**5.1.1 Lake inventory**

Delineation of water bodies is essential for the estimation of the water balance of the area. Water authorities need to know date, location, extent and variations of these water bodies. The test area covers a broad region while the transnational Prespa lakes basin is included. The problems that are faced are related to:


The 17th of January 2011 Landsat image has been used to make an inventory of all the lakes of the region at a scale of ~ 1:50000, Figure 3. The lake water surfaces have been extracted using classification of infrared bands & conversion of raster to vector techniques. There is up to date information which is readily available in a digital format for the whole of the translational

Monitoring Lake Ecosystems Using Integrated Remote

perform multi-temporal studies of lake surfaces.

North (A), South East ( B)and South West (C)

nearly 19.8 km² of its surface in the period 1973 to 2011.

Sensing / Gis Techniques: An Assessment in the Region of West Macedonia, Greece 193

countries. Water level also does not show the spatial variability of the water surfaces, as changes depend on the bathymetry, the amount of sediment input due to erosion or other factors like geomorphology / geology. Satellite and especially Landsat data can be used to

Data collection included the acquisition of lake coastlines as these are available by the national / local authorities or on the Web. The only readily available data for Vegoritis lake are those of maps provided by the Greek Geographic Service of the Army of 1970s while the boundary of Macro Prespa lake has been made available for a time period on the Web (Traborema EU project). The stored in the GIS database map coastlines have been used to assess changes in water surfaces. These coastlines have the same areal extend as these extracted from the Landsat MSS images of the 1974 and therefore are used as a reference. These lake surfaces / coastlines dated since the 70's have been compared to the ones

extracted from the multi-temporal Landsat images and stored as GIS vector layers.

Fig. 9. Incremental changes of Macro Prespa lake for the last ~ 30 years: Changes in the

Both Vegoritis and Macro Prespa lakes have lost their water surface area. A reduction of the surface area of Macro Prespa lake is evident, as estimates of its surface are as following: 20 November 1974 - ~276.5 km², August 1988 ~ 273.7 km², August 2000~265.2 km², 21 August 2008 ~257.2 km² and 17 January 2011~ 256.7 km². Macro Prespa lake has lost

Sharp drop of water level of Macro Prespa lake occured in 1975/1977 (1.2 m), 1987 /1990 (3.7m) and 2000/2002 (2.2m.) Figure 9. It is further evident that Macro Prespa lake is still losing its surface, even though the entire Prespa basin has been declared as a transboundary protected area, with the establishment of the "Prespa Park" by the Prime

Vegoritis lake has lost 30% of its surface (1970: 59.7 km² – 2011: 43.8 km²) in the last ~ 30 years. Changes on its coastline are observed in its southern part, Figure 10. This can be partly explained by its bathymetry as the waters are shallow in the southern part, while its deepest area is in its western part, Figure 5. Comparison with the multitemporal analysis of the other lakes of the area shows that Ohrid, Micro Prespa and Petron lakes have lost only a

Ministers of Albania, Greece and the FYR of Macedonia on 2 February 2000.

region. Extraction of surface areas / perimeter and spatial context of the location of the lakes is easily obtained. Relationships of the different lake water bodies are also obtained, Figure 7. Accurate mapping of surfaces of the Greek lakes in scales up to ~5000 (Figure 8) is obtained using the WMS - Web service of the Hellenic Cadastre, http://gis.ktimanet.gr/wms/ktbasemap/default.aspx . The acquisition dates of the aerial photography are in the time period of 2007 to 2009.

Fig. 7. Lake inventory from the 17th January 2011 image scene. Polygons of the water surface of the lakes have been extracted using classification techniques: A. Ohrid lake B. Macro / Micro Prespa lakes C. Vegoritis / Petron Lake D. kastoria Lake E. Chimaditis / Zazari lakes

Fig. 8. Overlay of the coastlines extracted from the 17th January 2011 image to the orthophoto of 0.5 m resolution A. south part of Vegoritis lake B. South part of Macro Prespa lake.

#### **5.1.2 Multitemporal analysis of change in surface area / size / shape of lakes**

Lakes are sensitive to both climate change and to anthropogenic influence. Drop of water level has been observed in both Macro Prespa and Vegoritis lakes, Figure 2. Time series water level data are available for both lakes even though these measurements are not comparable for Macro Prespa lake as different reference levels are used between the three

region. Extraction of surface areas / perimeter and spatial context of the location of the lakes is easily obtained. Relationships of the different lake water bodies are also obtained, Figure 7. Accurate mapping of surfaces of the Greek lakes in scales up to ~5000 (Figure 8) is obtained using the WMS - Web service of the Hellenic Cadastre, http://gis.ktimanet.gr/wms/ktbasemap/default.aspx . The acquisition dates of the aerial

Fig. 7. Lake inventory from the 17th January 2011 image scene. Polygons of the water surface of the lakes have been extracted using classification techniques: A. Ohrid lake B. Macro / Micro Prespa lakes C. Vegoritis / Petron Lake D. kastoria Lake E. Chimaditis / Zazari lakes

A B

orthophoto of 0.5 m resolution A. south part of Vegoritis lake B. South part of Macro Prespa

Lakes are sensitive to both climate change and to anthropogenic influence. Drop of water level has been observed in both Macro Prespa and Vegoritis lakes, Figure 2. Time series water level data are available for both lakes even though these measurements are not comparable for Macro Prespa lake as different reference levels are used between the three

Fig. 8. Overlay of the coastlines extracted from the 17th January 2011 image to the

**5.1.2 Multitemporal analysis of change in surface area / size / shape of lakes** 

lake.

photography are in the time period of 2007 to 2009.

countries. Water level also does not show the spatial variability of the water surfaces, as changes depend on the bathymetry, the amount of sediment input due to erosion or other factors like geomorphology / geology. Satellite and especially Landsat data can be used to perform multi-temporal studies of lake surfaces.

Data collection included the acquisition of lake coastlines as these are available by the national / local authorities or on the Web. The only readily available data for Vegoritis lake are those of maps provided by the Greek Geographic Service of the Army of 1970s while the boundary of Macro Prespa lake has been made available for a time period on the Web (Traborema EU project). The stored in the GIS database map coastlines have been used to assess changes in water surfaces. These coastlines have the same areal extend as these extracted from the Landsat MSS images of the 1974 and therefore are used as a reference. These lake surfaces / coastlines dated since the 70's have been compared to the ones extracted from the multi-temporal Landsat images and stored as GIS vector layers.

Fig. 9. Incremental changes of Macro Prespa lake for the last ~ 30 years: Changes in the North (A), South East ( B)and South West (C)

Both Vegoritis and Macro Prespa lakes have lost their water surface area. A reduction of the surface area of Macro Prespa lake is evident, as estimates of its surface are as following: 20 November 1974 - ~276.5 km², August 1988 ~ 273.7 km², August 2000~265.2 km², 21 August 2008 ~257.2 km² and 17 January 2011~ 256.7 km². Macro Prespa lake has lost nearly 19.8 km² of its surface in the period 1973 to 2011.

Sharp drop of water level of Macro Prespa lake occured in 1975/1977 (1.2 m), 1987 /1990 (3.7m) and 2000/2002 (2.2m.) Figure 9. It is further evident that Macro Prespa lake is still losing its surface, even though the entire Prespa basin has been declared as a transboundary protected area, with the establishment of the "Prespa Park" by the Prime Ministers of Albania, Greece and the FYR of Macedonia on 2 February 2000.

Vegoritis lake has lost 30% of its surface (1970: 59.7 km² – 2011: 43.8 km²) in the last ~ 30 years. Changes on its coastline are observed in its southern part, Figure 10. This can be partly explained by its bathymetry as the waters are shallow in the southern part, while its deepest area is in its western part, Figure 5. Comparison with the multitemporal analysis of the other lakes of the area shows that Ohrid, Micro Prespa and Petron lakes have lost only a

Monitoring Lake Ecosystems Using Integrated Remote

**(A)** In situ (red line and dots) and altimetric (blue line and dots) water level time series of Ohrid lake. Though absolute values differ for Y axis, vertical scales are identical for both.

lakes.

data.

Sensing / Gis Techniques: An Assessment in the Region of West Macedonia, Greece 195

to land influence in altimetric signal, but in general both in situ and altimetric observation are in good agreement, Figure 11. Time series water level measurements can be obtained through the process of radar altimetry and if it is combined with the estimated surface areas, lake bathymetry can give an indication of the quantitative characteristics of the

> **(B)** In situ (blue line and small dots for Stenje station and dark blue line and open circles for Nakolez station) and altimetric TPNO (red line with open circles) water level time series for Macro Prespa lake. Though absolute values differ for Y axis, vertical

scales are identical.

Fig. 11. Results of applying test for estimating water level of lakes from radar altimetry

Fig. 12. Seasonal changes of Macro Prespa lake shown on the Landsat images of the year 2010: A. 14 / November B. 4 / April C.7 / June & D./ E./F. 2 / 18 / 26 of August respectively.

small part of their surface area. Analysis of the space imagery of the years 1975 and 2011 respectively clearly revealed areas of shore line changes. It is now possible to draw accurate maps which look at the future incremental changes of Vegoritis / Prespa lakes. The modeling of this process is efficiently performed in the GIS.

Fig. 10. Changes of the Vegoritis lake surface area.

In the framework of the assessment of remote sensing techniques a small scale experiment has been carried out using radar altimetry techniques by Alexei Kouraev (Stefouli et al 2008). Results show that there are annual variations of Ohrid lake water level and these can be measured using radar altimetry. As Macro Prespa lake is hydraulically connected to Ohrid lake and located in higher altitude these could explain its drop of the water level. For some ENVISAT cycles estimates of water level have not been made due to quality control. The difference between the two time series can be up to 15-20 cm, apparently due

small part of their surface area. Analysis of the space imagery of the years 1975 and 2011 respectively clearly revealed areas of shore line changes. It is now possible to draw accurate maps which look at the future incremental changes of Vegoritis / Prespa lakes. The

A. & B.

C.

D.

In the framework of the assessment of remote sensing techniques a small scale experiment has been carried out using radar altimetry techniques by Alexei Kouraev (Stefouli et al 2008). Results show that there are annual variations of Ohrid lake water level and these can be measured using radar altimetry. As Macro Prespa lake is hydraulically connected to Ohrid lake and located in higher altitude these could explain its drop of the water level. For some ENVISAT cycles estimates of water level have not been made due to quality control. The difference between the two time series can be up to 15-20 cm, apparently due

The coastlines of 1988 (blue line) as well as that of 2011 (white line) have been plotted on the Landsat 1984 image scene.

South part of Vegoritis lake: The black line shows the coastline of the map. The transparent polygon of the lake surface has been outlined from the17th January 2011 Landsat image.

The 1988 coastline plotted on the orthophoto: Estimates of the land use change of the lake to agricultural land can be obtained

and used by authorities.

modeling of this process is efficiently performed in the GIS.

Fig. 10. Changes of the Vegoritis lake surface area.

to land influence in altimetric signal, but in general both in situ and altimetric observation are in good agreement, Figure 11. Time series water level measurements can be obtained through the process of radar altimetry and if it is combined with the estimated surface areas, lake bathymetry can give an indication of the quantitative characteristics of the lakes.

Fig. 11. Results of applying test for estimating water level of lakes from radar altimetry data.

Fig. 12. Seasonal changes of Macro Prespa lake shown on the Landsat images of the year 2010: A. 14 / November B. 4 / April C.7 / June & D./ E./F. 2 / 18 / 26 of August respectively.

Monitoring Lake Ecosystems Using Integrated Remote

Sensing / Gis Techniques: An Assessment in the Region of West Macedonia, Greece 197

Fig. 14. Synthetic map concerning coastal sediment concentrations surface currents in the

The Landsat and ASTER data have been analyzed for estimating differences of suspended sediment content in Vegoritis lake. The data of band 2 of Landsat images (A, B in Figure 15) and band 1 of the ASTER image (C in Figure 15) have been used in the analysis as they correspond to the same spectral region of 0.52-0.60 μm. The same color palette has been used for displaying the multi-temporal images. Blue-green colors show relatively low sediment content while red - yellow colours high content. The Vegoritis lake thermal regime is also displayed in Figure 15. Inflow patterns of sediments can be interpreted on the satellite imagery in the different acquisition dates. Numbers 1 to 4 show the location of the

form of gyres.

streams / canals that discharge into the lake.

Fig. 13. Surface currents as mapped on using visible part of the spectrum (left image) and the thermal bands (right image) of the summer Landsat images for the period 1988 – 2010.

The images have shown that wind-driven partial upwelling events occur at least throughout the summer stratified period, transporting water from intermediate depths to the surface. These are important events that contribute to the patchiness and heterogeneity that characterize natural aquatic systems. The circulation in Lake Prespa is typically dominated by the northern two-gyre pattern, especially in the summer. The north wind leads to a cyclone (a counterclockwise rotation gyre) in the southwest and an anticyclone (a clockwise rotation gyre) in the northeast.

Analysis shows that a well formed system of gyres is formed during summer D,E,F of the year 2010 while this is not apparent in other seasons of the year i.e. winter / spring or autumn. These results have also been confirmed from the lake surfaces extracted from the ~ 30 years time span. Inter annual changes of the surface currents have been also evaluated. Circular features have been mapped in summer season of every year while some results are shown in Figure 13. These prominent features have been identified in most of the Landsat images. Self organization techniques classification techniques of the visible part of the spectrum proved to be quite effective in mapping lake circulation patterns. Multitemporal data are stored in the GIS database, while synthetic maps can be produced, Figure 14.

 Fig. 13. Surface currents as mapped on using visible part of the spectrum (left image) and the thermal bands (right image) of the summer Landsat images for the period 1988 – 2010. The images have shown that wind-driven partial upwelling events occur at least throughout the summer stratified period, transporting water from intermediate depths to the surface. These are important events that contribute to the patchiness and heterogeneity that characterize natural aquatic systems. The circulation in Lake Prespa is typically dominated by the northern two-gyre pattern, especially in the summer. The north wind leads to a cyclone (a counterclockwise rotation gyre) in the southwest and an anticyclone (a clockwise

Analysis shows that a well formed system of gyres is formed during summer D,E,F of the year 2010 while this is not apparent in other seasons of the year i.e. winter / spring or autumn. These results have also been confirmed from the lake surfaces extracted from the ~ 30 years time span. Inter annual changes of the surface currents have been also evaluated. Circular features have been mapped in summer season of every year while some results are shown in Figure 13. These prominent features have been identified in most of the Landsat images. Self organization techniques classification techniques of the visible part of the spectrum proved to be quite effective in mapping lake circulation patterns. Multitemporal data are stored in the GIS database, while synthetic maps can be produced, Figure 14.

rotation gyre) in the northeast.

Fig. 14. Synthetic map concerning coastal sediment concentrations surface currents in the form of gyres.

The Landsat and ASTER data have been analyzed for estimating differences of suspended sediment content in Vegoritis lake. The data of band 2 of Landsat images (A, B in Figure 15) and band 1 of the ASTER image (C in Figure 15) have been used in the analysis as they correspond to the same spectral region of 0.52-0.60 μm. The same color palette has been used for displaying the multi-temporal images. Blue-green colors show relatively low sediment content while red - yellow colours high content. The Vegoritis lake thermal regime is also displayed in Figure 15. Inflow patterns of sediments can be interpreted on the satellite imagery in the different acquisition dates. Numbers 1 to 4 show the location of the streams / canals that discharge into the lake.

Monitoring Lake Ecosystems Using Integrated Remote

(b) Depth 0,5m

(b) Depth 5m

(a) Depth 0,5m

various dates of the year 2000

**5.1.3 Suspended sediments – chlorophyll** 

Dates

Sensing / Gis Techniques: An Assessment in the Region of West Macedonia, Greece 199

(c ) Depth 0,5m

21-03-2000 1,0 2,1 2,6 2,6 2,1 4-04.2000 2,20 2,80 2,60 2,5 16-04-2000 2,40 2,60 2,60 2,5 7-05-2000 0,50 1,00 2,20 2,40 1,5 22-05 2000 0,7 1,60 1,70 1,50 1,4 5-06-2000 0,5 1,80 2,80 2,40 1,9 21-06-2000 0,8 2,20 2,60 2,40 2,0 10-07-2000 0,5 2,10 2,30 2,20 1,8 16-07-2000 0,6 1,6 1,8 2 1,5 Table 1. Sechi measurements in locations a,b,c,d of Figure 15 with variable depth and in

Optical remote sensing of inland waters has become a task of increasing importance, since the availability of clean fresh water is one of the great environmental challenges. In particular natural lakes and artificial reservoirs have to be monitored on a regular basis to ensure the quality of the water. With its 300 m spatial resolution and 15 spectral bands the imaging spectrometer MERIS on ENVISAT can be used for monitoring of at least larger inland waters. However, the standard algorithms as used for open ocean or even coastal waters are not appropriate because different water constituents occur in particular different phytoplankton blooms with partly extreme high concentrations. To this end the CASE 2 REGIONAL (C2R)

A time series of MERIS full-resolution (300 m spatial resolution at nadir) imagery was obtained from ESA's rolling archive at ESRIN https://oa-es.eo.esa.int/ra/mer\_frs\_l1/index.php and processed using BEAM 4.9. Images were subset to a geographic region bounded by the lat/lon limits of the study area. The BEAM 4.9 C2R processor was applied to data to extract atmospherically corrected radiance and the algal product C2R Chl\_conc, according to the methods of Doerffer and Schiller (Doerffer and Schiller, 2008a, b). Default settings were accepted for all processing parameters. The algorithm used for the retrieval of water constituents is based on the Case-2-Water Bio-Optical Model. The input to the algorithm are the water leaving radiance reflectances (i.e. the output of the atmospheric correction ) of 8 MERIS bands. The algorithm derives data of the inherent optical properties total scattering of particles (total suspended matter, tsm) b\_tsm, the absorption coefficient of phytoplangton pigments a\_pig and the absorption of dissolved organic matter a\_gelb (gelbbstof) all at 443nm (MERIS band 2). Hence the concentrations of phytoplankton chlorophyll and of total suspended dry weight are determined. The algorithm is based on a neural network which relates the bidirectional water leaving radiance reflectances with these concentration variables. We estimated the concentrations of two parameters: chlorophyll and total suspended matter. As was already pointed the test area is a cross border area between 3 different countries so it is not easy to establish a classification scheme and find the suitable variables and classification limits for a common water quality classification system. However, a relative classification scheme can be created using MERIS images. According to results shown in

processor of the BEAM 4.9 (Envisat/Brockman Consult) has been developed.

(c) Depth 5m

(d) Depth 0,5m

(d) Depth 5m

Mean value

Fig. 15. Circulation patterns of Vegoritis lake a to d: field sampling points.

The high-spatial-resolution TIR images provide a detailed view of fine-scale processes, such as surface jets, that cannot be clearly resolved in moderate-resolution images, and they enable the accurate measurement of surface transport and circulation patterns.

The high spatial resolution of ASTER and ETM+ images allow the surface currents and general circulation in lakes and coastal environments to be accurately delineated. The vector field delineates three gyres as shown in Figure 14, Convergence and divergence zones and inflows can also be clearly resolved in the thermal patterns of the high-resolution TIR satellite images. The analysis enabled the characterization of wind-driven upwelling and the measurement of surface currents and circulation at lakes of West Macedonia. Trends during the last ~25 years of lake hydraulics, concerning surface currents, turbulence charactiristics and transport phenomena are identified.


Table 1. Sechi measurements in locations a,b,c,d of Figure 15 with variable depth and in various dates of the year 2000

### **5.1.3 Suspended sediments – chlorophyll**

198 Environmental Monitoring

Fig. 15. Circulation patterns of Vegoritis lake a to d: field sampling points.

and transport phenomena are identified.

enable the accurate measurement of surface transport and circulation patterns.

The high-spatial-resolution TIR images provide a detailed view of fine-scale processes, such as surface jets, that cannot be clearly resolved in moderate-resolution images, and they

The high spatial resolution of ASTER and ETM+ images allow the surface currents and general circulation in lakes and coastal environments to be accurately delineated. The vector field delineates three gyres as shown in Figure 14, Convergence and divergence zones and inflows can also be clearly resolved in the thermal patterns of the high-resolution TIR satellite images. The analysis enabled the characterization of wind-driven upwelling and the measurement of surface currents and circulation at lakes of West Macedonia. Trends during the last ~25 years of lake hydraulics, concerning surface currents, turbulence charactiristics Optical remote sensing of inland waters has become a task of increasing importance, since the availability of clean fresh water is one of the great environmental challenges. In particular natural lakes and artificial reservoirs have to be monitored on a regular basis to ensure the quality of the water. With its 300 m spatial resolution and 15 spectral bands the imaging spectrometer MERIS on ENVISAT can be used for monitoring of at least larger inland waters. However, the standard algorithms as used for open ocean or even coastal waters are not appropriate because different water constituents occur in particular different phytoplankton blooms with partly extreme high concentrations. To this end the CASE 2 REGIONAL (C2R) processor of the BEAM 4.9 (Envisat/Brockman Consult) has been developed.

A time series of MERIS full-resolution (300 m spatial resolution at nadir) imagery was obtained from ESA's rolling archive at ESRIN https://oa-es.eo.esa.int/ra/mer\_frs\_l1/index.php and processed using BEAM 4.9. Images were subset to a geographic region bounded by the lat/lon limits of the study area. The BEAM 4.9 C2R processor was applied to data to extract atmospherically corrected radiance and the algal product C2R Chl\_conc, according to the methods of Doerffer and Schiller (Doerffer and Schiller, 2008a, b). Default settings were accepted for all processing parameters. The algorithm used for the retrieval of water constituents is based on the Case-2-Water Bio-Optical Model. The input to the algorithm are the water leaving radiance reflectances (i.e. the output of the atmospheric correction ) of 8 MERIS bands. The algorithm derives data of the inherent optical properties total scattering of particles (total suspended matter, tsm) b\_tsm, the absorption coefficient of phytoplangton pigments a\_pig and the absorption of dissolved organic matter a\_gelb (gelbbstof) all at 443nm (MERIS band 2). Hence the concentrations of phytoplankton chlorophyll and of total suspended dry weight are determined. The algorithm is based on a neural network which relates the bidirectional water leaving radiance reflectances with these concentration variables. We estimated the concentrations of two parameters: chlorophyll and total suspended matter.

As was already pointed the test area is a cross border area between 3 different countries so it is not easy to establish a classification scheme and find the suitable variables and classification limits for a common water quality classification system. However, a relative classification scheme can be created using MERIS images. According to results shown in

Monitoring Lake Ecosystems Using Integrated Remote

the changes of the relief of the Vegoritis lake basin.

has been also integrated in the GIS database.

B. Emergent vegetation due to siltation

Sensing / Gis Techniques: An Assessment in the Region of West Macedonia, Greece 201

distributed hydrologic models and for the morphometric evaluation of river network structure. The analysis of the DEM resulted to the delineation of the hydrographic network of the area of the transnational Prespa basin. The ASTER DEM has been used to delineate

Geology plays a role in the region as it allows the interconnections of adjacent river basins, which is the case of Prespa and Ohrid lakes. Ground waters cannot be observed directly by existing EO satellites, however, location, orientation and length of lineaments can be derived from EO and can be used as input for studies of fractured aquifers (e.g. location of sites for water harvesting). Available geologic maps have been scanned, geo referenced, digitized for the whole region within the context of the GIS system, Figure 3. The original maps have been of different scales and information content. A great variety of rocks with varying age and lithology constitute the catchment areas. Available information on location of springs

Α1. 1988 Α2. 2000 A1 & A2. Impact of the implementation of Government policies after the 1990's as it shown on the multi temporal images of 1988 to 2000.

D. Red areas show burned by forest fires of 2007 overlayed on the Corine land cover map.

Fig. 17. Impact of anthropogenic factors to the lakes of the study area.

C. Mining activites. 3D representation of relief changes due to surface mining as it is mapped by the ASTER DEM & Landsat image of 2011.

Fig. 16, the quality of water in the lake Ohrid is the highest among all lakes. Then follows Macro Prespa, Micro Prespa and Vegoritis while Petron shows the worst water quality. This MERIS based relative classification of lakes coincides with the classification based on the available in situ data observations.

Fig. 16. Chl concentration and tsm
