Introduction

Wild forest berries, including bilberry (Vaccinium myrtillus) and cowberry (Vaccinium vitis-idaea), are among the most important non-wood forest products (NWFPs) in the Nordic countries (Turtiainen and Nuutinen 2012; Vaara et al. 2013). Bilberries and cowberries are picked for both household use and sale. Wild berries’ value as nutritional supplement is recognized, and their good reputation in terms of high content of bioactive compounds is used to market berry-based health products. Due to the increased demand for northern wild berries, especially in Asian countries (MARSI 2014), forest owners may be interested in knowing how much they could increase their income from the forest if they manage forests for berry production. Such knowledge on the optimal management for maximizing the profitability of joint production of timber and berries is, however, limited.

Earlier results suggest that valuing NWFPs may remarkably change the optimal stand management and increase the profitability of forestry (Palahí et al. 2009; de Miguel et al. 2014).When mushroom production was considered in Pinus sylvestris and P. nigra stands in Catalonia, thinning treatments, which usually are unprofitable, were included in the optimal management schedules to reduce stand density to a level that is favorable for mushrooms (Palahí et al. 2009). De Miguel et al. (2014) optimized the joint production of timber and pine honeydew honey in Pinus brutia stands in Turkey and found that especially on medium- and poor-quality sites, longer rotations considerably increase the incomes from the multi-product pine stands.

So far, stand management considering both timber and berries has been optimized only for bilberry (Miina et al. 2010). However, cowberry is economically more significant in Finland (MARSI 2014). Therefore, the joint production of timber and cowberry should also be studied. Miina et al. (2010) did not consider the costs of berry picking, and thus, the existing results for bilberry can be utilized mainly for developing management prescriptions for public forests, which are used for recreational berry picking. Including berries as a source of personal income to the forest landowner suits particularly to situations where the owner can control the utilization of berry yields. Due to the everyman’s right, this is seldom the case in Finland. Optimizing the joint production of timber and berries is the most meaningful in public forests where the aim should be to maximize the total benefit that the society gets from the forest. However, results on the joint production would help policy makers to develop instruments that promote forest management which is optimal from the society’s point of view.

Bilberry is an abundant plant in conifer-dominated forests of medium fertility, and cowberry on nutrient-poor mineral soil sites (e.g., Raatikainen 1978; Raatikainen and Raatikainen 1983; Raatikainen et al. 1984). However, annual berry crops fluctuate greatly, mainly due to weather conditions (Kardell and Eriksson 1990, 2011; Wallenius 1999; Selås 2000). The abundance and yield of bilberry and cowberry are mainly influenced by site conditions, and silvicultural operations affect the yields. Bilberry suffers from regeneration fellings and does not thrive in sparse seedling and sapling stands or in dense young stands. A moderate supply of light is needed for good bilberry crops, the optimal canopy cover being 10–50 % (Raatikainen et al. 1984). Due to different shading of tree species, the coverage and yield of bilberry is lower in spruce-dominated than in pine-dominated stands (Raatikainen et al. 1984; Miina et al. 2009). To increase bilberry yields, higher thinning intensities, more frequent thinnings, retention of more pine in mixed stands and longer rotation lengths could be applied in even-aged management system (Miina et al. 2010).

Cowberry needs more light than bilberry for maximum productivity. Although cowberry tolerates some shade and is adapted to grow under tree canopies, a good supply of light is necessary for abundant flowering and berry crops; the optimal crown cover is 1–40 % (Raatikainen et al. 1984). Cowberry also suffers from regeneration fellings combined with soil preparation, but the best yields are obtained a few years after regeneration cutting and again at the end of rotation, especially in sparse mature pine forests. Cowberry yields are low in dense and shaded stands.

Despite seemingly optimal site and stand characteristics for berry crops, not all forest stands supply good berry yields. According to Raatikainen (1978, see Table 1), the mean cowberry yield of mature stands on Vaccinium type sub-xeric heath site (Cajander 1949) was only 40 kg/ha, whereas the maximum cowberry yield measured was as high as 330 kg/ha. Bilberry can also produce good crops on Vaccinium site (Raatikainen 1978). In the same area in central Finland, the mean and maximum bilberry yields of mature stands growing on Myrtillus-type mesic heath site were 22 and 140 kg/ha, respectively (Raatikainen and Raatikainen 1983, see Table 1). On Myrtillus site, cowberry rarely produces harvestable crops (Raatikainen and Raatikainen 1983; Raatikainen et al. 1984). Due to high between-stand variation in berry yields, it may be profitable to modify the stand management for promoting berry yields mainly in those stands which are known to have high productivity. However, it is not known how stands producing both bilberries and cowberries should be managed in the optimal way.

The aim of this study was to optimize the even-aged management of stands in the joint production of timber and wild forest berries; bilberry (Vaccinium myrtillus L.) and cowberry (V. vitis-idaea L.). Multi-product management schedules were optimized for two Scots pine (Pinus sylvestris L.) stands and a mixed stand of Scots pine, Norway spruce (Picea abies (L.) Karst.) and birch (Betula pendula Roth) growing on sites most suitable for bilberry and/or cowberry in North Karelia, Finland. The management was optimized separately for an average and good berry stand. The market prices paid to berry pickers and yield-dependent berry picking costs were used in calculations. We analyzed also the effect of berry prices on the optimal stand management. The results are the most applicable to situations where the forest owner directly or indirectly benefits from managing the stands for both timber and berry yields.

Materials and methods

Overview

A simulation software was used in a deterministic way to predict the growth of trees in three initial stands. Due to a large annual variation in berry yields, stochastic simulations were used to predict the berry production during stand development. The incomes and costs of both timber production and berry picking were used to calculate the net present values of all future costs of and incomes from the multi-product management. The stand simulator was linked with an optimization algorithm to find out the management schedule for the stand (i.e., thinnings and rotation), which maximizes the profitability of the joint production of timber and berries (e.g., Palahí et al. 2009; Miina et al. 2010; de Miguel et al. 2014). Management was optimized separately for average and good berry stands.

Simulation of stand development and berry yields

The stand development was predicted using the models of Hynynen et al. (2002) for site index, dominant height development, diameter increment, crown height, height increment and tree survival. The taper models of Laasasenaho (1982) were used to calculate assortment volumes of removed trees. In all simulations, the temperature sum was set to 1200° days, the altitude to 100 m and the proportion of sea and lakes within 20 km to 0 and 20 %, respectively. The values represent the conditions of North Karelia, Finland.

The empirical berry yield models for bilberry (Miina et al. 2009) and cowberry (Turtiainen et al. 2013) were used to predict the berry production. The used model set first predicted the coverage of the berry species as a function of site conditions, e.g., site fertility, regeneration method, stand age and stand basal area. Then, it predicted the annual berry yields from the coverage of the berry species and some other predictors.

The bilberry yield was predicted using the following models (Miina et al. 2009):

$$\begin{aligned} Bilberry\;cover & = 100 \times \left( {1 + \exp \left\{ { - f(\cdot)} \right\}} \right)^{ - 1} \\ f(\cdot) & = - 3.8470 - 2.1815 \times OMaT - 0.4809 \times OMT - 0.4807 \times VT - 1.5053 \times CT + 0.1209 \times Pine \\& \quad - 0.4770 \times Birch \times OMT - 0.2588 \times Art\text{Re} gen - 1.4715 \times FormerAgrLand + 0.0029 \times Alt \\& \quad + \,0.0080 \times T - 0.0021 \times T^{2} /100 + 0.0947 \times G - 0.1916 \times G^{2} /100 \\ \end{aligned}$$
(1)
$$BilberriesP = \exp \{ - 0.5359 + 0.2398 \times Bilberry\;cover - 0.2812 \times Bilberry\;cover^{2} /100 + v + u\}$$
(2)
$$\begin{aligned} BilberriesS & = \exp \{ - 4.2024 + 0.3635 \times Bilberry \, cover - 0.4798 \times Bilberry \, cover^{2} /100 \\& \quad + 0.3742 \times G - 1.3447 \times G^{2} /100 + v + u\} \\ \end{aligned}$$
(3)

where Bilberry cover is the mean coverage of bilberry in the stand (%), BilberriesP and BilberriesS are the mean number of bilberries per m2 in pine- and spruce-dominated stands, respectively; T is stand age (years); G is stand basal area (m2/ha); Pine and Birch are indicator variables for pine and birch, respectively (reference: spruce); and Alt is altitude (m). The effect of the fertility of mineral soil site is described by indicator variables as follows (Cajander 1949): OMaT = Oxalis–Maianthemum type, OMT = Oxalis–Myrtillus type, VT = Vaccinium type and CT = Calluna type (reference: MT = Myrtillus type). ArtRegen and FormerAgrLand are indicator variables for artificial regeneration methods (reference: natural regeneration) and former agricultural land (reference: former forest), respectively, and v and u are random, normally distributed between-stand and between-year effects with zero means and constant variances.

The cowberry yield was predicted using the following models (Turtiainen et al. 2013):

$$Cowberry \, cover = 100 \times \left( {1 + \exp \left\{ { - f(\cdot)} \right\}} \right)^{ - 1}$$
(4)
$$\begin{aligned} f({\cdot}) & = - 4.7902 - 5.173 \times OMaT - 2.569 \times OMT - 0.4216 \times MT - 0.4185 \times VT \\& \quad - 2.0679 \times SpruceMireI - II - 0.7984 \times SpruceMireIII - 1.8198 \times PineMireI - III \\& \quad - 0.5644 \times PineMireIV - 1.7620 \times PineMireV - 0.9438 \times FormerAgrLand \\& \quad - 0.4327 \times Spuce \times (OMaT{\text{ or }}OMT{\text{ or }}MT{\text{ or }}SpruceMireI - III) \\& \quad - 0.7528 \times Birch \times (OMaT{\text{ or }}OMT{\text{ or }}MT{\text{ or }}SpruceMireI - III) + 2559.2/TS \\& \quad - 0.0039 \times Alt + 0.0106 \times T + 0.0157 \times G \\ \end{aligned}$$
$$\begin{aligned} Cowberries & = \exp \{ 6.7253 + 0.0966 \times Cowberry \, cover - 0.0837 \times Cowberry \, cover^{2} /100\\& \quad - 0.4716 \times \ln (G + 1) + 0.0071 \times Alt - 4626.4/TS + v + u\} \\ \end{aligned}$$
(5)

where Cowberry cover is the mean coverage of cowberry in the stand (%); Cowberries is the mean number of cowberries per m2; Spruce is indicator variable for spruce (reference: pine); TS is temperature sum (dd); and the peatland site for spruce (SpruceMire) and pine (PineMire) mire is defined as follows (Laine and Vasander 1993): I = eutrophic, II = herb rich (mesotrophic), III = Vaccinium myrtillus and tall sedge (meso-oligotrophic), IV = Vaccinium vitis-idaea and small sedge (oligotrophic), V = cottongrass and dwarf shrub (poor ombro-oligotrophic bogs). Other variables are defined as above.

The effects of the study years were originally included in Eqs. (2), (3), and (5) as indicator variables. We calculated the sample means of the year effects (0.1422, 0.5450 and 0.1849, respectively) and added them to the intercepts of the models. Berry yields measured in North Karelia, Finland (Kilpeläinen et al. 2016) and the berry yields reported by Turtiainen et al. (2005, Table 4) indicated the need to calibrate the berry yield models. It was found that the yield model of Miina et al. (2009) clearly under-predicted the bilberry yields, whereas the yield model of Turtiainen et al. (2013) over-predicted the cowberry yields, especially in the beginning of the rotation. Therefore, the bilberry yield predictions were multiplied by 2.3 on Myrtillus site and by 3.5 on Vaccinium site, whereas the cowberry yield predictions on Vaccinium site were multiplied by 0.8.

In addition, the positive effect of regeneration felling on berry production was overestimated since it takes several years before bilberry and cowberry are fully recovered from cutting (Atlegrim and Sjöberg 1996; Kardell and Eriksson 2011). Because this is not taken into account by the yield models used in this study, the bilberry and cowberry yield was assumed to be 0 and 20 %, respectively, of the yield prediction immediately after regeneration felling (i.e., at stand age of 1 year). Then, the yield reduction was assumed to decrease linearly until it was over in 20 years after regeneration felling for bilberry and 10 years for cowberry. The reduction affected mostly cowberry yield predictions since predicted bilberry yields are very low in young stands.

In a mixed stand, the number of bilberries was predicted first for each tree species using the total stand basal area as a predictor in the models. For birch, the number of bilberries was calculated using the model for pine-dominated stands (Eq. 2). Finally, the number of bilberries was calculated as the weighted average of the species-specific predictions. The proportions of tree species of stand basal area were used as weights.

The number of berries was converted into the berry yield (kg/ha) by multiplying it by the mean fresh weight of one berry; 0.35 g was used for bilberry (Eronen 2004) and 0.23 g for cowberry (Ihalainen et al. 2003). Finally, the berry yield predictions were multiplied by 0.75 assuming that 75 % of the total yield of the season would actually be harvested for sale (cf. Raatikainen and Niemelä 1983).

The annual berry yield was predicted in a stochastic way. For each year, the bilberry and cowberry yields were predicted 200 times by drawing the random between-year effects (u) from the normal distribution (Eqs. 2, 3 and 5), and the annual berry yields were computed as the mean of the 200 outcomes. The variances of the between-year effects (u) of Eqs. (2), (3) and (5) were 0.0712, 0.2004 and 0.0888, respectively. The stand was assigned as a good berry stand by setting the random stand effect (v) to 1.4 for bilberry (Eqs. 2 and 3) and 0.7 for cowberry (Eq. 5). The values were such that the yield predictions for good berry stands were comparable to the good yields measured by Raatikainen and Raatikainen (1983) for bilberry and by Raatikainen (1978) for cowberry. For an average berry stand, the random stand effects were zero.

Initial stands

Three 10-year-old initial stands representing North Karelia, Finland, were used as starting points for the simulations (Table 1). A pine stand and a pine–spruce–birch mixture represent Myrtillus site, and a pine stand represents Vaccinium site. Pure birch and spruce stands were not included because berry yield models for birch stands are missing, and in spruce stands, berry yields are mostly low and incomes from berries never play a major role in stand management (Miina et al. 2010).

Table 1 Main characteristics of the 10-year-old stands used as initial stands in the optimizations
Full size table

The initial stands were assumed to have been treated with a pre-commercial thinning. The initial growing stock was specified with 2 or 3 cohorts per species, and each cohort was represented by ten sample trees. The sample trees were generated by first predicting the diameter distribution of each cohort. The distributions were divided into ten diameter classes of equal width, and the midpoint tree of each class was taken as a sample tree. The frequencies of the sample trees were calibrated in the same way as in Pukkala and Miina (2005). Tree species, mean age, mean diameter and mean height in each cohort were given for each initial stand (Table 1).

Optimization method

The stand simulator was linked with an optimization program to study the effect of berry production on the optimal management of stands. The nonlinear programming algorithm of Hooke and Jeeves (1961) was used to find the optimal combination of decision variables defining the stand management. Since the Hooke and Jeeves’ direct search method may converge to local optima, each optimization problem was solved 20 times with different randomly generated initial values of decision variables, and the solution with the highest objective function value of these results was reported. Each direct search run started from the best of 200 random combinations of decision variables.

The objective function was the soil expectation value (SEV) calculated with 3 % discount rate (Price 1989). SEV was defined as the net present value of all future costs of and incomes from both timber and berry production. All costs and incomes of a full rotation were discounted to the beginning of the rotation. Future rotations were taken into account by assuming that the rotation will be followed by similar rotations to infinity.

The decision variables defining stand management were as follows: number of years since regeneration (first thinning) or first thinning (second thinning), removal percentage separately for each tree cohort (for both thinnings), and number of years between the last thinning and final cut. The thinning percentage was the same for every diameter class of a cohort. However, since the thinning percentages of tree cohorts were different decision variables, it was possible to optimize the type of thinning, in addition to thinning intensity. The number of thinnings (i.e., 1 or 2) was given to the simulation–optimization system before solving the problem.

The average unit costs of silvicultural treatments in 2013 were applied in the calculations (http://www.metla.fi/metinfo). In the beginning of the rotation, a regeneration cost of 1000 €/ha was assumed to accrue from planting on Myrtillus site, and 500 €/ha from natural regeneration on Vaccinium site. The pre-commercial thinning of young stand (350 €/ha) was assumed to have been done three years after stand height reached 1.3 m.

Roadside prices of timber assortments were used (Table 2). The mean roadside prices of 2013 in North Karelia were used in thinnings and final felling (http://www.metla.fi/metinfo). The entry cost of a cutting was taken as 100 €/ha. The variable costs of cuttings were calculated using the models of Rummukainen et al. (1995).

Table 2 Roadside prices and minimum top diameters of timber assortments
Full size table

Based on the statistics on the amounts of wild berries offered for sale in the last five years (MARSI 2014, Table 11), the average market price paid to pickers in eastern Finland was 2.4 €/kg for bilberries and 1.8 €/kg for cowberries. These prices were used in the calculations. In addition, zero as well as twofold and threefold pricing were used to analyze the effect of berry prices on the optimal stand management.

The berry picking costs included travel and harvesting costs. For travel costs, it was assumed that two berry pickers drive together a total distance of 20 km to pick and sell berries (Raatikainen 1978; Kangas and Markkanen 2001). Applying a travel cost of 0.4 €/km and a daily harvest of 40 kg per person, the travel cost per collected kilogram is 0.1 €. Using the wage of 6 € per hour and assuming that 12 kg of berries can be picked per hour, the harvesting cost is 0.5 €/kg. However, it was assumed that the harvesting cost of 0.5 €/kg (corresponding to the productivity of 12 kg/h) holds only when the berry yield is 40 kg/ha. The harvesting cost was multiplied by 41/(yield + 1), resulting in increased cost (i.e., decreased berry picking productivity) when berry yield <40 kg/ha and decreased cost in case of higher berry yields. The total cost of berry picking was 0.6 €/kg when the berry yield was 40 kg/ha. It was assumed that berries were collected only when the berry picking costs were less than the selling price of berries. As a result, years of poor berry yields were not included in the incomes and harvested berry yields.

Results

Optimal management of stands with varying berry prices

The effect of bilberry price (0, 2.4, 4.8 or 7.2 €/kg) on the optimal management of the pine and mixed stands on Myrtillus site was studied by maximizing the soil expectation value with 3 % discount rate. It was assumed that cowberries are not harvested on Myrtillus site. Without valuing bilberries, the optimal management schedule of the pine stand consisted of two thinnings removing 27 and 78 % of the basal area and a final felling at the age of 71 years (Table 3; Fig. 1). The mixed stand was thinned twice (removals 34 and 55 %) favoring spruce in the residual stand and clear-felled at the age of 73 years (Table 3; Fig. 2).

Table 3 Effect of bilberry price on the optimal stand management of stands with average berry yields
Full size table
Fig. 1
figure 1

Effect of bilberry price on the optimal management schedules of the pine stand on Myrtillus site with average and good bilberry yields. Cowberries are not harvested

Full size image
Fig. 2
figure 2

Effect of bilberry price on the optimal management schedules of the mixed stand on Myrtillus site with average and good bilberry yields. Cowberries are not harvested

Full size image

Assuming that the stand provides average bilberry yields and using the current market price of bilberry (2.4 €/kg), the optimal management of the pine stand on Myrtillus site changed only slightly, and the optimal management of the mixed stand did not change at all. With bilberry prices of at least 4.8 €/kg, bilberry picking increased the rotation length of the pine stand by about 20 years. In the mixed stand, a threefold bilberry price was needed to extend the rotation. More spruces were removed in thinnings, resulting in higher proportion of pine (Fig. 3).

Fig. 3
figure 3

Proportion of pine of the total stand basal area in the optimal management schedule of the mixed stand on Myrtillus site with average and good bilberry yields and different bilberry prices. Cowberries are not harvested

Full size image

In stands growing on Myrtillus site and having high bilberry yields, commercial berry picking had a strong impact on optimal management (Figs. 1, 2, 3). In both pine and mixed stand, berry picking affected the optimal management already with the current market price (2.4 €/kg), leading to longer rotation and heavier thinnings. In the mixed stand, it was optimal to make two heavy thinnings in which the stand was converted into a pine-dominated stand (price 2.4 €/kg) or into pine–birch mixture (4.8 and 7.2 €/kg) toward the end of the rotation (Fig. 3). Compared to timber-oriented management, multi-product management with the current market price of bilberry decreased the timber benefit by 8 % in the pine stand and 32 % in the mixed stand (Table 4).

Table 4 Effect of bilberry price on the optimal stand management of stands with good berry yields
Full size table

The management of the pine stand on Vaccinium site was optimized by maximizing the SEV of the incomes from timber, cowberry and bilberry. When bilberries were not harvested, the optimal management schedule of the pine stand with average cowberry yields changed only slightly, but the proportion of cowberry incomes of the total SEV was as high as 42–74 % (Table 3, Fig. 4). With good cowberry yields, the pine stand was thinned heavier (Table 4; Fig. 4). With the threefold cowberry price, only one heavy thinning was done and the rotation length was more than doubled. When both cowberries and bilberries were harvested, the optimal management of the average berry stand changed only when two- or threefold berry prices were assumed (Fig. 5). In the good berry stand, management changed even with the current market prices (cowberry 1.8 €/kg and bilberry 2.4 €/kg). For example, thinnings were heavier and the rotation length was 34 years longer. This change in stand management decreased the SEV calculated without berry income by 9 %.

Fig. 4
figure 4

Effect of cowberry prices on the optimal management schedules of the pine stand on Vaccinium site with average and good cowberry yields

Full size image
Fig. 5
figure 5

Effect of berry prices (cowberry/bilberry) on the optimal management schedules of the pine stand on Vaccinium site with average and good berry yields

Full size image

Berry harvests in the optimal management schedules

The highest bilberry harvests were obtained in mature stands at the end of the rotation. In the pine stand with average bilberry yields, the maximum annual bilberry harvests reached 30 kg/ha (Fig. 6). In the pine stand with good bilberry yields, the maximum annual bilberry harvests increased from 110 kg/ha to more than 130 kg/ha along with the increasing bilberry prices. In the mixed stand with average bilberry yields, the bilberry harvests reached hardly 10 kg/ha. In the stand where berry yields were high, the optimal management for joint production resulted in bilberry harvests of about 140 kg/ha (Fig. 7). With the market bilberry price of 2.4 €/kg, the mean annual bilberry harvest of the pine stand during the rotation (67 kg/ha/year) was higher than in the mixed stand (50 kg/ha/year), but with twofold berry price, the mean annual harvest was higher in the mixed stand (Table 4).

Fig. 6
figure 6

Predicted bilberry harvest in the optimal management schedules of the pine stand on Myrtillus site with average and good bilberry yields and different bilberry prices. Cowberries are not harvested. Note Bilberry harvests in the optimal zero-pricing schedules were calculated using bilberry price 2.4 €/kg

Full size image
Fig. 7
figure 7

Predicted bilberry harvest in the optimal management schedules of the mixed stand on Myrtillus site with average and good bilberry yields and different bilberry prices. Cowberries are not harvested. Note Bilberry harvests in the optimal zero-pricing schedules were calculated using bilberry price 2.4 €/kg

Full size image

The highest cowberry harvests of the pine stand on Vaccinium site were obtained in the beginning of the rotation and only a heavy thinning was able to increase markedly the cowberry harvests later during the rotation (Fig. 8). At the early stages of stand development, it was not optimal to thin the pine stand to promote cowberry harvests. When bilberries were also picked from the pine stand with good berry yields and current market prices, thinnings were heavier and regeneration felling was postponed resulting in increased berry harvests (Table 4). Heavier thinnings decreased the stand basal area to a level which enabled increased cowberry and bilberry harvests toward the end of the rotation (Fig. 9).

Fig. 8
figure 8

Predicted cowberry harvest in the optimal management schedules of the pine stand on Vaccinium site with average and good cowberry yields and different cowberry prices. Bilberries are not harvested. Note Cowberry harvests in the optimal zero-pricing schedules were calculated using cowberry price 1.8 €/kg

Full size image
Fig. 9
figure 9

Predicted bilberry and cowberry harvests in the optimal management schedules of the pine stand on Vaccinium site with average berry yields and different bilberry and cowberry prices. Note: Berry harvests in the optimal zero-pricing schedules were calculated using cowberry price 1.8 €/kg and bilberry price 2.4 €/kg

Full size image

Discussion

The results of this study are in line with earlier studies showing that valuing NWFPs may remarkably change the optimal stand management. The provision of products and services is diversified, and simultaneously, the overall monetary benefits from forest management are increased (Palahí et al. 2009; Miina et al. 2010; de Miguel et al. 2014). Therefore, knowledge on the multi-product stand management is useful for forest owners who want to optimize the joint production of timber and berries. As far as we are aware, berry picking costs were included for the first time in stand-level optimization in order to better meet the practical conditions and thus enhance the comparison of timber and berry production. This is also the first study where both cowberry and bilberry were taken into account in the optimization of stand management.

With the current market prices and picking costs of berries, the timber-oriented stand management was not or only slightly changed in stands that provide average berry yields. In these stands, the contribution of berries to the SEV varied a lot (3–46 %) and was sometimes significant. In good berry stands, heavier thinnings are recommended to reduce canopy shading to a level that is favorable for berry crops. In addition, longer rotations are favorable in joint production since they enable harvesting the good berry yields of mature stands. More pines should be left in the thinning treatments of mixed stands. The optimal stand management changed, and berry harvests increased most significantly in the mixed stand with good bilberry yields. With the current berry prices and assuming good berry yields, the total SEV calculated with berry incomes was more than twice as high as the SEV from timber production only.

Stand characteristics and management (e.g., thinning) greatly affect the development of berry yields along the stand development and the profitability of joint production of timber and berries. Because of the strong influence, the yield predictions of berries need to be reliable. In addition to model predictions, several other parameters, however, affect the profitability of berry production: harvesting rate, the size of berries, berry price, harvesting costs and discount rate. In this study, berry prices and production levels were varied; other parameters were fixed. Earlier, Miina et al. (2010) studied the sensitivity of the optimal stand management to berry price, discount rate and harvesting rate. With increasing values of all these parameters, it was generally more profitable to manage the stands in a way that promotes bilberry production. Besides stochastic berry production, variation in berry and timber prices as well as tree growth—including the effects of climate change—could also be considered in a stochastic way.

Turtiainen (2015) evaluated the berry yield models used in this study by comparing them with other models prepared for bilberry and cowberry in Finland. In addition, Kilpeläinen et al. (2016) recently evaluated the berry yield models by using independent data measured from North Karelia, Finland. The best available berry yield models according to Kilpeläinen et al. (2016) were used in this study after calibrating them on the basis of recent field data. Bilberry yields predicted by the model of Miina et al. (2009) correlated positively and statistically significantly with the predictions of other models. The main difference was that the model of Miina et al. (2009) predicted the highest bilberry yields in pine-dominated stands, while some other models predict the highest yields in spruce-dominated stands. Correlations between cowberry yields predicted by the model of Turtiainen et al. (2013) and other existing models were positive but only weakly significant. The model of Turtiainen et al. (2013) predicted good cowberry yields in the beginning of the rotation, while other models produced good yields for both openings and mature pine stands. That happened also in this study when the pine stand was thinned heavily and the rotation was long (Figs. 8, 9).

The berry yield predictions of this study were comparable to the berry yields measured in Finnish forests. For example, on Myrtillus site the mean annual bilberry harvests of the pine stand with average bilberry yields (12–16 kg/ha) correspond to the mean bilberry yields of 22 kg/ha measured by Raatikainen and Raatikainen (1983). On Vaccinium site, we predicted mean annual cowberry harvests of 33–34 kg/ha for the pine stand with average cowberry yields, whereas Raatikainen (1978) reported the mean cowberry yield on mature stands on similar site to be 40 kg/ha. Based on the literature survey of Turtiainen et al. (2005, Table 4), the mean annual bilberry yield of Myrtillus site in North Karelia was 14 kg/ha, and that of cowberry on Vaccinium site was 26 kg/ha. In our simulations, the maximum annual berry harvests were about 140 kg/ha for bilberry and 150 kg/ha for cowberry. According to Raatikainen and Raatikainen (1983), the maximum annual yield for bilberry was 140 kg/ha, and Raatikainen (1978) measured 330 kg/ha for cowberry. Our results were comparable also to those earlier studies in which thinnings were proposed to enhance bilberry yields in Scots pine forests in Scotland (Parlane et al. 2006) and in Scots pine and Norway spruce forests in Sweden (Kardell and Eriksson 2011).

To conclude, the results of this study verified that additional economic benefits can be achieved through modified stand management when a large part of the incomes are generated through utilization of berry yields. By picking berries, forest owners may get also other direct and indirect benefits (e.g., recreational and potential health benefits). Therefore, using the market price to measure the value of berry picking may underestimate the total benefits from the joint production of timber and berries. If the commercial utilization of berries and the recognition of other benefits related to berry picking increase in the future, the practical demand for multi-product stand management guidelines will increase.