e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/ecolmodel Balancing fuelwood and biodiversity concerns in rural Nepal Morten Christensen a,∗ , Santosh Rayamajhi a,b , Henrik Meilby a a b University of Copenhagen, Forest & Landscape Denmark, Rolighedsvej 23, 1958 Frederiksberg C, Denmark Tribhuvan University, Institute of Forestry, Pokhara, Nepal a r t i c l e i n f o a b s t r a c t Article history: An agent-based model is developed to explore the pattern of fuelwood collection in an Received 24 June 2008 1178 ha forest area in rural mountainous Nepal. The model relates fuelwood collection inten- Received in revised form sity and amount of dead wood available for collection to the diversity of polypore species, a 23 October 2008 group of strictly dead wood dependent fungi which can be used as indicators of the biodi- Accepted 24 October 2008 versity associated with dead wood. By analysing scenarios of increased collection the model Published on line 6 December 2008 shows that the relative impact on polypore diversity is rising more rapidly than the time used for collection. This indicates that better market access in the future could potentially Keywords: Agent-based modelling imply a major threat to biodiversity associated with dead wood. To assess the potential for biodiversity conservation we evaluated the effect of protect- Community-based management ing areas with high values of polypore diversity. The simulation results showed such area Firewood protection strategies to be effective for short-term protection of polypore diversity only in Polypore the event of a dramatic increase in the local market price of fuelwood. In case of smaller changes in fuelwood prices a collection quota system appeared to be the most suitable protection strategy. However, area protection is an important strategy for long-term protection of biodiversity associated with dead wood and, therefore, we conclude that a combination of small-protected zones and collection quotas seems to be the most promising strategy for protection of the forest. © 2008 Elsevier B.V. All rights reserved. 1. Introduction The balance between protection of biodiversity and the local use of natural resources is a challenge for most developing countries (Millennium Ecosystem Assessment, 2005) and exploration of new sustainable solutions to the dilemma is essential for rural development. Fuelwood for cooking and heating is one of the most important products harvested from the forests of most developing countries (Arnold et al., 2006; Cooke et al., 2008). In many areas a major source of fuelwood is dead wood generated by natural disturbances or natural competition, for which less restrictive legislation applies than for felling of living trees (Cooke et al., 2008). Dead wood is also a very important habitat for biodiversity in forests (e.g. Huston, ∗ Corresponding author. E-mail address: moc@life.ku.dk (M. Christensen). 0304-3800/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2008.10.014 1996; Lonsdale et al., 2007) and supports a wide range of organisms. Recent figures from boreal forests indicate that as much as one third of all forest-living organisms depend on dead wood in some, or all, stages of their lifecycle (Siitonen, 2001; Jonsson et al., 2006). Dead wood in a forest is generated in several ways (Christensen et al., 2005). In forest ecosystems without human disturbance dead wood is generated by the competition between the trees and by natural disturbances, such as wind throw, fire, flooding, ice break or attack by pathogens. Dead wood left in the forest will eventually be decomposed. The type and speed of decomposition is specific to each tree species and highly dependent on macro- and microclimatic conditions like temperature and humidity (Stokland, 2001). Polypores e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 constitute an important group of wood-decaying fungi often used as indicators for biodiversity conservation value (Norstedt et al., 2001; Christensen et al., 2004; Junninen and Kouki, 2006; Similä et al., 2006). Most species are strictly dependent on certain amounts and types of dead wood and removing dead wood completely from the forest floor is therefore a potential threat to them. The diversity of polypores is also known to represent the diversity of other dead-wood dependent organisms rather well (Similä et al., 2006). Compared to other categories of fungi polypores are useful as indicators because of their persistent fruit bodies, recognizable for a long period of the year. In rural areas of Nepal wood from natural forests is one of the most important sources of fuelwood. Per capita annual fuelwood consumption varies between 400 and 1500 kg fresh weight, mainly depending on altitude (Metz, 1994; Amacher et al., 1999; Stræde and Treue, 2006). In many forests fuelwood is primarily collected from pools of dead wood and dying trees and the large consumption poses a direct threat to the important part of the biodiversity depending on the dead-wood habitat. Community-based management of forest resources may potentially motivate sustainable fuelwood collection (Cooke et al., 2008). For the last two decades an important strategy for conservation of biodiversity in Nepal has been the establishment of integrated conservation and development projects (ICDPs) where local user groups are involved in forest management. The oldest ICDP is the nongovernmental Annapurna Conservation Area Project (ACAP) that has been in place since 1986 (Baral et al., 2007). Agent-based models where individual agents interact with each other and their environment to exploit a natural resource is a recently developed approach to explore scenarios of change (see Bousquet and Le Page, 2004; Matthews et al., 2007 for reviews). The total behaviour of a system depends on the aggregated results of individual decisions made by each agent (Matthews et al., 2007). So far, only few agentbased models address the direct impact of human behaviour on biodiversity (e.g. Linderman et al., 2005). In the present study each household acts as one agent who makes rational decisions regarding collection area, thereby minimising the time allocated to fuelwood collection. Their decisions are based on dynamic information on spatial distribution of dead wood (potential fuelwood). The diversity of wood-inhabiting polypores is used as a proxy for biodiversity related to dead wood. The model is used to assess the potential effects of two different fuelwood collection restrictions on conservation effectiveness and the economic consequences of such restrictions for forest users. 2. Methods 2.1. Research site A model was developed to describe fuelwood collection behaviour and dead wood distribution in part of Lete and Kunjo Village Development Committees (VDC, an administrative unit) of Mustang District, Central Nepal (83.58◦ E–83.66◦ E; 28.61◦ N–28.66◦ N). The mountain forest within the study area covers 1178 ha (Fig. 1). The altitude ranges from 2200 to 3000 m, 523 the average annual precipitation is 1242 mm (1971–2002) and the rainfall peaks in June to September. The yearly average temperature is 11.7 ◦ C (1976–1986) and the monthly average ranges from a maximum in July of 18 ◦ C to a minimum of 4 ◦ C in February (information from Department of Hydrology and Meteorology). The area is dominated by temperate coniferous and mixed forest. Pinus wallichiana, Tsuga dumosa and Rhododendron arboreum are main tree species. Cupressus torulosa, Abies spp., Ilex dipyrena, Taxus baccata, Betula alnoides and Acer spp. are less dominant or restricted to smaller parts of the forest area. Many species of smaller trees occur in the area. Among them particularly Coriaria nepalensis, Hippophae salicifolia, Lauraceae spp., and Viburnum erubescens are important sources of fuelwood. The forest area is used by inhabitants of eight small villages within the two VDCs. The total number of households is 220 and the total population is 1065. The western part of the area is strongly influenced by a major trekking and transport trail connecting low-lying parts of Nepal with remote areas near the Tibetan border. Twenty major tourist hotels operate in Lete. In 2006 these hotels served approximately 26,000 over-night visitors. All private households use fuelwood mainly for their cooking and heating. Some households and all hotels supplement fuelwood with liquefied petroleum (LP) gas, kerosene and electricity. Forest management is organised and implemented by two local functional conservation committees, one in each VDC, representing the local users in all eight settlements. The conservation committees are supervised by a ranger from ACAP. 2.2. Fuelwood resource mapping In total 123 permanent 20 m × 25 m sample plots were distributed using stratified random sampling and a MaxiMin strategy, implemented as a ‘Coffee-House’ strategy within each forest stratum (Müller, 2001), ensuring that plots were distributed evenly to all parts of each stratum (Fig. 1, see Meilby et al., 2006 for details). Dead wood was measured within 122 plots in November 2006, immediately before the main season for fuelwood collection. We therefore assume the amount measured to represent the maximum available amount of fuelwood in a year. Standing trees and snags taller Fig. 1 – Study area. Small squares are permanent sample plots. Black dots are human settlements (villages). 524 Table 1 – Applied decay classes based on Stevens (1997), Kruys et al. (1999), Fraver et al. (2002) and Ódor et al. (2006). Decay class Bark Twigs, needles and branches Softness, texture, penetrability Intact for all species Twigs, needles and branches present 2 Missing especially when sun exposed; Intact for Picea, Pinus, Cupressus, Betula Missing or partly intact for most species; Intact for Betula Twigs and needles absent. Branches (>3 cm) present. branch stubs are firmly attached Branches absent and branch stubs pull out easily 4 Missing for most species; Partly intact for Betula Absent 5 Missing Absent Soft and powdery 6 Missing Absent Soft and powdery, partly reduced to mould, only core of wood 3 Shape of log Shape of stump Standing dead tree/snag Wood is sound (hard) and cannot be penetrated with thumbnail and a knife penetrate 1-2 mm only Hard or partly soft, knife penetrate less than 1 cm Covered by bark, outline intact Circular Cylindrical With twig and branches Smooth, outline intact Circular Cylidrical, bark often lost Without twigs and branches Begins to soften, thumbnail penetrates readily, knife penetrates 1–5 cm Soft, thumbnails and knife penetrate readily, often blocky pieces Smooth or crevices present and small pieces lost, outline intact Circular Cylindrical but with crevices, Easy to turn over Very soft, easy to turn over Large crevices, small pieces missing, wood fragments often lost so the outline of the trunk is deformed Large pieces missing, outline deformed Outline hard to define Circular or elliptic Conical, very soft Not present (fallen) Flat elliptic, partly burried in the soil Hardly visible Not present (fallen) Flat elliptic covered by soil Not visible Not present (fallen) e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 1 Surface e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 than 1.3 m, lying stems and fallen branches with diameter ≥10 cm were measured within each 20 m × 25 m plot. Standing and fallen stems, and branches with diameter 4–9.9 cm were measured within an interior 10 m × 15 m sub-plot, and branches with diameter 2–3.9 cm were measured within a 10 m × 5 m sub-plot. For standing dead trees the breast height diameter and the height of the remaining part of the tree were measured. Fallen branches and stems were divided into sections of maximum 2 m and the exact length and the diameter at each end were recorded. Fallen stem/branch fragments less than 1 m long were not considered. We distinguished six decay classes defined by a set of criteria regarding the occurrence and state of bark, branches, twigs and foliage, texture and surface characteristics of the wood. Definitions of the six classes are presented in Table 1 and for classes relevant for logs (fallen trees), stumps and dead standing trees (snags) the visual appearance is sketched in Fig. 2. Analysis of fuelwood samples collected from the villages shows that the wood density of most fuelwood used for cooking and heating is similar to that of dead wood in decay classes 1–2. In the present study we therefore included dead wood in these two decay classes only. Conversion from dead-wood volume in the forest to oven dry weight was based on mean densities estimated for each decay class. The number of wood samples collected for the five decay classes was 272. For decay classes 1 and 2 the average densities were 446 kg/m3 (n = 57, S.D. = 82) and 331 kg/m3 (n = 73, S.D. = 108), respectively. 2.3. Biodiversity measures Polypores were surveyed within 57 of the permanent plots. Plots not included in the polypore survey were without woody vegetation or inaccessible in the rainy season. Moreover, 25 plots had not yet been established by the time that the survey started and therefore were not included. The plots were visited five times over 2 years (2005–2006) within the main fungal season (April, June, September, October and November). The fruit bodies were searched on fallen logs and standing dead and living trees, and on fallen branches and other smaller debris. The definition of polypores is according to Núñez and Ryvarden (2000, 2001) and only wood-inhabiting species is included. Identification was mainly based on Núñez and Ryvarden (2000, 2001). Voucher specimens are deposited in Japan (TNS) and Kathmandu (KATH). The biodiversity index (D) was defined as the total number of species observed within a 20 m × 25 m plot during the observation period (2005–2006). 2.4. Inventory of collection and consumption The private forest in the study area is negligible and almost all fuelwood is gathered in the common property forest for which the model was developed. Fuelwood is mainly gathered in the dry winter season when the wood is relatively dry and light and almost all fuelwood collected is dead wood. According to the household survey further described below, 86% of the fuelwood is collected from September to March. Additional informal information suggests that collection takes place almost exclusively from December to March due to the timing of agricultural and tourist activities. Extraction and processing in the forest is done by axe, hand saw and sickle. Transportation is almost exclusively done by humans, carrying loads with an average stated weight of 38 kg (n = 62, S.D. = 2.5 kg) corresponding to 24.7 kg dry weight (see below). In a sample of 42 randomly selected households from all eight settlements and all 20 major hotels/guesthouses fuelwood consumption was measured in four quarters of 2005–2006 (December, March, June and September). In each quarter fuelwood consumtion was measured over a period of 7 days. Sample households would use fuelwood from a pre-weighed stack only, and the remaining wood in the stack was reweighed every 24 h (e.g. Benjaminsen, 1993). The yearly fuelwood consumption per household ranged from a minimum of 2607 kg stored weight (corresponding to 2173 kg oven dry weight) to a maximum of 8825 kg stored weight (corresponding to 7354 kg oven dry weight). The average fuelwood consumption across all households in the study area was 5914 kg stored weight per household per year (corresponding to 4929 kg oven dry weight) (n = 62, S.D. = 1304 kg for stored weight, and S.D. = 1186 kg for dry weight) or 1222 kg stored weight per inhabitant (corresponding to 1018 kg oven dry weight). In the present study we used separate estimates for each of the eight settlements to avoid bias due to differences in consumption patterns. To enable conversion of the weight of fuelwood stored in households to dry weight we measured the water content of 586 samples from stored wood in stacks. Samples were randomly selected from the 62 intensively studied households during March, June and September 2006. Samples were taken to represent all major fuelwood species used in the study area. All samples were packed in sealed polybags and dried in an oven for 24 h at 100 ◦ C. The water content in stored wood was 20% (n = 586, S.D. = 8%). The water content of dead wood when gathered in the forest was taken from the literature (Metz, 1994) and set to 35% of fresh weight on average. In the model all weight measures are converted to oven dry weight. 2.5. Fig. 2 – Visual classification of dead wood (see also Table 1). 525 Modelling A spatial model was developed that included information on altitude, amount of dead wood and diversity of polypores for square 50 m × 50 m grid cells (0.25 ha) located in a spatial domain covering 8.2 km × 5.2 km. Two subsets of such cells, c, are defined: FOR includes the 4663 cells that are located in forest areas (1165.75 ha, i.e. slightly less than the mapped 1178 ha); RES are cells located in the forest for which fuelwood collection is restricted by a suggested conservation scheme (RES ⊆ FOR). For each cell the diversity of polypores, denoted 526 e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 by D, was estimated using a regression equation relating the richness of polypores to the amount of dead wood potentially available for fuelwood (kg dry matter per ha). The regression was prepared on the basis of the 57 plots that had been visited five times within the main fungal seasons of 2005 and 2006. Effects of altitude and aspect were tested but turned out not to be significant in regression models also including dead wood volume. To measure the consequences of particular exploitation patterns, i.e. their spatial distribution and intensity, we define a percentage measure of human impact (P) on the diversity of polypores. For the forest area as a whole the impact measure (PFOR ) is defined as PFOR = 100% ×  c ∈ FOR Dc− −  c ∈ FOR  D c ∈ FOR c− Dc+ (1) where Dc− is the predicted diversity index for polypores in cell c, had no fuelwood collection taken place, and Dc+ is the predicted diversity index when fuelwood collection is considered. Each of the 220 households in the study area is considered one agent in the model. Availability and accessibility are important variables determining the spatial patterns of fuelwood collection in the forest. Availability is in our study defined as the amount of potential fuelwood per hectare. Accessibility is defined as travelling time for the agent from village to collection site. The total time needed to extract a load of fuelwood (Ttotal , h), i.e. walk to a particular place in the forest, collect a load of fuelwood and walk home with the load, depends on accessibility and availability and can be expressed as Ttotal = (2 + a)Twalk + Tmin + Tabundance (2) where Twalk is the average time needed to walk to or from the site (half a round trip), a = 1 is a parameter expressing the assumed increase of time consumption when walking with a load, Tmin = 0.25 h is the assumed minimum time required to collect a load of fuelwood, i.e. the time needed at a site with a very high (∞) abundance of fuelwood, and Tabundance is the additional time needed due to lower availability. Based on our experienced travelling time to the 123 permanent forest plots in the research area we developed the travelling time regression (n = 123): Twalk = 23S + 8A 60 (3) where Twalk is walking time in hours, S is horizontal distance from house to collection site in kilometres, and A is the altitudinal difference in hectometres. Tabundance (h) was modelled as Tabundance = B 1 + M/1000 (4) where B is a parameter controlling the steepness of the function, and M is abundance of potential fuelwood in kg oven dry weight/ha. Thus, Tabundance = B when M = 0 and converges asymptotically towards zero when M increases. To enable prediction of fuelwood volume and polypore diversity for all 50 m × 50 m cells in the spatial grid we used kriging (Fig. 3). In both cases the empirical variogram suggested that an exponential variogram model would be suitable. In these models the estimated range parameters were 726 m for potential fuelwood and 247 m for polypore diversity. Predictions were based on local kriging including the 50 nearest points. The standard error of the kriging estimates ranged between 725 and 6837 kg/ha for potential amount of fuelwood (average S.E. was 2923 kg/ha) and between D values of 0.35 and 1.90 for polypore diversity (average S.E. 1.67). We simulated the annual extraction of fuelwood by i = 1, . . ., 220 households (agents). The exact spatial location of each household was not known and therefore the centre of each village was used to approximate the location of all households in the village. For each household the annual consumption of fuelwood was specified as a number of loads Ji . This number was determined as an average amount per household for each of the villages. For two smaller settlements the sample size was too small to allow estimation of separate consumption averages. In these cases the consumption values were estimated using a regression based on the consumption patterns observed in all eight settlements. The regression included information on household size, landholding and value of other financial assets. The simulation was done as follows (Fig. 4): One by one the households were allowed to collect one load of fuelwood; when all households had collected one load they were allowed, one by one, to collect their second load, their third load, their fourth, etc. When all households had collected all the loads (Ji ) that they needed the simulation terminated. Every time a load (38 kg fresh weight) of fuelwood was extracted from a cell the abundance of dead wood within the cell, M, was reduced by 98.8 kg/ha corresponding to 98.8 kg/ha × 0.25 ha = 24.7 kg dry Fig. 3 – Maps of (A) available fuelwood, 103 kg/ha and (B) diversity of polypores: D = species per 500 m2 plot. e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 527 Fig. 4 – Flow chart of the simulation model. Symbols: HH is a household; c is a 50 m × 50 m cell; M is the abundance of dead wood available in the cell; Twalk is the time required for a return trip to cell c; Tabundance is the time required to gather a load of fuelwood in cell c; Ttotal is the total time consumption to go and pick up a load in cell c; Tmin is the minimum total time required; Cmin is the cell for which Ttotal = Tmin . weight. This would lead to an increase in the time needed to collect additional loads within the cell (Tabundance ) and, due to the relationship between amount of deadwood and diversity of polypores, a reduction of the predicted diversity, D, and an increase of the impact index PFOR . Each load, ji = 1, . . ., Ji , was gathered in that particular cell which allowed the household to collect the load with the least possible effort, subject to previous extraction from the cell. Thus for each load, ji , the simulation solved: min c ∈ FOR\RES =  min  Ttotal (i, ji , c) c ∈ FOR\RES  (2 + a)Twalk (i, c) + Tmin + Tabundance (i, ji , c)  that the mean time consumption per load of fuelwood is 2.7 h, thereby matching the labour opportunity cost (Fig. 5). 2.6. Scenario analysis In the present dataset effects of altitude, distance to village and abundance of deadwood on polypore diversity are confounded. Since no true reference areas exist, conservation is evaluated with reference to current levels of fuelwood collection and biodiversity without assessing the consequences of current extraction levels. In other words, the present fuelwood collection, the abundance of deadwood and the associated (5) where Tabundance (i,ji ,c) depends on the initial abundance (Mc ) in c and the number of loads previously gathered in c, whereas Twalk (i,c) remains unchanged throughout the simulation. For each load of fuelwood the horizontal and vertical distances from the house, the time consumption (Ttotal (i,ji )), the dead-wood abundance at the time of collection (Mc,i,ji ) and the estimated biodiversity (Dc ) were recorded and used for calculation of summary statistics. After completing the simulation the impact measure (Eq. (1)) was computed for the forest as a whole (PFOR ). A range of alternatives for the value of B was tested (Fig. 5). In 2006 the average market price of one load of fuelwood was 70 Nepalese Rupees (approx. 1.00 USD). The average wage rate for unskilled labour in the area was approx. 25 Rupees per hour, indicating that one load can be collected in less than 3 h (70/25 = 2.8). Assuming that each household maximizes its utility a B value of 10 is applied in the simulations as it implies Fig. 5 – Distribution of time consumption per load for the total annual extraction simulated using three different values of the steepness parameter B (see Eq. (4)). 528 e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 diversity of polypores are assumed to be in equilibrium. However, the demand for fuelwood is likely to increase in the near future as Lete and Kunjo are about to be connected to markets further up and down the valley through a road presently under construction. In these markets the price of fuelwood is 140–180 Rupees, i.e. more than twice as much as in the study area, so fuelwood export is likely to become profitable once the road has been completed and transportation costs reduced. Improved road access is known to have increased the pressure on forest resources in many developing countries (e.g. Wilkie et al., 2000; Geist and Lambin, 2002), and has also been documented in Nepal (Gautam et al., 2004). The consequences of improved road access for extraction of fuelwood and other resources within the study area are yet to be seen. The simulation model is therefore used to predict the likely impact of various hypothetical extraction levels on biodiversity (PFOR ) and, as proxies for extraction cost, average and marginal time consumption for fuelwood collection. In addition, as a step towards identifying suitable conservation schemes, various restrictions on the area where fuelwood can be collected are introduced and the consequences for time consumption and human impact on biodiversity are examined. 3. Results 3.1. Dead wood The study area houses areas of remote virgin forest and a gradient in the amount of dead wood from 0 up to more than 400 m3 /ha. The average amount of dead wood useful for fuelwood was 12.2 × 103 kg dry weight per ha or approximately 14,500 × 103 kg in total. A geographical analysis shows a clear relation between the amount of dead wood and the distance to the nearest village, with high volumes only in remote places (Fig. 3a). Pinus wallichiana dominates the dead wood and accounts for more than 60% of the volume. Tsuga dumosa contributes 21% and Cupressus torulosa 8% of the volume of dead wood. Other important species are Rhododendron spp. (2%), Taxus baccata (<2%) and various deciduous trees and shrubs. 3.2. number of plots used in the regression was 34, the slope parameter was significant at the 1% level, the residuals were homogeneously distributed, and their normality could not be rejected (R2 = 0.21, RMSE = 0.55). 3.3. Fuelwood consumption From December 2005 to November 2006 the total annual consumption of fuelwood in the study area was estimated at 1408 × 103 kg stored weight in stacks, corresponding to 1126 × 103 kg oven dry weight. This corresponds to less than ten percent of the total available amount of dead wood. 3.4. Modelling of fuelwood and biodiversity The spatial distribution of the plots, the location of the villages and the altitude of the terrain are shown in Fig. 1. Villages are generally located in valleys and on plateaus. As will appear from the predicted abundance of potential fuelwood (Fig. 3A) areas close to the villages are in most cases characterised by low availability compared to areas further away. Areas with particularly high abundances are found in the eastern and south-western parts of the research area. The greatest diversity of polypores is found closer to the villages (Fig. 3B). Within large parts of the forest the estimated diversity of polypores (D) is about 3. 3.5. Modelling conservation strategies Increased extraction on fuelwood will lead to increased marginal time consumption per load as availability decreases. Biodiversity A total of 50 species of wood-inhabiting polypores were identified within the 57 plots. The average number of polypore species observed per plot was 2.8 (S.D. = 2.2) and in six plots none were found. The most frequent species were Heterobasidion insulare, Postia undosa, Trichaptum abietinum, and Gloeophyllum cf. abietinum. Less frequent species included Laetiporus sulphureus sl., Polyporus badius, Trametes versicolor, and Bondarzewia montana, which are used by local people for food and medicinal purpose. The number of polypore species per plot is positively correlated with the amount of dead wood. When estimating the impact of fuelwood collection on polypore diversity (PFOR ) we used a regression of polypore diversity index (D, species/ 500 m2 ) on fuelwood abundance (M, kg/ha): ln D = −0.2696 + 0.1760 ln M. The number of valid observations for this regression (M > 0, D > 0) was 36. Two outliers were excluded so the Fig. 6 – Results of simulations for different extraction scenarios, ranging from 0.1 to 5 times the present extraction level. Top: Mean and marginal time consumption per load. Bottom: Impact on biodiversity. Right-hand axes show impact and marginal time per load in per cent of present mean time consumption and present impact, respectively. e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 Fig. 7 – Results of simulations when areas with polypore diversity D ≥ 4 and D ≥ 3 are protected. The unprotected case is included for comparison. Top: Marginal time consumption per load vs. extraction. Bottom left: Impact on biodiversity vs. extraction. Bottom right: Impact vs. marginal time consumption. Right-hand axes show impact and marginal time per load in per cent of present mean time consumption and present impact, respectively. The protection strategy threshold is indicated by an asterisk. Dotted lines connect marginal time consumption, impact on polypore diversity and annual extraction at the threshold. 529 530 e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 We analysed the effect of extraction of up to 5 times the current extraction level and found that the marginal time consumption rises from 3.7 h for the last load at the present extraction level to 6.3 h at 5 times the current extraction (Fig. 6). To make this time consumption attractive at the present wage rate of 25 Rupees per hour the fuelwood price should rise to 158 Rupees. In the present-day scenario high transportation costs prevent fuelwood from reaching markets outside the study area. However, the ongoing road construction will reduce the price of transportation dramatically and the study area will become connected to suburban and urban markets at lower altitudes where the fuelwood price is about 140–180 Rupees per load. The relative effect of increased extraction on biodiversity (PFOR ) is greater than the relative effect on time consumption (Fig. 6, right-hand axes), and in the scenario with 5 times the current extraction the impact on diversity is 5.5 times the present impact, whereas the marginal time consumption per load is only 2.3 times the present average time consumption. This indicates that market mechanisms will not be able to conserve biodiversity if improved road connection to markets and reduced transport costs lead to increased collection. Quotas could be a possible way to regulate the extraction and protect biodiversity. An alternative strategy is to restrict the areas where fuelwood can be collected, prioritising areas with the highest D values. We simulated two scenarios of area protection, D ≥ 4 (corresponding to 120 ha or 10% of the forest area) and D ≥ 3 (corresponding to 349 ha or 30% of the forest area). The unrestricted scenario was used as baseline for this analysis. The area restriction implies that the marginal time consumption per load increases considerably (Fig. 7). Assuming that the market price of fuelwood determines the maximum time that can be spent on collecting a load, it emerges that for a marginal time per load of less than 5.5 h (twice the present average time per load) the area restriction has no effect on the impact measure PFOR . However, if the market price is high enough to allow collectors to spend more than 5.5 h per load, the area protection will be an efficient strategy. A time consumption of 5.5 h per load, corresponding to a price of about 140 Rupees, can be seen as a threshold between alternative biodiversity protection strategies. Below this value a quota system would be the only effective strategy, whereas above 5.5 h area protection seems most attractive due to the lower cost of implementation. Another important characteristic of the area protection strategy is that in the long run it can be expected that D values in protected areas will increase, gradually leading to a decreasing impact on polypore diversity in the area (PFOR ). 4. Discussion 4.1. Implication for protection of biodiversity Our model shows the complex interactions between extraction of natural resources and conservation of biodiversity. Today fuelwood extraction in Lete and Kunjo is exclusively determined by the local demand as transport is prohibitively costly. With increased road access connection will be made to markets in fuelwood-scarce areas in the south and north. We conclude that given dramatic changes in access to markets, leading to increased incentives for fuelwood collection, some sort of regulation will be necessary to protect the biodiversity related to this economically important natural resource. With regard to the fuelwood in Lete and Kunjo the consequences for biodiversity are exacerbated because the wood resource is found in a relatively small and accessible area, possibly making almost total extraction economically attractive. Establishment of protected areas is a common strategy for protecting biodiversity. Our simulation results show, however, that in the short-term area-based protection may potentially have little effect on the impact on biodiversity. This is because it induces increased fuelwood collection in areas with low densities of dead wood where small changes in dead-wood amount will have negative impacts on biodiversity. However, most likely the increasing marginal cost of collection caused by the area restriction will make fuelwood collection less attractive, thereby reducing the impact on biodiversity. The area protection only showed to be an efficient short-term tool for protection in case that the market price rises to a level higher than about 140 Rupees per load. The actual effect of fuelwood collection on polypore diversity is obviously more complex than our simple regression between amount of dead wood and species richness. The study area currently includes a few remote and inaccessible areas that could serve as potential refuges for polypores. Should dead-wood availability increase after a period of heavy extraction polypores would be able to recover relatively quickly due to their good dispersal abilities, at least over relatively short distances. Therefore, protection of small areas with high D-values may prove effective in the long term. Maintaining the current level of biodiversity in a scenario of increased extraction of fuelwood will only be possible with a combination of restricted areas and quotas. Although area restriction was observed to have little effect on overall human impact on biodiversity (PFOR ) in the short term, area protection may result in stable or improved conditions for polypores in the long term as the amount of dead wood increases in the restricted areas. A combination of area restriction and quotas has been suggested by Stefansson and Rosenberg (2005), but may be difficult to implement because of the cost of monitoring extraction. 4.2. Modelling and conservation strategies Agent-based modelling as presented in this paper can be an important tool in understanding the processes influencing the interaction between use and sustainability in areas with intensive extraction of natural resources. Lack of information on dead-wood accumulation is a main problem in our model and without time series it is not possible to address the sustainability of the extraction scenarios. Decay rates were not directly investigated in the present study but explored during fieldwork and discussed with local assistants. The rates differ between tree species. Tsuga, Pinus and Abies decay relatively fast, whereas Cupressus and Taxus decay more slowly. Apart from the variation between species, the size of the dead wood and the microclimatic conditions are also very e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 important for the speed of decay. To address these topics it is necessary to collect information over longer time where repeated measurement of dead wood would allow estimation of the average speed of decay. In this study we applied a simple model describing expected time consumption for fuelwood collection but most likely the real figures would show considerable variation. Dedicated fuelwood collectors can do the collection faster than inexperienced ones and a strong man can collect more wood faster than women and children. Fuelwood collection combined with other activities is also relatively important and can reduce the perceived time (cost) of collecting the fuelwood. We do not know exactly how much fuelwood is gathered while also doing other activities but during our fieldwork we observed collection of fuelwood combined with herding and mushroom collection. Given the large amount of fuelwood collected in the area, however, it is obvious that a considerable part of the fuelwood collection takes place as a primary activity. Another important limitation of the present study is the simplified assumptions regarding household behaviour and the nature of the fuelwood market. It is likely that faced with increased marginal time consumption for collection of fuelwood households will gradually reduce their consumption of fuelwood and partly switch to alternative fuel types, such as kerosene, LP-gas or electricity. 4.3. The role of community-based forest management The model has contributed towards an understanding of the interaction between fuelwood use and biodiversity that can guide future conservation efforts by local communities. Many local forest users are positive towards the protection of forest biodiversity and have in-depth knowledge about the conditions of their forest (Mehta and Heinen, 2001). Identification of valuable areas for protection of dead wood could be based on local knowledge, as could monitoring of the conditions using a participatory approach (Widmann, 2003; Subedi, 2006). Monitoring of polypores is relatively simple and could be supported by a short introductory training, booklets and posters. In support of conservation of biodiversity associated with dead wood designation of areas for protection of dead wood could be made a mandatory component of management plans. A combination of locally designated protected areas and additional fuelwood quotas could lead to a reduction of the human impact on biodiversity and would imply local ownership and promote awareness of the importance of biodiversity conservation. 5. Conclusion The present study developed a model describing fuelwood collection behaviour, dead-wood distribution and biodiversity patterns in a Nepalese mountain forest. The interaction between the natural resource and the human society is complex by nature and determined by its spatial heterogeneity and the economic behaviour of users. Conservation strategies for protection of biodiversity associated with dead wood must take this complexity into account. In the present study 531 area, a combination of area restrictions and quotas was shown most promising in terms of biodiversity conservation, while exclusively focusing on area restrictions did not lead to positive effects in most cases. The future market for fuelwood is unpredictable and will most likely be influenced by infrastructural changes. In case of major long-term changes in the local prices of fuelwood an area-protection strategy seems robust. Acknowledgements We would like to acknowledge Sanjeeb Bhattarai, Shiva Devkota, Giri Joshi, Arun Rijal, Somesh Das, and Basanta Pant for field work assistance, the people of Kunjo and Lete for providing information, ACAP for granting permission to work in the area, Jacob Heilmann-Clausen and Tumatso Hattori for assistance in polypore species identification, Helle O. Larsen for valuable comments during the writing process, DANIDA and ComForM for funding, and two anonymous reviewers for useful comments. references Amacher, G.S., Hyde, W.F., Kanel, K.R., 1999. Nepali Fuelwood production and consumption: regional and household distinctions. Substitution and successful intervention. J. Dev. Stud. 35, 138–163. Arnold, J.E.M., Kohlin, G., Persson, R., 2006. Woodfuels, livelihoods, and policy interventions: changing perspectives. World Dev. 34, 596–611. Baral, N., Stern, M.J., Heinen, J.T., 2007. Integrated conservation and development project life cycles in the Annapurna Conservation Area. Nepal: Is development overpowering conservation? Biodivers. Conserv. 16, 2903–2917. Benjaminsen, T.A., 1993. Fuelwood and desertification: Sahel orthodoxies discussed on the basis of field data from the Gourma region in Mali. Geoforum 24, 397–409. Bousquet, F., Le Page, C., 2004. Multi-agent simulations and ecosystem management: a review. Ecol. Model. 176, 313–332. Christensen, M., Heilmann-Clausen, J., Walleyn, R., Adamcik, S., 2004. Wood-inhabiting fungi as indicators of nature value in European beech forests. In: EFI Proceedings, vol. 51, pp. 229–238. Christensen, M., Hahn, K., Mountford, E.P., Ódor, P., Standovar, T., Rozenberger, D., Diaci, J., Wijdeven, S., Meyer, P., Winter, S., Vrska, T., 2005. Dead wood in European beech (Fagus sylvatica) forest reserves. For. Ecol. Manage. 210, 267–282. Cooke, P., Köhlin, G., Hyde, W.F., 2008. Fuelwood, forests and community management—evidence from household studies. Environ. Dev. Econ. 13, 103–135. Fraver, S., Wagner, R.G., Day, M., 2002. Dynamics of coarse woody debris following gap harvesting in the Acadian forest of central Maine. USA Can. J. For. Res. 32, 2094–2105. Gautam, A.P., Shivakoti, G.P., Webb, E.L., 2004. Forest cover change, physiography, local economy, and institutions in a mountain watershed in Nepal. Environ. Manage. 33, 28–61. Geist, H.J., Lambin, E.F., 2002. Proximate causes and underlying driving forces of tropical deforestation. BioScience 52, 143–150. Huston, M.A., 1996. Models and management implications of coarse woody debris impacts on biodiversity. In: McMinn, J.W., Crossley, D.A. (Eds.), Workshop on Coarse Woody Debris in 532 e c o l o g i c a l m o d e l l i n g 2 2 0 ( 2 0 0 9 ) 522–532 Southern Forests: Effects on Biodiversity. USDA Forest Service, Athens, GA, pp. 139–143. Junninen, K., Kouki, J., 2006. Are woodland key habitats in Finland hotspots for polypores (Basidiomycota)? Scand. J. For. Res. 21, 32–40. Jonsson, M., Ranius, T., Ekvall, H., Bostedt, G., Dahlberg, A., Ehnstrom, B., Nordén, B., Stokland, J.N., 2006. Cost-effectiveness of silvicultural measures to increase substrate availability for red-listed wood-living organisms in Norway spruce forests. Biol. Conserv. 127, 443–462. Kruys, N., Fries, C., Jonsson, B.G., Lämäs, T., Stähl, G., 1999. Wood inhabiting cryptogams on dead Norway spruce (Picea abies) trees in managed Swedish boreal forests. Can. J. For. Res. 29, 178–186. Linderman, M.A., An, L., Bearer, S., He, G.M., Ouyang, Z.Y., Liu, J.G., 2005. Modeling the spatio-temporal dynamics and interactions of households, landscapes, and giant panda habitat. Ecol. Model. 183, 47–65. Lonsdale, D., Pautasso, M., Holdenrieder, O., 2007. Wood-decaying fungi in the forest: conservation needs and management options. Eur. J. For. Res. 127, 1–22. Matthews, R.B., Gilbert, N., Roach, A., Polhill, J.G., Gotts, N.M., 2007. Agent-based land use models: a review of applications. Landscape Ecol. 22, 1447–1459. Mehta, J., Heinen, J., 2001. Does community-based conservation shape favorable attitudes among locals? An empirical study from Nepal. Environ. Manage. 28, 165–177. Meilby, H., Puri, L., Rayamajhi, S., Christensen, M., 2006. Planning a system of permanent sample plots for integrated long-term studies on community forest management. Banko Janakari 16, 3–11. Metz, J.J., 1994. Forest product use at an upper elevation village in Nepal. Environ. Manage. 18, 371–390. Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-being: Biodiversity Synthesis. World Resources Institute, Washington, DC. Müller, W.G., 2001. Collecting Spatial Data. Optimum Design of Experiments for Random Fields. Physica-Verlag, Heidelberg, 196 pp. Norstedt, G., Bader, P., Ericson, L., 2001. Polypores as indicators of conservation value in Corsican pine forests. Biol. Conserv. 99, 347–354. Núñez, M., Ryvarden, L., 2000. East Asian polypores, vol. 1. Fungiflora, Norway. Núñez, M., Ryvarden, L., 2001. East Asian polypores vol. 2. Fungiflora, Norway. Ódor, P., Heilmann-Clausen, J., Christensen, M., Aude, E., van Dort, K.W., Piltaver, A., Siller, I., Veerkamp, M.T., Walleyn, R., Standovár, T., van Hees, A.M., Kosec, J., Matocec, N., Kraigher, H., Grebenc, T., 2006. Diversity of dead wood inhabiting fungal and bryophyte assemblages in semi-natural beech forests in Europe. Biol. Conserv. 131, 58–71. Siitonen, J., 2001. Forest management, coarse woody debris and saproxylic organisms. Fennoscandian boreal forests as an example. Ecol. Bull. 49, 11–41. Similä, M., Kouki, J., Mönkkönen, M., Sippola, A.-L., Huhta, E., 2006. Co-variation and indicators of species diversity: can richness of forest-dwelling species be predicted in boreal forests? Ecol. Indicat. 6, 686–700. Stefansson, G., Rosenberg, A.A., 2005. Combining control measures for more effective management of fisheries under uncertainty: quotas, effort limitation and protected areas. Philos. Trans. R. Soc. B 360, 133–146. Stevens, V., 1997. The ecological role of coarse woody debris: an overview of the ecological importance of CWD in B.C. forests. Research Branch, B.C. Ministry of Forests, Victoria. Working Paper 30/1997. Stokland, J.N., 2001. The coarse woody debris profile: an archive of recent forest history and an important biodiversity indicator. Ecol. Bull. 49, 71–83. Stræde, S., Treue, T., 2006. Beyond buffer zone protection: a comparative study of park and buffer zone products’ importance to villagers living inside Royal Chitwan National Park and to villagers living in its buffer zone. J. Environ. Manage. 78, 251–267. Subedi, B.P., 2006. Linking plant-based enterprises and local communities to biodiversity conservation in Nepal Himalaya. Adroit., 244 pp. Widmann, P., 2003. Development of participatory biodiversity monitoring concept and methodology. Churia Forest Development Project PN 2001.2173.1. Wilkie, D., Shaw, E., Rotberg, F., Morelli, G., Auzel, P., 2000. Roads, development, and conservation in the Congo Basin. Biol. Conserv. 14, 1614–1622.