data acquisition methods for gis of serbian republic water

Transcription

data acquisition methods for gis of serbian republic water
4TH INTERNATIONAL CONFERENCE
RECENT PROBLEMS IN GEODESY AND RELATED FIELDS WITH
INTERNATIONAL IMPORTANCE
February 28 - March 2, 2007, Inter Expo Centre, Sofia, Bulgaria
DATA ACQUISITION METHODS FOR GIS OF SERBIAN
REPUBLIC WATER MANAGEMENT AUTHORITY
Vladimir Pajić, Miro Govedarica, Dubravka Bošković,
Dušan Jovanović, Srdjan Popov (SR)
ABSTRACT
Although there exist large amount of data for water resources in Serbia, GIS of Serbian Republic
Water Management Authority has to provide capabilities for data acquisition and change detection
for water resources. Remote sensing methods could serve that purpose. This paper presents methods
for detection of river basins and surface water bodies. Analysis of river basins is based on 90m
DEM for whole area of Serbia and 5m DEM for area of mountain Fruska Gora. MicroImages
TNTmips 7.0 software has been used for that purpose. Detection of surface water bodies has been
based on Landsat ETM+ images. Classification has been caried out by Leica ERDAS Imagine 9.0
supervised classifier and Feature Analyst 4.0.
KEYWORDS: Remote sensing, River basin, Water body, DEM, Landsat ETM+, Classification
INTRODUCTION
Development study of GIS of Serbian Republic Water Management Authority foresee that data
maintenance and data management will be one of the fundamental functional units of the system. It
includes next subunits: data acquisition, data maitenance, production of analyses and projects,
projects executions and management of objects on the field, and data distribution. This paper relates
to data acquisition.
Although there exist large amount of data for water resources in Serbia, GIS of Serbian Republic
Water Management Authority has to provide capabilities for data aquisition and change detection
for water resources. Our former paper [3] shows the layer structure requested by Water Framework
Directive (WFD) that GIS of Serbian Republic Water Management Authority will have to adopt.
This paper presents our analysis of methods for acquisition of data for layers RIVER BASINS,
SUB-BASINS and SURFACE WATERBODIES. Next chapters presents those methods in detail.
RIVER BASINS EXTRACTION
River basins analysis was based on 90m DEM. Four DEMs, which cover whole territory of Serbia,
are downloaded from http://srtm.csi.cgiar.org/ [1]. They cover area from 40N to 50N deegree of
latitude and 15E to 25E deegree of longitude in WGS84 coordinate system. Then DEMs were
imported in Leica ERDAS Imagine 9.0 and merged through mosaic process. Resulting DEM was
croped from 18N to 24N deegree of longitude, and projected to UTM, zone 34N.
Final step was to crop such projected DEM with Leica ERDAS Imagine 9.0 subset image process.
DEM was croped to following coordinates: ULX = 5267623m, ULY = 282181m, URX =
5269639m, URY = 627031m, LLX = 4623298m, LLY = 261006m, LRX = 4536581m, LRY =
739964m. In that way we get smaller area which also cover whole area of Serbia. That meant
shorter execution time for river basins extraction.
River basins extraction was carried out trough Microimages TNTmips Watershed process. This tool
evaluates and constructs watersheds using an input elevation raster. Terrain characteristics
represented by the elevation raster determine runoff regimes and stream flow.
The Watershed process addresses the influence of terrain on surface water hydrology by modeling
the movement of water over the land surface. The input data for the process is a DEM (Digital
Elevation Model), a regular grid of elevation values stored as a raster object. The Watershed
process computes the local directions of flow and the gradual accumulation of water moving
downslope across the landscape. From these intermediate results the process then computes the
stream network and the boundaries between watersheds, the areas drained by particular stream
systems. Watersheds can be further subdivided into basins associated with particular branches of the
stream network. The flow path network, watershed boundaries, and basins are created as separate
temporary vector objects. Several processing parameters can be adjusted to vary the level of detail
in these objects before saving the final results. Varied attribute information is also created and saved
with the flow paths and watersheds [5].
Time needed for extracting watersheds was, in case of filling all depresions, 2h44m46s. In that case
three large watersheds, which drains water into Black sea, Egean sea and Adriatic sea, were
detected. That can be seen on figure1.
Figure 1: Black sea, Egean sea and Adriatic sea drainage areas
Smaller drainage areas, basins, were also produced. That is shown on figure 2.
Figure 2: Smaller drainage areas
Next step is to aggregate resluting polygons for basins into larger areas which represents river
basins for main rivers in Serbia.
Data was produced in TNTmips internal format which is not siutable for further manipulation. For
that reason, data were exported to shape files which enabled their usage in other GIS software and
their importing into database.
SURFACE WATER BODIES EXTRACTION
For detection of surface water bodies LANDSAT ETM+ were used. LANDSAT ETM+ images are
downloaded from http://glcf.umiacs.umd.edu [2]. Images consist of eight bands, seven multispectral
and one pancrhomatic. Spatial resolution of multispectral bands is 30m and 15m of panchromatic
band.
Band combination 4,5,3, was chosen for detection of surface water bodies because it emphasis
land/water boundary. Both methods that were used, are based on classification of LANDSAT
ETM+ images. Tools used for classification are Feature Analyst 4.0 and Leica ERDAS Imagine 9.0
supervised classifier. Also, both methods are based on training set.
Feature Analyst’s machine learning approach to automated feature extraction incorporates software
agent technology which “learns” to find features like hydrology, vegetation, and other land cover
features based on user-specified examples. The software provides object-specific feature capture
technology using spatial context and advanced machine learning techniques that allow controlling
of feature extraction process rather than using hard-coded rule base [6].
The extraction process, in FA, starts with creation of training set. Training set included polygons
which represents examples of surface water bodies that can be seen on image. It is necessary to
create training set that, in the best way, represents shape and spatial distribution of extracted
objects. Next step, after the training set has been created, was to set up parameters of extraction
process. Those parameters include: image bands that were taken into account, spatial representation
of extracted objects, type of the result – vector or raster, and smallest area that will be extracted.
Band combination that was chosen for extraction process of surface water bodies were 4,5,3
because it emphasis land/water boundary. For input representation was chosen Manhattan, with
pattern width 3, that is shown in figure 3. Such spatial representation has been recommended for
extraction of surface waters. Minimal area that was taken into account was set to 20 pixels
(approximately 0,5 hectares). Results of extraction process are shown in figure 4.
Figure 3: Input representation.
Figure 4: Extracted surface water bodies
Other method used for extraction of surface water bodies is Leica ERDAS Imagine 9.0 supervised
classifier. Classification represents the process of sorting pixels into a finite number of individual
classes, based on their image values. If a pixel satisfies a certain set of criteria, then the pixel is
assigned to the class that corresponds to those criteria. The first part of the classification process is
training of the software system to recognize patterns in the data. Training represents the process of
defining the criteria by which these patterns are recognized. The result of training is a set of
signatures, which are criteria for a set of proposed classes.
Two methods are used for collecting signatures. First method is collecting signatures in feature
space view. That is shown in figure 5a, rounded area represents water in feature space view. Second
method is collecting signatures of water areas in image space view. This method was used to add
water areas that are not in rounded area in feature space view, figure 5b. After collecting sufficient
nuber of signatures, supervised classification was performed. Results of that process was in raster
form and we translated them in vector form for easier manipulation.
Figure 5: a) Feature space view; b) Image space view
Resluts of both classification methods, Feature Analyst 4.0 and ERDAS Imagine 9.0 supervised
classifier, have had minimal diferences. At the end both results are merged to get more complete
data for surface water bodies.
CONCLUSION
This paper presented methods for data acquisition of river basins and surface water bodies. In that
way, data acquisition for two layers required by WFD is covered.
Method for detection of river basins is based on analysis of DEMs. That analysis was performed
trough TNTmips Watershed process. That process produced two results: large watersheds, which
represents three large drainage areas for theritory of Serbia, and basins, which consist those larger
watersheds. Next step is to aggregate basins into larger areas that represents river basins for main
rivers in Serbia.
Methods for detection of surface water bodies, extracts all surface water bodies that exist on image.
Next necesary step is categorisation of surface water bodies in one of the following surface water
categories - rivers, lakes, transitional waters or coastal waters - or as artificial surface water bodies
or heavily modified surface water bodies. For each surface water category, the relevant surface
water bodies within the river basin district shall be differentiated according to type. These types are
those defined using either "system A" or "system B" [4]. We intend to perform that categorisation in
the future, because it is necessary step for creating database of water resources in Serbia.
REFERENCES
[1] Jarvis A., H.I. Reuter, A. Nelson, E. Guevara, 2006, Hole-filled seamless SRTM data V3,
International Centre for Tropical Agriculture (CIAT), available from http://srtm.csi.cgiar.org.
[2] Global Land Cover Facility, Institute for Advanced Computer Studies, University of Maryland,
College Park, USA, http://glcf.umiacs.umd.edu
[3] Defining Layer Structure for GIS of Serbian Republic Water Management Authority (Vladimir
Pajić, PH d Miro Govedarica, Dubravka Bošković, Dušan Jovanović, Srđan Popov) InterGeo East
2006, Conference for Landmanagement, Geoinformation, Building Industry, Envoironment
[4] Directive 2000/60/EC of the European Parliament and of the Council, Annex II
http://www.rec.hu/tisza/WFD_annex2.html
[5] Modeling Watershed Geomorphology, Randall B. Smith, Ph.D., MicroImages, Inc.,
http://www.microimages.com/getstart/watershd.htm
[6] Feature Analyst version 4.1 for Imagine, Reference Manual, Visual Learning Systems, Inc.,
http://www.featureanalyst.com/feature_analyst/platforms/erdas_imagine.htm
[7] ERDAS IMAGINE Tour Guide, 09 December 2005, Geospatial Imaging, LLC, Norcross,
Georgia
AUTHORS
degrees, name, family name: dipl. ing. Vladimir Pajić
organization / company: Faculty of Technical Sciences
address for contact: dr Ilije Đuričića 1, Novi Sad, Serbia
telephone, fax number, e-mail: +381637188904, +38121458873, pajicv@uns.ns.ac.yu
degrees, name, family name: Ph. D. Miro Govedarica
organization / company: Faculty of Technical Sciences
address for contact: dr Ilije Đuričića 1, Novi Sad, Serbia
telephone, fax number, e-mail: +381214852258, +38121458873, miro@uns.ns.ac.yu
degrees, name, family name: dipl. ing. Dubravka Bošković
organization / company: Faculty of Technical Sciences
address for contact: dr Ilije Đuričića 1, Novi Sad, Serbia
telephone, fax number, e-mail: +381214852260, +38121458873, dudab@uns.ns.ac.yu
degrees, name, family name: dipl. ing. Dušan Jovanović
organization / company: Faculty of Technical Sciences
address for contact: dr Ilije Đuričića 1, Novi Sad, Serbia
telephone, fax number, e-mail: +381214852260, +38121458873, dusanbuk@uns.ns.ac.yu
degrees, name, family name: Mr. dipl. ing. Srđan Popov
organization / company: Faculty of Technical Sciences
address for contact: dr Ilije Đuričića 1, Novi Sad, Serbia
telephone, fax number, e-mail: +381214852260, +38121458873, boromir@uns.ns.ac.yu