of major drivers of change in the use of wood resources. Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere, 2013. 43(4): p. 405-418.

27. Routa, J., et al., Forest energy procurement: state of the art in Finland and Sweden. Wiley Interdisciplinary Reviews-Energy and Environment, 2013. 2(6): p. 602-613.

28. Lundmark, R., Cost structure of and competition for forest-based biomass. Scandinavian Journal of Forest Research, 2006. 21(3): p. 272-280.

29. Wetterlund, E., et al., Optimal localisation of next generation biofuel production in Sweden. 2013, The Swedish Knowledge Center for Renewable Transportation Fuel.

30. Olsson, A., Modelling the Competition for Forest Resources: The Case of Sweden. Journal of Energy and Natural Resources, 2014. 3(2): p. 11.

31. Lundmark, R., D. Athanassiadis, and E. Wetterlund, Supply assessment of forest biomass – A bottom-up approach for Sweden. Biomass and Bioenergy, 2015. 75: p. 213-226.

32. Lundmark, R. and A. Olsson, Factor substitution and procurement competition for forest resources in Sweden. International Journal of Production Economics, 2015. 169: p. 99-109.

33. Commission, E., Directive 2009/28/EC of the European Parliament and of the Council of 23 April 2009 on the promotion of the use of energy from renewable sources., in Official Journal of the European Union, E. Commission, Editor. 2009. p. 16-62.

34. Egnell, G., H. Laudon, and O. Rosvall, Perspectives on the Potential Contribution of Swedish Forests to Renewable Energy Targets in Europe. Forests, 2011. 2(4): p. 578-589.

35. Skogsstatistiska årsboken (2014). 2014, Skogsstatistiska. 36. Taylor, P.J., Distance transformation and distance decay functions. Geographical Analysis, 1971.

8

37. Fotheringham, A.S., Spatial Structure and Distance-Decay Parameters. Annals of the Association of American Geographer, 1981. 71(3): p. 425-436.

38. Skov-Petersen, H., Estimation of distance-decay parameters - GIS-based indicators of recreational accessibility. 2001, Danish Forest and Landscape Research Institute.

39. Reggiani, A., P. Bucci, and G. Russo, Accessibility and Network Structures in the German Commuting. Networks & Spatial Economics, 2011. 11(4): p. 621-641.

40. Martínez, L.M. and J.M. Viegas, A new approach to modelling distance-decay functions for accessibility assessment in transport studies. Journal of Transport Geography, 2013. 26: p. 87-96.

41. Schaafsma, M., et al., Estimation of Distance-Decay Functions to Account for Substitution and Spatial Heterogeneity in Stated Preference Research. Land Economics, 2013. 89(3): p. 514-537.

42. Halás, M. and P. Klapka, Spatial influence of regional centres of Slovakia: analysis based on the distance-decay function. Rendiconti Lincei, 2015. 26(2): p. 169-185.

9

Appendix

Figure A1: County-level spatial distribution of productive forest land (in 1000 ha) (left), gross feeling volumes (in 1000 m3sk) (center) and gross feeling yield (in m

3sk ha

-1)

(right)

Data source: Skogstatistika årsboken, 2014 [35]

10

Figure A2: Spatial distribution of demand of sawlogs from final felling under the 10TWh (left) and

20TWh scenario (right)

Figure A3: Spatial distribution of demand of pulpwood from final felling under the 10TWh (left) and

20TWh scenario (right)

Figure A4: Spatial distribution of demand of branches & tops from final felling under the 10TWh (left)

and 20TWh scenario (right)

11

Figure A5: Spatial distribution of demand of stumps from final felling under the 10TWh (left) and 20TWh

scenario (right)

Figure A6: Spatial distribution of demand of sawlogs from thinning under 10TWh (left) and 20TWh

scenario (right)

Figure A7: Spatial distribution of demand of pulpwood from thinning under the 10TWh (left) and 20TWh

scenario (right)

12

Figure A8: Spatial distribution of demand of branches & tops from thinning under the 10TWh (left) and

20TWh scenario (right)

13

Figure A9: Spatial distribution of distance-decay adjusted demand of sawlogs from final felling under the

10TWh (left) and 20TWh scenario (right)

Figure A10: Spatial distribution of distance-decay adjusted demand of pulpwood from final felling under

the 10TWh (left) and 20TWh scenario (right)

Figure A11: Spatial distribution of distance-decay adjusted demand of branches & tops from final felling

under the 10TWh (left) and 20TWh scenario

14

Figure A12: Spatial distribution of distance-decay adjusted demand of stumps from final felling under the

10TWh (left) 20TWh scenario (right)

Figure A13: Spatial distribution of distance-decay adjusted demand of sawlogs from thinning under the

10TWh (left) and 20TWh scenario (right)

Figure A14: Spatial distribution of distance-decay adjusted demand of pulpwood from thinning under the

10TWh (left) and 20TWh scenario (right)

15

Figure A15: Spatial distribution of distance-decay adjusted demand of branches & tops from thinning

under the 10TWh (left) and 20TWh scenario (right)

16

Figure A16: Spatial distribution of price impacts for sawlogs from final felling under 10TWh scenario

(left) and 20TWh scenario (right) in Sweden (in %change from BAU)

Figure A17: Spatial distribution of price impacts for pulpwood from final felling under 10TWh scenario

(left) and 20TWh scenario (right) in Sweden (in %change from BAU)

Figure A18: Spatial distribution of price impacts for branches & tops from final felling under 10TWh

scenario (left) and 20TWh scenario (right) in Sweden (in %change from BAU)

17

Figure A19: Spatial distribution of price impacts for pulpwood from thinning under 10TWh scenario

(left) and 20TWh scenario (right) in Sweden (in %change from BAU)

Figure A20: Spatial distribution of price impacts for branches & tops from thinning under 10TWh

scenario (left) and 20TWh scenario (right) in Sweden (in %change from BAU)