Research problem focuses on studying dynamics of spatial distribution of the seagrass meadows with a case study of P. oceanica, using aerial and satellite imagery over the 10-years period. Characteristics of the spectral reflectance of seagrass enables its discrimination from other seafloor types. Raster images processing using RS methods is suitable for seagrass mapping. Current MSc research is based on various sources of data: fieldwork in-situ measurements, satellite imagery, aerial imagery and GIS layers (maps of Crete). Technically, research is based on using GIS and RS methods: ENVI and ArcGIS software.
Seagrass mapping and monitoring along the coast of Crete, Greece. Mid-Term Presentation of the MSc Thesis.
1. Seagrass mapping and monitoring
along the coast of Crete, Greece
Mid-Term MSc Thesis Presentation
University of Twente, Faculty of Earth Observation and Geoinformation (ITC)
CO9 - GEM - MSc - 09. Supervisors: V. Venus, B. Toxopeus
Enschede, Netherlands. November 16, 2010
Polina Lemenkova
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 1 / 46
2. Table of Contents
1. Research Problem
Seagrass: Facts
Fieldwork Area
2. Research Objectives
General Objectives
Specific Objectives
Research Scheme
Research Questions
Research Goals
Hypothesis Testing
3. Data
Data Sources
Data Types
Data Collection
4. Fieldwork
Fieldwork: Ligaria Beach
Study Area: Map
Sampling Design
Olympus ST 8000
Videographic Measurements
Underwater Measurements (1)
Underwater Measurements (2)
Devices
Seafloor Types
Upscaling
5. Methods
WASI
Color Discrimination
6. Data Analysis
Graphics
Bottom Reflectance
7. Image Processing
Landsat TM
Erdas Imagine
Unsupervised Classification
Maximal Likelihood (1)
Maximal Likelihood (2)
Supervised Classification (3)
Python Script
Image Grabbing
Google Earth
Image Overlay
8. Limitations
Uncertainties
Resolution
9. Significance and Justification
10. Expected Results
11. Acknowledgements
12. Bibliography
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 2 / 46
3. Research Problem
Research problem focuses on studying dynamics of spatial distribution of the
seagrass meadows with a case study of P. oceanica, using aerial and satellite
imagery over the 10-years period. Characteristics of the spectral reflectance of
seagrass enables its discrimination from other seafloor types. Raster images
processing using RS methods is suitable for seagrass mapping. Current MSc
research is based on various sources of data: fieldwork in-situ measurements,
satellite imagery, aerial imagery and GIS layers (maps of Crete). Technically,
research is based on using GIS and RS methods: ENVI and ArcGIS software.
Fig. 1. Crete Island: general overview
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 3 / 46
4. Seagrass: Facts
The most important facts about seagrass:
The endemic Mediterranean seagrass, P. oceanica is a main species
in marine coastal environment of Greece:
the largest
the most widespread
homogeneous
dense “mattes” forming meadows between 5-40 m
.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 4 / 46
5. Fieldwork Area: Overview
Seagrass sampling has been performed at two stations at a depth of 4 meters, in
the following selected areas:
1. Ligaria beach (Agia Pelagia district), 36◦
20’N 22◦
59’E
2. Xerokampos, 35◦
12’N 26◦
18’E
Heraklion Bay: satellite imagery of the study area. Study area: locations of
measurements. Source: Google Earth.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 5 / 46
6. General Objectives
The general research objectives of the MSc research includes GIS
and environmental analysis:
Mapping the extent of the spatial distribution of seagrass P.
oceanica along the northern coast of Crete
Monitoring environmental changes in seagrass meadows in the
selected fieldwork sites (Agia Pelagia, Xerokampos) over the
10-year period (2000-2010)
The research is based on three types of the data:
1. satellite
2. aerial images
3. fieldwork measurements
using methods of the remote sensing and GIS analysis.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 6 / 46
7. Specific Objectives
Technical objectives include the following points:
To apply aerial images from the Google Earth as well as
broadband remote sensing imagery Landsat TM , MSS, ETM+,
Ikonos, SPOT images for the seagrass monitoring along the
Cretan coasts.
To use supervised classification for the thematic mapping of the
seagrass distribution
Assessment of accuracy of seagrass mapping using GIS & RS
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 7 / 46
9. Research Questions
Mapping of spatial distribution of the seagrass using broadband RS
data:
to study properties of spectral ref lectance P.oceanica and
detect exact areas of its location along the Cretan coast
to detect dynamic in changes of P.oceanica seagrass
distribution along Crete during the past 10 years using series of
Landsat TM, MSS satellite images for 2000-2010
to study the heterogeneity of the seafloor
is there any difference between the spectral ref lectance in
diverse species of seagrasses (P.oceanica, Zostera, Cymodecea,
Halophila, Ruppia, etc)?
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 9 / 46
10. Research Goals
Underwater videometric measurements (UVM) for the up-scaling
mapping of meadows and mattes
mapping P.oceanica at different scales:
small-scaled mapping (ca 1: 30 000) of seagrass meadows,
based on satellite imagery and aerial photographs
large-scaled mattes mapping (ca 1: 1 000 or 1: 2 000) of
seagrass mattes, based on the UVM (more detailed)
using the RS UVM for the mapping of P.oceanica on the
mattes scale level and compare the results of the images
classifications on the meadows scale level
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 10 / 46
11. Hypothesis Testing
Hypothesis
A statistical testing is to compare spectral responses of different seagrass types, if they
are spectrally distinct and at least one pair is statistically different at every spectral
band. Hypothesis Ho: seagrass aquatic vegetation types are not spectrally distinct,
which means Ho: µ1 = µ2 = µ3... == µn. The alternative Hypothesis Ha claims the
opposite statement: seagrass aquatic vegetation types are spectrally distinct, Ho:
µ1 = µ2 = µ3... == µn
ANOVA
Another research question is to find out, whether there are changes in spatial distribution
of the seagrass. Hypothesis Ho: there are no changes between its spatial distributions.
Hypothesis Ha: the areas have reduced their area, i.e. how has seagrass distribution
changed during the research period of 10 years? The hypothesis testing is done by
ANOVA statistical test. The purpose of ANOVA is to visualize spectral differences
between the seagrass species and their spatial distribution. The key hypotheses to prove
whether the results are correct. The distribution of the spectral responses at every
spectral band is assumed to be normal and equality of the statistical variances.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 11 / 46
12. Data Sources
Data sources: primary and secondary. The research is based on the following
data sources: Primary source – data collected during the fieldwork :
Underwater videometric measurements of the Olympus cameras made
during the ship route: 9 total routes in the selected areas of the research
places, resulting in series of consequent images, completely covering the
area under the boat path.
Underwater imagery received by the Olympus cameras
Secondary – source data:
1. Imagery. Aerial imagery from the Google Earth Satellite images from the
open sources (mostly Landsat)
2. Maps. Detailed road map covering Crete island Raster map consisted of
satellite images (mosaic), whole Crete Topographic detailed map of Crete
island
3. Results of measurements. Data of the previous measurements received
during the last year fieldwork, to analyse whether P.Oceanica is spectrally
distinct from other sea floor types, using the differences in the spectral
signatures on the graphs in a WASI, the Water Colour Simulator software.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 12 / 46
13. Data Types
Data types:
1. Satellite images from the open sources (Landsat TM)
2. Aerial images, Google Earth
3. In-situ observations of the seagrass in selected spots for the validation of
the results, using measurement frame and underwater videographic
measurements of 3 cameras Olympus ST 8000 made during the boat route
(9 total in the selected areas of the research places) resulting in series of
consequent images, completely covering the area under the boat path
4. Arc GIS vector layers of Crete island and surroundings (.shp files)
5. Georeferenced raster maps covering the whole area of Crete island: road
map, mosaic of satellite images and topographic map
6. Data of the previous measurements received during the last year fieldwork,
to analyse whether P.Oceanica is spectrally distinct from other sea floor
types, using the differences in the spectral signatures on the graphs in a
WASI, the Water Colour Simulator software.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 13 / 46
14. Data Collection
Overview of the data collected during the fieldwork. Data types:
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 14 / 46
15. Fieldwork: Ligaria Beach
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 15 / 46
16. Study Area: Map
Map of the locations of the video measurements and GPS tracklogs, Ligaria
Locations of measurements
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 16 / 46
18. Olympus ST 8000
Footage
The series of images of the underwater
videographic measurements for further
analysis and classification according to
differences in the structure, colour, texture
and shapes of the depicted objects, in order
to receive information about the seafloor
cover types.
Measurements
Several boat routes, in a direction parallel
to the coast with videometric measurements
and photographs taken along the path.
Spot measurements in the selected
locations: frame, data on density, amount of
leaves per shoot and other health indicators.
Olympus ST 8000 camera. Source: Google
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 18 / 46
19. Videographic Measurements
The transect sampling method. Advantages:
Simplicity, objectivity and ease of comparison
Boat path covers the research area in most
complete way
Several (5-7) routes of the boat in each
sampling site perpendicular to the coast line,
150-200 m long each
1-2 routes parallel the coastline
9 measurements total
Underwater video cameras:
the underwater videographic measurements of
the seafloor.
a series of consequent overlapping images of
the seafloor under the boat path.
Adjustment of three cameras for the measurements
of depths. Source: courtesy of V. Venus.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 19 / 46
20. Underwater Measurements (1)
Examples of the underwater measurements:
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 20 / 46
21. Underwater Measurements (2)
Examples of the results of the underwater measurements.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 21 / 46
22. Devices
Measurement devices. Some examples of the results of the
underwater measurements (continue).
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 22 / 46
23. Seafloor Types
Selected snapshots of the video recordings: different seafloor types, Ligaria beach, Crete
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 23 / 46
24. Upscaling
Up-scaling: matte vs meadows
Seagrass meadows (left) and seagrass mattes (right)
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 24 / 46
25. WASI
WASI: RS reflectance; concentration of
phytoplankton at depths 0.5 - 8.0 m.
WASI (Water Color Simulator) software.
WASI is used to simulate changes in water
color, caused by presence of P. oceanica and
other factors.
WASI helps to perform color discrimination
and spectral reflectance of water under
various environmental conditions which
influence its color, namely :
Different bottom depths,
Concentration of suspended particles
in water column
Water temperature,
Sun angle
Concentration of Gelbstoff (colored
dissolved organic matter)
Concentration of phytoplankton
Aerosol scattering
Exponent of backscattering by small
particles
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 25 / 46
26. Color Discrimination
WASI water colour simulator; Concentration
of Gelbstoff at depths 0.5 - 8.0 m.
WASI water colour simulator software -
II From all different parameters for the
simulation of the P.oceanica spectrum,
the most valuable and important are the
following three:
bottom depths,
sun angle at zenith
concentration of Gelbstoff
(coloured dissolved organic
matter)
The concentration of Gelbstoff is a
primary factor affecting the absorption
on incident sunlight in coastal and
estuarine waters. Changes in these
parameters affects the accuracy of the
seagrass mapping.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 26 / 46
27. Graphics
Plots of bottom reflectance (right) and remote sensing reflectance (left).
The calculations are done for the spectrum 400-800 nm, covering the most important
part of the RS spectrum:
1. Blue-green 0.45 - 0.5µm
2. Green 0.5 - 0.6µm
3. Red 0.6 - 0.7 µm
4. Red-NIR 0.7 - 0.8µm
Remote sensing reflectance, depth 0.5-8.0 m. Bottom reflectance depth 0.5-8.0 m
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 27 / 46
28. Bottom Reflectance
Bottom reflectance of P.oceanica
WASI models of bottom reflectance are used to
calculate reflectance and radiance spectra in shallow
waters.
Data of reflectance of P.oceanica, silt and green algae
macrophyte, are read from the specific file (bottom.r)
of the WASI documentation
Data of reflectance of P.oceanica, silt and green algae
macrophyte, are read from the specific file (bottom.r)
of the WASI documentation
RS of seagrass underwater measurement at various
depths measured by a remote operated vehicle (ROV),
equipped with a SPECTRIX sensors.
Figure: Bottom reflectance of P.oceanica, 550 nm;
depth 0.5-8.0 m (upper). Figure taken from: Farmer
(2005). Bottom Albedo Derivations Using
Hyperspectral Spectrometry and Multispectral Video
(lower)
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 28 / 46
30. Landsat TM
Previews of the selected Landsat satellite images: Crete Island
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 30 / 46
31. Erdas Imagine
Supervised classification in Erdas Imagine. Areas of seagrass mattes within the image,
various in colour and form of mattes were classified as types of the seagrass (below,
screenshot of the classification of areas): Example of Bali area, northern Crete
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 31 / 46
32. Unsupervised Classification
Unsupervised classification, Agia Pelagia. Raster layer read into the ArcGIS project
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 32 / 46
33. Maximal Likelihood (1)
Results of the Supervised Classification (Maximal Likelihood, Erdas Imagine)
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 33 / 46
34. Maximal Likelihood (2)
Results of the Supervised Classification (Maximal Likelihood, Erdas Imagine)
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 34 / 46
36. Python Script
Google Earth images grabbing and
stitching - I.
The aerial imagery has been
downloaded using Python
scrip from the Google Earth
using script visualized left
(written on Python by Mr.
W. Nieuwenhuis).
The script file gdal merg.py
tool was used to stitch the
tiles into one big image.
Example of Google grabbing
process.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 36 / 46
37. Image Grabbing
Google Earth images grabbing and stitching - II. After downloading, the size of the
images was reduced, converted from .tif to .ecw format, using following script command
of FWTools2.4.7 (example):
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 37 / 46
38. Google Earth
Google Earth images grabbing and stitching - III. Images downloaded from Google Earth
using grabbing process (Crete, different areas):
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 38 / 46
39. Image Overlay
Google Earth images of the northern coast of Crete Island integrated
into the ArcGIS as layers. Google images read into GIS project.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 39 / 46
40. Uncertainties
Limitations of seagrass mapping - I. Seagrass mapping has certain difficulties and some
limitations:
Uncertainties of the spectral signature of the seagrass
Uncertainties during the classification process: noises, errors, misclassified pixels
Technical difficulties of underwater measurements comparing to terrain areas
Availability of necessary data, including up-to-date imagery
Uncertainties & errors can be caused by techniques used in data capture
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 40 / 46
41. Resolution
Limitations of seagrass mapping - II.
Bad resolution
Different reflectance of the seagrasses due to
the form, position and health of separate
leaves
Spots & noises on the images
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 41 / 46
42. Significance and Justification
Precise, correct and up-to-date information about the seagrass
distribution over the coasts is necessary for the sustainable
conservation of marine environment. Accurate mapping of the
seagrasses meadows enables
evaluating current distribution of the seagrass
analysis of the effects of environmental settings and geographic
locations on the seagrass distribution
analysis of its dynamics and changes over time
estimations of the degree of the deterioration aimed at coastal
management
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 42 / 46
43. Expected Results
The research work is expected to have following results :
Over the northern coasts of Crete: thematic maps showing
seafloor types and seagrass P.oceanica spatial distribution along
the coasts of Crete
Within the fieldwork locations, Ligaria beach: monitoring the
environmental changes, based on the classification of the
satellite and aerial imagery and fieldwork video camera footage
Within the fieldwork locations : maps of the sea floor cover
types, based on the fieldwork measurements and UVM
Results of the WASI spectral analysis illustrating graphs of the
spectral reflectance of different sea floor types (sand,
P.oceanica, rocky, etc) at various depths (0.5-4 m), based on
the results of 20.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 43 / 46
44. Thanks
Thank you for attention !
Acknowledgement:
Current research has been funded by the
Erasmus Mundus Scholarship
Grant No. GEM-L0022/2009/EW,
for author’s MSc studies (09/2009 - 03/2011).
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 44 / 46
45. Bibliography I
[1] S. Gauger, G. Kuhn, K. Gohl, T. Feigl, P. Lemenkova, and C. Hillenbrand. “Swath-bathymetric mapping”. In:
The expedition ANTARKTIS-XXIII/4 of the Res. Vessel ’Polarstern’ in 2006. Berichte zur Polar- und
Meeresforschung // Rep. on Polar and Marine Res. 557 (2007). Ed. by K. Gohl, pp. 38–45. issn: 1618-3193.
doi: 10.6084/m9.figshare.7439231. url: https://www.coldregions.org/vufind/Record/288392. In
English Ant. Acc. No.: 85104. CRREL Acc. No.: 63000887; illus., incl. sketch maps.
[2] K. Gohl, G. Uenzelmann-Neben, G. Eagles, A. Fahl, T. Feigl, J. Grobys, J. Just, L. V., N. Lensch, C. Mayr,
N. Parsiegla, N. Rackebrandt, P. Schluter, S. Suckro, K. Zimmermann, S. Gauger, H. Bohlmann,
G. Netzeband, and P. Lemenkova. Deep crustal refraction and reflection seismics. Crustal and sedimentary
structure and geodynamic evolution of the West Antarctic continental margin and Pine Island Bay.
Bremerhaven, Germany: Alfred Wegener Institute, 2006, pp. 20–30. doi: 10.6084/m9.figshare.7439243.
url: https://epic.awi.de/29852/1/PE_75.pdf. Expedition program No. 75 of ANT-XXIII/4 Cruise 11-12.
[3] K. Gohl, G. Uenzelmann-Neben, G. Eagles, A. Fahl, T. Feigl, J. Grobys, J. Just, V. Leinweber, N. Lensch,
C. Mayr, N. Parsiegla, N. Rackebrandt, P. Schloter, S. Suckro, K. Zimmermann, S. Gauger, H. Bohlmann,
G. L. Netzeband, and P. Lemenkova. Crustal and Sedimentary Structures and Geodynamic Evolution of the
West Antarctic Continental Margin and Pine Island Bay. Bremerhaven, Germany, 2006. doi:
10.13140/RG.2.2.16473.36961. url:
https://epic.Alfred%20Wegener%20Institute.de/29852/1/PE_75.pdf.
[4] G. Kuhn, C. Hass, M. Kober, M. Petitat, T. Feigl, C. D. Hillenbrand, S. Kruger, M. Forwick, S. Gauger, and
P. Lemenkova. The response of quaternary climatic cycles in the South-East Pacific: development of the opal
belt and dynamics behavior of the West Antarctic ice sheet. Bremerhaven, Germany, 2006. doi:
10.13140/RG.2.2.11468.87687. url:
https://epic.Alfred%20Wegener%20Institute.de/29852/1/PE_75.pdf.
[5] P. Lemenkova. “Using ArcGIS in Teaching Geosciences”. Russian. B.Sc. Thesis. Moscow, Russia: Lomonosov
Moscow State University, Faculty of Educational Studies, June 5, 2007. 58 pp. doi:
10.13140/RG.2.2.12357.70885. url: https://thesiscommons.org/nmjgz.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 45 / 46
46. Bibliography II
[6] P. Lemenkova. “Geoecological Mapping of the Barents and Pechora Seas”. Russian. B.Sc. Thesis. Moscow,
Russia: Lomonosov Moscow State University, Faculty of Geography, Department of Cartography and
Geoinformatics, May 18, 2004. 78 pp. doi: 10.13140/RG.2.2.25360.05122. url:
https://thesiscommons.org/bvwcr.
[7] H. W. Schenke and P. Lemenkova. “Zur Frage der Meeresboden-Kartographie: Die Nutzung von AutoTrace
Digitizer f¨ur die Vektorisierung der Bathymetrischen Daten in der Petschora-See”. German. In: Hydrographische
Nachrichten 25.81 (June 2008), pp. 16–21. issn: 0934-7747. doi: 10.6084/m9.figshare.7435538.v2.
[8] I. Suetova, L. Ushakova, and P. Lemenkova. “Geoecological Mapping of the Barents Sea Using GIS”. In:
Digital Cartography & GIS for Sustainable Development of Territories. Proceedings of the International
Cartographic Conference. ICC (July 9–16, 2005). La Coru˜na, Espa˜na, 2005. doi:
10.6084/m9.figshare.7435529. url: https://icaci.org/icc2005/.
[9] I. Suetova, L. Ushakova, and P. Lemenkova. “Geoinformation mapping of the Barents and Pechora Seas”. In:
Geography and Natural Resources 4 (Dec. 2005). Ed. by V. A. Snytko, pp. 138–142. issn: 1875-3728. doi:
10.6084/m9.figshare.7435535. url:
http://www.izdatgeo.ru/journal.php?action=output&id=3&lang_num=2&id_dop=68.
Polina Lemenkova Mid-Term Seagrass mapping and monitoring along the coast of Crete, Greece 11/2010 46 / 46