Sentinel-2 and Landsat-7 satellite images qualitative ...

COMPLEMENTARY CERTIFICATE IN GEOMATICS _________________________________________________________________

Sentinel-2 and Landsat-7 satellite images qualitative

comparison for evaluating advances in detecting lakes’

quality parameters

Presented by

Stamatina MAKRI

Supervisors

Dr. Giuliani Grégory, University of Geneva, EnviroSPACE, UNEP

Prof. Dr. Dao Hy, University of Geneva, Department of Geography and Environment

June 2016

Sentinel-2A true color composite of the west coast of Lake Leman. 08/2015

Abstract

Lakes are extremely important providing a wide range of ecosystem services, biodiversity,

climate change, drinking water supply, recreation and tourism. At the same time lakes are

under increasing pressure and extremely vulnerable to various perturbation decreasing their

quality. Water monitoring is indispensable for mitigating this problem; however there is a

constant need for cost-effective ways of monitoring, which spatial and temporal exigencies

increase as well. Remote sensing has proven to be an effective tool for evaluation of in-water

constituents of lakes. As some of these parameters can be determined and examined by

remote sensing, it is indispensable to keep pace with new advances.

This study examined the characteristics of the newcomer satellite sensor Sentinel-2A and

compared its performance with its heritage, mainly Landsat 7 and 8 missions. We provided a

general qualitative evaluation of two satellite scenes, one for the Sentinel-2A sensor and

another for the Landsat 7 ETM+. Images’ multispectral transformations revealed a

significantly higher quality of the Sentinel-2A scene. In particular, high spatial resolution

coupled with the improved and higher spectral resolution, resulted in high detailed color

composites and band ratios. The calculation of certain indexes for evaluating Chl-a

distribution in the lake, revealed the added value coming from the improved band placement

of the Sentine-2 sensor and in particular the advantage of the smaller bandwidth in red-edge

part of the spectrum.

Resumé Lacs sont extrêmement importants, offrant un large éventail de services écosystémiques, sur

la biodiversité et le changement climatique, comme unique source d’eau potable, comme

espace des loisirs et du tourisme. Au parallèle, les lacs sont sous des pressions croissantes et

extrêmement vulnérables aux diverses perturbations diminuant leur qualité. La surveillance de

l'eau est indispensable pour atténuer ce problème; cependant, il existe un besoin constant de

moyens rentables de surveillance, dont les exigences spatiales et temporelles augmentent

aussi. La télédétection est avérée d’être un outil efficace pour l'évaluation des constituants

dans l'eau des lacs. Comme certains de ces paramètres peuvent être déterminés et examinés

par télédétection, il est indispensable de suivre le rythme des nouvelles avancées.

Cette étude a examiné les caractéristiques du nouveau capteur satellite Sentinel-2A et

comparé ses performances avec son patrimoine, principalement les missions de Landsat 7 et

8. Nous avons fourni une évaluation qualitative générale de deux scènes satellites, l'une pour

le capteur Sentinel-2A et une autre pour le Landsat 7 ETM +. Les transformations

multispectrales des images ont révélé une qualité significativement plus élevé pour la scène

Sentinel-2A. En particulier, la haute résolution spatiale, couplée à la résolution spectrale

améliorée, a entraîné des compositions colorées de haute précision et qualité. Le calcul de

certains indices pour évaluer la distribution de Chl-a dans le lac, a révélé la valeur ajoutée

provenant de l'amélioration de placement des bandes spectrales du capteur Sentine-2A et en

particulier l'avantage porté par l’amélioration du largeur de bandes et l’ajout des bandes dans

la partie bord-rouge du spectre.

Table of Contents

Table of Figures ......................................................................................................................... 1

Table of Tables ........................................................................................................................... 3

1. Introduction ........................................................................................................................ 4

1.1 Motivation ................................................................................................................... 4

1.2 Aim and objectives ...................................................................................................... 5

2. Theoretical Background ..................................................................................................... 6

2.1 Optical water quality variables .................................................................................... 6

2.2 Water remote sensing .................................................................................................. 8

2.3 Sentinel-2 satellite missions ...................................................................................... 11

2.4 Comparison of Landsat-Sentinel technical characteristics ........................................ 13

3. Materials and Methods ..................................................................................................... 16

3.1 Study area .................................................................................................................. 16

3.2 Aimed optical active components .............................................................................. 16

3.3 Validation with in situ measurements ....................................................................... 16

3.4 Satellite datasets selection ......................................................................................... 17

3.5 Image preparations .................................................................................................... 18

3.6 Image transformations ............................................................................................... 19

3.6.1 Color composites ................................................................................................ 19

3.6.2 Band ratios .......................................................................................................... 20

3.6.3 Index calculations (NDVI, NDWI, Chl-a) ......................................................... 20

4. Results and primary interpretations .................................................................................. 23

4.1 Color composites ....................................................................................................... 23

4.2 Band ratioing and zonal statistics .............................................................................. 26

4.3 Indexes ....................................................................................................................... 31

5. Discussion ........................................................................................................................ 38

6. Conclusion ........................................................................................................................ 40

7. Bibliography ..................................................................................................................... 41

8. Appendix .......................................................................................................................... 43

A1. In situ data used for validation of satellites’ images outcomes ..................................... 43

A2. Available scenes of Sentinel-2 in August 2015 .............................................................. 44

A3. Available scenes of Landsat 7 and 8 in August 2015 .................................................... 46

A4. Summary details of Sentinel-2A scene selected ............................................................. 47

A5. Summary basic details of Landsat 7 ETM+ scene selected ........................................... 48

A6. Illustration of basic steps of images preparation in ArcMap 10.3 ................................ 50

A7. Realization of color composites in ArcMap 10.3 ........................................................... 54

A8. Realization of band ratios in ArcMap 10.3 .................................................................... 55

A9. NDVI calculation in ArcMap 10.3 ................................................................................. 56

A10. Chlorophyll index calculations in ArcMap 10.3 .......................................................... 58

A11. Overview of Zonal Statistics tables resulted in ArcGIS for the Sentinel-2A scene. ..... 59

A12. Overview of Zonal Statistics tables resulted in ArcGIS for the Landsat 7 ETM scene.

.............................................................................................................................................. 60

1

Table of Figures

Figure 1: General overview of a typical reflectance spectrum from a eutrophic inland water

body ............................................................................................................................................ 6

Figure 2: Absorption spectra of Chl-a and b. ............................................................................. 7

Figure 3: Depths of light penetration into the water .................................................................. 7

Figure 4: Illustration of light interactions before it is captured by satellite sensors .................. 8

Figure 5: Indicative illustration of spectral extent, resolution and density for Landsat and

MODIS satellites. ....................................................................................................................... 9

Figure 6: The electromagnetic spectrum .................................................................................... 9

Figure 7: Reflectance spectra of various types of land cover................................................... 10

Figure 8: Modelled Sentinel 2 coverage at full operation ........................................................ 11

Figure 9: MSI spectral bands distribution and width along with their spatial resolution. ....... 12

Figure 10: Landsat mission imaging the earth since 1972 until today and future. .................. 13

Figure 11: Comparison of Landsat 7 and 8 with Sentinel-2 bands placement ......................... 15

Figure 12: In situ sampling sites in Lake Geneva. ................................................................... 17

Figure 13: Sentinel-2 products folder format. .......................................................................... 18

Figure 14: Available file formats of Landsat 7 and 8 scenes. .................................................. 18

Figure 15: True color composite of the Sentinel-2A scene. ..................................................... 23

Figure 16: Large scale vision of the true color composite of the Sentinel-2A scene ............... 24

Figure 17: True color composite of Landsat 7 ETM scene. ..................................................... 24

Figure 18: Large scale view of the true color composite band of Landsat 7 ETM scene. ....... 25

Figure 19: False color NIR composite band of the Sentinel-2A scene.. .................................. 25

Figure 20: False NIR color composite of the Landsat 7 ETM scene.. ..................................... 26

Figure 21: Overview of the resulting rasters from band ratioing of Sentinel-2A scene. ......... 27

Figure 22: Column chart presenting zonal statistics outcomes of the Sentinel-2A scene, for

GE3 sampling point. ................................................................................................................. 28

Figure 23: Column chart presenting zonal statistics outcomes of the Sentinel-2A scene, for

SHL2 sampling point. .............................................................................................................. 28

Figure 24: Overview of the resulting rasters from band ratioing of Landsat 7 ETM scene. .... 30

Figure 25: Column chart presenting zonal statistics outcomes of the Landsat 7 ETM scene, for

the GE3 sampling point. ........................................................................................................... 30

Figure 26: Column chart presenting zonal statistics outcomes of the Landsat 7 ETM scene, for

the SHL2 sampling point. ......................................................................................................... 31

Figure 27: The NDVISRS index of the Sentinel-2A scene. ....................................................... 32

Figure 28: NDWISRS results distribution in the Sentinel-2A image. ........................................ 32

2

Figure 29: NDCIRIS index in Sentinel-2A image. .................................................................... 33

Figure 30: Large scale view of the NDCIRIS in Sentinel-2A image. ....................................... 34

Figure 31: NDVIRIS in Sentinel-2A image. .............................................................................. 34

Figure 32: NDBIRIS in Sentinel-2A image. .............................................................................. 35

Figure 33: NDVISRS result on Landsat 7 ETM scene. .............................................................. 36

Figure 34: NDWISRS result on Landsat 7 ETM scene. ............................................................. 36

Figure 35: NDBIRIS result on Landsat 7 ETM scene. ............................................................... 37

3

Table of Tables

Table 1: Wavelengths and bandwidths of the MSI system of Sentinel-2. ............................... 12

Table 2: Wavelengths and band resolution of Landsat 7 ETM+ .............................................. 14

Table 3: Wavelengths and band resolution of Landsat 8 ......................................................... 14

Table 4: In situ radiometric defined indexes used. ................................................................... 22

Table 5: In situ results of Chl-a, phosphates and nitrates in sampling site GE3. ..................... 27

Table 6: In situ results of Chl-a, phosphates and nitrates in sampling site SHL2. ................... 28

4

1. Introduction Lakes are extremely important, not only as key constituents of entire ecosystems but also on

peoples’ quality of life. Large water bodies, such as lakes, provide a diverse range of benefits

on the functioning and biodiversity of ecosystems, but also human consumption, agriculture,

fishing, recreation. Taking into account the large variety of factors affecting and degrading

inland waters’ quality, including urbanization, land use change, deforestation, population

growth, overexploitation and contamination from industries, we realize the need to observe

the condition and discover the constantly changing trends of their constituents. Water quality

monitoring is a fundamental tool in the management of freshwater resources and essential to

maintain good water quality and thus ecosystem and socio-economic development. Water

quality can be specified in terms of its physical (e.g. clarity, temperature), chemical (e.g.

nutrients, pollutants) and biological composition (e.g. algae biomass or other harmful algal

blooms). It is not an absolute but a rather relative measure of the condition of the water

according to the needs of constituent biota or concerned humans’ needs (Bartram & Ballance,

1996). Conventional methods for water monitoring include mainly in situ sampling and that is

how most existing monitoring programs work. Nonetheless, time and cost consuming in situ

sampling, the rising need of more and more widespread locations’ sampling, uneven

distribution and capabilities of laboratories, put pressure and hamper success story of

monitoring programs.

Satellite remote sensing of biogeochemical parameters in waters has been extensively applied

to lakes’ studies (e.g. Simis, et al.; Kutser, et al., 2005; Chen, et al., 2007). Water remote

sensing is based on the study of the spectrum of optical active components of the upper layer

of the water body, such as chlorophyll-a, turbidity, total suspended solids (Giardino, et al.,

2001; Kutser, 2004). Thus, earth observation satellites data of even daily basis renewal, can

contribute to augmenting challenges of water quality monitoring by supplementing in situ

surface sampling. Open and free earth observation satellites data (e.g. Landsat, Sentinel) offer

an important potential for large areas evaluation, however the quality or frequency of the

information they send varies. Nonetheless, satellite sensors operate using diverse spatial,

spectral and temporal resolutions. For more than 40 years, the Landsat satellite series provide

a unique record of land surface use and its modifications over time. Nonetheless, highly

demanding and complex questions require more frequent and spatially precise data.

Successful application of multi-spectral satellite sensors data, depend mainly on how adequate

and precise is attributed the reflectance spectrum and by improving accuracy and width of

spectral bands, as well as by increasing the correlations and the range of optical active

components retrieved (Dekker, 1993; Aurin & Dierssen, 2012). Recent advances in space

technology, coming with Sentinel series, have led to higher spatial resolution satellite sensors,

with differentiated placement of their spectral bands.

In this study we will try to make an evaluation of the contribution of the recent advances in

space technology, in lakes’ quality observation. For this, Lake Leman in Switzerland, a very

important inland water basin in Western Europe was chosen as case study.

1.1 Motivation Taking into account increasing stresses in water resources supplies, we realize the substantial

need to asses, maintain and protect water quality by all means. The use of advances tools,

5

such as remote sensing has proven to be an important supplementary input in water quality

monitoring. After the long presence of Landsat missions, recent advances in space technology

raise the need to explore and familiarize with ongoing new possibilities on water remote

sensing. Wide swath, high spatial resolution multi-spectral Sentinel-2 mission, recently

launched by the European Space Agency (ESA), opens the possibility of news avenues in

water remote sensing. Lake Leman in Switzerland constitutes one of the largest lakes in

Western Europe and drinking water supply for more than 800’000 people. However, over the

past few centuries, this lake as well as most freshwater bodies, has suffered strong quality

degradations due to increased anthropogenic stresses. Water quality of Lake Leman has been

monitored with conventional methods for almost 60 years, generating a huge database, used to

examine long-term changes and influences of its quality. Lately, the lake has also been the

subject of remote sensing technologies studies, mainly by hyperspectral remote sensing

technologies developed by the TOPO laboratory of EPFL, as part of the Leman-Baikal

project. Nonetheless, mission Sentinel-2 with disposal of free access high quality data, gives

the unique opportunity to explore new advances in satellite remote sensing and its potential in

more efficient monitoring of inland water bodies.

1.2 Aim and objectives In this study, we present Lake Leman remote sensing monitoring case study by retrieving,

analyzing and presenting both Landsat and Sentinel satellite sensors images, in order to reveal

their different potential in water quality assessment of the lake. A summary of their different

technical characteristics and new advances is presented. The robustness of satellites images

results is demonstrated by the use of in in situ data collected in the lake. The intention of this

study is to perform a qualitative comparison of Sentinel-2A and heritage Landsat satellites

images, in order to illustrate the power of new recent advances in remote sensing technology

for monitoring water quality. What is the power of new satellite remote sensing advances of

Sentinel-2 missions in inland water bodies’ quality assessment? What is their potential

application from both past and future satellite instruments?

6

2. Theoretical Background

2.1 Optical water quality variables Water remote sensing is based on the observation and measurement of the light reflected from

the water bodies. The most important optical water quality variables which can be estimated

with remote sensing are mainly chlorophyll pigments (Chl-a), suspended particulate matter

(SPM), dissolved organic matter and secchi disk (turbidity) (Laanen, 2007). The color of

surface waters is determined by the concentrations and color spectral characteristics of these

constituent compounds, which constitute also the most important water quality parameters.

Figure 1, gives us a general overview of the typical reflectance spectrum from a eutrophic

lake, showing how different water quality parameters influence water reflected spectrum. We

see that in the visible and Near Infra-red (NIR) region (~400-900 nm), active water quality

parameters interact with the received radiation, influence and modify the shape and amount of

reflected signal. Radiation with wavelengths longer than 900 nm is mostly absorbed by water.

Thus, our zone of interest stays in the visible spectrum.

Figure 1: General overview of a typical reflectance spectrum from a eutrophic inland water body showing spectrum

influence of different optical active constituent in waters (García, et al., 2016)

Chlorophyll-a, the main constituent of phytoplankton and very important bioindicator of

water quality, which will be under the microscope in this study, exhibits a unique spectral

absorption signature. This latter is marked by two distinctive peaks, one in the blue region of

the spectrum (~433 nm) and another in the red region of the spectrum (~686 nm) (Kirk,

1994). Figure 2 gives the pattern of absorption spectrum of Chl-a and b.

7

Figure 2: Absorption spectra of Chl-a and b (Raven, et al. , 1976).

Another important point is that we refer to surface waters signal of reflectance. When light

passes through water becomes attenuated by interaction with the water column. Penetration of

light in the water depends on its wavelength. Figure 3 gives us an overview of light

penetration of different wavelengths (blue, green, red). Red light is attenuated quite rapidly

reaching max 10-15 meters deep, while blue light penetrates much further reaching up to 30

m deep.

Figure 3: Depths of light penetration into the water (source Tom Morris, Fullerton University, http://goo.gl/hEbSxg)

http://goo.gl/hEbSxg

8

2.2 Water remote sensing Water remote sensing refers to water observation from distance in order to describe its color

or temperature. This is succeeded with the use of earth observations sensors or satellites.

Objects’ information is acquired via reflected electromagnetic signals. However, light and

color, to which we focus here, are complex features, influenced by absorption, scattering, and

emission. The light reaching water surface consists of direct sun light and diffused light after

interaction with the atmosphere. At the surface of the water this light is either reflected or

refracted. In the water column, the water itself and the different constituents of the water

column will interact and transform the light by transmission, absorption, and scatter.

Proportion of scattered light upwards, is captured and observed by satellite sensors. Figure 4

gives a schematic illustration of various light interactions involving air, water and substrate.

Figure 4: Illustration of light interactions before it is captured by satellite sensors (Dekker, 1993).

As explained in 2.1, in the visible range of light, chlorophyll, suspended particulate matter

and other optical quality parameters, interact and modify the shape and amount reflected

signal (Dekker, 1993). On the other hand, the received signal from satellites is strongly

dependent on the type of satellites, spectral response and band placement. In Figure 5 we can

see an indicative schematic illustration of satellites Landsat and MODIS spectral extent,

resolution and density. We can see along in the figure the reflectance of different kinds of

observed objects (water, soil, vegetation). As explained, for water observation we stay in the

visible range of wavelengths. We realize then, that different technical characteristics,

resolution, band number and placement between satellites play a key role on the received

information.

9

Figure 5: Indicative illustration of spectral extent, resolution and density for Landsat and MODIS satellites

superposed to reflectance of different observed objects (Dekker, 1993).

In order to translate measured spectral reflectance received from water bodies, into

concentrations of water quality variables, there has been developed a series of satellite images

algorithms (Laanen, 2007). These algorithms consist of multispectral transformations of

satellite data, whose aim is to convert measured radiances to thematic variables. The idea is to

decrease the quantity of available data by keeping only the useful proportion, according to

what we need to observe. These tools use radiances or reflectance of different channels, which

are represented by the different bands of multispectral satellite data. With reference to the

different observed objects’ reflectance proprieties, the different algorithms look for the best

combination of bands that will give the best possible and most accurate translation of

information. With respects to water observation, as discussed in 2.1, the bands of the visible

range of the spectrum (~400-700 nm) interests us the most, with respects to light interaction

with the constituent optical active components. However, a large number of multispectral

transformation are based on the bands corresponding to red (600-700 nm) and near infrared

(NIR) (700 nm-1mm) spectrum (Figure 6), since it’s between those two bands that we can

observe the biggest reflectance differences between vegetation and ground or water.

Figure 6: The electromagnetic spectrum (Lillesand & Kiefer, 1987).

In Figure 7 we can see an indication of reflectance spectra of chlorophyll and water, amongst

other features. The reflectance of clear water is generally low. Nonetheless, the reflectance is

10

maximum at the end of blue and decreases as the wavelengths increase. Water constituents

mostly interact with light in the visible range and water completely absorbs light for

wavelength longer than 900 nm. Turbid water, with dissolved non algal particles interacts

with light increasing the reflectance, which is higher outside the blue region that mostly

absorb. Reflectance of bare soil depends on its composition. Vegetation, whose main

constituent is Chl-a, has a unique spectral signature. Reflectance is low on both red and blue

regions (as chlorophylls have two absorption peaks between 400-500 nm, the blue region, and

600-700 nm, the red region as seen in Figure 2 before) and high at the green region (500-600

nm). Reflectance of vegetation in the NIR is much higher than in the visible due to its cellular

structure, which do not use wavelengths longer than 700 nm (Chin, 2001). Generally,

reflectance differences between the Visible and NIR can distinguish different vegetation

types. Finally, overlapping absorptions of dissolved non algal particles of the water and Chl-a

in the blue region, renders the blue to green bands inaccurate for evaluation of Chl-a. Hence,

relevant assumptions are based mainly on the red and NIR channels.

Figure 7: Reflectance spectra of various types of land cover (Chin, 2001)

In the current study we included two main satellite algorithms, the Normalized Difference

Vegetation Index (NDVI), the Normalized Difference Water Index (NDWI) and some

specific chlorophyll indexes based on literature. Besides these wildly used indexes, there is

the possibility to reveal required information from the arithmetic combination (division or

other) of different canals in the form of band ratios. Moreover, there is the possibility to use

band combinations in order to create RGB color composites, which are specifically tailored to

facilitate the identification of features of interest. Using these, we will tend to analyze data

received from satellites Sentinel-2A and Landsat, whose technical characteristics and details

are given hereafter.

11

2.3 Sentinel-2 satellite missions The European Space Agency is developing a new series of seven satellite missions under the

Sentinel program. The Sentinel missions consist of next-generation earth observation missions

with radar and multi-spectral imaging instruments, each focusing on a different aspect,

atmospheric, oceanic or land monitoring. Each mission is based on the implementation of two

sets of satellites in order to fulfill best revisit and coverage requirements, providing this way

robust datasets for use in many applications.

Sentinel-2 is a multispectral high-resolution imaging mission whose objective is land

monitoring, including vegetation, soil and coastal areas. It will be composed of two polar-

orbiting satellites, with the first Sentinel-2A having been already launched in June 2015 and

Sentinel-2B following, in the second half of 2016 (ESA, European Space Agency , 2016). The

twin satellites will be flying in the same orbit phased at 180º to each other, designed to give

high revisit frequencies. The mean orbital altitude of the satellites is 786 km and they will

acquire data over land and coastal areas including islands, the Mediterranean Sea, inland

water bodies and closed seas (Figure 8).

The Multispectral Instrument they use works passively collecting sunlight reflected from

earth. Thereafter, the incoming light beam is separated in two different assemblies one for the

visible and for the NIR bands and another one for the Short Wave Infra-Red (SWIR) bands.

High spatial and spectral resolutions of Sentinel satellites are of great interest, as their

combination is the most important advantage they have over heritage satellites. To be more

precise, Sentinel-2A, already in orbit, using as described a MSI sensor, measures earth

reflected radiance in 13 spectral bands, spanning from the visible and NIR to SWIR. The

spatial resolution of bands, meaning the detail in a photographic image visible to the human

eye, varies from 10 to 60 m as follows (Gatti & Bertolini, 2015):

Figure 8: Modelled Sentinel 2 coverage at full operation (ESA,

European Space Agency , 2016)

12

4 bands at 10 m: blue (490 nm), green (560 nm), red (665 nm) and NIR (842 nm)

6 bands at 20 m: 4 narrow bands for vegetation characterization (705nm, 740nm, 783

nm and 865 nm

2 larger SWIR bands (1610 nm and 2190 nm), for snow, ice, cloud detection or

vegetation moisture stress assessment

3 bands at 60m for cloud screening and atmospheric corrections (443 nm for aerosols,

945 for water vapor and 1375 nm for cirrus detection)

Table 1 provides a detailed description of spectral resolutions (bandwidth) and their ability to

resolve features in the electromagnetic spectrum.

Table 1: Wavelengths and bandwidths of the MSI system of Sentinel-2 (ESA, 2016).

Figure 9 shows MSI spectral bands of Sentinel-2A versus their spatial resolution.

Figure 9: MSI spectral bands distribution and width along with their spatial resolution. (Gatti & Bertolini, 2015)

Spatial Resolution (m)

Band Number Central Wavelength (nm)

Bandwidth (nm)

10 2 490 65 3 560 35 4 665 30

8a 842 115

20 5 705 15 6 740 15 7 783 20

8b 865 20 11 1 610 90 12 2 190 180

60 1 443 20 9 945 20

10 1 380 30

13

The system uses 12 optical detectors whose configuration results to a ground coverage swath

width of 290 km. Radiometric resolution, which is the capacity of the instrument to

distinguish light reflectance differences, is 12 bit, which enables an acquired image with a

range of 0 to 4095 light intensity values or pixels.

Temporal resolution is another important aspect of Sentinel-2 satellites, as mentioned. All

areas indicated before (Figure 8), will be revisited every 5 days in full operation of the two

satellites. Currently Sentinel-2A is in a 10-day repeat cycle.

Sentinel-2 missions build directly on a unique and proven heritage of Landsat missions

orbiting for more than 40 years. Spectral bands’ configuration of the Sentinel-2 satellites is

developed around services provided by Landsat 7, 8 and Spot wavelengths. In the next

chapter we will give a short comparison’s overview of the main characteristics of the two

different satellite missions.

2.4 Comparison of Landsat-Sentinel technical characteristics For more than 40 years, the NASA-USGS (United States Geological Survey) Landsat

missions have provided the longest temporal record of earth observation data. Overall they

have provided a remarkable unbroken record. Landsat 1 was launched in 1972, with an

unstopped continuity from Landsat 1 to 7 (Figure 10).

Figure 10: Landsat mission imaging the earth since 1972 until today and future (USGS, Landsat Missions, 2016).

Landsat 7 Enhanced Thematic Mapper (ETM+) launched in 1999, consists of 8 spectral bands

with a spatial resolution of 30 m, for bands 1 to 7 and 15 m for band 8 (Table 2).

14

Table 2: Wavelengths and band resolution of Landsat 7 ETM+ (USGS, Landsat Missions, 2016).

Enhanced Thematic Mapper

Plus (ETM+)

Landsat 7

Wavelength (micrometers)

Resolution (meters)

Band 1 0.45-0.52 30 Band 2 0.52-0.60 30 Band 3 0.63-0.69 30 Band 4 0.77-0.90 30 Band 5 1.55-1.75 30 Band 6 10.40-12.50 60 * (30) Band 7 2.09-2.35 30 Band 8 .52-.90 15

Landsat 8 launched in 2013, consists of 9 spectral bands with again a spatial resolution for

most bands of 30 m (Table 3). This satellite consists of two sensors providing improved

signal-to-noise radiometric performance. Band placement and width were also improved

enabling better characterization of land cover condition and changes.

Table 3: Wavelengths and band resolution of Landsat 8 (USGS, Landsat Missions, 2016).

Landsat 8 Operational Land Imager

(OLI) and

Thermal Infrared Sensor (TIRS)

Bands Wavelength (micrometers)

Resolution (meters)

Band 1 - Coastal aerosol 0.43 - 0.45 30 Band 2 - Blue 0.45 - 0.51 30

Band 3 - Green 0.53 - 0.59 30 Band 4 - Red 0.64 - 0.67 30

Band 5 - Near Infrared (NIR) 0.85 - 0.88 30 Band 6 - SWIR 1 1.57 - 1.65 30 Band 7 - SWIR 2 2.11 - 2.29 30

Band 8 - Panchromatic 0.50 - 0.68 15 Band 9 - Cirrus 1.36 - 1.38 30

Band 10 - Thermal Infrared (TIRS) 1 10.60 - 11.19 100 Band 11 - Thermal Infrared (TIRS) 2 11.50 - 12.51 100

Swath width resolution of both satellites is 185 km and their revisit frequency 16 days.

Radiometric performance of Landsat 7 is 8-bit, which translates into 256 levels of grey.

Landsat 8 radiometric performance as well as Sentinel-2 series, quantize over a 12-bit

dynamic range. This refers to 4096 potential grey levels.

The launch of Sentinel 8 ensured a continuous data availability, which is extremely important

for earth observation. Landsat 8 records are comparable to other Landsat records and as stated

Sentinel missions were built taking into account this legacy. However, since new missions are

relatively new, there are no comparison studies between the two so far. As already seen,

satellite sensors operate on different and various spatial, spectral and temporal resolutions.

Different band placements and sensitivities and different spatial resolutions may give different

results with reference to optical active constituents in the same water bodies. Figure 11 shows

the placement of Sentinel-2 bands, compared to Landsat 7 and 8 bands.

15

Figure 11: Comparison of Landsat 7 and 8 with Sentinel-2 bands placement (source: http://landsat.gsfc.nasa.gov/)

We can see that band configuration between Landsat 7 and 8 has changed, however with the

same spectral resolution. Sentinel-2 compared to Landsat 7 bands placement and spectral

resolution is a lot different, with Sentinel-2 having more bands, more careful placement of

spectral bands (placement and width), thus higher spectral resolution. A successful application

of the above differences in characteristics has the potential to lead on one hand, an increasing

number of optical active constituents that can be detected and on the other hand stronger

correlations with AOCs concentrations.

16

3. Materials and Methods

3.1 Study area Lake Leman is located on the north side of the Alps (46º26’N, 6º23’E), at an altitude of 372

m, on the border between Switzerland and France. It constitutes the biggest and most

important inland water basin in Western Europe with a surface of 580 km2, maximum length

of 72 km and maximum width of 14 km (CIPEL, 2016). The maximum and mean depths are

309 m and 152 m respectively. The lake contains a quantity of 89 km3 of water and supplies

with drinking water more than 800’000 people. It counts many tributaries and receives water

from both Switzerland and France, with its main tributary being the Rhone; it alone

contributes 75 % of the total inflow in the lake.

Lake Leman underwent a strong cultural eutrophication in the early 1960s and it has been

monitored since 1971. Phosphorous concentrations were at 90 µg L-1

, however with the

implementation of phosphorous reduction measures in 1972, levels dropped to almost 30 µg

L-1

in 2005 (Tandoléké, et al., 2009). Long-term monitoring of the lake for almost 45 years

has generated a unique in situ measurements database, available to examine changes in water

quality status of the lake. Many studies have analysed long-term changes in lakes productivity

and trophic level status (Anneville & Pelletier, 2000; Tandoléké, et al., 2009). However, this

study as it will be seen in detail hereafter was based on short-term observations due to the

recent availability of our examined satellite dataset.

3.2 Aimed optical active components As analyzed in 2.1 there are several optical active water quality variables, able to be retrieved

by satellites datasets. In the current study, which is based on a short-term satellite dataset, we

aimed mainly to observe general trends of phytoplankton density (represented by Chl-a

concentrations), which is generally influenced by higher content of lakes’ main nutrients

(phosphates and nitrates). The choice of aimed optical active components was basically

limited by the low availability of in-situ data due to the recent date of analyzed satellite

images.

3.3 Validation with in situ measurements Validation of satellite images treatment results was made using an independent unpublished in

situ dataset provided by Moisset (Moisset , 2016). It is important here to notice that since this

is a rather qualitative approach study than quantitative, available data will be used for simple

comparisons and not extensive and precise correlation.

Provided in situ dataset, consisted of two different sampling points in the lake, SHL2 and GE3

(Figure 12). It included measurements of Chl-a, phosphates (PO4) and nitrates (NO3) at

several lake depths and in two different dates in August 2015, on the 3rd

and the 14th

(full

table of data is given in Appendix A1). As analyzed in 2.1, light penetration gets attenuated

after the first meters of water depth, hence any validation was made taking into account mean

values of measurements included in the first 10 m of depth.

17

3.4 Satellite datasets selection Our satellite dataset consisted of two different satellite missions’ images, Sentinel-2A and

Landsat 7. According to the availability of in situ data (in August 2015) and the simultaneous

availability of both Sentinel and Landsat captures, we concluded to a common image

selection on the 29th

of August 2015. To be more precise, available captures of Sentinel-2 in

August 2015 were for the 12th

, 19th

, 22nd

and the 29th

of August. However, images of the 12th

and 22nd

were unsuitable due to the fact that they did not include the lake in total and capture

on the 19th

of August was covered by clouds (relevant information illustrations in Appendix

A2). Available Landsat captures in August 2015 were, for Landsat 7 on the 13th

and the 29th

of August and for Landsat 8 on the 5th

and 21st of August (Appendix A3). Thus, available

common captures for both satellite missions was that of the 29th

of August for Sentinel-2A

and Landsat 7 satellites. In situ data, as explained in 3.3, dated on the 3rd

and 14th

of August,

which has a time discrepancy with our satellite dataset but is still valid to draw conclusions.

Data acquisition for Sentinel-2A scenes is possible in the Scientific Data Hub platform, which

provides complete, free and open access to all Sentinel products (ESA, Sentinels Scientific

Data Hub, 2016). Sentinel data policy aims for maximum availability of data and indeed this

is a free access platform with friendly interface. Data format after downloading is in

SENTINEL-SAFE format, which has been designed as a common archiving and conveying

data within ESA. It basically wraps a folder containing image data in binary data format and a

metadata file in XML. Figure 13 shows Sentinel-2 product format. The folder contains a

manifest.safe file which holds the information in XML, a preview image in JPEG2000 format,

subfolders for measuring datasets (granules/tiles) in GML-JPEG2000 and datastrip level

information, auxiliary data subfolders and HTML previews. Selected Sentinel-2A scene’s

details can be found in Appendix A4.

Figure 12: In situ sampling sites in Lake Geneva (CIPEL, 2016).

18

Figure 13: Sentinel-2 products folder format (ESA, European Space Agency , 2016).

USGS decided an open and free availability of its archive since 2008. Landsat data are

available for immediate download, mainly in EarthExplorer platform (USGS, EarthExplorer,

2016). Landsat scenes products folders contain several files, including band and metadata

files. Landsat 7 folders include Gap Mask files for each band. An overview of Landsat scenes

folders format is shown in Figure 14. Basic details of Landsat 7 ETM+ scene selected can be

browsed in Appendix A5.

Figure 14: Available file formats of Landsat 7 and 8 scenes (USGS, 2016).

3.5 Image preparations Image preparation and treatments were done using exclusively ArcGIS software (version

10.3) (ESRI ArcGIS, 2014). We first created a geodatabase named Leman SMakri.gdb. We

then proceeded to images preparations and additional layers creation that we next used in

images treatments. The main steps of images preparation and additional layers creation were

as follows:

19

1) Using the “Editor” tool, we created a polygon defining our study area, which we used

next to extract the area of interest from each bands’ raster files. For that we used the

tool “Extract by Mask”, in order to cut all bands’ raster files of both Sentinel-2A and

Landsat 7.

2) For each cut scenes’ raster we preceded to radiometric enhancements of levels of grey,

using and modifying the histograms in each layer’s “Properties” window. This is an

image modification by changing pixel brightness values in order to improve its visual

impact.

3) We inserted two coordinates’ point layers for the two different sampling sites GE3 and

SHL2. For that we first created two excel files containing the coordinates of the two

different points. In ArcGIS we used the “Excel to Table” conversion tool to convert

the excel file to a table in ArcGIS. Using the path: File→ Add Data→ Add XY Data,

defined the x, y attributes according to the coordinates in the tables we created. It is

also important here to define the projection coordinate system using the Edit option, in

our case WGS_1984. This created two point layers containing our coordinates which

we exported by right clicking on the layer and by using the path: Data→ Export Data.

We saved the file as shapefile.

4) We created two buffer layers around the two different sampling sites. These layers

were used for pixels statistics, as we will in detail next. To create the two buffer layers

we used the “Buffer” tool. For Sentinel-2A images we created a buffer of 10 m, taking

into account used bands spatial resolution of 10 m, in order to measure at least 1

pixel’s statistics. For Landsat 7 scene we used a 30 m buffer around the sampling sites

points.

Further details including screen captures of main image preparation made, can be found in

Appendix A6. In the next chapter we can see incorporated treatments, in order to transform

and reveal measured spectral reflectance of our images.

3.6 Image transformations Satellite image transformations involve manipulations of multiple bands data in order to

highlight particular properties or features of interest within the study area, in a better and

more effective way than the original input images. These transformations produce “new”

images that have the potential to reveal better the electromagnetic radiation captured by the

satellite sensors. The multi-spectral nature of our images allows their transformation to new

images with a different set of components or bands. The idea here is to reduce the number of

dimensions and keep only those useful and more pertinent to get some more evident

information about the zones of interest. In our study we have incorporated image

transformation involving color composites, band ratios and indexes such as NDVI, NDWI and

Chl-a, which are all analyzed hereafter.

3.6.1 Color composites This transformation refers to the creation of new multispectral images that consist of the three

primary color bands (red, green and blue). Composites can be true, meaning that

corresponding to red, green and blue bands of the satellite image were assigned respectively

20

to the R, G, B colors for display, or they can also be false. False composites use an arbitrary

display assignment. In this later case the display color of the target feature does not

correspond to its true color; however some display schemes may be more suitable for

detecting certain objects. Both composite have as aim to attribute to different features

different colors and reveal more information.

As discussed in 2.2, with respects to water remote sensing and vegetation observations we are

particularly interested in the visible and NIR bands, where we can observe the biggest

reflectance differences between vegetation, water and other constituents. We discussed that

clear water has a maximum reflectance in the blue-end part appearing dark-bluish. Turbid

water has increased reflectance in the red-end region appearing more brownish. Chl-a in water

would give low reflectance in the blue and red and higher in the green part.

According to our aimed features we used, for the Sentinel-2A scene, we took the

corresponding bands 2 (blue), 3 (green), 4 (red) and 8 (NIR) (Figure 11). We performed the

true color composition R: 4 G: 3 B: 2 and the false NIR color composition R: 4 G: 3 B: 8,

using the tool “Composite Bands”, illustrations in Appendix A7.

For the Landsat image, we took the bands 1, 2, 3 and 4 which correspond, as seen in Table 2

in 2.4, to the blue, green, red and NIR region respectively. In ArcGIS, using again the tool

“Composite Bands”, we realized the true color composition R: 3 G: 2 B: 1 and the false NIR

color composition R: 3 G: 2 B: 4.

3.6.2 Band ratios Another image manipulation technique is band ratioing. This refers to different bands’

division, according to the desired features to observe. For each pixel we divide the digital

number value of one band to another band. This yields to a new set of pixel numbers for the

resulting rasters from 0 to 255 for Landsat 7 images and 0 to 4095 for Sentinel-2 images,

however the majority of resulting numbers are fractional. The aim of this manipulation is to

reveal differences between spectral reflectance of the features we intent to observe.

In the current study and according to aimed features, for the Landsat 7 scene we realized the

band ratios 4/1, 1/4, 2/3, 3/1, 3/2. In ArcGIS, this is done with the use of the tool “Raster

Calculator”. As the most possible is to have fractional numbers results, we incorporated the

option “Float” in the mathematical equation used (Appendix A8). For the Sentinal-2A scene,

we realized the corresponding ratios 8/2, 2/8, 3/4, 4/2, 4/3.

In order to evaluate ratios’ resulting values we realized zonal statistics calculations to the new

resulting band ratio rasters. In order to define the zone of calculations, we used the buffer

layers (as explained in 3.5), which defined a zone of 10 m and 30 m around the sampling

sites, for Sentinel-2A and Landsat 7 scenes respectively, according to their spatial resolution.

In ArcGIS for implementing zonal statistics we used the tool “Zonal statistics as table”, which

results to an excel table output (an example can be seen in Appendix A8).

3.6.3 Index calculations (NDVI, NDWI, Chl-a) Image ratioing may involve more complex ratios, than just the division of two bands. These

more complex ratios, called indexes, which may involve the sums or subtraction between

bands, have been extensively developed for monitoring various observed features. In the

current study we incorporated the NDVI index widely used for monitoring vegetation

21

conditions and the NDWI for revealing differences between water and other features, both

widely used in satellite remote sensing, as well as some Chl-a indexes, which are presented

hereafter.

The NDVI index is mostly used to reveal vegetation coverage of soils. It is sensitive to

chlorophyll contents in plants and given that plant leafs absorb the red and reflect the NIR; the

index tries to reveal the high NIR reflectance areas. The index tales values from -1 to 1, with

areas of low vegetation having negative values and the areas of high vegetation having high

values. The NDVI index is given by the following equation:

NDVISRS = (NIR – Red / NIR + Red) Eq.1

For the Landsat 7 scene we used the bands 3 (red) and 4 (NIR). For the Sentinel-2A scene we

used the corresponding bands 4 (red) and 8 (NIR). In ArcGIS, we used the specialized tool for

the calculation of the NDVI index, which can be found from the path Windows→ Image

Analysis and after having selected the relevant bands, clicking on the NDVI calculation tool.

We used the “Scientific Output” option to obtain values from -1 to 1 (Appendix A9).

The NDWI index is used for estimation of the water content and for soil moisture estimation.

As discussed in 2.2, water has a peak in reflectance in the green part of the visible spectrum

and its opposition to the NIR has the potential to reveal water covered areas. The NDWI index

is given by the equation:

NDWISRS = (Green – NIR / Green + NIR) Eq. 2

Values are similar to NDVI index, with low values 0 or -1, resulting to bright surfaces of no

vegetation and high values +1, to water content. This index was mainly used to validate the

accuracy and effectiveness of bands used for Landsat 7 and Sentinel-2A, for detecting water

content.

Besides these widely used satellite remote sensing indexes, several studies have worked on

developing specific Chl-a indexes by in situ radiometric measurements. Table 4 summarizes

some key, in situ radiometric defined indexes, chosen to be presented in this study, along with

the standard satellite indexes presented before. In table 4 we can find the details and the

equations that define each index, as well as the corresponding bands for each satellite.

Numbers in indexes equation represent the reflectance wavelengths defined by the in situ

reflectance. References are given in the table.

22

Table 4: In situ radiometric defined indexes used to evaluate satellite reflectance accuracy of Sentinel-2 and Landsat.

Index details and equation Landsat 7 bands

Sentinel-2A bands

References

NDCIRIS= R(709)-R(665) / R(709)+R(665) Eq. 3

Normalized Difference Chlorophyll Index

4-3 / 4+3 5-4 / 5+4 (Mishraa & Mishraa, 2012;

Augusto-Silva , et al., 2014)

NDVIRIS= R(748)-R(675) / R(748)+R(675) Eq. 4

Normalized Difference Vegetation Index

4-3 / 4+3 6-4 / 6+4 (Rouse, et al., 1974)

NDBIRIS= R(550)-R(675) / R(550)+R(675) Eq. 5

Normalized Difference Algal Bloom Index

2-3 / 2+3 3-4 / 3+4 (Xue, et al., 2015)

Indexes calculations were made using the “Raster Calculator” tool in ArcGIS, creating new

raster layers. An example of index raster calculation can be found in Appendix A10.

It is important to notice here the variability of bands in Sentinel-2A indexes bands. As

discussed in 2.3, for Sentinel-2 there are four narrow bands for vegetation characterization

(705nm, 740nm, 783 nm and 865 nm) (Figure 9). On the other hand, less variable, wider

bands of Landsat 7, resulted to the use of same bands for most indexes. The NDVISRS,

presented before, the NDVIRIS and the NDCIRIS used bands coincide. In addition, it is

importance remark that this a qualitative approach study and there was not any atmospheric

correction made. Hence, we cannot expect any precise results on the implementation of these

indexes. Our results are presented in the next chapter.

23

4. Results and primary interpretations

4.1 Color composites As discussed in 3.6.1, color composites give the combination of three raster bands, each

representing a different portion of the electromagnetic spectrum, in shades of red, green and

blue (RGB images). For visual analysis, color composites give the maximum use, compared

with human eye’s capabilities.

Figure 15 gives the true color composite of our Sentinel-2A scene. We can see that the image

offers a very close to natural color rendition. A larger scale view of the same color composite

can be seen in Figure 16, where we can visualize better the large variation of captured details.

The color of the lake, the constructed area around it, the vegetation area, even the atmospheric

perturbation by clouding and aircraft trails are very well figured. The image reveals very well

the area of the lake in deep blue color, an area of low less robust vegetation or constructed

populated area, around the lake, in brighter bare soil colors and lighter green as seen in the

large scale image. Outside this area there is an area of strong vegetation in deep green color.

Clouding and plane trains are clearly visible in white color. In addition, in the large scale

image (Figure 16), we can see that even in the low reflectance area of the water, details

especially near the coast are visible.

Figure 15: True color composite of the Sentinel-2A scene.

24

Figure 16: Large scale vision of the true color composite of the Sentinel-2A scene

Figure 17 presents the true color composite of the Landsat 7 ETM scene. The image contains

unfortunately a lot of noise. There is intensive stripping covering the area of the lake. In

Figure 18 we can observe a larger scale view of the image.

Figure 17: True color composite of Landsat 7 ETM scene.

25

We can see that the image appears in more intense red shades. Different features covering our

area are visible, but in different colors than in the Sentinel-2A image before. The lake area

appears in dark blue and green color and is difficult to discriminate any pattern details; the

populated and less robust vegetation area around the lake is in pink color, resulting probably

from a mix of red and blue band. The area of robust intense vegetation, further away from the

lake outside the populated area, appears rather dark grey again with some pink nuances. There

seems to be a perturbation of the reflectance captured by the satellite possibly due to

atmospheric interference.

Figure 19 presents the false color NIR composite of Sentinel-2A scene. The only difference

with the true color image is that the NIR band has been assigned to the blue shades.

Figure 19: False color NIR composite band of the Sentinel-2A scene. The NIR, band 8 as shown

in the legend, was attributed to the blue shades.

Figure 18: Large scale view of the true color composite band of Landsat 7 ETM scene.

.

26

Robust green vegetation parts of the true color composite before, here appear as blue, as

vegetation reflects infrared light. The parts of bare soil, construction, clouding or snow of

high reflectance appear yellowish. Water appears dark as it absorbs NIR radiation.

Figure 20 presents the same false color composite for the Landsat 7 ETM image. Here we can

see a visually better rendition than in the true color composite. Vegetation areas reflecting the

NIR look bluish, while barren soils and clouding look brighter. The lake area looks dark as

expected. Again we have the pinkish areas resulting from a misinterpretation in the blue and

red region.

4.2 Band ratioing and zonal statistics As discussed in 3.6.2, the band ratioing transformations we conducted had as aim to reveal

spectral reflectance of various features that we intended to observe. Figure 21 presents an

overview of band ratios new rasters received for Sentinel-2A. In order to evaluate ratios

outcomes, we implemented zonal statistics calculations around our two sampling points in the

lake GE3 and SHL2 (Figure 12 in 3.3). Zonal statistics results were plotted in column charts

for each band ratio and each of the two sampling sites. Figure 22 and 23 demonstrates the

results for Sentinel-2A image, for the two different sampling sites. A full overview of the

resulting tables from ArcGIS is given in Appendix A11. An evaluation of zonal statistics

results is attempted using the in situ retrieved results, as discussed in 3.3. The in situ results

Figure 20: False NIR color composite of the Landsat 7 ETM scene. Band 4 representing the NIR was attributed to

the blue shades.

27

can be seen in Table 5 and 6 hereafter. For the latter, we took into account average values for

depths up to 10 m, which is presented in the tables in red.

Table 5: In situ results of Chl-a, phosphates and nitrates in sampling site GE3. In red we can see the average value for

measurements up to 10.

Date Station Coordinates Depth Chl.a PO4 NO3

(Swiss system) (m) (µg/L) (mg/L) (mg/L)

10.08.2015 GE3 506.100/128.040 0 4.34 0.011 0.296

10.08.2015 GE3 506.100/128.040 -10 21.29 0.009 0.327

12.82 0.01 0.31

Figure 21: Overview of the resulting rasters from band ratioing of Sentinel-2A scene.

28

Table 6: In situ results of Chl-a, phosphates and nitrates in sampling site SHL2. In red we can see the average value

for measurements up to 10.

Date Station Coordinates Depth Chl.a PO4 NO3

(Swiss system) (m) (µg/L) (mg/L) (mg/L)

03.08.2015 SHL2 534.700/144.950 0 1.90 0.011 0.336

03.08.2015 SHL2 534.700/144.950 -10 5.49 0.012 0.334

3.69 0.01 0.33

Figure 22: Column chart presenting zonal statistics outcomes of the Sentinel-2A scene, for GE3 sampling point.

Figure 23: Column chart presenting zonal statistics outcomes of the Sentinel-2A scene, for SHL2 sampling point.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

S2A8/2

S2A2/8

S2A3/4

S2A4/2

S2A4/3

S2A4/8

Val

ue

s

Band ratios

Zonal statistics S-2A GE3 point

Average number of ratiovalues

0

1

2

3

4

5

6

7

S-2A8/2

S-2A2/8

S-2A3/4

S2A4/2

S2A4/3

S2A4/8

Val

ue

s

Band ratios

Zonal statistics S-2A SHL2 point


29

From the charts in Figures 22 and 23 we can see that the majority of band ratios give lower

average values around the SHL2 sampling point, except for the S2A 2/8 and the S2A 3/4

ratio. The GE3 sampling point is in a less deep part of the lake, richer is dissolved substances

and algae, which we would expect to increase total reflectance of the water and result to

overall higher ratios. Nonetheless this depends on the ratio, the S2A 2/8 ratio is expected to

have higher values in a part of clearer, deeper, open water area. Additionally, as mentioned

before, the fact that we did not implement any atmospheric correction to our images probably

results to misleading measured reflectance, especially in the band of red, for the S2A 3/4

ratio.

The band ratio S2A 2/8 (blue/NIR) gives, relatively to other bands, higher values for both

sites, since water absorbs in the NIR and its reflectance is higher in the blue of the visible

range. However, the ratio gives higher values in the SHL2 site which is a more open and clear

water part of the lake, than in the GE3 less profound and close to the shore. The lower values

of the same ratio in the GE3 site is expected, as it is richer in Chl-a (as presented in Tables 5

and 6) and part of the blue light is absorbed by algae.

The ratio S2A 4/8 (red/NIR) is expected to show a satisfying correlation with Chl-a content

and we would expect to have lower values for the sampling point richer in Chl-a. Indeed, we

see that the ratio showed a lower value at the GE3 sampling point, correlating with our in situ

dataset.

From the rest of the bands, the ratio S2A 3/4 (green/red) would be expected to also correlate

and reflect Chl-a content in the two sites , with higher values in parts of the lake with higher

Chl-a content, as it absorbs in the red and has higher reflectance in the green. However, the

ratio gives higher values in the SHL2 site (2.14), where we found lowest Chl-a content from

the in situ measurements. In GE3 site the ratio shows lower average pixel values (1.42). We

see then that due to the overlapping absorbance in the blue and green, this ratio does not

conclude to a good Chl-a representative. Nonetheless, it is indispensable to apply atmospheric

correction for retrieving accurate vegetation results. Additionally, it is important to notice that

Sentinel-2A band configuration, offers 4 distinctive narrow bands for vegetation evaluation,

which we did not use in this analysis. We only used the wider bands, closer to the Landsat

band configuration.

Figure 24 presents band ratios’ rasters, resulting from the Landsat 7 ETM scene. Here, we can

see the lower level of visual quality results compared to Sentinel-2A band ratioing. There is

always the stripping noise which is actually no data values and is important to be noticed at

this point. Figures 25 and 26 hereafter, present the column charts of zonal statistics outcomes

derived from the Landsat 7 ETM scene. Also here, the ratio ETM 1/4 (blue/NIR), shows

higher values (1.69) in the SHL2 site (less turbid clearer water), and lower values (1.47), in

the GE3 sampling point (more turbid higher Chl-a content), as expected. The ratio ETM3/4

(red/NIR) shows also correlation with the in situ data Chl-a content, giving a lower value in

the GE3 sampling point. The ratio green/red ETM 2/3 is an interesting observation here,

giving higher values for the GE3 point (4.04) and lower values at the SHL2 point (1.37),

which is different to what we acquired in the Sentinel-2A scene. We see thus an inconsistency

in the green, blue bands of the two satellites.

30

Figure 25: Column chart presenting zonal statistics outcomes of the Landsat 7 ETM scene, for the GE3 sampling

point.

Figure 24: Overview of the resulting rasters from band ratioing of Landsat 7 ETM scene.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

ETM4/1

ETM1/4

ETM2/3

ETM3/1

ETM3/2

ETM3/4

Val

ue

s

Band ratios

Zonal statistics ETM+ GE3 point


31

Figure 26: Column chart presenting zonal statistics outcomes of the Landsat 7 ETM scene, for the SHL2 sampling

point.

4.3 Indexes Indexes are very useful tools in remote sensing for retrieving optical active components

concentrations in water bodies. The indexes presented in 3.6.3, were applied separately in our

Sentinel-2A and Landsat 7 scenes.

With respects to Sentinel-2A image, Figure 27 shows the result of classic satellite remote

sensing NDVISRS index.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

ETM4/1

ETM1/4

ETM2/3

ETM3/1

ETM3/2

ETM3/4

Val

ue

s

Band ratios

Zonal statistics ETM+ SHL2 point


32

The index gave values from -0.50 to +0.88. Lower values are presented in blue, close to zero

values in yellow and higher positive values in red. This index, as discussed, incorporates

bands 4 (red) and 8 (NIR) and is mostly used in retrieving soil vegetation. Strong vegetation

absorbs most of visible wavelengths and reflects most of the NIR. Positive values close to 1

are considered as vegetated areas, values close to -1 represent water or humid areas and

values close to 0 are generated from features of high brightness, bare soil or clouds.

We can see that the ratio shows indeed higher values in the parts of intensive vegetation. For

the areas around the lake where vegetation type changed, with less robust vegetation, crops

and constructed areas the index is close to 0 painting these areas yellow. The areas of strong

vegetation the NDVI takes high positive values painting them red in the figure. With respects

to the water body area, which is our main interest, we wouldn’t expect any color variations

from this index. The lake appears blue, with index values very close to -1 as water absorbs

most of the NIR radiation and has low reflectance in the red overall.

Figure 28 presents the outcome of NDWISRS index in the Sentinel-2A scene. The index

incorporated the bands 3 (green) and 8 (NIR). Then index is important for estimating water

contents and soil moisture. Lower values are typically non water features, terrestrial

vegetation, bare soil etc., and higher values are open water or humid areas.

Figure 28: NDWISRS results distribution in the Sentinel-2A image.

Figure 27: The NDVISRS index of the Sentinel-2A scene.

33

The index ranged from -1 to 0, with negative values representing non water areas and 0 water

and more humid areas. The ratio, as we can see, delineates perfectly the lake, revealing also

some other humid areas and watercourses in the region around the lake.

We will now present the three chlorophyll index, which as discussed in 3.6.3 are based on situ

reflectance measurements, with application on satellite remote sensing scenes. These indexes

used the narrow vegetation bands of Sentinel-2A band configuration. Figure 29 shows the

NDCIRIS index in Sentinel-2A image.

Figure 29: NDCIRIS index in Sentinel-2A image.

The index took values from 0 to 87, with higher values representing higher content of

chlorophyll. The ratio was built to predict chl-a content in waters, using corresponding bands

in order to eliminate other constituents such as colored dissolved organic matter or other

detritus that affect spectral channels in the blue and green part. This ratio used mostly red

bands for eliminating interference derived from increased turbidity. We used band 5, which is

included in the four red edge vegetation bands of Sentinel-2A and band 4. As we can see in

Figure 29 and with respects to the water body area, we can distinguish some color variations.

Better visible in the large scale Figure 30. However, since we haven’t implemented any

atmospheric correction calculation on the image is hard and very unsafe to draw any

conclusions.

34

The same is also for the indexes NDVIRIS, which used the red-edge vegetation band 6 and

band 4, and the NDBIRIS that used bands 3 and 4, presented in Figures 31 and 32.

Figure 31: NDVIRIS in Sentinel-2A image.

Figure 30: Large scale view of the NDCIRIS in Sentinel-2A image.

35

Figure 32: NDBIRIS in Sentinel-2A image.

The indexes NDVISRS, NDWISRS and NDBIRIS were also calculated for the Landsat 7 ETM

scene. The indexes NDCIRIS and NDVIRIS, in the case of Landsat 7 band configuration

coincided with the normal NDVISRS bands.

Figure 33 show the outcome of normal NDVISRS in the Landsat 7 scene. The index took

values from -1 to 1. The overall quality result of the index is very low. Compared with the

NDVISRS of Sentinel-2A image in Figure 27, we can see that even high vegetation

delimitation shows differences.

Figure 34 present NDWISRS result on Landsat 7 ETM scene. Again as we can see that the

index fails to represent accurately the limits of water content in the area.

Finally, in Figure 35 we can see the NDBIRIS result on Landsat 7 ETM scene. This index

seems to have a better visual result, in defining some type of soil vegetation.

36

Figure 33: NDVISRS result on Landsat 7 ETM scene.

Figure 34: NDWISRS result on Landsat 7 ETM scene.

37

Figure 35: NDBIRIS result on Landsat 7 ETM scene.

38

5. Discussion As seen in the results, we can estimate that overall Sentinel-2A scene gave better quality

results and revealed high potential for retrieving more adequately optical active components

from water bodies, with its low reflectance whatsoever.

To be more precise and if we take a closer look to our results this consideration has a base.

True color composite of Sentinel-2A scene attributed almost completely natural colors. The

image was of great quality overall, at 10 m pixel size, offering really impressive views and

remarkable detail even in the low reflectance water area. In the large scale figure of the same

composite we were able to visualize the captured detail. This fact verified the high spectral

sensitivity of the Sentinel-2A MSI sensor and its very realistic band placement. The false

color NIR composite also verified the accurate placement of this band.

Band ratioing results were again satisfying with very good quality resulting rasters for the

Sentinel-2A scene. We concluded good response of ratios of certain bands, the blue and the

NIR, however less effective red band response. High radiometric resolution of this sensor

coupled with the fact that we did not applicate any atmospheric correction before image

treatments is probably the answer to this drawback. In fact, many studies have shown the

importance of atmospheric correction before interpretation of satellite images and their large

source of noise for accurate surface reflectance measurements (Bagheri, 2011; Honkavaara, et

al., 2009). Such errors can increase uncertainty depending on the spectral channel. In

particular, with respects to vegetation ratios and indexes, influences can be significant and

deviations from being able to draw real reflectance of the constituent feature, large

(Hadjimitsis, et al., 2010).

Indexes implementation revealed the added value and high potential of the narrow red-edge

bands of Sentinel-2A. Again the need for efficient atmospheric correction was obvious,

nonetheless there is an unquestionable advantage drawn from the narrower bandwidth, more

careful band placement and pixel size over the heritage satellites. All characteristics were

reflected in the transformed rasters of our study.

It is possibly unfair to make any direct comparisons with Landsat 7 ETM scene, as there is

also Landsat 8 in orbit, a more updated and improved version of it, which unfortunately did

not have any available scenes for the aimed period. However, since both are considered as

heritage of the newcomer Sentinel-2 missions, differences must be considered. The true color

composite of the Landsat 7 scene was overall of inferior precision and quality. There seemed

to be a failure in visible color accounting from the reflectance received. There was a

perturbation of the reflectance captured by the satellite possibly due to atmospheric

interference. Other than that, color differences to natural nuances could be due to bad spectral

resolution. The problem seemed to be mostly in the differentiation of different types of

vegetation, in the red region. The sensor gets real reflectance for the parts where the

absorption-reflectance is more intense i.e. in the forest parts beyond the crop and populated

area around the lake. The difference came out as pink, a mixture of blue and red. The

implementation of different indexes, made obvious the great disadvantage by the non-

appropriate band placement and large bandwidth. There was problems differentiating

reflectance, mainly in the visible part of the spectrum where reflectance is basically low with

respects to water bodies.

It is important though to notice that this study was based on a qualitative evaluation of the

analyzed scenes, without any further atmospheric correction calculation before, which made

39

difficult to detect in precision new advances in water optical constituent capturing. However,

overall evaluation of Sentinel-2A scenes made obvious its high potential, as it gave overall

better reflectance values, better color intensities, higher resolution by revealing more detail.

All these characteristics coupled with its higher temporal resolution, make a cogent new

sensor package. However, added value of these characteristics always depends on observed

features. In any case, for inland waters Sentinel-2 seems to present significant better

characteristics than its heritage.

40

6. Conclusion Increasing stresses on lakes all over the world have generated the need for cost effective and

quick water monitoring techniques. Satellite remote sensing has proven to give better results

in both temporal and spatial scale. However, the extent and complex interactions of light,

which constitutes the base for remote sensing on one hand and the large variability of lake

constituents on the other hand, makes indispensable the use of accurate sensors, of high

sensitivity, with the best possible band location and bandwidth. These are the variables that

mostly create noise to signals received from satellites.

In this study, we made obvious that improvements of these variables has a high added value

and give a high potential to reach very low levels of received noise for the signals received

from satellites. Sentinel-2A, with its improved characteristics enlightened this way and the

great potential it carries.

41

7. Bibliography

Anneville, O., & Pelletier, J. P. (2000). Recovery of lake Geneva from eutrophication: quantitative response of phytoplankton. Arch.Hydrobiol. 148, 608-624.

Augusto-Silva , P., Ogashawara, I., Barbosa, C. C., de Carvalho, L. A., Jorge, D. S., Fornari, C. I., & Stech, J. L. (2014). Analysis of MERIS Reflectance Algorithms for Estimating Chlorophyll-a Concentration in a Brazilian Reservoir . Remote Sens. 6, 11689-11707.

Aurin, D. A., & Dierssen, H. M. (2012). Advantages and limitations of ocean color remote sensing in CDOM-dominated, mineral-rich coastal and estuarine waters. Remote Sensing of Environment 125, 181-197.

Bagheri, S. (2011). Nearshore water quality estimation using atmospherically corrected AVIRIS data. Remote Sens. 3, 257-269.

Bartram, J., & Ballance, R. (1996). Water Quality Monitoring - A Practical Guide to the Design and Implementation of Freshwater. London: for UNEP/WHO by E&FN Spon.

Chen, Z., Hu, C., & Muller-Karger, F. (2007). Monitoring turbidity in Tampa Bay using MODIS/Aqua 250-m imagery. Remote Sensing of Environment 109, 207-220.

Chin, L. S. (2001). Optical Remote Sensing. Retrieved from Center for Remore Sensing (CRISP): http://www.crisp.nus.edu.sg/~research/tutorial/optical.htm

CIPEL. (2016, June 10). Commission internationale pour la protection des eaux du Léman. Retrieved from http://www.cipel.org/

Dekker, A. G. (1993). Detection of optical water quality parameters for eutrophic waters by high resolution remote sensing (Phd Thesis). Amsterdam: Universiteit Amsterdam.

ESA. (2016, June 10). Retrieved from European Space Agency : https://sentinel.esa.int/web/sentinel/missions/sentinel-2

ESA. (2016, June 10). Sentinels Scientific Data Hub. Retrieved from https://scihub.copernicus.eu/

ESRI ArcGIS. (2014). ArcMap 10.3. Redlands, California, United States: http://www.esri.com/software/arcgis.

García, L. E., Rodríguez, D. J., Wijnen, M., & Pakulski, I. (2016). Earth Observation for Water Resources Management. Washington: International Bank for Reconstruction and Development / The World Bank.

Gatti, A., & Bertolini, A. (2015). Sentinel-2 Products Specification Document. France: Thales Alenia Space.

Giardino, C., Pepe, M., Brivio, P. A., Ghezzi, P., & Zilioli, E. (2001). Detecting chlorophyll, Secchi disk depth and surface temperature in a sub-alpine lake using Landsat imagery. The Science of the Total Environment 268, 19-29.

Hadjimitsis, D. G., Papadavid, G., Agapiou, A., Themistocleous, K., Hadjimitsis, M. G., Retalis, A., . . . Clayton, C. R. (2010). Atmospheric correction for satellite remotely sensed data intended for agricultural applications: Impact on vegetation indices. Natural. Hazards Earth Syst. Sci. 10, 89-95.

Honkavaara, E., Arbiol, E., Markelin, L., Martinez, L., Cramer, L., Bovet, S., . . . Veje, N. (2009). Digital airborne photogrammetry—A new tool for quantitative remote sensing?—A state-of-the-art review on radiometric aspects of digital photogrammetric images. Remote Sens. 1, 577-605.

42

Kirk, J. T. (1994). Light photosynthesis in Aquatic ecosystems. Cambridge, UK: Cambridge University Press.

Kokaly, R. F., Despain, D. G., Clark, R. N., & Livo, E. K. (2007). Vegetation Cover and Microbial Communities in Yellowstone National Park Using AVIRIS Data. In L. A. Morgan, Integrated Geoscience Studies in the Greater Yellowstone Area. Volcanic, Tectonic, and Hydrothermal Processes in the Yellowstone Geoecosystem (pp. 463-499). U.S. Geological Survey.

Kutser, T. (2004). Quantitative detection of chlorophyll in cyanobacterial blooms by satellite remote sensing. Limnology and Oceanography 49, 2179-2189.

Kutser, T., Pierson, D. C., Kallio, K. Y., Reinart, A., & Sobek, S. (2005). Mapping lake CDOM by satellite remote sensing. Remote Sensing of Environment 94, 535-540.

Laanen, M. (2007). Yellow Matters Improving the remote sensing of Coloured Dissolved Organic Matter in inland freshwaters (Phd Thesis) . Amsterdam: Universiteit Amsterdam.

Lillesand, T. M., & Kiefer, R. W. (1987). Remote Sensing and Image Interpretation. Toronto: Sec. Ed., John Wiley and Sons.

Mishraa, S., & Mishraa, D. (2012). Normalized difference chlorophyll index: A novel model for remote estimation of chlorophyll-a concentration in turbid productive waters. Remote Sensing of Environment 117, 394-406.

Moisset , S. (2016, April 25). Personnal communication.

Raven , P. H., Evert , R. F., & Curtis , H. (1976). Biology of Plants. New York: Worth Publishers.

Rouse, J. W., Haas, R. H., Schell, J. A., & Deering, D. W. (1974). Monitoring vegetation systems in the Great Plains with ERTS. NASA Spec. Publ., 351,309.

Simis, S. G., Peters, S. W., & Gons, H. J. (2005). Remote sensing of the cyanobacterial pigment phycocyanin in turbid inland water. Limnology and Oceanography 50, 237-245.

Tandoléké, R. D., Lazzarotto, J., Anneville, O., & Druart, J. C. (2009). Phytoplankton productivity increased in Lake Geneva despite phosphorus loading reduction. Journal of phytoplankton research 31, 1179-1194.

USGS. (2016, June 10). EarthExplorer. Retrieved from http://earthexplorer.usgs.gov/

USGS. (2016, June 10). Landsat Missions. Retrieved from http://landsat.usgs.gov//about_project_descriptions.php

Xue, K., Zhang, Y., Duan, H., Ma, R., Loiselle, S., & Zhang, M. (2015). A Remote Sensing Approach to Estimate Vertical Profile Classes of Phytoplankton in a Eutrophic Lake. Remote Sens. 7, 14403-14427.

43

8. Appendix

A1. In situ data used for validation of satellites’ images outcomes

Date Station Coord. Depth Chl.a PO4 NO3

(swiss system) (m) (µg/L) (mg/L) (mg/L)

10.08.2015 GE3 506.100/128.040 0 4.34 0.011 0.296

10.08.2015 GE3 506.100/128.040 -10 21.29 0.009 0.327

10.08.2015 GE3 506.100/128.040 -20 11.18 0.011 0.626

10.08.2015 GE3 506.100/128.040 -30 8.57 0.011 0.789

10.08.2015 GE3 506.100/128.040 -40 3.32 0.010 0.837

10.08.2015 GE3 506.100/128.040 -50 1.67 0.013 0.795

10.08.2015 GE3 506.100/128.040 -60 1.48 0.015 0.787

Date Station Coord. Depth Chl.a PO4 NO3

(swiss system) (m) (µg/L) (mg/L) (mg/L)

03.08.2015 SHL2 534.700/144.950 0 1.90 0.011 0.336

03.08.2015 SHL2 534.700/144.950 -10 5.49 0.012 0.334

03.08.2015 SHL2 534.700/144.950 -25 8.59 0.014 0.186

03.08.2015 SHL2 534.700/144.950 -60 0.40 0.013 0.735

03.08.2015 SHL2 534.700/144.950 -100 0.28 0.020 0.773

03.08.2015 SHL2 534.700/144.950 -150 0.27 0.031 0.791

03.08.2015 SHL2 534.700/144.950 -200 0.35 0.047 0.654

44

A2. Available scenes of Sentinel-2 in August 2015

46

A3. Available scenes of Landsat 7 and 8 in August 2015

47

A4. Summary details of Sentinel-2A scene selected Summary

Date: 2015-08-29T10:37:05.000Z

Filename:S2A_OPER_PRD_MSIL1C_PDMC_20160408T105241_R108_V20150829T103705_20150829T103705.SAFE

Identifier:S2A_OPER_PRD_MSIL1C_PDMC_20160408T105241_R108_V20150829T103705_20150829T103705

Instrument: MSI

Satellite: Sentinel-2

Size: 6.77 GB

Product

Cloud cover percentage: 9.188053333333333

Degraded MSI data percentage: 0

Degraded ancillary data percentage: 0

Footprint: <gml:Polygon srsName="http://www.opengis.net/gml/srs/epsg.xml#4326"

xmlns:gml="http://www.opengis.net/gml"> <gml:outerBoundaryIs> <gml:LinearRing>

<gml:coordinates>46.94616825397537,4.314035280374384 46.90124657599177,4.31296436636968 ……

46.94616825397537,4.314035280374384</gml:coordinates> </gml:LinearRing> </gml:outerBoundaryIs> </gml:Polygon>

Format: SAFE

Format correctness: PASSED

General quality: PASSED

Generation time: 2016-04-08T10:52:41.000132Z

Geometric quality: PASSED

Ingestion Date: 2016-04-08T16:30:20.690Z

JTS footprint: POLYGON ((4.314035280374384 46.94616825397537,4.31296436636968

46.90124657599177,……46.94616825397537))

Mission datatake id: GS2A_20150829T103026_000963_N02.01

Orbit number (start): 963

Pass direction: DESCENDING

Processing baseline: 02.01

Processing level: Level-1C

Product type: S2MSI1C

Radiometric quality: PASSED

Relative orbit (start): 108

Sensing start: 2015-08-29T10:37:05.000Z

Sensing stop: 2015-08-29T10:37:05.000Z

Sensor quality: PASSED

Instrument

Instrument abbreviation: MSI

Instrument mode: INS-NOBS

Instrument name: Multi-Spectral Instrument

Platform

NSSDC identifier: 2015-000A

Satellite name: Sentinel-2

Satellite number: A

48

A5. Summary basic details of Landsat 7 ETM+ scene selected

Landsat Scene Identifier LE71960282015241NSG00

Sensor Mode BUMPER

Station Identifier NSG - Neustreliz, Germany

Day/Night DAY

WRS Path 196

WRS Row 028

Date Acquired 2015/08/29

Start Time 2015:241:10:22:42.3251874

Stop Time 2015:241:10:23:09.0791250

Date L1 Generated 2015/08/29

Image Quality VCID 1 9

Image Quality VCID 2 9

Processing Software Version LPGS_12.6.1

Calibration Parameter File L7CPF20150701_20150930.04

Cloud Cover 0.25

Cloud Cover Quad Upper Left 0.23

Cloud Cover Quad Upper Right 0.26

Cloud Cover Quad Lower Left 0.22

Cloud Cover Quad Lower Right 0.31

Sun Elevation 50.0541954

Data Type Level 1 ETM+ L1T

Sun Azimuth 150.61244202

Full Aperture Calibration N

Gain Band 1 L

Gain Band 2 L

Gain Band 3 L

Gain Band 4 L

Gain Band 5 L

Gain Band 6 VCID 1 L

Gain Band 6 VCID 2 H

Gain Band 7 L

https://lta.cr.usgs.gov/landsat_dictionary.html#entity_id

https://lta.cr.usgs.gov/landsat_dictionary.html#sensor_mode

https://lta.cr.usgs.gov/landsat_dictionary.html#distribution_site

https://lta.cr.usgs.gov/landsat_dictionary.html#day_or_night

https://lta.cr.usgs.gov/landsat_dictionary.html#wrs_path

https://lta.cr.usgs.gov/landsat_dictionary.html#wrs_row

https://lta.cr.usgs.gov/landsat_dictionary.html#acquisition_date

https://lta.cr.usgs.gov/landsat_dictionary.html#start_time

https://lta.cr.usgs.gov/landsat_dictionary.html#stop_time

https://lta.cr.usgs.gov/landsat_dictionary.html#date_l1_generated

https://lta.cr.usgs.gov/landsat_dictionary.html#image_quality_f1

https://lta.cr.usgs.gov/landsat_dictionary.html#image_quality_f1

https://lta.cr.usgs.gov/landsat_dictionary.html#processing_software_version

https://lta.cr.usgs.gov/landsat_dictionary.html#cpf_name

https://lta.cr.usgs.gov/landsat_dictionary.html#cloud_cover

https://lta.cr.usgs.gov/landsat_dictionary.html#cloud_cover_quadrant_upper_left

https://lta.cr.usgs.gov/landsat_dictionary.html#cloud_cover_quadrant_upper_right

https://lta.cr.usgs.gov/landsat_dictionary.html#cloud_cover_quadrant_lower_left

https://lta.cr.usgs.gov/landsat_dictionary.html#cloud_cover_quadrant_lower_right

https://lta.cr.usgs.gov/landsat_dictionary.html#sun_elevation

https://lta.cr.usgs.gov/landsat_dictionary.html#data_type_l1

https://lta.cr.usgs.gov/landsat_dictionary.html#sun_azimuth

https://lta.cr.usgs.gov/landsat_dictionary.html#full_aperture_calibration

https://lta.cr.usgs.gov/landsat_dictionary.html#image_gain_bd_1








49

Browse Exists Y

Data Category NOMINAL

Gap Phase Source DE

Gap Phase Statistic 15.145389

Elevation Source GLS2000

Output Format GEOTIFF

Ephemeris Type PREDICTIVE

Panchromatic Lines 14281

Panchromatic Samples 16161

Reflective Lines 7141

Reflective Samples 8081

Thermal Lines 7141

Thermal Samples 8081

Ground Control Points Model 181

Geometric RMSE Model 4.74

Geometric RMSE Model X 2.683

Geometric RMSE Model Y 3.908

Map Projection L1 UTM

Datum WGS84

Ellipsoid WGS84

UTM Zone 31

Grid Cell Size Panchromatic 15

Grid Cell Size Reflective 30

Grid Cell Size Thermal 30

Scan Gap Interpolation 2

Orientation NORTH_UP

Resampling Option CUBIC_CONVOLUTION

https://lta.cr.usgs.gov/landsat_dictionary.html#browse_exists

https://lta.cr.usgs.gov/landsat_dictionary.html#data_category

https://lta.cr.usgs.gov/landsat_dictionary.html#gap_phase_source

https://lta.cr.usgs.gov/landsat_dictionary.html#gap_phase_statistic

https://lta.cr.usgs.gov/landsat_dictionary.html#elevation_source

https://lta.cr.usgs.gov/landsat_dictionary.html#output_format

https://lta.cr.usgs.gov/landsat_dictionary.html#ephemeris_type

https://lta.cr.usgs.gov/landsat_dictionary.html#panchromatic_lines

https://lta.cr.usgs.gov/landsat_dictionary.html#panchromatic_samples

https://lta.cr.usgs.gov/landsat_dictionary.html#reflective_lines

https://lta.cr.usgs.gov/landsat_dictionary.html#reflective_samples

https://lta.cr.usgs.gov/landsat_dictionary.html#thermal_lines

https://lta.cr.usgs.gov/landsat_dictionary.html#thermal_samples

https://lta.cr.usgs.gov/landsat_dictionary.html#ground_control_points_model

https://lta.cr.usgs.gov/landsat_dictionary.html#geometric_rmse_model

https://lta.cr.usgs.gov/landsat_dictionary.html#geometric_rmse_model_x

https://lta.cr.usgs.gov/landsat_dictionary.html#geometric_rmse_model_y

https://lta.cr.usgs.gov/landsat_dictionary.html#map_projection_l1

https://lta.cr.usgs.gov/landsat_dictionary.html#datum

https://lta.cr.usgs.gov/landsat_dictionary.html#ellipsoid

https://lta.cr.usgs.gov/landsat_dictionary.html#utm_zone

https://lta.cr.usgs.gov/landsat_dictionary.html#grid_cell_size_panchromatic

https://lta.cr.usgs.gov/landsat_dictionary.html#grid_cell_size_reflective

https://lta.cr.usgs.gov/landsat_dictionary.html#grid_cell_size_thermal

https://lta.cr.usgs.gov/landsat_dictionary.html#scan_gap_interpolation

https://lta.cr.usgs.gov/landsat_dictionary.html#orientation

https://lta.cr.usgs.gov/landsat_dictionary.html#resampling_option

50

A6. Illustration of basic steps of images preparation in ArcMap 10.3

1) We created a geodatabase CGEOM Leman SMakri.gdb and created a shapefile

cut_leman.shp. We keep the same coordinate projection system WGS_1984

We used the “Extract by Mask” tool to cut from each band layer our zone of interest:

Landsat band layer ex.

51

2) Radiometric enhancements of greys.

Band 1 Sentinel ex.

Before After

3) We inserted 2 coordinates points for each site of sampling GE3 and SHL2 as a table

and then save them as shapefiles

In ArcGIS we use the excel to table tool to convert the excel to a table in ArcGIS

52

Then we go to File-> Add Data-> Add x,y Data and we define what is x and what Is y. we

note that the coordinate system is correct and click ok

This creates two point layers. We right click on the points and Data-> Export data

Save as shapefile

53

4) Buffer creation around the 2 sites GE and SHL2 in the lake to be used after for the

zonal statistics

Ex. 10 m buffer for SHL2 sampling point

54

A7. Realization of color composites in ArcMap 10.3

Ex. of color composition 4-3-2 for the Sentinel-2A scene:

55

A8. Realization of band ratios in ArcMap 10.3

Ex. of band ratio 4/1 for the Landsat 7 scene:

Ex. of zonal statistics implementation for 4/1 band ration around the GE3 sampling point of

the Landsat 7 scene.

56

A9. NDVI calculation in ArcMap 10.3

Ex. NDV I calculation of Landsat 7 scene:

Bands to use 3 (red) and 4 (NIR)

In ArcGIS: Windows -> Image analysis

In the layers cascade we choose (in blue) layers 3 and 4

57

We also check the option scientific output so we can have values from -1 to 1

Following we use the leaf icon

58

A10. Chlorophyll index calculations in ArcMap 10.3

NDBIRIS calculation for the Sentinel-2A scene using the “Raster Calculator” tool in ArcMap

10.3

59

A11. Overview of Zonal Statistics tables resulted in ArcGIS for the Sentinel-2A scene.

Zonal statistics S2A GE3 sampling point:

Zonal statistics S2A SHL2 sampling point:

60

A12. Overview of Zonal Statistics tables resulted in ArcGIS for the Landsat 7 ETM

scene.

Zonal statistics ETM GE sampling point:

Zonal statistics ETM SHL2 sampling point:

Sentinel-2 and Landsat-7 satellite images qualitative ...

Documents

Transcript of Sentinel-2 and Landsat-7 satellite images qualitative ...