IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
MINEO IST–1999-10337 Assessing and monitoring the environmental impact of mining activities in Europe using advanced Earth Observation techniques
MINEO (central Europe) environment test site in Germany Contamination /impact mapping and modelling – Final report
Project funded by the European Community under the “Information Society Technology” Programme (1998-2002)
GTK/RS/2004/3 IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Primary authors of this report are:
Christoph Dittmann and Peter Vosen Deutsche Steinkohle AG Division for Engineering Survey & Geoinformation Services (DIG) Karlstrasse 37-39 D-45661 Recklinghausen Germany
Andreas Brunn (Christian Fischer, Wolfgang Busch) Institute of Geotechnical Engineering and Mine Surveying Technical University of Clausthal Erzstrasse 18 D-38678 Clausthal-Zellerfeld Germany
Front cover illustration:
HyMap hyperspectral remote sensing data acquired in August 2000 during the MINEO hyperspectral survey. False colour composite of principal components transformation (R: PC4; G: PC9; B: PC1). The open water body (in red) originates from the relative uplift of the ground-water table due to mining-caused subsidence movements. IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Abstract
The MINEO project is an EC 5th framework research and development project. It aims to develop methods and define new parameters supported by remote sensing data, which will help identify, assess and monitor the environmental impact of the mining industry. The project involves collaboration of seven European geological surveys and representatives from mining industry. Deutsche Steinkohle AG (DSK) is the only industrial undertaking to be involved as a full project partner.
The remote sensing data analysed at the German test site were acquired on 24th August 2000. The data were recorded by HyMap, an airborne, hyperspectral scanner with 126 channels across the wavelength region 0,45 – 2.5 µm. HyMap provides almost contiguous spectral coverage, except the atmospheric water vapour bands. Additionally to the 2000 data set an archival HyMap data set from August 1998 was used for change detection tasks.
DSK operates an environmental monitoring concept which takes the special needs of the environment and of the local population in those areas which are affected by underground mining into account. The results of the MINEO project are intended to be incorporated in this concept in order to improve the quality of the environmental monitoring action.
The Kirchheller Heide (Kirchhellen Heath) north of Bottrop (Ruhr area) was chosen as the test site because high dynamic changes in the environment occur due to the mining activities. The main aim at this test site was to detect vegetation stress caused by mining induced subsidence movements. To achieve this objective actual land cover / land use have been mapped and several parameters from the vegetation spectrum have been derived, which were used to describe the plant health status.
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Content 1 Introduction...... 1 1.1 Motivation for participation in MINEO...... 1 1.2 Co-operation: partners and contractors...... 2 2 Description of the test site ...... 3 2.1 Geography of the test site ...... 3 2.2 Geological outline of the test site ...... 4 2.3 Soils ...... 6 2.4 Hydrological features...... 6 2.5 Vegetation and land use...... 7 2.6 Climate...... 7 2.7 History of exploitation...... 8 2.8 Related environmental problems ...... 9 2.9 Expected contribution of hyperspectral data and GIS modelling ...... 11 3 Hyperspectral data set...... 12 3.1 Data acquisition survey 2000...... 12 3.2 Quality Control ...... 14 4 Other available relevant environmental data...... 16 5 Field spectroradiometry campaign...... 17 5.1 Brief description of the spectroradiometer used ...... 18 5.2 Description of main spectral features representing vegetation and vegetation stress...... 19 5.3 Endmember selection...... 20 5.4 Feeding MSL, spectra categories included in MSL...... 22 6 Data pre-processing ...... 23 6.1 Geometric corrections...... 23 6.2 Atmospheric corrections...... 25 7 Description of image processing procedures and algorithms used in impact mapping 31 7.1 Objective...... 31 7.2 Procedures, algorithms and toolboxes ...... 31 7.2.1 Red edge Analysis ...... 31 7.2.2 Features from continuum removed spectra...... 33 7.2.3 Software tool for the extraction of the features at continuum removed spectra ...... 34 7.2.4 Derivative Analysis ...... 34 7.2.5 Software tool for the extraction of the red edge and derivative analysis features ...... 36 7.3 Land cover & use: forest stands mapping...... 38 7.4 Vegetation stress detection ...... 39 7.5 Change Detection...... 43 7.5.1 Data and Data Preparation for change detection...... 44
I IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
7.5.2 Change Detection: pre-classification...... 46 7.5.3 Change Detection: post-classification ...... 49 7.6 Description and assessment of the maps produced vs. expected results...... 50 7.7 Invaluable contribution of hyperspectral imagery vs. conventional sensors ...... 51 7.8 Generic character of the procedure/algorithm ...... 52 8 Description of the GIS database ...... 53 8.1 Digital Terrain Model (DTM)...... 53 8.2 Colliery and subsidence data ...... 53 8.3 Digital Ground Water Model (DGWM) ...... 53 8.4 Biotope types and actual land use...... 54 9 GIS modelling...... 55 9.1 Description of the conceptual environmental model ...... 55 9.2 Objectives of GIS modelling and expected results...... 55 9.3 Description of GIS derived layers ...... 55 9.4 Generation of a high resolution Digital Terrain Model based on different input data..... 56 9.5 Updating of the Digital Terrain Model with calculated / measured subsidence data ...... 57 9.6 Calculation of ground-water isobath at different times ...... 60 9.7 Hyperspectral data as additional and independent GIS layers...... 60 10 Monitoring in German hard coal mining-legal aspects & DSK’s monitoring concept. 63 10.1 Legal aspects...... 63 10.1.1 The Environmental Impact Assessment in the mining context...... 63 10.1.2 Special features of the planning of mining projects...... 63 10.1.3 The Environmental Impact Study...... 64 10.1.4 Environmental monitoring...... 65 10.2 DSK monitoring concept ...... 66 10.2.1 Detailed thematic concept...... 66 10.2.1.1 Subsidence monitoring ...... 66 10.2.1.2 Surface monitoring ...... 67 10.2.1.3 Hydrological monitoring...... 67 10.2.1.4 Biomonitoring...... 68 10.3 Integrated use of classical, GIS and Remote Sensing methods ...... 70 11 Conclusion, assessment of results ...... 72 11.1 Assessment of results...... 72 11.2 Results versus user demand ...... 72 11.3 Future plans ...... 72
II IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
List of figures
Figure 2.1: Location of the test site...... 3
Figure 2.2: Geological overview of the Münsterland area (DROZDZEWSKI, 1995)...... 4 Figure 2.3: Geological map showing details of the Rhine-Westphalia coalfield and areas adjoining to
the north (WALTER, 1995; according to GLA Krefeld, 1988)...... 5 Figure 2.4: Overview of the coal production and the personnel development in German coal industry
since 1945 (source: STATISTIK DER KOHLENWIRTSCHAFT E.V.)...... 8 Figure 3.1: Airborne data acquisition over the Kirchheller Heide test site. The border of the EIS area is shown in red; yellow crosses delineate the flight paths...... 13 Figure 3.2: Signal to noise ratio for each HyMap channel...... 14 Figure 5.1: Geocoded subset of HyMap image (r:0,890 µm; g:0,662 µm; b:0,507 µm) with ground reference measurements (A). Examples of recorded ground reference spectra (B); the
shaded vertical bars indicate regions of strong atmospheric water vapour and CO2 absorption...... 17
Figure 5.2: Factors influencing the spectral curve of green leaves (BACH, 1995)...... 19 Figure 5.3: Image spectra derived for a single forest class (left) and statistically selected min., mean and max. spectra...... 20 Figure 5.4: 2D scatter plot of Spruce trees...... 21 Figure 5.5: Endmembers plot for different kinds of Pine trees (Pinus Sylvestris)...... 21 Figure 6.1: Pre-processing workflow established for geocoding and atmospheric correction of HyMap data using PARGE & ATCOR-4...... 23 Figure 6.2: Example of geocoded HyMap image combined with GIS vector data sets...... 25 Figure 6.3: Schematic sketch of radiation components: 1) path radiance; 2) reflected radiation from the viewed pixel; 3) adjacency radiation; 4) terrain radiation reflected to the pixel...... 25 Figure 6.4: Schematic workflow of atmospheric correction with ATCOR-4 ...... 27 Figure 6.5: Water surface reflectance (Lake) retrieved after the single target calibration of field P-05 with its ground-reflectance and the modified water surface reflectance (Lake(mod)) as used for the inflight calibration; the vertical shaded bar marks the region with the best fit after single reference target calibration over field P-05...... 29 Figure 6.6: Comparison of ground-reflectance spectra with the corresponding HyMap reflectance spectra for different fields; the shaded vertical bars mark regions of strong atmospheric
water vapour and CO2 absorption...... 29
III IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 7.1: Typical vegetation reflectance spectrum highlighting the red edge between 0,68 and 0,80 µm...... 31 Figure 7.2: HyMap Spectrum (left) with its first (middle) and second (right) derivative...... 32
Figure 7.3: Continuum-removed example of a chlorophyll absorption in vegetation , (RENCZ, 1999) (a) and an example of the absorption band parameters: ‘band depth’, ‘band width’ and ’FWHM’ (b)...... 33 Figure 7.4: Tool for the determination of the features from continuum removed spectra...... 34 Figure 7.5: Original spectrum and median filter results using different window sizes...... 36 Figure 7.6: Area of the first derivative over the whole spectral range (left) and Area of the first derivative in the wavelength range between 540 and 780 nm, which is an interesting range for plant analysis (right)...... 36 Figure 7.7: IDL toolbox to calculate the spectrum feature parameters associated with the red edge...... 37 Figure 7.8: Examples of plant stress mapping results using different promising features...... 41 Figure 7.9: Comparison of plant stress on pines overlaid with iso-lines of changing depth of ground water level (line labels indicate the number of change in meters)...... 42 Figure 7.10: Ground control points spread over the image...... 45 Figure 7.11: Registration results...... 46 Figure 7.12: Visual interpretation of three data sets recorded over a period of five years...... 47 Figure 7.13: NDVI images derived from the HyMap data sets. These two pictures show the alteration of the vegetation vitality within the subsidence lake. High values (white) indicate high vitality, low values (black) indicate low vitality...... 48 Figure 7.14: Visual comparison of the two hyperspectral data sets. False colour composite (R:PC4, G:PC9, B:PC1). See text for details...... 48 Figure 7.15: Change detection NDI(PC1) image. This image is colour coded to facilitate the visual interpretation. Bluish colours indicate no or little change, yellow and red show areas of high change degree. Areas I and II show detected changes. The size accretion of the subsidence lake are highlighted at II; the field at I laid idle in year 2000 in contrast to year 1998; the red pixels at III results either due to miss-registration of the images or due to different illumination conditions, causing shadow effects...... 49 Figure 7.16: Picture on the left shows the result derived from the 2nd derivative between 550 and 740 nm, scaled image to change factor, increasing plant stress from blue to red; the right picture shows the same result scaled for pine trees and overlaid on a HyMap channel, increasing stress from green to red...... 50
IV IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 9.1: A simplified conceptual model of environmental impacts due to subsidence movements at the German MINEO test site...... 55 Figure 9.2: Digital terrain model (3D view, elevation from 40 m a.s.l. (yellow) to 100 m a.s.l. (light brown)) with integrated watercourse network (blue lines). Wooden areas with old trees are shown as olive green blocks)...... 56 Figure 9.3: Longitudinal section – derived from the four-dimensional watercourse network...... 57 Figure 9.4: DTM with integrated subsidence data. Area affected by subsidence is shown by black dashed line, subsidence isolines are depicted in blue with centres of max. subsidence...... 58 Figure 9.5: 3D elevation view (from green = 30 m a.s.l. to grey = 70 m a.s.l.), watercourse network (blue lines) and lakes (light blue). Estimation of sensitive areas at risk of ground water- logging. The dark blue spots indicate sites of the modelled ground water logging for 1993 (left) and 2019 (right)...... 59 Figure 9.6: Updated land use map, here forest stands (light green areas), overlaid on the HyMap data (R = 0.890 µm, G = 1.685 µm, B = 0.662 µm) in 3D-view draped over the DTM...... 59 Figure 9.7: Detail of the terrestrial land cover and use map, showing the relevant area marked with AA02 = beech forest...... 61 Figure 9.8: Detail of HyMap derived forest stands, showing the distribution of the different tree species within the same area as shown in figure 9.7...... 61 Figure 9.9: Change of ground water table (yellow areas = lowering, light lilac areas = no change, dark lilac areas = increasing less than 1.75 meters, area with red outline = increasing more than 1.75 meters) and plant stress on pines (increasing stress from green to red pixels)...... 62 Figure 10.1:DSK-EIS-areas in the Ruhr District (ca. 750 km²) – including the MINEO test site Kirchheller Heide (“PH” – Prosper-Haniel mine) ...... 64 Figure 10.2: Process of environmental planning and monitoring...... 66 Figure 10.3: DSK monitoring concept – basis for optimized data processing...... 70
V IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
List of Tables
Table 3.1: Spatial configuration of the HyMap sensor (COCKS, 1998) ...... 12
Table 3.2: Spectral configuration of the HyMap sensor (COCKS, 1998)...... 12 Table 3.3: List of delivered items acquired during the hyperspectral data acquisition survey 2000...... 13 Table 4.1: Selection of remote sensing data records held by DSK...... 16 Table 5.1: Technical characteristics of employed spectrometer devices during the field campaign (source: ASD Inc.: http://www.asdi.com; GER: http://www.ger.com GER Mark V IRIS users guide, respectively)...... 18 Table 6.1: Modular structure of ATCOR-4...... 26 Table 7.1: Plant Indicators for moisture and Nutrient content of the soil. Higher values indicate, that the plant usually prefers locations with higher content. i indicates indifferent and means,
that the plant usually lives on all locations.(ELLENBERG, 1992) ...... 39 Table 7.2: Example of predefined feature values for pines (pinus sylvestris)...... 40 Table 7.3: Comparison of features inside and outside the sphere of influence...... 43 Table 7.4: Accuracy assessment of the maximum likelihood classification; all-over accuracy reaches 73 % and the kappa coefficient is 0,6519...... 50 Table 7.5: Accuracy assessment of the maximum likelihood classification. continuation of Table 7.4.....51
VI IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
List of appendices
Appendix I: Land use & cover map Appendix I a: Land use & cover map, Subarea I Appendix I b: Land use & cover map, Subarea II Appendix II: Vegetation stress map Appendix III: Vegetation stress changes on Pine stands, August 1998 to August 2000 Appendix IV a: Workflow procedure for HyMap data pre-processing “KH_GEOC” and “KH_ATMC” Appendix IV b: Workflow procedure developed for vegetation stress detection “KH_VEGS” Appendix IV c: Workflow procedure developed for change detection “KH_CH_DET_VEGS” Appendix IV d: Workflow procedure developed for DTM generation “KH_GIS_MAKE_DTM” Appendix IV e: Workflow procedure developed for DTM update “KH_GIS_UPDATE_DTM” Appendix IV f: Workflow procedure developed for isobath calculation “KH_GIS_ISOBATH” Appendix V: List of persons working over the test site.
VII IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
1 INTRODUCTION Mining industry provides the raw materials for a wide spectrum of industrial branches, from metal industries through energy to chemical industries with its all further processing branches and associated lines of business. The world economy is based upon the mining products considerably and last but not least the European mining and extractive industry contributes about 7% of the gross domestic product of the EU from this resource. But according to its nature, mining industry leads to changes in environment inevitably and that is why the European mining industry faces increasing environmental pressure and regulatory controls. Therefore a sound management of mining activities is of high concern to preserve nature and minimise the environmental impacts. Furthermore understanding and monitoring the environmental impact processes in mining areas is crucial for sustainable management of Earth’s environment.
The shared-cost RTD project MINEO under the programme of Information Societies Technology with the contract number IST-1999-10337 offers an excellent platform on European scale to investigate and develop new techniques and methods using the Earth Observation technology with integrated systems for coherent environmental management, supporting the entire cycle from prevention, identification and mitigation of mining-induced impacts.
This final report from the Central European environment test site Kirchheller Heide in Germany describes the works carried out and presents the results for the objectives as they have been defined for the work package WP2-4 in the Annex 1 “Description of Work” of the IST-Contract. The aims for the WP2-4 can be shortly outlined as follows (ref.: Annex1: Description of Work):
- development of a conceptual link between mining activities and hydrological and ecological parameters, - development of effective remote sensing methods for assessing the environmental impact of mining operations; this includes the description and localisation of the mining impacts, e.g. detection of water logging areas, assessment of vegetation vitality based on change detection and multi-temporal image analyses, - development of guidelines for long-term monitoring, taking remote sensing methods into account.
1.1 Motivation for participation in MINEO The coalfield in the northern Ruhr district contains extensive mine workings which are now under the management of Deutsche Steinkohle AG (DSK). Longwall mining tends to predominate. For economical and technical reasons no stowing material is used in the newly-created workings. Local rock pressure causes the overlying strata to collapse into the mine cavities, which are almost completely filled as a result. This caving system creates a breakage zone directly above the coal extraction area which extends upwards to reach the surface as a flexure of the strata. The result is a surface subsidence trough. The extend of the resulting subsidence is determined by local geology, working seam thickness, winning depth, the area being worked and the extraction intensity.
The mining licensing procedure, combined with environmental constraints and mining law, claims that DSK has to record and monitor such phenomena extensively and methodically to provide accurate and reliable forecasts of their anticipated impact. For that reason special emphasis has been placed on the development of operationally effective working methods exerting remote sensing (RS) and geographical information system (GIS) technologies on the establishment of parameters for these methods of environmental description and assessment. This is intended to be incorporated as a new back-up data level for DSK’s environmental monitoring concept.
Deutsche Steinkohle AG Division for Engineering Survey 1 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
1.2 Co-operation: partners and contractors Deutsche Steinkohle AG (DSK) was in charge of the work package WP2-4, which comprises all tasks related to the Central Europe environmental test site Kirchheller Heide. Besides broad and extensive collaboration with all partners of the MINEO consortium the following partners and contractors co-operated with DSK on different tasks during the MINEO lifetime:
- French Geological Survey (BRGM): overall co-ordination of the MINEO project, - Danish Geological Survey (GEUS): overall organisation and co-ordination of the HyMap Sensor aerial survey for every MINEO test site, - HyVista: subcontracted for HyMap survey, - German Federal Institute for Geosciences and Natural Resources (BGR): overall spectro- radiometric reference field survey for every MINEO test site and additional laboratory measurements on soil and rock samples form the German test site, - Geo-Research-Centre Potsdam (GFZ-Potsdam): subcontracted for additional spectro-radiometric field reference measurements at the German test site, - German Space Agency (DLR): consulting service for Hymap data pre-processing (atmospheric correction), - Technical University Clausthal, Institute of Geotechnical Engineering and Mine Surveying (TUC- IGMC): scientific consulting service and support for the German test site during the project life time.
Deutsche Steinkohle AG Division for Engineering Survey 2 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
2 DESCRIPTION OF THE TEST SITE The Central Europe environmental test site Kirchheller Heide is situated in western Germany, in the northern part of the Ruhr district, one of the most congested urban areas in Europe with about 7,5 million inhabitants. The size (ca. 120 square kilometres) and location of this test site follows the area which has been defined for the environmental impact assessment (EIA) for the mine field of the mine Prosper-Haniel. About 40% of the test site area is covered with forests of different types and 50% is used for agricultural land; about 10% is urban area (Kirchhellen village). Because of the rural character this test site has a high ecological value for that region and a great importance for the inhabitants in terms of recreation. The test site is surrounded by the following towns: Bottrop and Oberhausen in the south, Dinslaken in the west, Hünxe in the north and Gladbeck in the east.
The choice for Kirchheller Heide as the Central European MINEO test site has been driven by several factors. The main factor was the active coal mining in this area. Though the coal is produced in depths of 700 m and deeper, the impacts of this activity reach the surface as a subsidence occurrence and claim to be monitored and managed in terms of mitigation these influences and preserve the hydrological and ecological balance in this area. Changes in ecology, especially vegetation vitality in forest stands, and alterations in biotope type composition are of high relevance not only for the mine itself but also for other parties involved in this area: agriculturists, forestries and last but not least for the inhabitants of the Ruhr district, for whom this area became an important recreation area with its nature and landscape reserves. Additionally extensive and updated records of GIS and remote sensing data sets covering this area were applied to carry out the tasks and achieve the objectives laid down for the MINEO project.
2.1 Geography of the test site The landscape of the test site is characterised by the sandy plate region of the lower Rhine. In topographic terms the area is bounded in the west by a slight downward slope to the Dinslaken Rhine plain. In the south the Emscherland is located and the test site is expanded into this area. In the east and north the Kirchheller Heide is bordered by the Münsterland. The interface between the south-west Münsterland and the eastern lowlands of the lower Rhine is characterised by a morphologically weak formed relief. The area of the test site is largely occupied by coherent remnants of the main Rhine terrace, which is known as the Königshard main terrace plate. To the north the area adjoins part of the Hünxe-Gahlen undulations, which have a maximum height of 50 metres above sea level. The eastern offshoots of the test site are formed by eolian sand deposits.
Hamburg Berlin Kirchheller Heide
Munich
Figure 2.1: Location of the test site.
Deutsche Steinkohle AG Division for Engineering Survey 3 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
2.2 Geological outline of the test site
Figure 2.2: Geological overview of the Münsterland area (DROZDZEWSKI, 1995).
The investigation area is composed of sediments from the Upper Cretaceous, Tertiary and Quaternary Series (Figure 2.2). The basement rocks comprises the folded strata of the Upper Carboniferous Series, which is overlain discordantly by the 1,000 metres-thick cretaceous sediments.
The coal worked in the test area is derived from the Bochum, Essen, Horst and Dorsten strata, which can be ranked stratigraphically into Westphalian A, B and C series. The surrounding rock comprises sandstone, sandy shale and shaly clay. The sedimentation process is characterised by cyclic repeating of sedimentation material from coarse (sandstone, sandy shale) to fine grained material (shaly clay) with intermediary peat bogs and coal seams respectively and back again from fine grained to more coarse grained material. After FÜCHTBAUER (1988) these sediments have been deposited in a river delta environment. After the deposition of the Westphalian series the coal layers underwent folding during the Asturian phase of Variscian orogenesis and were pushed up to form flat overthrusts. Most of these structures take the form of upfolds of varying degrees of steepness, these being intermediary
Deutsche Steinkohle AG Division for Engineering Survey 4 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
alternated by low synclines several kilometres in width. The coal deposits being extracted beneath the Kirchheller Heide lie inside the Lippe syncline (Figure 2.3).
Figure 2.3: Geological map showing details of the Rhine-Westphalia coalfield and areas adjoining to the north (WALTER, 1995; according to GLA Krefeld, 1988).
The deposits of the Cenomanian and Turonian Series form the foot wall of the cretaceous sediments. These deposits are overlain by the Emscher marl (Coniacian and Santonian Stages), which is a silty clay-marl rock with a distribution of fine sand. As the hanging wall adjoin a sequence of solid fine- sand marlstones (the Recklinghausen sandy marls) from the Upper Middle Santonian and marly fine to medium-grained sandstones with calcareous sandstone beds (the Osterfeld sands) from the Upper Santonian (MESSER, 1996). The most recent of the cretaceous sediments is formed by the Bottrop
Deutsche Steinkohle AG Division for Engineering Survey 5 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
marl, which consists of glauconitic clay-marl to fine-sand marlstone dating from the Lower Campanian Stage. The Upper Cretaceous beds are followed by the Tertiary deposits, which are located in the Oligocene strata. The foot wall of the Tertiary sequence starts with the fine-sandy Walsum marine sands, which are overlain by the Ratinger clays - mainly calciferous mudstones some 8 to 10 metres in thickness. This is followed in the hanging wall by the Lintorf beds, which are composed of argillaceous fine-sand silts and silty fine-sands up to 100 metres thick (MESSER, 1996).
The Tertiary strata and Bottrop marl beds die away in the east of the test zone, such that the Osterfeld sands in this area are directly overlain by the Quaternary deposits. To the north-west and south-east the area is crossed by the Kirchhellen cretaceous anticline, which becomes the Bottrop cretaceous syncline in the south-west and the Dorsten cretaceous syncline in the north-east. The Bottrop marls, Osterfeld sands and Recklinghausen sandy marls are absent from the centre of the anticline, with the result that in this area the Walsum marine sands bear directly against the Emscher marl.
The oldest Quaternary sediments in the area of investigation are formed by the Lower Pleistocene main terrace of the Rhine, which is composed primarily of sands and gravels. This main terrace varies from 3 to 5 metres in thickness (DWORSCHAK et al., 1998). As a glacial sediment, the Kirchheller Heide area also contains base moraine from the Saale Ice Age, which can be assigned to the Drenthe Stage. In the Rotbach valley the base moraine is up to 12 metres thick, while in the northern part of the site it is only present as a thin cover less than 3 metres in thickness.
The youngest of the Pleistocene sediments comprise thin beds of eolian sand. These are the most widely spread of all the Quaternary deposits, being frequently blown up into dune systems. The Holocene deposits are composed of lowland alluvial plain deposits up to 5 metres thick, together with occasional areas of low moorland and transition moorlands. The present morphology has been formed during the Holocene, whereas nowadays cultural activities changes the landscape significantly.
2.3 Soils The various beds of initial geological substrates, combined with other soil-forming factors, have resulted in a considerable differentiation of the soil layers. In the area under consideration terrestrial and semi-terrestrial soils alternate with each other. Gleysols have formed on the low-permeable sediments. This can be observed in the dispersion area of the Saale glacial clay. Gleysols are particularly prevalent in the south of the investigation zone. Perched ground water has resulted in the formation of gleysols whose distribution extends primarily to the alluvial zones and spring outlets.
Podzol soils with transition to brown soils tend to be found especially in the higher layers, which have mainly been formed from the main-terrace sediments, and also in the cover-sand distribution zone. Eutrophic brown soils mainly occurs in the north-eastern emergence point of the cretaceous sands and marls (Recklinghausen sandy marls and Osterfeld sands).
The test site also contains isolated Plaggen-type soils - anthropogenic soils with a thick humus cover - which manifest that the area was used for agriculture in earlier times (DWORSCHAK et al., 1988).
2.4 Hydrological features The test site is traversed typically by meandering lowland and bottom-land streams which drain into the three water courses of the Lippe, Emscher and Rhine. The main part of the investigation area is drained by the river Rotbach and its tributaries. While the upper course of the Rotbach, along with its tributary - the “Schwarze Bach” - is still in a natural state, the middle course of this river has been extensively developed. In the north the catchment areas of the “Bruckhauser Mühlenbach” and “Gartroper Mühlenbach” are found. The south-eastern part of the area is drained into the Emscher by
Deutsche Steinkohle AG Division for Engineering Survey 6 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
the Boye and its tributaries. In addition to the flowing water courses the “Kirchheller Heide” contains a large number of springs (usually outcrop springs) and stillwater areas. The latter generally comprise water-filled excavation pits (e.g. the “Kirchheller Heidesee” and the “Heidhofsee”), as well as fish ponds.
The investigation area contains three ground-water storeys, which in some parts are in hydraulic contact with each other because of the geological situation. The first ground-water storey lies in the Quaternary aquifer. Because of the heterogeneous development of the deposits the water flow conditions vary considerably. The main terrace sediments of the Rhine constitute the most important pore aquifer. In addition, ground water also occurs in the thin cover sands and in the flood plain deposits of the streams and brooks. The Saale glacial clay acts as an aquifer and separates the ground- water storeys from each other.
The first storey is separated from the second, situated in the Walsum marine sands, by the Lintorf and Ratinger beds. The third storey is located in the beds of the Recklinghausen sandy marls and Osterfeld sands. This ground-water storey is separated from the second storey by the Bottrop marls. The third storey is underlain by the Emscher marl, which prevents any hydraulic contact with the aquifer of the Turonian limestones and carboniferous strata. Because of the geological situation the three ground-water storeys in the east of the investigation area are in hydraulic contact with each other, which leads to a reduction in the number of storeys. The number of storeys is also reduced in the centre of the Kirchhellen cretaceous anticline. The second storey is not present in this area.
2.5 Vegetation and land use The Kirchheller Heide is characterized by alternating woodland and agricultural land. The large interconnected areas of natural woodland comprise “Köllnischer Wald”, “Hiesfelder Wald”, “Scholtenbusch”, “Krummbeck”, “Schlägerhardt”, “Hohe Wart”, “Bruckhauser Wald” and “Hünxer Wald”. However, the natural woodland often alternates with forest areas containing non-local species of trees. The natural alders which grow along the sides of the streams and brooks are especially valuable from an ecological point of view; these tree areas, which line the upper courses of the Rotbach and Schwarzer Bach, have partly been reduced by local drainage operations.
Because of the large areas of perched ground water the heathland has traditionally seen a large percentage of pasture land use, though local drainage projects have converted much of the area to agriculture (RHINELAND CHAMBER OF AGRICULTURE, 1993). This farming land is almost exclusively devoted to the production of grain and maize for fodder. The growing of sugar-beet, potatoes and rape is now of secondary importance.
Large areas of sand and gravel are extracted from the area under investigation. This material comprises gravel beds from the main Rhine terrace and sands from the eolian sand cover. The extraction activities are concentrated along a strip west of Kirchhellen which runs north to south along the L621 road. After the extraction pits have been exhausted the cavities are either filled in or left as open expanses of water, such as the Heidesee.
2.6 Climate The test site is located in the transition zone between maritime and continental climate which belongs to the west wind zone. This climate is generally characterised by cool, rainy summers and mild winters. Occasionally a continental influence sets in with longer periods of high barometric pressure, with the result that in summer with easterly and south-easterly winds the temperature level is higher and the weather conditions are dry, warm and summery. In winter the continental weather conditions
Deutsche Steinkohle AG Division for Engineering Survey 7 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
are frequently associated with cold spells. The local climate conditions of the test site represents also the contrast between the conurbation with urban and industry areas to the south and the rural characterised space in centre and to the north. This contrast can be seen in the records of higher temperature for the southern part and lower for the northern.
2.7 History of exploitation The first historical documents describing the coal mining and its usage in the Ruhr area are dated back to 13th century. In this time hard coal was mined in the southern part of the Ruhr district where it reaches the surface. In the following centuries a slow development of the mining technology took place enabling mining in first shallow underground mines. The first shafts have been built in the 16th century. Mining of coal in this region on big scale began first with the industrialisation at the beginning of 19th century. In this time there were already 229 mines in the Ruhr coalfield with approx. 400 000 tons per year produced coal. The increasing needs of coal for steel and iron production made the mines spread all over the Ruhr area; the coal production grew considerably. In the year 1873 there were 262 mines with a production of ca. 16 million tons per year. At the end of the 19th century the mines in the Ruhr area were organised in a centrally managed market sharing cartel, the Rhine- Westphalia Coal Syndicate, which was replaced in 1946 by the North German Coal Distribution Office (STATISTIK DER KOHLENWIRTSCHAFT E.V.).
Coal production in Germany since 1945 160,0 150,0 145,6 173 160
120,0 146 113,7 s e on in i n
t 120 87,9 mi 71,0 f
oduc 80,0 80 illion tons er o
pr 54,2 l m 47,3 69 41,9
40,5 mb 34,3 40,0 u Coa 27,9 40 N 39 27 19 17 15 15 12 11 0,0 0 1957 1960 1970 1980 1990 1995 1997 1998 1999 2000 2001 Year Coal production Number of mines
Personnel development in German coal mines 700 8.000
600 7.000
500 6.000 5.000 on man x 1000 400 4.000 300 ft [kg] hi
3.000 s / rsonnel
200 producti e 2.000 P
100 1.000 coal 0 0 1945 1955 1965 1975 1985 1995 Year Personnel Performance man / shift
Figure 2.4: Overview of the coal production and the personnel development in German coal industry since 1945 (source: STATISTIK DER KOHLENWIRTSCHAFT E.V.).
After World War II the mining industry underwent a small renaissance in Germany. In 1954 there were 183 mines spread over Germany, most of them in the Ruhr coalfield. The coal production increased until it reached its peak with 151 million tons in the year 1956 and the total amount of employees in the mines was more than 600.000 in year 1957. These few numbers reflect the socio-
Deutsche Steinkohle AG Division for Engineering Survey 8 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
economical importance of the mining industry in Germany and especially in the Ruhr district. Since the sixties the amount of mines, the number of employees and the production continuously declines in consequence of increasing production costs due to the difficult production conditions. Nevertheless the production performance for one man per shift is still increasing reflecting the strong efforts having been put in the development and application of high technology in coal underground mining (ref. Figure 2.4: Overview of the coal production and the personnel development in German coal industry since 1945 (source: STATISTIK DER KOHLENWIRTSCHAFT E.V.).
Decreasing coal production has brought along changes in the industry's organisation. Two major producers, Saarberg AG and Ruhrkohle Bergbau AG, merged in 1998 to form Deutsche Steinkohle AG (DSK), which accounts for 100% of German coal production. Currently (year 2001) DSK operates 11 mines with 27,9 million tons/a coal production, one cocking plant and 2 briquette plants. The major part of the produced coal is used in power plants (73,8%) and ironworks (22,5%).
The mine Prosper-Haniel operating within the MINEO test site Kirchheller Heide came into being in 1974 after merging two previous mines: Prosper and Jacobi/Franz Haniel. These mines were working since 1861 and 1871 respectively. The Prosper-Haniel mine produces currently 3,6 million tons of coal per year with 4478 employees from depths between 630 m up to 1050 m below the surface. The mining claim is 165, 4 km² in size, which roughly represents the test site in size and position.
The coal production is based on very high advanced mining technology. The most applied technology is the longwall mining technique. Here, the deposits are subdivided into strips with a width of approx. 250 to 350 m and a length in some cases of more than 2000 m (“panels”) and than mined in sequence. The coal seam is cut or peeled applying shearer loader or planing machines respectively. For technical and geological as well as economical reasons (additional costs about 30 millions EUR/a) generally no stowing material is used. The average daily production of coal is about 14000 up to 15000 tons, which cumulate in an annual production of ca. 3,6 million tons. The coal mined at the Prosper-Haniel mine is of high quality and therefore used for electricity Figure 2.5: Example of the layout of the generation. mining areas for underground coal-mining (longwall mining).
2.8 Related environmental problems Underground mining inevitably produces cavities in the rock. These cavities are either may filled with ‘imported’ stowing material, or the overlying roof strata in non-cohesive rock collapses and fills the cavity (ref. Figure 2.6). A backfill with stowing material would reduce the absolute subsidence referred to each mined seam panel (seam area) up to 40 %. However a prevailing part of the seams partially or at its whole are below the minimum seam thickness for operation with pneumatic stowing technique. For the period under consideration - here in the case of Prosper-Haniel mine until year 2019
Deutsche Steinkohle AG Division for Engineering Survey 9 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
– the total reduction of subsidence at the surface with stowing would diminish by 15 %. This small amount of subsidence reduction would not justify the high costs for application the stowing method and therefore the caving method without stowing is applied. The longwall mining customary in the Ruhr coalfield, mining without stowing accounts for a portion of 88,9% of production in 1990 (DEUTSCHE STEINKOHLE AG BERGWERK PROSPER-HANIEL, 1999). In the case of the caving method of mining, a dislocation zone which transforms in the upward direction into flexure of the strata occurs directly above the mined area. A subsidence is formed on the surface. The shape of this subsidence will depend, inter alia, on depth of mining, the dip and thickness of the seams, and the length and width of the mined area. Depending on the thickness of the overburden and the extent of the area mined, the impacts on the surface start to occur after around six months. At the end of approximately two years, around 80% of the maximum anticipated subsidence has occurred, and the final condition is achieved after about five years. In the Ruhr coalfield, the amount of subsidence varies between a few meters in the older, more southerly mining districts, and 10 to 20 m in the modern mining region between the Emscher und Lippe river.
Figure 2.6: The occurrence of a subsidence (schematic, not to scale).
During active and future mining activities subsidence movements will occur. This phenomenon leads to specific problems, especially related to the hydrological situation. Water-logging zones may occur. Changes of the ecological conditions in this small-parcelled area can lead to influences and changes of the ecological situation. Alterations in the biosphere system due to human activities are related to different dynamics: ground water situation, watercourse and landscape ecology.
Meandering small streams are very sensitive related to changes in slopes and watersheds. Topographic changes lead to different parameters of the draining regime. While these effects can be reduced for the stream network by engineering techniques, spatial changes of the soil wetness and changes of the infiltration capacity can not be evaluated and corrected easily. Furthermore changing the topographic conditions affect different parameters of the ground-water situation, which is reflected by the changes of the spatial situation of the ground-water and its subsurface watersheds. Different ground-water recharge rates and changes of the ground-water table will result as long-lasting effects. Besides the serious consequences related to water resources, these alterations have a strong influence to the growing situation and ecology for specific vegetation types, e.g. forests, at different locations. Vegetation stress and alteration in plant community composition may be the consequence. As this area serves on recreation on a considerable extent and comprises nature preserve areas, human support is needed to minimise negative effects of mining activities. It is important to detect the spatial and temporal variations in each specific ecosystem.
Deutsche Steinkohle AG Division for Engineering Survey 10 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
2.9 Expected contribution of hyperspectral data and GIS modelling The spatial and chronological dynamics of the impacts which occur in the area under observation is determined by the extent of the present soil movement, groundwater situation and the vegetation. A change over time occurs additionally as a result of changes in land use e.g. by the construction or rededication of sites, and by changing in agricultural planting methods. These background conditions result in the necessity for a forecast of the potential impacts of individual mining-plan in the context of supervision and monitoring of current mining activities and approval procedures within ongoing projects. Mining Projects are subject to corresponding legal regulations based on §§ 50 ff. of the Bundesberggesetz (BBERGG, 1980 = Federal Mining Act) (ref. to chapter 10), in which approval and supervision of these projects, including the Environmental Impact Assessment (EIA) are regulated.
This Environmental Impact Assessment can take advantage of regularly updated environmental data base layers related to mining environments. Earth Observation (EO) techniques can meet this demand. Remote sensing data provide information over large area quantitatively and non-destructively. EO data, when integrated into Geographic Information Systems (GIS) and combined with other data relevant to environmental concerns, have been proven valuable in the environmental impact assessment of mining. Imaging hyperspectral sensors produce data that can characterise the chemical/mineralogical composition and biophysical parameters of the imaged ground surface. In particular they can be used in the production of environmental impact maps around mining areas.
In the case of the Kirchheller Heide test site the environmental impact is more of indirect nature. The mining induced subsidence in combination with the shallow ground-water level leads to ground-water- logging occurrences, which on his part may result in vegetation stress and diminished vegetation vitality. Vegetation vitality estimation and vegetation stress detection caused by mining operations is expected by using imaging spectroscopy data. Furthermore applying such data sets from different dates a change detection can be conducted in terms of environmental impact monitoring and supervision. Detailed description of applied image processing algorithms and methods is given in chapter 7.
Deutsche Steinkohle AG Division for Engineering Survey 11 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
3 HYPERSPECTRAL DATA SET In order to achieve the objectives defined for the test site Kirchheller Heide in the framework of the MINEO project an airborne hyperspectral data survey was performed in August 2000. The hyperspectral data acquisition was co-ordinated by the Danish Geological Survey (GEUS), as for every MINEO test site. The hyperspectral image data collection was accomplished by HyVista with the airborne HyMap sensor. The HyMap sensor records a digital image of the earth’s sunlit surface in a large number of wavelengths. The recorded wavelength range reach from 450 nm to 2480 nm. These wavelengths are recorded in four spectrometers: VIS, NIR, SWIR1 and SWIR2. The HyMap records an image by using a rotating scan mirror which scans the image line by line using the aircraft’s motion speed. This sensor is currently one of the most advanced hyperspectral scanners for not military purposes. It is designed for earth resource remote sensing. Some characteristics of the sensor are given in the tables below.
Spatial configuration IFOV 2,5 mr along track 2,0 mr across track FOV 61,3 degrees (512 pixels) Swath 2,3 km at 5 m IFOV (along track) 4,6 km at 10 m IFOV (along track) Table 3.1: Spatial configuration of the HyMap sensor (COCKS, 1998) .
Module Spectral range Bandwidth across Average spectral module sampling interval VIS 0,45 – 0,89 µm 15 – 16 nm 15 nm NIR 0,89 – 1,35 µm 15 – 16 nm 15 nm SWIR1 1,40 – 1,80 µm 15 – 16 nm 13 nm SWIR2 1,95 – 2,48 µm 18 – 20 nm 17 nm Table 3.2: Spectral configuration of the HyMap sensor (COCKS, 1998)
To fulfill the tasks laid down for the German test site two HyMap data sets has been used. First data set was acquired during the MINEO Hyperspectral Airborne Survey on August 24th 2000. The second data set was acquired on August 20th 1998 in the frame of an DSK internal R&D project no. 50-EE- 9652 (BENECKE ET AL., 2000). Both HyMap data sets cover the same area. The data pre-processing, like geocoding and atmospheric correction was performed on both data sets as described in chapter 6. To achieve the objectives outlined above, the analysis work has been focused on the HyMap2000 data. Nevertheless the HyMap1998 data also have been processed, primary in terms of change detection techniques.
3.1 Data acquisition survey 2000 On August 24th 2000 the MINEO central European test site was flown by an aircraft equipped with the HyMap scanner and with a RMK camera. The size of the area recorded by the HyMap sensor covers the environmental impact study (EIS) area of the mine Prosper-Haniel (ref. Figure 3.1).
The flight altitude of 2200 m allowed to record HyMap data with a spatial ground resolution of 5x5 m. Seven flight lines have been flown, with seven strips recorded and delivered to DSK. Whereas the sixth strip has been flown twice due to unexpected aircraft motion. This strip has not been used for the image analysis tasks. The resulting six strips of hyperspectral data were of good quality. The aerial photographs were taken with a scale of 1:13000. The data have been delivered end of December 2000. The photographs were delivered as black and white negatives instead of CIR photographs as originally
Deutsche Steinkohle AG Division for Engineering Survey 12 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
planned. The table below lists the delivered items acquired during the hyperspectral data acquisition survey 2000.
Figure 3.1: Airborne data acquisition over the Kirchheller Heide test site. The border of the EIS area is shown in red; yellow crosses delineate the flight paths.
Hyperspectral data. Description Strips 1-6a + 6b. List of files. RADIANCE.BIL 126 band image data file in BIL format; converted to physical units of radiance [µW*(cm²srnm)-1 ] RADIANCE.HDR associated header file with the image data file *.log Record of various messages during data acquisition. *.cal On-board calibration of internal lamp image file. Recorded in DN. This file is used by HyVista for radiometric pre-processing. Delivered with associated header file. *.drk Dark current image data file. 128 channels by 10 pixels. This file is used to examine the dark current statistics. Delivered with associated header file. *.jpg Quicklook image file. *.out Multi-column ASCII file derived form the *.log file. It contains the geo- parameters data: UTC time, navigation data, IMU data, GPS data. It is used for geocoding the image data. Aerial photographs Description BW negatives Scale: 1:13000; 68 analogue prints with 60% overlap long track and 30% across track. RMK mission Flight logs, photo centre map, annotation data format, calibration record. documents Table 3.3: List of delivered items acquired during the hyperspectral data acquisition survey 2000.
Deutsche Steinkohle AG Division for Engineering Survey 13 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
The weather conditions during the overflight were good, no clouds with some haze. For further details please refer to the MINEO HYPERSPECTRAL AIRBORNE SURVEY.
3.2 Quality Control The quality of a hyperspectral data-set is closely linked with the noise content of the data. ROTHFUSS (1994) pointed out, that the standard deviation in a homogeneous box is a dimension for the noise of a sensor and that noisy channels should be eliminated from the data-set.
The signal is the part of the image which has no noise. All interfering signals are called noise. Noise is caused by several reasons, mainly by reasons of the scanning mechanism and can be of various shapes. Noise falsifies the signal significantly and therefore complicates information extraction. A simple statistical model, the additive noise model, can be described for each pixel as: DN p int[a p n p ].
To know about the relative amounts of signal to noise ratios is necessary to assess noise reduction and information extraction methods. A device for the Signal to Noise Ratio is the SNR. It is a number with no dimension and hence independent from the data dimension. A disadvantage of the SNR is that it can not be appointed directly but has to be estimated from the data. The estimation for an image with random noise can be done by calculating the contrast ratio between a bright and a homogeneous image part:
C SNR Signal Cnoise
The image parts can be selected manually out of the image. For MINEO an operational method for estimating Signal to Noise Ratios has been developed. This method searches the brightest and most homogeneous image part automatically using a floating window (GAO, 1993). The ratios of each image channel can be seen in the figure below.
Figure 3.2: Signal to noise ratio for each HyMap channel.
Deutsche Steinkohle AG Division for Engineering Survey 14 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
According to the results of this algorithm all HyMap channels with a low Signal to noise ratios (below 200) have been eliminated from the data-set from further processing. For the German HyMap Survey from 2000 10 channels had to be eliminated due to high noise and negative values in the data set. These channels were the numbers 0, 30, 62, 63, 64, 65, 67, 93, 94, 124 and 125. This are mostly channels at the transition between two detectors devices of the HyMap sensor. Additionally some more channels in the wavelength region around 940 nm had to be separated in the data from year 1998 due to high water-vapour absorption in the atmosphere , what leads to implausible reflectance values in the data-set (BRUNN, ET AL., 2001).
Deutsche Steinkohle AG Division for Engineering Survey 15 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
4 OTHER AVAILABLE RELEVANT ENVIRONMENTAL DATA The active mining projects in Germany are not only obligated by law to be well documented and monitored but also an extensive EIA is required for every area where active mining takes place. The EIA rests upon an extensive environmental database. The environmental information are usually held in digital form on DSK’s Geographic Information System (GIS) and is regularly updated at frequent intervals. Because of the nature of the problem (subsidence monitoring) and the resulting environmental impacts (groundwater-logging, vegetation stress and alteration of biotopes) the analysis of the HyMap data is closely associated with GIS-supported analyses and modelling. Within the framework of the environmental impact assessments a wide variety of inventory data and project information on the test site is required for the EIA, the most important are:
- topography - digital ground water model (DGWM) - geology - planned extraction areas and derived ground movements - soil - actual land use - hydrology - cultural treasures and tangible assets - flowing water and stagnant water - biotope types - animal habitats - digital terrain model (DTM).
The investigation of vegetation vitality and the detection of changes using the HyMap data is supported by a link-up with these GIS data sets. The GIS data sets were also used for validation of the HyMap data interpretation and analysis results. More detailed description of the relevant GIS-based environmental data is given in chapter 8.
In addition to the data mentioned above DSK has access to a diversity of remote sensing data, which are used in the MINEO project too. Table 4.1 presents an overview of these data records. These data are used for different purposes (e.g. derivation of DTM, change detection) to fulfil the tasks laid down for the Central European MINEO test site.
Data Date recorded Description Scale /accuracy aerial photograph 1993 RMK camera; RGB 1:6000; 60% overlap DPA (ms) 1997 airborne scan GIFOV 0,7 m HyMap 1998 airborne scan GIFOV 7 m aerial photograph 2001 RMK camera; RGB 1:4000; 60% overlap Table 4.1: Selection of remote sensing data records held by DSK.
Deutsche Steinkohle AG Division for Engineering Survey 16 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
5 FIELD SPECTRORADIOMETRY CAMPAIGN Target Target description Numbers Recording of samples instrument P-01 white fine sand 17 FieldSpec Pro P-02 concrete 12 FieldSpec Pro P-03a corn crop 20 FieldSpec Pro P-03b corn crop 5 FieldSpec Pro P-04 corn crop 27 FieldSpec Pro P-05 dry grass 26 FieldSpec Pro P-08 tree nursery 24 FieldSpec Pro (deciduous trees) P-09 potatoes 24 FieldSpec Pro P-10 bare soil 11 FieldSpec Pro P-12a tree nursery 20 FieldSpec Pro (deciduous trees) P-12b tree nursery 31 FieldSpec Pro (deciduous trees) P-16 dry grass 18 FieldSpec Pro P-17 meadow 12 FieldSpec Pro P-18 bare soil 14 FieldSpec Pro P-19a corn crop 20 FieldSpec Pro P-19b corn crop 19 FieldSpec Pro P-20 potatoes 24 FieldSpec Pro P-26 corn crop 18 FieldSpec Pro P-117 bare soil 13 FieldSpec Pro P-118 bare soil 22 FieldSpec Pro BGR-a white fine sand 12 Mark V BGR-b clay 3 Mark V BGR-c medium sand; Fe 3 Mark V enriched (A)
(B) Figure 5.1: Geocoded subset of HyMap image (r:0,890 µm; g:0,662 µm; b:0,507 µm) with ground reference measurements (A). Examples of recorded ground reference spectra (B); the shaded vertical bars indicate regions of strong atmospheric water vapour and CO2 absorption.
On August 24th 2000 the Central Europe environmental test site Kirchheller Heide was flown by an aircraft equipped with the HyMap scanner. At the same time a field campaign took place on the ground. The main purpose of this field campaign was the collection of spectral reference
Deutsche Steinkohle AG Division for Engineering Survey 17 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
measurements of selected targets which display different materials and cover types and can be used for the radiometric correction and inflight calibration of the HyMap sensor. The simultaneous recording of the reference spectra ensures the same illumination and atmosphere conditions for the two independent recording systems, the HyMap sensor and the field spectrometer. Selecting the reference targets focus has been given on homogeneous areas with a suitable size aiming in pixels with as small variations as possible over this areas. Figure 5.1(A) shows the targets recorded during the field campaign.
The spectroradiometry field campaign was done in close co-operation with German Federal Institute for Geosciences and Natural Resources (BGR), Geo-Research-Centre Potsdam (GFZ-Potsdam) and the Technical University of Clausthal-Zellerfeld, Institute for Geotechnical Engineering and Mine Surveying (TUC-IGMC). TUC-IGMC gave scientific support and guidance in the target selection.
BGR, like for all other MINEO test sites, was assigned to take spectral measurements on reference targets. BGR took these measurements at three different sites, namely: - a sand pit with white fine to medium-grained sand (Pleistocene drift sand) - a clay pit (Ratinger clay of the Tertiary Series) - a sand and gravel pit (sediments from the main Rhine terrace). These measurements has been done using the IRIS GER Mark V spectrometer. A brief description of this device is given in chapter 5.1.
GFZ-Potsdam used an ASD FieldSpec Pro FR spectrometer for its measurement programme. A brief description of this device is also given in chapter 5.1. The employment of a second spectrometer for the field campaign made it possible to take spectral references over several different sites during the aeroplane overflight and thus enabled us to get more reference data representing the spectral diversity of the test site. The targets chosen for this handheld spectrometer consists mainly of agricultural land, such as pasture, fields of maize, potatoes and also bare soil. Some additional measurements were also taken in reforestation areas containing two recently-planted stocks of trees (tree nurseries). In total 18 different sites have been measured with the ASD FieldSpec Pro FR spectrometer. Figure 5.1(B) shows some examples of the recorded reference spectra during the field campaign.
5.1 Brief description of the spectroradiometer used As mentioned above during the field campaign two spectrometer devices have been used to record the reference spectra of different materials and land cover types. The following table gives an overview of the technical characteristics of these devices.
Name FieldSpec Pro FR GER Mark V IRIS Spectral Range 350 – 2500 nm 300 – 3000 nm Spectral Resolution 350-700 nm: 3 nm 2 nm, 4 nm, 6 nm; spectral resolution 700-1400 nm: 10 nm selectable 1400-2500 nm: 10 nm Channels 296 depending on selected spectral resolution Scanning Time 100 msec 10 sec Detectors 1 Si photodiode (350-1000 nm) 2 Si photodiodes 2 InGaAs photodiodes (1000-2500 nm) 2 PbS photodiodes Foreoptics / Input 8°, 25° FOV foreoptics optional Weight total: 7,2 kg total: 20 kg Table 5.1: Technical characteristics of employed spectrometer devices during the field campaign (source: ASD Inc.: http://www.asdi.com; GER: http://www.ger.com GER Mark V IRIS users guide, respectively).
Deutsche Steinkohle AG Division for Engineering Survey 18 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
5.2 Description of main spectral features representing vegetation and vegetation stress. Since DSK’s objective in this project is to analyse the influences of water-logging occurrences to the environment, the efforts have been focused on vegetation vitality analysis, vegetation stress detection, and description of the biotope type alteration process.
Spectra of vegetation appear generally similar, because all plants are made of the same basic components. This fact requires adopting a different strategy for plants than the direct spectral feature and band shape matching used for geology applications. The spectral shape of plant leaves is controlled by absorption features of specific molecules and the cellular structure (ref. Figure 5.2). Foliar absorption is primarily caused by photosynthetic pigments in the visible spectrum (VIS) (400- 700 nm). The main pigments controlling the absorption are chlorophyll a and b and -carotene. The amount of absorbed radiation is dependent on the pigment concentration. The higher the concentration, the higher the green peak (around 550 nm) and lower the reflectance in the red range (670 nm). The absorption in the blue (400-500 nm) and red (600-700 nm) light range is directly proportional to the photosynthetic activity (PA) (STRASSBURGER, 1991). The light in the near-infrared (NIR) range (700-1300 nm) is not absorbed by the photosynthetic pigments which results in a sharp rise of the reflectance spectrum curve. This inflection point, so-called “red edge-wavelength”, occurs at the transition between the VIS and NIR wavelengths. The red edge position occurs typically between 690 and 740 nm, it is predominantly determined by the interaction between chlorophyll absorption of the red light and the internal scattering process on the leaf (CURRAN et al, 1990). The shape of the reflectance spectrum in the mid-infrared range (MIR) (1300-2500 nm) is controlled by leaf-water-content, lignin, cellulose and other carbon-based compounds.
According to FILELLA & PEÑUELAS (1994) chlorophyll concentration is usually an indicator of photosynthetic capacity, developmental stage and consequently an indicator for vegetation stress. Remote sensing of chlorophyll is therefore expected to be a valuable tool in the evaluation of plant status both in agricultural and natural plant communities. One of the best remote sensing descriptors of chlorophyll concentration has been found to be the red edge. The wavelength position of the red edge is dependent on several factors, but the most important are the chlorophyll concentration and leaf water status. Shifts in the red edge position indicate an increase or decrease in chlorophyll content. The shift towards shorter wavelengths (blue-shift) means a decrease of chlorophyll content and can be interpreted as an increase of vegetation stress.
Figure 5.2: Factors influencing the spectral curve of green leaves (BACH, 1995).
Deutsche Steinkohle AG Division for Engineering Survey 19 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
In literature different methods are described which seem to be suitable for the estimation of plant stress in spectrometer data, but most of them were tested and evaluated only with high resolution field- or laboratory-spectra. DSK with its subcontractor Technical University of Clausthal, tested these methods with airborne imaging spectrometer data (HyMap). The most promising methods were the parameterisation of the red edge position (e.g. BACH, 1995) and the geometric analysis of spectra features (e.g. KRUSE, 1995). A detailed description of the algorithms used in vegetation and vegetation stress mapping will be given in chapter 7.
5.3 Endmember selection The image derived endmember (EM) definition has been done for forest stands, since only perennial plants, like trees, show stress effects. In accordance to the mapping key for the biotope maps the EM were defined for tree species dominant in this test site and their corresponding stage of life (ref. section 8.4). The stages of life are grouped in three classes: - pole forest, stem diameters at breast height (DBH) < 15cm, - DBH 15 – 50 cm, - DBH > 50 cm.
Suitable endmembers are a fundamental assumption for a proper analysis of imaging spectrometer data. In the following a multilevel method for the selection of endmembers from hyperspectral image data which combines GIS and image-processing methods is described.
The first step of the endmember selection is the GIS analysis of the ground truth data. For each forest class one separate layer was produced which was clipped by 30 m around its particular border to avoid mixed pixels caused by misregistration of the image data.
These GIS layers were used to produce ENVI Regions of Interest (ROI’s) for each forest class to find the minimum, maximum and the three middle spectra. We did not use the ENVI methods, which calculate synthetic encasing spectra of the whole data using minima, maxima and mean from each separate band but developed a method which enabled us to select real spectra from the maximum, minimum and mean of the range of dispersion of the spectra.
Figure 5.3: Image spectra derived for a single forest class (left) and statistically selected min., mean and max. spectra.
Second for each forest class a Minimum Noise Fraction (MNF) transformation was calculated. It is necessary to calculate this transformation for each region of interest separately instead of using one transformation with the according mask, because the result of the MNF and the image statistic is dependent on each pixel contained in the image. Using the MNF statistic files the spectra calculated
Deutsche Steinkohle AG Division for Engineering Survey 20 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
for each forest class were transformed in the MNF space, too. This step was necessary to calculate the Mixture Tuned Matched Filtering (RSI, 1997). The result of this operation is a greyscale image for each inserted spectrum which represents the relative portion of this spectrum from the whole pixel (score) and a second greyscale image which represents the reliability of this procedure (infeasibility).
The last step is the selection of the purest pixels using the 2D-scatter plot. In this plot for each spectrum the score and infeasibility values were displayed and the “purest” pixel selected. These purest pixels were defined as endmembers and stored as an ENVI spectral library.
The image on the left (Fig. 5.4) shows a 2D scatter plot for the Spruce trees in class DBH < 15cm. As endmembers the points with the highest score and the lowest infeasibility (marked in red) were selected.
Figure 5.4: 2D scatter plot of Spruce trees.
The plot below (Fig. 5.5) shows the representation of the endmember spectra which were selected for different kinds and specifications of pines (pinus sylvestris).
Figure 5.5: Endmembers plot for different kinds of Pine trees (Pinus Sylvestris). Code description: ak0= Pinus Sylvestris, alt4= life age (here: DBH 15-50 cm), II_c2= subclasses.
Deutsche Steinkohle AG Division for Engineering Survey 21 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Pine forests will serve as an example for the detection of stress in the further work, because they are present in a suitable amount in each part of the test site and because they are sensitive concerning water availability in their living space.
5.4 Feeding MSL, spectra categories included in MSL The MINEO Spectral Library (MSL) represents one of the major outputs of the MINEO Project. It is a database system for the management of spectra together with their environmental information. It is designed to handle spectra coming either from laboratory analysis, field spectroradiometry or extracted from hyperspectral images. The main purpose of MSL is the support of the hyperspectral image analysis and consequently assists the mapping of mining-related contamination and environmental impacts.
Since DSK produces coal from ca. 800 m below the surface the occurring effects at surface are not related to toxic materials like it might be expected for open cast mines. The environmental impacts caused by deep coal mining in this coalfield have a more indirect nature and are related to surface subsidence movements and the combined relative uplift of ground water level and water-logging occurrences. These alterations may lead to vegetation stress and undesirable changes in the plant communities. Having this as background, DSK’s contribution to the MSL is two folded. On the one hand the reference spectra from the field spectrometry campaign have been incorporated into the MSL, and on the other hand image-derived endmember for specific plant species were defined. A special attention have been given to plant communities which show high sensitivity to the changed environmental conditions.
To ensure that every spectrum stored in the MSL has a unique name, the prefix ‘m3’ has been defined for the Kirchheller Heide test site. The rest of the spectra name relates to the acquisition point in the field. All spectra are stored in ASCII format.
The reference spectra recorded during the field campaign are stored in the MSL in their original form and should reflect the real conditions in the test site. For each spectrum the related environmental information like geoposition of measurement point, type of material recorded, date of acquisition, cloud cover, instrument used and processing type are stored. Using these indicative specifications, the spectra can be grouped in three major classes: - sediment spectra (clay and sand), - soil spectra (bare soil on agriculture fields, harvested fields with plant remnants and high organic matter portion), - vegetation spectra (corn field and pasture land, tree nurseries).
Figure 5.1 shows the location of the reference spectra recording points as well as some examples of these spectra. Selected reference spectra stored in the MSL were used during the atmospheric correction for inflight calibration of the sensor, please refer to section 6.2: ‘Atmospheric correction’.
As already stated in section 5.3 the image derived endmember (EM) definition has been done for forest stands for the dominant tree species and their corresponding stage of life. The tree stage of life definition corresponds to the definition used in the GIS database for biotope maps, please refer to section 8.4: ‘Biotope types and actual land use’. These EM were used to compute the spectral parameters which describe the vegetation stress. Please refer to section 7.4: ‘Stress detection’.
Deutsche Steinkohle AG Division for Engineering Survey 22 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
6 DATA PRE-PROCESSING Previous to the thematic analysis of the hyperspectral imagery a pre-processing stage is required in order to bring these data into an appropriate shape for further information extraction working steps. This pre-processing stage consists primarily of: geometric corrections, atmospheric corrections and mosaicing. The pre-processing is necessary on the one hand for integration of the data into the GIS data base for further processing and analysis steps with other related data sets (geo-correction) and on the other hand for making the hyperspectral images from different dates comparable to each other, i. e. all atmospherically spurious effects like haze, water vapour, various aerosols, which vary from date to date, are attempted to be removed (atmospheric correction). The geocoded and atmospheric corrected data are then mosaiced and are ready for GIS integration and further analysis. The established workflow using the PARGE procedure, which applies the parametric approach, and the physically- based model ATCOR-4 (RICHTER & SCHLÄPFER, 2000) is delineated in Figure 6.1. Detailed descriptions of these two pre-processing steps are given below. The corresponding MINEO generic procedure ‘PRE_PROCESS’ can be found in Appendix IVa.
Input:
hyperspectral image (radiance data)
geocoding geocoded hyperspectral DTM (PARGE) image data
IMU-data & GCP-points atmospherical data: • date of image acquisition, atmospheric correction geocoded & atmosphere • atmosphere profile (ATCOR-4) corrected hyperspectral (e.g. from radio tube reflectance image data ascent), • visibility
mosaicing spectral reference measurements from the ground geocoded, corrected & mosaiced image data set
Figure 6.1: Pre-processing workflow established for geocoding and atmospheric correction of HyMap data using PARGE & ATCOR-4.
6.1 Geometric corrections Airborne image data acquisition in contrast to spaceborne data is characterised with lower system stability. This may result in geometric distortions due to variations in flight path and attitude of the plane. The distortions cannot be corrected with the traditional georeferencing procedures, since the plane movements, given by roll, pitch, yaw and heading angles, cannot be described satisfactorily by polynomial transformations of the image. The parametric approach is a favourable solution for high resolution airborne remote sensing data. It recalculates the flight scan geometry pixelwise applying
Deutsche Steinkohle AG Division for Engineering Survey 23 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
physically measured auxiliary data. The three major types of auxiliary data are: image/scanner general information, navigation data (IMU and GPS) and a digital terrain model for the equalisation. The applied procedure PARGE (PARametric GEocoding) is developed by ReSE Applications Schlaepfer and the Remote Sensing Laboratories (RSL) of the University of Zurich (SCHLAEPFER, 2001). PARGE is an IDL based software package and thus able to read and write the ENVI image data format. The accuracy of the georeferencing procedure applying PARGE reaches subpixel accuracy theoretically. The geocoding results are of reliable accuracies of down to 2 pixels (SCHLAEPFER, 2001). But the quality of the geocoding results are highly dependent on the input data, especially the auxiliary navigation data. Detailed information about the PARGE procedure and the applied algorithms can be found in the PARGE User Guide or in the web (http://www.rese.ch).
The geometric correction of the HyMap images for the Central European test site in Germany Kirchheller Heide was performed using the PARGE version 1.2. Besides the image data the following auxiliary input data sets were required: - IMU data - DTM - Ground Control Points (GCP). The IMU data (internal monitoring unit data) contains all necessary information about the sensor position during the flight, i.e. the roll, pitch and heading angles, the GPS coordinates and the UTC time for every scanned image line. These data have been delivered by HyVista in raw format (*.log file) and as a reformatted ASCII file (*.out file). The first step of the geocoding process was the co- ordinate transformation, since all additional environmental data relevant to this test site are stored in the Gauß-Krueger coordinate system. The recorded geographical coordinates have been read from the *.out file and transformed into the Gauß-Krueger coordinate system and written back into the *.out file. The input DTM for the geocoding procedure is a high quality DTM derived photogrammetrically which serves as a basic data set for the EIA in the Kirchheller Heide test site, too. The pixel size has been resampled to 4x4 m. To ensure the high georeferencing accuracy, the DTM should be of higher resolution than the image data. The resolution of the HyMap images is 5x5 m, so the resolution of 4x4 m for the DTM has been stated as capable.
The geocoding procedure was performed separately for each image strip. For each strip GCP’s have been set. The amount of the GCP’s is individual for the strips, since it highly depends on the imaged surface. Characteristic points such as road and way crosses have been used to set a GCP. To ensure good geocoding accuracy, it was attempted to set the GCP well of nadir (i.e. as far as possible from the central axis of the image strip) and spread them equally over the image. Although the algorithm used in PARGE theoretically needs only one GCP to estimate the roll/pitch or x/y values and perform the georeferencing, in practice 1 GCP per 100 image lines is needed to ensure good geocoding accuracy. In addition it should be noted that the source GCP are extracted from should be of higher resolution than the image data to be geocoded. The coordinates for the GCP were taken from a digital copy of the German Base Map (DGK5) with a scale of 1:5000 with the accuracy of M1 m. The accuracy assessment of the geocoding procedure has been carried out computing the RMS error. The RMS error has been calculated using independent GCP’s. The calculated RMS error is 1,5-2 pixel varying between the strips. Transformed into meters, the RMS error does not exceed the displacement of 10 m. For the available data set the achieved accuracy can be stated as satisfactorily as shown in Figure 6.2.
Deutsche Steinkohle AG Division for Engineering Survey 24 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 6.2: Example of geocoded HyMap image combined with GIS vector data sets.
6.2 Atmospheric corrections The task of earth observation is the mapping of surface properties. However varying atmospheric conditions, differences in the sun geometry and topographic effects strongly influence the recorded signal. Figure 6.3 describes different scattering mechanisms. Theses influences modify the true spectral behaviour of the ground features. The objective of atmospheric correction is the elimination of atmospheric and illumination effects and the data conversion from radiance to reflectance. This is necessary to retrieve physical parameters from the surface. Especially for quantitative analysis and change detection applications with images from different periods and sensors an accurate atmospheric correction is an essential part of pre-processing and a prerequisite for the derivation of products needed for the subsequent image processing and analysis steps.
Figure 6.3: Schematic sketch of radiation components: 1) path radiance; 2) reflected radiation from the viewed pixel; 3) adjacency radiation; 4) terrain radiation reflected to the pixel.
For atmospheric correction several different approaches are used which generally can be divided in two main groups, these are the image statistics based approaches and physically-based models which consider the local atmospheric conditions. The image statistics based approaches calculate the
Deutsche Steinkohle AG Division for Engineering Survey 25 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
reflectance either from the image directly (e.g. the Flat Field Calibration, Dark-Object-Subtraction model (DOS)) or from ground reflectance spectra using statistical methods (e.g. the Empirical Line Calibration). These models ignore the true atmospheric conditions during the image acquisition and some additional effects like illumination direction and topography of the represented surface. The second group, physically-based models, considers the atmospheric conditions using radiative transfer codes such as MODTRAN4 (BERK at al 2000) or 6S. In ATCOR-4, which is an extension of the satellite version of ATCOR (ATCOR-2: flat terrain; ATCOR-3: rugged terrain) for wide field-of-view airborne optical sensors in addition the terrain shape obtained from a digital terrain model is taken into account. ATCOR-4 performs the atmospheric and topographic correction accounting for the angular dependence of the atmospheric correction functions (RICHTER, 2000).
The choice of ATCOR-4 to accomplish the atmospheric correction for the German test site was driven by several aspects listed below: - the physically-based approach of ATCOR-4, using the MODTRAN4 radiative transfer code, enables the modelling of the real atmosphere conditions during the overflight incorporating own atmosphere data and thus increase the accuracy of the procedure; - enhanced capability handling of hyperspectral imagery; - implemented characteristics of HyMap sensor; - IDL based with very good interface between ENVI and PARGE data formats.
ATCOR-4 has been developed by Dr. R. Richter at German Aerospace Center DLR. Detailed information about the ATCOR-4 procedure and the applied algorithms can be found in the ATCOR-4 User Guide or in the web (http://www.rese.ch). ATCOR-4 has a modular structure (ref. Table 6.1). The architecture of these modules follows roughly the recommended workflow.
Module Description SUNNY1 calculate solar zenith and azimuth angle RESOLAR resample the solar irradiance ATLUT calculate atmospheric LUT RESLUT resample atmospheric LUT to sensor wavelength table IFCALI in-flight calibration and consideration of ground spectra SHADOW & SKYVIEW calculate the self shadowing and cast shadow for DTM or the sky view factor respectively SPECL spectral emissivity classification BBCAL blackbody calculation of temperature/radiance ATCOR-4 atmospheric & topographic correction of image cube Table 6.1: Modular structure of ATCOR-4.
The atmospheric correction of the HyMap data was done in close co-operation with DLR on geocoded and orthorectified image data (BRUNN, DITTMANN et al., 2001). The geocoding was done applying the parametric geocoding approach in PARGE as described above. Before the atmospheric correction with ATCOR-4 can be performed some preparation of input data are required. This includes the resampling of the field reference spectra (refer to chapter 5) to the sensor-specific spectral resolution and the calculation of atmospheric look-up tables (LUT) based upon the aerological data from the radiosonde. The use of radiosonde profiles is preferred to the standard atmospheric models if such data is available. In our case the data of a radiosonde ascent were used. The radiosonde ascent took place during the flight time and was carried out by the German National Meteorological Service (Deutscher Wetterdienst, DWD) office in Essen, which is about 20 km away from the test site. The generation of the LUT’s was performed with the ATLUT module of ATCOR-4. ATLUT calculates scan angle dependent atmospheric look-up tables using the radiative transfer code MODTRAN4. ATLUT uses
Deutsche Steinkohle AG Division for Engineering Survey 26 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
the MODTRAN4 atmospheric transfer code to calculate appropriate atmospheric lookup tables for the situation during the data acquisition (RICHTER, 2000). The output of ATLUT can be stored in a permanent spectral atmospheric database. RESLUT resamples the spectral LUT with the band-specific response function of the selected sensor, in this case HyMap.
INPUT IMAGE geocoded Radiance Image DEM (from PARGE) elevation scan & azimuth slope angles aspect (PARGE) (SHADOW, Reflectance Image SKYVIEW) (1st. stage) ATM Database Inflight Calibration (ATLUT, Ground Reference (IFCALI) RESLUT,) Spectra Adjacency Correction
OUTPUT IMAGE geocoded Reflectance Image (final stage)
ATCOR4
Figure 6.4: Schematic workflow of atmospheric correction with ATCOR-4
The atmospheric correction was performed after computing the atmospheric LUT and resampling the reference ground spectra. Based on the aerological data the LUT’s were calculated for the visibilities of 15, 25 and 50 km. Prior to the correction a quality check of atmospheric input data and the radiometric calibration have been done using the IFCALI module within ATCOR-4. The IFCALI module enables the comparison of the image retrieved reflectance with ground reference spectra. This comparison of the image retrieved reflectance spectra was done with the standard calibration file, provided by Integrated Spectronics Inc. It showed the best match in the spectral region below 1 µm for a visibility of 25 km. For each of the six HyMap strips LUT’s for a 25 km visibility were used subsequently. However, significant deviations up to 50% were present in the spectral regions beyond 1.5 µm. For that reason a recalibration for the HyMap 2000 data was necessary. The IFCALI module has two modes of operation: the “reflectance” and the “calibration” mode. If the “reflectance” mode is activated, an evaluation of the atmospheric input data and the radiometric calibration can be performed (see above). The “calibration” mode serves for the calculation of the calibration coefficients c0, c1 for each reflective band that relates the measured digital numbers (DN) to the at-sensor radiance. Applying the reference spectra a new calibration file can be computed. The inflight calibration was performed successively. First a calibration was done using the ground spectra BGR-a and BGR-b, which primarily were planned to serve as calibration reference fields. These targets are two flat open pits and the material consists of white fine-grained sand and clay. Problems existed because the spatial resolution of the DTM was not sufficient to follow the steep slopes of the terrain. Also, the homogeneity of these targets was not sufficient for the inflight calibration procedure. Therefore, the calibration was performed with single field targets over the most homogeneous fields P-03, P-04, P-05 and P-20 (refer to chapter 5, Figure 5.1). As a last test a regression analysis was calculated over the
Deutsche Steinkohle AG Division for Engineering Survey 27 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
fields P-03, P-04, P-05 and P-20. The best results have been achieved for the single target calibration at the field P-05. Besides the comparison with the reference spectra, a known reflectance, e.g. water surface, has been extracted from the image to prove the results.. The resulting reflectance had the best match with the reference data in the range of 0.4-0.7 µm. However image retrieved water surface reflectance still showed too high values within the NIR spectral range (approx. 4%). Therefore, a synthetic water reflectance spectrum with NIR reflectance values close to 1% was used for the improved recalibration. (Figure 6.5). Applying the synthetic water reflectance and the reference reflectance of field P-05 a new calibration file was generated. It was employed for the atmospheric correction of all six image strips.
The formula below describes the correction algorithm and the dependencies in the ATCOR-4 procedure.
2 d c0 c1DN x, y Lp z, v , 1 i x, y (6.1) * i v z, v b x, y Es s z cos x, y Edif x, y, z Eg z, r terrainVterrain x, y
The terms are defined as: 1 i x, y reflectance value at coordinate x, y, calculated iteratively x, y horizontal coordinate, corresponding to the georeferenced pixel coordinates z vertical coordinate, containing the elevation information from the DTM DN(x,y) digital number of georeferenced pixel Lp z, v , path radiance, dependent on elevation and view geometry v z, v ground to sensor view angle transmittance, direct and diffuse components s z sun to ground beam (direct) transmittance x, y angle between the solar ray and the surface normal (illumination angle) b binary factor: b = 1 if pixel receives direct solar beam, otherwise b = 0 Es extraterrestrial solar irradiance * Ed x, y, z diffuse solar flux on an inclined plane Eg z global flux (direct plus diffuse solar flux on a horizontal surface at elevation z) 0 terrain initial value of average terrain reflectance i terrain x, y locally varying average terrain reflectance, calculated iteratively Vterrain x, y terrain view factor (range 0-1)
HyMap data have been atmospherically corrected using ATCOR-4. The model is suitable to perform an accurate correction in combination with ground reference data. An essential part is the possibility to check spectra before processing the image cube and to calculate an updated calibration file using ground reference spectra. The accuracy of the atmospheric correction method described here depends on several factors: the calibration accuracy of the sensor, the accuracy of the radiative transfer code (MODTRAN4), the quality of the digital terrain model, and the image ortho-rectification relying on auxiliary information such as attitude and GPS/DGPS. The deviation of ground measured reflectance and retrieved reflectance does not exceeds 3% in our cases (ref. Figure 6.6). The retrieved values are within the error margin of ground measurements except for the few HyMap channels with noise problems and the regions of strong water vapour absorption.
Deutsche Steinkohle AG Division for Engineering Survey 28 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 6.5: Water surface reflectance (Lake) retrieved after the single target calibration of field P-05 with its ground-reflectance and the modified water surface reflectance (Lake(mod)) as used for the inflight calibration; the vertical shaded bar marks the region with the best fit after single reference target calibration over field P-05.
Figure 6.6: Comparison of ground-reflectance spectra with the corresponding HyMap reflectance spectra for different fields; the shaded vertical bars mark regions of strong atmospheric water vapour and CO2 absorption.
Deutsche Steinkohle AG Division for Engineering Survey 29 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
The geometric correction using the parametric approach have been successfully applied also at other MINEO test sites. Atmospheric correction using ATCOR-4 was tested, evaluated and compared with different methods of atmospheric correction also at other test sites, e.g. Erzberg test site in Austria. The successful application of these methods described above emphasises the generic nature of the data pre-processing stage and should always be considered in the future remote sensing data analyses for monitoring purposes.
Deutsche Steinkohle AG Division for Engineering Survey 30 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
7 DESCRIPTION OF IMAGE PROCESSING PROCEDURES AND ALGORITHMS USED IN IMPACT MAPPING 7.1 Objective Mining of natural resources leads usually to an impact on the environment. The type of this impact depends highly on the mined deposits and on the mining methods applied. In contrast to mining of ore deposits, where high concentrations of elements such as FE, Cu, Pb, S, AG, AU etc. are present and where toxic materials may remain after purification and refinement process, underground mining of coal does not cause a direct toxic pollution to the environment. The environmental influence caused by underground mining is more of indirect nature, therefore no conventional contamination maps of specific pollutant distribution can be produced. Rather maps of the reaction of vegetation to this changes will be generated. These maps contain the development and distribution of vegetation stress and the alteration of biotope types in time series. This kind of maps are intended to be used for environmental monitoring tasks and landscape development management. To meet the objectives defined for the Central Europe MINEO test site a multi part procedure has been realized. Firstly an over-all map of forest stands within the test site has been produced. This map in conjunction with GIS database was then used to assess areas with water-logging sensitive tree species. Secondly spectra feature parameters were tested for their ability for vegetation stress detection. Subsequently algorithms and computational tools were developed to calculate this parameters in areas of interest, susceptible to water-logging. In the next step change detection analyses of multi temporal data sets were performed in order to visualise and evaluate areas of occurred changes. To meet this tasks workflow procedures have been developed and attached in appendices IVb and IVc.
The following subsections describe the developed procedures, methods and tools which have been applied for vegetation mapping, stress and change detection.
7.2 Procedures, algorithms and toolboxes
7.2.1 Red edge Analysis
Figure 7.1: Typical vegetation reflectance spectrum highlighting the red edge between 0,68 and 0,80 µm.
Figure 7.1 shows a typical reflectance spectrum for vegetation between 0.42 and 2.48 µm. Between 680 and 800 nm, in the near infrared region, the spectrum rises steeply. This steep region is the edge of vegetation reflectance, where the chlorophyll pigment loses its ability to absorb energy (ref. section
Deutsche Steinkohle AG Division for Engineering Survey 31 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
5.2). HORLER, BARBER and BARRINGER 1980, showed, that this red edge shifts to shorter wavelength for stressed vegetation. SINGHROY ET AL. (1991) found out, that stress induced shifts in the red edge may occur in both directions, towards the longer and shorter wavelengths. The red edge wavelength is defined as the position of the reflectance curve, where the second derivative of the strong increase between red and infrared wavelength equals zero (ref.. Figure 7.2). All the features, that work out details of the wavelength regions between 0,570 and 0,750 µm are influenced as well by the concentration of chlorophyll as by the condition of the chloroplast membranes on which the chlorophyll is bound. The broadening of the chlorophyll absorption bands is attributed to disruption of these membranes. SINGHROY ET AL (1991) suggested that the direction of the red edge shift direction is dependent on whether the membrane disruption or the chlorophyll loss process predominates.
Using a field or laboratory spectroradiometer with a very high spectral resolution it has been shown, that shifts can be retrieved effectively using polynomial functions. Instead of the up to 500 channels, these instruments deliver for this wavelength, common airborne imaging spectrometers like HyMap deliver only 8 up to 12 channels in this wavelength region. This is the reason why methods for red edge parameterisation had to be tested and developed. For the determination of the red edge wavelength several different analytical methods can be used (BACH, 1995). A well known and documented method is the fitting of an inverted Gaussian Model to measured vegetation reflectance data (BONHAM, CARTER, 1988). A second approach is the parameterisation by linearization which is described in detail in GUYOT & BARET, (1988) and BACH (1995). The model, implemented for this work, uses a combined method of interpolation and linearization. First an interpolate of the run of the curve with cubic spline functions was performed (RSI, 2001). These functions can be derived twice to calculate the inflection point of the reflectance curve. Because the algorithm for spline interpolations only delivers accurate derivative values of all measured points (but not for interstices between these points) it is only possible to calculate the two ambient points. After that we interpolate linearly between these two points and calculate the zero-point of the 2nd derivative from this linear interpolation, which represents the red edge wavelength.
Figure 7.2: HyMap Spectrum (left) with its first (middle) and second (right) derivative.
An important advantage using this spline interpolation is that splines directly touch every base of the spectrum that should be interpolated and influence the whole run of the interpolated curve. This characteristic of spline interpolations makes it possible to consider even subtle differences in the run of the reflectance curve. In the context of this red edge analysis five spectral features are relevant, which are: S wavelength of the red edge position, S the minimum in red (reflectance), S the minimum in red (wavelength), S maximum reflectance near 800 nm,
Deutsche Steinkohle AG Division for Engineering Survey 32 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
S wavelength distance between the minimum in red and the red edge. The features “red edge”, “wavelength min. in red” and the “wavelength distance between the min. in red and red edge” are measured in wavelength units (mostly µm or nm) and are therefore independent from scaling differences and pixel brightness. This is most important when average pixel brightness changes due to the interaction with canopy shape, illumination direction and distance from nadir directions. RENCZ (1986) determined the red edge features using the Inverse Gauss Model and showed systematic variations in the shape and positions of the red edge. He notes, that the red edge differs between species, and that “blue shift” phenomena related to vegetation stress must be separated carefully from “blue shifts” resulting from species differences.
7.2.2 Features from continuum removed spectra Many studies have shown that stressed vegetation will respond with changes in spectral reflectance. Next to the changes in the red edge region described above these changes have been observed at the green reflectance peak near 0,57 µm, the chlorophyll absorption maximum near 0,68 µm and in the region of the infrared reflectance shoulder between 0,75 µm and 1.1 µm (SINGHROY ET AL, 1991). This approach was applied by FISCHER (2001) to detect plant anomalies on different forest stands under investigation in this area.
The geometric analysis of spectral features is primarily based on a geometric analysis of continuum removed spectra. Continuum removal is the ubiquitous first step of this feature extraction process. The continuum of a spectrum is a continuous, convex hull draped over the source spectrum at its highest points. Dividing the source spectrum by its continuum spectrum results in a continuum removed or normalised spectrum with values from 0.0 to 1.0 (ref. Figure 7.3). Absorption features, which commonly occur superimposed on a background slope in the source spectrum, are transformed into features with a uniform, flat background of 1.0 in the continuum removed spectrum. This allows each absorption feature in a spectrum to be mathematically analysed with respect to a consistent reference plane.
a) b) Figure 7.3: Continuum-removed example of a chlorophyll absorption in vegetation , (RENCZ, 1999) (a) and an example of the absorption band parameters: ‘band depth’, ‘band width’ and ’FWHM’ (b).
The feature extraction algorithm contains several spectral attributes. The most common are the band position of the maximum absorption depth, the band depth, the FWHM (full width at half maximum) and the symmetry factor (ref. Figure 7.3). The symmetry factor is defined as the sum of the reflectance values for feature channels to the right of the minimum divided by the sum of the reflectance value for feature channels on the left (ref. Figure 7.3b). The logarithm of this value is taken to linearize the values. Symmetrical bands have a symmetry value of zero. Bands that asymmetrical towards longer
Deutsche Steinkohle AG Division for Engineering Survey 33 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
wavelength have positive asymmetries. Using this definition, the 0.67 µm chlorophyll absorption band characteristically has negative asymmetry. Applied to HyMap data the maximum absorption depth (band depth) is correlated well to visible plant stress. This band depth decreases dramatically, when the plants show any damage caused by distress e.g. by water.
7.2.3 Software tool for the extraction of the features at continuum removed spectra For the determination of this features a tool has been developed (ref. Figure 7.4), which calculates the features mentioned above over arbitrary parts of the spectrum, which makes it possible to consider the leaf water and canopy structure parameters in the wavelength region between 1000 and 1350 nm. Also the analysis of reflectance maximum is implemented in this tool. The tool is written in C++, due to the complexity of the calculations on large image data. Using C++ makes this tool faster than executing it in IDL.
Figure 7.4 shows the processing windows for that tool. The upper window is the data-handling window which gives the possibility to select the image file, shows a preview window, makes the user select the wavelength region in which he wants to work and gives the option to decide whether a reflectance peak or absorption peak should be analysed. The tool can calculate the features on continuum removed spectra either for one spectrum separately (chosen out of the preview image) or for the whole image. If the whole image possibility is selected the program delivers a generic binary file (.bsq) and an ENVI Header file. The lower image of figure 7.4 shows the spectral analysis window which gives the possibilities to select the feature to calculate and represents the according spectrum. Because the program is realized with C++ it only runs on Windows NT, 2000 and XP platforms. For Windows 95 derivatives an older Version is available which shows the same functions, but is restricted to file sizes below 2 GB. The tool is especially adapted to the needs of the German Test Site (according to data format and spectrum shape) and is available in German language only. Tests with spectra from the Greenland dataset have shown, that extensive work for data adaptation has to be carried out to apply the algorithms for this data. Additionally the Software delivers plausible values for vegetated areas only, the results for sparse vegetated areas are highly dependent on mixture ratios of vegetation and soil/rock spectra and not very reliable for plant status estimation.
Figure 7.4: Tool for the determination of the features from continuum removed spectra.
7.2.4 Derivative Analysis The derivative Analysis originally developed for spectrometry (TALSKY, 1994) it is particularly promising for the use with hyperspectral remote sensing data. Derivatives not only emphasise subtle
Deutsche Steinkohle AG Division for Engineering Survey 34 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
spectral details as well as changes but also suppress the mean level. They even minimise illumination and atmospheric effects (SCHOWENGERDT, 1997, PHILPOT, 1991). Although lot of advantages the derivative analysis needs to be applied with great attention, because it is very sensitive to data noise. Therefore it may be necessary to perform smoothing filters to minimise the random noise in the image data.
One of the main differences between remote sensing imaging spectroscopy and the conventional spectroscopy is that imaging spectroscopy delivers data in a subsequent manner over a widespread wavelength region, but not real continuous data. Only few researchers have tried to employ approaches commonly used in spectroscopy and have manipulated their data as truly spectrally continuous data.
Not all methods used in spectroscopy can be directly adopted to remote sensing analysis because there are significant differences between these two types of data. Data used in spectroscopy are usually collected under controlled laboratory conditions with full control over the intensity and spectral distribution of the illumination as well as viewing geometry. In contrast to that remote sensing has the lack of a defined preparation of the target substance, reflection and absorption standards etc. but the most important deficiency of remote sensing data is that the spatial resolution for a single pixel is greater than several meters in diameter (often several tens of meters), and it is rare for a single object or target feature to fill any one pixel. Thus, the characteristics of any pixel can rarely be considered truly homogenous.
As mentioned above the derivative analysis may be very sensitive to noise in the spectrum. Although HyMap is a sensor with very little noise portion it is necessary to smooth the spectra prior to the further analysis. Perhaps the most common method for smoothing is the Savitzky Golay filter (SAVITZKY and GOLAY, 1964), which provides a simplified least square procedure for simultaneously smoothing and differentiating of data. The method implicitly assumes that random noise has similar characteristics throughout the spectrum and that it can be handled by an invariant procedure over the spectrum. Some scientists argued justly that, since random noise usually varies over the spectrum, polynomial curve fitting might alter the signal waveform instead of eliminating the noise. Another commonly used smoothing algorithm is the median filter. This filter smoothes data locally in a predefined smoothing window but without any curve fitting. Instead, a median filter uses the mean value of samples within the local smoothing window as the new value of the middle sampling point in the smoothing window. During the image data analysis for the MINEO project the straight forward and fast processing median filter with a window size of 5 was used. The Figure 7.5 shows the results of the median filter using different window sizes.
Because we didn’t use the Savitzky Golay Algorithm which calculates the derivative in one step with the smoothing it was necessary to implement a different method for the derivative calculation. For this work a 3-point, Lagrangian interpolation was used, which is implemented in the IDL programming language (RSI, 2001).
To extract the features applying the derivative analysis to the HyMap imagery the derivative curves were integrated and the areas calculated. For the first group of features the integration was done over an interesting part of the spectrum. For the determination of plant features the well known wavelength around the red edge between 540 and 740 nm was considered and the areas of the first 5 derivatives of the spectrum were calculated. For the second group of features the integration over the whole spectrum range was calculated. The figure 7.6 shows the expanse of the areas concluded by the function curves of the first derivative of a spectrum as an example (left: whole spectrum, right: 540 – 740 nm).
Deutsche Steinkohle AG Division for Engineering Survey 35 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 7.5: Original spectrum and median filter results using different window sizes.
Figure 7.6: Area of the first derivative over the whole spectral range (left) and Area of the first derivative in the wavelength range between 540 and 780 nm, which is an interesting range for plant analysis (right).
7.2.5 Software tool for the extraction of the red edge and derivative analysis features A toolbox has been developed, which calculates these values (together with the derivative spectra and the areas described in chapter 7.2.3). The input spectra for this toolbox may either be a separate spectrum read out directly from the image or from an ASCII file or these values can be computed for a whole image. Figure 7.7 shows the GUI of this developed toolbox for computing the spectrum feature parameters associated with the red edge.
Deutsche Steinkohle AG Division for Engineering Survey 36 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 7.7: IDL toolbox to calculate the spectrum feature parameters associated with the red edge.
Figure 7.7 gives an example of the application of this software tool. The upper left image shows the data handling window of the program. Here the user can define the image or an image-part to analyse. Then a preview- and a detailed Zoom image of the data-set are loaded. Besides this the user can define spectra to analyse in detail. The lower image shows the detailed analysis window for the selected spectrum. Here the user can define the area which should be integrated for the derivative analysis, a first guess for the definition of the red edge can be fined and the values for all these features are displayed. Generally this window can be used to get a first idea of the spectral features of an area and to define the pre-knowledge values that have to be provided for the detection of the inflection point. The upper right figure shows the selection window for the calculation of the spectra over the whole (or parts of the) image dataset. The user can define the wavelength for which the area should be calculated, the degree of the derivative that should be calculated and the user has to define the filter
Deutsche Steinkohle AG Division for Engineering Survey 37 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
size, which should be applied for the derivative analysis. For the red-shift analysis a first guess that may change for different data characteristics (e.g. different for the German and Greenland dataset, where the procedure was tested on) has to be provided. This Analysis tool was realised using IDL by our subcontractor IGMC at the Technical University of Clausthal and runs on ENVI-Versions newer than 3.5.
7.3 Land cover & use: forest stands mapping As already stated in previous sections no conventional contamination maps were produced, but maps showing the vegetation cover in the test site. As a first mapping step a forest stands map was produced, which contains the information of tree species distribution (Appendix I). This map in conjunction with GIS database was then used to assess areas with water-logging sensitive tree species within the forest stands. For specific areas of interest, like zones of increased risk of water-logging, more exact version of this map was generated (Appendix Ia and Ib). This version includes in addition the tree life stages classes, since the life stage is closely linked to the sensitivity to changed environmental conditions in these sites.
The test site Kirchheller Heide is characterised by fast changes in the landscape. Most of them can be attributed to human activity, e.g. changes of the areas under cultivation and annual field crops, movements and changes in dimensions of the gravel and sand opencast pits as well as rededication of land for settlement areas. Analysing remote sensing data these changes often superimpose the mining caused subsidence impacts to the environment. The environmental influence of the mining induced subsidence does not appear rapidly, but it is rather a slow and continuous process with duration of more than one year. Therefore it is essential to focus on areas with as little as possible alteration in the land use, land cover and environment inventory over longer time periods. As such, the forest stands have been chosen. The forest stands suit very well to fulfil the tasks, since a large part of the test site is covered by forests (ca. 40%) and zones with high water-logging risk are situated within the forest stands.
The map of tree species distribution within the forest stands is based on analysis of the imagery acquired during the MINEO flight campaign in summer 2000. The analysis have been performed on a forest mask, i.e. areas outside the forest have been excluded, “masked out”, during the analysis. To achieve a map of good quality several mapping and classification methods have been tested. Firstly primacy has been given to the hyperspectral mapping methods like Spectral Angle Mapper (SAM), Linear Spectral Fit (LS-Fit) and others. These mapping algorithms use the endmembers (EM) as input spectra. These spectra are assumed to be 100% pure and form the classes of the image. Then they try to match spectra found in the image to the predefined classes. Each of these hyperspectral methods is based on different approach, but all of them need the EM as an input. The EM’s extracted from the image for the different tree species show too low spectral diversity between the tree species classes and too high variance within one class. Therefore these algorithms did not lead to the desired classification results.
Secondly the maximum likelihood classification was applied. As input data the first six principal components (PC) were used. Since the PCA was calculated exclusively on the forest mask, the data dimensionality became reduced significantly. Only the first six PC’s contain target information, the following PC’s are dominated by noise in the data. The information content can be estimated from the eigenvalues calculated during the principal components transformation.
Thus the forest stands map (Appendix I) will be incorporated into DSK’s geodatabase, the class definition for this map follows the standards defined for the biotope type map (ref. section 8.4). The map key was merely changed to emphasise the distribution of the dominant tree species. The forest
Deutsche Steinkohle AG Division for Engineering Survey 38 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
stand map shows ten dominant tree species in the test area. Coverage with other tree species can be neglected due to very small areas and decreased environmental relevance. Some of the mapped fields referred to a one species contain also other tree species, but the area comprises of more than 70% of the dominant species. Mixed forests with less than 70% of dominant tree species have been defined as a separate class, because they occupy a considerable area in size and thus form a distinct biotope type. Additionally to the tree species classification also a subclasses distinction have been performed. The subclasses contain the information about the life stages of the tree stands (Appendices Ia and Ib). For presentation reasons the overview map does not contain the life stages classes of the mapped tree stands, they can be seen in the more detailed maps of subareas with increased risk of water-logging and thus vegetation stress.
7.4 Vegetation stress detection Because of the different sensitivities of each particular plant species (ref. Table 7.1) the further examinations were applied only on Pine and Beech trees. In the EIA nearly no Beech stands are located in high water-logging risk areas, so that plant stress maps are mostly presented at examples from pine trees.
Specie (used name / scientific name) Moisture-Indicator Nutrient-Indicator Pine / Pinus Sylvestris 3 2 Birch / Betula Pendula i i Alder / Alnus 7 – 9 i Oak / Quercus Robur i i Fir / Abies alba i i Spruce /Picea Abies i i Beech /Fagus Sylvatica 5 i Maple /Acer 5 – 6 6 – 7 Table 7.1: Plant Indicators for moisture and Nutrient content of the soil. Higher values indicate, that the plant usually prefers locations with higher content. i indicates indifferent and means, that the plant usually lives on all locations.(ELLENBERG, 1992)
One of the main subjects of the works on the central European test site was the detection of vegetation stress caused by changes of the water balance, in the case of DSK especially in the ground-water balance. As described above plant stress is spectrally expressed in a change of shape and wavelength of the vegetation edge at the transition between red and infrared wavelength. This shape was dressed in eighteen features which were described above. These features are (together with common short cuts for each feature): 1. Full width at Half maximum (FWHM) 2. maximum absorption depth (MaxAb) 3. Symmetry (Sym) 4. Red edge Wavelength (red edge) 5. Minimum in red (min in red) 6. Wavelength of the minimum in red (wl min in red) 7. reflection maximum near 800 nm (shoulder near 800, sh800) 8. Wavelength region between the minimum in red and the red edge wavelength (wl red red edge) 9. – 13. Area of the 1st. – 5th. derivative between 550 and 740 nm (derpart1-derpart5) 14. – 18. Area of the 1st. – 5th. derivative between 420 and 2500 nm (derges1-derges5).
Because changes of the plant status is expressed in the value of a difference from a standard value but not in the direction of this difference the following procedure for the stress detection have been
Deutsche Steinkohle AG Division for Engineering Survey 39 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
applied on different forest stands. First a standard value for each feature and each forest class with no influence of the mining caused subsidence movements had to be defined. This has been done calculating the features for the predefined endmember spectra. These values are stored in a table and used as predefined standard values.
Feature Min Max Mean Stdev FWHM 138,000000 169,000000 166,269067 6,545560 maxAb 0,484480 0,903873 0,846213 0,020815 sym -1,974790 0,571716 0,016514 0,166395 red edge 723,053040 796,503906 727,043228 24,328910 min in red 8,000000 56,000000 17,609867 3,205766 wL min in red 662,000000 677,000000 662,532000 2,774712 sh800 47,000000 334,000000 181,479200 41,552500 wl red red edge -10676,000000 134,503906 53,845628 350,474326 derpart1 28,663843 206,325516 114,874497 23,814633 derpart2 38,769428 105,946159 73,629655 7,644259 derpart3 -54,159782 -1,071044 -29,916728 6,770229 derpart4 -5,949154 41,025124 19,153514 6,082839 derpart5 -7,192945 4,402609 -1,473295 1,626689 derges1 -27,509701 56,585484 -7,733156 6,688159 derges2 -33,898972 39,539581 -2,351159 10,945812 derges3 -28,063385 63,125366 17,949706 12,184478 derges4 -38,706009 17,723446 -11,520743 7,952942 derges5 -5,068799 15,933497 6,154050 2,795740 Table 7.2: Example of predefined feature values for pines (pinus sylvestris).
After that all the features for the hyperspectral image dataset were calculated. The stress evaluation took place by calculating the deviation from the standard feature values for each forest class separately, and the stress factor was defined as the absolute value of this deviation. For visualisation the last step was a threshold classification (density slice). Figure 7.8 shows the results of this stress detection algorithm for pine trees with the most promising features, which were the maximum absorption depth, the red edge wavelength and the area below the second derivative in the wavelength region between 550 and 740 nm.
Because a proper ground truth for stressed vegetation is not available at DSK and was not acquired during the MINEO Project, the verification has to be done using the subsidence amounts and ground- water table depth for the areas. The Figure 7.8 shows the derived stress for pine stands computing the area below the second derivative for regions where the subsidence was more than one meter and the total amount of the groundwater depth is less than one metre. The Figure 7.9 shows, that most of the plant stress (orange and red colours) is concentrated in the parts with the lowest groundwater depth. Higher stress values outside these areas are very rare and can be caused by misregistration between the image dataset and the used GIS-layer (especially at the edges of the allotments, where influences of mixed pixels with the adjacent land use get more frequent) and by generalisation of the GIS-layer (spread over the whole area, not every tree in a forest which is mapped as a pine forest is a pine and small clearings of the forest are unmapped).
Deutsche Steinkohle AG Division for Engineering Survey 40 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
a.) b.)
c.) a.) thresholded image showing the maximal absorption depth deviation
b.) thresholded image showing the deviation in the red edge wavelength position
c.) thresholded image showing the deviaton of the area of the 2nd derivative between 550 and 740 nm
Legend:
Increasing plant stress
Figure 7.8: Examples of plant stress mapping results using different promising features.
Deutsche Steinkohle AG Division for Engineering Survey 41 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 7.9: Comparison of plant stress on pines overlaid with iso-lines of changing depth of ground water level (line labels indicate the number of change in meters).
A statistic comparison of the extracted features from inside and outside the sphere of influence is given in the table below.
The table points out, that especially the features maximum absorption depth (maxab), red edge wavelength (red edge) and the area of the second derivative between 550 and 740 nm are suitable features for the assessment of plant stress on forest areas.
It can be shown that as well the maximum absorption depth as the derivative feature has declined significantly in areas with the influence of mining ground-water alterations. The red edge wavelength value decreased as expected only slightly by averaged 0.03 nm. BONHAM AND CARTER (1988) pointed out, that this is a usual value for the change of the shift of the red edge wavelength.
Deutsche Steinkohle AG Division for Engineering Survey 42 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Inside sphere of influence Outside sphere of influence FEATURE Min Max Mean Min Max Mean FWHM 0 184 162,66 138 169 165 maxAb 0,13 0,909 0,79 0,51 1 0,85 sym -0,79 1,067 0,097 -2,42 0,57 -0,14 red edge 723 795,440 726,29 723,00 796,08 726,32 min in red 0,007 0,112 0,022 0,003 0,052 0,014 wL min in red 662 692 662 662 677 662 sh800 0,048 0,381 0,161 0,041 0,391 0,161 wl red red edge -10661 133,44 24,90 -10661 134,088 35,809 derpart1 0,021 0,234 0,10 0,023 0,240 0,10 derpart2 -0,008 0,102 0,06 0,037 0,117 0,07 derpart3 -0,060 0,045 -0,024 -0,072 0,003 -0,02 derpart4 -0,009 0,035 0,018 -0,008 0,048 0,02 derpart5 -0,006 0,010 -0,001 -0,007 0,006 -0,001 derges1 -0,031 0,075 -0,009 -0,035 0,05 -0,011 derges2 -0,030 0,034 0,001 -0,040 0,045 0,002 derges3 -0,036 0,056 0,009 -0,037 0,061 0,010 derges4 -0,029 0,033 0,005 -0,022 0,035 0,005 derges5 -0,014 0,010 -0,002 -0,015 0,007 -0,003 Table 7.3: Comparison of features inside and outside the sphere of influence.
7.5 Change Detection A big concern in DSK’s monitoring process is the documentation of changes in land use and land cover. Especially the mining induced changes in perennial cultures are of great interest. Remote sensing with the methods of change detection offers a big potential in the assessment of these changes.
Digital change detection methods have been broadly divided into either pre-classification spectral change detection or post-classification change detection methods.
S In post-classification change detection two images from different dates are independently classified and labelled. The area of change is then extracted through the direct comparison of the classification results. The advantage of post-classification change detection is that it bypasses the difficulties in change detection associated with the analysis of images acquired at different times or by different sensors. The main disadvantage of the post-classification approach is the high dependency of the land cover change results on the individual classification accuracies. Post-classification methods can be divided in two types which are: (1) methods based on two independent spectral or spectral spatial classifications, and (2) methods based on two independent true land cover class classifications. For the first type supervised or unsupervised classification techniques can be used. For the true land cover classification in (2) a supervised classification has to be performed (YUAN ET AL., 1998). S Spectral change detection techniques (pre-classification) rely on the principle that land cover changes result in persistent changes in the spectral signature of the affected land surface. These techniques involve the transformation of two original images into a new single-band or multi-band image in which the areas of spectral change are highlighted. The spectral change data must be further processed by other analytical methods, such as classifier, to produce a labelled land cover change product. Most of the spectral change detection techniques are based on some style of image differencing or image rationing (YUAN ET. AL., 1998). For these techniques proper atmospheric corrections and image normalisations are essentials for good
Deutsche Steinkohle AG Division for Engineering Survey 43 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
results. YUAN ET AL. (1998) suggests methods like band to band regression or principal component analysis to simultaneously perform the image to image equalization and the detection of change areas.
Because this spectral change detection is based on pixel by pixel or scene by scene comparison of the reflectance data the coregistration of the different images is the most essential prerequisite for the quality of the results of the change detection technique. TOWNSHEND ET AL. (1991) pointed out, that the error equivalent for the comparison of two NDVI images was greater than 50 % of the actual differences in the NDVI by a misregistration of only one pixel. To achieve an error of only 10 %, registration accuracies of 0.2 pixels or less are required (TOWNSHEND ET. AL. 1991). The spectral methods can be further divided in two strategies, which are (1) pixel – by pixel operations and (2) scene by scene operations.
1. The pixel by Pixel methods are the simplest change detection techniques. In this category coregistered pixel from two dates are processed together to produce an output image in which areas of spectral change can be distinguished from background areas where little or no change has occurred. Some examples for that type are Raw Image Differencing, Change Vector Analysis, Image Rationing or Vegetation Index Differencing. 2. The scene wise methods were developed to minimize the effects of using different sensors, or different illumination conditions or atmospheric scattering for change detection. All these conditions can be viewed as scene to scene radiometric differences, which can be addressed and largely removed prior to spectral change analysis. Based on these considerations methods like Normalized Image Differencing, Radiometrically Normalized Image Differencing, Albedo Differencing or Principal Component Comparison have been developed ( YUAN ET AL., 1998).
The spectral change identification methods and the classification based methods can be combined in various ways to minimize errors in land cover change analysis. An example of these hybrid methods may be that radiometrically normalized image differencing is used to identify the area of significant spectral change, and then apply a post-classification comparison method within areas where spectral change was detected to obtain class information.
For the MINEO works as well the pre-classification as the post-classification approach were tested. Firstly a visual comparison of the available data sets from different sensors have been performed. During the subsequent pre-classification change detection vegetation indices and principal components of the two available HyMap data sets were compared and analysed. The post-classification approach was applied to image derived spectral parameters, where each particular feature from both dates were compared.
7.5.1 Data and Data Preparation for change detection For the works on change detection two HyMap datasets were available. The first was a dataset from August 1998. It has a geometric resolution of 7 m and 128 spectral channels between 420 and 2481 nm. It was atmosphere corrected by DLR using a preliminary version of ATCOR 4. A Parametric Geocoding could not be performed because the sensor was not mounted on an IMU Platform during the flight and no GPS Data was available for the flight. The second dataset was the one flown during the MINEO project. It has a geometric resolution of 5 m and 126 spectral channels between 435 and 2485 nm. Next to a parametric geometric correction using PARGE (refer to chapter 6.1) the atmospheric correction using ATCOR 4 was done by DSK (see chapter 6.2). To be able to compare the two images pixel by pixel, the pixel grids of each image must conform to the other image in the data base. Rectification is the process of making an image conform to another
Deutsche Steinkohle AG Division for Engineering Survey 44 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
image. For the works in MINEO it was necessary to register the geometrical uncorrected image from 1998 with geocoded 2000 image. This was done using the polynomial approach implemented in ERDAS/Imagine 8.5. Accurate Ground Control Points (GCP´s) are essential for an accurate registration of the images. From the ground control points, the rectified coordinates for all other points in the image are extrapolated. Because of proper results the data was fragmented in two parts which were registered separately. For both parts together approx. 400 GCP´s spread all over the image were defined. The distribution of these Ground Control Points is given in Figure 7.10 as an example.
Figure 7.10: Ground control points spread over the image.
Then a 5th degree polynomial registration was performed. In the same step the 1998 HyMap data were resampled to a pixel spacing of 5 m (for visual results refer to Figure 7.11).
The image below demonstrates the registration results for a subset of the whole image. It shows the channels 20 (r), 40 (g) of the 2000 image and the channel 80 (b) of the 1998 image in one colour composite, so that misregistration gets visible. It can be seen that on this sample part of the image no major misalignments can be identified. Statistical checks using independent control points delivers medial values below one pixel (for this sample part).
Deutsche Steinkohle AG Division for Engineering Survey 45 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 7.11: Registration results.
7.5.2 Change Detection: pre-classification Change detection analysis was focused on areas with increased risk of water-logging occurrences. In the following a description is given on an exemplary sub area in the test site, where a development of a ground-water lake can be studied over a period of five years. Although DSK holds a variety of remote sensing data sets, this works have been performed only on the HyMap images, since a high spatial resolution is needed for this small parcelled area. Merely an additional DPA (Digital Photogrammetric Assembly) data set, acquired 1997, with ground resolution of 50x50 cm per pixel was applied for visual interpretation.
Examining the DPA image from year 1997, one can see, that the water-logging occurrence resides in an initial phase. No open water can be seen, just a thinned vegetation cover is distinguishable, represented by the bluish colour within the rectangle in the Figure 7.12A. The HyMap image from 1998 shows already a ground-water lake represented by dark colours originating from a relative ground-water uplift due to surface subsidence. This water body was once again recorded in year 2000 during the MINEO hyperspectral air-survey. This image documents an increased size of this lake, manifesting the continuous subsidence movement at this site.
The water-logging affects the surrounding area. This results in alterations of plant community composition, arising of new wetland biotopes and rededication of land use. The perennial plants, e.g. trees with already developed root system, suffer stress, since they cannot adapt easily to the changed environmental conditions. This can be observed in decreased photosynthetic activity (PA), loss of leaves and consequently forest dieback (ref. Figures 7.12 and 7.13). Looking at Figure 7.12C, one can see, that the brownish colour of the trees within the lake turns over greenish into dark colour due to suffered stress, leave loss and results consequently in death of the trees.
Deutsche Steinkohle AG Division for Engineering Survey 46 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
A Interpreting the principal components of the HyMap images, the changed vitality status of the trees within the subsidence lake can be easily recognised. The Figure 7.14 shows two false colour composites of the HyMap data sets. The forests are represented in green colour whereas the open water body appears enhanced in red colour. The image from 1998 shows green trees in the lake and also green-yellow areas, indicating photosynthetic active vegetation in wet conditions. The yellow spots can be interpreted as mixture of reflective response from vegetation and water, since yellow means a transition from green into red colour.
As already mentioned above, after the visual interpretation of B the images shown in Figure 7.12 , 7.13 and 7.14 not only the vitality status decreased within the two years can be concluded, but also the subsidence lake size increment can be identified. In order to enhance the area of the water body increment, the two HyMap data sets were compared together. A normalised difference image (NDI) was computed using the first principal components (PC) from both HyMap data sets.
PC1_ Hy00 PC1_ Hy98 NDI(PC1) (7-1) PC1_ Hy00 PC1_ Hy98
According to CHAVEZ (1989) and SABINS (1997) the first PC C (PC1) contains the most significant information from all channels of the hyperspectral data sets. Utilising these first principal components (PC1’s) from the two images the difference between these data sets will be enhanced and not the different spectral response between the single channels.
Figure 7.15 shows the colour-coded NDI(PC1)-image. The blue colours indicate areas of either no or little change, whereas yellow and red shall be interpreted as areas of high change. However it has to be pointed out, that this kind of image operation requires co-registered image pairs and is high sensitive to mis-registration, since the NDI computation bases upon the pixel-by-pixel approach. Figure 7.12: Visual interpretation of three data sets recorded over a Figure 7.12: Visual interpretation of three period of five years. data sets recorded over a period of five years. Picture A (DPA image; CIR) shows an initial stadium of water-logging occur- rence. Picture B and C (HyMap images; false colour composites) show open water body due to subsidence. For details see text.
Deutsche Steinkohle AG Division for Engineering Survey 47 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
HyMap ‘98 HyMap ‘00
Figure 7.13: NDVI images derived from the HyMap data sets. These two pictures show the alteration of the vegetation vitality within the subsidence lake. High values (white) indicate high vitality, low values (black) indicate low vitality.
HyMap ‘98 HyMap ‘00
Figure 7.14: Visual comparison of the two hyperspectral data sets. False colour composite (R:PC4, G:PC9, B:PC1). See text for details.
Deutsche Steinkohle AG Division for Engineering Survey 48 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
III
I
II
Figure 7.15: Change detection NDI(PC1) image. This image is colour coded to facilitate the visual interpretation. Bluish colours indicate no or little change, yellow and red show areas of high change degree. Areas I and II show detected changes. The size accretion of the subsidence lake are highlighted at II; the field at I laid idle in year 2000 in contrast to year 1998; the red pixels at III results either due to miss-registration of the images or due to different illumination conditions, causing shadow effects.
7.5.3 Change Detection: post-classification For the detection of land cover status changes the 18 features (ref. to chapter 7.4) implemented for stress detection were calculated for both time pieces of the data. Afterwards a difference between the features in 1998 and 2000 was calculated. At last a change factor was defined as the absolute value of each particular feature. The Figure 7.16 shows a scaled image of this change factor for the area of the second derivative between 550 – 740 nm.
The left image of Figure 7.16 shows increasing status changes from blue to red colours. It can be seen that many red colours are agricultural fields, which were planted in 1998 and led idle in 2000 or vice versa. Strong changes are to be found on roads and houses, two. The reason for that can be found in the misregistration of the two images. The detected changes on forest areas are real changes of plant health. The right image of Figure 7.16 shows real changes (increasing from green to red) of perennial plants (here pine trees). One can see that the most significant changes (red and orange colours) are concentrated round a lake which came into existence in the last years caused by subsidence movements.
Deutsche Steinkohle AG Division for Engineering Survey 49 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 7.16: Picture on the left shows the result derived from the 2nd derivative between 550 and 740 nm, scaled image to change factor, increasing plant stress from blue to red; the right picture shows the same result scaled for pine trees and overlaid on a HyMap channel, increasing stress from green to red.
7.6 Description and assessment of the maps produced vs. expected results Land cover and land use maps were produced using supervised classification approach. The maximum likelihood classifier was applied. The forests cover ca. 40% of the test site, about 72% of total forest coverage fall to deciduous trees and 28% to conifer trees. The aim was to obtain a maps of forest stands which describe the tree species and their actual live stage. The maps includes classes of ten dominant tree species and subclasses describing their live stages (ref. App. I, Ia and Ib). Coverage with other tree species can be neglected due to a very small areas and decreased environmental relevance. Mixed forests have been defined as a separate class, because they occupy a considerable area in size and therefore form a distinct biotope type.
The achieved over all mapping accuracy reaches 73% (kappa coefficient = 0,6519). The accuracy assessment is done by calculating the confusion matrix and is shown below in Table 7.4 and 7.5. The accuracy of the classification process of the tree species and their subclasses of live stage is quite sensitive. The more classes will be defined, the lower the over all accuracy will be, therefore a well balance between the thematic accuracy, the amount of classes and subclasses and the over all accuracy is needed.
Class C Commission Omission Commission Omission Class P Prod, Acc, User Acc, Prod, Acc User Acc, (Percent) (Percent) (Pixels) (Pixels) (Percent) (Percent) (Pixels) (Pixels) Beech 19,05 50,55 42319/222132 183778/363591 Beech 49,45 80,95 179813/363591 179813/222132 Oak 44,69 53,61 37766/84516 54034/100784 Oak 46,39 55,31 46750/100784 46750/84516 Alder 38,7 59,05 42227/109127 96477/163377 Alder 40,95 61,3 66900/163377 66900/109127 Birch 83,95 45,2 35748/42 5835637/12472 Birch 54,8 16,05 6835/12 4726835/42583 Poplar 75,38 45,98 56614/75103 15739/34228 Poplar 54,02 24,62 18489/34228 18489/75103 Ash tree 94,87 36,44 85662/90 2932655/7286 Ash tree 63,56 5,13 4631/7 2864631/90293 Maple 60,52 57,38 13313/21998 11695/20380 Maple 42,62 39,48 8685/20 3808685/21998 Spruce 63,99 58,87 49894/77970 40178/68254 Spruce 41,13 36,01 28076/68254 28076/77970 Pine 12,35 19,1 47324/383143 79270/415089 Pine 80,9 87,65 335819/415089 335819/383143 Larch 72,54 69,6 18863/26003 16347/23487 Larch 30,4 27,46 7140/23 4877140/26003 mixed deciduous trees 2,74 1,05 11954/436082 4501/428629 mixed deciduous trees 98,95 97,26 424128/428629 424128/436082 mixed conifer trees 0,12 10,88 41/33 2094051/37219 mixed conifer trees 89,12 99,88 33168/37219 33168/33209 Table 7.4: Accuracy assessment of the maximum likelihood classification; all-over accuracy reaches 73 % and the kappa coefficient is 0,6519.
Deutsche Steinkohle AG Division for Engineering Survey 50 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Table 7.5: Accuracy assessment of the maximum likelihood classification. continuation of Table 7.4.
However it has to be said, that the biotope type map, which have been used as ground truth, bear some inaccuracies since it has been produced for different purposes. The map contain information about the biotope types and not about the plant distribution. Biotope is defined as a specific plant species community, i.e. besides the dominant species also additional species may occur, which may be detected by the remote sensing sensor, but compared with the biotope type map (ground control, in this case) they will be assigned as misclassification. This situation may also happen in converse manner too.
The aim of vegetation stress detection was to detect this stress in areas affected by water-logging occurrences, and the second aim was to detect this stress already in the initial stage. Therefore the focus has been given to the perennial plants with high sensitivity to changing environmental conditions. As such the trees, here pines, have been found. But since water do not stress plants as such, on the contrary it is rather essential element for vegetative life, it was hard to detect plant stress at the beginning stage caused by water-logging. Nevertheless we were able to detect plant stress in advanced stage (ref. Figure 7.8, Figure 7.16 and Appendix II). Figure 7.9 shows, that the detected vegetation stress fits very well with the water-logging occurrence and further this can be interpreted as result of mining caused subsidence movements.
7.7 Invaluable contribution of hyperspectral imagery vs. conventional sensors The use of hyperspectral data at this test site to study the vegetation stress caused by water-logging has been invaluable to the detecting and mapping this stress occurrences. In particular the high spectral resolution of the applied data sets was the key attribute which enabled us to perform the analyses described in previous sections. The use of data sets from the conventional optical sensors, like LANDSAT, SPOT, IRS, IKONOS or other space- or airborne pan chromatic or multispectral instruments would not allow to reach the objective defined for this test site.
However it has to be pointed out, that the application of such data type needs to be supported with an extensive data base of auxiliary data sets and still needs to be accompanied by field works and ground truth, e.g. spectral field surveys.
Deutsche Steinkohle AG Division for Engineering Survey 51 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
7.8 Generic character of the procedure/algorithm The procedures and methods developed and used for this test site can be applied to other test sites with similar environmental conditions. This procedures and methods are set up as a generic workflow model (ref. to Appendix IV b and IV c), which can be followed to obtain the desired results. The generic character of the developed methods and tools is manifested in the physical rules, which control the spectral behaviour of the material under study. However it has to be said, that this methods and procedures are extremely sensitive to the quality of the records and conditions accompanying each hyperspectral survey, since lot of factors may influence the data record and thus lead to different, and sometimes erroneous, results. Therefore all these proposed workflow procedures are designed to be used as flexible as possible. The topics and problems this study deals with are a matter of current research world wide, and therefore they should be understood as a contribution to this research work. This procedures resides still under current development and improvement processes.
Deutsche Steinkohle AG Division for Engineering Survey 52 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
8 DESCRIPTION OF THE GIS DATABASE As already mentioned in chapter 2 and 4 an extensive GIS database is held by DSK for the Kirchheller Heide test site. The GIS database is build up from the late eighties on and is frequently updated to reflect the actual status of the mining site itself and the environment around the mine. The GIS database contains all relevant layers necessary to perform mining operations and, in particular, to design the Outline Operating Plan (OOP) and conduct the Environmental Impact Assessment (EIA). In the following subsections only those GIS layers are described, which were relevant for processing the HyMap imagery during the MINEO project, either as input data and “ground truth” or as additional information source to derive new information layers for this test site.
8.1 Digital Terrain Model (DTM) The shape of the test site is best described using the digital terrain model (DTM). DSK uses such DTMs to incorporate the topographical information of every area where an EIA has to be conducted. The DTM is based upon an aerophotogrammetric survey. These surveys are done at regular times to obtain the occurring topographical changes in the affected areas. The aerial photographs used to generate the DTM for the test site Kirchheller Heide were taken in 1993 with a scale of 1:6000. The workflow procedures developed for DTM generation and DTM update are described in chapter 9. The resulting DTM is of high quality with an accuracy of 5 cm in the horizontal direction and 20-30 cm in the vertical direction. Zones with permanent changes, like active mine heaps and gravel open pits are consciously excluded from the DTM.
8.2 Colliery and subsidence data As already described in chapter 2.8 underground mining inevitably produces cavities in the rock body in which the overlying strata collapse and leads consequently to subsidence movements at the surface. Applying the deposits volume and planned coal production, the geological data, such as seam bedding and tectonic conditions along with the geology of the overlying strata, the expected subsidence at the surface can be computed for any desired date. Underground mining represents a dynamic system, i.e. the spatial and temporal changes may occur regarding to the plans as the coal production proceeds and the knowledge of the deposits bedding conditions increase. Therefore a constant update of the data is needed to recalculate the subsidence movements. Applying this subsidence amounts, the DTM from 1993 of the Kirchheller Heide can be updated reflecting the actual topographical situation in the study area. The DTM-93 has been corrected applying the subsidence data computed for August 2000, time of the MINEO hyperspectral survey at the German test site.
This sophisticated and well calibrated subsidence computation model enables also a forecast of anticipated subsidence movements in the future. Thus the subsidence data are one of the key GIS data layer, since they serve as input data-set for updating the DTM and the DGWM, and in consequence important for the environmental impact assessment.
8.3 Digital Ground Water Model (DGWM) The high complexity of the hydrogeological situation in the test site requires the application of an digital ground water model (DGWM) in terms of describing and analysing the changes in the ground water (GW) conditions. The workflow procedure developed for DGWM generation is described in chapter 9.
Deutsche Steinkohle AG Division for Engineering Survey 53 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
8.4 Biotope types and actual land use A crucial point in the EIA is the description and assessment of the mining-caused environmental impacts at the Kirchheller Heide test site. Several items for preservation have been defined, e.g. biotopes, water and water courses, soil, landscape and others. Biotope and land use maps play a central role in this assessment. Mapping of the biotope inventory was accomplished by Institut für Vegetationskunde, Ökologie und Raumplanung (IVÖR) in year 1993. It was updated in year 1995 by Institut für Landschaftsentwicklung und Stadtplanung (ILS). The maps were produced in scale of 1:5000. In summer 2000 near to the HyMap overflight the actual land use was mapped in selected parts of the test site. These parts correspond to the areas with increased risk of water-logging. These sub-areas were also subject of the extensive analysis of the hyperspectral data sets.
The biotope and land use maps were mapped under the guidelines of Landesanstalt für Ökologie, Bodenordnung und Forsten (LÖBF) [Regional Office of North-Rhine-Westphalia for Ecology, Land Division and Forestry] and adjusted by ILS for the EIA demands performed by DSK.
The biotope maps in the forest stands mapping units have predefined map key, which includes the following attributes: - tree species, - life stage, - habitat moisture level, - index of natural vegetation.
The life stages are divided into the following levels: - 1: tree free (clearings, fallow lands), - 2: reforestation up to 2 m of tree high, - 3: pole woods (DBH <15 cm), - 4: DBH 15-50 cm, - 5: DBH > 50 cm. The habitat moisture level are divided in the following classes: - 1: wet, - 2: moist, - 3: fresh, - 4: dry. The index of natural vegetation is levelled as follows: - 1: natural vegetation (vegetation community which develops without human interaction), - 2: forest stands with single tree species for timber production, - 3: Niederwald (“lower forest”, young forest, mixed, used intensely), - 4: Mittelwald (“middle forest”, middle aged forest, mixed, with some older trees, the use is not very intensive).
The biotope maps, having the rich information layers in the background, serve as a good data source to assess the environmental status at the test site. However, they bear some inaccuracies when used as ‘ground truth’ for remote sensing data analysis. The fact, that in an given mapping unit, e.g. pine stand, also trees from different species may occur leads to some inaccuracies while classifying the remote sensing data, in particular, when using high spatial resolution imagery (ref. to chapter 7.6). Also an additional information layer of canopy closure would be desirable for the analysis of remote sensing imagery.
Deutsche Steinkohle AG Division for Engineering Survey 54 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
9 GIS MODELLING 9.1 Description of the conceptual environmental model The underground mining of coal causes subsidence movements and thus leads to specific environmental impacts on the surface. At the German MINEO test site no contaminant distribution or land/water/air pollution is taking place. The impacts here result in alteration of the hydrological situation, land use rededication, wetland development and biotope type alteration. Figure 9.1 shows a simplified conceptual environmental model which depicts the causales, the affected media and the expected results.
Groundwater
Impacts to ecosystems and Subsidence Surface water urban areas :
water- loggings agricultural land use rededication residential areas Wetlands development of new wetlands
Figure 9.1: A simplified conceptual model of environmental impacts due to subsidence movements at the German MINEO test site.
9.2 Objectives of GIS modelling and expected results The GIS data sets described in chapter 8 are used to model the environmental conditions like they may be anticipated in the future. This is one of the key objectives of the EIA which has to be performed on every mining site managed by DSK. As stated in section 2.9 and chapter 10 DSK is compelled by the Federal Mining Act to perform an EIA which contain an impact forecast of the planed mining activities to the environment and a land management plan. To fulfil these objectives the coal production plans are incorporated into this extensive GIS data base. The resulting models describe the foreseen environmental conditions. For the Kirchheller Heide study area the EIA prescribes three time points, 2004, 2009 and 2019, where the conditions have to be forecasted and the models recalibrated with updated input data if necessary. The mining operation is escorted by the EIA during its whole lifetime, so the models are updated concurrently and their accuracy increases in time.
9.3 Description of GIS derived layers During the GIS analysis and modelling the following layers were derived: - subsidence updated DTM’93 for 2000, 2004, 2009, 2019, - DGWM modelled for 2004, 2009 and 2019, - updated groundwater level conditions for 2000, 2004, 2009, 2019, - water-logging occurrences modelled for 2004, 2009 and 2019, - updated land cover and land use map for 2000.
Deutsche Steinkohle AG Division for Engineering Survey 55 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
9.4 Generation of a high resolution Digital Terrain Model based on different input data To obtain occurring topographic changes of affected areas aerophotogrammetric surveys are done at regular times. They facilitate 3D-measurements with high accuracy using analytical and digital photogrammetric plotters. Based on these measurements detailed digital terrain models (DTM) are generated using the ArcGIS TIN module.
A two-stage procedure is agreed upon for the measurement of the digital terrain model. Firstly, the DTM stage 1 is reconstructed by measuring a regular grid of 5 metres mesh size and build the subsequent triangular mesh (ArcGIS TIN). This DTM, initially roughly structured, is available to the specialist bodies involved within a very short time, for example, for the calibration of the ground water model.
In order to reproduce the morphology of the terrain surface, including the areas in which the rough grid cannot record the fine-structured terrain forms, the next stage involved the measurement of additional stations along breaklines. Once the extent and nature of the object types to be recorded in the second stage, e.g. artificial breaklines or slopes along the banks of drainage channels, have been agreed on with those responsible, the DTM stage 2 can be calculated (Figure 9.2). Use of terrestrial surveys in areas not visible from the air, information about the beds of flowing watercourses, analogue map material, and the analytical or digital photogrammetric images all ensured an elevation accuracy of 20 – 30 cm.
Figure 9.2: Digital terrain model (3D view, elevation from 40 m a.s.l. (yellow) to 100 m a.s.l. (light brown)) with integrated watercourse network (blue lines). Wooden areas with old trees are shown as olive green blocks).
By extracting all hydrologic elements of relevance from the DTMs a three-dimensional water course network (route system) representing the present water flow status is reconstructed. This is done using the ArcGIS Dynamic Segmentation module. This module also generates event tables labelled with descriptions of the watercourse.
Deutsche Steinkohle AG Division for Engineering Survey 56 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
However, the ecological and hydrological assessment considers not only the current situation of the watercourse system, but also forecasts the status at various intervals during the next two decades by pre-calculating subsidence and merging it with the watercourse network. Binding watercourse development plans are incorporated at the relevant time intervals. This "four-dimensional" watercourse network is then used to generate site plans – including the route-measured watercourse system - and longitudinal sections (Figure 9.3) - illustrating the watercourse bottoms, breaklines on slope tops and hydrological relevant topographical elements at different intervals. These longitudinal sections form the basis for the assessment of water flow changes resulting from underground extraction and for the planning of appropriate measures.
Figure 9.3: Longitudinal section – derived from the four-dimensional watercourse network.
The corresponding MINEO generic procedure ‘’KH_GIS_MAKE_DTM’ can be found in Appendix IV d.
9.5 Updating of the Digital Terrain Model with calculated / measured subsidence data As already described in chapter 2.8 underground mining inevitably produces cavities in the rock body in which the overlying strata collapse and leads consequently to subsidence movements at the surface. Applying the deposits volume and planned coal production, the geological data, such as seam bedding and tectonic conditions along with the geology of the overlying strata, the expected subsidence at the surface can be computed for any desired date. Underground mining represents a dynamic system, i.e. the spatial and temporal changes may occur regarding to the plans as the coal production proceeds and the knowledge of the deposits bedding conditions increase. Therefore a constant update of the data is needed to recalculate the subsidence movements. Applying this subsidence amounts, the DTM from 1993 of the Kirchheller Heide can be updated reflecting the actual topographical situation in the study
Deutsche Steinkohle AG Division for Engineering Survey 57 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
area. The DTM-93 has been corrected applying the subsidence data computed for August 2000, time of the MINEO hyperspectral survey at the German test site.
This sophisticated and well calibrated subsidence computation model enables also a forecast of anticipated subsidence movements in the future. Thus the subsidence data are one of the key GIS data layer, since they serve as input data-set for updating the DTM and the DGWM, and in consequence important for the environmental impact assessment.
By combination of these DTMs with subsidence data the topographical changes can be modelled and predicted for the next two decades.
Figure 9.4 shows a DTM 2019 with incorporated subsidence data. The DTM which is based upon aerophotogrammetric surveys from 1993 has been lowered by the subsidence amounts derived from the colliery data. The resulting DTM exhibits the actual terrain elevation of the study site during summer 2000.
70m a.s.l.
30m a.s.l.
Figure 9.4: DTM with integrated subsidence data. Area affected by subsidence is shown by black dashed line, subsidence isolines are depicted in blue with centres of max. subsidence.
Figure 9.5 shows ground water-logging occurrences modelled for year 1993 (left picture), the starting point of observations for the EIA and modelled for 2019 incorporating the coal production plans along with the corresponding estimated subsidence movements and the changed hydrological situation resulting from the subsidences.
Deutsche Steinkohle AG Division for Engineering Survey 58 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Figure 9.5: 3D elevation view (from green = 30 m a.s.l. to grey = 70 m a.s.l.), watercourse network (blue lines) and lakes (light blue). Estimation of sensitive areas at risk of ground water-logging. The dark blue spots indicate sites of the modelled ground water logging for 1993 (left) and 2019 (right).
Figure 9.6 shows the land use GIS data sets updated during the field survey in summer 2000. For the demonstration reasons only the forest stands are shown in this Figure.
Figure 9.6: Updated land use map, here forest stands (light green areas), overlaid on the HyMap data (R = 0.890 µm, G = 1.685 µm, B = 0.662 µm) in 3D-view draped over the DTM.
Deutsche Steinkohle AG Division for Engineering Survey 59 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
The corresponding MINEO generic procedure ‘’KH_GIS_UPDATE_DTM’ can be found in Appendix IV e.
9.6 Calculation of ground-water isobath at different times The generation of the DGWM has been subcontracted to Ingenieurbüro GKW. The DGWM is a three dimensional, stationary ground-water-current-model generated using the finite elements method. Usage of this DGWM enables not only the description of the actual GW conditions, but also the simulation of the future hydrological situations for defined points of time. The main objective is to present, estimate and forecast the trend and dimension of GW level alterations taking different subsidence settings into account. The complex hydrogeological conditions, especially the first GW storey in the thin layered quaternary sediments which is partly in hydrological contact with the second and third GW storey made necessary the deployment of a 3D model with five levels in the vertical direction (RÜBER 1997). The influence of the receiving water course on the GW conditions has been taken into account applying a leakage approach. The DGWM is based upon the following input data sets: - Geological maps: Sheets Bottrop (4407) and Dinslaken (4406) in scale of 1:25000 - Hydrogeological maps of the Rhine-Westfalia coalfield in scale of 1:10000, following sheets were used: - Bottrop (1966), - Feldhausen (1967), - Sterkrade-Königshardt (1968), - Hünxe (1978), - Kirchhellen (1966), - Dorsten (1972), - Lohberg (1967), - Damm (1978), - borehole logs, - digital terrain model , - land use data, - precipitation rate data, - water course system data, - ground water recharge data, - data from ground water measuring stations.
The model enfolds an area of 153 km², consequently it covers an area which is much larger than the test site. Summer 1993 has been stated as a status quo for the GW conditions, i.e. without mining induced influences. From this date the DGWM was frequently updated with new, actual input data. The integration of the updated DTM with different subsidence settings enables the forecast of hydrological conditions for any desired date in the future.
Precise ground-water models (GWM) were developed to derive parameters of the geohydrological situation. Spatial and temporal changes of the hydrological situation result from the amount of subsidence, local ground-water situation and actual land-use. To estimate changes of the dynamic system “subsidence-ground-water-vegetation”, the ground-water recharge rate and isobars of the ground-water table are calculated in detail. This is done for different times up to 2019 as a prediction.
The corresponding MINEO generic procedure ‘’KH_GIS_ISOBATH’ can be found in Appendix IV f.
9.7 Hyperspectral data as additional and independent GIS layers
There are three main results from the hyperspectral image processing that can be used as new thematic layers in the DSK GIS database used for environmental planning and monitoring:
Deutsche Steinkohle AG Division for Engineering Survey 60 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
- forest stand maps - vegetation stress maps - change detection maps
The forest stand maps derived from Hymap data can be used for refinement (concerning the tree species) and update of the already existing land cover and land use map. Because this map is based on topographical maps in scale 1:5000 it is not possible to derive new accurate delimination lines from the Hymap data, that meets the needed accuracy. But the Hymap forest stand maps can be used as a “hot spot” map to indicate areas with changed vegetation since the last terrestrial mapping or to give a more detailed species view to wooden areas. The update of the GIS database has to be done by screen digitizing using the Hymap forest stand maps as background layer.
Figure 9.7: Detail of the terrestrial land cover and use map, showing the relevant area marked with AA02 = beech forest
Figure 9.8: Detail of HyMap derived forest stands, showing the distribution of the different tree species within the same area as shown in figure 9.7.
Deutsche Steinkohle AG Division for Engineering Survey 61 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Because the vegetation stress on the central European test site is caused mainly by changes of the ground water balance a ground water model has been developed and the ground water situation has been modelled and forecasted for different time periods up to 2019. The hyperspectral vegetation stress maps can be used to validate, whether the change of the ground water level really has an effect on the vitality of the vegetation.
Figure 9.9 shows a combination of changes of the ground water table – derived from ground water model, digital terrain model and subsidences – and vegetation stress – derived from hyperspectral data. The area with significant stress detection fits perfectly to the area with the highest increase of ground water table.
Figure 9.9: Change of ground water table (yellow areas = lowering, light lilac areas = no change, dark lilac areas = increasing less than 1.75 meters, area with red outline = increasing more than 1.75 meters) and plant stress on pines (increasing stress from green to red pixels).
Deutsche Steinkohle AG Division for Engineering Survey 62 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
10 MONITORING IN GERMAN HARD COAL MINING - LEGAL ASPECTS & DSK’S MONITORING CONCEPT 10.1 Legal aspects §§ 50 ff. of the Bundesberggesetz (BBERGG, 1980 = Federal Mining Act) states that mining operations may be started, conducted and discontinued only on the basis of operating plans. The responsible authorities are the Federal States mining authorities. All further remarks relate to the legal situation and the administrative practice of the mining authorities of the State of North Rhine-Westphalia. Different state provisions may apply to mining operations in other federal states, and in particular to DSK mines in the Saarland.
In terms of the registration and monitoring of the environmental impacts of mining activities, the Outline Operating Plan (OOP) is of decisive importance. The OOP contains for a prolonged period (generally fifteen to twenty years) the general information regarding the intended project, its technical implementation and the probable chronological sequence (§ 52, No. 1 BBergG). More specific details are then provided in the Main Operating Plan, which governs in details the setting-up and conduct of a mining operation for a period of two years in each case.
Particularly under certain background conditions, the performance of an Environmental Impact Assessment is necessary within the outline operating plan procedure and it is frequently necessary in the context of approval of the outline and main operating plan to fulfill subsidiary provisions which stipulate, in particular, monitoring of the land surface.
10.1.1 The Environmental Impact Assessment in the mining context In 1985, the EU issued a directive on Environmental Impact Assessments (EIAs) (EU, 1985), which was implemented in the national law of the member states in the following years. In Germany, the Environmental Impact Assessment Act (UVP-G, 1990) requires that the impacts of human interference in the natural habitats of plants, animals and humans must be regulated in the context of Environmental Impact Assessments (EIAs) with the objective of maintaining the ecological equilibrium.
Under the Environmental Impact Assessment Act, mining projects under BBergG are subject under certain background conditions to an EIA. More specific details on the requirements are provided in the implementation ordinances of the various federal states (KLEINSCHMIDT, 1993). In North Rhine-Westphalia, the background conditions are defined in § 1 of the Ordinance on Environmental Impact Assessments in the mining industry (UVP-V, 1992): - if subsidence of the surface of 3 m or more occurs or - if subsidence of the surface of 1 m to 3 m with significant impairment of watercourses, groundwater, soil, protected cultural assets or similar assets may be anticipated.
10.1.2 Special features of the planning of mining projects Mining projects necessitate a long planning and exploration phase, and therefore long-term planning certainty in view of the necessary investment. The tied nature of mining operations to the deposits and the changing knowledge of geological conditions during mining necessitate a dynamic mode of operation in which mining planning can only be finalized in detail during ongoing mining (KNÖCHEL, 1992 and SALEWSKI, 1991). The period of planning for mining projects in underground coal-mining is at least twenty years (HANSEL, 1993) and is characterized by numerous changes and revisions to the plans.
Deutsche Steinkohle AG Division for Engineering Survey 63 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Due to the long-term nature of such planning, only the outline operating plan is suitable for long-term planning accompanying the entire project and, therefore for integration of the EIA. DSK is currently engaged in an outline operating plan procedure with an integrated EIA for each of its mines in the Ruhr region. The corresponding approvals are anticipated between the years 2001 and 2005.
10.1.3 The Environmental Impact Study The central element of the EIA is the Environmental Impact Study (EIS). DSK, as the project sponsor, is obliged under § 6 UVP-G to present in detail the impacts of the project on the environment. This is accomplished in the EIS. It should be noted in this context that variants of or amendments to plans frequently occur as a consequence of dynamic modification of mining planning, with the result that underground mining of coal is at all times at both the planning and the implementation stage simultaneously (KNÖCHEL, 1992). Registration of the actual impacts of mining-induced subsidence on the surface is therefore possible only during ongoing mining. Unlike the EIA for static projects with a once-only utilization of land (e.g. highway construction projects), the EIA in the case of mining projects is obliged to assess a continuous and dynamic process. Special attention is devoted to this fact in the EIS by examining the anticipated impacts of the project on the environment in the planning territory at a range of times and in a range of periods. The EIS for mining projects can be subdivided into the following components (ILS, 1997): - Description and assessment of the environment and its constituent parts; - Forecast for the environment and its constituent parts without the planned project ("Status Quo" forecast); - Analysis of the sensitivity of the assets affected; - Elaboration of impact forecasts not taking account of control provisions; - Description of any significant anticipated impacts of the project on the environment.
Figure 10.1: DSK-EIS-areas in the Ruhr District (ca. 750 km²) – including the MINEO test site Kirchheller Heide (“PH” – Prosper-Haniel mine)
Deutsche Steinkohle AG Division for Engineering Survey 64 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Within the scope of the EIA, the EIS constitutes the results of an expert investigation complete with evaluation proposals and recommendations (WYCISK, 1993), which serves the responsible approval authority as a basis for decision on approval of the project. A comprehensive determination and description of the impacts of the planned project must be performed in this context, taking account of possible interactions between the individual environmental factors. It is therefore necessary to follow a comprehensive evaluation concept orientated around the balance of nature. In view of the large number of EIA procedures which DSK is obliged to pursue as a project sponsor, a specimen requirement profile has been developed in order to standardize EIS contents and methods (ILS, 1995). As a result of the large range of information to be covered within the scope of the EIS, a geo-information system has, in addition, also been used from the start for complete processing of all EIS data. Corresponding computer-based processing concepts have been elaborated, in order to assure uniformity (BENECKE ET AL., 1999A; HENTRICH & VOSEN, 1999; VOSEN, 1997).
The first step in accordance with the specimen requirement profile for the EIS takes the form of delineation of the investigation area and of the relevant spatial units, in order to provide a working basis. In a second step, the bases for the individual assets are registered and an appraisal vis-à-vis of the anticipated changes is performed, taking into account any existing damage. In the next step, the forecasts for the area under observation are determined both with and without the impacts of mining, and an analysis of the probability of change is performed. These results provide the basis for an evaluation of the predicted changes in condition within the context of a conflict analysis. In addition, provisions which can be applied to avoid or at least reduce negative impact on the environment are formulated. Attention is drawn to the relevant literature (see, for instance, GASSNER & WINKELBRANDT, 1992, PFAFF-SCHLEY, 1992, FROELICH & SPORBECK, 1995) for a detailed examination of the factors to be investigated and the procedures applied in an EIS.
A comprehensive analysis of the interactions within and between the individual elements of the landscape balance impact structure would be desirable for ecological reasons, but is in practice not feasible, due to the complexity of the interrelations involved. The performance of the natural ecosystem can, therefore, only be evaluated in terms of specific formulated targets. Qualitative analysis useful for the purpose of environmental monitoring can thus be accomplished only by means of modelling orientated around the targets.
The results of the EIS must then be taken into account in the decision on the permissibility of the mining project by the responsible authority. Inevitably, forecasts of anticipated environmental impacts (succession forecasts for the plant communities, prediction of subsidence, etc.) contain uncertainties, as a result of the dynamic process described above. Also to be added are the uncertainties which derive from modelling inaccuracies. This forecasting uncertainty is taken into account as early as the ongoing outline operating plan procedure by means of definition of environmental quality targets which must be met (HANSEL, 1993). The procedure terminates, following appropriate involvement of the specialist authorities affected and of the public, and following consideration of all relevant interests, with approval of the project, naturally taking account of environmental impact. Depending on the individual case, various conditions or subsidiary provisions may result from this.
10.1.4 Environmental monitoring In order to ensure adherence to the conditions and subsidiary provisions to be imposed by the responsible authority in conjunction with approval of an outline operating plan incorporating EIA, continuous environmental monitoring in the area affected by mining and, where necessary, also beyond it, by the mining companies, etc., will be necessary. The purpose of this environmental monitoring is that of observing and registering relevant changes in the balance of nature during and
Deutsche Steinkohle AG Division for Engineering Survey 65 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
after mining. It must be possible to differentiate mining-induced environmental impacts from changes to the balance of nature caused by other factors, in order to ensure effective environmental monitoring. Ultimately, mining-induced impacts must be registered, evaluated and compared against the relevant forecasts. This environmental monitoring activity is supervised by a permanent work group headed by the Federal States mining authorities, in which all the authorities and other persons and bodies affected are represented.
Figure 10.2: Process of environmental planning and monitoring.
10.2 DSK monitoring concept To manage and to accomplish the monitoring process DSK has to develope a detailed thematic concept, to build up an optimized project management (e.g. integration of all partners, communication, workflow) and to develope an optimized data processing concept by use of GIS and remote sensing.
10.2.1 Detailed thematic concept The essential basis for environmental monitoring of hydrological and ecological parameters is, firstly, the registration of the geometrical changes on the surface (as described in chapter 2.8). In addition it is also necessary to observe the resultant changes in the water balance (groundwater and watercourses) and, ultimately, the impacts on the other entities affected (vegetation, animals, the soil, etc.).
10.2.1.1 Subsidence monitoring Reliable measurements of soil movement are necessary for checking, in particular, of the impact forecasts made, including, where necessary, updating of mathematical forecasting models. In urban areas, extremely precise registration of three-dimensional movements (subsidence and horizontal shifts) by means of measurements of defined points is also necessary. In rural areas, registration of subsidence of the surface is generally sufficient (BENECKE ET AL., 1999B).
Deutsche Steinkohle AG Division for Engineering Survey 66 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
It was firstly necessary to describe functional interactions before soil movements could be predictively calculated. A formula for the pre-calculation of subsidence was developed in 1934 (KEINHORST, 1934): m = seam thickness mined area S ( ) P m P a P z a = subs idence factor affected area z = time factor
The introduction of computer-based data processing brought with it the development of improved models for the calculation of soil movements using complex analytical methods. A computer-assisted pre-calculation method is used at DSK for forecasting of anticipated mining-induced soil movements (subsidence, horizontal movement, tilting and changes in lengths). The anticipated movements are calculated in this process taking account of mining geometry, seam thickness, depth of the mined area and specific factors for the plot against time and the critical angle between defined points in time. This model inevitably contains inaccuracies in calculation of the soil-movement elements, because - mining frequently deviates from planning; - the dynamics of the soil-movement process cannot be precisely modelled; - a simplified geological model of the overlying strata and the overburden is used for calculation, without taking detailed account of the tectonics and the actual sequence of geological rock strata.
If the data for actually completed mining activities is available, a post-calculation of soil movement using the same computation method can be performed. Individual modification of the calculation parameters is then possible using the known mining-area geometry and seam thickness combined with assessment against measurements, with the result that the accuracy of post-calculations is significantly greater than that of predictive calculations.
Natural, anthropogenic and technically induced movements of the surface may be superimposed on mining-induced soil movements. It is therefore important for any analysis of measured soil movements to know the causes of such non-mining-induced movements in as much detail as possible. The elevation-change components of such movements may range in order of magnitude from a few millimeters up to several meters.
10.2.1.2 Surface monitoring To obtain occurring topographic changes of affected areas aerophotogrammetric surveys are done at regular times. They facilitate 3D-measurements with high accuracy using analytical and digital photogrammetric plotters. Based on these measurements detailed digital terrain models (DTM) are generated using the ArcGIS TIN module (refer to chapter 9.4 and 9.5).
By combination of these DTMs with subsidence data the topographical changes can be modelled and predicted for the next two decades.
10.2.1.3 Hydrological monitoring Precise ground-water models (DGWM) were developed (refer to chapter 9.6) to derive parameters of the geohydrological situation. Spatial and temporal changes of the hydrological situation result from the amount of subsidence, local ground-water situation and actual land-use. To estimate changes of the dynamic system “subsidence-ground-water-vegetation”, the ground-water recharge rate and isobars of the ground-water table are calculated in detail. This is done for different times up to 2019 as a prediction.
Deutsche Steinkohle AG Division for Engineering Survey 67 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
Further the ecological and hydrological assessment considers the changes of the watercourse system, The longitudinal sections (refer to chapter 9.4) form the basis for the assessment of water flow changes resulting from underground extraction and for the planning of appropriate measures.
10.2.1.4 Biomonitoring
The aim of monitoring flowing waterways is to identify early on the possible effects of subsidence on abiotic or biotic conditions such as becoming muddy, a change in water quality, a change in the symbiosis and, if necessary, to introduce preventative measures in time. Furthermore, the monitoring also serves to record the ecological effects of hydro-structural measures to regulate drainage channels.
The monitoring of flowing waterways comprises a description of the sample points, an ecomorphological mapping, the recording of hydrological data and of the water and bank vegetation, a chemical-physical water analysis, the recording of the macrozoobenthos and of the fish fauna.
As the dragonfly fauna by flowing waterways cannot be adequately recorded using macrozoobenthos collections, an additional recording of dragonflies is undertaken by observation along some sections suitable as dragonfly habitats.
For the evaluation of the macrozoobenthos, calculations including the saprobes index are carried out to determine and illustrate the biocoenotic characteristic of a survey section independently from the species composition, number of species and number of individuals, since these can naturally vary so greatly that they may eclipse any changes to the biocoenotic structures caused by subsidence.
Within the context of these calculations, for example, the flow preference distribution, the habitat preference distribution and the feeding preference distribution, are calculated as percentages according to the known autecological requirements of the organisms found, and illustrated in graphs. The effects of changes caused by subsidence, such as an increase or decrease in flow rate, increased sedimentation or a loss of running waterway characteristics, can thus be shown to a large extent independently of species and individuals.
Additionally, in connection with the other eco-hydrology studies, at a number of sample points along the larger waterways the flow patterns close to the channel floor are determined according to the FST- hemispheres method. The flow patterns close to the channel floor are characteristic for a variety of waterway types ("sandy flowing waters", "gravelly flowing waters" etc.) and drainage situations (low tide, mean tide etc.). They definitively determine the composition of the biocoenosis of the watercourse. This makes them a particularly suitable aid in putting together the reproducible description of changes to the hydraulic conditions above ground, e.g. a change in the channel floor gradient caused by subsidence or in the drainage of a watercourse. Since the waterway biocoenosis reacts to such changes with a certain delay, this method makes it possible to forecast changes within the waterway biocoenosis and thus, if necessary, introduce preventative measures in good time.
All data obtained are collected in an electronic database which is directly linked to the GIS.
In addition to the detailed status recordings at the sample points, sections of waterway where considerable changes have been forecasted must be patrolled in order to record the extent of the areas undergoing change. This step is also supplemented by an analysis of the remote sensing data, as well as the present gradient conditions.
Deutsche Steinkohle AG Division for Engineering Survey 68 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
The monitoring of terrestrial areas takes place at 2- to 5-year intervals, using the same methods as those employed for the status recording in the environmental compatibility survey.
In this way, a core profile 1.0 – 1.5m deep is drawn in each monitoring area. The description and measurements of the soil profile are subdivided according to top soil, subsoil and substratum, thus taking into consideration differences occurring in the horizon sequence. The measurements record the site factors relevant to plant ecology, namely ground moisture, oxygen supply at root level and nutrient supply. The values are classified individually for each of the four factors in 5- to 7-level scales. The combination, resembling a telephone number, of the value levels of the site factors is referred to as the soil's eco-key. It describes the character of the present status of the soil at this site.
Further, the plants in the monitoring areas are recorded according to the Braun-Blanquet method. A database provides details of the site requirements of individual species concerning ground moisture, acidity level, oxygen and nutrient supply, also classified in 5- to 7-level scales. From this information, different combinations are used to determine the ecokey of the present plant cover. Whilst the ecokey of the soil has the character of a random sample as regards site location and time, the ecokey of the plant cover provides an integrating value over a longer period and a larger area. Not until the two partial ecokeys are put together do we have a broadly validated description of the sample site. For this, the ecokey of the plant cover is compared with that of the soil. This takes into account the causal interactions between soil and vegetation as parts of the site as a whole. The integrated ecokey obtained by comparing the partial ecokeys for a specific area is an indicator for the development of the site.
At the same time as the ecological assessments, a photogrammetric remote sensing of the assessment area is carried out. At a number of ground water measurement points, the ground water isobaths are measured and applied to the entire area using a ground water model and a digital terrain model derived from the remote sensing data.
During the course of the monitoring, the following successive steps are necessary: - Definition and setting up of the long-term areas, which will be composed as follows: - Long-term areas in the areas expected to be affected. - Long-term areas in suspected areas, for which no visible biotope changes have been forecasted, but where changes cannot be entirely ruled out, due to the subsidence or ground water data. - Control areas which represent the most important site conditions and biotope types of the affected areas, which themselves are however in no way influenced by the effects of subsidence. - Reference areas in which the effects of subsidence are already visible. - The control and reference areas make it possible to differentiate between changes due to subsidence and processes with other causes (e.g. acidic precipitation or climatic fluctuations), and provide valuable information for the further planning and implementation of the monitoring (advanced development of the reference areas). - In the zones surrounding the long-term areas, the biotope type mapping is updated to determine the extent of the areas undergoing change. This is facilitated and supplemented by an evaluation of the remote sensing data in the GIS. - In areas where clear biotope changes are identified, the fauna must also be monitored, if possible in the same year or, at the latest, in the following year.
Deutsche Steinkohle AG Division for Engineering Survey 69 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
This must include an investigation of the animal species groups which were identified in the affected areas within the environmental compatibility survey or which, with a high probability, were able to colonize the changed biotopes (e.g. amphibians in newly created wet biotopes).
10.3 Integrated use of classical, GIS and Remote Sensing methods To fulfil the extensive demands of the thematic monitoring concept DSK needs – besides an optimised workflow – an optimised data processing concept by use of GIS and remote sensing.
The basis for this optimised data processing are the „classical“ methods („go and see“, e.g. terrestrial sampling of plants and soils by experts), GIS methods and Remote Sensing methods.
The „classical“ methods used for the monitoring are the same used for the EIS. They are part of the DSK monitoring concept and part of the monitoring guidelines by the authorities. GIS methods have been developed to integrate these data and methods in the central geodatabase. Remote sensing data and methods have been analysed during several research projects by DSK (e.g. MINEO) for including thematic as well as geometric aspects.
Figure 10.3: DSK monitoring concept – basis for optimized data processing.
To reach the aim of optimized data processing on this basis DSK started a new internal R&D project (2002 – 2004) called “Integrated use of classical, GIS and Remote Sensing methods for a long-term monitoring of mining related effects”. Main objectives are the use of remote sensing data methods and development of extended ArcGIS methods/models for efficient monitoring support. Improvement for the monitoring database by integration of remote sensing data/methods includes to determine the best fitting remote sensing methods (e.g. airborne vs. Satellite, hyperspectral vs. multispectral), the definition of needed remote sensing algorithms (including the MINEO results concerning classification and vegetation stress) and the integration of these data and methods with ArcGIS and the central geodatabase ArcSDE/Oracle. Besides that the project will lead to an extended structure of the
Deutsche Steinkohle AG Division for Engineering Survey 70 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
geodatabase by developing methods for storing and analyzing spatial objects considering “time” as a forth dimension.
Deutsche Steinkohle AG Division for Engineering Survey 71 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
11 CONCLUSION, ASSESSMENT OF RESULTS 11.1 Assessment of results The results of this study a very encouraging. It cold be shown that HyMap data can be used to detect vegetation stress caused by water-logging occurrences as well as to map plant species, for perennial plants, here trees, also the live stage could be mapped with quite good accuracy.
The derivation of spectral features from vegetation spectra, like red edge wavelength position, the 1st and 2nd derivatives and other features described in chapter 7 enabled us to detect the subtle changes in the vegetation spectrum which are closely correlated with vegetation stress
However, remote sensing will always struggle to give perfect results in vegetation stress detection since the plants seem to withstand the changed conditions for long time and than react very fast showing decreased photosynthetic activity (PA), leaves loss and dieback. Therefore only the advanced stage of vegetation stress could be detected with satisfying results. The initial stage of vegetation stress could not be mapped with the same contentment. While detecting the vegetation stress, different factors have to be taken into account. One of them is, that not every species react in the same way to the changed hydrological situation. Plant species with increased PA may occur close to them with decreased PA which results in mixed pixels. These pixels represent an average spectral response over a given area, here 5 metres, which aggravates the image data interpretation, especially in areas with mixed forests and in initial stage of vegetation stress.
11.2 Results versus user demand The procedures and tools to detect vegetation stress as well as mapping methods of the tree species within the forest stands were developed and applied having the repeatability in mind. Although the developed tools for stress detection are based upon a well documented and tested algorithms (refer to chapter 7), their application with satisfying results is limited at this stage only to environments with similar environmental conditions. This can be found at the other mine sites under DSK’s management, so they are intended to be implemented into the monitoring concept. This will enhance the database and thus improve the EIA.
The developed workflow models will speed up the image data interpretation process and should be understood as the first step towards the operational status.
11.3 Future plans As already mentioned, the results achieved in this study and the developed methods are intended to be incorporated into the DSK monitoring concept. However this requires enhancements and improvements of these methods. As such not only the adaptation of these tools and procedures to new data sets from spaceborne sensors is meant, but also improvements of the tools in terms of application in different environmental conditions. This can be achieved by cross validation of this techniques at other test sites with vegetation stress hazards, not necessarily related to active mining sites.
Deutsche Steinkohle AG Division for Engineering Survey 72 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
REFERENCES
BACH, H.: Die Bestimmung hydrologischer und landwirtschaftlicher Oberflächenparameter aus hyperspektralen Fernerkundungsdaten, Münchener Geographische Abhandlungen, Reihe B, Band 21, 1995. BBERGG (1980): Bundesberggesetz vom 13. August 1980 (BGBI. I S. 1310), zuletzt geändert durch Gesetz vom 26. Januar 1998 (BGBI. I S. 164). BERK, A. ET AL: Modtran 4 users manual. Air Force Research Laboratory, Hanscom MA, 2000. BENECKE, N., BRANDT, S., FISCHER, C., SPRECKELS, V., VOSEN, P.: Überwachung der Tagesoberfläche im Gebiet des Steinkohlenbergbaus – Nutzung von GIS, Photogrammetrie und Fernerkundung. In: GIS, Heft 1, S. 34-39, 1999 (a). BENECKE, N., BUSCH, W., JACOBSEN, K., STREERATH, A.: Digitale Photogrammetrie und Geoinformationssysteme zur Erfassung und Analyse bergbaubadingter Bodenbewegungen. In: Wissenschaftliche Schriftreihe im Markscheidewesen, DMV e.V., 1999 (b). BENECKE, N.: Project final report :„Nutzung von hochauflösenden Satellitendaten zur großräumigen Überwachung der Umweltauswirkungen bergbaulicher Tätigkeiten im Ruhrgebiet.“; DSK internal RD project No. 50-EE-9652, DSK-DG, Bottrop, 2000. BOHNAM-CARTER, G.F.: Numerical procedures and computer program for fitting an inversed gaussian model to vegetation reflectance data. In: Computers & Geoscience, Vol. 14, Nr. 3, pp:339-356, 1988. BRUNN, A., BUSCH, W.: Hyperspectral Imagery in Mine Waste Management. In: Proceedings of the International Geoscience and remote sensing symposium, Sydney, IEEE, 2001. BRUNN, A., DITTMANN, C., FISCHER, C., RICHTER, R. : Atmospheric Correction of 2000 HyMap Data in the Framework of the EU-Project Mineo, Proceedings of the SPIE Workshop, Toulouse, September, 2001. BVOSCH: Bergverordnung über den Abbau unter Schiffahrtsstraßen (BVOSch) vom 20. Februar 1970. Amtsblatt Düsseldorf 1970 Nr. 17 – Sonderbeilage; 1970. CHAVEZ, JR. P.S.: Extracting Spectral Contrast in Landsat Thematic Mapper Image Data using Selective Principal Component Analysis. Photogrammetric Engineering and Remote Sensing, Vol. 55, No. 3, March 1989, pp. 339-348. CHRISTIANSEN, G. ET AL.: Monitoring of environmental changes caused by hard coal mining in the Ruhr district of Germany. In: Mander, Ü., e.a. (ed.): Conference Proceedings of the IALE European Conference 2001. Publicationes Instituti Geographici Universitatis Tartuensis, Bd. 92, pp. 270 – 274, Tartu, 2001. COCKS, T. ET AL.: The HyMap airborne sensor: the system, calibration and performance. 1st EARSEL Workshop on Imaging Spectrometry, Zürich 1998.
Deutsche Steinkohle AG Division for Engineering Survey 73 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
CURRAN, P. ET AL. : The effect of a red leaf pigment on the relationship between red edge and chlorophyl concentration. In: Remote Sensing of Environment, Vol.35, pp. 69-76, 1995. DEUTSCHE STEINKOHLE AG BERGWERK PROSPER-HANIEL: Erläuterungsbericht zum UVP-pflichtigen Rahmenbetriebsplan für den untertägigen Abbau bis Ende 2019. Herne, 1999. DROZDZEWSKI, G: Geologie des Münsterlandes. GLA-NRW, Krefeld, 1995. DWORSCHAK ET AL.: Key to the 1:5000 scale soil map of the woodland areas around TK25 4306 Hünxe; GLA-NRW, Krefeld, 1988. EIN-BV: Bergverordnung über Einwirkungsbereiche (Einwirkungsbereichs-Bergverordnung- EinwirkungsBergV) vom 11. November 1982 (BGBI. I S. 1558), 1982. ELLENBERG, H.: Zeigewerte für Pflanzen in Mitteleuropa. Göttingen 1992. EU (1985): Richtlinie des Rates über die Umweltverträglichkeitsprüfung bei bestimmten öffentlichen und privaten Projekten vom 27. Juni 1985 (ABI. EG Nr. L 175/40). FILELLA, I., PEÑUELAS, J. :The red edge position and shape as indicators of plant chlorophyll content, biomass and hydric status, Int. J. Remote Sensing, Vol.15, No. 7, pp.1459-1470, 1994. FISCHER, C.: Eignung hyperspektraler Fernerkundungsdaten zur Erfassung von Umweltauswirkungen des Steinkohlenbergbaus. Doctoral thesis, IGMC-TU Clausthal, Clausthal-Zellerfeld, 2001. FISCHER, C., BUSCH, W.: Monitoring of environmental changes caused by hard coal mining. Proceedings of the SPIE Workshop, Toulouse, September, 2001. FISCHER, C.: Use of GIS in multitemporal imaging spectrometer data for modelling and mapping environmental changes in mining areas. In: Proceedings of the ISPRS Commission IV Symposium “Geospatial Theory, Processing and Applications”, Vol. 34, Part 4, Commission IV, 8-12 July 2002, Ottawa, ISSN 1682-1750, pp.460-464. FRÖHLICH UND SPORBECK: Musteranforderungsprofil für Umweltverträglichkeitsprüfungen. Interne Studie im Auftrag der DSK. Essen, 1995. FÜCHTBAUER, H. (Ed): Sedimente und Sedimentgesteine. 4th Edition, Stuttgart, 1988. FUAN T., WILLIAM P.: Derivative Analysis of Hyperspectral Data for Detecting Spectral Features, IGARSS 1997. GANGKOFER, U. : Methodische Untersuchungen zur Vor- und Nachbereitung der Maximum Likelihood Klassifizierung optischer Fernerkundungsdaten, Münchener Geographische Abhandlungen, Reihe B, Band 24, 1996. GAO, B.-C., An Operational Method for estimating signal to noise ratios from data acquired with imaging spectrometer, Remote Sensing of Environment, Vol. 43, No. 1, pp. 23-33, 1993 GASSNER, E., WINKELBRANDT, A.: UVP – Umweltverträglichkeitsprüfung in der Praxis. REHM Verlag, 2. überarb. und erw. Auflg., München, 1992. GUYOT, G., BARET, F.: Utilisation de la haute resolution spectrale pour suivre l’etat des couverts vegetaux, In.: Proc. 4th Coll. On Spectral Signature of Objects in Remote Sensing, Aussois, France, ESA SP- 287, S. 279 – 286, 1988 GUYOT, G.; ET AL: Imaging spectroscopy for vegetation studies. In: Toselli & Bodechtel (Ed): Imaging spectroscopy: fundamentals and prospective application. Kluwe Academic Publishers, pp 145 – 165, Dordrecht, 1992.
Deutsche Steinkohle AG Division for Engineering Survey 74 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
HALLAWELL, J. M.: Development for a rational of monitoring. In: LIKENS, E. (Ed.): Monitoring for Observation and Ecology; Chapman & Hall, London, 1992. HANSEL, G.: Genehmigungsverfahren mit Umweltverträglichkeitsprüfung für den Steinkohlenbergbau. In: Das Markscheidewesen, Heft 1, S. 346-350, 1993. HENTRICH, S. und VOSEN, P.: Kohle und Wasser: Konzept und Methodik zur GIS-gestützten Prognose, Analyse und Überwachung des Einflusses des untertägigen Steinkohleabbaus auf Oberflächengewässer. In: Proceedings of the 14th ESRI European User Conference, München, 15.-17.11.1999. HILBIG, P.: Der Bergschaden als geodätisch-markscheiderisches Problem. In: Veröffentlichungen des Geodätentages in Essen, 1956. HORLER, BARBER, BARRINGER: Effects of Heavy metals on the absorbance and reflectance spectra of plants: Intern. Journal of Remote Sensing, v. 1, p. 121-136. ILS: Musteranforderungsprofil für die Umweltverträglichkeitsstudien. Internes Gutachten im Auftrag der DSK – Entwurf. Essen, 1995. ILS: Umweltverträglichkeitsstudie zum Steinkohlendabbau-Vorhaben – Kirchheller Heide / Hünxer Wald. Internes Gutachten im Auftrag der DSK, Essen, 1997. KLEINHORST, H.: Betrachtunge des Bergschädenfrage. In: Glückauf, S. 149-155; 1934. KLEINSCHMIDT, K.: Begriff, Anspruch und Bedeutung der Umweltverträglichkeitsprüfung (UVP). In: Mitteilung aus der NNA3, 1993. KNÖCHEL, H.: Die Umweltverträglichkeitsprüfung bei Vorhaben des untertägigen Steinkohlenbergbaus. In: Nordrhein-Westfälische Verwaltungsblätter (NWBI), 6. Jahrgang, Heft 4, S. 117-121, 1992. KRUSE, F.: Mapping spectral variability of geologic targets using Visible/Infrared Imaging Spectrometer (AVIRS) data and a combined spectral feature/unmixing approach, Proceedings of SPIE, 1995. LEHMANN, K.: Bewegungsvorgänge bei der Bildung von Pingen und Trögen. In: Glückauf, heft 20, 1919. MBVO : Verordnung über markscheiderische Arbeiten und Beobachtungen der Oberfläche (Markscheider-Bergverordnung – MarkschBergV) vom 19. Dezember 1986 (BGBI. I S. 2631). MESSER, J.: Impact of urbanisation on ground-water recharge, taking particular account of the Castrop upland area and the urban district of Herne. Doctoral thesis from Clalusthal Technical University; Clausthal-Zellerfeld, 1996. MINEO: MINEO Assessing and monitoring the environmental impact of mining activities in Europe using advanced Earth Observations techniques; Annex 1: Description of Work. Shared-cost RTD, Contract no.: IST-1999-10337. MINEO: MINEO Hyperspectral Airborne Survey. Survey Report. http://www.brgm.fr/mineo; 2000. PFAFF-SCHLEY, H.: Die Umweltverträglichkeitsprüfung als Planungsinstrument. Planungs-UVP, Anlagen-UVP. Eberhard Blöttner Verlag, 1992.
Deutsche Steinkohle AG Division for Engineering Survey 75 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
PHILPOT, W.D.; The derivative Ratio algorithm: Avoiding Atmospheric effectas in Remote Sensing,“ IEEE Transactions in Geoscience and Remote Sensing, 29(3):350-357, 1991.RENCZ, A. N. (Ed. 1999): Remote Sensing for the Earth Sciences. In: Manual of remote sensing, 3rd ed., vol. 3, -Wiley & Sons, New York, 1999. RENCZ, A. BONHAM-CARTER, G.F., VAN DER GRIENT, C., MILLER, J.R., HARE, F.W., Preliminary results from modelling vegetation spectra derived from MEIS data, Algonnquin Park, Ontario: Proc. 10th Canadian Symposium on Remote Sensing, Edmonton, May, 1986. RENCZ, A. (EDITOR): Remote Sensing for the Earth Sciences. In: Manual of Remote Sensing, 3rd ed., Vol. 3, Wiley & Sons, New York, 1999. RHINELAND CHAMBER OF AGRICULTURE: Agrarstrukturgutachten für den rheinischen Bereich des Steinkohlenbergabus; Bonn, 1993. RICHTER, R.: Atmospheric/Topographic Correction for Wide FOV Airborne Imagery:Model ATCOR4 (Version 2.0 October 2000), report DLR-IB 564-04/00, Wessling, 2000. RICHTER, R. AND SCHLÄPFER, D.: A Unified Approach to Parametric Geocoding and Atmospheric / Topographic Correction for Wide FOV Airborne Imagery; Part 2: Atmospheric Correction. In Proc.2nd EARSeL Workshop on Imaging Spectrometry, Enschede, 2000. ROTHFUSS, H.: VErarbeitung und Einsatz Abbildender Spektrometerdaten (GER) mit unterstützenden Bodenmessungen zur Erkundung einer landwirtschaftlich genutzten Fläche. Fortschrittsbericht Reihe 15, Nr. 132. VDI, VDI-Verlag, Düsseldorf, 1994. RSI, Research Systems Inc.: ENVI Users’s Guide, Version 3.4 and IDL Reference Guide, Version 5.4. Research Systems Inc. 2001, Boulder, USA. RÜBER, O. : Grundwasserströmungsmodell Kirchheller Heide; - Simulation des Einflusses von Bergsenkungen auf den Grundwasserhaushalt im Bereich der Kirchheller Heide anhand eines 3-dimensionalen Strömungsmodells. Gutachten im Auftrag der Deutschen Steinkohle AG; GKW-Ingenieurgesellschaft mbH Bochum, 1998. RUND: Geimeinsamer Runderlaß des Ministeers für Arbeit, Gesundheit und Soziales – III A 7 – 8603.4 – (III Nr. 2/83), des Ministers für Ernährung, Landwirtschaft und Forsten – III C 7 – 8300/2 und des Ministers für Wirtschaft, Mittelstand und Verkehr III/A3 – 40 – 01 (65/82) vom 11.02.1983, Fernleitung zum Transport gefährdender Stoffe – Überwachung der Fernleitung im Einwirkungsbereich des Bergbaus M. BI. NW – Nr. 28; 1983. SABINS, F.: Remote Sensing. Principles and interpretation. New York, 1997. SALEWSKI, E. M.: Die Möglichkeiten markscheiderischer Aussagen über Bodenbewegungen und Lagerstättenverhältnisse als Grundlage für eine bergmännische Planung unter Berücksichtigung der Belange des Landschafts- und naturschutzes. Dissertation, TU Clausthal, 1991. SAVITZKY, AND GOLAY,: SMOOTHING and differentiatiation of data by simplified least square procedure. In: Analytical Chemistry, Vol. 36, S. 1627-1638, 1964 SCHLAEPFER, D.: PARametric GEocoding, User Guide V. 1.3.4.; ReSe Applications & RSL University of Zürich, Zürich, 2001.
Deutsche Steinkohle AG Division for Engineering Survey 76 & Geoinformation Services
IST – 1999 - 10337 August 2002
MINEO Central European Contamination/impact mapping and environment test site in Germany modelling – Final report
SCHOWENGERDT, R.: Remote Sensing: Models and Methods for Image Processing, Second Edition, Academic Press, San Diego, 1997. SINGHROY, V.H. AND KRUSE, F.: Detection of metal stress in boreal forest species using the 0.67µm chlorophyll absorption band. In: Proc. 8th thematic conference on Geologic Remote Sensing, Exploration, Engineering and Environment, Denver, 29 April – 2 May 1991, pp 361 – 372, 1991 STATISTIK DER KOHLENWRITSCHAFT E.V.: Zur Lage des Kohlenbergbaus in der Bundesrepublik Deutschalnd, 1. Vierteljahr 2002; Essen/Köln, 2002. STRASSBURGER, E.: Lehrbuch der Botanik. 32. Aufl. Gustav Fischer Verlag, Stuttgart, 1991. TALSKY, G.: Derivative Spectrophotometry, Weinheim, 1994. TOWNSHEND, J., JUSTICE, C., LI, W., GUERNEY, C., AND MCMANUS, J.: Global Land cover classification by remote sensing: present capabilities and future possibilities, Remote Sensing of Environment, 35, 243-255, 1991 TOWNSHEND, J., JUSTICE, C., GUERNEY, C., AND MCMANUS, J.: The impact of Misregistration on Change Detection, In: IEEE Transactions on Geoscience and remote sensing, Vol. 30, No. 5, September 1992 UVP-G (1990): Gesetz über die Umweltverträglichkeitsprüfung (UVP-G) vom 12. Februar 1990 (BGBI. I S. 205), zuletzt geändert durch Gesetz von 18. August 1997 (BGBI. I S. 2081). VOSEN, P.:Methodisch-konzeptionelles Vorgehen zur GIS-gestützten Bearbeitung von Umweltverträglichkeitsstudien im Steinkohlenbergbau. In: Wissenschaftliche Schriftenreihe im Markscheidewesen, DMV e. V., Heft 17, S. 95-104, 1997. WALTER, R.: Geologie von Mitteleuropa, 6th edition, Stuttgart, 1995. Wycisk, P.: Die Umweltverträglichkeitsprüfung (UVP) – konzeptioneller Rahmen einer vorsorgenden Umweltgeologie: Beispiel Deponistandortsuche. In: Zeitschrift der Deutschen Geologischen Gesellschaft, S. 308-318. Hannover, 1993. YUAN, D., ELVIDGE, C., LUNETTA, R.: Survey of multispectral methods for land cover change analysis, In: Lunetta, Elvidge, Remote Sensing change detection, Ann Arbor Press, Chelsea, 1998
Deutsche Steinkohle AG Division for Engineering Survey 77 & Geoinformation Services
Appendix I: Land use & cover map
Appendix Ia: Land use & cover map, subarea 1
Appendix Ib: Land use & cover map, subarea 2
Appendix II: Vegetation stress map
Appendix III: Plant stress change detection map on pine trees; August 1998 to August 2000
Appendix IV a: Workflow procedure for HyMap data pre-processing
Step 1: Geometric correction (KH_GEOC) with parametric approach using ParGe
KH_GEOC_1: Input data - hyperspectral image data (radiance data) - DTM (spatial resolution shall be better than the image data) - IMU–data
KH_GEOC_2: GCP definition
KH_GEOC_3: Geometric correction, first run
KH_GEOC_3a: Parameter adjustment
KH_GEOC_4: Final geocorrection - Output: - geocoded image data - scan angles file (*.sca-file)
KH_GEOC_5: Results assessment
Step 2: Atmospheric correction (KH_ATMC) using ATCOR4
KH_ATMC_1: Input data - geocoded hyperspectral image data (radiance data) - atmospherical data: - date of image acquisition - atmosphere profile (e.g. from radio tube ascent) - visibility - spectral reference data from field spectral campaign - scan angles file from ParGe (*.sca-file) - DEM / DTM - elevation - slope - aspect
KH_ATMC_2: Building of atmospheric look up tables for the specified sensor - input of atmospheric data and building ATM-LUTs with ATLUT module of ATCOR4
KH_ATMC_3: Inflight calibration with IFCALI module of ATCOR4 - input of ATM-LUT - input of atmospheric data from the day of image acquisition - generation of new calibration file using the ground reference spectra - parameter adjustment - inflight calibration.
KH_ATMC_4: Atmospheric correction with ATCOR module of ATCOR4 - Output: geocorrected reflectance image data
Appendix IV b: Workflow procedure developed for vegetation stress detection
Vegetation Stress in the Kirchheller Heide coal mining area (KH_VEGS)
KH_VEGS_1: Input data 1 - HyMap reflectance
KH_VEGS_ 2: Enhancement of HyMap data for visual estimation of extension vegetation stress effects - Vegetation indices calculation (NDVI, NDWI,...)
KH_VEGS_ 3: Input data 2 - GIS-data (Biotope map, GW level, changes of GW level, Subsidence data, DTM,... ): - Definition stressors responsible for vegetation stress, e.g. water-logging occurrences (knowledge based) - estimation of stressed vegetation types (sensitivity of a given species to a given sources of stress, their sensitivity gradient depending on the age of the given individuals or stands; knowledge based), - spatial distribution of possible stressed vegetation (e.g.: GIS query for vegetation stress in forest stands caused by changes of the GW level). - Due to the high variability of the vegetation, a refinement may be needed: e.g. separate GIS query for each sensitive deciduous and conifer tree species.
KH_VEGS_ 4: Input data 3 - Spectra of stressed vegetation - if no ground reference available: input of vegetation spectra from HS image (sensitive species):
KHVEGS_ 4a: Collecting of spectra (endmember definition) - HS image masking with single tree species stands applying the appropriate GIS data (e.g. Biotope types map). The GIS data may be buffered introvert to avoid extremely mixed pixels at the area edges; - MNF transformation, data compression; - n-Dimensional classification for each vegetation type in the given age stage to determine spectral differences within the vegetation type;
- storing new classes as new nD-ROI’s; - statistical analysis on the nD-ROI’s; - provisional Endmembers for stressed and not stressed vegetation (photosynthetic active vegetation, PAV), not photosynthetic active vegetation, NPAV (e.g. Wood), shade.
KHVEGS_ 4b: Endmember definition for stressed and not stressed vegetation - MTMF computation applying the provisional Endmembers - MTMF results enhancement (rationing) for estimation of spatial distribution of stressed vegetation. - definition of final Endmembers by analysis the MTMF results.
KH_VEGS_ 5: Enhancement of Spectra for estimation of vegetation stress effects - analysis of vegetation spectra features (red edge position, parameters of absorption, - reflectance features (e.g. depth, symmetry, FWHM), - calculating of derivations, ...
KH_VEGS_ 6: Spatial characterisation of stressed areas
KH_VEGS_ 7: Enhancement of stress characterisations to less sensitive vegetation species (ref. to KHVEGS_2 to KHVEGS_6)
KH_VEGS_ 8: Verification of the results
Appendix IV c: Workflow procedure developed for change detection
Vegetation Stress change detection in the Kirchheller Heide coal mining area (KH_CH_DET_VEGS)
KH_CH_DET_VEGS_1: Input data 1 - hyperspectral data from first date (here 2000)
KH_CH_DET_VEGS_2: Geocorrection of the image data
KH_CH_DET_VEGS_3: Input data 2 - hyperspectral data from second date (here 1998)
KH_CH_DET_VEGS_4: Geocorrection of the image data - the image data have to be corrected with high accuracy and stored in the same coordinate system as the first data set.
KH_CH_DET_VEGS_5: Visual interpretation - enhancement of HyMap data by calculation of PCA, MNF, vegetation indices, spectra feature parameters,...
KH_CH_DET_VEGS_6: Pre-classification change detection - computation of Normalised Difference Image using the first Principal Component (NDI_PC1) from the two input images - colour coding for visual enhancement of change affected areas
KH_CH_DET_VEGS_7: Post-classification change detection - rationing of the spectra feature parameters images (e.g. 2nd derivative images) calculated from the two input images.
KH_CH_DET_VEGS_8: Verification of the results
Appendix IV d: Workflow procedure developed for DTM generation
Generation of a high-accurate Digital Terrain Model (DTM) of the Kirchheller Heide coal mining area (KH_GIS_MAKE_DTM) using ArcGIS
KH_GIS_MAKE_DTM_1: Input data 1
- Aerophotogrammetric flight (scale 1:4000 to 1:6000)
- High accurate scan of aerial photos (14 – 21 µm)
- Automatic measurement of a regular grid (ground resolution 5 meters) using digital photogrammetry
- Manual editing of the regular grid – deleting unusable points, measuring additional points
KH_GIS_MAKE_DTM_2: Generating DTM stage 1
- Use of ArcGIS TIN to build a triangular network
- Converting TIN to a regular lattice using ArcGIS Grid
KH_GIS_MAKE_DTM_3: Input data 2 - Refining the DTM by incorporating line elements (breaklines): - by analytical photogrammetric measurements (e.g. artifical breaklines, watercourse axis, flood-relevant slope tops, lakes) - by integration of terrestrial measurements (e.g. GPS, tachymetric surveying) - by integration of analogue map data (longidutinal sections, site plans)
KH_GIS_MAKE_DTM_4: Generating DTM stage 2 - Use of ArcGIS TIN to build a triangular network representing the morphology of the terrain surface at the date of fly - Converting TIN to a regular lattice using ArcGIS Grid - Construction of isolines and elevation zones
KH_GIS_MAKE_DTM_5: Watercourse network - Extraction of a three-dimensional watercourse network from the DTM-TIN - Use of ArcGIS Dynamic Segmentation to build a stationed three-dimensional route system representing the water flow status at the date of fly - Generating event tables labelled with descriptions of the watercourse
KH_GIS_MAKE_DTM_6: Longitudinal sections - Creation of longitudinal sections including floor and terrain slope edges using ArcGIS
KH_GIS_MAKE_DTM_7: Thematic watercourse maps - Joining the route system with ecological and hydrological data - Creation of high quality thematic watercourse maps using ArcGIS (e.g. ecological status maps)
Appendix IV e: Workflow procedure developed for DTM update
Generation of forecasted high-accurate Digital Terrain Models of the Kirchheller Heide coal mining area (KH_GIS_UPDATE_DTM) using ArcGIS
KH_GIS_UPDATE_DTM_1: Input data 1 - DTM-TIN stage 2 representing the morphology of the terrain surface at the date of fly - calculated subsidences for different time periods until the end of coal extraction (e.g. 2020) - watercourse developing plans (e.g. deepening, displacement)
KH_GIS_UPDATE_DTM_2: Update the DTM and watercourse network - Lowering the DTM by combination of DTM-TIN stage 2 and subsidences - Retaining all linear structures (breaklines) - Update DTM and watercourse network by incorporating planned watercourse changes due to developing plans - Converting TIN to a regular lattice using ArcGIS Grid - Construction of isolines and elevation zones
KH_GIS_UPDATE_DTM_3: Longitudinal sections - Creation of longitudinal sections containing floor and terrain slope edges for several time periods using ArcGIS
KH_GIS_UPDATE_DTM_4: Thematic watercourse maps - Joining the route system with ecological and hydrological data - Creation of high quality thematic watercourse maps using ArcGIS (e.g. ecological forecast maps)
Appendix IV f: Workflow procedure developed for isobath calculation
Generation of isobath lines of the Kirchheller Heide coal mining area (KH_GIS_ISOBATH) using SPRING and ArcGIS
KH_GIS_ISOBATH_1: Input data 1 - Geological and hydrogeological data - Digital Terrain Model - Land use data - Precipitation rate data - Water course system data - Ground water recharge data - Data from ground water measuring stations
KH_GIS_ISOBATH_2: Generation digital ground water model (GWM) - generation of a three dimensional, stationary ground water current model using the finite elements method (SPRING modeling software)
KH_GIS_ISOBATH_3: Generation of isobath lines - Combination of digital terrain model and digital ground water model using ArcGIS TIN and Grid - Generation of isobath lines - Creation of high quality thematic ground water maps using ArcGIS (e.g. isobath status maps)
KH_GIS_ISOBATH_4: Generation of forecasted isobath lines - Lowering the DTM and GWM by combination with subsidences - Combination of lowered digital terrain model and digital ground water model using ArcGIS TIN and Grid - Generation of forecasted isobath lines - Creation of high quality thematic ground water maps using ArcGIS (e.g. isobath forecast maps)
Appendix V: List of persons working over the test site
Peter Vosen Deutsche Steinkohle AG (DSK) Division for Engineering Survey and Geoinformation Services (DIG) Head of GIS Team (DIG 3.2) MINEO project leader Responsibilities: MINEO project management GIS data management
Christoph Dittmann Deutsche Steinkohle AG (DSK) Division for Engineering Survey and Geoinformation Services (DIG) Photogrammetry / Remote Sensing (DIG 2.2) GIS and RS data analyst Responsibilities: remote sensing data analysis, GIS analysis, MINEO project co-management
Andreas Brunn Technichal University of Clausthal Institute of Geotechnical Engineering and Mine Surveying Responsibilities: scientific consulting, hyperspectral data analysis, IT tool development
Christian Fischer Technical University of Clausthal (TUC) Institute of Geotechnical Engineering and Mine Surveying (IGMC) Responsibilities: scientific consulting, hyperspectral data analysis
Hartmut Mollat, German Federal Institute for Geosciences and Natural Resources Uwe Schäfer, (BGR) Kurt Oppermann Responsibilities: spectral field campaign, development and maintenance of MSL
Martin Schodlock Geo-Research-Centre Potsdam (GFZ-Potsdam) Responsibilities: spectral field campaign
Rudolf Richter German Aerospace Center (DLR) Responsibilities: consulting services for HyMap data pre-processing (atmospheric correction)