The Intermediate Report On Project "Land Cover/Land

advertisement
The Final Pre-Launch Report On Project "Land Cover/Land Use Mapping
And Monitoring In Russia" (Convention N 95/CNES/0399- "Programme
VEGETATION")
Milanova E.V. – Principal Investigator
Kalutskova N.N., Kotova T.V., Lioubimtseva E.Yu., Solntsev V.N.,
Tcherkashin P.A., Yanvareva L.F.,
(Moscow State University, Faculty of Geography)
Kazantsev N.N.,
(GIS Center, Institute of Geography, Russian Academy of Sciences)
Anisimova N.V., Kalibernova N.M., Katenina G.D.,
Dr. Kholod S.S., Khramtsov V.N., Dr. Volkova E.A.
(Department of Vegetation Cartography and Geography of Komarov
Botanical Institute, St.-Petersburg, Russia)
Moscow State University
March, 1998
1
Table of Contents
SUMMARY
3
INTRODUCTION
4
OBJECTIVES OF THE INVESTIGATION
5
OVERVIEW OF DATA SOURCES
6
The state of land-cover and vegetation mapping of Russia
6
Thematic Maps
7
Satellite data
8
DESCRIPTION OF THE STUDY AREA
9
METHODOLOGICAL
19
APPROACHES AND TECHNOLOGY OF INVESTIGATION
19
1. Remote sensing for land-cover and landscape mapping
19
Image data set
21
Data processing
23
Simulation techniques and their impact on data quality
24
2. GIS data base as a reference for landscape mapping of European 24
Russia
26
Landscape database for image interpretation
27
Spatial layers in ARC /INFO for landscape mapping of European Russia
28
Landscape size and map resolution
29
Landscapes of European Russia.
29
3. Study on vegetation/land use dynamics of European Russia
Reconstruction of long-term land-cover and land-use trends
34
Analysis of land cover trends of Russia since 1972 based on temporal
series of remotely sensed information
35
4.Interpretation and mapping landscape pattern in forest and forest- 35
steppe zones of Russia using remote sensing
36
Outline of landscape approach
37
Land cover classification
41
Landscape pattern interpretationon the satellite images
Landscape heterogeneity and image resolution
44
Landscape hierarchy and image resolution
49
RESULTS
AND DISCUSSION ON REPRESENTATIVITY DUE TO USE OF
SIMULATED DATA
49
49
1.Remote sensing applications for land-cover and landscape mapping
51
Country (macroregional) level
51
Natural zonal structure
52
Anthropogenic transformation of natural zones structure
52
Regional level
53
Visual analysis of AVHRR data from EROS Data Centre
55
Elaboration of geographical and landscape meaning of determined 56
cluster categories.
57
Correction of published cartographic data on land cover by results of this
58
study
Determination of major tendencies of land use/ cover dynamics
Application of other images
63
Local level - high reslution data from RESURS-F/MK-4 and AFATE-20
64
2. Landscape pattern analysis in forest and forest-steppe zones of
Russia using remote sensing
67
3. Analysis of land cover transformation on EPR in 1970-1992
69
4. First assessment of the specific features of VGT matching the
objectives of investigation, and/or related problems
FUTURE WORK PLANNED FOR THE POST- LAUNCH PERIOD
70
73
PROJECT BIBLIOGRAPHY
REFERENCES
2
SUMMARY
The research project, carried in the framework of VEGETATION Preparatory Programme,
aims assessment and mapping land use, vegetation and landscape cover of the European Part of
Russia using combination of remote sensing and in-field data of different spatial and temporal
resolution as geographical indicators of environmental status and dynamics. Coarse, medium
and high-resolution imagery from several satellite systems (NOAA-AVHRR, RESURS-01-3/MSUSK, RESURS-F/MK-4,) have been processed, interpreted and analysed for compilation of
reference maps in the GIS database on the study area. Complimentary information such as
cartographic and statistical data were incorporated into geographic information system (GIS)
and analysed in order to aid qualitative and quantitative interpretation of satellite imagery.
The resultant database contains multiple information layers, including the remote sensing
data, cartographic ancillary data layers, and statistics. Feasibility study had been undertaken
using the database to elaborate methodology for scale-dependent applications and study of land
cover dynamics under ongoing changes in land use system in Russia.
Analyses of AVHRR and RESURS-01 series allowed to develop algorithms of processing and
interpretation VEGETATION/SPOT-4 data for land cover and land-cover/use change mapping.
Original methodology of landscape interpretation based on spatial relations of land-cover
pattern derived from satellite imagery has been proposed for hierarchical modelling landscape
heterogeneity. The methodology developed during the pre-launch phase and the resultant
database will be used for validation and analyses of VEGETATION temporal series, land-cover
classification and landscape modelling based on VGT and SPOT-4/HRVIR data.
3
INTRODUCTION
In order to understand environmental problems and their solutions scientists and decision
makers must obtain precise and credible data on the background of vegetation, soils and land use.
Detailed information on the present status and trends of evolution of natural and anthropogenic
landscapes is necessary for sustainable land use, rational management of natural resources, and
nature conservation.
All land cover/ land use information for Russia come from various sources that differ in
accuracy, level of detail, time of compilation, etc. Since Russia is going through the process of
agricultural restructuring, spatial and statistical data on land use/cover must be evaluated before it
can be applied to any scientific research. One of the objectives of this research is to collect and
analyse all currently available data (local and regional maps, satellite imagery, detailed groundtruth field observations in key areas) in the framework of a GIS in order to determine its
applicability for land use/cover modeling and change analysis.
Remote sensing data, which have proved a powerful tool of land-cover mapping (Tucker et al.,
1984; Malingreau et al., 1989; Townshend et al. 1992; Defourny et al., 1994), provide a lot of
information not only on spectral properties of different elements of landscape but also on their
organisation in space, composition, configuration, shape, size, connectivity, neighborhoods and
other pattern features of landscape mosaic. A number of landscape pattern characteristics can be
quantified through several rather simple indices and derived from satellite images classified for
land cover.
This research is one of the first attempts of mapping and analyses of land use/cover and
landscape dynamics of Russia using combination of remote sensing and in-field data of different
spatial and temporal resolution. Coarse, medium and high-resolution imagery from several
satellite systems (NOAA-AVHRR, RESURS-01-3/MSU-SK, RESURS-F/MK-4,) have been
processed, interpreted and analysed. Collection, generalisation and analyses of all currently
existing data (local and regional maps, satellite imagery, detailed ground-truthing field
4
observations in key areas) will allow preparation of basic digitised map of land cover and land use
structure of Russia (status of mid-1990s), as well as a set of middle-scale and local-scale thematic
maps (biodiversity, standing biomass, species composition, carbon storage, land degradation,
trends of vegetation and land use changes). Complimentary information such as cartographic and
statistical data were incorporated into geographic information system (GIS) and analysed in order
to aid qualitative and quantitative interpretation of satellite imagery.
Remote sensing information from several Earth observation systems was processed interpreted
in order to update existing cartographic data, to fill information gaps, to identify new areas of
change, and to reveal the degree of vegetation disturbance. RESURS-F high-resolution regional
coverage images were used in combination with highly repetitive coarse resolution data from
NOAA/AVHRR. The later were used to study interannual (phenological) dynamics and long-term
trends of land cover changes. False colour photographic images (made by MK-4 onboard
RESURS F) have proved efficient for detailed land use / land cover classification and definition of
territorial units,and coarse-resolution data validation.
Because landscape aggregations show strong association between land cover and terrain
characteristics, configuration and land-cover content of landscape pattern can be sufficient
indicator for studying internal structure and functioning of landscapes. Evaluation and mapping of
landscape mosaic, its heterogeneity and patches are necessary for understanding landscape
functions and ecological processes.
OBJECTIVES
OF
THE
INVESTIGATION
(MILANOVA
E.V., TCHERKASHIN
P.A.
AND
LIOUBIMTSEVA E.YU.)
The main objective of this research is development methods and algorithms of application
remote sensing data, and more specifically VEGETATION/SPOT-4 data, for land cover, landcover change, vegetation, land use, and landscape mapping and GIS modelling throughout the
European Part of Russia.
The second specific objective is to study feasibility of remote sensing applications for
operational landscape mapping at different scales. This would allow us to evaluate efficiency of
5
VGT/HRVIR product for multiscale hierarchical landscape modelling.
In addition to these main objectives several methodological, thematic, and practical goals
should be considered.
In order to systemize ancillary data for image interpretation and simulation VGT/SPOT data
various cartographic thematic information was combined with data from Russian satellites
(RESURS-F, RESURS-01) and NOAA/AVHRR in the GIS environment. This phase of research
was necessary to address several thematic and scaling problems, such as:

How do spatial contours agree/disagree on traditional hand-drawn maps and maps
produced by automatic interpretation of remotely sensed images from different Earth
observation systems? and

What is the physical/ geographic meaning of disagreements in the contours of maps
produces by different techniques?
The thematic goals of the investigation address the present status and trends of land cover in
the European part of Russia. Retrospective analysis of land cover change since the last century
and especially the 2 last decades (while remote sensing data become available) is important for
modelling amplitudes of land-cover change on the future temporal series of the earth
observations.
As a baseline to studying and monitoring of Russia’s environment it was essential to update
currently available materials and to compile a GIS spatial database, embracing various data on
vegetation cover and land use. The final product of this study is to help in better understanding
the ways and mechanisms of the vegetation cover evolution and dynamics. It also has to provide
some new ideas on appropriate structure of land use and management of forest, agricultural and
pastoral resources of Russia.
There are several tasks of this study:
1 Assessment and mapping of landscapes of Russia using coarse, medium and high
resolution satellite imagery from several satellite systems (NOAA AVHRR, RESURS/MSUSK, RESURS-F/MK-4) for development of procedure and algorithms of analyses landscape
6
status, change and dynamics.
Currently being developed procedure of data processing from several different sensors
would allow establishing and refining the methods of processing VEGETATIONand
SPOT/HRVIR data.
1 Development a system of landscape survey based on application of remote sensing data,
which facilitate qualitative description and quantitative analysis of landscape units of research
area.
2 Evaluation the possibility of applying pre-processed data available from global archive
for regional-level landscape study and assess reliability of data available from the global 1-km
AVHRR data archive available for distribution by the United States Geological Survey, EROS
Data Center and the European Space Agency under the guidance of the International
Geosphere-Biosphere Programme.
OVERVIEW OF EXISTING DATA (ANISIMOVA N.V., KALIBERNOVA N.M.,
E.YU.)
The state of land-cover and vegetation mapping of Russia.
AND
LIOUBIMTSEVA
The vegetation of European part of Russia is studied rather well in cartographic and botanicalgeographic aspects. The detail of showing vegetation variety at the maps has increased in tens
times from the end of XIX century at the first vegetation maps created by Korzhinskiy and
Tanfiliev till the latest "Vegetation map of European part of the USSR" in 1: 2 500 000 scale
edited by T.I Isachenko and E.M Lavrenko in 1979.
This map is the most significant one, because it is a result of botanical-geographic
investigations of this region to 70th. It is also the most detailed and informative map. The
previous "Vegetation map of European part of the USSR" created by Iljinskiy in 1937 and
published in the Atlas of the World contains only 39 subdivisions. The following map of this
region edited by E. Lavrenko and V. Sochava in 1948 shows 71 mapped units. The legend of the
latest map (1979) contains 248 units so its content increases in 3-4 times. Many letter symbols
and supplementary signs are used for showing regional peculiarities of the vegetation at this map.
7
So more than 500 subdivisions of vegetation cover are shown for the area of European part of the
former USSR.
It is of great value that not only primary types of vegetation but also the secondary ones are
presented at the map. All secondary types are co-ordinated with the native ones. The map was
supplemented by explanatory text (Gribova, Isachenko and Lavrenko,1980) were some analytic
maps of distribution of the main vegetation types are presented : tundras, boreal dark-coniferous
forests, pine forests, nemoral bread-leafed forests, steppes, xerophytic open woodlands, mountain
xerophytic vegetation, deserts, flood-plain vegetation, paludal vegetation.
On the base of this map the educational vegetation map in 1: 2 000 000 scale was published in
1987.
In connection with intensive development of non-chernozem zone of Russia the vegetation
map of this area (1976) and the map of vegetation protection (1977) were compiled in 1: 1 500
000 scale. The map of vegetation regionalization of non-chernozem zone was also published in
1989.
Among the newest survey maps the "Vegetation map of the USSR" (1990) in 1:4 000 000
scale (for higher school) is of great interest. Many scientists from Komarov Botanical Institute,
Siberian Institute of Geography (Irkutsk), Laboratory of complex cartography of Moscow State
University were carried out the content of this map. It synthesizes the data on vegetation cover of
the former USSR during the last 25-30 years. It presents geographical and typological
peculiarities of the vegetation of large area (including the European part). The legend of the map
is made in two forms: textual and tabular. Some new data on plant communities distribution,
zonal boundaries, anthropogenic dynamics, vegetation structure as well as new cartographic
methods make this map very informative.
The latest achievement of geobotanical mapping is the "Map of reconstructed vegetation cover
of Central and Eastern Europe (1996) compiled by the large group of scientists of the former
USSR and Eastern European countries, edited by S.A. Gribova and R.R. Neuhausl. Such criteria
of vegetation cover as floristic composition, typological, ecological, geographical peculiarities,
structure and dynamics are analyzed for the most part of Europe.
8
The largest subdivisions in the legend correspond to zonal types of vegetation cover (tundra,
coniferous forests, steppes, deserts, etc.) but their typological rank is various. Some of them are
vegetation types (tundra, steppe, desert), the others are classes of formations (mesophytic
coniferous forests, mesophytic summer-green broad-leaved forests).
The next subdivisions of the legend reflect subzonal and altitudinal features of vegetation
units. The following differentiation is based on ecological relations of plant communities,
especially edaphic ones. Psammo-, petro-, galo- and gigrophytic variants of zonal vegetation
types are presented for tundra, steppe and desert vegetation.
One of the most important criterion is geographical one, which shows provincial (longitudinal)
peculiarities of plant communities. For every mapped category botanical-geographic type is
determined, so geographical position of every unit is marked.
The main mapped units for homogeneous vegetation cover are typological ones (associations,
their groups or classes, according to Russian geobotanical school). Dominant, subdominant and
differential species are used for botanical characteristic of typological categories. Types of
territorial units (complexes, series, ecological-dynamic ranges, etc.) are working out for
heterogeneous vegetation cover.
Middle-scale vegetation maps were also compiled for some regions of the European part of
Russia. Some of them correspond to the lists of topographical maps in 1 : 1 000 000 scale: O-36,
O-37 (Isachenko & Lavrenko, 1975), O-39 (Karpenko & Lavrenko, 1975) - for the central part of
Russia; Kansko-Timanskaya and Malozemelskaya tundra (Gribova, Lavrenko, 1975) - for the
north part of Russia. There are also some vegetation maps for administrative regions: Rostov
region in 1: 600 000 scale (Gorbachev, 1973),
Novgorod region (Isachenko, 1982), Pskov region (Karpenko & Shabalina, 1969), Moscow
region (Isachenko, 1964), Leningrad region (Isachenko & Katenina, 1967) and others, created for
regional atlases.
Among the latest middle-scale maps "Vegetation map of Moscow region" in 1: 200 000 scale
(Ogureeva, 1996) should be mentioned. This map shows typological diversity of plant
9
communities, its relation with landscape structure of the area and spatial regularities of
distribution as well as vegetation dynamics under anthropogenic impact. The map allows to
define the modern state of the vegetation of Moscow region, to estimate degree of its disturbance
for the purposes of nature economy and nature protection.
Thematic Maps
Most of recently published thematic maps of landscape components of Russia and the former
USSR together with other materials (such as topographic maps, satellite imagery, statistical and
remote sensing data) served as references and sources of information in spatial data bases
assembled in this study. However, the legends and classificatory principles of vegetation, soil and
landscape maps of Russia produced by different institutions at different time in different scales
are poorly compatible, that leads to certain contradictions between the maps. Some of these data
suffer from a certain lack of accuracy and need an update. Finally, almost none of existing maps
have a digital version, which makes correction and update of information extremely difficult.
After preliminary analyses of relevant bibliography and cartographic materials on the study
area, as well as personal communications with experts - authors and editors of above mentioned
maps, we selected the following maps as principal sources of basic information on the present
day landscapes of European Russia (Table 1):
Table 1 Thematic maps used in the research
TITLE
SCALE
YEAR
Vegetation of the USSR
1:4M
1990
Land Use of the USSR
1:4M
1991
Vegetation of the former USSR
1:8M
1993
Phytomass ( former USSR)
1:8M
1993
Mortmass (former USSR)
1:8M
1993
10
Annual production (former USSR)
1:8M
1993
Soils of the USSR
1:4M
1995
Ecological map of Moscow oblast.
1:350 K
1993
Ecological- geographical map of Russia,4 sheets
1: 4 M
1996
Land cover of USSR, 4 sheets
1:4 M
1988
1:2.5 M
1980
Landscape map of USSR 4 sheets.
1:4 M
1988
Geological map of RSFSR
1:2 M
1988
Vegetation map of Kanino-Timanskaya and
1 : 1M
1975
Map of Landscapes of the USSR
Malozemelskaya tundra
Vegetation map of Moscow region // Atlas of 1 : 1 500 K
1964
Moscow region.
Vegetation map of Leningrad region // Atlas of
1 : 1 500 K
1967
1:500 K
1984
Leningrad region
Map of potential erosion of Non- Chernozem
11
zone of RSFSR (excluding Ural and Zauralie)
Map of engineer- geological conditions of Non-
1:1.5 M
1984
1:1.5 M
1984
1:1.5 M
1984
1:1.5 M
1984
1:1.5 M
1984
Map of vegetation of Moscow oblast
1:200 K
1996
Map of soil-geographic regionalisation of Non-
1:1.5 M
1984
Chernozem zone of RSFSR (excluding Ural and
Zauralie)
Map of pit areas of Non- Chernozem zone of
RSFSR (excluding Ural and Zauralie)
Map
of
permafrost
conditions
of
Non-
Chernozem zone of RSFSR (excluding Ural and
Zauralie)
Map of agricultural land use of Non- Chernozem
zone of RSFSR (excluding Ural and Zauralie)
Map of vegetation protection
of
Non-
Chernozem zone of RSFSR (excluding Ural and
Zauralie). Ed. E.M.Sergeeva.
Chernozem zone of RSFSR (excluding Ural and
12
Zauralie).
Map
of
fresh
underground
water
storage
1.5 M
1984
1:500 K
1969
1:350 K
1993
1 : 1 500 K
1964
1 : 600 K
1973
conditions of Non- Chernozem zone of RSFSR
(excluding Ural and Zauralie)
Vegetation map of Pskov region // Atlas of Pskov
region
Moscow oblast. Pollution of environment, soils
Vegetation map of Yaroslavl region. // Atlas of
Yaroslavl region
Vegetation map of Rostov region
Climatic data were taken from two sources: Global Climatic Data base by IMAGE Project
(Leemans and Kramer, 1994) and Climatic Atlas of the USSR (1987).
Compilation of reference maps and updating currently existing data, required also processing
and consulting coarse and medium-resolution satellite imagery (NOAA and RESURS) over the
EPR (years 1992-93) in order to evaluate maps' reliability, to fill information gaps, and to
evaluate visually the present-day landscape pattern of the area.
Besides broad-scale thematic maps, medium and large-scale topographic maps were analyzed
in order to refine broad-scale map contours, especially in cases of contradictions between
13
different data sources (e.g. contradictory contours on Vegetation and Land Use reference maps).
More than a hundred of map sheets of 1:200 K and dozens of 1:500 K sheets were consulted in
the course of this work.
Satellite data
Three types of remote sensing data were analysed in order to test their suitability to detect
landscape pattern at different scales (table 2).
Table 2 Satellite data used in the research
N Satellite
N
Instrument Coverage
1
AVHRR
Lon Acq. date Resol. spectral
used
bands
3500 x 2000 55
km
N
43
E
21.0531.05.92
1 km
3500 x 2000 55
km
N
43
E
21-30.
06.92
1 km
3500 x 2000 55
km
N
43
E
110.05.92
1 km
RESURS- MSU-SK
01-3
600 x 600 57
km
24’
N
42
48’
E
08.10.93
160
m
0.5-0.6,0.6-0.7
RESURS- MSU-SK
01-3
600 x 600 37
km
87’
N
58
13’
E
09.06.92
160
m
0.5-0.6,0.6-0.7,
6
RESURS- MK-4
F
60 x 80 km 37
12
N
55
88’
E
08.06.91
25 m
0.460-0.505,
0.515-0.565,
0.635-0.690 m
7
RESURS- MK-4
F
60 x 80 km 38
04’
N
57
96’
E
23.05.93
45 m
0.460-0.505,
0.515-0.565,
0.635-0.690m
8
RESURS- MK-4
F
60 x 80 km 39
07’
N
51
76’
E
04.08.90
35 m
0.460-0.505,
0.515-0.565,
0.635-0.690m
2
3
4
4
NOAA
Lat
NOAA
NOAA
AVHRR
AVHRR
0.58-0.68
0.725-1.1 m
0.58-0.68
0.725-1.1 m
0.58-0.68
0.725-1.1 m
0.7-0.8,
1.1m
0.8-
0.7-0.8,0.81.1m
14
High-resolution photographic images acquired by MK-4 camera on board of Russian satellite
RESURS-F (also widely known as COSMOS series) were used for large-scale landscape
mapping.
MK-4 is a four-channel photographic system comprising four cameras (f = 300 mm and frame
format 18 x 18 cm), so that from the 200-km orbit it captures 120 x 120 km area in 1:600 000
scale. Photo cameras work in four automatically selected from six possible spectral bands (0.4600.505, 0.515-0.565, 0.635-0.690, 0.810-0.900, 0.400-0.700, 0.580-0.890 m). Image of Moscow
(acquired on 08.06.91 at the altitude 237 km) was available with spatial resolution of 25 m.
Darwin (04.08.90) and Voronej (23.05.93) scenes, acquired on 281 km and 245 km altitude
respectively have 45 and 35-m pixel resolution approximately.
Four additional scenes of MK-4 within forest and steppe zones of European Russia were
analysed for validation of medium and coarse-resolution data.
Medium-resolution data from MSU-SK/RESURS-01 instrument were interpreted together
with MK-4 imagery for all test sites in order to evaluate and compare landscape aggregation and
structure at different scales.
RESURS-01 are Russian resource observation capturing systems, which are launched to sunsynchronous orbit at 650 m altitude, inclination 98° and period of rotation about 98 min.
Currently working RESURS 01- 3 (previously used name -COSMOS 1939) carries MSU-SK
wide-swath, medium resolution instrument with a conical scan. It has five spectral channels: 0.54
- 0.60 km, 0.60-0.72 km, 0.72-0.82 km, 0.81-1.00 km and 5 channel - 10.30-11.75 km.
Width of observation zone is 600 km, and on land resolution for 1-4 channels is 150 - 160 m. An
MSU-SK image acquired on 08.10.93 covering all test sites (center 57°24 N, 42°48 E) was
analysed only in red and infra-red channels. Since healthy green vegetation generally reflects 40
to 50% of the incident near-infrared energy (0.7 to 1.1 micrometers (µm), and absorb
approximately 80 to 90% of the incident energy in the visible (0.4 to 0.7 µm), particularly in the
red (0.6 to 0.7 µm) part of the spectrum, ratios of MSU-SK bands 4 to 2 was found the most
effective for land-cover mapping in forest and forest-steppe zones.
NOAA/AVHRR 10-day composite scenes were processed only in two spectral bands (red
15
(0.58-0.68 m) and near-infrared (0.725-1.1 m) channels). NOAA data were used for two
purposes: to evaluate the lost of information on landscape structure with decrease of data
resolution and to test the possibility of tracing landscape frame at the macroregional level.
Landscapes of the three test areas of about 60 x 80 km were analysed for pattern interpretation
at the microregional scale based on high, medium and coarse resolution images, topographic and
field data. The area of 500 X 600 km, which includes all of them, was studied in order to analyse
possibilities of mesoregional mapping. Interpretation of landscape pattern based on coarseresolution data was undertaken for the whole of European Russia in order to test the possibility of
landscape mapping at the meso- and macroregional scale.
DESCRIPTION OF THE STUDY AREA (SOLNTSEV V.N. AND KALUTSKOVA N.N.)
A polygon between 33-42 eastern longitude and 53-60 northern latitude was chosen as study
area on regional level. This is a central part of Russian plain with the center in Moscow (Fig. 1).
The following features defined selection of this territory as key study area for the
investigation:
1. Consequential diversity of types of land cover and their territorial combinations;
2. Good coverage (comparing with other parts of Russia) with traditional and remotely
sensed data;
3. Huge amount of attribute and spatial data from traditional sources (topographic and
industrial maps, government fund materials, fieldwork results, etc.).
This gives us an opportunity to investigate the characteristics of land cover through remote
sensing and check the accuracy and efficiency of the results with a set of independent data
sources.
Russian plain is known for "classic" picture of latitude zonality of vegetation, soils and
landscapes. Up to 16 consecutive sub-latitude stripes (Isachenko, 1985). The following natural
zones that the study polygon covers determine large diversity of landscapes (from north to south):
1. South-taiga landscape zone with dark coniferous forests and thin forests on wetlands
16
on podzol and turf soils
2. Semi-taiga landscape zone (mixtures of coniferous with broad-leaf and small-leaf
forests) on podzol and carbonated soils
3. Broad-leaf forested landscapes (oak-lime and pine forests)
4. Forested Steppe landscapes (combinations of grasslands with areas of oak and oakpine forests)
5. Northern Steppe landscapes (steppe grasslands in combination with forested ravines
and valleys on original chernozem soils)
Since amount of precipitation in semi-continental climate of the polygon is approximately the
same (450-650 mm) the change of landscape zones from north to south is caused by solar
radiation supply. Due to its change moisturising conditions shift from superfluous in the north to
optimal in the center of the polygon and insufficient in the south.
Latitude zonal variances in natural conditions are overlaid with provincial conditions, related
to diversity in geological foundation. Upper and lower plains, formed on easy banks of
sedimentary covering of East European geological platform, alternate within the region. Plains
have different genesis - from glacier (moraine) to erosion-accumulative. Diversity of deposits
settle variety of soil textures (sand, clay, loess, etc.) and drainage conditions, that in some cases
provoke waterlog. Diversity of deposits also produce variety in soil fertility.
Variance in landscapes of the territory is even more complex due to active human interaction.
The center of Russian plain had been long ago domesticated. Now it’s a region with highest in
Russia population density, essential industrial community and highly developed transport
infrastructure. Anthropogenic development of natural resources of the territory had even changed
the picture of landscape zones. The less changed zone is south taiga due to low cultivation rate
(10-30%). Changes in sub-taiga zone, where capital region is, are tremendous: now it’s a
combination of large urban areas with arable lands and garden-plots and isolated areas of forests.
Zone of broad-leaf forests is now looking the same as forested steppe due to high cultivation (up
to 90% of total area). Almost all the forests had been destroyed in northern steppe zone.
17
Anthropogenic changes of regional structure are also very important. They are related to large
industrial installations, urban areas or past specialisation of certain parts of the area. On the study
polygon there are part of Volga-Baltic navigable system (Sheksnink and Rybinsk water
reservoirs), Moscow water supply system, including Moscow-Volga channel and several water
reservoirs on Volga river. Landscapes are significantly changed also near Moscow, where open
excavations of coal and turf is taking place. Several administrative units (Vladimir, Kasimov and
Suhinich) with most fertile lands are almost entirely ploughed up.
Another important feature of the study area is presence of protected areas. The biggest include
Darvin, Okskiy, Prioksko-Terrasniy and Central-Chernozem conservations.
The final reason, which makes this territory most interesting as a polygon for testing various
remote-sensing techniques, is high mobility of the economic processes that take place here during
the last decade caused by transition in political and economic directions. Changes in ownership
and industrial specialisation, stagnation in agriculture, development on modern lines of transport
infrastructure - all these reasons cause complex natural-anthropogenic transformations of the
landscapes.
Five test areas with distinctively different land-cover composition and configuration were
selected in European Part of Russia (fig. 1):
Darwin nature reserve. The test area belongs to the Upper Volga province of Russian plain
and is featured by low accumulative relief (absolute altitudes 120-150m) and cool humid
temperate climate. The area is dominated by spruce and spruce-hardwood forests (Picea albies x
P.obovata) with high participation of birch and pine forests (Betula czerepanovii, Pinus silvestris)
and meadows and peat bogs. Although configuration and dynamics of the present-day landscapes
are partly defined by creation of artificial Rybinsk water reservoir, a western part of the test area
belongs to Darwin reserve with quasi natural landscapes, while the rest of it experience
agricultural and forestry impact restricted to coastal zone and river valleys.
Moscow region belongs to Smolensko-Moscovskaya province of the Russian plain and
featured by hilly moraine relief with relatively high altitudes (220-310m), temperate climate and
rather contrast potential landscapes. In course of long anthropogenous impact primary landscapes
18
of coniferous and hardwood forests on sod-podzolic soils have been totally replaced by secondary
birch and fur-burch forest. The major parts of the test area is occupied by urban and rural lands
with very dense built-up areas and infrastructure and by arable lands, improved pastures, hay
meadows and gardens.
Kulikovo Pole site is featured with hilly plain landscape with forest-steppe and steppe
vegetation cover heavily transformed by agriculture. Most of the area is represented by arable
and pastoral lands with small islets of oak forests. Gully relief results from widely spread soil
erosion.
Eastern coast of the Gulf of Finland – coastal plain covered with pine dwarfshrub-moss
forests; mixed pine-birch-spruce moss forests with near 80% forests are transformed by fires and
forests felling; 20% of the area is occupied by secondary small-leaved forests; present
agricultural lands occupy small areas (mainly hay-meadows); selective felling is everywhere.
Voronej region is situated in the Oka-Don province. The test area is featured by flat relief on
the forested eastern bank of the Voronej River (absolute altitudes 150-160 m) and typical gully
relief, dendrite in plan throughout the rest of the area. Subhumid temperate climate with often
draughts caused development of forest-steppe and steppe potential landscapes with hardwood
forests on grey forest soils and meadow and typical chernozem steppe. However, almost the
whole area is occupied nowadays by agricultural lands with exception of the Voronej nature
reserve with oak-lime forest. Gully valleys are occupied by bairak hardwood forests and shrubs.
Agricultural lands are often protected by antierosional shelter belts.
METHODOLOGICAL APPROACH AND TECHNOLOGY OF INVESTIGATION
1. Remote sensing for land cover and landscape mapping (Tcherkashin P.A. and Milanova
E.V.)
Data set
We found that the most sufficient source of NOAA-AVHRR NDVI images for continentalscale feasibility study was Global Ecosystems Database (GED), Version 1.0 (on CD-ROM) by
19
EPA Global Climate Research Program, NOAA/NGDC Global Change Database Program
(Kineman and Ohrenschall, 1992). Monthly Global Vegetation Index (GVI) from Gallo Biweekly
Experimental Calibrated Global Vegetation Index (April 1985 - December 1990) with 10’
resolution was used for this study (principal investigator: Kevin P. Gallo, USGS EROS Data
Center and the NOAA National Environmental Satellite, Data, and Information Service). This
data set contain Normalised Difference Vegetation Index (NDVI) from AVHRR sensor on
NOAA-9 and NOAA-11 satellites (Kidwell, ed., 1990).
For this particular research the total of 60 images over the period 1986-1990 with a 10 -km
resolution were obtained from the database. The territory of the whole former Soviet Union was
chosen as an object of study (40°-80° N, 20°-180° E) representing high variability of land cover
on continental scale and is familiar to the authors.
Two principle sources of satellite imagery were used for VEGETATION data simulation on
regional level over Central Russia: AVHRR images of 1-km resolution over the whole area of
study and photographic high and medium resolution images acquired by KATE-1000, MK-4 and
MKF-6 instruments on board of Russian satellites of RESURS-F1 series and orbital stations
Soyuz and MIR. Complimentary ancillary data incorporated into geographical information
system were analysed in order to aid image interpretation.
The Advances Very High Resolution Radiometer (AVHRR) 10-day composite data were
visually analysed and most representative were selected for the area of study in 1992 to study
seasonal and phenological dynamics of land cover. These data were the principle source of
information on surface biophysical parameters, as well as on thematic surface characteristics,
such as land cover and land use.
These data were obtained from the global data base of EROS Data center where geometric and
radiometric composition was done according the IGBP requirements (Eidenshink & Faundeen,
1994). Internet WWW service was used to obtain the data.
Four test sites were selected in different land-cover types: middle taiga, southern taiga, mixed
forest, and forest-steppe. Throughout the selected areas high resolution data were processed.
They comprise: black-and-white aerial photographs in 1 : 25 000 scale were acquired in July
20
1989 by AFATE-20 apparatus (altitude of 5050 m) over the Karelian test site (middle taiga) and
7 multispectral photographic images acquired in 1989-1992 by MK-4 and MKF-6 cameras on
board of RESURS-F1. The later cover Mozhaisk site (mixed forest), Kulikovo Pole (foreststeppe) and Darwin reserve (southern taiga).
Complimentary thematic data (cartographic and statistical) were incorporated into Geographic
Information System (GIS) and analysed in order to aid interpretation of remote sensing data.
Data processing
Landscape units have relatively homogenous characteristics of topography, soil, vegetation
and climate within a region. Because landscape type boundaries are often clearly represented on
remotely sensed images, they can be easily detected, interpreted and sketched for land cover
mapping. Compilation of GIS based reference maps requires additional data to evaluate data
reliability, to fill information gaps, to identify new areas of change, and to reveal the degree of
vegetation disturbance. RESURS-F/MK-4 and RESURS -01/MSU-SK high and medium
resolution regional coverage images are used in combination with highly repetitive coarse
resolution data from NOAA. The later prove to be extremely helpful for study of interannual
(phenological) dynamics and long-term trends of land cover changes.
Particular attention is given to analysis of vegetation indices. In the study the basis indicator to
assess land cover status is NDVI (Normalised Difference Vegetation Index) which can be defined
as a difference of channels 1 and 2 of AVHRR.
Multi-temporal NDVI is a useful tool for monitoring the dynamics of natural ecosystems on a
regional or national basis (Eidenshink and Haas, 1992). Single data analyses, especially using
AVHRR data, are frequently inadequate for discriminating land cover types, because disparate
cover types can share spectral characteristics (Loveland et al., 1991). One-kilometer AVHRR
data have been used less often than GAC (Global Area Coverage) or GVI (Global Vegetation
Index) data because they have not been generally available.
The following computer techniques are currently found to be most efficient for different types
of land categorisation based on AVHRR data: classification (supervised and unsupervised) and
21
principal component analysis (PCA). Both supervised and unsupervised classification methods
are based upon statistical parameters such as mean and standard deviation.
Unsupervised classification uses a composite image from 3 bands of data to create image
category groups. No knowledge of the area is required to create an unsupervised classification,
however ground information, such as vegetation, land use, and topographic maps or local
knowledge, is required to match created groups with real land use/land cover classes.
Unsupervised classification requires three steps - the creation of the composite image, the
classification of the composite image and the interpretation of the classification results.
Moreover, high and very high-resolution satellite data (e.g. images by MK-4 onboard
RESURS-F) are used as complementary to field-based survey in order to obtain radiometric
identification for each vegetation category and for low-resolution data validation. It is envisaged
that detailed observations and in-field experiments are undertaken to collect local data at the case
study level and to aid the interpretation of remotely sensed images. Observations in the chosen
representative case studies also provide better understanding of land use/cover changes in relation
to various climatic, geomorphologic and socio-economic conditions.
Emissive or thermal infrared radiation - from about 7 m to 13 m had proved to be an
important source of information about land use/cover as well. Variations in emitted energy in the
far infrared provide information concerning surface temperature and thermal properties of soils,
rocks, vegetation, and man-made structures.
The thermal landscape is a composite of the familiar elements of surface material, topography,
vegetation cover and moisture. Various rocks, soils and other surface materials respond
differently to solar heating. The thermal behavior of surface materials is also influenced by other
factors. For example, slopes that face the sun will tend to receive more solar radiation than slopes
that are shadowed by topography. Such differences are of course combined with those arising
from different surface materials. Also, the presence and nature of vegetation alters the thermal
behavior of the landscape. Vegetation tends to heat rather rapidly, but it can also shade areas,
creating patterns of warmth and coolness.
Water tends to retain heat, in contrast many soils and rocks (if dry) tend to release heat rapidly
22
at night and to absorb heat quickly during the daytime. Even small or modest amount of moisture
can therefore greatly alter the thermal properties of soil and rock. Therefore thermal images can
be very effective in monitoring the presence and movement of moisture in the environment.
However it is often possible to isolate the effect of some of these variables and that way to derive
useful information concerning, for example, movement of moisture or the patterns of differing
surface materials.
In our case timing of acquisition is unknown since images that are available are 10-day
composites. Thus we have to rely on those thermal properties of land cover objects that are more
or less stable from day to day.
From this point of view the following vegetation cover objects are thought to be determined
through thermal structure of the surface:

Big cities and industrial zones - through «thermal islands». This must be especially
evident on winter images.

Territories of extraordinary humidity regime - extra dry or wet lands - through
absorption of energy by water in mid infrared zone of spectra.

Anthropogenic modifications of landscapes characterised by thermal pollution of
atmosphere and water bodies.
Simulation techniques and their impact on data quality
Analyses of AVHRR data in 1 and 2 channels and NDVI calculation and classification of land
cover, using both supervised and unsupervised (clustering) approaches allowed us to identify and
interpret the signal of principal types of vegetation cover throughout Central Russia.
High resolution photographic imagery were processed first in their full spatial resolution (40m
and 15m respectively) in order to validate classification based on AVHRR data, and than
degraded to 1 km -resolution in order to estimate the losses of information due to resolution
differences. Thus, two complimentary classifications based on imagery of coarser and finer
resolution were carried out.
23
Classification of land cover over the selected test sites was validated also by ground-truthing
in Karelian site, Darwin reserve, Moscow-Mozaisk region and Kulikovo Pole, where in-field
land-cover observations confirmed classification results and brought data for certain corrections
of land-cover types definitions.
After the launch of SPOT4, these methodological developments will be tested on actual
VEGETATION and HRVIR data that would allow developing the routine procedure of their
processing.
Besides remote sensing data complimentary secondary information, such as the present and
old topographic and thematic maps and statistical records, were organised in the data base of
Geographical Information System. The later allowed us to apply
methods of cartographic
modelling to landscape and land cover analyses and simulation. First, assemblage of a number of
multidisciplinary information layers and simple overlay classification allowed geoecosystem
mapping, which in conjunction with remote sensing data helps better understanding and
interpreting the signal. Second, incorporation into GIS both temporal series of satellite-born and
cartographic data on different time intervals (including those on the last century) allowed us to
trace and model trends in land-cover change. Combination of remote sensing data with ancient
maps and statistical is the only one possible approach to reconstruct long-terms changes of land
cover (Boiko et al, 1991). Cartographic modelling, however, is a key for understanding and
quantitative estimating future land-cover changes, based on the actual parameters derived from
remote sensing and knowledge of principal factors of change.
Area of study in Central Russia would be classified according to its sustainability and risks of
land cover change in order to identify zones susceptible to transformation. The results of
modelling are to be cross-validated with identification of such zones on AVHRR images.
2. GIS data base as a reference for landscape mapping of European Russia (Lioubimtseva
E.Yu)
Landscape data base for image interpretation
The initial phase of this study required assembling, storing and bringing to the same format
24
various data on landscapes and building GIS database on European Russia.
Building a landscape database required extraction of information on landscape components
from available thematic maps. Each thematic layer (coverage) was digitised separately so that the
following thematic 16 vector layers were created: hydrological network, relief (6 classes);
phytomass (7 classes); mortmass (9 classes), production (7 classes); typologic and
ecophytocenologic subdivisions of climax vegetation (48 types and 127 groups respectively);
pedological types and genetic groups (25 and 41 respectively), land-use categories (57 classes);
railways, roads (3 categories); urban centers (5 categories), and administrative regions. These
primary data were necessary for interpretation and mapping physiographic, ecological and socioeconomic landscape components and form a core information in the database.
All cartographic data stored and processed in Arc/Info GIS environment (Arc/Info, ver.7,
1995).
In order to provide complete superimposition of all information layers they were all
transformed to the same projection (conic equidistant projection Nefedova). Deformations,
initially introduced to thematic maps by imperfect registration and geodetic control, were
geometrically corrected by edge matching, and computing links for adjusting erroneous data
using a considerable number of control points.
Data layers based on typological classifications of one of the landscape parameters (e.g.
vegetation, soils, and land use), where qualitative characteristic are expressed by alphanumeric
codes were processed in vector format. Topological organisation of vector data is ideal to keep
information on horizontal spatial relationships between territorial units and their elements
(polygons, arcs and points), what is very important for analysing spatial organisation of
horizontal landscape structure. However, vector cartographic model brings an assumption that
polygon contours are sharp and not fuzzy as in the reality. In our case, when almost all data were
taken from hand-drown broad-scale maps, mainly with qualitative parameters, such assumption is
quite acceptable and a vector model is more appropriate. Layers containing continuous qualitative
information (more specifically, climatic parameters and satellite imagery) were acquired and
stored in a grid format. The later seems to be more convenient when arithmetic and logical
25
operations on data should be undertaken.
The database structure meets the logic of landscape differentiation by various factors and
components. The layers are organised as «themes» - directories, containing thematic vector
coverages, grids, and associated attribute and database tables. In turn, each directory includes
subdirectories and files, containing information on layer’s thematic content, geographical and
topological information.
Spatial layers in ARC /INFO for landscape mapping of EPR
Assemblage and analyses of cartographic, field and remote sensing data on landscape
components of European Russia allowed us classification and mapping of the present-day
landscapes of this macroregion at the reference scale of 1:4-1:8 M. A series of natural and
anthropogenous factors define the landscape structure of the area: climate, relief, hydrological
network, potential vegetation and soils, and human impact. Role of these factors varies at
different hierarchical levels of landscape differentiation. However, strong association of climatic,
terrain and land-cover features within a landscape makes it possible to detect and map both
present and potential landscape patterns at each hierarchical level.
Hierarchical landscape classification designed as a tree of criteria associated with knowledge
base was coupled with cartographic database in GIS. The system allows selection of the
landscape parameters required for landscape mapping (fig.2 ).
26
Fig.2 Conceptual model of the landscape hierarchy in the GIS database.
Although the proposed methodology of landscape mapping was tested only on the EPR the
same approach can be applied elsewhere on the regions where sufficient thematic and
cartographic information is available. Availability of previously prepared data bases or digital
maps of potential and actual landscape components and classified remote sensing data could
considerably facilitate the procedure of landscape mapping. The main advantages of landscape
maps produced by GIS tools that they can be easily corrected, updated or modified according to
the requirements of specific applications. As a difference from traditional hand-drawn maps
thematic information on landscape components organised as a multi-layer cartographic data base
proves useful to analyse not only horizontal spatial relationships between landscape polygons
(such as neighbourhood, mosaic pattern, transitional zones and corridors) but also vertical links
and mutual impacts and dependencies between landscape components. GIS landscape data base is
27
an open system and in contrast with traditional landscape maps allows import and export of
information, modification of classificatory principles and great flexibility in the use of spatial
variables of interest.
A serious disadvantage of semi-authomatic mapping compare to expert’s maps is that certain
specific landscape combinations associated with environmental anomalies, which require
individual analysis, can escape from adequate mapping. The automatically imposed degree of
polygon generalisation can cause the loss of some landscape elements and, contrary, to produce
polygons with erroneous geographical meaning and/or level of landscape aggregation. That is
why we believe that the proposed computer-based mapping should be carried only under
supervision of expert with good knowledge of the study area. Another noteworthy point is that
cartographic source data incorporated to the GIS are never completely free from geometrical and
thematic errors, gaps and subjectivity. That is why including of satellite imagery into the data
base is very useful correct landscape mapping.
Landscape size and map resolution
Each overlay operation in the chain of generating landscape maps was accompanied by
elimination of polygons with area below precision of reference maps. Because all source data
are heterogeneous the minimum allowable size of landscape polygon was defined
individually for each level of map compilation. Precision of 1:4M source maps is 10-12 km².
It decreases in fact with errors brought in course of digitalisation and especially
superimposition of layers coming from different maps. On 1:8 M source maps precision is
already 25-30 km². However, elimination of such polygons does not affect precision of
broad-scale landscape mapping because it is still lower than the minimum possible size of
landscape unit at the lowest level of suggested taxonomy, which is n x10² km².
Differently from traditional drawn landscape maps, the GIS is an open system, which can
easily import new data sets. In case of more specific thematic applications, new variables and
their parameters linked with respective coverages may be added to the database.
28
Landscapes of European Russia.
Let us briefly describe an overall landscape pattern of European Russia (fig.3 and 4).
The major part of the EPR’s landscapes belongs to temperate system except a narrow strip
of subpolar landscapes along the Arctic coast and a small subtropical enclave on the southwestern slope of the Caucasian Mountains. Within the temperate zone four subsystems:
humid, subhumid, semiarid and arid progressively change from north-west to south-east of
the country.
The major part of EPR is dominated by regular horizontal landscape zoning on medium
and low accumulative and denudation-accumulative plains. Strong accumulation occurs on
the Pricaspian lowland, whose major part lies below the sea level. The mountain class is
represented by the low Hibini Mountains. The later are featured by very complex altitudinal
spectrum including 7-8 layers, such as steppe, coniferous, broad-leaved and mixed forests,
woodlands, shrublands, alpine meadows, up to permanent snow layer, varying with slope
exposition and steepness.
Landscape diversity on the plains of humid temperate zone appears mainly on the subtype
level and shows high dependence of forest composition and soils on topography and drainage
conditions. For example, dark coniferous forests (Picea obovata, P. abies)on peat podzolic
and gelic gleisols dominate vast lowlands of the Onega, Mezen and Pechora river basins.
Mixed broad-leaved and light coniferous forest (Pinus silvestris, Quercus robur, Tilia
cordata, Acer platanoides, Corpinus betulus) on sod-podzolic soils are spread on high and
medium plains of the Volga-Oka-Dnepr watershed.
Anthropogeneous modifications of landscapes show rather distinctive zonal pattern,
generally following that of the natural climatic and vegetation zones. Most of tundra and
forest-tundra landscapes are used as low productive tundra pastures for domestic reindeers.
The northern taiga zone southward from approximately 65°-68°N to 59-61°N is
dominated by modified forest landscapes with limited exploitation (forests of the II group) in
29
the western part of Russian plain, and with active industrial exploitation of forests (III group)
in the eastern part of the region. Arable and pastoral modifications appear only in the river
valleys.
In the mixed forest zone all natural landscapes are replaces by several types of
anthropogeneous landscape complexes, with various proportion of agricultural, pastoral and
silvicutural land use.
The share of agricultural landscapes progressively increases in the forest-steppe and
steppe zones. One can trace a certain correlation between share of forests in every landscape
zone and size of forest massifs in this zone. For example, forest-steppe zone of the EPR
where the share of forests does not exceed 4-10% is featured by islet small-size forest
massifs of 0.1-5 km2. In zones of broad-leaved and mixed forests and southern taiga,
corresponding to forest’s bioclimatic optimum, the share of forests in land cover increases to
25-40% and the size of areas varies from 0.5 to 10 km 2. Big forest massifs - from 20 to 200
km2 are rather rare and can be met in middle and northern taiga on the north-east of the EPR,
where forests make 40-70% of the area.
Meadow-steppe and typical steppe landscapes to the south of 52-53°N on the west and
55°N on the east of the EPR are replaced everywhere by non-irrigated arable modifications
with small islands of protected forests and woodlands, wind-forest strips, and relatively small
massifs of irrigated arable lands (about 1-5 km2). In the dry steppe zone on the south-eastern
and southern part of the region (to the south of approximately 48°N) the share of irrigated
agricultural lands increase progressively up to 20-30%.
Semi-desert and desert zone of south-eastern part of the Russian plain (the Pricaspian
lowland) is mainly occupied by low productive pastoral modifications with extensive
nomadic breeding of sheep and goats. The Volga valley and delta are entirely occupied by
irrigated arable lands.
3 Study on vegetation/land use dynamics of European Russia (Yanvareva L.N., Kotova
T.V., Kazantsev N.N., P.A. Tcherkashin)
Reconstruction of long-term land-cover and land-use trends
30
Nowadays European part of Russia is the most highly cultivated part of the country. During
the last 100 years vegetation cover has gone through natural evolutionary changes, but alteration
under human impact - mostly agriculture and timber industry, was considerably higher.
Retrospective analysis is used to study land use for time against last 20 years (before remote
sensing data became available), that involve cartographic and statistical data of different level of
reliability.
Major methodological steps include data collection, analysis and assessment of reliability for
different time periods. This requires investigation of cartographic routine used during that period.
Description of these routines may be found in literature (Gedymin, 1960, 1964; Goldenberg,
1980, Postnikov, 1980)
The following methodological approaches are used to analyse land cover dynamics:
 Statistical
 Cartographic
 Case-studies
The following data have been collected:
(1) Statistical data of land use parameters (area of forests, arable lands, pastures) on
1881 and 1979 for 2 levels:
a) gubernia (1881) and oblast (1979) level as large administrative units;
b) uezd (1881) and rayon (1979) as small administrative units.
(2) Traditional (on paper) maps of administrative divisions for the respective years
(1881, 1979)
(3) As additional source - traditional maps of land use ( 1862 and 1985).
The maps of administrative divisions have been digitized and transformed to the map
projection using now in Russia. Then the statistical data have been input in the relational tables
for each unit for each year and thematic maps have been designed. The patterns of administrative
units for these time slices were different, but changes were also different in different parts of the
31
regions. So it was defined to design the grid of formal polygons 30’ to 30’ and recalculate all the
data to this grid taking into account the source data type.
The grid was automatically designed and the data are being recalculated.
Land use on the territory of European Russia was developing in centrifugal manner. Central
regions are characterised by longest exploitation of lands (up to 1000 years), southern, southeastern and northern - by shortest (100-150 years). Starting from the end of the 19th century
square of agricultural lands has doubled (from 34.6% to 63.4% of total area).
Data comparison had brought out some patterns of land cover distribution. In particular,
significant variances were determined in the level of agricultural development between different
landscape zones as well as provincial variance inside zones. Analysis also showed variance in
directions of the process of agricultural development in different regions.
Preliminary investigation allowed to accomplish maps on land use dynamics (for arable lands,
forests and pastures), fig. 5, 6 and 7.
Major divergence between land use types is defined by natural landscape zones and sub-zones.
Inside these zones differences in land use are called out by economic, political and social factors.
At the same time they are related to such natural factors as relief, lithology, hidrological
conditions, etc.
Natural conditions of European Russia are rather diverse and are represented by a whole
spectra of landscapes from arctic tundra to dry deserts. Northern part of the continent is occupied
by tundra and forested tundra with low level of solar energy supply, long severe winter, immense
spread of permafrost, and low biological productivity. Permafrost predetermine instability of
tundra landscapes and their vulnerability to anthropogenic impact. Vegetation cover is
represented by moss, lichen, and scrub and complement in the southern areas with undersized
low-productive birch-spruce and pine forests; 4% of the territory are occupied by marshes.
Due to severe natural conditions development of agriculture is almost impossible and the
biggest part of used lands is represented by deer pastures. Nowadays their exploitation is very
intensive. Many of the pastures are degraded, most valuable moss-lichen pastures are displaced
32
by less valuable grass pastures. Simultaneously the process of pasture waterlogging is spreading.
Situated to the south from tundra zone, taiga is characterised by significant annual thermal
regime contrasts, surplus damping during most of the year and predominance of needle-leaf
forests in vegetation cover. Forests, that occupy 70-80% of total cover, consist of birch, spruce
and pine. In the southern taiga broad-leaf species, like oak, linden, and maple appear among
common species.
Taiga landscapes have suffered consequential transformation under human impact during the
last century. The most significant factor of these transformations was intensive and irrational
clearance. Anthropogenic transformations influenced not only distribution of forests, but their.
In Moscow, Kalinin, Novgorod, Pskov and several other districts (sub-taiga zone) small-leaf
forests occupy ½ of the forested areas while 7/10 of these areas consist of immature trees.
According to General Survey data (1766-1884) forested areas occupied more than 75% of
northern taiga, 50-75% of central taiga, 30-50% of southern taiga, and 15-30% of broad-leaf
zone. In 1993 forested areas occupied 50-70% of northern and central taiga, 35-50% of southern
taiga, and 10-30% of broad-leaf zone. In northern and central taiga reduction of forested area was
caused by continuous clearance. In the south the major factor of forested areas decrease is
expanse of arable lands and pastures. Northern taiga is virtually not developed from agricultural
point of view.
Southern taiga and sub-taiga are areas of historic development of agriculture. Against the
beginning of 17th century arable lands already occupied more than 40% of the territory.
Broad-leaf forest zone and zone of forested steppe were areas of intensive agricultural
development. Forested steppe zone is characterised by high biological productivity. During the
period of anthropogenic influence natural vegetation cover has changed considerably. Areas to
the south of Oka river were ploughed already in 15-16th century. Current area of arable lands
come by 60-70%. Natural vegetation is only preserved in coombs.
Steppe zone is highly supplied with solar energy. Humidity conditions vary from semi-dry to
semi-wet. This natural landscape zone was one of the latest to perceive arable pressure - only at
33
the end of the 19th century. Only 24.5% of steppe were ploughed in 1863, the rest was occupied
by pastures. Against 1881 share of arable lands enlarged to 48.5%. Nowadays almost all the lands
in this zone are ploughed; pastures are only preserved in river valleys. Broad-leaf forests, mostly
oak, form minor solids on watersheds and river bottomlands.
Semi-desert and desert landscape zone include dry lands with very high solar energy supply.
Biological productivity of natural vegetation is rather low, however, in the areas with optimal
seasonal damping productivity is very high.
In agriculture mostly bottomlands of river valleys and firths are used. Mean forest cover is less
than 0.7% and in some areas (Kalmykiya, Astrahan region) is as low as 0.2%. Agricultural lands
occupy 83% of the zone, however arable lands constitute only 5% - the rest are pastures. Large
areas of former arable lands are bare sands nowadays. A lot of pastures are degraded that cause
decline of biological productivity of grass from 5 to 8 times.
Analysis of consistent patterns in land cover transformation for landscape zones has brought
out areas of most unstable land cover in southern taiga and sub-taiga. At the same time dynamics
of lands in western and eastern parts of these zones contrast from each other essentially.
In the west (Pskov, Novgorod, Tver regions) the tendency of neglecting arable lands and their
overrun with scrub is noticed. Area of forest here is the same and even higher than during the
times of General Survey. The reason is shift of rural population into big cities, complexity of
mechanisation of agriculture on small fields, almost total absence of technical and social
infrastructure.
In the east (Perm, Vyatka regions, Udmurtia and Tataria republics) set-up is opposite: expanse
of arable lands and pastures over forests is noticed.
Accomplishment described above is a provisional basis for investigation of vegetation cover
dynamic. Cartographic and statistical data has been collected describing transformation of land
use patterns in the European part of Russian Empire and the Former USSR. Collected materials
were analysed in order to create a preliminary land cover/land use regionalization map. Regions
of the map were examined and described, case study regions were selected for more detailed
34
investigation. Computer technique for compilation of time series of land cover maps was
developed. Future work includes investigation of accuracy level of historic maps of land cover
and development of maps of land cover transformation for the last 100 years for the case studies.
It is expedient to conduct data collection for the 50’s of this century since current series
includes data from General Survey till 1979 with a gap in the middle. Data for three years: 1881,
1887, 1917 is used in the research in order to assess its level of reliability, since during this
period institution of public statistics was in the process of developing and characterised by
diversity of data sources that sometimes contradicted each other.
Analysis of land cover transformation on EPR between 1970-1992
On this stage of investigation our major tasks were:
1. To create multi-level spatial database to serve as a basis for land cover change
investigations, study of dynamics and developing methodology for up scaling. Such a
database that includes spatial and attribute data from different sources and different time
periods for well-studied and monitored territory will serve as a good training polygon for
methodological studies.
2. Given described database and ability to check our estimations by field research to
estimate the possibility to analyze land cover dynamics for given period. Basically, the most
important would be changes in forest cover on this territory during the period of transition to
another land use system. Forest cover had undergone dramatic changes on the study area
during the last 10 years. Would these changes be seen on regional level?
The database created in ARC/INFO environment includes all available digital information on
the study area.
As a separate part of the model was created a database with control ground-truth points inside
the study area with precise knowledge of land cover, land use, past and future trends. Several of
them were chosen on territories with more or less stable land cover (natural reserves). These
points were used for error correction during overlay operations and quality control for remotely
sensed data. The others were selected during analysis of dynamics and land cover mapping
35
process - to check the results. This database was geocoded so that all the points might be overlaid
with the rest of spatial data. Appropriate database records for ground points were attached to
point geographic features in spatial model using so-called "hot-link" technique.
Partly integrated into the model were also false-colour photo images from Resource-F Russian
satellite. Since images were obtained as hard-copy prints only meta data about them was included
into the model in form of "footprints". These images were of great value to solve uncertainties
that occurred on regional level while analyzing changes in land cover.
The model described above was used then for the following purposes:

Investigate if traditional Russian mapping approach might be used in conjunction with
remotely sensed data for regional level study of land cover changes.

Develop and test methodological approach for scale-dependant modeling, analyze
distortions that may occur when moving from local to regional level.

Study actual changes in land cover in the center of European Russia during the period
of economic transition using data from different time periods.
4. Interpretation and mapping landscape pattern in forest and forest-steppe zones of Russia
using remote sensing (Lioubimtseva E.Yu.1)
Outline of landscape approach
Landscape approach, world-wide recognised today as a powerful method of multidisiplinary
environmental mapping (Bunce et al., 1984; Delbaere & Gulinck, 1994; Meeus, 1995) provides a
basis for the perception of the surface area as a system of interrelated territorial samples
(landscape units) with specific environmental characteristics. Experience of using SPOT and
Landsat-TM data for large-scale landscape mapping and ecological regionalisation were recently
reported by several authors (e.g. Nellis and Briggs, 1987; Gulinck et al., 1991,1993; Gossens et
* This part of research was carried in UCL and partly funded by OSTC (Belgium) 1 This part of research was carried out in UCL
and funded by SSTC (Belgium)
36
al., 1993; Farjon and Thunissen, 1994; Haines-Young and Bunes, 1994; Quattrochi and Pelletier,
1993). Photographic images from RESURS-F and Salyut were applied by Glushko and Tikunov
(1994) for qualitative broad-scale landscape mapping. Methodology of using remote sensing data
in landscape studies, however, is still far from being developed. Although it is generally
understood that remote sensing can be successfully used for mapping landscape framework, there
are no commonly accepted approaches to step from land-cover to landscape mapping. There is
neither consensus on which spectral or geometric land-cover properties can be used as diagnostic
tools for landscape interpretation, and how much they are effective at different scale levels.
Land cover classification
Basic data used for landscape mapping are classified images of high (25-45 m) and medium
(170 m) resolution from RESURS-F/MK-4 and RESURS-01/MSU-SK instruments, and ancillary
data from topographic and thematic maps, where the emphasis was done on landscape pattern
interpretation.
Satellite images were radiometrically enhanced with crisp filter in order to facilitate visual
analyses and interpretation land-cover and of landscape pattern and then classified according
land-cover properties.
Two methods of land-cover classification were applied: per pixel maximum likelyhood
algorithm and ISODATA algorithm imbedded to ERDAS IMAGINE image processing system.
While unsupervised ISODATA classification turned to be quite effective for MSU-SK data,
maximum likelihood supervised classification with 8-12 training polygons per class gave much
better results for photographic images (78-87% accuracy). The main sources of errors are
confusion between parts of water and wet bare soil pixels (Voronej case study region - MK-4
and MSU-SK scenes) and pastures, hay meadows and part of agricultural crops (Moscow case
study region MK-4).
In order to avoid lost of information on land-cover texture and landscape pattern configuration
due to resampling and to keep land-cover proportions within landscapes, we carried first landcover classification and than geometric correction and rectification of classified data to
37
geographic projection afterwards.
Landscape pattern interpretation on the satellite images
Analysis of pattern and internal spatial structure of landscape are crucial for understanding
different landscape functions, such as ecological niches, barriers, corridors for species migration,
channels, sources and destinies of energy, material or water flows, etc. (Zonneveld, 1989;
Forman, 1991). The pattern of various elements within the landscape, in addition to the landcover features, is an important determinant of landscape existence and functioning.
There are strong pattern - process relationships in landscape which link pattern configuration
(which may be very diverse: circular, dendrite, radial, chess-board, etc.) and land-cover
composition with ecological processes and energy, information and matter fluxes. Therefore,
pattern is a key for understanding functions of different landscape elements.
For interpretation landscape pattern, derived from land-cover classification on satellite images
we use the patch-matrix-corridor landscape model, commonly known in landscape ecology
(Forman and Gordon, 1986).
Landscape at any level of spatial aggregation is heterogeneous radiometrically and appears on
the satellite image as a combination of a number of particularly organised in space land-cover
categories in different proportions. The higher is the level of landscape aggregation and the
broader is the scale, the higher the heterogeneity of a landscape unit is.
Proportion and spatial organisation of land-cover parcels within a landscape are indeed
important criteria of landscape delineation. For example, agricultural landscape on MK-4 scene
of Voronej region is a combination of arable lands, gardens, built areas and small woody massifs
and shelter belts. On MK-4 scene of Moscow region agricultural landscape is represented by
quite different combination of arable lands, forest patches, urban areas, and meadows with
another proportion of land-cover types and shape of patches. Forest-agricultural landscape on the
same image has its own combination and configuration of forested areas, arable lands and
meadows, urban landscape is a combination of built massifs with small vegetated parcels and
road network, etc.
38
That is why landscape is recognisable first of all not by radiometry (because it is mixed) of
land cover of its patches but mainly by its texture and geometric configuration. Linear objects
even on very high-resolution images are always represented by mixed- pixel and can be
recognised by their shape rather than classified according to spectral properties.
After preliminary delineation of landscapes on classified images according to visual
assessment of their pattern and analyses of ancillary data (topography, vegetation, land-use maps)
the following criteria were evaluated in order to analyse configuration and composition of
landscape pattern.
1) Land cover:
 dominant land-cover type(s);
 proportion of land-cover categories in landscape;
 land cover diversity - number of land-cover types within landscape unit
2) Patches
 patch size
 patch shape
 patch connectivity - linkage of patterns by corridors
3) Matrix
 matrix porosity
 matrix shape
4) Corridors:
 corridor rate - proportion of corridor’s area in the landscape
 connectivity - link of corridors with patches
4.3 Impact of spatial resolution on landcover proportion.
39
300000
250000
200000
urban built-up
parkland
coniferous forest
150000
water
agricultural
orchards
hardwood secondary
forest
100000
grassland
mixed forest
area, ha
50000
0
25m
100m
300m
600m
900m
resolution, m
1200m
1500m
Fig. 8 Land-cover proportions at increaingly coarsing spatial resolutions.
40
Patch shape is a key feature of landscape pattern since it is closely related with content and
ecological functions of patches. For shape evaluation we used the following shape index
V= (4Sp*100)/Pp ,
(1)
where Sp and Pp are patch area and patch perimeter respectively. Utility of this index for
evaluating shape of landscape pattern on classified SPOT images was shown by Gulinck with
colleagues (1993).
Measurements of patch areas and perimeters were carried with a help of «seed growing» tool,
which allow to grow polygons including all pixels meeting the specified criteria - belonging to
the same land-cover category and connected by 8-pixel neighbourhood.
Matrix porosity can be calculated as W=(1-Sm/A)*100
(2)
where Sm is a matrix area and A is a total landscape area. It can be also calculated as
W= (Sp + Sc)/A ,
(3)
where Sc and Sp are respectively area of patches and area of corridors and A is the total
landscape area. Because such linear objects as roads, linear fellings, hedges and small gullies are
usually represented by mixed pixels even on images of high resolution and their width does not
exceed pixel size Sc can be represented as k Lc, where Lc is a corridor length and k is a pixel
size.
Analyses of shape index calculated for all patches in each landscape on MK-4 scenes shows
that for most land-cover categories it has a general trend to decrease with increase of patch size.
Small patches ( less than 1000 m²) with relatively simple rectangular shape have the maxima
values of shape index - 80-90 (shape index of circle is 100). In contrast, large patches of widely
extended land-cover types often tend to have very complex shape with index values. Practically
for all land-cover types in all tested landscapes shape index of patches bigger than 6-7*105 m² is
below 5 or 6. However, the relationship between patch size and shape seems to be more complex
41
and is individual for each type of landscape pattern. While in relatively "natural" landscapes this
relation features all types of patches, in highly anthropogenous urban landscape (urban landscape
of Moscow, suburban landscapes in Moscow agglomeration, urban landscape of Voronej town),
and relatively high values of shape index (40-50) also feature relatively big urban and vegetated
patches (2-4*105m²)
Several specific combinations of landscape patterns were revealed for each test site area.
On the images of Darwin reserve there are five following types of landscapes:

swampy coastal plains with spruce and birch forest and peat bogs;

coastal plains with arable lands and grasslands;

low plains with sparse birch and aspen woodlands and swampy meadows;

low swampy grasslands with numerous lakes and bogs;

swampy spruce forest with peat bogs.
All these landscapes are featured by rounded shapes of small patches and very complex
configuration of patches in all land-cover types (shape index varies between 0.005 and 20). The
highest values of shape index occur for small lake and bog patches but even their shapes are
rather complex (index values 12-15) (fig.9 a,b). Matrix porosity is low for forest (less than 10%)
and relatively high in landscapes dominated by woodlands and grasslands (more than 70 %).
In four landscapes mapped on the scene of Moscow region (fig. 10) shape index shows the
maximum variance. As it was mentioned it behaviour differs mainly in urban and non-urban
landscapes. All landscapes are featured by high density of corridors (roads and railways) and
connectivity between non-forest land-cover patches. Forest, who it is a matrix land cover in two
landscapes shows extremely high matrix porosity (91% in case of low plains with forestagricultural landscape to the west of Moscow). In contrast, porosity of anthropogenous landscape
matrixes, such as arable and urban lands are lower - 59 and 58% respectively.
The major part of Voronej region test area is occupied by landscape of dissected plains with
arable lands and dense dendrite gully network with hardwood forest and shrub. It can be easily
42
visually recognised as a combination of very simple rectangular large patches of arable lands and
extremely complex and highly connected forest and shrubland patches (shape index values 0.050.2). Matrix porosity and density of corridors in this type of landscape are very high. The later
include hedges (shelter belts around field and garden patches and along the roads), roads and
network of gullies. This type of landscape with typical dendrite configuration makes contrast
with landscape of the Voronej-river valley with patchy mosaic of arable lands, gardens, urban
and rural built areas, parks and forest, as well as with relatively homogenous landscape of
watershed with hardwood and mixed forest of the Voronej region. Again, like in case of Moscow
region urban and rural built patches show generally the highest values of shape index (40-50 for
areas of 6-8*105m²). The size of forested patches is the maximum and while this land cover is a
matrix of landscape it shows relatively low porosity (32%).
Landscape heterogeneity and image resolution
Each landscape is featured by certain heterogeneity objectively observable in field or on
remote sensing data. Spatial heterogeneity can be regarded as a measurable expression of the
overall spatial complexity or variety of an area (Wiens, 1992). The notion of heterogeneity
comprises those of patchiness, porosity or land-cover proportions and can be evaluated in many
different ways.
Landscape heterogeneity measurable on a satellite image can by different from that observable
in field, because it depends on spatial resolution of the image. It is therefore different for the
same landscape while measured at different scales (resolutions). For example, in an area of
heterogeneous land covers studied on the image of 30-m resolution, spectral responses for
different objects within a 30-m pixel are averaged and aggregated into a composite spectral
response for any particular pixel that falls over a specific area on the ground.
Thus, choices of resolution adequate to the type of landscape pattern and research objectives
are particularly critical in landscape studies.
In order to test possibility to detect of landscape patterns on images of different spatial
resolution we used two types of data: real (RESURS-01/MSU and NOAA/AVHRR) and
43
simulated - MK-4 images degraded to 150 m and 1000 m resolution.
It was shown by Milne (1992) that relation between landscape pattern represented at one scale
versus another can be evaluated as Z= Ni/Ni*k
(4),
where Ni is a proportion of the study area occupied by a given patch type at the first scale scale
and Ni*k is a proportion of the study area occupied by the same patch type at the second scale.
We evaluated the same parameters of landscape pattern: land-cover proportions, patch size
and shape, and matrix porosity for landscapes of the same areas with coarser resolutions in order
to understand to which extend medium and coarse resolution data can be appropriate for mapping
landscape structure.
At the images of 150-160 m resolution, both acquired by RESURS-01/ MSU-SK, and
degraded RESURS-F/MK-4, the same landscapes as defined on the real MK-4 images can be still
detected visually by their specific patterns but their internal structure and configuration are
simplified (fig.10). Number of patches in each land cover category and matrix porosity decrease
and patches smaller than 3*105m² are not anymore detectable. Small neighbour patches with the
same land-cover are aggregated in bigger ones (e.g. several lakes of 2-5*105 m on MK-4 image
appear as one lake of 1.4 km²). Relatively big homogenous landscape patches (we have analysed
all grassland and woodland patches bigger than 4 km², detected both on MK-4 and MSU-SK
images) show much higher values of shape index (12-23) with 150 rather than with 45 m
resolution (index values 0.5-1.2) for the same landscape patches. This difference is even higher
for lakes: 70-95 with 150 m resolution (very simple configuration) against 2-5 with 45 m
resolution. Landscape corridors (except river valleys) can not be detected on these images.
On coarse- resolution imagery -AVHRR - there is no really homogenous land covers and-all
objects are represented by mixed pixels. Each 1-km pixel is a generalisation of combination of a
number of land cover types and represents one or several landscapes. Combination of landscapes
of land-cover is responsible for pixel’s signal.
If a landscape dimension of the level perceptible on high and medium resolution imagery varies
44
within n*10-n*10² km², one 1km pixel of the AVHRR scene is either one landscape (which is a
particular mixture of several land-cover types) or combination of several landscapes. As a matter
of fact, in the conditions of rather patchy heterogeneous land cover of European Russia each
land-cover category detected with 1-km resolution represents a mixed pixel mosaic resulting
from patch combination within one or several landscapes.
Mapping of the same landscapes with 1-km resolution leads to very high lost of information not
only on internal landscape pattern but also on configuration of the whole landscape units. It leads
to reduction of landscape diversity within a pixel of bigger size to dominant type of landscape
type at the expense of accompanying types and considerable simplification of the landscape
structure compare to the higher resolution level. Possibility to recognise visually the same
landscape units with coarse resolution depends on their patchiness. Let us illustrate it with a few
examples.
Comparison of land-cover classification carried with three different resolutions on the test area of
Darwin reserve shows that from five landscape types detectable by their pattern on MK-4 and
MSU-SK images only two can be visually recognised on AVHRR image: swampy dark
coniferous forest and coastal plain with agricultural lands.
In the first case the landscape is almost homogenous and represented by one highly connected
matrix (porosity is less than 20% on MK-4 and less than 11% on MSU-SK scene). That is why
practically no lost of information occurs with decrease of spatial resolution.
In the second case (coastal plain with arable lands, grasslands, woodlands and built-up areas) we
deal in contrast with very heterogeneous patchy landscape formed by a variety of land-cover
types of small size (5*102-1*105 m²) and simple shape (12- 60 on MK-4 scene). Since patchiness
of this landscape is far below the size of grid cell (1.1 km²) of the AVHRR image, the landscape
is well detectable as one homogenous land-cover type - representing highly heterogeneous
mosaic.
Differently from this case the three other landscapes in this region have relatively big patches
(generally 0.8-2 km² for grasslands and woodland while measured on MK-4 image), which is
comparable with the dimension of AVHRR grid. Shape of the patches is quite complex (0.1-6)
45
and matrix porosity varies from 52 to 74 % from one landscape to another. So most of patches
drop into several neighbour 1-km cells and each cell results from aggregating signal of "pieces"
of patches with different land cover. As a result neither patch nor landscape borders can not be
always detected. The landscape of swampy grasslands with lakes and bogs was only partly
identified on AVHRR scene and partly mixed with a landscape of plains with sparse woodlands
and swampy meadows (Fig. 12).
a
b
c
Fig.12 Three cases of landscape patchiness versus spatial resolution:
a) patch size is far over cell size - landscape is detectable;
b) patch size is comparable with cell size - landscape is not detectable;
c) patch size is far below cell size - landscape is detectable.
Table 3 Landscape elements at different spatial resolutions
Land-cover type of
LANDSAT-
landscape element
TM*
SPOT* RESURS-F
RESURS01/
NOAA/
MSU-SK
AVHRR
/MK-4
Areal elements
Forest
y
y
y
y
y
bare soil
y
y
y
y
n
water
y
y
y
y
y
46
grasslands
y
y
y
y
y
shrublands
y
y
y
y
n
urban agglomerations
y
y
y
y
y/n
sparse built-up areas
y
y
y
n
n
rivers
y
y
y
y
y
roads
y
y
y
n
n
hedges
y
y
y
y/n
n
Linear elements

- according to Lioubimtseva, 1997

* according to Farjon & Thunissen, 1994
For efficiency in cost, data processing time, and analysis, it is always desirable to chose the
broadest scale data available for identifying landscape characteristics of interest (Quattrochi &
Pelletier, 1994). However, mechanistic generalisation of data to coarser resolution can lead to
very high overestimation and misevaluation of areas, while such features as land-cover
proportions in land-cover mosaic and patch configuration are much more essential in landscape
studies than in land-cover or land-use mapping.
Landscape hierarchy and image resolution
Besides heterogeneity of landscape pattern at a given spatial level, the suitability of resolution
to landscape mapping depends on hierarchical level of landscape aggregation. Landscape
hierarchy concept is useful for understanding of different processes and landscape functions,
which occur at different spatial levels and require different scales of study. Hierarchical
conceptual model helps us to see different landscape features and functions at different spatial
levels, and suggests a certain abstraction. Fine detail information which may be important for
instance at the elementary lowest hierarchical level (facies (n m²) become irrelevant at the higher
47
levels of landscape aggregation (stows (n104m², landscape regions (n106 m²). Matter, energy or
information flows and other ecological processes which occur at the continental or planetary
scale (e.g. biochemical cycles, species migrations, temperature variations) differ from those on
the level of region or stow. Therefore, different spatial resolutions of remote sensing images can
be useful for studying landscape phenomena at different levels.
Landscape aggregations of the high hierarchical level (landscape zones), which can be mapped
with coarser resolution are also combinations of several land-cover types but of the much broader
scale: e.g. landscape of coniferous boreal forest on the plain; forestry-agricultural landscape with
urban lands; high-mountain landscape; plain desert landscape with irrigated agricultural lands.
Each landscape zone which can be delineated at the meso- or macroregional level is also
heterogeneous and represents a combination of several types of patches, where each land-cover
category on the image results from averaged signal of one or several landscapes of the
subordinated hierarchical level.
In order to test usefulness of coarse-resolution data for broad-scale landscape mapping (on the
level of climatic zones, major landforms and dominant vegetation and land-use types), the results
of land-cover classification derived from NOAA/AVHRR were compared with classified MSUSK and MK-4 images (three test areas discussed above and four complimentary areas in forest
and steppe zones of European Russia) and superimposed with recently compiled digital landscape
map of the European Part of Russia.
Association of some land-cover categories derived from NOAA/AVHRR, RESURS-01/MSUSK and RESURS-F/MK-4 with landscape units of Central Russia is shown in Table 4.
Table 4 Selected land cover and landscape categories of Central Russia
land cover category
landscape zone
AVHRR
MSU-SK
MK-4
climate
land
form
I. mixed
and
deciduou
s forest
1.swamps;
2.hardwood
forest;
3.mixed
and
(from different test sites)
temperate
low
plain
1a) sphagnum swamps humid
with sparse mixed woody
vegetation
vegetation
land use
1.(110b) herbaceous (88) no impact
bogs with Carex nigra,
C.rostrata, Equisetum
middle
fluviatil, Menyanthes (32) protected and
plain
trifoliata;
recreational forest
1b) eutrophic fens with
48
hardwood
forest; 4.
sparse
mixed
woodland;
5.overgro
wing
fellings
and
burnts;
birch-fur woodlands and
bushes
2. oak and lime forest
3a) mixed birch-fir forest
3b) mixed pine-smallleafed forest
3c)small-leafed
aspen) forest
(birch-
6. urban 4a)
sparse
lands
woodland
birch
4b) new pine plantations
4c) mixed (birch, aspen,
fir) woodland
5a) fellings
5b) old burnts
6a) urban lands
(115d)
Sphagnum
oligotrophic bogs with
Pinus
Silvestris,
Ledum
palustre,
Sphagnum magellanica
with
hydromorphicpodzolic soils
of the I group;
(33) forests of
limited
exploitation of the
II group
(36)
protected
forests
in
2.(58a)broad-leaved
combination with
forest with Quercus
pastoral
and
robur, Tilia cordata
arable lands (not
Acer platanoides and
less than 20%);
nemoral species on
grey and dark grey (37) woodlands in
forest soils;
combination with
pastures
and
3.(32a) fir forest (Picea
bushes
abies, P.obovata) with
burch and lime (Betula (4) arable lands
pendula, Tilia cordata) with forests and
on
not-improved
pastures
sod-podzolic soils
6b) suburban lands
6c) arable lands
II.
steppes
and
meadows
1. sparse
woodland;
2.
grasslands
heathland
s
and
meadows,
1 sparse broad-leaved temperate
(lime, oak, ash) woodland
subhumid
2a) typical grass-and-forb
steppe
3
agricultur
al lands
2c) flood-plain meadows
2b) forest-steppe with
rare small tree massifs
2d) meadows on burnts
and fellings
2e) shrub heathlands
3) arable lands with crops
III.
1.
agricultur agricultur
al lands
al lands
with
crops,
2
meadows
1a). agricultural
with crops
lands
1b) improved pastures
and
hay-making
meadows
1c) shelter belts
2a) flood-plain meadows
heathlands
and
2b) shrub heathlands
grasslands
2c) abandoned lands
middle (66a)
Grass-form
plain
mesophytic
or
xeromesophytic
steppes and steppic
meadows
(Festula
valesiaca, sp. Stipa,
Bromopsis,
Helictotrichon,
Phleum, Poa) with
Quercus
publescens
and
mid-european
species on leached
chernozems
and
meadow-chernozemic
soils
(1) arable lands;
(32) protected
woodlands and
forests of I group
(2) arable lands
with not improved
pastures and hay
meadows
(1) arable lands
(2) arable lands
with not improved
pastures and hay
meadows
(4) arable lands
with forests and
not-improved
pastures
(76)improved
49
3.
open 3) open soils
soils
4a) urban lands
4. urban
4b) suburban lands
lands
4c) arable lands
gardens
pastures
and
meadows
with
arable lands (not
less than 20%)
and
Some land cover categories, which are mapped as homogenous on course resolution image e.g. mixed and deciduous forest appear to be a mosaic mixture of at least 6 different land-cover
classes within the same area on medium-resolution MSU-SK image and 14 classes on MK-4.
While high resolution data are more suitable for mapping landscape aggregations of low
hierarchical levels (stows, landscape regions), territorial units of subcontinental level, such as
landscape zones are better detectable on course-resolution images. Indeed, high resolution
information, which allows detailed mapping of relatively homogenous land cover categories with
good separability of spectral signatures, becomes a «noise» and needs generalisation for mapping
landscape structure defined by macroregional climatic, landform and land-use regularities.
Among the limitations of applicability of course and medium resolution data for land-cover
mapping is the mixed pixel problem caused by the low spatial resolution of data. That is why this
type of data is more suitable for large spectrally homogenous areal elements. However, in case of
broad-scale landscape mapping aggregation of spectral heterogeneity in coarser pixels can be
useful because it helps to abstract from unnecessary details caused by landscape patchiness of
subordinated hierarchical levels. Observation of landscapes at different scales let us suggest that
the following relations between spatial resolution of remote sensing data and landscape hierarchy
should be considered :
Table 5. Sensors’ suitability to landscape mapping at different levels:
sensor
NOAA/AVHRR
spatial
landscape
resolution
size
landscape aggregation
1.1 km
104-106 km² 10²-10³ km²
landscape zone
10²-104 km²
land or group of stows
RESURS01/MSU-SK 160-170m
patch size
1-10 km²
hierarchical
level
of
50
RESURS-F/MK-4
20-40m
10-2-10 km²
10²-105 m²
stow, landscape region
RESULTS AND DISCUSSIONS ON REPRESENTATIVITY DUE TO USE OF SIMULATED DATA
1.Remote sensing applications for land-cover and landscape mapping (Tcherkashin P.A.,
Solntsev V.N., Kholod S.S., Khramtsov V.N.)
Country (macroregional) level
IDRISI Ver. 4.1 software package for IBM PC hardware platform was used for data
visualisation and analyses (Eastman, 1992). Attribute information, traditional mapping sources,
and expert knowledge were used to assign land cover types to the resulted land categories.
Annual land cover maps were combined to produce a map of land cover for the period 1986 1990 (Fig. 13). Principal Component Analysis (PCA) was carried out for each year during this
period using IDRISI PCA routine . 12 monthly generalised images were used as an input for each
year and 4 components were derived. Images were compared to each other and to traditional
maps. Since now adequate ground-truth data existed on vegetation phenology, expert assessments
were used to recognize geographical meaning of the images. We believe that the following
phenological characteristics of land cover are represented by principal components (Table 6):
Table 6 Phenological characteristics of land cover.
Component Name
Phenological Meaning2
PCA #1
Total annual NDVI value. This proved to represent
annual production of biomass (Refer to next section)
PCA #2
2
Early year vegetation activity (for each pixel of the
Values on the resulted images are not correlated to any quantitative categories, since this
requires a lot of attribute statistical data. Nevertheless, they allow to fulfill spatial
comparison of different territories
51
image the higher is value, the earlier is vegetation
onset)
PCA #3
Duration of growing season (for each pixel of the
image the higher is value, the longer is total duration
of vegetation activity period)
PCA #4
Late year vegetation activity (for each pixel of the
image the higher is value, the later is vegetation
offset)
Some of this information may be directly used for vegetation analysis, other will be used for
land cover stratification by other computer techniques. Additional research is needed to correlate
pixel values of the PCA images to real phenological characteristics, for example, time of
vegetation onset/offset, duration of growing season, etc.
Broad classification results were not efficient for land cover classification but proved to
correlate with global climatic conditions. Since NDVI reflects vegetation behaviour, these zones
represent to some extent global vegetation growth conditions formed by climate (solar energy and
humidity) and local landscape peculiarities (relief, soils, etc.).
Around 40 seasonal clusters were produced by fine classification scenario for each of the
images. These clusters were then analysed to determine land cover type they belonged to.
Temporal NDVI profiles for the whole year and graphs of attribute information for each cluster
as well as data from traditional mapping and descriptive sources were used for that purpose.
Practical implementation has shown that only 15-16 first clusters really carry efficient
information covering about 90% of total variability of covers. Higher clusters represented
insignificant variations in dry grassland areas and had fallen into one class. On the level of major
biomes (forest and grassland) land cover could be clearly determined for most of the clusters.
Those with mixed phenological characteristics formed a third land cover class - forest/grassland.
Experiments with present-day landscape mapping and classification done on the basis of
remote sensing at macroregional (for the whole Russia) level have shown the following :
52
Natural zonal structure
According to our analysis of vegetation biomass seasonal dynamics based on 10 km
resolution data set, zonal stratification of forest landscapes on the north of Eurasia differs from
that which is traditionally drown on vegetation and landscape maps. Two zones are well visible
on a 10 km resolution composite image. They may be determined basing on humidity of the
landscapes: a) extra-humid forests zone that mainly corresponds to the zone of northern taiga
but also expands throughout tundra and central taiga zones; b) zone of optimal humidity forests
corresponding with the zone of southern taiga and mixed forests. Both of these natural zones
stretch in sub-latitudinal sense throughout almost all of the Eurasian continent that does not
correspond to the pattern of traditionally subdivided natural zones. It is also possible that steppe
zone should be also stratified according to the balance of humidity and warmth (e.g. normal
humidity, dry and extra dry). These zones should not properly correspond to the generally
accepted structure of vegetation and landscape zones because there differentiation is based on
NDVI classification and not on floristic characteristics of the land cover.
Anthropogenic transformation of natural zones structure
Global land use dynamics and evolution can be studied through analysis of satellite imagery
and cartographic data at different time periods. Preliminary experiments in this field show that
human activities have changed the pattern of natural zones in many regions of Russia. Main oil
and gas exploration region of Russia in Western Siberia is one of examples. Severe exploitation
of natural resources in this region causes irreversible global scale changes of vegetation and land
cover.
Another area of active anthropogenic influence that was detected on the NDVI scene is an
area of intensive irrigation northward from the Black Sea along the Dnieper and Kuban river
valleys. As far as vegetation activity is concerned landscapes of sub humid steppe here greatly
differs from dryer steppes of the Volga region and West Siberia, although traditionally they are
shown on vegetation and land cover maps as the same natural zone without further subdivision.
As a final example, well-known area of ecological disaster around the Aral sea can be easily
identified with the use of remote sensing indicators.
53
Regional level
Visual analysis of AVHRR data from EROS data center
Even a brief look at the AVHRR 10-day composite images of 1-km resolution showed that not
all of them were suitable for the research, since many of them contained more or less visible
errors and artefacts, like borders of images from different passes due to badly applied mosaicing
technique, or cloud influence.
During the experiments images were visually analysed to determine the following:

What images were reliable for this particular research

What additional (to NDVI) information could be provided by thermal infrared spectral
bands

What phenological information did images of different seasonal periods provide, how
clearly various land cover types could be visually interpreted on them
The following 4 images were chosen for the research basing on how clear they are from
different artefacts and errors described above and how representative they are from seasonal
vegetation phenology point of view: April 21-30 1992, June 1-10 1992, September 1-10 1992 and
February 21-28 1993. All 5 spectral bands were obtained for each image as well as NDVI. All the
experiments described below were fulfilled with these images.
Since phenological and ecological meaning of NDVI was widely examined, we are not
including results of its visual interpretation here, but focus on information that thermal bands
(AVHRR 4,5) provide from land cover point of view.
Visual evaluation of thermal imagery showed that thermal data that is available now in this
research is not suitable for direct determination of land cover. However, thermal characteristics
can be estimated for classes derived from standard NDVI-based classification procedure. This
information than be used as knowledge for further investigation using thermal data in land cover
assessment.
On regional scale 5 major cluster categories were determined using 1-km seasonal vegetation
54
activity data from AVHRR by applying unsupervised classification described above (Fig.14). By
visual interpretation and analysis of attribute data and case studies these clusters were primarily
defined as following land cover types:
1. Water Bodies
2. Agricultural Lands
3. Deciduous Forests
4. Coniferous Forests
5. Non-Vegetated Areas
The image containing these clusters was then analysed using vast traditional cartographic and
text data about land cover particularity of the study area as well as land use on it. In general this
analysis had brought out landscape structure of the study area that is composed of several naturalanthropogenic regions basing on specific for each region combination of cluster patterns.
By comparison analysis of land cover data from traditional and remotely sensed sources for
each of these natural-anthropogenic regions the following objectives were achieved:
Elaboration of geographical and landscape meaning of determined cluster categories.
Cluster «Agricultural Lands» is apparently rather reliably determine arable lands under
cereals, pasturage and meadows. However wide spread of «Non-vegetated areas» cluster in the
south of the study area, for example on the Middle-Russia Height, points to the fact that some of
gardens and potato plantations felt into this cluster. Moreover, some of the agricultural lands fall
into «Coniferous Forests» category especially on the watersheds of the southern part of the area
with highly intensive agriculture - pine forest are very rare here and only appear on sandy
terraces of large rivers. This cluster also appears in low-forested areas of Dneper-Desna and
Meshera lowlands. This can be explained by landscape meaning of «Coniferous Forests» cluster
(see below).
«Deciduous Forests» cluster is determining this type of forests rather clearly, while combining
both small-leaf forests, widely spread in the north, with broad-leaf forests preserved in the south.
55
But besides that it includes almost all mixed spruce/small-leaf forests, that are often shown on
currently existing vegetation maps as pure spruce forests. This cluster is dispersed on the Valday
Height and to the west from Moscow - these areas are determined on existing maps as continuous
spruce forests. Furthermore in Bryansk region this cluster includes oak-pine-spruce forests on
sandy river terraces. Geographical perception of these facts could be the following: cluster
category, primarily determined as «Deciduous Forests» actually shows relatively dry well drained
environments with young forested vegetation cover.
«Coniferous Forests» cluster is determining pine forests very clearly on most of the territory,
except for Bryansk region. However, analysis of ground truth data had shown that patterns of this
cluster on the image does not correspond directly to forests on traditional maps but also include
extensive wetlands covered with thin pine and birch-aspen forests that are classified on traditional
maps as non-forested territories. This cluster also includes, as it was described earlier, some
agricultural lands in almost non-forested southern areas. Apparently this cluster determines
relatively wet, poorly drained environments that are covered with pine forests and wet thin forests
in the north and center of study area and non-forested wet agricultural lands.
Finally, «Non-vegetated areas» cluster is clearly determining urban territories. Due to this all
the major cities are easily defined on the image, for example Moscow, Tula, Tver, Yaroslavl,
Vladimir, Bryansk, Kaluga, Volgograd, etc. Besides of that some agricultural lands have fallen
into this cluster as well. Those are gardens and vegetable plantations. This is demonstrated by
wide spread of this cluster on Oka-Don Height and gravity of this cluster to large urban
agglomerations that are usually circled by so-called «garden belt». It must be mentioned that
determination of these territories as «non-vegetated» may be caused by specific dates of data
collection: in April crops are not yet vegetating, in June they are not as developed as natural
vegetation and cereals and in September they are already harvested.
Thus, itemising selected cluster categories we may describe them as following:
1. water bodies
2. dry agricultural lands
56
3. dry forests
4. wet forests and agricultural lands
5. low-vegetated urban and rural lands
Correction of published cartographic data on land cover by results of this study
Comparison analysis of classified image with ground truth data had shown that on most of the
study area patterns of major land cover types coincide. This allows us to conclude that
accomplished classification is reliable for regional study of land cover state and dynamics.
However, there are sites where cluster patterns disperse with available data from traditional
sources.
Cluster image has several significant advantages: first of all it is compiled by impersonal
computer technique, that is the same for the whole study area; then it shows up-to-date
information comparing to traditional maps. Differently from traditional maps, they are less
affected by subjective factors related to procedures of ground data collection and generalisation.
Moreover, «age» of information on traditional maps takes decades. This means that cluster
composite images may be successfully used to verify and correct landscape meaning of land
cover categories on traditional maps. Comparing traditional forest maps with composite cluster
image some significant shortcomings of the first could be noticed:
1. These maps often show thin forests as non-forested areas due to low economic value of
wood
2. Contours on traditional maps often do not show real correspondence of forests and nonforests due to manual generalisation algorithm while they usually correspond to variations in
soils, relief and land use and thus show landscape characteristics of the territory.
3. Traditional forest maps often distort species composition to improve economic value of
different forests. Due to this many of mixed forests are marked as pure coniferous, while thin
forests on wetlands are not even counted as forests.
4. Even most recently verified hand-written maps are outdated for decades already,
57
especially during the time of economic restructuring of agriculture.
Time gap between real changes in agriculture and land cover and traditional mapping process
explains faults of topographic and land use maps and thus landscape and ecological maps as well.
Here is just a few examples:
These maps do not trace overgrow of brushwood on place of abandoned arable lands in areas
where lands are degrading.
From the other side traditional maps do not catch degradation of forest cover related to
industrial clearance and agricultural expansion.
Continuing incursion of urban, rural, transport and industrial lands over arable and forested is
not clearly determined on traditional maps as well.
It is evident that all described faults of traditional maps may be dislodged by application of
time series of remotely sensed data classified using proposed technique.
Determination of major tendencies of land use/ cover dynamics
Mutual analysis of traditional and remotely sensed data allows to determine the following,
sometimes contradictory, patterns of land use/cover dynamics:
1. Extemporaneous recovery of forests
2. Anthropogenic recovery of forests
3. Controlled deforestation around agglomerations and transport channels
4. Changes in species composition of forests (through clearance of valuable wood)
5. Urban development over agricultural lands
6. Waterlogging of arable and forest lands next to wetlands
All the processes mentioned above together with industrial and agricultural pollution overlay
with real diversity of natural and anthropogenic processes to form actual dynamics of land
use/cover and may to some extent be determined by classification of time series of space
imagery.
58
Application of other images
Additional information about land cover may be obtained from original NDVI images. For
example, winter image (February 1993) clearly determines forested and open areas with their
internal geometrical structure. This, in particular, helped to determine that areas in Oka and
Desna valleys determined on cluster classification as «coniferous forests» were open arable lands.
Very specific information may be obtained from original thermal imagery for different time
periods. In particular, April image shows contrast in thermal regime of major landscape
provinces. This disclose that these provinces are characterised not only by similar genetic
structure, but also seasonal dynamic, that come out in alike regime of snow accumulation and
melting.
Extremely interesting are «warm islands» on place of major industrial centres and cities - the
biggest one is in place of Moscow, that give ground to attempts to develop procedures of
automatic detection of thermal anthropogenic pollution through use of moderate resolution
imagery like AVHRR.
Local level
Two case studies were analysed in detail using very high resolution false colour images.
Preliminary experiments with test sites in Mozhaisk test site of Moscow region (mixed forest)
and Kulikovo Pole (forest steppe) have turned to be antipodes in terms of interrelationships
between natural and human components in their landscape structure.
In Mozhaisk test site natural landscape differentiation is emphasised by anthropogenic (mostly
agricultural) activity. Land use structure well fits into the natural landscapes pattern. In-field data
should be used however for further analysis of this idea.
In Kulikovo Pole test site region natural landscape pattern is mainly masked by long-term
intensive agricultural activity (mostly collective form of agriculture in Soviet period). Land use
boundaries do not follow natural landscape structure which never was taken into account while
planning economic specialisation of different parts of the area. Some natural features of the
region (like total spreading of cover loams and homogenous relief) promoted the process of
59
hiding natural landscape structure.
For local level some experiments were carried out on vegetation large- scale mapping in
North-Western part of the EPR, ground- truth of data, as well as comparison and unification of
different- scale land cover classification systems (Table 7).
For example, vegetation map in 1 : 25 000 scale is compiled for the area of 18 km 2. It is the
index plot situated on the islands of Gulf of Finland south-west of Vyborg. Vegetation cover of
this region is typical for coastal middle-taiga landscapes of the Karelian Isthmus. This plot is also
standard owing to the factors of anthropogenic impacts on the vegetation : forest felling, fires,
melioration, arable lands, recreation pressure.
Dependence between the types of plant communities and the character of anthropogenic
transformations is determined. Distinctions between the impact of upper and ground fires, clear
and selective felling are revealed. Time of the impact and corresponding period of
reestablishment is also of great importance. There are two gradations for felling (more than 40-50
years ago and contemporary ones) and fires (more than 10 years ago and the last years).
Analysis of the map lets to define the ratio between the areas with different degree of
vegetation disturbance and to estimate the present ecological potential of this region. Vegetation
cover of two largest mapped islands differs in degree of human impact.
Pine forests, strongly and average transformed by forest felling and fires, predominate in the
Vysotskiy island. Long-term stable spruce forests and secondary mixed forests of birch, spruce
and pine with rather strong edificator role of spruce are typical for the Krepysh island where
settlements are absent and there are no present forest felling and fires.
Potentialities of vegetation restoration are revealed and dynamic series of plant communities
are worked out. This results let us make the prognosis of vegetation regeneration under different
human impact.
Analysis of potential of aerial photographs for interpretation of vegetation and landscapes
made for this region allowed us to draw some preliminary conclusions.
60
Table 7 Physiographic features Land-cover test sites
Name of index-plot
Locatio
Geographic
n,
al zone,
Coordi
subzone
Climatic conditions
Prevalent types of
Prevalent types of
Degree of vegetation
Historical
Comparison with the maps: 1.
soils
plant communities
transformation
peculiaritie
Vegetation of European part of the
s of land
USSR (1974) S. 1: 2 500 000
nates,
use
2. Reconstructed vegetation cover of
absolut
Central and Eastern Europe (1989)
e
height,
the area
of the
plot
I.
Vuva-river—
Salnye
tundry
(Lapland biospheric
reserve)
Kola
North taiga,
peninsu
mountain
la
birch
6805’6810’
n.lat.
crooked
forests
For the taiga zone:
t (av.winter) - 10
illuvial-
Spruce
bilberry-
humus and illuvial-
empetrum
ferruginous soils in
birch
t (av. summer) +15;
taiga zone;
empetrum
annual presipitation –450-500mm;
mountain tundra soils
mountain
tundra
Podzolic
duration of vegetal period – 132
3117’-
days (24 May-02 Oct.)
3124’
For mountain tundra:
(dark podbury).
Moderate
1: north taiga light spruce forests with
forests predominate;
forest
dwarfshrub cover in the north part of
bilberry–
partly
felling and
the plot; north taiga pine forests in the
crooked
birch
deer
south part of the plt (don’t correspond
grazing
to our data); crooked forests are absent
forests;
Primary
spruce
secondary
forests
after
forests;
fires;
low birch thickets
partly
(yernik)
tundra is transformed
and
at the map (because of the scale);
dwarfshrub
dwarfshrub and yernik tundra.
cowberry-licken-
in licken (Cetraria)
2: light spruce forests with dwarfshrub
moss tundra
tundra
and mosaic lichen-moss cover; birch
e.long
under
deer
grazing
t(av.winter) -11
crooked forests; dwarf shrub and shrub
tundra;
Habs.
200-
t (av. summer) +12;
850 m
annual presipitation 450mm;
S=
duration of vegetal period – 89 days
95
km2
North-
Middle
Krepysh
East
taiga
part of
subzone
islands
Finland
Gulf
6034’6040’
forests
in
south
part
don’t
correspond to our data.
(09 June–05 Sept.)
II. B.Vysotskiy and
(Leningrad region)
pine
t (av. winter) - 8,4
t (av.summer) +17,6
 active t 1787
annual precipitation 785 mm;
duration of vegetal period – 166
Weak- and medium-
Pine
podzolic
moss forests;
ferruginous
soils;
illuvial
sandy
peat-podzolic
and peat-gley soils
Mixed
dwarfshrub-
pine-birch-
spruce moss forests
Near 80% forests are
Till
transformed by fires
years
dwarfshrub forests, often long-term
and forests felling;
Finnish
stable on the place of spruce forests
20% of the area is
occupied
secondary
by
small-
40th
arable lands
using
for
growing
folder and
1: middle- and south taiga pine moss
2: spruce forests with dwarfshrub-grass,
locally moss cover.
n.lat.
days
leaved forests;
grain crops
2830’-
present
agricultural
2840’
lands occupy small
e.long
areas (mainly haymeadows); selective
Habs.
felling is everywhere.
were
widelydestributed
on
the
islands.
0-50 m
S=18
km2
III. Primorsk region
Coastal
Middle
area in
taiga
north-
subzone
t (av. winter) -8,6
part of
t (av.summer) +17,0
annual precipitation 811 mm
Finland
and
peat-podzolik
 active t 1701
east
Weak-podzolic
Pine
dwarfshrub-
moss forests;
illuvial-ferruginous
sandy
soils;
Pine-spruce
moss
forests
cultivated sod gley
Old
1: middle-and south taiga pine moss
forests
are
ovegrowing
dwarfshrub forests, often long-term
transformed by fires
arable lands
stable on the placed of spruce forests.
and forest felling;
(mainly
Near 30% of the area
soils
is
duration of vegetal period – 167
Gulf
The most part of
occupied
secondary
days
by
small-
6020’-
leaved
forests;
6023’
present
agricultural
n.lat
lands
(mainly
pastures
2840’-
2: pine dwafshrub, lichen, moss forests.
Finnish)
which were
used
for
growing
folder and
grain crops.
and
haymeadows),
2850’
settlements and fur-
e.long.
breeding farm occupy
Habs.
10% of the area;
0-50 m
selective felling is
everywhere,
S - 30
clear
cuttings of the last
km2
years are also take
place.
IV. Kurgal natural
Kurgal
South taiga
reservation
peninsu
subzone
la
in
south
part of
t (av. winter) - 11
Podzolic
illuvial-
ferruginous
t (av.summer) +16
 active t 1700
soil;
humus-podzolic
gley;
peat-podzolic
Pine
dwarfshrub-
moss forests
Spruce
bilberry-
moss forests,
Near 30% of the area
Formly in-
1: arable lands, fallows, low woods,
is
tensive
birch forests on the place of south taiga
secondary
agriculture,
spruce and pine forests;
coniferous-birch
timber
occupied
forests;
10%
by
-
2: spruce dwarfshrub-grass; pine grass-
cutting,
62
Finland
Gulf.
5944’-
annual precipitation 600 mm
duration of vegetal period 175 days
soils
Spruce forests with
agricultural
broad-leaved
(kitchen-gardens,
industry
hay-meadows).
were
trees
and nemoral grasses
lands
5947’
Forests are in the
n.lat.
regime
of
natural
fishing
dwarfshrub, locally moss forests.
developed
at this area
reservation.
2804’2813’
e.long.
Habs.
15-25
m
S = 35
km2.
63
2. Landscape pattern analysis in forest and forest-steppe zones of Russia using remote
sensing (Lioubimtseva E.Yu.)
Efficiency of application satellite images to studying and mapping landscape pattern depends
on several factors, such as landscape heterogeneity, size and shape of landscape patches and
hierarchical level of landscape aggregation. The choice of spatial resolution of remote sensing
data suitable for landscape mapping should correspond to spatial extension and scale of
ecological problem or phenomenon of interest.
In the nature landscape is always heterogeneous to some extent at any spatial level or scale but
while it is remotely sensed and observed with a given spatial resolution we always have to deal
with a certain level of abstraction assuming homogeneity at a given scale. Configuration and land
cover content of landscape pattern are important criteria of landscape zoning at any scale since a
pattern of landscape reflects ecological functions and dynamics of the landscape.
The results of landscape interpretation at three different scale in forest and forest-steppe zones
of European Russia showed that configuration features of landscape pattern (shape of patches and
matrix porosity as well connectivity of patches by corridors), whose interpretation may be
extremely important for understanding horizontal flows of matter and energy within and among
landscapes, may be determinant criteria for choice of an adequate resolution. More complex is
the shape of landscape patches, higher is porosity of landscape matrix and more patches are
connected by corridors at a given scale of interest, higher spatial resolution is needed to detect
landscape pattern. At the same scale of study a landscape, whose patches have typically shape
index closer to 100 is less demanding in terms of resolution than a landscape with sophisticated
shape of patches (whose index is closer to 0).
As a general rule for most patch types shape index has inverse proportion with size but not
always. Patch size itself should be also considered for choice of resolution. As it was shown by
examples coming from different landscape s at different scales - far bigger is typical patch size
compare to size of pixel, better the chosen resolution is adapted to mapping landscape's
heterogeneity. When a patch size is far below the resolution of image - landscape will be mapped
as homogenous. When typical patches have dimension comparable with resolution of the image -
64
landscape can not be interpreted.
Homogenous model of heterogeneous landscape can be useful at a given scale of study when a
certain level of generalisation is needed for landscape classification by one or several dominant
land-cover types of patches. For studying internal structure of landscape it must be represented on
the image with sufficient heterogeneity, that means that a pixel size should be small enough
compare to the smallest meaningful patches.
Finally, hierarchical level of landscape mapping is a very important factor of choice of scale of
mapping and adequate spatial resolution of satellite data. At any hierarchical level from
microregional to global landscape can be always understood as a combination of several patch
types with particular configuration and pattern. Higher is the hierarchical level, broader is the
scale - more complex is heterogeneity of a landscape, because elements of its internal structure
carry their own internal heterogeneous structure of the subordinated hierarchical level with
patches of the next lower hierarchical level and so on. Thus for the purposes of broad scale
mapping it is useful to "be deliberated" of irrelevant details on landscape pattern of the lower
hierarchical levels, which become a "noise" for landscape mapping. Choice of resolution coarser
than a size of landscape patches out of interest of the levels below than the level of study is very
useful in this case because it allows to find the optimum level of abstraction in heterogeneity of
the landscape.
3. Analysis of land cover transformation on EPR in 1970-1992 (Kazantsev N.N., Yanvareva
L.N., Tcherkashin P.A.)
Two maps were overlaid using OVERLAY module in ARC/INFO to produce map of
difference in forest contours on the maps:
1. Forests map derived from AVHRR data
2. Traditional map of forests
15 years occurred between the maps compilation while all the dramatic changes have taken
place. Only one layer from each map is compared – forest contours (Fig.15). Though land cover
65
on this territory is very fractional due to intensive agricultural use, more than 80% of the total
area on both layers coincide. However, there is a significant shift on both layers in forest/non
forest area interpretation.
Evidently, differences in contours of the maps may be caused by the following factors:
1. Erroneousness in data collection or processing, overlay operations, projection
transformation, etc. Most of these differences may be determined heuristically.
2. Differences in understanding what is forest. Ambiguity arises with respect to the
definition of forest (Forest map of Europe, 1992). Definitions vary from country to country
and depend on the purpose of the forest inventory. Additionally to that, classification may
rely on physical characteristics of the surface (reflectance, as with remote sensing) or
composition and structure of vegetation cover as with traditional mapping approach. Both are
approximations.
3. Real changes in vegetation cover that took place on the surface. For example, forest
cuts or, from the other side, regeneration of forests.
Apparently, all three factors influence the final picture produced as a result of this work. We
just have to check to what extent these factors may be separated in order to obtain geographically
meaningful information.
A series of on-ground checks took place in key points of disagreement between the layers of
information to determine the factors of disagreement.
These checks have shown that significant changes have taken place in the land cover of the
center of European Russia during the last 2 decades due to changes in agriculture that
investigations like this can reveal. In the framework of this preliminary investigation the
following changes were detected:
Deforestation – there are certain regions, that may be easily interpreted on the attached map,
that traditional maps show as forest cover and more recent remotely sensed images show as open
land. We have investigated only several biggest areas of possible deforestation in the area near
Moscow. The major factor for possible change of land cover in these areas is tremendous level of
66
urban construction that took place in these areas in the last 20 years. After citizens of the Moscow
metropolis obtained a possibility to build a country summerhouse, thousands of small
households, usually from 600 to 2000 sq. meters in size, formed a «dacha belt».
Since according to the law arable lands could not be used for summerhouse construction, this
affected total area of forests. In early 1990s when economic conditions in agriculture were rather
severe, some of the local authorities have sold out significant portions of their forested territories
to summerhouse construction as the only way The most intensive construction took place to the
west of Moscow, where recreational conditions were the best.
Forest regeneration – there are areas within the study territory that had experienced secondary
forest regeneration during the last 2 decades, according to the results of map overlay. Mostly this
process take places in the marginal arable areas and was mostly intensive in early 1990s during
the period of stagnation. At that time agricultural development of the lands that were not fertile
and situated far away from industrial infrastructure was not efficient and these lands were not
cultivated for several years. But these areas are rather small in size and are not very well
determined at regional scale. However, on high-resolution images, these areas are easily
determined by intensive vegetation of newly renovated forested cover.
Comparison of spatial information on vegetation cover from different sources in the
framework of GIS could serve as an important tool for investigation of vegetation cover
dynamics. However, the results of preliminary experiments have to be further investigated in
order to get deep understanding of vegetation cover dynamics in the European part of Russia.
4. First assessment of the specific features of VGT matching the objectives of investigation,
and/or related problems (Lioubimtseva E.Yu. and Tcherkashin P.A.)
The results of the first phase of research revealed several problems seriously and inevitably
affecting data quality.
This study is the first experience of use data from AVHRR-NOAA instrument in conjunction
with photographic imagery from Russian satellites (RESURS F series) for land-cover mapping of
whole of Central Russia, some technical problems are related with limitations of these data. Most
67
of them, however could be solved, while using data from VEGETATION instrument.
One of the most serious problems is combination of AVHRR 1 km-resolution data with highresolution from RESURS and data scaling. It was shown already by a number of researchers
(Malingreau & Belward, 1991; Goossens et al., 1995) that such jointure of high and low
resolution is necessary for validation of coarse-resolution data. A major advantage of using
VEGETATION data is possibility of their direct combination with high-resolution data of the
HRVIR instrument, carried on the same platform and allowing direct combination of high and
low resolution views of the same target. The VEGETATION approach would provide support to
statistical sampling procedures of several scale and resolution levels.
Although we are using AVHRR data from the Global data set, which were already
geometrically corrected, not all of them have a quality directly suitable for land-cover mapping
application. Geometric distortion of these data varies: on the western part of Russia close to atnadir images are available, while data on its eastern part and especially Siberia, where there is no
LAC AVHRR data until now, suffer of strong distortion, which does not allow to use this data
without addition correction. The later requires numerous resampling of images inevitably
bringing high RMS error. This problem exists because global AVHRR data set was assembled by
collection of data from a large number of widely dispersed receiving stations all over the world
(Eidenshink & Faundeen, 1994). The VEGETATION Programme approach would bring a
solution of this problem thanks to centralised acquisition and formatting of VEGETATION data
(Malingreau, 1995). This will assure production of global data set of uniform high quality.
Mapping and analyses of vegetation cover, its seasonal dynamics and evolution, was based on
analyses of NDVI (Normalised Difference Vegetation Index), calculated as normalised difference
of the calculated as normalised difference of channels 2 (0.725-1.1 m) and 1 (0.58-0.68 m).
Photographic data from MK-4 instrument (RESURS-F), which were used on test sites, have three
visible and near- infrared channels (four operationally selected from six available spectral bands
(0.460-0.505,
0.515-0.565,
0.635-0.690,
0.810-0.900,
0.400-0.700,
0.580-0.890
m).
Composition of spectral channels of VEGETATION instrument (RED (0.61-0.68), NIR (0.780.89), SWIR (1.58-1.75) and BLUE (0.43-0.47) are better adapted to characterisation of the
68
vegetation cover and its dynamics. We expect that experimental BLUE channel would first,
considerably improve atmospheric correction of data, and second, will add new capabilities of
mapping coniferous boreal forest vegetation, which is a dominant biome throughout northern
Russia. Onboard channel calibration of VEGETATION data also makes them considerably more
useful than visible and infrared channels of AVHRR.
Moreover, automatic co-registration of VGT and HRVIR data may find promissing
application in landscape hierarchical modelling, and in particular - heterogeneity modelling,
where scaling aspect is important. Geostatistical analysis of co-registered landscape frames
derived from imagery of different spatial resolution will help to understand relations between
different scales of landscape fragmentation.
Processing of AVHRR imagery over most part of European Russia usually meets difficulties
caused by cloud contamination. In fact cloud contaminated pixel screening of 10-day data is
much more difficult than cloud screening on one-day images. This is because compositing
algorithm currently applied to AVHRR data in the global data set mask or distorts the signal so
that it is almost to identify clouds in visible and NIR channels. Analysis of thermal channel 4
partly solves the problem of cloud screening. It is essential that this problem should be solved in
VGT 10-day product, especially because thermal channels are absent in VGT instrument.
FUTURE WORK PLANNED FOR THE POST- LAUNCH PERIOD (MILANOVA E.V.)
The results of the pre-launch phase allow us to formulate our tasks and requirements for the
post-launch period. The experience acquired during the preparatory period will be used in order
to test efficiency of VEGETATION data for land-cover/use and landscape applications. The
following VEGETATION products will be required by our team for further research:
Series of VGT 10-day composites (all bands) during one year (full coverage on European part
of Russia) ;
one-day composites for the same area for selected periods;
co-registered low-resolution and high-resolution VEGETATION and HRVIR scenes for five
selected sample areas.
The following research activities are planned during the post-launch period:
More accurate small-scale land-cover mapping using new VGT data and experience of
AVHRR and RESURS interpretation;
69
Analyses of land-cover phenological dynamics using VGT data; validation through GIS-stored
ancillary data;
Analyses of land-cover/use trends using new VGT data and experience of AVHRR
interpretation and GIS ancillary data ;
Feasibility study of application VGT/HRVIR simultaneous data for hierarchical landscape
fragmentation.
70
BIBLIOGRAPHY
Kalibernova N. M., Schukin A. K., Lokal level integrated landscape management in Boreal
Forest, Regions: Kurgalsky Peninsula studies // Intern. Boreal Forest Research Assoc. USA (in
press in English).
Kholod S.S. Coenotic approach in study spatial heterogeneity of tundra vegetation // Botanical
journal. T. 82. N 8. 1997, 48-62. (in Russian)
Khramtsov V. N., Volkova E. A., Kalibernova N. M. An experience in compiling
enviromental maps using program support of geoinformational systems (GIS) // Computer data
bases in botanical. St-Petersburg, 1997, 100-107 (in Russian)
Lioubimtseva E. – Landscape and land-cover mapping of European Russia, Report to OSTC,
July 1997, UCL- Louvain-la-Neuve, Belgium
Lioubimtseva E., 1998. - Interpretation and Mapping Landscape Pattern in Forest and ForestSteppe Zones of Russia Using RESURS-F/MK-4 and RESURS-01/MSU-SK Satellite Imagery,
Proceedings of the 24th Annual Conference of the Remote Sensing Society, Greenwich, UK
Lioubimtseva E. 1998, Interpretation and Mapping Landscape Pattern in Forest and ForestSteppe Zones of Russia Using Remote Sensing, Proceedings of the European IALE Congress,
UK
Lioubimtseva E., Milanova E.V. and Tcherkashin P.A. - Applications of Remote sensing to
Land Use/ Cover Mapping of Russia - Abstracts of the 28th Intern.Geogr.Congress «Land, Sea
and Human Impact» 6-11 Aug.1996, Hague
Lioubimtseva E. and E.V. Milanova -Remote Sensing and GIS Applications to Landscape
Mapping of Central Russia. – Abstracts of Intern.Conf.of Land Degradation, June 1996, Adana,
Turkey
Lioubimtseva E. and P.Defourny - Landscape Mapping of European Russia using GIS
Knowledge-based Approach - submitted to Landscape and Urban Planning, 1998
Lioubimtseva E. and P.Defourny - Remote sensing applications for landscape mapping in
forest zone of Russia (in preparation)
Lioubimtseva E.Yu., Yanvareva L.N., Kazantsev N.N., and E.V.Milanova, 1998. Landscape
Status and Changes in the European Part of Russia, Proceedings of the European IALE Congress,
UK
Milanova E.V., N.N.Kazantsev, E.Yu.Lioubimtseva, A.V.Kushlin. Landscape approach to
71
studying land use/land cover changes: mapping and GIS . In: Global Change and Geography.The
IGU Conference, Moscow,Russia,August 14-18,1995.Abstracts. Moscow,1995, p.237
Milanova E.V., V.N.Solntsev, N.N.Kalutskova,P.A.Tcherkashin 1995 Landscape studying and
assessment through remote sensing technique (NOAA AVHRR indices). Landscape studying and
assessment through remote sensing technique (NOAA AVHRR indices): case study in Central
Russia. In: Global Change and Geography.The IGU Conference, Moscow, Russia, August 14-18,
Abstracts. Moscow, 238
Milanova E.V. Landscape methodology of land use/cover change assessment through remote
sensing technique.Taskforce meeting on Modelling Land use and Land cover Changes in Europe
and Northern Asia. Abstracts, IIASA, Laxenburg, 3-5 April 1995
Milanova E.V. 1995Landscape approach to regionalisation of Russia for planning and
development of tourism. In: World Conference on sustainable tourism, Lanzarote, Spain
Milanova E.V. Land cover/Land use mapping and monitoring of Russia.-In: VEGETATION
Preparatory Programme. Kick off meeting. Abstracts and transparencies. IRSA/JRC, 1995
Milanova, E.V., P.A.Tcherkashin and E. Lioubimtseva 1997- Land Use and Land Cover data
validation and requirement of Russia, LUCC Data Requirements Workshop, Barcelona,
November 1997
Milanova E.V. Land Use and Land Cover Change. Sci/research Plan. IGBP # 35/ HDP #7,
1995, 132 pp. (contribution)
Milanova E.V. N.N.Kalutskova, V.N.Solntsev. Geoecological expertise of agrolandscapes on
the base of GIS for Moscow region. In: Environmental change. Ed. A.N. Gennadiev,
E.V.Milanova, M., MSU Publishers,1997, 178- 187
Volkova E.A., Fedorova I.I. Map of ecological functions of vegetation cover of Russia //
Geobotanical mapping. L. 1995, 51-57 (in Russian)
72
73
REFERENCES
Agroclimatic Atlas of the USSR, 1978 GoskomGidromet, Moscow
ARC/INFO, vers.7 for UNIX and Open VMSTM 1995. Understandin GIS, the ARC/INFO
method, ESRI, Redlands, California
ARC/INFO, vers.7 for UNIX and Open VMSTM 1995. Data Management, ESRI, Redlands,
California
ArcView The Geographic Information System for EveryoneTM, ESRI, Redlands, California
Atlas of the Environment and Health of the Population of the Czech and Slovak Federal
Republic, 1992, Brno-Praha
Bazilevich N.I., A.V. Drozdov, and R.I.Zlotin 1993 Geographical features of productive and
destructive processes in the landscapes of Northern Eurasia, Izvestia Russian Academy of
Science, vol 4, pp.5-20
Bazilevich N.I. 1995 Biomass and biologic production of vegetation formations of the former
USSR (in Russ.) Biomassa i bioproductivnost rastitelnih formatsij. Moscow: Nauka
Boiko F.F., V.I.Kravtsova and I.V.Kuksa (1991) The mapping of forest distribution dynamics
using space photo and old maps (with reference to Central Tataria), Soviet Journal of Remote
Sensing, vol. 8 (6), pp.973-983
74
Bridgewater, P.B. 1993. Landscape ecology, geographic information systems and nature
conservation. In: Landscape Ecology and GIS, edited by R. Haynes Young, D. Green and S.
Cousins, 23-36. London: Taylor and Francis
Bunce, R.G.H and Heal, O.W. 1984. Landscape evaluation and the impact of changing land use
of the rural environment: the problem and an approach. In: Planning and ecology, edited by R.D.
Roberts & T.M. Toberts, 164-188. London: Chapman and Hall
Cousins S.H. 1993. Hierarchy in ecology: its relevance to landscape ecology and geographic
information systems. In: Landscape Ecology and GIS, edited by R. Haynes Young, D. Green and
S. Cousins, 75-86. London: Taylor and Francis
Delbaere B. and Gulinck H. 1994. A review of landscape ecological research with specific
interest to landscape ecological mapping. In:
Remote Sensing in Landscape Ecological
Mapping, edited by B. Delbaere and H.Gulinck, 3-29 IRSA JRC European Comission
Ecological map of Moscow oblast.1:350 000. M, Ltd «Letopisetz»,1993
Ecological- geographical map of Russia. 1: 4 000 000. Omsk, 1996, 4 sheets
English Nature, 1993. Strategy for the 1990s: Natural Areas, Setting Nature Conservation
Objectives. A Consultation Paper. English Nature, Peterborough
Farjon J.M.J.and Thunissen H.A.M., 1994 Remote sensing and landscape ecology: state of art in
the Netherlands. In Remote Sensing in Landscape
75
Ecological Mapping, edited by B. Delbaere and H.Gulinck, 3-29 IRSA JRC European Comission
Forman R. , 1991 Ecologically Sustainable Landscape: The Role of Spatial Configuration/ In:
Changing Landscapes: An Ecological Perspective, eds: I.S.Zonneveld & R.Forman, pp.261-275
Forman R.T.T. and Gordon M. 1986. Landscape ecology. John Wiley and Sons, New York.
Geological map of RSFSR. 1:2 000 000. M,GUGK,1988
Geological map of Russian platform and its fringes. Ed. D.V.Nalivkin. 1: 1 500 000.L.,
Vsegei,1965
Geologo-ecological investigation in Moscow and Moscow oblast. Series of maps. M, Science
center,1992
Gorbachev B.N. Vegetation map of Rostov region. 1 : 600 000 s. M. 1973.
Gossens R., T.Ongena, E.D'Haulin and G.Larnoe 1993 The use of remote sensing (SPOT) for the
survey of ecological patterns, applied to two different ecosystems in Belgium and Zaire. In:
Landscape Ecology and GIS, edited by R. Haynes Young, D. Green and S. Cousins, 75-86.
London: Taylor and Francis, pp.147-158
Gribova S.A., Lavrenko E.M. Vegetation map of Kanino-Timanskaya and Malozemelskaya
tundra. 1 : 1 000 000 s. M. 1975.
76
Griffiths G.H., J.M.Smith, N.Veitch and R.Aspinall, 1993 The ecological interpretation of
satellite imagery with special reference to bird habitats In: Landscape Ecology and GIS, edited by
R. Haynes Young, D. Green and S. Cousins, 75-86. London: Taylor and Francis, pp.256-271
Gulinck H., O.Walpot, P.Janssens and I.Dries., 1991 The vizualization of corridors in the
landscape using SPOT data In: Nature Conservation 2: The Role of Corridors, ed. D.A.Saunders
& R.J.Hobbs, Surrey Beatty & Sons, pp.9-17
Gulinck H., H.Dufourmont, M.Brunovsky, A.Andries, P.Wouters 1993 Satellite Images for the
Detection of Changes in Rural Landscapes: a Landscape-Ecological Perspective.
EARSeL
Advances in Remote Sensing, Vol.2, N.3-XL, pp.84-90
Gulinck H., O.Walpot and P.Janssens
1993 Landscape structural analysis of central Belgium
using SPOT data. In: Landscape Ecology and GIS, edited by R. Haynes Young, D. Green and S.
Cousins, 75-86. London: Taylor and Francis, pp.129-139
Haines-Young R.H., 1992; The use of remotely-sensed satellite imagery for landscape
classification in Wales. Landscape Ecology, vol.7, n;4,p.253-274
Haines-Young R.H. and Bunce B., 1994 Remote sensing, land classification and environmental
survey, In Remote Sensing in Landscape Ecological Mapping, edited by B. Delbaere and
H.Gulinck, 3-29 IRSA JRC European Comission
Isachenko A.G. 1991. Assessment and mapping of ecologic potential of Russia's landscapes (in
77
Russ.) Ocenka i kartografirovanije ecologicheskogo potenciala landshaftov Rossii. Proceeding of
Russian Geographic Society, vol.6, 457-471
Isachenko T. I. Vegetation map of Moscow region 1 : 1 500 000 s. // Atlas of Moscow region.
M. 1964.
Isachenko T. I, Katenina G. D. Vegetation map of Yaroslavl region. 1 : 1 500 000 s. // Atlas of
Yaroslavl region. M.1964.
Isachenko T. I, Katenina G. D. Vegetation map of Leningrad region. 1 : 1 500 00 s. // Atlas of
Leningrad region. M. 1967.
Karpenko A. S., Shabalina E. A. Vegetation map of Pskov region. 1: 1 500 000 s. // Atlas of
Pskov region. M. 1969.
Kibalchich A.O.(1991) Structural-territorial changes in agricultural land use of European part of
Russian Federation. Izvestia Russian Academy of Science, Vol.2, pp.56-71
Kondratyev K.Ya., A.A. Grigoryev, G.A. Ivanin, and G.A. Putintseva (1993) Maps of spectral
brightness coefficient for the USSR territory and their statistical analysis International Journal of
Remote Sensing, vol. 14, No.3, pp.521-535
Lambin E.F. and A.H. Strahler (1994) Indicators of land-cover change for change-vector analysis
in multitemporal space at coarse spatial scales International Journal of Remote Sensing, vol.15,
No10, 2099-2119
78
Lambin E.F.and D.Ehrlich (1996) The surface temperature-vegetation index space for land cober
and land-cover-change analysis. International Journal of Remote Sensing, vol.17, N3, pp.463-487
Land Use of the USSR 1:4M, 1991. edited by Yanvareva L.N. Moscow State University,
Moscow
Landscape map of USSR. Ed. A.G.Isachenko. 1:4 000 000. M., GUGK,1988, 4 sheets.
Land cover of USSR. Ed. L.F.Yanvareva. 1: 4 000 000. M.,GUGK,1989,4 sheets
Landscape map of USSR. Ed. I.S.Gudilin. 1:2500 000. L, VSEGEI,1985, 16 sheets
Leemans R. and Kramer , 1994 IMAGE Project Global Climatic Data Base
Lipsky Z. and V.Kremsa 1994 Landscape (ecological mapping in the Czech Republic. In:
Remote Sensing in Landscape Ecological Mapping, edited by B. Dlebaere and H.Gulinck, 3-29
IRSA JRC European Comission
Loveland T.R., Merchant, J.W., Ohlen, D.O., and Brown J. (1991) Development of a land-cover
database for the conterminous U.S. Photogrammetric Engeneering and Remote Sensing, 57,
1453-1463
Luri D.I., 1989 Ecotone between forest and steppe as a membrane ecosystem (in Russ.) Bull. of
79
Russian Acad.Science, Geogr., n 6, pp.16-27 Malingreau et al., 1989;
Malingreau, J.P., Tucker, C.J., and Laporte, N. (1989) AVHRR for monitoring global tropical
deforestation. International Journal of Remote Sensing, 10, 855-867
Map of Landscapes of the USSR, 1:2.5 M, 1980. Ministry of Geology of the USSR, edited by
I.S. Goudilin, Moscow
Milanova E.V. A.V.Kushlin, E.Yu. Lioubimtseva and N.N.Kazantsev 1995.- Landscape approach
to studying land use/cover changes: mapping and GIS The IUG Conf. Global Change and
Geography, Aug. 14-18, 1995 Moscow
Meeus J.H.A., M.P. Wijermans and M.J. Vroom 1990. Agricultural landscapes in Europe and
their transformation. Landscape and Urban Planning 18, 289-352
Meeus, J.H.A. 1995 Pan-European landscapes. Landscape and Urban Planning 31, 57-79
Moscow oblast. Pollution of environment,soils. 1:350 000. Cartographic appendix to magazine
«Lik», issue1. M., 1993
Map of USSR.1: 2 500 000. M., GUGK,1984,16 sheets
Map of potential erosion of Non- Chernozem zone of RSFSR (excluding Ural and Zauralie). Ed.
E.M.Sergeeva. 1:1 500 000. M., GUGK,1984,4 sheets.
80
Map of engineer- geological conditions of Non- Chernozem zone of RSFSR (excluding Ural and
Zauralie). 1:1 500 000. M., GUGK,1984, 4 sheets.
Map of pit areas of Non- Chernozem zone of RSFSR (excluding Ural and Zauralie). Ed.
E.M.Sergeeva. 1:1 500 000. M., GUGK,1984,4 sheets
Map of permafrost conditions
of Non- Chernozem zone of RSFSR (excluding Ural and
Zauralie). Ed. E.M.Sergeeva. 1:1 500 000. M., GUGK,1984,4 sheets
Map of soil-geographic regionalisation of Non- Chernozem zone of RSFSR (excluding Ural and
Zauralie). Ed. E.M.Sergeeva. 1:1 500 000. M., GUGK,1984,4 sheets
Map of geomorphologic- neotectonic regionalisation
of Non- Chernozem zone of RSFSR
(excluding Ural and Zauralie). Ed. E.M.Sergeeva. 1:1 500 000. M., GUGK,1984,4 sheets
Map of regionalisationon the base of chemical composition of ground water of Non- Chernozem
zone of RSFSR (excluding Ural and Zauralie). Ed. E.M.Sergeeva. 1:1 500 000. M.,
GUGK,1984,4 sheets
Map of fresh underground water storage conditions of Non- Chernozem zone of RSFSR
(excluding Ural and Zauralie). Ed. E.M.Sergeeva. 1:1 500 000. M., GUGK,1984,4 sheets
Map of agricultural land use of Non- Chernozem zone of RSFSR (excluding Ural and Zauralie).
Ed. E.M.Sergeeva. 1:1 500 000. M., GUGK,1984,4 sheets
81
Map of vegetation protection
of
Non- Chernozem zone of RSFSR (excluding Ural and
Zauralie). Ed. E.M.Sergeeva. 1:1 500 000. M., GUGK,1984,4 sheets
Map of vegetation of Moscow oblast.1:200 000. Ed. G.N.Ogureeva. M., 1996, 4 sheets
Nellis M.D.and Briggs J.M., 1987 The effect of spatial scale on Konza landscape classification
using textural analysis. 1987, Landscape Ecology, vol.2, n 2, pp. 93-100
Nelson, R (1989) Regression and ratio estimations to integrate AVHRR and MSS data. Remote
Sensing and Environment, 30, 201-216
Olson J.S., H.A Pfunder and
Chan Y. 1983. Carbon in life vegetation of major world
ecosystems. Oak Ridge Natl. Lab., ORNL 5862, 152 p.
Petch J.R. and J. Kolejka 1993. The tradition of landscape ecology in Czechoslovakia. In:
Landscape Ecology and GIS, edited by R. Haynes Young, D. Green and S. Cousins, 39-56,.
London: Taylor and Francis
Politico-Administrative Map of the USSR, 1:8M, 1992 GUGK, Moscow
Quattrochi D.A.and Pelletier R.A., 1991 Remote sensing for analysis of landscapes: an
introduction. In: , ch.3 In: Turner & Gardner eds:Quantitative methods in landscape
ecology.Springer Verlag, pp;51-76
82
Remote Sensing in Landscape Ecological Mapping, edited by B. Dlebaere and H.Gulinck, 3-29,
IRSA JRC European Comission
Rybinsk water reservoir:Topographic map.1:200 000. M,GUGK,1993
Soil map of Russian Federation for high school. Ed. I.Yu.Kulbatskaya. 1:4 000 000. Omsk,
Cartographic enterprise,1993
Soils of the former USSR 1996. edited by Glazovskaja M.A., Moscow State University, GUGK,
Moscow
Suchevsky A.G.(1993) - Forest's role in environmental change, Izvestia Russian Academy of
Science, ser.geogr., vol.2, pp.60-72
Townshend, J.R., Justice, C.O., Curney, C., and MacManus, J., (1992) The impact of
misregistration on the detection of changes in land-cover. I.I.I.E. Transactions on Geosiences and
remote Sensing, 30 (5), 1054-1060
Turner M.G.and Gardner R.G., 1991 Quantitative methods in landscape ecology, an introduction:
ch 1 In: Turner & Gardner eds:Quantitative methods in landscape ecology.Springer Verlag, pp.314
Tucker, C.J., Holben, B.N., and Goff, T.E. (1984) Intensive forest clearing in Rondonia, Brazil,
as detected by Satellite Remote Sensing. Remote Sensing of Environment, 15, 255-261
83
Vegetation of the USSR 1:4M, 1991. Komarov Institute of Botany, Russian Academy of Science,
Leningrad
Vegetation map of Europaean part of the USSR and the Caucasus. 1 : 2 000 000 s. (for higher
school). M 1987.
Yurkovskaya T.K. Vegetation map of Karelia. 1 : 2 000 000 s. // Atlas of Karelskaya SSR. M.
1989.
Wiens J.A., 1995 Landscape mosaics and ecological theory. In: Mosaic Landscapes and
Ecological processes. Ed.L.Hansson, L.Fahrig and G.Merriam, Chapman & Hall, London
World Map of Present-Day Landscapes 1992. Moscow State University & UNEP, edited by
Milanova E.V., A.V. Kushlin and N.J. Middleton, Moscow
Zonneveld I.S., 1989 The land unit - a fundamental concept in landscape ecology, and its
applications. Landscape Ecology, vol.3, N.2, pp.67-89
Figure Captures:
1.
Study area and coverage by satellite data
2.
Conceptual model of the landscape hierarchy in the GIS database
3.
Present-day landscapes of European Russia
4.
Present-day landscapes of Central Russia
5.
100-years dynamics of arable lands
6.
100-year dynamics of pastures
84
7.
8.
100-year dynamics of forests
Land-cover proportions at increasingly coarsing spatial resolutions
9.
Shape index and size of landscape patches
10.
Land cover of Moscow area (derived from MK-4 scene)
11.
Landscapes of the Oka-Don waterdshed (derived from the MSU-SK scene)
12.
Three cases of landscape patchiness versus spatial resolution: a) one patch (matrix) is
far over grid cell - landscape is detectable by land cover b) patch size is comparable
with grid cell - landscape is not detectable by land cover; c) patch size is far below
grid size - landscape is detectable by land-cover).
13.
Land cover of Northern Eurasia (derived from the AVHRR composites)
14.
Land cover of Central Russia
15.
10-year dynamics of forests
Tables:
1.
Thematic maps used in the research
2.
3.
Satellite data used in the research
Landscape elements at different spatial resolutions
4.
Selected land cover and landscape categories of Central Russia
5.
Sensors’ suitability to landscape mapping at different levels:
6.
7.
Phenological characteristics of land cover
Physiographic features of land-cover test sites
85
Download