ВІСНИК ЛЬВІВ. УН-ТУ Серія географічна. 2004. Вип. 31. С. 215–222 VISNYK LVIV UNIV Ser.Geogr. 2004. №31. Р. 215–222 УДК 911 DRIVING FORCES OF EVOLUTION OF LANDSCAPE SPATIAL PATTERN A. Khoroshev Moscow State University, Moscow, 119992 Russia akhorosh@orc.ru Landscape science deals with delineation of holistic natural systems with dense interdependencies among components. Degree of adaptation between abiotic components (deposits, topography) and mobile components (plant cover, soil) is assumed to be a criterion of equilibrium or non-equilibrium state of landscape. We analyse reasons for discrepancy between properties of abiotic and mobile components with focus on selfdevelopment of landscape when it gradually becomes less and less dependent on lithogenic controls. We used notion of uncertainty in inter-component relations evaluated by means of discriminant analysis to evaluate degree of landscape equilibrium. The purpose is to evaluate contribution of self-development to evolution of landscape spatial patterns, to reveal driving forces of self-development. Low uncertainty (high concordance) is assessed as equilibrium units that can be considered holistic systems. Interpretation of high uncertainty involves both processes within the patch and properties of environment described by means of space image. The results enable to separate patterns with different degree of determinism between abiotic and mobile components as well as to identify diverging and converging landscape units. Key words: equilibrium, landscape components, relations, self-development, uncertainty Introduction. Identification of holistic landscape systems has always been a task of crucial importance for landscape ecology due to close connection with the problem of forecasting anthropogenically induced changes [12, 9]. Response of landscape to disturbances is expected to be the same in patches with similar interrelations among components (namely, deposits, landforms, water, soil cover, plant cover). East European tradition in landscape research distinguishes so called vertical and horizontal (or morphological) landscape structure [3] related to topological and chorological relations in landscape ecology [16]. The former means a set of interacting landscape components like mentioned above, while the latter is related to a set of landscape units being in interaction by means of lateral flows of matter, energy and information. The notion “horizontal structure” is close to “landscape pattern” in landscape ecology [2]. Landscape components have various temporal scales, so their mutual adaptation to each other requires certain time. Deposits and landforms are treated as inert components as compared with relatively mobile components like soil and plant cover. This holds true for plain regions, while in mountains characteristic time scales can be comparable for plant cover and soil-forming deposits, e.g. in mudflow cones. Equilibrium in relations between landscape components is a topic of great concern for the researcher who is interested in problems of landscape resilience. Getting insight into degree of equilibrium enables us to ________________________ © Khoroshev А., 2004 216 A. Khoroshev predict diversity of possible responses of landscape to disturbances. We assume that mobile components tend to adapt step by step to constraints imposed by inert components, equilibrium relations being the result. In reality this is not more than particular case. Abiotic constraints components can be more or less wide. Under certain abiotic conditions development of soil and plant cover can follow this or that trend depending on occasional events. The result is diversity of mobile components states under similar lithological and geomorphological conditions (so called “lithogenic conditions”) and vice versa. In landscape ecology absolute and relative space are distinguished [6]. Absolute space involves an Euclidean point of view usually based on defined grid system while relative space is defined by functions, processes and rates [7]. In other words, relations between landscape components and landscape units are assumed to be more important than geographical distance from the relativistic point of view. In landscape mapping evaluation of these relations is critical for hierarchy identification as well as for choice of management strategy. In this article we propose to identify landscape patterns based on relationship between inert and mobile components. Holistic landscape structures are formed under the influence of a number of factors. Lithogenic factor is believed to be the most powerful in comparison with hydroclimatic and biotic ones [12]. Conventional landscape mapping is based on few fundamental principles having roots on deterministic approach: a) mobile components reflect properties of lithogenic base on landscape; b) landscape boundaries and boundaries of geological and geomorphological units are in one-to-one correspondence; c) each hierarchical level of landscape organization is controlled by specific principal factor of differentiation [14]. For example higher order landscape units are identified according to landforms and the lower order ones – according to diversity of bedrock. Mobile components a priori are believed to differ in perfect concordance with lithogenic diversity. Different approach is applied by F.N.Milkov [8] who separates landscape units of the same hierarchical order according to genesis. V.B.Sochava opposed identification of low order landscape units according to geomorphological properties [11] and argued that a set of interactions between components should be considered. Fruitful findings by V.N.Solntsev [12] provide an approach to identify landscape structures according to interactions of landscape components. His concept assumes that three types of relatively independent landscape systems emerge at the same hierarchical level being shaped by lateral flows, properties of deposits or solar energy. We make an attempt to delineate landscape units that possess holistic properties due to perfect adaptation of mobile components either to lateral flows or to radial flows. Lateral flows are controlled by degree of relief dissection and slope gradients. Radial flows depend on texture of deposits affecting soil water capacity and permeability. Classification of landscape units by intensity of lateral (say, geomorphological classes) and radial flows (say, lithological classes) is performed in order to discriminate properties of landscape units between lithogenic classes. This procedure provides identification of landscape units in which a set of soils and plant cover properties is specific and differs from all other units in perfect concordance with geomorphological and/or lithological contrasts. In other words, relative space concept is applied: groups of landscapes are composed in such a way that distances between groups are the largest possible both by inert and mobile components. Groups consist of landscape units with equilibrium relations between inert and mobile components since their properties do not intersect. Possible evolution trends cannot be multiple. Resilience in relation to disturbances is high. In parallel, a number of landscape units fail to fall definitely into a certain class sin- DRIVING FORCES OF EVOLUTION OF LANDSCAPE SPATIAL PATTERN 217 ce properties of mobile components (e.g. abundance of plant species) tend to be included into several classes with similar values of posterior probabilities. We regard this result as uncertainty in inter-component relations. Indefinite concordance between inert and mobile components may be caused by few reasons discussed below. However, in general case resilience is lower as compared with equilibrium units since bifurcation is possible after disturbance. The objective of the research is to separate equilibrium units with lithogenically determined vertical structure from the units with uncertain relations between inert and mobile components as well as to reveal driving forces of uncertainty and self-development. Quantitative information on density of intercomponent relations is applied to identification of holistic landscape patterns and self-developing landscape units. Material and methods. Study area is located in the middle taiga region in North European Russia (the Vaga river basin, 6053’ N 4320’ E). The research has been performed since 1994 [1, 4, 5]. Interrelations among landscape components are studied at a fine scale on transect 8050 m long with descriptions regularly spaced over interval 25 m as well as in coarser scale with 500 sample plots being dispersed over the area 200 km 2. Landscape is typical for plains with moraine loam covered by surface sandy deposits. During Holocene epoch landforms were shaped by erosion which followed network of neotectonic joints. Lithogenic diversity is striking due to different thickness of sand cover and moraine carbonate loam lying over Permian marlstone. Exposures of marlstone occur on river valleys slopes. Emergence of alkaline groundwater within joints zones is essential for geochemical diversity: typical acid Podzolic soils are in close neighbourhood with neutral Umbric Albeluvisols and Umbrisols. Plant cover consists of primary spruce forests (Picea obovata) alternating with secondary pine (Pinus silvestris) forests on sandy soils and birch (Betula pendula) and aspen (Populus tremula) forests on clayey soils. Low shrubs layer is typical with dominance of Vaccinium myrtillus, Vaccinium vitis-idaea. Moss layer composition varies in concordance with water supply: sequence of communities with dominance of Sphagnum sp., Polytrichum commune and Pleurozium schreberii is common for catenas. Central sections of watershed areas poorly drained being occupied by oligotrophic mires. Arable lands are located in the area with the most dense river network. Exposures of marlstone on steep slopes favour formation of fertile Umbrisols cultivated intensively. Field material includes description of plant cover (abundance of species, canopy density and coverage for each layer), soils (thickness of horizons, color according to Munsell charts, texture, depth of carbonate horizon), landforms (genesis, shape, slope gradient, aspect). Texture of deposits measured up to the depth of 150 cm with records over the interval 5 cm is used as the control over intensity of radial flows in soils. Landsat 7 space images as well as map of Quaternary deposits and digital elevation model DEM (scale 1:50 000) were used to identify landscape units. ArcView 3.2a software was applied as a tool to calculate characteristics of relief dissection, namely: total length of valleys around each sample plot, variance of elevations, diversity of aspects, distance to the closest stream, slope angle. These indices are believed to characterize intensity of lateral transfer at the site and general drainage conditions. The approach proposed to identify self-developing landscape units and to interpret driving forces includes following procedures. 1) Principal components analysis and multidimensional scaling are performed for sets of data characterizing soils, plant cover, relief and deposits. Factor values obtained are used at the following steps. 218 A. Khoroshev 2) Classification of sample plots according to lithogenic conditions. Two types of classification were performed using k-means method. Classification by texture of soilforming deposits characterizes conditions for radial flows. Classification by topographic variables characterizes drainage conditions. 3) Discriminant analysis is performed in order to assess proportion of patches that differ by soils and plant cover in perfect concordance with classes of topography or deposits identified before. Posterior probabilities are computed for each sample plots. 4) Shannon formula is applied to calculate uncertainty measure for each sample plot: H=-ΣPiLog(Pi), where H – uncertainty measure, Pi – probability that plant cover and soils fall into a certain class (i) of topography or deposits. Uncertainty measure shows degree of concordance in relations between mobile components and lithogenic conditions. 5) Interpretation of uncertainty measures is performed using regression models: H=b0+b1*v1+b2*v2+b3*v1*v2+b4*v1^2+b5*v2^2+… , where v1, v2, … - factor values characterizing landscape properties. The question solved is: which factor is responsible for equilibrium/unequilibrium in inter-component relations. 6) Interpolation of uncertainty measures over the study area allows us to identify areas with perfect adaptation of components and areas with continual transitions between contrast equilibrium units. 7) Estimation of space scale relevant for analysis of inter-component relationship. To solve this question we compared uncertainty measures calculated in relation to topography and deposits calculated for landscape scale with 600 m as radius (H1) and for local scale with 300 m (H2) as radius around each sample plot. If uncertainty in intercomponent relations decreases as larger environment is taken for description of topography (H1-H2>0), broad space scale is more relevant. In case of increase of uncertainty measure for broader scale (H1-H2<0), we conclude that the properties of the certain sample plot needs smaller environment in order to be explained by topography. The concept worked out is aimed at identification of both lithogenically determined and self-developing landscape units based on statistical modelling. Low uncertainty value shows evidence that lithogenic and mobile components are in equilibrium. Landscape units of this type are quite holistic and emerged most likely as a result of divergence within landscape. High uncertainty values are more exciting. Interpretation involves testing several alternative hypotheses. 1. Uncertainty of mobile components in relation to texture of deposits (Ut) is compensated by low uncertainty in relation to drainage conditions (Ud) and vice versa. Regression models Ut=f(Ud) provide tools for testing this hypothesis. 2. Uncertainty on higher hierarchical level can be explained by relations on lower level. To test this alternative data set and grid size are changed, uncertainty measures for few scales are compared. 3. Uncertainty is explained by unfinished recovery succession. Statistical models of linkage between uncertainty value and various properties of soils and plant cover enable to reveal processes responsible for increase of uncertainty. 4. Self-developing redistribution of matter in soils generates emergence of new properties of landscape and new landscape units being relatively independent on abiotic conditions. This phenomenon is known as deterministic uncertainty [10]. The tools include statistical modelling of linkage between uncertainty value and properties of environment measured on space image in moving window, namely diversity (H) and deviation from properties of environment (D): Ut,d=f(H, D). DRIVING FORCES OF EVOLUTION OF LANDSCAPE SPATIAL PATTERN 219 The approach proposed deals with problem of translation of spatial information from one scale to another which is considered to be one of the most important challenges in landscape ecology [15]. Note that uncertainty Ut and Ud is calculated for macroscale, since it is based on 800 sample plots comprising heterogeneous area of 200 km 2. H and D reflect both abiotic and biotic heterogeneity in mesoscale. Thus, the procedure enables to evaluate contribution of mesoscale factors to macroscale uncertainty by means of coefficient of determination in regression equation. Let us give the simplest example. Plant cover in the landscape unit is in poor compliance with deposits, namely alder is not typical for boreal forest on loamy moraine plain. However high Ut value calculated for macroscale can be easily explained if we consider high diversity of close surroundings of the unit due to neighbouring small fen 200 m wide. Penetration of alder into common spruce forest is a result of mesoscale process. If uncertainty is significantly related to mosaic properties of environment, most probable trend of spatial evolution can be revealed: either convergence or divergence of landscape units or stable equilibrium relations between lithogenic and mobile components. To make decision we analyze various combinations of uncertainty measure, diversity (entropy) of environment and uniqueness measure of the patch compared with environment. Taking diversity of environment into consideration we obtain opportunity to evaluate contribution of ecotone factor to uncertainty measure. In case there is no connection between uncertainty measure and properties of space in close surroundings we conclude that uncertainty cannot be interpreted as a result of self-development of spatial pattern. Properties of environment were calculated in two scales with moving windows 500 and 1200 m. Results and discussion. Spatial pattern of the landscape is shaped by a network of joints stretching north-eastward and north-westward. River valleys, streams, bog expanses indicate joints of various rank orders. We identified seven neotectonic blocks based on density and depth of joints and separated by the most pronounced lineaments. Discriminant analysis shows that 70 % of sample plots differ significantly given that they are located within this or that block. Thus, blocks are treated as macroscale holistic landscape systems. Apparently, neotectonic and palaeolandscape events were specific within each block and imposed important constraints over possible trends of landscape structure development. The lower hierarchical level of landscape organization is determined by diversity of conditions for lateral and radial flows. At the macroscale drainage conditions (relief dissection) impose constraints on diversity of components: 60 % of sample plots are discriminated correctly taking herb layer properties as dependent variables, 40-45 % taking trees, low shrubs, moss layers or soils as dependent variables, i.e. as responses to lithogenic contrasts. Total set of mobile components properties is discriminated correctly up to 50 % between drainage classes. Figures for responses to texture classes are similar. Thus, lithogenic factors in macroscale control totally no more than half of variability of soils and plant cover. Topographically determined and lithologically determined holistic landscape systems with perfect one-to-one correspondence between lithogenic classes and mobile components were identified based on posterior probabilities. It turned out that in deeply dissected areas plant cover and soils have much in common despite lithological diversity Holistic structures are determined by intensive lateral flows rather than lithology resulting in convergent self-organization. On the contrary, in poorly dissected areas mobile components respond to texture of deposits and ignore, sometimes, lateral drainage conditions. In case where radial flows have highest intensity like sandy deposits equilibrium landscape units emerge despite this or that pattern of neighbouring river valleys. High 220 A. Khoroshev uncertainty areas in relation both to deposits and relief dominate in flat watershed areas favouring self-development of spatial pattern. One of the possible explanations of high inter-component uncertainty involves scale effect. Finer scale calculation showed decrease of uncertainty for soils and increase of uncertainty for plant cover layers. Scale effect is especially well manifested in opposite hydrological conditions. In swamped areas soils and herbs are sensitive to mesorelief within 600 m as radius being insensitive to microrelief within 300 m as radius. Trees, shrubs and mosses, on the contrary, respond to the closest environment (300 m). In the driest pine-dominated habitats mosses and soils are sensitive to drainage conditions in local (300 m) environment while trees, shrubs and herbs depend on drainage conditions within 600 m. Nature of uncertainty in relations between components in study area turned out to depend on several soil and phytocoenotic processes. Illuviation in soil profile affects equilibrium of soils in relation to deposits. Emergence of carbonate-saturated groundwater distorts equilibrium typical for boreal landscapes. This phenomenon is manifested as humus accumulation in soil which forces plant cover to become insensitive to deposits owing to paradox combination of oligorophic and mesotrophic species. Equilibrium of tree storey in relation to both deposits and topography depends on stage of recovery succession, being lower in climax spruce-dominated associations. Lithogenic contrasts are masked by strong influence of spruce canopy on phytocoenotic and soil processes. Low shrubs, herbs and mosses in climax associations depend on canopy trees rather than on properties of deposits and topography. At initial stages of succession or in second-growth forests plant cover is more diverse and responds to abiotic conditions much better. To reveal self-developing landscape units we related uncertainty measures explained above to properties of spatial pattern in surroundings of sample plots. High diversity of environment can show evidence of ecotone effect as explanation of uncertainty of relations between plant cover and abiotic conditions while homogeneous environment makes us to exclude this factor. Unique patch inside homogeneous matrix may testify either divergence of spatial pattern or gradual disappearance of the patch given that equilibrium is either high or low. We revealed the following variants. 1. Positive linkage between uncertainty and diversity and negative linkage with uniqueness is a result of convergence in mosaic environment. Neighbourhood of small landscape units in unequilibrium state is typical. Ecotone effect is possible. This type of relations is typical for flat watershed surfaces with sandy or two-layered deposits. 2. Positive linkage of uncertainty with both diversity and uniqueness characterises dependence of patch state on distance to landscape boundaries. Equilibrium patches are formed in core sections of uniform landscape units with size about 500 m. The closer to boundary, the higher uncertainty is. At the same time uncertainty is high if the patch is located within larger uniform landscape unit with size about 1200 m. Thus, large uniform landscape unit is treated as a result of self-development because mobile components in core sections cease to obey lithogenic conditions. This is typical for large mire expanses in watershed depressions. 3. If uncertainty is negatively related to both diversity and uniqueness we conclude that the patch emerged either as a result of convergence within large uniform unit or as a result of divergence within mosaic envirionment. Ecotone effect is impossible. This is true for flat watershed areas with sandy deposits. In depressions mosses and low shrubs can also easily adapt to mosaic environment being in equilibrium with deposits even if the patch is small. Then equilibrium is a kind of adaptation to mosaic environment by means of self-isolation. High uncertainty in this variant is a result of self-development, decrease of linkage with lithogenic condi tions. DRIVING FORCES OF EVOLUTION OF LANDSCAPE SPATIAL PATTERN 221 4. Negative linkage between uncertainty and diversity and parallel positive one with uniqueness is revealed for gentle slopes with sandy or loamy deposits. This type of relations testifies start of divergence or degradation of the patch in alien environment, e.g. small forest “island” surrounded by arable lands. Conclusion. The results show evidence that starting conditions of landscape evolution are influenced by block organization of the territory. Inter-component relations have certain degree of freedom for self-development. Approximately half of soils and plant cover variability is determined by topography and deposits. The densest linkages are typical for slope units, the least dense ones with great opportunities for self-development – for humid flat watershed areas. Contribution of local factors to uncertainty of lithogenic-biotic relations is significant in watershed depressions and wide flat and slightly inclined watershed areas. On slopes and narrow watershed areas properties of mobile components are determined mostly by lithogenic conditions. Compact areas with high uncertainty of intercomponent relations are located of flat and wide watershed areas and in mire depressions. Principal factors of uncertainty are identified, namely: emergence of alkaline ground waters, radial redistribution of substance in soils, peat accumulation, interactions among vegetation layers, stage of recovery succession. Combined analysis of uncertainty values and structure of local environment enabled to prove phenomena of convergence and divergence of low-order landscape units as a result of self-development. Self-development capacity is related to size of landscape unit. Homogeneity of environment frequently encourages self-development and increase of independence of mobile components on lithogenic base of landscape. The research was supported by Russian Foundation for Basic Research (grant 0105-64822). ________________________ 1. 2. 3. 4. 5. 6. 7. 8. 9. Dyakonov K.N., Puzachenko Yu.G., Khoroshev A.V., Abramova T.A. Evolution of middle taiga moraine landscape of eastern European type in Holocene // Landscape ecology: Theory and applications for practical purposes. The Problems of landscape ecology, Vol. VI. Warsaw, 2000. Forman R.T.T., Godron M. Landscape ecology. New York: John Wiley and Sons, 1986. Kalesnik S.V. Sovremennoye sostoyaniye ucheniya o landshaftakh (Modern concept in landscape science). – Proceedings of the 3th Congress of All-Union Geographical Society. Leningrad: VGO Publishing House, 1959. – 17 p. (in Russian). Khoroshev A.V. Origin of intralandscape spatial variability at the local level // Landscape ecology: Theory and applications for practical purposes. The Problems of landscape ecology, Vol. VI. Warsaw, 2000. Khoroshev A.V. Linear interrelationship between landscape geocomponents // Publicationes Instituti Geographici Universitatis Tartuensis. Vol. 92. Development of European Landscapes. Conference Proceedings. Vol. I. Tartu, 2001. Marceau D.J. The scale issue in social and natural sciences // Canadian Journal of Remote Sensing. 1999. Vol. 25. – No. 4. Meentemeyer V. Geographical perspectives of space, time, scale // Landscape Ecology. 1989. Vol. 3. No. 3/4. Milkov F.N. Obsheye zemlevedeniye (General Earth science). – Moscow, 1990. Naveh, Z. What is holistic landscape ecology? A conceptual introduction // Landscape and Urban Planning. 2000. 222 A. Khoroshev 10. Phillips J.D. The relative importance of intrinsic and extrinsic factors in pedodiversity // Annals of the Association of American Geographers. 2001. Vol. 91. No 4. 11. Sochava V.B. Vvedeniye v Ucheniye o Geosistemakh (Introduction to Geosystems Science). Nauka, Novosibirsk. (in Russian) 1978. 12. Solntsev, N.A. Prirodniy geograficheskiy landshaft i nekotoriye yego obshchiye zakonomernosti (Natural geographical landscape and some of its general regularities). // Trudy Vtorogo Vsesoyuznogo Geograficheskogo Syezda (Proceedings of the Second All-Union Geographical Congress). Vol 1. Leningrad, 1948. 13. Solntsev V.N. Sistemnaya organizatsiya landshaftov (System organization of landscape). Moscow, 1981. 14. Vidina A.A. Metodicheskiye ukazaniya po polevym krupnomasshtabnym landshaftnym issledovaniyam (Methodics of field fine-scale landscape research). – Moscow, 1962. 15. Wu J., Hobbs R. Key issues and research priorities in landscape ecology: An idiosyncratic synthesis // Landscape Ecology. 2002. Vol. 17. No 4. 16. Zonneveld I.S. The land unit – A fundamental concept in landscape ecology, and its applications // Landscape ecology. 1989. Vol. 3. No 2. РУШІЙНІ СИЛИ ЕВОЛЮЦІЇ ПРОСТОРОВОЇ СТРУКТУРИ ЛАНДШАФТУ А. Хорошев Московський державний університет, Москва, 11992 Росія akhorosh@orc.ru Ландшафтознавство займається виділенням холістичних природних систем із сильними взаємозалежностями між компонентами. Ступінь пристосованості між абіотичними компонентами (геологічні відклади, рельєф) та мобільними компонентами (рослинний покрив, ґрунт) вважають критерієм зрівноваженості ландшафту. Аналізуються причини неузгодженості між абіотичними та мобільними компонентами з наголосом на саморозвиток ландшафту, коли він стає все менш залежним від літогенного чинника. Використано уявлення про невизначеність міжкомпонентних взаємозв’язків, яка оцінюється за допомогою дискримінантного аналізу. Мета полягає в оцінці впливу саморозвитку на еволюцію просторової структури ландшафту для виявлення рушійних сил саморозвитку. Низька невизначеність (висока узгодженість) вказує на рівновагу одиниці, яку можна розглядати як холістичну систему. Інтерпретація високої невизначеності охоплює як процеси всередині одиниці, так і властивості середовища, які описуються просторовим зображенням. Це дає змогу виділяти просторові структури з різним ступенем детермінованості абіотичних і мобільних компонентів, а також визначати ландшафтні одиниці, що дивергують або конвергують. Ключові слова: рівновага, саморозвиток, невизначеність. ландшафтні компоненти, взаємозв’язки, Стаття надійшла до редколегії 28.03.2004 Прийнята до друку 16.06.2004