the influence of past human activity gradients on present ...for the whole park and chosen study...

14
The influence of past human activity gradients on present variation of NDVI and texture indices in Zabory Landscape Park M. Kunz Nicolaus Copernicus University, Faculty of Biology and Earth Science, Institute of Geography, Torun, Poland A. Nienartowicz Nicolaus Copernicus University, Faculty of Biology and Earth Science, Institute of Ecology and Environment Protection, Torun, Poland Keywords: NDVI, forest landscape, texture indices, fractal dimension, Tuchola Forest ABSTRACT: Landscape structures can be analyzed in a few ways using different methods, researching tools, and source data. One of these methods is remote sensing and imageries gained by it. This technology is successfully used in landscape ecology to estimate the forest landscape structure modified by economic activity of man. To estimate those changes landscape ecology creates many indicators and spatial structure measures and they are estimated on the basis of satellite imageries. The present state of the landscape is the result of its preceding states and is modeled as well by man as by natural environmental processes. The structure of forest landscape was analysed on the basis of Landsat satellite imageries of the Zabory Landscape Park from 1975–2003. NDVI index was calculated for the whole park and chosen study areas within this territory. On the basis of images presenting spatial variation of NDVI, the following parameters of the texture were defined: diversity (H), fragmentation (F), fractal dimension (FD). In the calculation of diversity and fragmentation indices, the size of area kernel was 5 5 pixels. The fractal dimension was calculated with the Triangular Prism Method (TPM). NDVI index was also calculated for chosen sample plots. These were squares with 300 300 meters sides representing all age classes of trees growing in the habitat of fresh and dry forest. The created data base concerning the subsection of surface division, and reproduction of reforestation history of this area from 1796, allowed for detailed analysis of results with reference to the use of this area in the past. In a comparative analysis of three study areas on the basis of structure indicators, the graphic method of Jentys- Szaferowa was used, and methods of numerical taxonomy. 1 INTRODUCTION Human activity is the main factor causing changes in the structure of ecosystems and landscapes. In the forest landscape analysis historically researched, the influence of this factor is observed among others on changes in plant types, ecosystem ages, and species composition, congestion and dendrometric parameters of trees in it total, utilizable and green biomass, and canopy structure. Spatial changes in plant structure caused by 171 New Developments and Challenges in Remote Sensing, Z. Bochenek (ed.) ß2007 Millpress, Rotterdam, ISBN 978-90-5966-053-3

Upload: others

Post on 24-Jan-2021

2 views

Category:

Documents


0 download

TRANSCRIPT

  • The influence of past human activity gradients on presentvariation of NDVI and texture indices in ZaboryLandscape Park

    M. KunzNicolaus Copernicus University, Faculty of Biology and Earth Science, Institute of Geography,Torun, Poland

    A. NienartowiczNicolaus Copernicus University, Faculty of Biology and Earth Science, Institute of Ecology andEnvironment Protection, Torun, Poland

    Keywords: NDVI, forest landscape, texture indices, fractal dimension, Tuchola Forest

    ABSTRACT: Landscape structures can be analyzed in a few ways using differentmethods, researching tools, and source data. One of these methods is remote sensingand imageries gained by it. This technology is successfully used in landscape ecologyto estimate the forest landscape structure modified by economic activity of man. Toestimate those changes landscape ecology creates many indicators and spatialstructure measures and they are estimated on the basis of satellite imageries. Thepresent state of the landscape is the result of its preceding states and is modeled aswell by man as by natural environmental processes. The structure of forest landscapewas analysed on the basis of Landsat satellite imageries of the Zabory LandscapePark from 1975–2003. NDVI index was calculated for the whole park and chosenstudy areas within this territory. On the basis of images presenting spatial variation ofNDVI, the following parameters of the texture were defined: diversity (H),fragmentation (F), fractal dimension (FD). In the calculation of diversity andfragmentation indices, the size of area kernel was 5� 5 pixels. The fractal dimensionwas calculated with the Triangular Prism Method (TPM). NDVI index was alsocalculated for chosen sample plots. These were squares with 300� 300 meters sidesrepresenting all age classes of trees growing in the habitat of fresh and dry forest. Thecreated data base concerning the subsection of surface division, and reproduction ofreforestation history of this area from 1796, allowed for detailed analysis of resultswith reference to the use of this area in the past. In a comparative analysis of threestudy areas on the basis of structure indicators, the graphic method of Jentys-Szaferowa was used, and methods of numerical taxonomy.

    1 INTRODUCTION

    Human activity is the main factor causing changes in the structure of ecosystems andlandscapes. In the forest landscape analysis historically researched, the influence of thisfactor is observed among others on changes in plant types, ecosystem ages, and speciescomposition, congestion and dendrometric parameters of trees in it total, utilizable andgreen biomass, and canopy structure. Spatial changes in plant structure caused by

    171

    New Developments and Challenges in Remote Sensing, Z. Bochenek (ed.)

    �2007 Millpress, Rotterdam, ISBN 978-90-5966-053-3

  • anthropogenic factors are determined by changes caused by topographic and habitatfactors. The current landscape structure is the result of its previous conditions and iscreated by man as well as by natural processes.

    Landscape structures can be analyzed in a few ways using different methods,researching tools, and source data. One of these methods is remote sensing andimageries gained by it. This technology is successfully used in landscape ecology toestimate the forest landscape structure modified by economic activity of man. Toestimate those changes landscape ecology creates many indicators and spatial structuremeasures and they are estimated on the basis of satellite imageries.

    Altobelli et al. (2001) state that a good index of human influence on a vegetationcover and at the same time on the structure of a landscape is a fractal dimensioncalculated on the basis of NDVI map created according to satellite imagery.According to these authors also the NDVI values of separate pixels, or mean NDVIvalues of systematically distributed areas covering many pixels, show the intensity ofhuman influence on vegetation and landscape very well. It is concluded from ourprevious researches (Kunz et al. 2000) that diversity and fragmentation indices are agood measure of past as well as contemporary human influences on the structure ofthe landscape. In this paper these indices were used to characterise the landscapestructure of the Zabory Landscape Park. Spatial and time variation of NDVI wasanalysed in relation to the history of the use of this area read from old topographicand economic maps, textual cataloguing data, and the series of Landsat satelliteimageries.

    The aim of this research was to define how places exploited agriculturally withvarious intensity in the past, in spite of afforestation, influence spatial variety of greenbiomass value, and heterogeneity of the contemporary landscape.

    2 STUDY AREA

    The research took place in the area of the Zabory Landscape Park (ZLP) in the Bory-Tucholskie natural district, covering one of the biggest forest complexes in Poland. Thearea is situated in Northern Poland in Pomeranian voivodship. By order of theadministration this is the area of Chojnice and Brusy communes, and it is situatedrespectively to the north and west from these cities (Figure 1). Today forests (71%) andaquatic ecosystems, mainly rivers and lakes (13%) are dominating in the area of thepark. This area was finally shaped by the youngest Baltic glaciation. This is why verydifferentiated post-glacial forms of the sculpture of the Earth surface are also here asvalleys, channels, and small lakes, which together with hummocks of dunes create afully differentiated landscape.

    The spatial variability of landscape created by natural factors is covered byheterogeneity introduced by 130 years of intensive forest economy which took place inthis area. Forests are managed by The Regional Management of State Forests in Torun.In spite of strong environmental changes of forest communities, the nature of this area isof such good values that in 1996 in the southern part of the landscape park the NationalPark ‘‘Bory Tucholskie’’ was created of the area of about 48 km2. Except for the nationalpark, there are six nature reserves in the researched area.

    172 M. Kunz & A. Nienartowicz

  • The influence of past human activity gradients on present variation of NDVI and texture indices 173

    Figure 1. Situation of study area (A – Kruszyn study area, B – Laska study area, C – PNBT

    study area).

  • The subject of analysis is the area of the whole Zabory Landscape Park. In itsseparate parts in the past, forest economy was realised with various intensity (Figure 1).To estimate the influence of forest exploitation on spatial changes in the landscape, threebig study areas have been chosen in the gradient of human activity:

    * the land on the north from the Zbrzyca river in the region of Kruszynskie lake, whereat the end of 19th century poor sheep pastures were dominating; those areas werereforested after the creation the Prussian Oberforstereich Zwangshoff and in the nextfew years it was the area of the most intensive forest activity in the whole researchedarea (Kruszyn study area),

    * the land on the north from the line of Witoczno–Lackie–Dybrzk lakes;according to data from old topographic maps there were always forestcomplexes in the past; the area is mainly covered by big nature reserves andthe oldest treestands (Laska study area),

    * the land situated within the ‘‘Bory Tucholskie’’ National Park, the exploitationof forest products has been very limited here since 1996 (PNBT study area).

    Each study area is in the shape of a square with a side of 5 040 meters. It is built of168� 168 pixels (total 28 224 pixels 30� 30 metres). Areas delimited in this way arethe basis for analysis of texture indices. 25 test areas (sample plots) measuring300� 300 meters have been chosen in each study area (for Kruszyn study area 26 testareas have been chosen). They represent different age classes of treestands, and twobasic types of habitats; these are a fresh forest and a dry forest.

    3 METHODS

    In the presented papers, satellite imageries Landsat registered in 1975–2003 (Table 1)are the main sources of information about time and spatial changes of a landscapestructure. Imageries were processed and maps were created with GIS technology withthe use of ArcView program together with its extensions, IDRISI and MatLab. Systemsof processing imageries (Image Analyst, IDRISI) made it possible to manage interactiveoperations on satellite imageries. It was particularly making a radiometric correction,geometrization of imageries, localization of researched areas in imageries, definition ofelectromagnetic radiation reflection to different treestands (spectral response) and otherelements of the landcover.

    174 M. Kunz & A. Nienartowicz

    Table 1. List of used satellite imageries.

    Data Satelite Scene No

    30.05.2003 Landsat ETMþ 190-2205.05.2000 Landsat ETMþ 190-2328.07.1990 Landsat TM 191-2205.04.1981 Landsat MSS 206-2212.10.1979 Landsat MSS 205-2226.05.1978 Landsat MSS 205-2209.09.1975 Landsat MSS 205-22

  • For the whole park and chosen study areas NDVI was calculated (Table 2) accordingto the formula proposed by Rouse et al. (1973):

    NDVI ¼ ðIR� RÞ=ðIRþ RÞ;

    where IR means near-by infra-red radiation, and R – red radiation.The list of the used canals for an NDVI calculation for different types of scanners is

    in Table 2.In the case of the park and study areas, calculations were done twice. Except NDVI

    calculated at the level of the whole study area, this indicator was also defined withouttaking into account the area of lakes and marshlands.

    On the basis of the received NDVI indices, diversity (H), fragmentation (F) andfractal dimension (FD) of separate study areas were calculated. Calculating two firsttexture data kernel 5� 5 pixels was used. Each area was characterised by average valuesof H and F. These were arithmetic averages from values gained for the group of squares,whose central area was each pixel of the researched area in the sequence.Diversity of the researched part of the landscape was calculated using Shannon’sformula:

    H ¼ �X

    Pi ln Pi;

    where pi means the share of i – this type of pixels (i ¼ 1; 2; 3; . . . ; n), n – number ofpixel categories in the square (i.e. 25 pixels).Fragmentation was calculated according to the formula:

    F ¼ ðn � 1Þ=ðc � 1Þ;

    where n means, much the same as in the previous formula, number of pixels categories,and c – number of pixels in the square (25 pixels).

    The detailed description of the above methods was written by Monmonier (1974),Murphy (1985) and Turner (1989).

    To calculate a fractal dimension, the so called Triangular Prism Method (TPM) wasused. This method was proposed by Clarke (1986) and Jagii et al. (1993). It was used byamong others Lam (1990), Olsen et al. (1993), De Jong and Burrough (1995) andAltobelli et al. (2001). In calculations – sommaco – script was used. It was used inMatLab program by dr Altobelli (University of Trieste, Italy).

    To specify dependences between spectral characteristics of researched areas andgrowing their types of plant communities on chosen sample plots in separate forest

    The influence of past human activity gradients on present variation of NDVI and texture indices 175

    Table 2. List of used bands of chosen satellites for calculationof NDVI index.

    Satellite RED IR

    Landsat MSS MSS 5 MSS 7Landsat TM TM 3 TM 4Landsat ETMþ ETM 3 ETM 4

  • types, NDVI index analysis took place in 2000. The size of sample plots was 300� 300meters, which is 100 pixels of satellite imageries Landsat TM and Landsat ETMþ. Thecomparison of spatial landscape changes, the influence of various intensity of gainingwood by clear cuttings, introduction of line elements such as roads, dividing lines,borders of isolations was made by comparing the above mentioned factors.

    By using other levels of information from created Geographical Information System(GIS) of the Zabory Landscape Park (Kunz 1999) the history of this area afforestationwas defined and the dependence of spectral response on the age of treestands and thetype of habitat was described.

    Surface researches were also compared on the basis of the value of 6 features with thenumerical taxonomy method. Hierarchic classification took place with UPGMA methodusing squared Euclidean Distance. The numerical ordination was made with thecorrespondence analysis method. The MVSP package (Kovach 1993) was used incalculations.

    4 RESULTS

    Because of differences in spatial resolution of the used satellite imageries (80 and 30meters) the analysis of results was divided into two time periods: 1975–1981 and 1990–2003. Gained results of NDVI values for all study areas are in Table 3. In analysescovering the first time period (1975–1981) the highest NDVI values were gained for theKruszyn study area, and the lowest for the PNBT study area. The highest differencesbetween these areas were stated in 1976–1981. At the same time it must be underlinedthat differences between study areas appeared in case of the calculation of NDVI indexfor the whole study areas as well as in analyses which took place without taking intoconsideration aquatic and marshy ecosystems. In the second case, the biggest differenceswere noted in 1981.

    The PNBT study area, which gained the lowest NDVI values in the first years, in thesecond time period (1990–2003), analyzed by Landsat TM and ETMþ satellites,achieved the highest values of these parameters. For sure, the lowest values in thiscomparison concerning the period from 1990, were gained by the Kruszyn study area.The exception to the above dependence is the year 2000, when for the Laska study areavalues were higher than for the PNBT study area. Similarly, there appeared differencesin NDVI values in the cycle of the calculation which took place without taking intoaccount aquatic ecosystems in the researched area.

    The highest fragmentation index for the whole time horizon exists for the Kruszynstudy area, and the lowest for the PNBT study area. A similar variation between thecompared objects is characteristic for the diversity, which achieves the lowest values incase of the PNBT study area. These results show that giving up cuttings and thedomination of one exploitation form, which are forests in the area of the national park,decreases the heterogeneity of the landscape.

    The texture index – fractal dimension achieved the highest values for the Kruszynstudy area in 1990–2003. The lowest FD values were typical for the PNBT study area.For the period 1975–1981 the opposite dependence was observed, highest values weregained by the PNBT study area. Limited cuttings in the area of the present national park

    176 M. Kunz & A. Nienartowicz

  • The influence of past human activity gradients on present variation of NDVI and texture indices 177

    Tab

    le3

    .V

    alu

    eso

    fN

    DV

    Ian

    dte

    xtu

    rein

    dic

    eso

    fst

    ud

    yar

    eas

    in1

    97

    5–

    20

    03

    .

    Sat

    elli

    te

    Stu

    dy

    MS

    SM

    SS

    MS

    SM

    SS

    TM

    ET

    ET

    area

    Ind

    ex(0

    9.0

    9.1

    97

    5)

    (26

    .05

    .19

    78

    )(1

    2.1

    0.1

    97

    9)

    (05

    .04

    .19

    81

    )(2

    8.0

    7.1

    99

    0)

    (05

    .05

    .20

    00

    )(3

    0.0

    5.2

    00

    3)

    ND

    VI

    0.3

    73

    30

    .14

    13

    0.1

    77

    90

    .06

    57

    0.6

    88

    50

    .29

    49

    0.6

    90

    5m

    axN

    DV

    I0

    .66

    67

    0.4

    80

    50

    .42

    37

    0.2

    39

    41

    .00

    00

    0.8

    66

    70

    .96

    29

    ND

    VI*

    0.3

    89

    50

    .14

    77

    0.1

    85

    80

    .07

    33

    0.6

    92

    60

    .32

    52

    0.7

    00

    7K

    rusz

    yn

    H2

    .81

    48

    2.7

    92

    02

    .64

    83

    2.6

    52

    82

    .63

    73

    2.9

    03

    12

    .83

    69

    FD

    2.9

    01

    42

    .91

    45

    2.9

    05

    52

    .91

    53

    2.8

    24

    72

    .91

    76

    2.8

    44

    3F

    0.8

    79

    80

    .86

    49

    0.7

    69

    40

    .77

    41

    0.6

    84

    80

    .85

    26

    0.7

    92

    2

    Las

    ka

    ND

    VI

    0.3

    56

    20

    .12

    21

    0.1

    48

    40

    .04

    42

    0.6

    98

    20

    .35

    33

    0.7

    49

    6m

    axN

    DV

    I0

    .67

    44

    0.5

    18

    90

    .45

    45

    0.2

    52

    70

    .93

    75

    0.8

    73

    01

    .00

    00

    ND

    VI*

    0.4

    09

    70

    .14

    97

    0.1

    83

    60

    .06

    49

    0.7

    34

    30

    .44

    37

    0.7

    59

    3H

    2.8

    02

    42

    .81

    04

    2.6

    49

    72

    .60

    91

    2.6

    47

    12

    .90

    80

    2.7

    85

    5F

    D2

    .89

    47

    2.9

    15

    02

    .90

    49

    2.9

    07

    72

    .82

    69

    2.8

    98

    82

    .83

    63

    F0

    .87

    33

    0.8

    78

    10

    .76

    77

    0.7

    41

    30

    .68

    83

    0.8

    55

    10

    .77

    20

    PN

    BT

    ND

    VI

    0.3

    62

    50

    .09

    60

    0.1

    44

    00

    .01

    62

    0.7

    18

    00

    .30

    97

    0.7

    49

    0m

    axN

    DV

    I0

    .69

    62

    0.4

    13

    00

    .38

    98

    0.1

    73

    30

    .94

    20

    1.0

    00

    01

    .00

    00

    ND

    VI*

    0.4

    10

    90

    .13

    08

    0.1

    86

    10

    .04

    43

    0.7

    41

    90

    .39

    96

    0.7

    63

    7H

    2.8

    01

    72

    .80

    39

    2.6

    37

    02

    .55

    59

    2.5

    77

    02

    .86

    89

    2.7

    74

    9F

    D2

    .90

    94

    2.9

    20

    62

    .90

    75

    2.9

    17

    02

    .80

    96

    2.8

    96

    02

    .83

    63

    F0

    .87

    29

    0.8

    74

    00

    .76

    43

    0.7

    17

    80

    .64

    93

    0.8

    29

    50

    .75

    45

    To

    tal

    ND

    VI

    0.2

    92

    50

    .12

    31

    0.1

    41

    50

    .03

    85

    0.6

    33

    80

    .23

    78

    0.7

    04

    8ar

    eao

    fm

    axN

    DV

    I0

    .72

    97

    0.6

    15

    70

    .47

    94

    0.3

    26

    51

    .00

    00

    1.0

    00

    01

    .00

    00

    ZP

    KN

    DV

    I*0

    .35

    37

    0.1

    42

    30

    .16

    30

    0.0

    47

    10

    .67

    78

    0.2

    93

    60

    .72

    36

    *m

    ean

    ND

    VI

    ind

    exca

    lcu

    late

    dd

    isre

    gar

    din

    gaq

    uat

    icec

    osy

    stem

    s.

  • lead to the uniformity of the age structure of tree stands, and in consequence to thedecrease of FD index in the forest landscape.

    The comparison analysis of these study areas took place on the basis of the 6following features: 1 – mean value of NDVI, 2 – max NDVI, 3 – mean value of NDVIexcluding aquatic ecosystems, 4 – H, 5 – FD, 6 – F. The calculation results of thesefeatures for three study areas were compared graphically with Jentys-Szaferowa (1948)method used mostly for the comparison of sizes and shapes of plants. The comparisonunit could be arithmetic average from the value of features of all samples or the featurevector of one sample. In this paper the second case takes place. The comparison unit isthe feature of the PNBT surface. The diagram is drawn on the basis of the gainedquotients. It represents the deviation of the Kruszyn and Laska areas from the comparisonunit. The location of points representing features is marked in relation to the scaleshowing the value of the quotient. On the diagram, the comparison unit is expressed by astraight vertical line. The deviation to the left shows that features of the comparedsamples are lower, and to the right – higher than the comparison unit (Figure 2). Theresult shows that the area of the present national park was the territory of the intensiveforest economy in 1970s. It is also proved by the low value of NDVI during this time.

    Analyzing sample plots (300� 300 meters) distributed in the Kruszyn study area, itwas stated that in areas on the habitat a dry forest had higher NDVI values than sampleplots localized on the habitat of a fresh forest. Analyzing the spatial variability of NDVIindex in the year 2000 in relation to the range of forests in the past it can be noticed thatthe researched areas, which were covered by the forest (dark grey color in the Figure 3a)all the time have higher NDVI values than researched areas marked on formerly arablelands (grey color in Figure 3a).

    Analyzing the spatial distribution of NDVI index on the Laska study area in 2000 inrelation to the range of arable lands in the past it can be noticed that the researched areaswhich were arable lands 126 years ago (grey color in Figure 3b) show lower NDVI valuesthan areas marked on lands covered all the time by forests (dark grey in Figure 3b).Lower NDVI values for sample plots used in the past as arable lands in relation to forestareas were gained for all analyses which took place on the basis of Landsat TM andETMþ satellite.

    All sample plots marked in the PNBT study area are at present covered bytreestands of higher age classes on the habitat of the fresh pine forest. Sample plots,which were arable lands over 200 years ago, have lower NDVI values than thesewhich were covered by forests all the time. In Figure 3c, areas which were arableland in the past, are marked with a dark color. Lower NDVI values for sample plotswith the arable past were gained for all considered periods of Landsat TM andETMþ satellite registration.

    No correlation was noticed between the average age of tree stands and the meanNDVI value of sample plots of all study areas.

    The result of the comparison of index values for three study areas, gained forsatellite imageries with higher resolution (TM, ETMþ) is presented in Figure 2.Curves drawn with the graphic method of Jentys-Szaferowa show that in all cases thelowest NDVI values were gained by the Kruszyn study area. In analyses of ETMþ inthe year 2000 differences among areas were the biggest. The Laska area gained thehighest value only in this analysis. There was a big similarity between the Laska and

    178 M. Kunz & A. Nienartowicz

  • PNBT areas in analysis of ETM+ in the year 2003. In Figure 2 their lines nearly covereach other.

    Big similarity between the Laska and PNBT areas (and at the same time difference ofthe Kruszyn area) proved the classification results of numerical taxonomy. The Laskaand PNBT areas create a common cluster (Figure 4).

    The influence of past human activity gradients on present variation of NDVI and texture indices 179

    TM 1990

    0,9

    1

    1,1

    654321

    Kruszyn Laska PNBT

    ETM+ 2000

    0,8

    0,9

    1

    1,1

    1,2

    654321

    Kruszyn Laska PNBT

    ETM+ 2003

    0,9

    1

    1,1

    654321

    Kruszyn Laska PNBT

    Figure 2. The comparison of indices of three sample plots analyzed by the graphical method

    of Jentys-Szeferowa. PNBT (vertical line) is the comparison unit. Notation of features: 1 – NDVI,

    2 – max NDVI, 3 – NDVI disregarding aquatic ecosystems, 4 – diversity (H), 5 – fractal dimen-

    sion (FD), 6 – fragmentation (F).

  • Data in Table 3 and the location of points indicating individual features in Figure 2show that lower NDVI values correspond to lower H, FD and F indices. Especially thelast one gains the high value at the low NDVI. This dependence is shown very well inFigure 5. There is NDVI value (from the series of calculations disregarding aquaticecosystem) and FD index.

    Kruszyn

    0

    0,1

    0,2

    0,3

    0,4

    0,5

    0,6

    0,7

    me

    ad

    ow

    Ps7

    F

    me

    ad

    ow

    Ps3

    7F

    Ps1

    03

    F

    Ps2

    6F

    me

    ad

    ow

    Ps3

    1F

    Ps9

    8F

    Ps6

    3F

    Ps8

    3F

    Ps6

    3F

    Ps6

    3F

    Ps3

    5F

    Ps5

    3F

    Ps1

    02

    F

    Ps8

    8F

    Ps3

    8F

    Ps6

    3F

    Ps1

    03

    D

    Ps4

    5D

    Ps4

    3F

    Ps6

    4D

    Ps5

    8D

    Ps5

    7F

    Laska

    0

    0,1

    0,2

    0,3

    0,4

    0,5

    0,6

    0,7

    me

    ad

    ow

    Ps9

    2F

    Ps6

    3F

    Ps4

    0F

    Ps9

    8F

    Ps8

    4F

    Ps3

    6D

    Ps4

    0F

    Ps7

    3F

    Ps9

    8F

    Ps9

    8F

    Ps8

    3F

    Ps7

    8F

    Ps1

    18

    F

    Ps7

    8F

    Ps4

    2F

    Ps2

    7D

    Ps1

    38

    F

    Ps4

    4F

    Ps6

    8F

    Ps3

    9D

    Ps9

    1F

    Ps1

    01

    F

    Ps1

    23

    F

    PNBT

    0

    0,1

    0,2

    0,3

    0,4

    0,5

    0,6

    Ps1

    11F

    Ps5

    9F

    Ps9

    0F

    Ps3

    8F

    Ps5

    7F

    Ps1

    00F

    Ps7

    7F

    Ps8

    8F

    Ps1

    01F

    Ps9

    0F

    Ps3

    0F

    Ps5

    3F

    Ps5

    1F

    Ps1

    03F

    Ps1

    03F

    Ps5

    9F

    Ps3

    1F

    Ps3

    9F

    Ps6

    3F

    Ps8

    8F

    Ps9

    8F

    Ps5

    7F

    Ps3

    9F

    Ps6

    3F

    Figure 3. Mean NDVI value (year 2000) of researched areas within study areas (a dark grey

    color marks the areas on which there where forest ecosystems in 1874); marks: Ps – Pinus

    sylvestris, age if trees, F or D – fresh or dry forest.

    180 M. Kunz & A. Nienartowicz

  • The influence of past human activity gradients on present variation of NDVI and texture indices 181

    TM 1990

    Eucildean D

    istanceS

    quared

    00.0008

    0.00160.0024

    0.00320.004

    0.0048

    ETM+ 2000

    Eucildean D

    istanceS

    quared

    00.0012

    0.00240.0036

    0.00480.006

    0.0072

    ETM+ 2003

    Squared E

    ucildean Distance

    Laska

    PN

    BT

    Kruszyn

    Laska

    PN

    BT

    Kruszyn

    Laska

    PN

    BT

    Kruszyn

    00.001

    0.0020.003

    0.0040.005

    0.006

    Figure 4. Dissimilarity between plots Kruszyn, Laska and PNBT.

    PCA

    Kruszyn

    Laska

    PNBT

    -0.009

    -0.018

    -0.027

    -0.036

    -0.045

    0.009

    0.018

    0.027

    0.036

    0.045

    NDVI

    H

    F

    Figure 5. Ordination of three sample plots analyzed on the basis of Landsat ETMþ from 2000.

  • On the basis of ETMþ from the year 2003, the Kruszyn study area, coveringsecondary forest on formerly arable lands, gains low NDVI, high FD and the PNBTstudy area gains high NDVI with low FD value.

    5 DISCUSSION AND CONCLUSION

    Made on the basis of satellite imageries comparative analyses of study areas showeddifferences in spatial structure of landscape in these areas. Chosen study areas, inspite of the fact that today they are all covered by forests and represent onedominating type of landscape – forest landscape, in the past the represented differenttypes of phytocenosis with the domination of variable categories of the land cover.Except the difference which is the result of the way the landscape was used andmanaged in the past, at present the analyzed study areas are treated differently byforms and range of protection of their natural values. This determines the intensity offorest economy in separate areas.

    It is concluded from the comparisons of ranges and mean NDVI values in separateareas and historical cartographic materials that NDVI values registered on presentsatellite imageries are determined by the way the area was used in the past. In forestareas localized in the northern part of the Zabory Landscape Park, where the Kruszynstudy area was marked, there are mostly pine forests restored on the formerly arableareas. Taking into consideration species composition, they refer to dry forests. Thissimilarity is mainly a consequence of lower share of undershrubs, and high frequencyand cover of overground lichens. In this part of the landscape park, a considerable area iscovered by Pinus-Calluna community distinguished by Boinski (1992). The heatherforest was mainly created as the result of reforestation of poor pastures on the turn of the19 century. In this community, apart from the common heather there were alsooverground mosses and lichens.

    Other forest communities dominate in the southern part of the researched area, it isthe place where the national park was created. Because of the fact that this area was notagricultural in the past, degeneration of forest communities was mainly based onintroduction of pine monocultures to the cuttings, today’s fresh forests dominate here,and they are classified as Leucobryo-Pinetum association.

    Because of the lower state of green biomass and lower content of chlorophyll in thearea unit which is noticed in forests corresponding to dry forests, NDVI in formerlyarable complexes, in which such a phytocenosis is dominating, is lower. Kunz et al.(2000) stated it using methods of the remote sensing control.

    From data given by Barcikowski (1992) it is concluded that lowered values of NDVIfor areas covering forests which are intensively exploited, are influenced by the presenceof cuttings and pine crop which are a few years old. Biomass and chlorophyll indices insuch ecological groups are very low. Changes of both parameters in dependence of theage of treestands are not considerable in phytocenosis for about 20 years. This is why thearea of today’s national park, where age differentiation covers first of all II / III – VIclasses, is characterised not only by higher NDVI value but there is also a lower varietyand fragmentation of the landscape calculated on the basis of spatial variability of thisparameter.

    182 M. Kunz & A. Nienartowicz

  • On the areas covering intensively exploited forests, because of the presence of manycutting areas and a few years old crops, NDVI decreases and variability and fragmentationincreases. The direction of changes of the last two parameters is according to calculations,which were done by Franklin and Forman (1987) on theoretical models of the forestlandscape structure with the participation and various distribution of cutting areas.

    In comparisons made in this paper satellite imageries were used from 1975–2003.The contemporary state was researched from before and over 8 years after the foundationof The National Park ‘‘Bory Tucholskie’’. After the foundation of the national parktimber logging was considerably limited and clear cuttings do not take place there at all.This is why differences between this part of the forest complex and its northern part havebecome bigger, although in areas of intensive forest economy clear cuttings are limited,and if they take place, on these clearings biogroups of trees are left which are seedbearing. The foundation of the national park contributed to smaller diversity andfragmentation of the forest landscape in this part of Tuchola Forest. However, what arethe results of the use of new forest technologies will be possible to define by thecomparison of satellite imageries from the following decade, e.g. from 2010.

    In this analysis it was possible to notice the influence of two essential factors on thepresent landscape variability of NDVI in ZLP. The first one is the past agricultural use ofthis area which is manifested in considerable distinctness of the Kruszyn area (pastheathlands and pastures) and similarity of the Laska and PNBT areas, on which cuttingswere immediately reforested in the past. The second factor is intensity of the currentforest economy. The Laska area with economic forests (used today as well as in theKruszyn area) has NDVI and FD parameters more approximate to the PNBT area than tothe Kruszyn area. NDVI values of this area, which in the past was used agriculturally,however, are lower than NDVI of the national park in which wood is not cut, and forestthickets do not exist.

    The opposite dependence between NDVI and the fractal dimension is according todependencies given by Altobelli et al. (2001) for areas covering bigger number oflandscape types. It is similar to what we have stated in this document that humanactivities in the forest landscape cause the decrease of NDVI and the increase of thefragmentation. This analysis showed considerable usefulness of numerical taxonomymethods to the analysis of data interpretation gained from satellite imageries.

    The considerable age variability of tree stands makes a spectral characteristic of plantcover difficult in the Zabory Landscape Park. For a more precise definition of theinfluence of agricultural land use on the condition of restored forest ecosystems it wouldbe well to make a similar analysis with the use of high resolution satellite imageries andbigger groups of pixels, which would cover cultivations of the same age, existing on thesame habitat and in places of similar land use in the past.

    REFERENCES

    Altobelli, A., Feoli, E., Ourabia, L., 2001. An overview of landscape structure through theapplication of fractal dimension to remotely sensed images using GIS technology. [In:]Nienartowicz, A., M. Kunz (eds), GIS and remote sensing in studies of landscape structuresand functioning, NCU, Torun.

    The influence of past human activity gradients on present variation of NDVI and texture indices 183

  • Barcikowski, A., 1992. Differentiation in the structure and energy flow in phytocenosis withprimary and secondary succession. [In:] Bohr, R., Nienartowicz, A., Wilkon-Michalska,J. (ed.), Some ecological processes of biological systems in North Poland. pp. 35–58, NCU,Torun.

    Boinski, M., 1992. Osobliwosci szaty roslinnej Borow Tucholskich. Tow. Milosnikow BorowTucholskich, Torun.

    Franklin, J.F., Forman, R.T.T., 1987. Creating landscape pattern by forest cutting: Ecologicalconsequences and principles. Landscape Ecology 1: 5–18.

    Monmonier, M.S., 1974. Measures of pattern complexity for choropleth maps. The AmericanCartographer 1(1): 159–169.

    Murphy, D.L., 1985. Estimating neighborhood variability with a binary comparison matrix.Photogrammetric Engineering & Remote Sensing 51(6): 667–674.

    De Jong, S.M., Burrough, P. A., 1995. A fractal approach to the classification of mediterraneanvegetation types in remotely sensed images. Photogrammetric Engineering & RemoteSensing 61(8): 1041–1053.

    Kovach, W.L. 1993. MVSP – A Multi Variate Statistical Packade for IBM PC’s, version 2.1.Kovach Computing Services, Pentraeth, Wales, UK.

    Jentys-Szaferowa, J., 1948. Graficzna metoda porownywania ksztaltow roslinnych. Kosmos: 66.Kunz, M., 1999. System Informacji Geograficznej (GIS) Zaborskiego Parku Krajobrazowego,

    [In:] Barcikowski, A., Boinski, M., Nienartowicz, A. (eds), Wielofunkcyjna rola lasu.Ochrona Przyrody-Gospodarka-Edukacja, Wyd. UMK, Torun.

    Kunz, M., Nienartowicz, A., Deptula, M., 2000. The use of satellite remote sensing imagery fordetection of secondary forests on post-agricultural soils: A case study of Tuchola forest,northern Poland. [In:] Casanova, L. (ed.), Remote Sensing in the 21st Century: Economicand Environmental Applications. JA. A. Balkema/Rotterdam/Brookfield, pp. 61–66.

    Lam, M.S.N., 1990. Description and measurement of Landsat TM images using fractals.Photogrammetric Engineering & Remote Sensing 56 (2): 187–195.

    Olsen, E.R., Ramsey, R.D., Winn, D.S., 1993. A modified fractal dimension as a measure oflandscape diversity. Photogrammetric Engineering & Remote Sensing 59(10): 1517–1520.

    Rouse, J.W., Haas, R.H., Schell, J.A., Deering, D.W., 1973. Monitoring vegetation systems inthe great plains with ERTS, ERTS Symposium. NASA SP-351 I: 309–317.

    Turner, M.G., 1989. Landscape ecology: the effect of pattern on process. Ann. Rev. Ecology andSystematic 20: 171–197.

    184 M. Kunz & A. Nienartowicz