© 2019 г.

A.A. Dulina¹, S.I. Chumachenko²

¹Center for Forest Ecology and Productivity of the RAS

Profsoyuznaya st. 84/32 bldg. 14, Moscow, 117997, Russia

²MB of Bauman Moscow State Technical University

1^st Institutskaya street, 1,

141005, Mytischi, Moscow region, Russia

E-mail: anna_dulina@bk.ru

Received 25 November 2018

The article analyzes the most favourable conditions for the growth and fruit bearing of berries and mushrooms; studies yield indices of food resources; examines Russian and foreign yield capacity models of forest food resources. It is proposed to use illumination at the level of ground cover, which is an informative parameter, for modelling the dynamics of yield capacity of food resources. Illumination estimation will help us to more accurately assess the yield of food resources not only in single-storeyed stands with understorey trees and undergrowth, but also in multi-storeyed stands.

Key words: forest food resources, modelling, illumination, yield

Currently, the area of forest lands in the center of the European Russia that are not leased for forest management is decreasing. The possibility of income growth only through the sale of wood by increasing the area of woodland holdings is almost used up, so it’s necessary to find a way to get higher return on each forest plot. Multi-purpose forest management might be the optimal solution. Harvesting of wood and non-wood resources in the same forest area is not only possible, but also necessary. It is beyond dispute that harvesting of timber which is a crucial material for construction, the raw material for the pulp and paper industry and many other industries, plays an important role in the Russian economy. Yet besides wood, forests have huge resources of various products used for food and as raw material for industrial processing. Forest Code of the Russian Federation, 2006, divides these types of forest management into four groups: harvesting of galipot; harvesting and picking of non-wood forest resources including stumps, birch bark, bark of trees and shrubs, kindling, browsable forage, spruce, fir, pine or other coniferous trees boughs for the Christmas and New Year holidays, moss, forest litter, cane, reed etc.; harvesting of food forest resources including wild fruits, berries, nuts, mushrooms, seeds, birch sap etc. and the collection of medicinal plants; use of forests for hunting (Forest Code of the Russian Federation, 2006).

Past research and practice shows that the income from the use of non-wood resources in certain types of forests is several times higher than the income from timber harvesting (Telishevsky, 1986). For example, comparative cost estimates of wood and cloudberry in sphagnum pine forest show that the income from the sale of berries per 1 ha was 11.5 times higher than the income from logging (Kositsyn, 1996, 1998). In the Komi Republic, researchers compared the cost of 120-year-old pine wood and the annual yield of ceps in the same area. It was found that in just 5 years, the income from the sale of harvested mushrooms from such a plot will be comparable to the cost of wood harvested at the age of 120 years. If the plot is used to collect mushrooms throughout the period of forest growth, the profit from those mushrooms will be several times higher than the profit from logging (Kozubov, Taskayev, 2000). Therefore, the use of non-wood forest resources increases the economic potential of the forest industry and provides population with organic products (Bolshakov, 2013).

Interest in the use of non-wood resources recently revived because forest users want to get the maximum benefit from a woodland holding. Of all possible forest resources, besides timber, food resources (berries and mushrooms) are the most in-demand on the market. Previously, forest management methods around the world have traditionally focused on wood production; however, with the transition to multi-purpose forest management it was found that assessing and forecasting the dynamic pattern of non-wood resources is a very relevant issue (Vacik et al., 2016; Huber et al., 2016). Computational modeling is one of the modern approaches to solving this problem.

The objective of this paper is to analyze the current state of assessment and forecasting of the dynamic pattern of non-wood resources and the rationale for the choice of factors and indices for modelling forest food resources for the central part of Russia.

To achieve this objective, it is necessary to (1) assess the growing conditions of the most common berries and mushrooms in the European part of Russia, (2) assess the existing models and (3) assess the possibilities of using taxation data and propose other informative indicators to calculate the yielding power of mushrooms and berries.

ENVIRONMENTAL GROWING CONDITIONS OF THE MOST COMMON BERRIES AND MUSHROOMS

Among the non-wood food resources of Russia the most important ones in the context of economic meaning of food fruits are forest plants belonging to the Vacciniaceae family – cranberry (Oxycoccus paluslris Pers), cowberry (Vaccinium vitis-idaea L.), bilberry (Vaccinium myrtillus L.) and to the Rosaceae family – raspberry (Rubus idaeus L.) and cloudberry (Rubus arcticus). The most valuable mushrooms are cep (Boletus edulus Fr. ex Bull.), rough boletus (Boletus scaber Frex Bull.), orange-cap boletus (Boletus aurantiacus Fr.), and chanterelle (Cantharellus cibarius Fr.) (Egoshina, 2005).

Bilberry is a low-growing prosrate shrub with a wide distribution area: from the meadow-steppe zone to the marshes. It is rather common, growing from semi-open spaces to particularly shady forests. The prosrate shrub grows mainly on moist poor sandy soils, usually acidic, with a small amount of lime (Landolt, 1977; Tsyganov, 1983). Intensely dislikes direct sunlight. The optimum light regime is achieved with an average closure of canopy. Bilberry growing in pine or spruce forests is of industrial importance. After selection felling the fruiting of bilberry is increased by 2.5 times within 5 years, and then returns to the usual amount. After clear cuts, in the first 2–3 years degradation of the shrubs is observed, followed by their complete withering away that is only reversed 40–50 years later (Zworykina, 1972; Kurlovich et al., 2015).

Cowberry is an evergreen shrub with a wide area of distribution from the meadow-steppe to the swamp, forest and meadow zones. It is a light-requiring plant that grows especially well in fresh coniferous forests. This berry grows on poor acidic fine-sanded soils; cranberry prefers less moist places than bilberry (Landolt, 1977; Tsyganov, 1983). It fruits well in stands with low closure of the tree canopy and in open areas. Cowberry yield increases by 1.5 times in the first 5–7 years after the selection felling. Clear cuts result in abundant fruiting for 6–8 years followed by the depression of the berry field which restores 30–40 years later (Cherkasov et al., 1988; Kurlovich et al., 2015).

Cranberry is an evergreen prosrate shrub with an area of distribution from wet forest-meadow to marsh zones. Cranberry habitats are raised (oligotrophic) and transitional (mesotrophic) bogs. In the forest zone, it grows on moderately humid habitats on microhills, in the forest tundra it is widespread at swamps. This prosrate shrub likes acidic poor peat lands, grows in open and semi-open areas (Landolt, 1977; Tsyganov, 1983). Cranberry yield is closely connected to forest density – the highest yield is seen in low-density stands (Cherkasov et al., 1981).

Raspberry­ is a perennial semishrub of forest-meadow type. It grows on acidic and neutral rich soils. It can be found in both open and shaded areas (Landolt, 1977; Tsyganov, 1983). Good yields are seen in open and semi-open habitats: burnt areas, clear cut areas, and forest fringes. It takes raspberry 2–3 years to grow in fresh cut areas, and in 3–8 year-old clear cuts it reaches its maximum yield. In old clear-cut areas raspberry is inhibited by the growth of deciduous tree species (Kazantseva, Miryaminova, 2017). After the closure of the tree canopy it can be found in forests of different types in depressed state, bearing almost no fruit.

Cep grows in deciduous forests on soils rich in humus, overgrown with grass and moss. Ceps are usually found in stands with an average density. Fruiting of ceps is significantly increased in complex multi-storeyed stands (Telishevsky, 1986).

Rough boletus is the cep’s closest relative growing in deciduous forests which necessarily contain birch. It has a variety of forms represented in its typical conditions. Stands with average density are best suited for the growth of these mushrooms. Rough boletus is most productive in middle-aged stands.

Table 1. Ranges of changes in fruiting conditions for the most valuable berries and mushrooms in Russia (Landolt, 1977; Tsyganov, 1983; Telishevsky, 1986; Cherkasov et al., 1981)

Orange-cap boletus grows in deciduous forests, aspen and birch forests on clay soil. It can also be found in stands with an average density, as well as on forest fringes and in clearings. It is believed that this mushroom most often grows together with aspen and poplar, more seldom with willow, and can also form mycorrhiza with oak, beech, hornbeam, and birch.

Chanterelle is found in coniferous and deciduous, not very dense mossy forests. It grows throughout the summer and autumn (in June – October) in clumps in coniferous, birch and mixed forests in moist places on all types of soils. It forms mycorrhiza with various trees, most often with spruce, pine, oak, and beech.

Based on the literature data, we put together the understandings of the most favourable forest conditions for achieving the industrial-scale yield of berries and mushrooms (Table 1). Fruiting of species in different types of forest conditions (FCT) is presented in Table 2.

Table 2. Fruit formation (prosrate shrubs) and formation of fruit bodies (mushrooms) in different types of forest conditions (FCT), +- fruiting

OVERVIEW OF MODELS OF FOREST FOOD RESOURCES YIELDS

One of the well-known approaches to assessing the yield of non-wood resources in Russia is using tables of biological and industrial yields presented in the books Taxation manual to forest resources of Russia (besides wood) (Kurlovich, Kositsyn, 2018), and Guide to accounting and evaluation of secondary forest resources and by-products of forest management (Kurlovich et al., 2003). They include more than 500 standard and reference tables on the main types of non-wood resources of Russian forests, classified by forest vegetation zones and forest regions they are made up of; these conditions determine the growth of specific tree species as well as underbrush, grass and shrub storey and ground cover.

The tables were obtained by numerous estimates with the foundation of experimental plots to calculate the yielding power of non-wood resources. Using the tables, it is possible to forecast the yield of a non-wood resource in almost any region. If data on productivity in the studied area are lacking, the authors recommend using tables for the area with similar growing conditions.

The manual and the guide contain data on the regions of the Russian Federation, making it possible to determine the yield of food resources adjusted for the area of growth. Therefore these tables can be used for the entire territory of Russia.

Main variables influencing yields are silvicultural and taxation characteristics of forest areas; based on these data it is possible to predict the productivity of non-wood resources with a certain degree of accuracy.

The first indicator determining the possibility of growth is the type of forest conditions. It is underpinned by edaphic factors that reflect soil fertility and moisture. For example, cranberry is an endemic species for A5 and B5 and does not grow in any other conditions. Predominant species has considerable influence on growing of certain types of resources. It is especially true for mushrooms. Some species, for example rough boletus, live in symbiosis with only one tree species, birch, whereas ceps prefer mixed forests (Pautov, 2009).

The age of stand also determines the possibility of mushroom and berry growth. The majority of forest berries (cowberry, blueberry, cloudberry, and bilberry) reach the industrial-scale amounts in forests where the stand is more than 40–60 years old. This is due to optimal light conditions and less competition with herbage plants. At the same time, mushrooms are most productive in forests where the stand is 15–40 years old. This is due to forest litter thickness: the thinner it is, the faster it warms up, while mushrooms need heat to grow (Zhukova et al., 2008).

Density of forest stand is one of the limiting factors for the growth of berries and mushrooms and the one on which the yield depends. Density influences the amount of light, heat and precipitation getting beneath the forest canopy as well as the composition and development of plants that can compete with berry prostrate shrubs (Malinovskikh, 2017). The explanation may be that if the stand is very dense, plants can’t grow because of lacking light and impossibility to photosynthesize. If the density is low, there is a large number of competitive species that crowd out berry prostrate shrubs. For instance, bilberry does not stand direct sunlight, the optimal light mode for it is created at the density of 0.6–0.8 (Nikitenko, 2016). Cowberry, cloudberry and cranberry are more light-requiring, they bear fruit well at the stand density of 0.3–0.5 and in open places (Gorobets, Slavsky, 2013).

Stands with the density of 0.7–0.8 are optimal for fruiting of all types of mushrooms (Yurkina, 2017). Mushrooms also need light during spore maturation. Thus, the majority of berries and mushrooms depend on a certain light mode, which is taken into account in the taxation manual in the form of stand density. The tables of the manual, however, ignore complex stands with several storeys. The flow of light reaching the ground cover will be several times lower in a multi-storeyed stand than in a pure stand of the same density, which means, that the yielding power of berries and mushrooms will not be the same. Therefore it might be not quite right to use the indicator of density which influences the yielding power of berries and mushrooms, and not to take light into account.

On the basis of stand density and species composition of the 1st and 2nd storeys, understorey trees and underbrush it is possible to determine illumination as a more informative factor influencing the development of berries and mushrooms. The light intensity will then be considered not at the level of the upper storey, but at the level of ground cover, which will allow us to assess the yielding power and growth prospects of the grass and shrub storey more accurately and to give recommendations on how to improve the conditions for forest management focusing on many resources.

We should not ignore the influence of forest management activities on various forest food resources. Forest felling is one of the most important industry-related and anthropogenic factors in terms of its impact on forest biogeocenoses. Partial or complete removal of trees affects light and temperature conditions not only causing significant changes in the tree canopy but also leading to the disturbance of all other storeys of plant and soil cover (Kurlovich et al., 2015). Adequate cuttings can help to increase the yielding power of the resource of interest.

There are species that achieve industry-scale yield only at open areas (clear cut areas, burnt areas, forest fringes), including i.e. European raspberry. This is why raspberry is actively growing and bearing fruit on plots affected by clear cuts.

For berries that grow beneath the forest canopy the optimal measure would be forest thinning to the density at which the greatest yields are achieved. It is necessary to pay attention to the pattern of regeneration after felling. For fruiting of all types of mushrooms mixed stands with the density of 0.7–0.8 are optimal (Bulgakov et al., 1987). Therefore thinnings, especially liberation and disengagement cuttings are to be done on time.

Thinnings help increasing light and heat inflow to the lower phytocenosis storeys, while increasing soil moisture contribute to a microclimate favourable for mushroom fruiting. In complex fresh and moist subors liberation and disengagement cuttings of young stands let grasses, especially cereals, grow, which hinders fruit bearing of mushrooms. When thinning, the density should be reduced to 0.6–0.7, leaving up to 3 deciduous plants in coniferous stands (Klyuchnikov, 2005). After clear cuts of coniferous or deciduous forests their light and temperature conditions on the soil surface change dramatically. In the second year already the plot is overgrown with herbs. Most forest mushrooms start to grow later, when a young forest emerges on the clear cut area. Therefore, in the setting of multi-purpose forest management it is possible to control the yield of food resources (Obydennikov, 2002).

Growing interest in multi-purpose forest management in order to increase the profitability of a forest plot means that it is necessary to make joint forecasts on the use of both wood and non-wood food resources, which is impossible without modelling.

The transition to multi-purpose forest management has already begun around the world. In Finland, there are a lot of studies on bilberry yields in different growing conditions, taking into account a large number of factors (precipitation, temperature, humidity, forest type, stand age) that affect productivity (Wallenius, 1999; Turtiainen et al., 2007, 2012). The dependence of the yield on these factors is demonstrated: middle-aged stands with optimal humidity were found to be best suited for berries to bear fruit. Also, the yielding power is affected by accumulated temperatures.

Empirical models of berry yielding power developed before 2009 were based on regional data on yields in North Karelia (Ihalainen et al., 2003, 2005). Models for Bilberry Yield and Models for Cowberry Yields are used to estimate the yields of bilberry and cowberry. The input parameters are stand age and type. Then the models for bilberry and cowberry yields were calibrated using field data on the yields in southern Finland. These models, however, did not take into account the projective cover of bilberry as a parameter determining the yield. Using the Stand simulator MOTTI made it possible to model the yield of bilberry on different soils in northern and southern Finland (Hynynen, 2003, 2005).

A set of empirical models was developed in 2009 (Miina et al. 2009) predicting first the area of bilberry projective cover – Model for the Percentage Coverage of Bilberry (Model 1) and then the annual yield of berries depending on the stand features – Models for Bilberry Yield (Model 2). Models also allow us to see annual fluctuations in yield forecasts. The models focused exclusively on mineral soil plots (Miina et al., 2010). A model which includes not only mineral but also peat soils was developed in 2014. Then, annual fluctuations of bilberry yield in 2001-2014 were modelled using a much larger data set from MASI, data on permanent experiment areas in Finland. The Finnish name MASI literally means berries (MArja) and mushrooms (SIeni) system, and it consists of a number of different databases. Model 1 was designed to predict the average area of bilberry projective cover based on constant indicators – age, bonitet, and FCT. Model 2 predicts the average amount of bilberry in a stand. Input data include the area covered by bilberry, height of stand and average effective temperature. In 2015, the average annual yield of bilberry for 1997–2013 was calculated using data from MASI (Turtiainen et al., 2011, 2016). Similar models making forecasts in two stages – first the projective cover area and then the yield – were also developed for cowberry: Model for the percentage coverage of cowberry and Model for cowberry yield (Turtiainen et al., 2005, 2013).

Other countries have developed a number of models for species of the Vaccinium genus. For the French forests, the Model the cover/abundance of Vaccinium myrtillus was used that takes into account climatic and edaphic factors (temperature, precipitation, solar radiation, water balance, soil pH and C/N ratio) to predict bilberry yield (Coudun, Gegout, 2007). In Canada, bilberry and blueberry models have been developed to predict yields using climate variables (Hall et al., 1982). In Sweden, various ecosystem services including bilberry production have been modelled in order to make forecasts using stand features, climatic and edaphic factors (Gamfeldt et al., 2013).

Mushrooms are an equally important food resource. Empirical models for forecasting the yield of forest mushrooms: Empirical models for predicting the production of wild mushrooms (Bonet et al., 2008) were designed. Main factors affecting the yielding power are taxation features of forest stands (species – common pine, age, height, area, number of trees per hectare, average diameter, age class). However, mushroom production also depends on weather conditions such as time and amount of precipitation which cannot be accurately predicted in a long-time forecast.

Then, several more factors are added to the models. Terrain affects the predicted production of mushrooms, likely due to the presence of groundwater. Water is the greatest growth limiting factor in the forests of southern Central Pyrenees, where light is abundant and temperatures are mild all year round. Moreover, the model shows the dependence of the mushroom yielding power on the nature of forest restoration: mushroom yield is higher in case the regeneration is natural (Bonet et al., 2010).

In 2014, an expert model of cep yield in the northern Karelian spruce forests was developed in Finland – Expert model for Boletus edulis. 25 experts in cep yield assessment were involved. In addition to the environmental factors (precipitation, temperature and nutrients, ground vegetation) and the state of the stand, the experts also visually assessed the forest based on forest stand pictures (Tahvanainen et al., 2014).

In Spain, a model of forecasting the yield of mushrooms of the Russulaceae family (Lactarius) with two stages of modelling was developed. At the first stage, the probability of mushrooms appearing in the stand is analyzed – Models for the probability of mushroom occurrence, and at the second stage their yield is forecasted – Models for mushroom yield conditional on the probability of mushroom occurrence. Growth conditions for mushrooms are based on the features of climate, soil, stand composition and age. The amount and distribution of precipitation as well as soil pH are the main factors influencing the yielding power (Tay et al., 2016).

The Russian model FORRUSS-S (FORest of RUSsia – Stand) is designed to forecast the dynamic pattern of taxation features of multispecies all-aged stands in an area of up to hundreds of thousands of hectares (Chumachenko, 1993; Chumachenko et al., 2003). This model is an ecological-physiological (explanatory), spatially-oriented model of stands under different scenarios of forest management. The model generates new taxation descriptions (with a 5-year interval), on the basis of which external modules allow evaluating many additional characteristics including forest food resources.

Comparison of the features of non-wood forest resource models is presented in Table 3.

Table 3. Classification of non-wood resource models

The vast majority of the presented foreign models (Finland) is designed and function mainly with single-species stands. No methods of work proposed for multispecies stands could be found in the literature. Stands change over time, and stand density, composition, age and other parameters will change accordingly (Isaev, 2013). Therefore, it is necessary to model the yield of non-wood resources taking into account the dynamics of forest stand features including changes caused by anthropogenic factors. It is thus necessary to integrate the model of dynamics of stand features and the model of evaluation of non-wood forest resources.

It is planned to use the FORRUS-S model of dynamics of the main taxation indicators of multi-species all-aged forests to assess the changes in commercially available forest resources of the central part of European Russia.

Let us describe the FORRUS-S model in detail. The input data of the model are: standard taxation descriptions of strata, stand plans, bio-ecological data bases (growth rates at different ontogenetic states, the attitude to the light, seed and coppice productivity, etc.). A 5-years interval was chosen for modeling, based on the data of population biology about the time needed for the noticeable changes to occur in the growth rate, development and relations of woody plants to light in youth. The Modelling block consists of two models: Natural development and Exogenous influences.

The Natural development model of stands imitates the essential processes that take place in forest stands: natural regeneration, growth, and spontaneous thinning. Modelling makes it possible to forecast changes in average taxation parameters of stands by forest cohorts (elements), i. e. height, diameter, density, stock volume etc. as well as changes in the species and age composition of each forest stratum. Increments are calculated based on light conditions in the three-dimensional simulated space, taking into account the position of the cohort in the forest canopy and potential growth rates determined by the type of forest growth conditions.

An important role is played by the Exogenous effects block which allows simulating a large number of biotic and abiotic effects that are external to the forest cenosis. Abiotic effects include: most types of forestry activities (clear cuts and selection fellings, planting and taking care of forestry crops, thinnings of various intensity etc.), hydrotechnical melioration, fertilization, etc.

The model can be expanded with additional modules, including a module on food resource accounting which we plan to use in our work.

CONCLUSION

Currently, the trend towards multi-purpose forest management has increased. The reason is that the management aimed at obtaining wood can also result in additional income from other forest products, for example through picking wild berries and mushrooms. This is why there is a need to model joint production of wood and forest food resources.

This article presents the conditions of growth and fruiting of the most valuable mushrooms and berries of the center of the European Russia: they turned out to be FCTs А2, А3, В2, and В3. The influence of forestry activities such as clear cuts and selection fellings of mature and overmature stands as well as thinnings on the yield of certain species of berries and mushrooms is considered.

The analysis of Russian and foreign literature showed that the yielding power of forest food resources within one forest area is closely related to the type of forest conditions, species, age and density, i.e. taxation data, the dynamics of which is predicted by the majority of forest models. The problem of non-wood resources estimation in mixed stands was revealed, as the considered models of calculating mushroom and berry yield are only true for pure same-age stands. Evaluation of food resources in mixed stands requires taking into account the share of species that form the forest stand.

Density of the tree storey is not always a significant factor of berry and mushroom yielding power. The dependence of yielding power on the density is disturbed in the presence of the second storey, understorey trees or underbrush; therefore when creating models of food resource dynamics it is proposed to take into account the illumination level under the trees’ crowns, which determines the economic yielding power of the stock of berries.

ACKNOWLEDGEMENTS

The study was carried out as part of the FP7 ERA – Net Sumforest-POLYFORES project with the financial support of the Ministry of science and higher education of the Russian Federation (project unique identifier RFMEFI61618X0101), the results of the study were processed by of state assignment «Methodical approaches to the assessment of the structural organization and functioning of forest ecosystems»№ АААА-А18-118052400130-7.                                                                     

 REFERENCES

Bol’shakov B.M. Sostojanie i perspektivy ispol’zovanija nedrevesnyh resursov lesa (State and prospects for the use of non-timber forest resources), Pushkino: VNIILM, 2014, pp. 7-11.

Bonet J.A., Palahi M., Colinas C., Pukkala T., Fischer C.R., Miina J., Martínez de Aragón J. Modelling the production and species richness of wild mushrooms in pine forests of Central Pyrenees in northeastern Spain, Canadian Journal of Forest Research, 2010, Vol. 40, No. 2, pp. 347-356.

Bonet J.A., Pukkala T., Fischer C.R., Palahi M., Martínez de Aragón J., Colinas C. Empirical models for predicting the production of wild mush-rooms in Scots pine (Pinus sylvestris) forests in the Central Pyrenees, Annals of Forest Science, 2008, Vol. 65, No. 2, pp. 1-8.

Bulgakov N.K., Koz’jakov S.N., Fesjuk A.V. Tehnologija zagotovki i pererabotki nedrevesnyh resursov lesa (Technology of harvesting and processing of non-timber forest resources), Moscow: Lesnaja promyshlennost’, 1987, 224 p.

Cherkasov A.F., Butkus V.F., Gorbunov A.B. Kljukva (Cranberry), Moscow: Lesnaja promyshlennost’, 1981, 214 p.

Cherkasov A.F., Shutov V.V., Mironov K.A. Vosstanovlenie zaroslej brusniki i cherniki posle sploshnyh rubok (Restore thickets of lingonberry and blueberry after clear cutting), Lesovedenie, 1988, No. 4, pp. 42-48.

Chumachenko S.I. Bazovaja model’ dinamiki mnogovidovogo raznovozrastnogo lesnogo cenoza (The basic model of the dynamics of multiple species of different age forest cenosis), Voprosy jekologii i modelirovanija lesnyh jekosistem, Moscow: MLTI, 1993, No. 248, pp. 147-180.

Chumachenko S.I., Korotkov V.N., Palenova M.M., Politov D.V. Simulation modelling of long-term stand dynamics at different scenarios of forest management for coniferous — broad-leaved forests, Ecological Modelling, 2003, Vol. 170, No. 2-3, pp. 345-361.

Coudun C., Gegout J. Quantitative prediction of the distribution and abundance of Vaccinium myrtillus with climatic and edaphic factors, Journal of Vegetation Science, 2007, No 18, pp. 517-524.

Cyganov D.N. Fitoindikacija jekologicheskih rezhimov v podzone hvojno-shirokolistvennyh lesov (Phytoindication of ecological regimes in the subzone of coniferous-deciduous forests), Moscow: Nauka, 1983, 196 p.

Egoshina T.L. Nedrevesnye rastitel’nye resursy Rossii (Non-timber plant resources of Russia), Moscow: NIA-Priroda, 2005, 80 p.

Gamfeldt L., Snäll T., Bagchi R. Higher levels of multiple ecosystem services are found in forests with more tree species, Nature Communications, 2013, Vol. 4 No.1340, pp. 1-8.

Gorobec V.A., Slavskij V.A. Nedrevesnaja produkcija lesa (Non-wood forest products), Voronezh: VGU imeni G.F. Morozova, 2013, 169 p.

Hall I.V., Aalders L.E., McRae K.B. Lowbush blueberry production in eastern Canada as related to certain weather data, Canadian Journal of Plant Science, 1982, Vol. 62, No.3, pp. 809-812.

http://www.consultant.ru/document/cons_doc_LAW_64299/ (2018, 17 November).

Huber P., Kurttila M., Hujala T., Wolfslehner B., Vacik H. Managing for NWFPs — an assessent on the forest holding level // Wild Forest Products in Europe. Barcelona, 13–14 October, 2016, Barcelona: European Forest Institute, 2016. p. 9.

Hynynen J., Ahtikoskib A., Siitonen J., Sievaonen R., Liski J. Applying the MOTTI simulator to analyse the effects of alternative management schedules on timber and non-timber production, Forest Ecology and Management, 2005, Vol. 207, No. 1-2. pp. 5-18.

Hynynen J., Ojansuu R. Impact of plot size on individualtree competition measures for growth and yield simulators, Canadian Journal of Forest Research, 2003, Vol. 33. No. 3, pp. 455-465.

Ihalainen M., Pukkala T., Saastamoinen O. Regional expert models for bilberry and cowberry yields in Finland, Boreal Environment Research, 2005, No 10, pp.145-158.

Ihalainen M., Salo K., Pukkala T. Empirical prediction models for Vaccinium myrtillus and V. vitis-idaea berry yields in North Karelia, Silva Fennica, 2003, Vol. 37, No. 1, pp. 95-108.

Isaev A.S., Rysin L.P., Smirnova O.V. Raznoobrazie i dinamika lesnyh jekosistem Rossii (Diversity and dynamics of forest ecosystems of Russia), Moscow, Tovarishhestvo nauchnyh izdanij KMK, 2013, 478 p.

Jurkina E.V. Resursnyj potencial nedrevesnoj produkcii lesa (Resource potential of non-timber forest products), Syktyvkar: SLI, 2017, 240 p.

Kazanceva M.N., Mir’jaminova L.R. Plodonoshenie maliny obyknovennoj (Rubus idaeusl.) v lesah na juge tjumenskoj oblasti (Fruiting raspberry ordinary (Rubus idaeusl.) In the forests in the south of the Tyumen region), Aktual’nye problemy lesnogo kompleksa, 2017, No. 47, pp.1-4.

Kljuchnikov L.Ju. Pobochnoe lesopol’zovanie (Secondary forest management) Moscow: GOU VPO MGUL, 2005, 68 p.

Kosicyn V.N. Jekonomicheskaja ocenka resursov dikorastushhih jagodnikov (Economic valuation of wild berry resources), Rastitel’nye resursy, 1998, Vol. 34, No. 4, pp. 76-81.

Kosicyn V.N. Stoimostnaja ocenka resursov moroshki v podzone juzhnoj tajgi (Cost estimate of cloudberry resources in the southern taiga subzone), Lesohozjajstvennaja informacija, 1996, No. 12, pp. 18-23.

Kozubov G.M., Taskaev A. I. Lesnoe hozjajstvo i lesnye resursy Respubliki Komi (Forestry and forest resources of the Komi Republic), Moscow: “Dizajn. Informacija. Kartografija”, 2000, 512 p.

Kurlovich L.E, Nikolaev A.F., Cherkasov V.N., Kosicyn G.V. Rukovodstvo po uchetu i ocenke vtorostepennyh lesnyh resursov i produktov pobochnogo lesopol’zovanija (Guidance on accounting and valuation of secondary forest resources and secondary forest products), Pushkino: VNIILM, 2003, 315 p.

Kurlovich L.E., Kosicyn V.N. Taksacionnyj spravochnik po lesnym resursam Rossii (za iskljucheniem drevesiny), Pushkino: VNIILM, 2018, 282 s.

Kurlovich L.E., Pankov V.B., Kivileva I.M. Vlijanie lesohozjajstvennoj dejatel’nosti na sostojanie i produktivnost’ pishhevyh i lekarstvennyh rastenij (The impact of forestry activities on the state and productivity of food and medicinal plants), Lesohozjajstvennaja informacija, 2015, No. 2, pp. 24-34.

Landolt E., Okologische Zeigerwerte zur Schweizer Flora, Veroffentlichungen des Geobotanischen Instituts der ETH, Stiftung Rubel, Zurich, 1977, Vol. 64, pp. 208.

Malinovskih A.A. Vlijanie urovnja osveshhennosti pod pologom lesa na urozhajnost’ cherniki v uslovijah Sredne-obskogo bora Altajskogo kraja (Possession of the lightness of illumination under the forest canopy on the yield of blueberries in the conditions of the Middle Ob forest of the Altai Territory), Vestnik Altajskogo gosudarstvennogo agrarnogo universiteta, 2017, Vol. 6, No. 152, pp. 87-92.

Miina J., Pukkala T., Hotanen J.-P., Salo K. Optimizing the joint production of timber and bilberries, Forest Ecology and Managemen, 2010, Vol. 259, Issue 10, pp. 2065-2071.

Miina J., Hotanen J.-P., Salo K. Modelling the abundance and temporal variation in the production of bilberry (Vaccinium myrtillus L.) in Finnish mineral soil forests, Silva Fennica, 2009, Vol. 43, No 4,  pp. 577-593.

Nikitenko E.B. Nedrevesnye resursy lesa (Non-timber forest resources), Irkutsk: Izd-vo BGU, 2016, 222 p.

Obydennikov V.I., Avdeev A.N., Avdeev Je.N. Ispol’zovanie i vosproizvodstvo resursov jagodnikov v svjazi s rubkami v sel’skih lesah Novgorodskoj oblasti (Use and reproduction of berry resources in connection with logging in rural forests of the Novgorod region), Lesohoz. Inform, 2002, No. 10, pp. 15-21.

Pautov J.A., Zasuhin D.P. Rekomendacii po vydeleniju uchastkov massovogo sbora gribov i jagod mestnym naseleniem (Recommendations for the allocation of sites for the massive collection of mushrooms and berries by the local population.), Syktyvkar: Komi regional’nyj nekommercheskij fond «Serebrjanaja tajga», 2009, 17 p.

Tahvanainen V., Kurttila M., Miina J., Salo K. Predicting the yields of commercially important mushrooms in Finland, Symposium on wild forest mushrooms and other NTFPs: Innovations and perspectives, La Pocatiеre, 27-28 August 2014, La Pocatiеre: Centre for Forest Research, 2014, рр.1-11.

Taye Z.M., Martínez-Peña F., Bonet J.A., Martínez de Aragón J., De-Miguel S. Meteorological conditions and site characteristics driving edible mushroom production in Pinus pinaster forests of Central Spain, Fungal Ecology, Vol. 23, 2016, pp. 30-41.

Telishevskij D.A. Kompleksnoe ispol’zovanie nedrevesnoj produkcii lesa (Integrated use of non-timber forest products ), Moscow: Lesnaja promyshlennost’, 1986, 261 p.

Turtiainen M., Miina J., Salo K., Hotanen J.-P. Empirical prediction models for the coverage and yields of cowberry in Finland, Silva Fennica, 2013, Vol. 47, No. 3, pp. 1-22.

Turtiainen M., Miina J., Salo K., Hotanen J.-P. Modelling the coverage and annual variation in bilberry yield in Finland, Silva Fennica, 2016, Vol. 50, No. 4, pp.1-12.

Turtiainen M., Nuutinen T. Evaluation of information on wild berry and mushroom markets in European countries, Small-scale Forestry, 2012, Vol. 11, No. 1, pp. 131-145.

Turtiainen M., Salo K., Saastamoinen O. Model-based estimates of regional and national bilberry and lingonberry yields on mineral soils in Finland, Finland: Research Notes, 2005, 44 p.

Turtiainen M., Salo K., Saastamoinen O. National and regional estimates of blueberry (Vaccinium myrtillus L.) and lingonberry (V. vitis-idaea L.) yields on peatlands in Finland, Finnish Peatland Society Helsinki, 2007, Vol. 58, No. 3–4, pp. 87-98.

Turtiainen M., Salo K., Saastamoinen O. Variations of yield and utilisation of bilberries (Vaccinium myrtillus L.) and cowberries (V. vitis-idaea L.) in Finland, Silva Fennica, 2011, Vol. 5, No. 2, pp. 237-251.

Vacik, H., Huber P., Kurttila M., Hujala T., Wolfslehner B., la Sánchez-González M., Pasalodos-Tato M., de Miguel S., Bonet J.A., Marques M., Borges J.G., Enescu M.C., Dinca L. Comparing the potential of on-Wood Forest Products across case studies in Europe, Conference on Non-Timber Forest Products and Bioeconomy, Rovaniemi, 28–30 November, 2017, Helsinki: Natural Resources Institute Finland, 2017, pp. 201-204.

Wallenius T.H. Yield variations of some common wild berries in Finland in 1956–1996, Annales Botanici Fennic, 1999, No 36, pp. 299-314.

Zhukova A.I., Grigor’ev I.V., Grigor’eva O.I., Ledjaeva A.S. Lesnoe resursovedenie (Forest Resource Management), St. Petersburg: SPb GLTA, 2008, 215 p.

Zvorykina K.V. Vlijanie vyrubki na urozhajnost’ cherniki (Impact of logging on blueberry yield), Kirov: Vsesojuznyj nauchno-issledovatel’skij institut ohotnich’ego hozjajstva i zverovodstva, 1972, pp.17-19.

Reviewer: DSc in biology Chertov O.G.

© 2019 г.

        E.N. Sochilova¹, N.V. Surkov^{1, 2}, D.V. Ershov¹, V.A. Khamedov³

¹Center for Forest Ecology and Productivity of the RAS

Profsoyuznaya st. 84/32 bldg. 14, Moscow, 117997, Russia

² Lomonosov Moscow State University, Moscow, 119991, Russia

³ Yugra State University, Khanty-Mansiysk, 628012, Russia
E-mail: elena@ifi.rssi.ru
Received 30 October 2018

The paper describes assessment of spatial biomass of top wood layer based on combination of high-resolution Landsat-8 satellite images and selected ground forest inventory data measurements. Test area is one of forestry of Khanty-Mansiysk region. Segmentation of satellite images for spectral homogeneous land sites (segments) mapping is applied. Land category, dominated specie, age and wood stock volume for these sites are defined. Ground forest inventory data and segments used for selection of segments for dominated specie classification and validation of obtained map. The first, nine types of land cover are classified, four of them belong to forest cover with dominating of pine, spruce, cider and birch. The reference sample is updated by segments of such non-forest classes as fires, cuts and other non-forested lands, swamps, water internal bodies. Twelve spectral metrics are used for classification: reflectance in blue, green, red and near-infrared bands of Landsat-8. There are following vegetation seasons: and of winter, beginning of spring and middle of summer. The most significant informative metrics are the reflectance in the NIR band of the spring image, also green and red bands of the summer image. Random Forest algorithm is applied for training classification. The total accuracy of land categories and dominated species classification is 86,3%. Cross-validation of the classification based on the control sample was 0.712. In the second stage, we used regression models to relate the reflectance in the red band of the winter image with the taxation characteristics of the wood stock and age of the forest species in the selected reference segments. The level of relationship between the reflectance and wood stock values were equal to 0.80 for pine, 0.56 for dark coniferous species and 0.73 for birch. Between the reflectance and the specie height is following 0.75 for pine, 0.61 for birch and 0.64 for dark coniferous species. A check with control data showed that the error in estimating the wood stock above 250 m³ / ha for birch is 15.4%, for pine — 19.0% and for dark coniferous species — 5.5%. We used regional growth tables and the mean heights reconstructed from the regression equations for calculation mean specie ages. Then the age groups (according regional felling age) for each species are determined and the wood stocks are converted into wood biomass by conversion coefficients. As a result, maps of mean ages, heights, wood stock in m³/ha and biomass in t/ha were created. Based on these maps quarter assessments of the areas and stocks of the main dominated forest species of our test area, including felling age forest stands, were carried out.

Key words: stand biomass, wood stock volume, remote sensing data, Landsat-8, forest classification, Random Forest, forestry

Combining multi-season and multi-temporal images with forest inventory data allows updating the spatial information on the distribution of the main land categories and predominant tree species in the study area, as well as identification the recent changes associated with fires, loggings and other impacts on forests. This paper presents the methods and results of the spatial assessment of the top layer wood biomass using Landsat 8 satellite multi-season images of high spatial resolution and selected forest inventory data. A plot for research was selected on the territory of Sovetskoye forestry of Khanty-Mansiysk Autonomous Okrug.

Traditionally, it is believed that there are correlations between reflectance characteristics in different spectral ranges and some properties of the vegetation cover (Vinogradov, 1984; Poso et al., 1987), in particular the forest inventory characteristics of trees. Among the revealed regularities a link between the reflectance in the red spectral range and the stocks of stem wood in the study area was mentioned (Kuusela, Poso, 1970). In this paper it is also proposed to use the above-named pattern for the assessment of wood stocks and it is showed that the best results are achieved in winter, in the absence of tree leaves. Such studies were developed on a broad scale when mass inflow of similar data from Landsat satellites began in 1972 in the United States and Northern Europe. These data have been widely used to obtain tree stand characteristics such as height, projective forest cover and leaf area index (LAI) (Hall et al., 2003; Zhang et al., 2014). Many studies have focused on gaining information on green biomass volume and vegetation productivity (Hame et al., 1997; Hall et al., 2006; Ji et.al., 2012; Zhang et al., 2014). Use of optical remote sensing data for the quantification of forest fuel stocks (Arroyo et al., 2008; Gharun et al., 2017) or the volume of stem wood for logging purposes (Fazakas et al., 1999; Zheng et al. 2014) is of greater interest. Such use of this type of data is, however, much less common. Information on stands obtained from ground-based studies can be used both to construct models for remote assessment of wood stocks (Tokola, 2000; Hall et al., 2006; Chirici et al., 2008; Mura et al., 2018) and to verify the accuracy of the results (Tokola, Heikkilä, 1997; Halme, Tomppo, 2001). In Russia, more attention is paid to the use of remote sensing optical range data for the classification of forest species composition, but there are also papers on the assessment of wood stocks, for example in the southern part of the Valdai Hills (Kozlov et al., 2007) where a digital elevation model was used together with Landsat 5 and Landsat 7 data, as well as in Krasnoyarsk Krai on the basis of Landsat 7 data (Sochilova, Ershov, 2012) and in Primorsky Krai on the basis of the medium spatial resolution data of PROBA-V satellite (Zharko et al., 2018).

STUDY OBJECT AND INPUT DATA

The study area of 2.300 km² is located within Sovetsky district of Khanty-Mansiysk Autonomous Okrug, to the east of Agirish settlement, to the north of the regional center, Sovetsky town, and to the south-east of «Malaya Sosva» nature reserve. The research area includes parts of three local forestry areas – Torskoye, Esskoye and Zelenoborskoye, which are subunits of the Sovetskoye forestry. In physical and geographical terms, the territory is located within the West Siberian Plain, on the southern spurs of the North Sosva Upland with absolute heights up to 233 m at the raising of Malososvinsky amphitheatre. Through these spurs passes the watershed between the basins of the Severnaya Sosva (flows into the Ob) and Konda (flows into the Irtysh) rivers. The lowest level of the water line is at an absolute height of 72 m. The relief of the western part of the territory is an elevated wavy morainic plain, with deeply embedded valleys of the Ess river tributaries. The eastern part is a flat fluvioglacial plain, with a much lower degree of dissection (Gvozdetsky, Mikhailov, 1978).

The climate of the studied area is continental, rather humid; in summer it is largely determined by the western cyclonic transport of Atlantic air masses, in winter – by the Asian continental anticyclone. The average temperatures in January are around minus 18 ^°C, in July – plus 17.5 ^°C. The frost-free period lasts for 90 days. Annual precipitation is up to 500 mm but due to the flat terrain and poor drainage the area is swampy (Geographical conditions…, 2018).

The study area belongs to the Severnaya-Sosva landscape province of the northern taiga forest-swampy zone of the West Siberian Plain. Forests cover 76% of the territory of the Sovetskoye forestry (Forest management regulations…, 2018). The most typical are pine green moss and shrub forests with an admixture of larch. They often have a small (8-12 m) height and a broken canopy, may turn into waterlogged open woodlands. The share of pine forests in the Sovetskoye forestry is 74%, where active logging takes place. Spruce-cedar plantations occupying less than 10% of the area are present on the drained slopes of river valleys. Flat parts of the interstream areas have a significant share of birch forests with admixed aspen and pine, often also waterlogged. In addition to forests, large areas are occupied by raised sphagnum bogs (Atlas of forests…, 1973, Gvozdetsky, Mikhailov, 1978).

Input materials are cloud-free Landsat 8/OLI scenes with a spatial resolution of 30 m obtained at different seasons: late winter (23.02.2014), beginning of spring vegetation (19.05.2016) and mid-summer (20.06.2016) (Fig. 1). We used reflectance data in blue, green, red and near infrared bands. Satellite data values are unified by converting the data into dimensionless absolute values of reflectance in the range from 0 to 1 showing the ratio between the amount of the stream of sun falling onto the object and the amount of reflected light (Markham, Barker, 1986; Belova, Ershov, 2012). The paper specially focuses on the investigation of winter reflectance in the red band to assess the stand volume through the intensity of sun beams reflected from the snow between tree crowns.

We used also selected ground-based inventory data with the information on species composition, heights, age and stocks of stem wood for each forest stand area.

Figure 1. Spring (a), summer (b), winter (c) Landsat-8 satellite images of the test area

RESEARCH METHODS

Estimation of biomass stocks of forest species using satellite images of high spatial resolution includes following several steps:

• segmentation of combined multiband data for the summer season into spectrally-homogeneous areas of the earth surface;

• segment selection using inventory data to train the classifier and to assess the accuracy of classification results;

• classification of the land cover, analysis of performance of spectral metrics and evaluation of the accuracy of the obtained map;

• construction of regression relationships between reflectance values in the winter red band with forest-inventory characteristics (stock of stem wood and height) on selected reference segments;

• mapping of stem wood stocks, assessment of stock estimation accuracy;

• mapping of average heights of the stands, assessment of height determination accuracy;

• calculation of average age of stands using tables of forest allometric equations , determination of age groups for each species;

• conversion of wood stocks into biomass using conversion coefficients of equations.

As a result we created maps of average ages, heights, wood stock (m³/ha) and biomass (t/ha) of forest stands.

The procedure of supervised classification requires a prepared training datasets. In this case, spectrally-homogeneous segments are considered as training plots. We used Full Lambda Schedule (FLS) segmentation technique (Redding et al., 1999) for the assessment of land categories, predominant species, age and growing stock. Dominated forest specie stand plots derived from the forest inventory data with the following parameter– at least seven units in the top layer of the stand. Overlapping of segments and taxation plots makes it possible to create reference plots with the area of more than 1 hectare with one predominant species. In addition, to eliminate accidental artifacts or grave mistakes, visual expert analysis is carried out, using information about the typical conditions of certain tree species growth, as well as possible anthropogenic impacts. The reference sample of forest species is supplemented by segment plots of non-forest classes: burnt areas, clear cut areas, other non-forested areas, swamps, and water bodies.

To reach our goal, the following steps were defined: (1) classification of the land cover using all spectral metrics without exception; (2) analysis of informational value of spectral metrics and looking for their optimal number for classification; (3) evaluation of the accuracy of the results using an validation sample plots (error matrix).

Classification and analysis of performance of spectral reflectance characteristics for the classification of dominated forest species is performed using the Random Forest statistical method (Breiman, 2001). The method is based on constructing a large number of decision trees from the initial training sample. This algorithm also includes the optimal number of metrics, with which a tree divides classes most clearly, trying to ensure that each tree leaf contains observations of only one class. The resulting classification variants with different numbers of metrics are compared by the kappa coefficient (Tyurin, Makarov, 1998). The higher the value is, the more consistent the separation is. In the course of solving this task, the best consistency was achieved using seven metrics. For internal verification purposes, the entire Random Forest algorithm was executed several times, in our case – five times. The final classification version is based on a simple vote of the classifiers defined by individual trees (Chistyakov, 2013).

The Random Forest technique makes it possible to evaluate performance of the metrics using the «importance measure» of the metric for classification proposed by Breiman (Breiman, 2001). For each metric, a random forest is constructed and the probability of an OOB (Out-Of-Bag) misclassification is estimated. Then, in OOB samples, a random permutation of the metric values is performed and the probability of an erroneous classification with the modified OOB samples is estimated. The performance is estimated by averaging the difference of error probability values for all trees before and after mixing these values. Error values are standard deviation—normalized (Chistyakov, 2013). This parameter is conventionally called the «mean decrease in accuracy» (MDA). The more pixels start to leave the class defined at the beginning and mix with others, the greater MDA value will result from random mixing of the sample, which means that the metric whose values are mixed is more informative (Guyon, Elisseeff, 2003).

Using Random Forest, it is possible to assess the classification accuracy by comparing it with a set of control pixels randomly pre-selected from the original sample of reference pixels. The initial sample is divided into two equal subsets: the first is used to train the model, the second – to check the classification accuracy. The result of this analysis is a class confusion matrix showing the overall accuracy of classification.

For the analysis of performance of the metrics the classification is carried out several times with the consequent exclusion of individual spectral metrics with the lowest MDA. Each time, a confusion matrix is constructed and the number of correctly recognized pixels and the classification error of the first and second kind are estimated. Basing on the obtained errors, a diagram of classification accuracy versus the number of used metrics is constructed. This allows using the maximum value of correctly recognized pixels to determine the optimal number of informative metrics for species recognition in the study region.

As a result of the classification, a map of the land cover of the study area is created.

The next step is the construction of regression relationships between brightness values in the winter red band, with forest inventory characteristics (stock of stem wood and tree height) of the selected reference segments, to estimate the stock of wood and average age of forests (Sochilova, Ershov, 2012). The process of stocks and heights reconstruction begins with preparing of training and validation datasets, i.e. plots of stand with the dominated species and a different stock of stem wood. We used the same datasets as in the case of the land cover types classification, divided in half on training and validation subsets. If too small data are obtained, visual image analysis is used to select additional standards of a species. Then, average reflectance values for the training dataset are extracted from a Landsat 8 image scene. Equations of correlation between reflectance values of each species and forest inventory characteristics within homogeneous stand areas are then reconstructed. The estimation of the wood stock of dark coniferous species may be difficult because of the small number of pure reference stands of young trees with low stocks. A possible solution to this problem may be selection of additional support areas of deciduous stands with dark coniferous species in the second layer.

The age of stand is determined by height using regional tables of forest allometric equations and tables of stand productivity. These tables of modal stand were analyzed (Shvidenko et al., 2008; Nagimov, 2011; Sekerin, 2015; Onuchin, 2017). Tabular data and equations of correlation are compared with the actual data on the height and age of the stands of the study area. The equations that are closest to the actual ones are selected and used to further determination of the stands age. The age determined in this way allows proceeding to the age groups of forest stands. Wood biomass values for each age group of individual species are calculated using the stocks of stem wood and conversion coefficients (Zamolodchikov et al., 2003). Thematic raster maps of the stem wood stocks and biomass, heights and age are constructed.

RESULTS AND DISCUSSION

Classification of land cover. Based on the analysis of forest inventory materials, four predominant forest species forming stands of sufficient purity and area were identified: Scotch pine, spruce, Siberian pine (cedar) and birch. Stands of maple, linden and aspen presented in the test area forests less than seven units in a stand. So it was not possible to form a sample of training datasets for them. The summer Landsat 8 image (bands 2-5) was divided into homogeneous spectral segments. Overlapping between the segments and selected taxation stand areas made it possible to form reference plots for training the classifier and result monitoring. The reference sample of forest species was supplemented with segments for non-forest classes: burnt areas, clear cut areas, non-forested lands, swamps, and water bodies. As a result, the reference sample for training and validation of classification accuracy consisted of 1,811 segments for four classes of forest species and 252 segments for five non-forest classes.

At the first stage, the territory is classified into 9 types of land cover, four covered with forest classes with dominant species: pine, cedar, spruce, and birch. 12 spectral metrics are used for classification: reflectance in the blue (#2), green (#3), red (#4) and near infrared (#5) bands of the OLI Landsat 8 system for winter, spring and summer. Random Forest algorithm carried out the land cover classification based on the reference sample using all 12 metrics (Fig. 2).

Overall accuracy of forest species classification was 86.3%, error of the first kind (underestimation) – 15.1% and of the second kind (overestimation) – 12.2% (Table 1).

The largest underestimation or skipping validation pixels was seen in spruce (41%, mainly due to its confusion with cedar stands), the most significant overestimation or false classification was also observed in spruce (35%). For cedar, these values are 34 and 28%, for pine – 30 and 18%, and for birch – 8 and 7%, respectively.

Using the Breiman’s importance measure, the performance value of each metric was determined by the mean decrease in accuracy (MDA) before and after random mixing of the values of each metric in the modified sample, which makes it possible to rank the metrics by their importance. Figure 3 shows that the highest MDA with sample mixing is found for reflectance in the NIR band of the spring image, followed by the green and red bands of the summer image. The least informational value is shown by the green band in winter and spring. The evaluation of metric perfomance value was carried out based on these data, by their successive exclusion from the classification and comparison of the results with the validation sample. Initially, all 12 metrics were used for classification. The cross-validation of classification results using the validation sample gave the value of 0.712 (Fig. 4). Then, the winter green band was excluded from the analysis, which reduced the cross-validation accuracy to 0.614. Further successive exclusion of the green and blue bands (spring) and the blue band (winter) did not change the dynamics of the classification accuracy change, reducing its accuracy to 0.56 with eight metrics. Thus, all twelve spectral metrics are informative for the classification of the land cover of the study area.

Figure 2. The land cover map of the test site

Table 1. Error matrix of land cover class recognition

Figure 3. Assessment of performance metrics for classification of land cover

Figure 4. Change of accuracy of classes recognition depending on number of the used metrics

Determination of stem wood and biomass stocks . Test plots – segments in accordance with the inventory database for coniferous and deciduous stands – were selected in such a way as to reflect the largest range of the stocks of wood. In total, the sample included 350 plots of pine stands, 112 plots of birch and 77 plots of dark coniferous species. The latter include spruce and cedar. The area of the plots is from 1.8 to 9.5 hectares. Estimation of the stock of dark coniferous wood is difficult in terms of finding standards for young stands with low stocks. To solve this problem, additional support plots of birch stands with dark coniferous species in the second layer were selected, where the number of units of dark coniferous plants in the second layer was to be 7 or more. In the study area, the selection of 28 additional plots of this kind allowed us to increase the level of confidence of the reconstruction of dark coniferous species stocks from 0.56 to 0.63. For the training sample from the Landsat 8 winter scene, pixel spectral-reflectance brightness values are extracted and average value for each area is calculated. The obtained regression relationships between the reflectance values in the winter red band and forest inventory characteristics (stock and height) look like exponential curves (Fig. 5). The level of stocks correlation was 0.80 for pine, 0.56 for dark coniferous species and 0.73 for birch; heights correlation — 0.75 for pine, 0.64 for dark coniferous species and 0.61 for birch. The validation sample was used to assess the quality of the results. The error in determining the stocks of birch was 39.5% for small stocks (up to 50 m³/ha inclusive) and 15.4% for stocks exceeding 250 m³/ha, of pine – 50.4% (up to 50 m³/ha inclusive) and 15.5% (above 350 m³/ha), for dark coniferous species 22.2% (up to 150 m³/ha) and 2.7% (above 300 m³/ha), respectively (Fig. 6).

Figure 5. Relationship between the reflectance in the red band of the winter image and wood stock values (m3/ha) – top pictures; Relationship between the reflectance in the red band of the winter image and forest stand height (m) – down ones; (a) — Evergreen dark-coniferous species, (b) – Pine; (c) – Birch

Figure 6. The error (percent) in estimating of the wood stock (top pictures) and heights (down ones) for forest species.
(a) — Evergreen dark-coniferous species, (b) – Pine; (c) — Birch

Using the coefficients of the regression equations, the reflectance values were converted into stem wood stocks and heights of pine, dark coniferous species and birch in accordance with the classification results. The result is a map of the average wood stock in m³/ha in each pixel (Fig. 7), and the heights of stands (Fig. 8).

Figure 7. Map of wood stock values, m3/ha

Figure 8. Map of forest heights (m)

Before determining the biomass stock, stand age was determined. Biomass is calculated by stem wood stocks and conversion coefficients determined for each tree species, latitudinal zone and age group of plants (Zamolodchikov et al., 2003). Age groups for different species are different and are accepted according to forestry regulations of the forestry of interest (Table 2). Age is determined by growth curves. The actual data best corresponded to the dependencies given in the work of A. Z. Shvidenko et al. (2008): for pine of bonitet (site index) V in the northern taiga of West Siberia, birch of bonitet IV in the middle taiga of Siberia and Siberian pine of bonitet V in the taiga of the northern and middle Urals in hydromorphic conditions. The age of plants of the respective species is calculated for each image pixel that contains the height data, and a map of plant ages of the studied area is constructed based on these three equations (Fig. 9), so as to determine age groups afterwards (Fig. 10). Stocks of wood are reconverted in biomass using conversion coefficients and the map of biomass overall stocks is created (Fig. 11) within each plant age group.

Table 2. Plant age groups according to the regulations of the Sovetskoye forestry (years)

Figure 9. Relationship between ages and heights for dark-coniferous species(a), pine (b); birch (c).
It is numbered: 1 – values according to allometric tables of growth of trees and forest productivity (Shvidenko et.al., 2008), 2 — the restored values of specie ages

Figure 10. Maps of mean ages of forest stands, years

Forest stand and quarters polygons, in which more than 50% of the area has tree logging age are shown in yellow colour borders and in red colour hatching, consequently

Figure 11. Map of total forest wood biomass in t/ha

CONCLUSION

The present paper has shown the possibility to estimate forest canopy biomass stocks, to determine heights and ages of stands on the basis of classification of Landsat 8 satellite multi-season images of high spatial resolution and selected forest inventory data. Our research resulted in creating thematic maps of average ages, heights, growing stocks in m³/ha and stocks of biomass in t/ha. On the basis of maps quarterly estimates of the areas and stocks of the main forest-forming species of the test area, including forest plots of the felling age, were carried out. The proposed estimates can be used as relevant information for forestry enterprises, as well as for estimation of the stocks of forest fuels in the upper canopy of the study area.

ACKNOWLEDGEMENTS

The research methods were developed in connection with the governmental assignment given to the Center for Forest Ecology and Productivity of the RAS in the line of the «Concept of satellite monitoring of the state and dynamics of forest ecosystems» (No. 0110-2018-0001), the program of the Presidium of the RAS «Development of methodological approaches to the assessment of ecosystem functions and services of forests» (No. 0110-2018-005), state assignment «Methodical approaches to the assessment of the structural organization and functioning of forest ecosystems» (No. АААА-А18-118052400130-7).

REFERENCES

Atlas lesov SSSR (The Atlas of Forests in USSR), Moscow: GUGK, 1973, 222 p.

Belova E.I., Ershov D.V., Metod predvaritel’noi obrabotki stsen Landsat-5/7 s izobrazheniem snezhnogo pokrova (The method for processing of Landsat-5/7 scenes with snow cover), Sovremennye problemy distantsionnogo zondirovaniya Zemli iz kosmosa, 2012, Vol. 9, No. 4, pp. 9-14.

Vinogradov B.V., Aerokosmicheskii monitoring ekosistem (Ecosystem monitoring from air and space), Moscow: Nauka, 1984. 320 p.

Gvozdetskii N.A., Mikhailov N.I., Fizicheskaya geografiya SSSR. Aziatskaya chast’ (Physical geography of USSR. Asian part), Moscow: Mysl’, 1978, 512 p.

URL://m-sosva.ru/? page_id=154

Zharko V.O., Bartalev S.A., Egorov V.A., Issledovanie vozmozhnostei otsenki zapasov drevesiny v lesakh Primorskogo kraya po dannym sputnikovoi sistemy Proba-V (Possibilities of timber stocks inventory using Proba-V satellite data in Primorsky Krai), Sovremennye problemy distantsionnogo zondirovaniya Zemli iz kosmosa, 2018, Vol. 15, No. 1, pp. 157-168.

Zamolodchikov D.G., Utkin A.I., Chestnykh O.V., Koeffitsienty konversii zapasov nasazhdenii v fitomassu dlya osnovnykh lesoobrazuyushchikh porod Rossii (Coefficients for conversion of timber stock to phytomass for main tree species in Russia), Lesnaya taksatsiya i lesoustroistvo,2003, Issue 1 (32), pp. 119-127.

Kozlov D.N., Puzachenko M.Yu., Fedyaeva M.V., Puzachenko Yu.G., Kartografirovanie zapasov drevostoya eli v biogeotsenozakh yuzhnoi taigi (yuzhnaya chast’ Valdaiskoi vozvyshennosti) na osnove distantsionnoi informatsii Landsat-7 i tsifrovoi modeli rel’efa. (Spruce timber stocks mapping using Landsat-7 and DEM data in the South part of Valdai upland) Aerospace methods and gis–technologies in forestry and forest management: Reports of the IV International Conference, Moscow, 17-19 April 2007, Moscow: Izdatel’stvo Moskovskogo gosudarstvennogo universiteta lesa, 2007, pp. 197-201.

Nagimov V.Z., Rost i nadzemnaya fitomassa drevostoev sosnyaka lishainikovogo v podzone severnoi taigi Tyumenskoi oblasti. Avtoreferat diss. cand. s.-kh. n. (The growth and phytomass of lichen pine forest in northern taiga forest in Tumen region. Extended abstract of candidate’s agriculture sci. thesis), Ekaterinburg: Ural’skii gosudarstvennyi lesotekhnicheskii universitet, 2011, 23 p.

Onuchin I.E., Lesovodstvenno-taksatsionnaya otsenka kedrovykh nasazhdenii na territorii Khanty-Mansiiskogo avtonomnogo okruga – Yugry. Avtoreferat diss. cand. s.-kh. n. (Forestry examination of Siberian stone pine forests in the Khanty-Mansijsk Autonomous District -YGRA. Extended abstract of candidate’s agriculture sci. thesis), Ekaterinburg: Ural’skii gosudarstvennyi lesotekhnicheskii universitet, 2017, 23 p.

Sekerin E.M., Puti povysheniya lesoobrazovatel’noi roli sosny sibirskoi v podzone yuzhnoi taigi Urala. Avtoreferat diss. cand. s.-kh. n. (Ways to raise forest-forming role of Siberian stone pine in the South Ural taiga forests. Extended abstract of candidate’s agriculture sci. thesis) Ekaterinburg, Ural’skii gosudarstvennyi lesotekhnicheskii universitet, 2015, 22 p.

URL:http://ugrales.ru/files/7-Deyat/leshoz/lhr/16032018/Sovetskoe_29_03_2018__18-np.pdf

Sochilova E.N., Ershov D.V., Analiz vozmozhnosti opredeleniya zapasov drevesnykh porod po sputnikovym dannym Landsat ETM (Possibility analysis of stem volume of forests assessment using Landsat ETM data), Sovremennye problemy distantsionnogo zondirovaniya Zemli iz kosmosa, 2012, Vol. 9, No. 3, pp. 277-282.

Tyurin Yu.N., Makarov A.A., Statisticheskii analiz dannykh na komp’yutere (Statistical analysis of data with the computer), Moscow: INFRA-M, 1998, 528 p.

Chistiakov S.P., Sluchainye lesa: Obzor (Ramdom Forests: An overview), Trudy Karel’skogo nauchnogo tsentra RAN, 2013, No. 1, pp. 117-136.

Shvidenko A.Z., Shchepashchenko D.G., Nil’sson S., Bului Yu.I., Tablitsy i modeli khoda rosta i produktivnosti nasazhdenii osnovnykh lesoobrazuyushchikh porod severnoi Evrazii (normativno-spravochnye materialy) (Tables and models of growth and productivity of forests of major forest forming species of Northern Eurasia (standard and reference materials), Moscow: Federal’noe agentstvo lesnogo khozyaistva, Mezhdunarodnyi institut prikladnogo sistemnogo analiza, 2008, 886 p.

Arroyo L.A, Pascual C., Manzanera J. A., Fire models and methods to map fuel types: The role of remote sensing, Forest Ecology and Management, 2008, No. 256, pp. 1239-1252.

Breiman L., Random forests, Machine Learning, 2001, Vol. 45, No.1, pp. 5-32.

Chirici G., Barbati A., Corona P., Marchetti M., Travaglini D., Maselli F., Bertini R., Non-parametric and parametric methods using satellite images for estimating growing stock volume in alpine and Mediterranean forest ecosystems, Remote sensing of Environment, 2008, Vol. 112, Issue 5, pp. 2686-2700.

Fazakas Z., Nilsson M., Olsson H., Regional forest biomass and wood volume estimation using satellite data and ancillary data, Agricultural and Forest Meteorology, 1999, No. 98 (1), pp. 417-425.

Gharun M., Possell M., Jenkins M.E., Poon L.F., Bell T.L., Adams M.A., Improving forest sampling strategies for assessment of fuel reduction burning, Forest Ecology and Management, 2017, Vol. 392, pp. 78-89.

Guyon I., Elisseeff A., An Introduction to Variable and Feature Selection, Journal of Machine Learning Research, 2003, Vol. 3 (1), pp. 1157-1182. 

Hall R.J., Davidson D.P., Peddle D.R., Ground and remote estimation of leaf area index in Rocky Mountain forest stands, Kananaskis, Alberta, The international journal of remote sensing, 2003, No. 29, pp. 411-427.

Hall R.J., Skakun R.S., Arsenault E.J., Case B.S., Modeling forest stand structure attributes using Landsat ETM+ data: Application to mapping of aboveground biomass and stand volume, Forest Ecology and Management, 2006, No. 225, pp. 378-390.

Halme M., Tomppo E., Improving the accuracy of multisource forest inventory estimates by reducing plot location error – a multicriteria approach, Remote sensing of Environment, 2001, No. 78, pp. 321-327.

Hame T., Salli A., Andersson K., Lohi A., A new methodology for the estimation of biomass of coniferdominated boreal forest using NOAA AVHRR data, International Journal of Remote Sensing, 1997, Vol. 18, No. 15, pp. 3211-3243.

Ji L., Wylie B.K., Nossov D.R., Peterson B., Waldrop M.P., McFarland J.W., Rover J., Hollingsworthe T.N., Estimating aboveground biomass in interior Alaska with Landsat data and field measurements, International Journal of Applied Earth Observation and Geoinformation, 2012, Vol. 18, pp. 451-461.

Kuusela K. Poso S., Satellite pictures in the estimation of the growing stock over extensive area, The Photogrammetric Journal of Finland, 1970, Vol. 4, No. 1, pp. 3-9.

Markham B.L., Barker J.L., Landsat MSS and TM post-calibration dynamic ranges, exoatmospheric reflectances and at-satellite temperatures, Landsat Technical Notes, Vol.1, No. 1, 1986, P. 3-8.

Mura M., Bottalico F., Giannetti F., Bertani R., Giannini R., Mancini M., Orlandini S., Travaglinia D., Chirici G., Exploiting the capabilities of the Sentinel-2 multi spectral instrument for predicting growing stock volume in forest ecosystems, The International Journal of Applied Earth Observation and Geoinformation, 2018, No 66, pp. 126-134.

Poso S., Paananen R., Simila M., Forest inventory by compartments using satellite imagery, Silva Fennica, 1987, Vol. 2, No. 1, pp. 69-94.

Redding N.J., Crisp D.J., Tang D., Newsam G.N., An efficient algorithm for Mumford-Shah segmentation and its application to SAR imagery, Proc. Conf. «Digital Image Computing: Techniques & Applications» (DICTA-99), Perth, 1999, pp. 35-41.

Tokola T., The Influence of field sample data location on growing stock volume estimation in Landsat TM-based forest inventory in Eastern Finland, Remote sensing of environment, 2000, Vol. 74, Issue 3, pp. 422-431.

Tokola T., Heikkilä J., Improving satellite imagebased forest inventory by using a priori site quality information, Silva Fennica, 1997, No. 31 (1), pp. 67-78.

Zhang J., Huang S., Hogg E.H., Lieffers V. Qin Y., He F., Estimating spatial variation in Alberta forest biomass from a combination of forest inventory and remote sensing data, Biogeosciences, 2014, No. 11, pp. 2793-2808.

Zheng S., Cao C., Dang Y., Xiang H., Zhao J., Zhang Y., Wang X., Guo H., Retrieval of forest growing stock volume by two different methods using Landsat TM images, The international journal of remote sensing, 2014, No. 35, pp. 29-43.

Reviewer: DSc in engineering, рrofessor Lebedev Yu.V.

© 2019 г.

A.V. Gornov

Center for Forest Ecology and Productivity of the RAS

Profsoyuznaya st. 84/32 bldg. 14, Moscow, 117997, Russia

E-mail: aleksey-gornov@yandex.ru

Received 15 November 2018

Classification of forests of Karelia and the Karelian Isthmus was carried out using the field guide of the forest types of the European Russia developed by L. B. Zaugolnova and V. B. Martynenko. The studied forests were divided into five main sections: lichenous, green moss, sphagnous, grass and grassmarsh. The most common sections were the green moss and the sphagnous one. They include five groups of forest types each. The least diverse is the lichenous section: it consists of only two groups of forest types. In each group of forest types individual forest types were identified. The most common types of forest were bilberry-green moss pine and spruce forests.

Key words: classification of forests, field guide of forest types of the European Russia, Karelia, Karelian Isthmus, northern taiga, middle taiga

One of the first stages of studying forest vegetation is classifying it. A number of approaches to forest classification has now been developed. They tend to be either dominant or floristic (Braslavskaya, Zaugolnova, 2010). The dominant approach is based on the analysis of similarities/differences of forest communities only by predominant species from different layers taking into account a limited number of species. The floristic approach implies the analysis of similarities/differences between forest communities based on the entire species composition of each layer (Braun-Blanquet, 1964; Alexandrova, 1969; Mirkin et al., 1989; Mirkin, Naumova, 1998; Bulokhov, Solomeshch, 2003). Both approaches have their advantages that make them important for the practical use of classifications. For example, the relative simplicity of the diagnostic rules of the dominant approach makes it useful for solving practical problems. The ecological and floristic classification based on the species association can be used to assess the ecological regime of communities. All of the above encouraged some scientists to develop the ecological-coenotic approach to forest classification (Zaugolnova, Morozova, 2006). As a result, L. B. Zaugolnova and V. B. Martynenko created an electronic field guide of the forest types in the European Russia (http://cepl.rssi.ru/bio/forest/index.htm). This field guide allows describing the typological diversity of the forests, according to the practice of domestic forest phytocenology and foreign approaches to plants classification. This helps to set up correspondences between syntaxons in ecological and floristic classification and forest types in the dominant classification. In this regard, the aim of our study was to classify the forests of Karelia and the Karelian Isthmus using the field guide of the forest types of the European Russia.

Materials and methods

The material was collected in the forests of Karelia and the Karelian Isthmus as part of the ICP Forests international program (Lukina et al., 2013). The study area is fully located in the taiga zone. Karelia is divided into two unequal parts – the middle and northern subzones of the taiga. In the middle taiga of Karelia 42 permanent lookout points (PLPs) were established, and in the northern taiga – 58 lookout points (Fig. 1). The Karelian Isthmus was subject to various geobotanical zonings: it belonged entirely to the southern taiga (Yurkovskaya, Elina, 2009) or was divided into southern and middle subzones (Geobotanical…, 1989). Despite the fact that the territory of the Karelian Isthmus is much smaller in area than the middle and northern taiga parts of Karelia, 47 PLPs were established there. The geographical location of the territory to a large extent explains the species and cenotic diversity of forest ecosystems, as well as their dynamics; diversity indicators of are considered separately for the Karelian Isthmus, the middle and northern subzones of the Karelian taiga.

The ground cover study was performed by the forest monitoring method according to the ICP Forests international program (2008). 4 plots of 100 sq. m. each were randomly made within the PLPs. The plots were usually square in shape. If due to local conditions it was impossible to establish a square-shaped site, it was made rectangular. The estimation plots were located as far as possible from each other within the boundaries of the PLP or its buffer zone (Fig. 2) to ensure that the long-term similarity (if any) between the estimation plots occurs not only due to spatial coincidences. Geobotanical descriptions are made for 147 PLPs.

Geobotanical descriptions were made by T. Yu. Braslavskaya (Center for Forest Ecology and Productivity of the RAS), N. V. Genikova (Forest Institute of the Karelian scientific centre of the RAS), E. E. Kostina (Forest Institute of the Karelian scientific center of the RAS) and N.V. Ivanova. Geobotanical description of plots for estimation included a complete floristic list taking into account the layer structure of the vegetation. The moss and lichen layer (D) includes ground-storey bryophytes and lichens; the grass and dwarf shrub layer (C) includes all herbaceous and woody plants of over 0.5 m high (dwarf shrubs, dwarf semishrubs, seedlings and young trees); the layer of shrubs and understorey of trees (B) includes shrubs and young trees of more than 0.5 but less than 5 m high, and all vines climbing up to this height; the tree layer (A) includes woody plants higher than 5 m and all the vines climbing up to this height. In the tree layer as well as in the layer of shrubs and understorey of trees, crown density of all participating species was determined, and in the grass-dwarf shrub and moss-lichen layers the projective cover was found and recorded as the whole number of percent. The main requirement to the description of an estimation plot was the fullest possible identification of the species composition of the vegetation present. Latin names of vascular plants are given according to S. K. Cherepanov (1995).

Classification of vegetation of PLPs. Based on the descriptions of the plots for estimation, each PLP was attributed to a specific forest type. The classification was carried out using the electronic field guide of forest types in the European Russia developed by L. B. Zaugolnova and V. B. Martynenko (http://www.cepl.rssi.ru/cepl.htm). In addition, each PLP was provided with a summary geobotanical description containing the information from the four estimation plots.

The field guide covers slightly and moderately disturbed forests with well-developed ground cover (both grass-dwarf shrub and moss-lichen). The principles of both dominant and floristic classifications of the forest communities are taken into account when compiling the field guide. To describe forests within the framework of the dominant approach, the dominance of both individual species (mainly trees) and ecological-coenotic groups of species (for plants of ground layers) was used. A section that unites the communities by the nature of the ground layers is the largest unit in the typological scheme. A name of a section corresponds to the predominant life form of plants in the ground layers. Sections and subsections are named by traditional Russian geobotanical names. A subsection is an association of forest communities within one section, differing by the structure of the grass-dwarf shrub layer. The set of subsections in its most general form reflects the ecological series of zonal complexes of forest vegetation; they can become the basis for the ecological series of individual landscapes and catchment areas to be formed. The main typological unit is a group of forest types. It is a subdivision of subsection communities depending on the predominant species of the tree layer: each group of forest types is characterized by the predominant species of the upper layer of the forest stand combined with a certain ratio of ecological-coenotic groups of species as part of grass-dwarf shrubs and moss-lichen layers. A group of forest types is a set of smaller units, i.e. forest types. A forest type is the forest community the size of which most corresponds to the association as seen in the dominant approach and it is the smallest unit in this typology.

Evaluation of the floristic diversity. For forests species diversity assessment the Whittaker indicators that have become classical in modern ecology: alpha and beta diversity (Mirkin et al., 1989), were used. Geobotanical descriptions made on the estimation plots of the PLPs served as the material for the analysis of floristic diversity.

Alpha diversity is indicative of the richness of species in particular communities. Key indicators of alpha diversity: species richness is the total number of species in a community, and species density is the average number of species per unit area. Simultaneous recording of species richness and species density results in comparable estimates of species diversity in the analysis of different communities such as forest types.

Beta diversity is indicative of the variability of alpha diversity indicators in space during the transition from one type of forest to another. Beta diversity was estimated using the similarity index (Jaccard index) and the heterogeneity index (Whittaker index). In addition to these indices, vegetation beta diversity can be characterized by the set and number of forest types represented in a particular area.

Figure 1. Location of permanent lookout points (PLPs) on the territory of Karelia and the Karelian Isthmus

Figure 2. Layout examples of plots for estimation in the territory of PLPs

Ecological-coenotic structure. The structure of species diversity was evaluated by the ratio of species belonging to different ecological-coenotic groups (ECGs). According to A. A. Nitsenko (1969) and G. M. Zozulin (1973), ecological-coenotic groups are large groups of ecologically related species, which are in their origin associated with different types of communities. The study used the classification of ecological-coenotic groups of vascular plant species of the European Russia (Eastern European…, 2004) developed on the basis of ecological-coenotic suites by A. A. Nitsenko and historical suites by G. M. Zozulin. Estimation of the ecological-coenotic structure of communities was carried out according to the general list of species found on all plots for estimation of the PLPs. The generalized version of ECGs was used for the analysis of the forest vegetation: nemoral (Nm), boreal (Br), nitrophilous (Nt), pine forest (Pn), meadow and meadow-marginal (Md), wetland (W).

Results and discussion

According to the geobotanical descriptions of the PLPs, the studied forests of the northern and middle taiga can be divided into 5 main sections depending on the general coverage of the moss-lichen layer and the ratio of lichens, green and sphagnum mosses in its composition: lichenous section (cladinosa), green moss section (hylocomiosa), sphagnous section (sphagnosa), grass section (herbosa), grass and marsh section (uliginosoherbosa). Let us have a more detailed look at them.

Green moss section (hylocomiosa)

Communities of the green moss section are most common in the studied forests: the green moss section comprises 70% of all established PLPs. 40% of these forests are forests of the northern taiga, and 60% are forests of the middle taiga. They are mainly various pine forests. The forests of the green moss section are characterized by a closed tree layer and well-developed moss and lichen cover. As a rule, the projective cover of the moss-lichen layer is at least 50% with green mosses dominating therein. Due to the formation of leaf litter, the moss cover can fall to 40% and less, such communities are a transitional form between the green moss and grass sections. The grass-dwarf shrub layer consists mainly of boreal species: Vaccinium myrtillus, V. vitis-idaea, V. uliginosum, Empetrum hermaphroditum, E. nigrum, Linnaea borealis, etc. At the same time, species of the Empetrum genus are most often found in the forests of the northern taiga.

The green moss section includes forests of five groups of forest types: dwarf shrub-green moss birch forests, small grass-green moss spruce forests, dwarf shrub-green moss spruce forests, dwarf shrub-green moss pine forests and small grass-green moss pine forests (Fig. 3). The most common group of forest types is the group of dwarf shrub-green moss pine forests. They occupy about 50% of the PLPs belonging to the green moss section. The second most common group includes dwarf shrub-green moss spruce forests (about 30% of the PLPs of the green moss section). The share of small grass-green moss spruce and pine forests as well as dwarf shrub-green moss birch forests is approximately the same. Let us consider the groups of forest types of the green moss section.

Figure 3. Ratio of PLPs by forest type groups. Green moss section. The X-axis shows groups of forest types (1 – dwarf shrub-green moss pine forests, 2 – dwarf shrub-green moss spruce forests, 3 – small grass-green moss spruce forests, 4 – small grass-green moss pine forests, 5 – dwarf shrub-green moss birch forests), the Y-axis shows the percentage

Dwarf shrub-green moss pine forests are the forests where the tree layer is dominated by Pinus sylvestris (geobotanical description – see Table 1) with Betula pendula, B. pubescens, Picea abies, etc. as common admixtures (Fig. 4).

Figure 4. Dwarf shrub-green moss pine forest

The shrubs layer and the understorey consist of the predominant younger generation of Betula pendula, B. pubescens, Picea abies, Pinus sylvestris trees and accompanying shrub species Juniperus communis, Sorbus aucuparia, Salix caprea, etc. Vaccinium myrtillus and V. vitis-idaea dominate in the grass-dwarf shrub layer both in the northern and middle taiga. Constant species with low coverage: Avenella flexuosa, Linnaea borealis, Lycopodium annotinum, Luzula pilosa, Equisetum sylvaticum, etc. At the same time, dwarf shrub-green moss pine forests of the northern taiga differ from the same forests of the middle taiga in the composition of the grass-dwarf shrub layer. In the northern taiga the coverage and occurrence of some dwarf shrubs are relatively high: Empetrum nigrum, Ledum palustre, and Calluna vulgaris. In the pine forests of the middle taiga there are species that do not occur or are very rare in pine forests of the northern taiga: Convallaria majalis, Dryopteris carthusiana, Fragaria vesca etc.; Calamagrostis arundinacea can be seen rather constantly. In the middle taiga the fern Pteridium aquilinum can also be found which is a marker of frequent ground fires. The moss-lichen layer is well developed. It is dominated by green mosses: Hylocomium splendens, Pleurozium schreberi, species of the genus Dicranum. Polytrichum commune can have low coverage, but is relatively common. The share of lichens in modest.

Within the dwarf shrub-green moss pine forest group of forest types five forest types are identified: bilberry-green moss pine forests, cowberry-green moss pine forests, cowberry-heather-green moss pine forests, bilberry-ledum-green moss pine forests, and heathberry-bilberry-green moss pine forests. Of those, bilberry-green moss pine forests are predominant, with approximately the same share in the northern and middle taiga (Fig. 5). The northern taiga has two types of forests that are not found in the middle taiga: bilberry-ledum-green moss and heathberry-bilberry-green moss pine forests.

Dwarf shrub-green moss pine forests are almost ubiquitous and are very common in the northern and middle taiga. These forests usually result from repetitive fires. If there are no fires for a long time, the forests turn into pine stands with a large share of spruce, and later – into spruce forests with singular pine trees.

Figure 5. The ratio of forest types in the group of dwarf shrub-green moss pine forests. The shaded columns stand for the northern taiga, the unshaded columns – for the middle taiga. The X-axis shows forest types (1 – bilberry-green moss pine forests, 2 – cowberry-green moss pine forests, 3 – cowberry-heather-green moss pine forests, 4 – bilberry-ledum-green moss pine forests, 5 – heathberry-bilberry-green moss pine forests), the Y-axis shows the percentage

Small grass-green moss pine forests were seen only in PLPs located in the forests of the middle taiga (geobotanical description – see Table 2). In this forest types group, the predominant species of the tree layer is Pinus sylvestris. Betula pendula, B. pubescens and others are usually found as admixtures. Sometimes a second substorey is formed by Picea abies and species of the genus Betula. The layer of shrubs and tree understorey consists of the predominant (Sorbus aucuparia, Frangula alnus, Picea abies) and the accompanying species (Padus avium, Juniperus communis); almost no younger generation of Pinus sylvestris is observed. In the grass-dwarf shrub layer Vaccinium myrtillus, Avenella flexuosa, Oxalis acetosella and others are codominants. In some places, Pteridium aquilinum becomes predominant showing the fire origin of such communities. Calamagrostis arundinacea, Dryopteris carthusiana, Fragaria vesca, Trientalis europaea, and Rubus saxatilis show high occurrence but insignificant abundance. Nemoral species are observed: Convallaria majalis, Lathyrus vernus, and Melica nutans. The moss-lichen layer is poorly developed: its coverage depends on the power of the grass-dwarf shrub layer. Pleurozium schreberi, Hylocomium splendens and species of Dicranum genus are often codominants here.

Table 1. Some geobotanical parameters of the dwarf shrub-green moss pine forests

Two forest types are subdivided within the group of dwarf shrub-green moss pine forests: bilberry-oxalis-green moss pine forests (8 PLPs) and hair grass-bracken-green moss pine forests (1 PLP).

Small grass-green moss pine forests grow on flat parts and slopes of lake terraces, on the slopes of the dividing ridge on the outwash sandy loam or sandy moraine.

Table 2. Some geobotanical parameters of the small grass-green moss pine forests

Figure 6. Dwarf shrub-green moss spruce forest

Dwarf shrub-green moss spruce forests (geobotanical description – see Table. 3). Picea abies is dominating in the tree layer with the admixture of Betula pubescens (Fig. 6). The layer of tree understorey and shrubs consists of Picea abies, Betula pubescens, Sorbus aucuparia, Juniperus communis, etc. The grass-dwarf shrub layer of spruce forests of the northern and middle taiga is usually dominated by Vaccinium myrtillus and V. vitis-idaea. Constant species: Avenella flexuosa, Solidago virgaurea, Linnaea borealis, etc. At the same time, the forests of the northern taiga have a relatively high coverage and occurrence of some dwarf shrubs: Empetrum nigrum and Vaccinium uliginosum. The moss-lichen layer is formed by Pleurozium schreberi, Hylocomium splendens, and Dicranum scoparium. Polytrichum commune is a constant species with modest abundance. Microdepressions are home to some species of Sphagnum genus. The share of lichens within this layer is small.

Three types of forests are subdivided within the group of dwarf shrub-green moss spruce forests: bilberry-green moss spruce forests, cowberry-green moss spruce forests and dwarf shrub-green moss spruce forests. Of those, bilberry-green moss spruce forests are predominant, most often to be seen in the middle taiga (Fig. 7).

Table 3. Some geobotanical parameters of the dwarf shrub-green moss spruce forests

Dwarf shrub-green moss spruce forests are virtually ubiquitous and rather common in the northern and middle taiga. Both spruce and pine forests often have post-fire origin.

Small grass-green moss spruce forests (geobotanical description – see Table 4). In the entire network of PLPs located both in the northern and middle taiga this group of forest types was found on ten plots: one in the forests of the northern taiga and nine – in the middle taiga. The tree layer is formed by Picea abies with a small admixture of Betula pubescens, B. pendula, Pinus sylvestris, and less frequently – Alnus incana (Fig. 8). A mixed tree layer of Picea abies and Pinus sylvestris may occasionally be formed. In this case, the structure and composition of lower layers may resemble similar groups of spruce and pine forests, depending on the advancement of secondary (usually post-fire) successions from pine forests to spruce forests. The layer of shrubs and understorey consists of Picea abies, Sorbus aucuparia, Betula pubescens, B. pendula, Alnus incana, etc. Regeneration of Picea abies is associated with microhills and the periphery of the «windows» in the tree canopy. The grass-dwarf shrub layer is codominated by Oxalis acetosella, Vaccinium myrtillus, Avenella flexuosa, sometimes Gymnocarpium dryopteris; Calamagrostis arundinacea, Maianthemum bifolium, Equisetum sylvaticum, and Trientalis europaea can be found with high constancy. Several nemoral species of low abundance are typically present: Convallaria majalis, Stellaria holostea, Lathyrus vernus, Melica nutans, etc. The share of short grasses in the grass and dwarf shrub layer is higher as compared to dwarf shrub-green moss spruce forests. The degree of development of this layer depends on the closeness of tree crowns and the development of the layer of shrubs and the understorey. The moss and lichen layer is characterized by the codominance of Hylocomium splendens, Pleurozium schreberi, and Dicranum species. Microdepressions are home to some species of Sphagnum genus. The share of lichens in the composition of the layer is insignificant.

Figure 7. The ratio of forest types in the group of dwarf shrub-green moss spruce forests. The shaded columns stand for the northern taiga, the unshaded columns – for the middle taiga. The X-axis shows forest types (1 – bilberry-green moss spruce forests, 2– cowberry-green moss spruce forests, 3 – dwarf shrub-green moss spruce forests), the Y-axis shows the percentage

Dwarf shrub-green moss spruce forests are virtually ubiquitous and rather common in the northern and middle taiga. Both spruce and pine forests often are of post-fire origin.

Small grass-green moss spruce forests (geobotanical description – see Table 4). In the entire network of PLPs located both in the northern and middle taiga this group of forest types was found on ten plots: one in the forests of the northern taiga and nine – in the middle taiga. The tree layer is formed by Picea abies with a small admixture of Betula pubescens, B. pendula, Pinus sylvestris, and less frequently – Alnus incana (Fig. 8). A mixed tree layer of Picea abies and Pinus sylvestris may occasionally be formed. In this case, the structure and composition of lower layers may resemble similar groups of spruce and pine forests, depending on the advancement of secondary (usually post-fire) successions from pine forests to spruce forests. The shrubs layer and the understorey consist of Picea abies, Sorbus aucuparia, Betula pubescens, B. pendula, Alnus incana, etc. Regeneration of Picea abies is associated with microhills and the periphery of the «windows» in the tree canopy. The grass-dwarf shrub layer is codominated by Oxalis acetosella, Vaccinium myrtillus, Avenella flexuosa, sometimes Gymnocarpium dryopteris; Calamagrostis arundinacea, Maianthemum bifolium, Equisetum sylvaticum, and Trientalis europaea can be found with high constancy. Several nemoral species of low abundance typically present: Convallaria majalis, Stellaria holostea, Lathyrus vernus, Melica nutans, etc. The share of short grasses is higher in the grass and dwarf shrub layer as compared to dwarf shrub-green moss spruce forests. The degree of development of this layer depends on the closeness of tree crowns and the development of the shrubs layer and the understorey. The moss and lichen layer is characterized by the codominance of Hylocomium splendens, Pleurozium schreberi, and Dicranum species. Microdepressions are home to some species of Sphagnum genus. The share of lichens in the composition of the layer is insignificant.

Figure 8. Small grass-green moss spruce forest

Within the group of small grass-green moss spruce forests three types of forests are identified: oxalis-fern spruce forests, spruce bilberry-small grass-green moss spruce forests, and fern-green moss spruce forests. Of these, bilberry-small grass-green moss spruce forests dominate, being common in the middle taiga (Fig. 9).

Table 4. Some geobotanical parameters of the small grass-green moss spruce forests

Table 5. Some geobotanical parameters of the dwarf shrub-green moss birch forests

Small grass-green moss spruce forests often grow on drained habitats; they can be found on the slopes of moraine hills, on flat sections of dividing ridges, on the smooth slopes of the foothills on both noncalcareous and calcareous moraine loams. This group of forest types has traditionally been regarded as representing zonal vegetation of the middle taiga; these forests, having higher floristic diversity differ from dwarf shrub-green moss spruce forests.

Dwarf shrub-green moss birch forests (geobotanical description – see Table 5). The forest stand is formed by Betula pubescens with a small admixture of B. pendula, Pinus sylvestris and Picea abies (Fig. 10). The layer of understorey of trees and shrubs consists mainly of Picea abies. Other species (Salix caprea, Pinus sylvestris, Sorbus aucuparia, Juniperus communis, etc.) show low abundance. Vaccinium myrtillus, V. uliginosum, V. vitis-idaea dominate in the grass-dwarf shrub layer both in the northern and middle taiga. Linnaea borealis, Lycopodium annotinum, Trientalis europaea, etc. demonstrate high constancy but insignificant abundance. Sometimes Gymnocarpium dryopteris is present. The presence of Empetrum nigrum is typical for the PLPs located in the northern taiga. The moss-lichen layer is dominated by Pleurozium schreberi, Hylocomium splendens and Polytrichum commune. Microdepressions are home to some species of Sphagnum genus. The share of lichens in the composition of the layer is insignificant.

Figure 9. The ratio of forest types in the group of small grass-green moss spruce forests. The shaded columns stand for the northern taiga, the unshaded columns – for the middle taiga. The X-axis shows forest types (1 – bilberry-small grass-green moss spruce forest, 2 – oxalis-fern spruce forest, 3 – small fern-green moss spruce forest),
the Y-axis shows the percentage

Two forest types are subdivided within the group of dwarf shrub-green moss birch forests: dwarf shrub-green moss birch forests and bilberry-green moss birch forests. The first forest type is found in the forests of the northern taiga, and the latter one – in the middle taiga (Fig. 11).

Dwarf shrub-green moss birch forests are usually a derivative post-fire or post-felling version of spruce forests in this section. The ratio of the dominant species of the ground layers and the degree of their development reflect different stages of secondary successions.

Figure 10. Dwarf shrub-green moss birch forest

Figure 11. The ratio of forest types in the group of dwarf shrub-green moss birch forests. The shaded columns stand for the northern taiga, the unshaded columns – for the middle taiga. The X-axis shows forest types (1 – dwarf shrub-green moss birch forests, 2 – bilberry-green moss birch forests), the Y-axis shows the percentage

Figure 12. Alpha diversity of the green moss section forests. The shaded columns stand for the total number of species in the forest type group (species richness); unshaded columns – for the average number of species per PLP (species density). The X-axis shows forest types groups (1 – dwarf shrub-green moss birch forests, 2 – dwarf shrub-green moss spruce forests, 3 – small grass-green moss spruce forests, 4 – dwarf shrub-green moss pine forests, 5 – small grass-green moss pine forests), the Y-axis shows the number of plant species (vascular plants, mosses, lichens)

Table 6. Some parameters of the floristic diversity of the dwarf shrub-green moss pine forests

Floristic diversity of the forests of the green moss section was estimated using alpha and beta diversity.

Evaluation of alpha diversity. Species density in different forest type groups of the green moss section varies within a rather broad range – from 24 to 47 species (Fig. 12). At the same time the majority of PLPs were found to have 20 to 40 plant species (Tables 6–10). The highest floristic diversity is typical for small grass-green moss pine and spruce forests: up to 80 plant species per PLP. Such forests represent more advanced stages of succession (Morozova et al., 2008). Dwarf shrub-green moss pine forests are the poorest in species density, as most often they result from repetitive fires. The number of plant species per PLP in such forests rarely exceeds 20. Species richness depends primarily on the number of PLPs belonging to a certain group of forest types, but generally it correlates with species density.

Evaluation of beta diversity. Beta diversity (Tables 11 and 12) within the green moss section is assessed using Jaccard index and Whittaker index. At pair-wise comparison of forest type groups of the green moss section Jaccard similarity index varies in the range from 0.2 to 0.6. Dwarf shrub-green moss pine and spruce forests demonstrate the biggest similarity (0.6), so do small grass-green moss pine and spruce forests (0.6); the lowest similarity is found between dwarf shrub-green moss birch forests and small grass-green moss spruce forests (0.2). The value of Whittaker’s index (10.2) here is significantly higher than for different groups of forest types within the section (2.7–5.9).

Table 7. Some parameters of the floristic diversity of the small grass-green moss pine forests

Table 8. Some parameters of the floristic diversity of the dwarf shrub-green moss spruce forests

Table 9. Some parameters of the floristic diversity of the small-grass-green moss spruce forests

Ecological-coenotic structure of forests. The results of the floristic list–based calculation of the ecological-coenotic structure of species diversity have shown that the forests of the green moss section have a large number of ecological-coenotic plant groups (Fig. 13). The section is dominated by forest species (boreal, nemoral, nitrophilous and pine-forest) of all forms of life: trees, shrubs, dwarf shrubs and grasses. The share of meadow and forest-edge plants is quite high. This is due to the fact that most of the forests of the green moss section are pine forests with the canopy transmitting enough light for the light-demanding plants of open habitats to exist. The share of nitrophilous species, indicating of an advanced successional status of communities, is low in the studied forests (Zaugolnova, Morozova, 2004).

Table 10. Some parameters of the floristic diversity of the dwarf shrub-green moss birch forests

Table 11. Jaccard index values for the forest type groups of the green moss section

Table 12. Values of beta diversity for the forests of the green moss section

Figure 13. Ecological-coenotic structure of forests of the green moss section. The X-axis shows ecological-coenotic plant groups (Br – boreal, Nm – nemoral, Pn – pine-forest, Md – meadow and meadow-forest edge, Nt – nitrophilous, W – wetland),
the Y-axis shows the percentage

The forests of the green moss section are very common both in the northern and the middle taiga. They are mostly various monodominant pine forests with a relatively high tree canopy density, the predominance of one or two types of dominants in the grass-dwarf shrub layer, and well-developed moss cover of boreal green mosses. In the long-term absence of fires or felling, spruce understorey can enter the tree layer, and pine forests with spruce are replaced by spruce forests with singular pine trees. At the same time, spruce forests quite often demonstrate post-fire origin as well (Volkov, 2008; Zaugolnova, Morozova, 2004). Birch forests usually represent a post-fire or post-felling derivative version of pine and spruce forests of this section, which is proven by the floristic similarity of these forests.

Lichenous section (cladinosa)

Classification and characteristics of forests. Communities of the lichenous section are not common in the studied forests: of all the established PLPs only 8% belong to the lichenous section. The majority of them was found in the forests of the northern taiga. They are different pine forests with a well-developed moss-lichen cover: as a rule, the coverage of the layer is at least 50%. The lichen coverage exceeds or is almost equal to that of mosses.

Only two forest types groups are represented in the lichenous section: lichen pine forests and lichen-green moss pine forests. Their share is equal. Let us look at the groups of forest types of the lichenous section.

Lichen pine forests (geobotanical description – see Table 13) were found only in the forests of the northern taiga. The tree layer is dominated by Pinus sylvestris with a small admixture of Betula pubescens (Fig. 14). The understorey and shrubs is usually sparse. A young generation of trees (Betula pubescens, Picea abies) and shrubs (Juniperus communis, Salix caprea, etc.) is observed here. The grass-dwarf shrub layer is codominated by Calluna vulgaris, Vaccinium vitis-idaea, V. myrtillus, Empetrum nigrum, etc. The moss-lichen layer is characterized by the predominance of lichens: Cladonia arbuscula, C. rangiferina, C. stellaris, etc. Mosses are represented by Pleurozium schreberi, Polytrichum commune, Dicranum polysetum, Dicranum scoparium, etc.

Figure 14. Lichen pine forest

Table 13. Some geobotanical parameters of the lichen pine forests

Three types of forests are subdivided within the group of lichen pine forests: heather-lichen pine forests, heathberry-cowberry-lichen pine forests and cowberry-lichen pine forests, with the domination of the heather-lichen pine forests (Fig. 15).

Figure 15. The ratio of forest types in the group of lichen pine forests. The shaded columns stand for the northern taiga. The X-axis shows forest types (1– heather-lichen pine forests, 2 – heathberry-cowberry-lichen pine forests, 3 – cowberry-lichen pine forests), the Y-axis shows the percentage

Green moss-lichen pine forests (geobotanical description – see Table 14). The tree layer consists of Pinus sylvestris with a small admixture of Betula pubescens and Picea abies. The shrubs layer and the understorey are relatively poorly developed (Fig. 16). It consists of a young generation of trees (Pinus sylvestris, Betula pubescens and Picea abies) and shrubs (Sorbus aucuparia, Salix caprea, Juniperus communis, etc.). The grass-dwarf shrub layer is formed by the predominant dwarf shrubs: Vaccinium myrtillus, V. vitis-idaea, Calluna vulgaris, Ledum palustre, Empetrum nigrum, E. hermaphroditum, etc. There are singular boreal grasses: Equisetum sylvaticum, Carex globularis, Avenella flexuosa, etc. The moss-lichen layer consists of lichens (Cladonia arbuscula, C rangiferina, etc.) and green mosses (Pleurozium schreberi, Hylocomiun splendens, etc.).

Three types of forests are subdivided within the group of green moss-lichen pine forests: green moss-lichen heather pine forests, green moss-lichen bilberry pine forests, and green moss-lichen cowberry pine forests. There are only a few PLPs for each forest type. Predominant forests are green moss-lichen heather pine forests, which are found both in the northern (2 PLPs) and middle (1 PLP) taiga (Fig. 17).

Table 14. Some geobotanical parameters of the green moss-lichen pine forests

Figure 16. Lichen-green moss pine forest

Figure 17. The ratio of forest types in the group of green moss-lichen pine forests. The shaded columns stand for the northern taiga, the unshaded columns – for the middle taiga. The X-axis shows forest types (1 – green moss-lichen heather pine forests, 2 – green moss-lichen cowberry pine forests, 3 – green moss-lichen bilberry pine forests), the Y-axis shows the percentage

Floristic diversity of forests of the lichenous section was evaluated using alpha and beta diversity.

Evaluation of alpha diversity. Species density in different forest type groups of the lichenous section varies within a small range: from 22 to 29 species (Fig. 18). At the same time at the majority of PLPs from 20 to 30 plant species were found (Tables 15 and 16). Lichen pine forests had the lowest values of floristic diversity. For example, in the heather-lichen pine forests, which predominate in this group of forest types, there are only 16 to 19 plant species per PLP. The floristic diversity of the green moss-lichen pine forests is higher than that of the lichen pine forests. For example, in the green moss-lichen cowberry and bilberry pine forests there are up to 36 species per PLP.

Evaluation of beta diversity. Beta diversity (Tables 17 and 18) within the lichenous section is estimated using the value of Jaccard index and Whittaker’s index. When comparing lichen pine forests and green moss-lichen pine forests, the value of Jaccard index of similarity was high (0.7). This means there is a significant floristic similarity of these groups of forest types. Whittaker’s index value (3.3) for the lichenous section is very close to the equivalent parameter for the groups of forest types within a section: 2.4 (lichen pine forests) and 2.5 (green moss-lichen pine forests).

Table 15. Some parameters of the floristic diversity of lichen pine forests