The reconstruction of vegetation change in Nyíres-tó mire (ne Hungary)

(1)

First published online 19 September, 2009

THE RECONSTRUCTION OF VEGETATION CHANGE IN NYÍRES-TÓ MIRE (NE HUNGARY):

AN IMAGE-SEGMENTATION STUDY

D. CSERHALMI¹, J. NAGY², D. NEIDERT³and D. KRISTÓF⁴

1Department of Botany, Institute for Biology, Faculty of Veterinary Science Szent István University, H-1077 Budapest, Rottenbiller u. 50, Hungary

E-mail: szeltolo@gmail.com

2Institute of Botany and Ecophysiology, Faculty of Environmental and Agricultural Sciences Szent István University, H-2103 Gödöllő, Páter K. u. 1, Hungary

3Department of Zoology, Plant Protection Institute HAS, H-1029 Budapest, Adyliget Nagykovácsi u. 26–30, Hungary

4Institute of Geodesy, Cartography and Remote Sensing (FÖMI) H-1149 Budapest, Bosnyák tér 5, Hungary

(Received 2 March, 2009; Accepted 30 June, 2009)

The study area is the peaty bed of Nyíres-tó mire which is situated in the northeastern Alföld on the Bereg Plain. For this paper we used a digital photo interpretation method with which we reconstructed the former vegetation from black and white aerial photos, and made chronosequence of vegetation maps. The image segmentation method dissolves the photo into different objects (segments) by spectral and textural parameters. The segments consist of similar pixels, representing a unique ground object. We made the segmentation with the Definiens Inc. eCognition software. The stability of the mire was calculated with GRID-files. The historical vegetation maps show, that after many arid years, the tree or shrub dominant associations increased until the eighties. Later, the sufficient precipitation and the artificial flooding stabilised the tree covering. The analysis of GRID-files shown, that 45.77% of the pixels get code 1 (stable), 44.32% get code 2 (slightly changeable) and only 9.91% get code 3 (changeable). It means that almost half of the mire’s vegetation is the same as in 1952.

Key words: aerial photo, Bereg Plain, eCognition, image interpretation,Sphagnum

INTRODUCTION

Nyíres-tó was discovered by Tibor Simon in 1952 (Simon 1953), who later related the vegetation of the mire (Simon 1960). The importance of his research became evident when Nyíres-tó, Báb tava and Navad-patak were found to be the first mires have raised bog plant association in the Alföld. Later Vozáry (1957) made a pollen analysis. Simon used his results, and described the succession of Nyíres-tó (Simon 1968). After this, there was not any regular botani-

(2)

cal monitoring in the area for almost twenty years. In the early nineties Simon did a floristical comparison, where he pointed out the increase of the rate of weeds, but he did not describe the vegetation changes on a map (Simon 1992).

The regular monitoring was started in the middle of the nineties, and the sec- ond vegetation map was prepared by Nagy (1999). The actual vegetation map was created in 2006, where aerial photos and GIS methods were applied (Nagy 2006). Nagy used the macrofossil (Jakab and Magyari 2000) and palinological (Magyari 2002) data of Báb-tava and palinological data of Nyires-tó mire (Sümegi 1999) together with his present phytocoenodynamical knowledge to describe and draw the reconstruction of primary succession of these mires (Nagy 2007).

For the reconstruction of the vegetation change in the last fifty years we have only the starting state and the present state of the mire, so the interstates have to be revealed.

To monitor ecosystem changes, multitemporal methods are highly fa- vourable (Coppinet al.2004). Aerial photos are available from the fifties, and the application of aerial photographs in the botanical research has accelerated in the last few decades. However, high resolution images (IKONOS, Quick- bird) are only available from the end of the nineties, so long-term change detection cannot be performed. On the other hand, SPOT and Landsat satellites collecting images over large areas since the seventies, but their relatively low resolution makes them inadequate to investigate the small mires (Cserhalmi and Kristóf 2007). Digital image interpretation methods became popular (Mast et al.1996, Kadmon and Harari-Kremer 1999, Pellerin and Lavoie 2003, Sickel et al.2004, Langakeet al.2007), but the application of panchromatic photos was rejected because of their low information content. However, the usefulness of these images is still contentious.

Image segmentation can be performed panchromatic, coloured or multi- spectral photos. The image segmentation method dissolves the photo into different objects (segments) by spectral and textural parameters. Image objects are contiguous regions of an image (Benzet al.2004). Segmentation helps when drawing vegetation maps (Mathieuet al.2007), which can provide as equally detailed maps as produced by manually interpreting aerial photographs, or drawn with a ground-base method.

In this paper we used a digital photo interpretation method with which we can reconstruct the former vegetation from black and white photos, and make chronosequence of vegetation maps.

(3)

MATERIALS AND METHODS

The study area (48°10’58”N, 22°30’06”E) is a C-shaped, approx. 750 metres long and on average 80 metres wide oxbow (Szurdoki and Nagy 2002).

It belongs to the Samicum phytogeographical region on the northeastern part of Pannonicum that lies on the southeastern part of the Holarcticum (Soó 1964). In the centre parts oligotrophic bog communities (mostlyEriophoro vagi- nati-Sphagnetum) were present. The mire belongs to the administration terri- tory of the Hortobágy National Park.

Panchromatic aerial photos taken from the study area were bought from the Ministry of Defence Mapping Company. The black and white photos were taken in the vegetation period, but in different periods of the day, which means different shading which can cause problems during the digital interpretation. The first photo was taken in 1956, the others with a ten-year period- icity until 2002.

The first step of any image analysis is the segmentation of the image into unclassified basic image objects. The image segmentation method dissolves the photo into different objects (segments) by spectral and textural parameters. The segments consist of similar pixels, representing a unique ground object. We made the segmentation with the Definiens Inc. eCognition software.

Different commercial softwares for segmentation can be used, but a comparison between the programs showed the best result can be calculated with eCognition and InfoPACK (Meinel and Neubert 2004).

To create the polygons we defined the scale parameter. This is an abstract term which determines the maximum heterogeneity of the resulting image objects. In heterogeneous data the resulting objects for a given scale parameter are smaller than in more homogeneous data. A larger scale parameter value means larger image objects. Obtaining optimal objects from a multiresolution segmentation in most cases depends on a suitable choice of scale parameter which was counted by considering to the average size of the actual vegetation units (Definiens AG 2006).

Final work on the segments was done using ArcView 3.1. The first map was drawn on the photo taken in 2002. The next segmented image (1997) was fit to the previous one that showed which association can be found in 2002 in a polygon from the 1997 photo. Step by step the vegetation map can be drawn for every year.

AsGlyceria,TyphaandCarexspecies have similar structure from above, we used higher coenotaxon categories. It means that in Phragmitetea australis the following associations are included:Glycerietum maximae,Typhetum latifo- liae,Phragmitetum australis,Caricetum ripariae,Sparganietum erecti.

(4)

The stability of the mire was calculated with GRID-files. In this raster file, every pixel covers a unit, and has a value, called thematic code. This code can constitute the vegetation type, soil type, or other groundcover type (Czimber 2002). We labelled every association with a numeric value (thematic code), and every pixel of a GRID-file has the value of the association, which means that mathematical calculations can be performed. Subtracting the earlier image from the later results in a new image, where the zero value means those pixels, where were not any changes between the associations (between the codes). Af- ter this we reclassify the image, where zero (0) means “non-changed” area, and any other value (1) means “changed” area. Summing these “change- maps” the pixel’s values of the image will range from 0 to 5. Zero (0) means that the pixel has the same value in the whole research period, from 1956 to 2002. One (1) means that the pixel value was modified once during the whole research period. Five (5) means that the pixel changed its value within every period. We separated three different categories, where 0–1 means stable; 2–3 means slightly changeable; and 4–5 means changeable area.

RESULTS

In the late fifties (Fig. 1) theEriophorum vaginati-Sphagnetumassociation could be about 40–45 metres wide, and the surrounding alder trees formed only a narrow stripe. The raised bog associations were totally surrounded by Salix cinereafrom the east side and the north part of the mire was covered with CarexandGlyceria maximajust as Simon described (1960). There can only have been a few shrubs ofSalix cinerea. As the climate was very dry, and the artificial well had not built yet, the southern part of the mire was presumably covered by a moderately wet grassland, probablyAlopecuretum pratensisandAgroste- tum albae, which are still quite common nowadays.

The succession slowly proceeded forward, the alders getting bigger, and started to spread into the central region, partly covering the raised-bog associations (Fig. 2). The dry climate has helped the forest formation, but can endan- ger the Sphagnum-dominated associations (Simon 1992). The cover of Salix cinereahas also increased and at the south side of the mire solitary willow and oak trees have started to grow.

More and moreSalixshrubs appeared among theGlyceriaandCarextufts, and the alder stripe almost reached its present width (Fig. 3). Presumably due to the dry conditions the peat moss coverage diminished, and the extension of Eriophoro vaginati-Sphagnetumalso decreased. The southern grass associations started to decline, because the trees slowly displaced them.

(5)

The northern part was a mixture of sedge and willow, like at the Navad- patak today (Nagy et al.2008), but in some patches the cover ofSalix cinerea was higher than that ofGlyceria, so on the photo from 1988 we recorded it sepa- rately (Fig. 4). Till the eighties, the weather conditions and the precipitation supply were fair, resulting in the former changes getting into stagnate period.

Fig. 1. The vegetation map of Nyíres-tó in 1956

(6)

TheAlopecurusdominated patch dramatically decreased and the tree cover ex- tended. In the middle of the bed, the sufficient water supply inhibited the tree expansion.

Until 1997 the state of the mire stabilised and the vegetation was very similar to the present state (Fig. 5). The artificial water supply helped to eliminate

(7)

the effect of rainfall fluctuation. A new, young alder group appeared at the northwest side of theSphagnum carpet. Alder trees are typical fast growing trees, so they quickly enclosed the northern part until 2002 (Fig. 6). The mezofil grass ecosystems disappeared from the mire, the dominant herb plants are Glyceria maximaandCarex riparia.

(8)

Most of the Phragmitetea australis associations have changed at the southern part of the mire: the majority ofGlycerietum maximaeandCaricetum ripariae stands have been replaced withSparganietum erecti,Typhetum latifoliae, Lem-

(9)

netea and Bidentetea stands as wild boars have destroyed the deeper zone of the mire-bed by their treading and wallowing in the last five years. These changes are not detectable from the aerial photographs.

(10)

DISCUSSION

The segmentation was a suitable method in the reconstruction process of the Nyíres-tó mire. Comparing the segmentation to the common manual mapping techniques, segmentation results in almost the same accuracy and the-

(11)

matic resolution than the manual one, but the handmade map is less detailed, and the edges are less natural (Cunningham 2006). Besides this, older photos cannot be interpreted manually due to the lower quality; it is very difficult to separate homogeneous patches by eye. Whereas panchromatic photos are

Fig. 7. Result map of the GRID analysis

(12)

ideal for long-term change detection in small areas, like mires (Cserhalmi and Kristóf 2007), so the image segmentation method is suitable for vegetation reconstruction. The method is partly automated, but with accurate and regular ground control data, and with the knowledge of succession patterns the vegetation can be reconstructed for the last fifty years.

Analysing the GRID-files, it can be shown that 45.77% of the pixels get code 1 (stable), 44.32% get code 2 (slightly changeable) and only 9.91% get code 3 (changeable) (Fig. 7). It means that almost the half of the mire’s vegetation is the same as in 1952. The reason the vegetation could remain in the origi- nal state being that Nyíres-tó already has its whole buffer zone of forest and meadow (Szurdoki and Nagy 2002). One of the most stable areas is the southern, broader part of the alder bog, and a narrow stripe at its northern side. The big willow carr at the northern end of theEriophoro vaginati-Sphagnetumwas also present in the fifties, just like a triangular area at the northern side, where tall sedge associations covered and still cover the land. The most changeable areas are in the lagg zone of the mire.

Our work showed that the information content of panchromatic photos was more valuable and useful than we could have imagined. With digital photo interpretation methods we could get such information that could not be achieved with the manual methods.

*

Acknowledgements – Our research was helped by the Institute of Botany and Eco- physiology, the Department of Cartography, Geoinformatics and Remote Sensing, the PhD School of Biological Sciences (SZIE), the PhD School of Environment Sciences (SZIE), the management of Hortobágy National Park and the Ministry of Environment and Waters’

KAC applications. We would like to thank István Fintha (†), Attila Molnár (Hortobágy Na- tional Park), for their information helping us accomplishing this paper, and for Stephanie Veitch for the English language review.

REFERENCES

Benz, U. C., Hofmann, P., Willhauck, G., Lingenfelder, I. and Heynen, M. (2004): Multi-resolution, object-oriented fuzzy analysis of remote sensing data for GIS-ready information. –J. Photogram. & Remote Sensing58(3–4): 239–258.

Coppin, P., Jonckheere, I., Nackaets, K., Muys, B. and Lambin, E. (2004): Digital change detection methods in ecosystem monitoring: a review. –Int. J. Remote Sensing25(9):

1565–1596.

Cserhalmi, D. and Kristóf, D. (2007): Vegetation change detection on mires with digital aerial photographs. –Cereal Res. Comm.35(2): 329–332.

Cunningham, M. A. (2006): Accuracy assessments of digitized and classified land cover data for wildlife habitat. –Landscape and Urban Planning78: 217–228.

(13)

Czimber, K. (2002):Geoinformatika. Elektronikus jegyzet. – NymE, Sopron, 101 pp.

Jakab, G. and Magyari, E. (2000): Új távlatok a magyar lápkutatásban: szukcessziókutatás paleobryológiai és pollenanalitikai módszerekkel. (Progress in the Hungarian mire studies the use of paleobryological and palinological techniques in the reconstruction of hydroseres). –Kitaibelia5(Suppl. 1): 17–36.

Kadmon, R. and Harari-Kremer, R. (1999): Studying long-term vegetation dynamics using digital processing of historical aerial photographs. –Remote Sensing Environment68:

164–176.

Langake, T., Burnett, C. and Lang, S. (2007): Assessing the mire conservation status of a raised bog site in Salzburg using object-based monitoring and structural analysis. – Landscape and Urban Planning79: 160–169.

Magyari, E. (2002):Climatic versus human modification of the Late Quaternary vegetation in East- ern Hungary. – PhD thesis, University of Debrecen, Debrecen.

Mast, J. N., Veblen, T. T. and Hodgson, M. E. (1996): Tree invasion within a pine/grassland ecotone: an approach with historic aerial photography and GIS modelling. –Forest Ecol. Manag.93: 181–194.

Mathieu, R., Aryal, J. and Chong, A. K. (2007): Object-based classification of Ikonos imagery for mapping large-scale vegetation communities in urban areas. –Sensors7: 2860–2880.

Meinel, G. and Neubert, M. (2004):A comparison of segmentation programs for high resolution remote sensing data. – Int. Archives of the ISPRS, 35, pp. 1097–1102.

Nagy, J. (1999): Research report of the botanical investigations of Nyíres-tó and Báb-tava 1993–1998.– Archives of the Hortobágy National Park Dir., Debrecen, 68 pp.

Nagy, J. (2006):Vegetation maps of Báb-tava, Navat-patak, Nyíres-tó, Zsid-tó, Kis-tó, Nagy-tó and Hamvas-tó. The detailed description of their plant communities with the list of the plant spe- cies.– In: Magura, T. (ed.): “Complex habitat rehabilitation of the Central Bereg Plain, Northeast Hungary”. Archives of the Hortobágy National Park Dir., Debrecen, 59 pp.

Nagy, J. (2007):Tavi szukcesszió – úszóláp szukcesszió.(Lake succession – floating mire succession). – In. Botanika III. (Botany III). Nemzeti Tankönyvkiadó Zrt., Budapest, pp.

591–593.

Nagy, J., Cserhalmi, D. and Gál, B. (2008): The reconstruction of vegetation change in the last 55 years on a mire of Bereg plain (Hungary). –Acta Bot. Hung.50(1–2): 163–170.

Pellerin, S. and Lavoie, C. (2003): Reconstructing the recent dynamics of mires using a multitechnique approach. –J. Ecol.91: 1008–1021.

Sickel, H., Ihse, M., Norderhaug, A. and Sickel, M. A. K. (2004): How to monitor semi-natural key habitats in relation to grazing preferences of cattle in mountain summer farm- ing areas: an aerial photo and GPS method study. –Landscape and Urban Planning67:

67–77.

Simon, T. (1953): Torfmoore im Norden des Ungarischen Tieflandes. –Acta Biol. Acad. Sci.

Hung.4: 249–252.

Simon, T. (1960): Die Vegetation der Moore in den Naturschutz-Gebieten des Nördlichen Alföld. –Acta Bot. Acad. Sci. Hung.6: 107–137.

Simon, T. (1968): Die Torfmoor-Gesellschaften Ungarns. –Acta Geogr. Debr.7: 201–206.

Simon, T. (1992): Vegetation change and the protection of the Csaroda relic mires, Hungary.

–Acta Soc. Bot. Poloniae61(1): 63–74.

Soó, R. (1964):A magyar flóra és vegetáció rendszertani-növényföldrajzi kézikönyve. I. (Synopsis systematico-geobotanica florae vegetationisque Hungariae. I).– Akadémiai Kiadó, Buda- pest, 591 pp.

(14)

Sümegi, P. (1999):Reconstruction of flora, soil and landscape evolution, and human impact on the Bereg Plain from late-glacial up to the present, based on palaeoecological analysis. – In: Ha- mar, J. and Sárkány-Kiss, A. (eds): The Upper Tisa Valley, Tiscia Monograph Series, Szeged, pp. 173–204.

Szurdoki, E. and Nagy, J. (2002): Sphagnum dominated mires and Sphagnum occurrences of North-Hungary. –Folia hist.-nat. mus. Matr.26: 67–84.

Vozáry, E. (1957): Pollenanalytische Untersuchung des Torfmoores “Nyírestó” im Nord- osten der Ungarischen Tiefebene (Alföld). –Acta Bot. Acad. Sci. Hung.3(1–2): 123–134.