New method of landscape typology in the Czech Republic

(1)

New method of landscape typology in the Czech Republic

D�šan Romportl�, Tomáš C��man, Zdeněk Lipský

�epart�ent of Physical Geography and Geoecology, Faculty of Science, Charles University

�lbertov 6, 128 43 Praha 2 e�ail: dusan@natur.cuni.cz

________________________________________________________________________________

Abstract. �he paper introduces a new �ethodological syste� of a co�plex landscape typology. In co�parison with the for�er typologies the basic difference is that the presented typology is based on exact, easily quantified data covering both natural and cultural landscape conditions, which can be classified in GIS. By co�bining chosen the�atic layers a new unique raster dataset was created where each pixel has a specific co�bination of selected characteristics. �ll pixels of the sa�e su��ary characteristics represent a particular landscape type. �he unique landscape types were generalized and co�bined with si�ilar ones in polygons defined by eCognition seg�entation process. �he output of the used �ethodology is a �ap representing preli�inary landscape types of the conte�porary Czech landscape that will be the sub��ect of further

�odifications and interpretations.

key words: landscape typology, cultural landscape, eCognition, GIS

Introduction

Classification, regionalization and typology of landscapes represent one of the �ost i�portant sub��ects of study in geoecology and co�plex geography. However, landscape as a dyna�ic co�plex spatial syste� is very changeable in the ti�e; therefore the landscape classification represents a difficult but necessary task for landscape scientists. It is i�portant both fro� theoretical (as basic research) and practical points of view. � core of practical applications of landscape typologies and classifications consists in a wide spectru� of �ethods of landscape planning, assess�ent, evaluation and �anage�ent. �lthough different landscape typologies were developed in the Czech Republic in the past (e.g. �e�ek et al. 1977, �tlas životního prostředí a zdraví obyvatel ČSFR 1992, �ole��ka, Lipský 1999), none of the� is widely used and applicable in recent situation. Latest landscape typology developed by research group of Löw et al. was not published co�pletely yet, therefore could not be reviewed critically. Whereas unified co�plex landscape typologies exist co��only abroad (e.g. �tlas kra��iny SR 2002), there is nu�ber of different landscape classifications based on sub��ective expert approaches in the Czech Republic.

�he paper introduces a new �ethod of co�plex ob��ective typology for the conte�porary Czech landscape. �he basic difference is that presented typology is based on exact, easily quantified data covering both natural and cultural landscape conditions, which could be classified in GIS. Prepared �aps will be published in new �tlas of Landscape of the Czech Republic.

Met�ods

New �ethodical syste� of landscape typology is based on four sequential steps as follows:

Klasyfikacja krajobrazu. Teoria i praktyka. Problemy Ekologii Krajobrazu. 2008, t. XX. 315-320.

(2)

i. the choice of landscape-character representatively expressing components

�he usage of exact quantifiable datasets is the �ain difference to other approaches to landscape classification in the Czech Republic. �he incorporation of all environ�ental and socioecono�ic variables that deter�ine the landscape character would be difficult due to different weight of each variable and due to different ti�e and spatial scale. �herefore the first �ost i�portant step was to deter�ine co�ponents, fro� the hierarchical syste� below (fig.1), that representatively express the landscape character. �he selection had to reflect the i�portance of each co�ponent in deli�itation of landscape types, however the data quality and accessibility was a constraint. In the Czech Republic cli�ate, relief and geological substrate were selected as the �ain co�ponents describing the natural environ�ent (pri�ary landscape structure) and land use with landscape heterogeneity were selected as the best available co�ponents describing the secondary landscape structure.

�he input data were as follows:

1. Environ�ental data

• cli�ate (derived fro� cli�atic regionalization according to Quitt 1971)

• parent geological substrate (derived fro� geological database GEO CR 500 and Soil �ap of the ČR)

• relief expressed by altitude and vertical heterogeneity (derived fro� elevation grid �E� �rcCR 500) 2. Secondary landscape structure

• land use (derived fro� CORINE Land Cover 2000 database)

•landscape secondary structure heterogeneity (derived fro� CORINE Land Cover 2000 database)

Fig.1. Selection of representative natural and cultural co�ponents for typology (�fter �ücher et al. 2003, �odified)

ii. Datasets pre – processing: generalization and reclassification

Firstly all datasets were generalized according to their the�atic content i�portant fro� classification point of view. In case of climate there were left all original 13 classes (�2 - CH7) of �ezzo - cli�atic regionalization after Quitt (1971) as it shown in table 1.

Next the�atic layer – parent geological substrate - was created by synthesis of geological (GEO ČR 500, 1:500 000, Czech geological survey) and pedogeographical datasets (Map of Soil Types of the Czech Republic, 1:200 000, Němeček et al.). Original 19 classes of geological dataset were �erged into 6 categories; consecutively 3 classes covering specific sedi�ent substrates were generated fro� Soil �ap of CR. Resultant the�atic layer represents generalized but accurate dataset of 9 basic types of parent �aterial (tab. 2).

(3)

Other infor�ation layers i�portant for typology of natural landscape were derived fro� digital elevation �odel (�E�) with pixel size 200x200�, that is part of �opographical database ArcČR 500. Si�ple categorization of landscape into 6 altitude levels covering landfor�s fro� lowlands to �ountains is one of the basic inputs (tab.

3).Vertical� �eterogeneity was derived fro� the sa�e database by geostatistical function standard deviation, which describes dataset variety – altitudes in the net of 1k�² square size. �his output was co�pared to traditional vertical heterogeneity evaluation. Final results of database co�putation were reclassified into 4 categories representing basic types of landscape surface according to vertical heterogeneity fro� flat to undulating (tab.

4).�dditional typological process was �anaged by using land use / land cover infor�ation derived fro� database CORINE Land Cover 2000. Original 28 categories of CORINE’s no�enclature, �apped in the Czech Republic,

�able 1. Categories of cli�atic regionalization (after Quitt, 1971)

Code Type Category o� CLIM

1 �4 War� cli�atic region type 4

2 �2 War� cli�atic region type 2

3 ��11 �oderate cli�atic region type 11

11 CH7 Cold cli�atic region type 7

�able 2. Categories of parent geological substrate (derived fro� GEO ČR 500; Soil �ap, Ně�eček et al.)

Code Type Category o� SUB

1 v Vulcanites

2 p Plutonites

3 � �eta�orphites

4 s Sedi�ents of covered for�ation

5 b Specific sedi�ents of Barrandien

6 k �arsts sedi�ents

7 c �esozoic sedi�ents

8 q Quaternary sedi�ents

9 a �lluvial sedi�ents

�able 3. Categories of altitude levels (derived fro� �E�, �rcČR 500)

Code Type Category o� DEM Altitude levels

1 L Lowlands 0 - 250

2 � �owns 250 – 500

3 H Highlands 500 - 750

4 U Uplands 750 - 1000

5 � �ountains 1000 - 1250

6 A �lpine �ountains 1250 - 1600

(4)

were generalized to 10 classes significant for classification of landscape types as it is shown in table 5.

�oreover the sa�e dataset was used for deriving land cover heterogeneity infor�ation by applying geostatistical function “Variety” – which evaluates nu�ber of different land cover categories in given space – here square of 1k�2 net again. �ccording to the results 4 types of landscape heterogeneity fro� co�pletely ho�ogeneous (1 - 2 classes per 1k�²) up to intensely heterogeneous (7 – 8 classes) were generated (tab. 6).

�ataset created this way were transfor�ed to raster type layers with identical pixel size 200x200�. �ll classes of each the�atic layer hold unique code nu�ber, therefore a co�plex of unified and co�parable datasets was developed and processed by �ethods of raster algebra in GIS environ�ent. Unique code of each pixel is built by following co�putation:

�YPE_N = CLI� + SUB + �E� + V�R + LC + HE�

e.g. �YPE_N = �2.q.L.f.a.c �YPE_N = 2.8.1.1.2.1

iii. rGB synthesis of selected datasets

One of the keynote steps in proposal typology is RGB synthesis representing original way of co�bination of selected input layers in �rcGIS environ�ent. For this purpose pri�ary – not derived – layers were used: cli�ate, parent geological �aterial, altitude levels and land use. �hree of these layers carry color channel infor�ation, the fourth one is without any infor�ation. Spectral characteristic of synthesized scene is changeable according to input layers integration (for exa�ple: RE� channel – cli�ate, GREEN – parent �aterial, BLUE – altitude, NO

�able 4. Categories of vertical heterogeneity (derived fro� �E�, �rcČR 500)

Code Type Category o� VAR Interval of standard deviation of altitude per 1km²

1 f Flat 0 – 10

2 u Undulate 10 - 25

3 h Hilly 25 - 50

4 � �ountainous 50 - 150

�able 5. Categories of land cover (derived fro� CORINE Land Cover 2000)

Code Type Category of LC

1 x �nthropogenic areas

2 a �rable land

3 k Per�anent cultures

4 s Pastures

5 h Heterogeneous agriculture areas

6 c Coniferous forests

7 d �eciduous forests

8 o Open bare spaces

9 t Wetlands

10 w Waters

�able 6. Categories of land cover heterogeneity (derived fro� CORINE Land Cover 2000)

Code Type Category o� HET Number of classes

1 c Co�pletely ho�ogeneous 1 - 2

2 � Ho�ogeneous 3 - 5

3 t Heterogeneous 5 - 6

4 h Co�pletely heterogeneous 7 - 8

(5)

CH�NNEL – land use). Use of four channels representing particular layers is i�portant for parallel typological processing of both natural and cultural environ�ent.

Firstly cultural infor�ation is reduced due to delineation types of natural landscapes, further the weight of natural factors is set low and existing scene is seg�ented ��ust according to land cover infor�ation. Si�ilarly as it shown above in the �ethod of raster calculation each pixel is characterized by unique code – bands co�bination. For exa�ple pixel with RGB values 933 is part of highland area in �oderate cli�atic region type 3 with �eta�orphites as parent �aterial. �he infor�ation about land cover could be gained by shifting channels.

�he final result is an ER��S I�agine file, visually si�ilar to satellite scenes, which could be classified by re�ote sensing �ethods.

iV. Delineation of landscape types polygons by segmentation of rGB scene

Second keynote step of typology process is use of controlled seg�entation of RGB scene. It’s allowed in the environ�ent of software eCognition, which is unique ob��ect oriented software, where i�age classification is based on attributes of i�age ob��ect rather than on attributes of individual pixels (�ücher et al. 2003).

RGB co�posite i�age could be changed by linking different bands to different input levels, therefore �ultiple seg�entation on various hierarchical levels is possible. �he basic principle of this process is �erging of related pixels of si�ilar spectral characteristics. Groups of these related pixels represent specific landscape types, whose delineation is done by process of �ultiresolution seg�entation. Weight of each input layer could be set differently and so influence of particular natural or cultural factors could be reduced or accentuated. �his �ethod solves the �ost proble�atic step in typological process – definition of landscape types – in an ob��ective and independent way. eCognition software also allows export of created polygons into GIS environ�ent.

In case of landscape typology of the Czech Republic firstly seg�entation based on natural data (CLI�, SUB and �E�) was done, in order to delineate types of natural landscape. �s the second step - next seg�entation process on lower hierarchical level based on land cover data was perfor�ed in resulting natural units.

Characterization and interpretation of delimited landscape types

�he deli�ited polygons were i�ported to the �rcGIS 9.1 software. � unique code was assigned to each polygon and the zonal statistics was carried out within each polygon using Spatial �nalyst for �rcGIS. �he statistics considered above �entioned input raster datasets that enabled us to characterize each polygon fro�

the viewpoint of cli�ate, relief and geological substrate. When characterizing the geological substrate and cli�ate the algorith� “�a��ority” was chosen reflecting the prevailing kind of substrate or cli�atic region within each polygon. When characterizing the relief �ean altitude was calculated. �he obtained characteristics were cartographically expressed.

Conclusion

�here were 45 unique natural landscape units deli�ited within the Czech Republic in the highest scale level, however the exact nu�ber of deli�ited landscape units was not the �ain ai�. �he �ain target, of the applied approach to landscape typology, was to suggest and test a widely applicable �ethod (choice of datasets, preprocessing of datasets) of landscape typology. �he next and the �ost i�portant step is further testing, considering spatial scale, before the final list of landscape types of the Czech Republic is published.

References

�tlas kra��iny Slovenske�� republiky, 2002. 1. vyd., Bratislava: �ŽP SR; Banská Bystrica. Slovenská agentúra životného prostredia. 344.

�tlas životního prostředí a zdraví obyvatel ČSFR, 1992. GgÚ ČS�V Brno.

(6)

�e�ek J., Quitt E., Raušer J., 1977. Fyzickogeografické regiony ČSR, Sborník ČSSZ, roč. 1977, č. 2, sv. 82.

89 – 99.

�ole��ka J., Lipský Z., 1999. �apy současné kra��iny. Geografie - Sborník ČGS, 104, 3, s.161-175 Löw J., �íchal I., 2003. �ra��inný ráz. Lesnická práce, �ostelec nad Černý�i lesy. 552.

�ücher C. �., Bunce R. H. G., Jong�an R. H. G., �li��n J. �., �oo�en �. J. �. et al., 2003. Identification and characterisation of environ�ents and landscapes in Europe. �lterra – rapport 832, Wageningen. 120.

Quitt E., 1971. �li�atické oblasti ČSSR. Studia Geographica 16. Geografický ústav ČS�V, Brno.1 – 74.