Internship Report

20.August - 12. October 2018

Jorrit Scholze

Curtin UniversityRemote Sensing and Satellite Research Group

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Thetis Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Measurement Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3.1 Remote Sensing Reflectance . . . . . . . . . . . . . . . . . . . . . . . . . . 43.2 Conductivity, Temperature, and Pressure . . . . . . . . . . . . . . . . . . 53.3 Fluorescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63.4 Measurement Relationships . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4 Match Up Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84.1 Remote Sensing Reflectances . . . . . . . . . . . . . . . . . . . . . . . . . 9

5 Chlorophyll . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1

List of Figures

1 Geographical location of the deployed Thetis in Western Australia . . . . . . . . 32 Scatterplots of various calibration methods, (a) RRS1 vs. RRS4, (b) RRS4 vs.

RRS5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Daily RRS spectra of (a) 16-09-2017 and (b) 08-03-2018 . . . . . . . . . . . . . . 54 Temperature and Salinity profiles in 2017 and 2018 over depth and time . . . . . 55 Fluorescence profiles in 2017 and 2018 over depth and time . . . . . . . . . . . . 66 Scatterplots of ALH and Fluorescence in (a) 2017 and (b) 2018 . . . . . . . . . . 77 Various water parameter and constituents for 2018 over depth and time . . . . . 88 Temporal matching Thetis and OLCI Level 2 RRS . . . . . . . . . . . . . . . . . 99 Daily RRS spectra of (a) 16-09-2017 and (b) 08-03-2018 for OLCI and Thetis data 10

List of Tables

1 Used ocean quality and science flags for masking pixel . . . . . . . . . . . . . . . 8

2

Internship Report, J. Scholze

1 Introduction

The Thetis buoy is an innovative multi-parametric moored underwater profiler that provideskey information on phytoplankton primary productivity (PP), phytoplankton blooms, and waterquality in the coastal waters between Rottnest Island and the Rottenest Island canyon in West-ern Australia (Figure 1). The aim is to better understand changes in phytoplankton PP andabundances busing observations from the equipped instruments one to several times a day fromthe surface to the seafloor. Furthermore a comparison to earth-orbiting ocean color satellitesdata is a purpose of the Thetis to provide a better understanding of the coastal environmentsof Western Australia to the Australian research community and pursue the research of oceancolor remote sensing with the help of ground truth data.

Figure 1: Geographical location of the deployed Thetis in Western Australia

2 Thetis Instruments

The Thetis is a submersible vertically profiling platform for use in coastal marine environments.The purpose f the platform is to sample the water column distributions of physical, biological,chemical and optical properties at a fixed geographical location. The Thetis used for this projectis equipped with a unique physical and optical sensor payload, including:

• Seabird Conductivity, Temperature, and Pressure Sensor (SBE 49 FastCAT)

• 2x WET Labs ECO Triplet BB2FLs

– Backscattering measurements at 470nm, 532nm and 700nm

– 370nm/460nm (CDOM)

– 470nm/695nm (Chlorophyll)

• WET Labs AC-S hyperspectral reflective tube spectrophotometer

– a and c at 400-730nm ( 4nm intervals)

Seite 3 von 10

Internship Report, J. Scholze

• 2 x Satlantic HyperOCR (Ed, Lu) for profiling radiometry

• Seabird Dissolved Oxygen SBE43

3 Measurement Data Analysis

3.1 Remote Sensing Reflectance

Remote sensing reflectance (RRS) from the Thetis Satlantic HyperOCR sensors was estimatedusing different exponential extrapolation methods for downwelling irradiance. Five calibrationmethods were applied for the extrapolation:

1. denotes surface tilt-filtered surface irradiance (ES) data, with Diffuse Attenuation Co-efficient (Kd) derived from single parameter ES fit. Z90 weighted exponential upwelligradiance (LU ) fit

2. denotes exponential extrapolation using full cast subsurface and surface ES where appli-cable, with Kd derived from double parameter exponential fit. Z90 weighted exponentialLU fit

3. denotes exponential extrapolation using Z90 data (subsurface downwellig irradiance (Ed)and surface ES where applicable), with Kd derived from double parameter exponentialfit. Z90 weighted exponential LU fit

4. denotes exponential extrapolation using full subsurface Ed data only. Z90 weighted expo-nential LU fit

5. denotes exponential extrapolation using Z90 subsurface Ed only. Z90 weighted exponentialLU fit

The scatterplots in figure 2 show selected comparisons between the different methods.Whereas (a) RRS1 and RRS4 show deviations with a flatter trend, the comparison between(b) RRS4 and RRS5 show a high correlation between each other. These results can be trans-ferred to other comparisons. RRS1, RRS2 and RRS3 show higher similarities and RRS4 andRRS5 likewise. This can be explained due to the use of the calibration methods and the maindifferences between them.

Figure 2: Scatterplots of various calibration methods, (a) RRS1 vs. RRS4, (b) RRS4 vs. RRS5

Seite 4 von 10

Internship Report, J. Scholze

Furthermore daily RRS plots were generated for detecting outlying spectra. In figure 3 thedifferences between the calibration methods can be observed. For some days the extrapolation ofthe RRS failed because of corrupt data. These spectra were neglected in the further processing.

Figure 3: Daily RRS spectra of (a) 16-09-2017 and (b) 08-03-2018

3.2 Conductivity, Temperature, and Pressure

Data derived from the Seabird Conductivity, Temperature, and Pressure Sensor (SBE 49 Fast-CAT) was processed without calibration. Measured profiles were filtered for corrupt data andevaluated over time and depth. For the filtering a moving window was applied removing outliersevery 10m of depth throughout the profile. 2017 had three profiles a day at dusk, midday anddawn and in 2018 just one profile at midday. The temperature and salinity distribution can beseen in figure 4. In general a higher temperature and salinity can be seen in 2018, due to thesummer months. No optical correlation can be observed between both parameters.

Figure 4: Temperature and Salinity profiles in 2017 and 2018 over depth and time

Seite 5 von 10

Internship Report, J. Scholze

3.3 Fluorescence

Estimating the chlorophyll content in the water column two approaches were applied. The firstmade use of the ECO Triplet Fluorometer, a sensor which measures the fluorescence directly.The second approach was the determination of the Absorption Line Height (ALH) calculatedfrom the WET Labs AC-S spectrophotometer using bands centred at 651nm, 677nm and 700nm.

3.3.1 Fluorometer

For the fluorometer moving window to filter outliers was applied. The filtering made use ofthe quantiles of 10m sections of the profile. The remaining data was interpolated and can beobserved in figure 5. CTD similar results can be seen, with lower concentrations in 2017 andincreasing concentrations towards May 2018. Structures in depth show a increase with depthin October for 2017, for 2018 this structure can not be observed. Striking features can beobserved in mid November at 40m depth with a concentration of up to 1.6 µg/L. The highestconcentration in 2018 can be observed in mid April for all depths.

Figure 5: Fluorescence profiles in 2017 and 2018 over depth and time

3.3.2 Absoption Line Height

The determined ALH was analyzed in comparison to the fluorescence measured with the ECOTriplet Fluorometer. The AHL determination is effective in removing the contributions toabsorption by coloured dissolved organic matter and non-algal particles. Furthermore the AHLis not sensitive to incident irradiance, in particular non-photochemical quenching. For the AC-S Profiles from 2017 severe quality failure occurred due to rusting flow tube collars. Figure6 (a) shows the remaining data for three different day times. The analysis was performed fordifferent depths. A expected quenching effect for the midday data set was expected with lowerfluorescence values, but could not be recognized. Furthermore the CPG at 551nm was used asa fouling estimator and are considered in the analyses in the last two plots. They show strongdifferences. For CPG lower than 0.2 a linear relationship can be observed. Figure 6 (b) showsthe remaining data for 2018. Due to one profile a day only midday data is available. For alldepths, a linear relationship can be assumed. At all depths above 40m, the ALH is significantlyand positively related to the ECO measured fluorescence. Below 40m depth a deviation for alow AHL occurs. The CPG at 551nm does not show a major impact on the data set.

Seite 6 von 10

Internship Report, J. Scholze

Figure 6: Scatterplots of ALH and Fluorescence in (a) 2017 and (b) 2018

3.4 Measurement Relationships

As a last analysis various water constituents and parameter were optically compared to eachother to find possible correlations and interactions. All parameters were filtered and corruptdata was removed. Furthermore a interpolation was applied. Figure 7 shows temperature,salinity, fluorescence and backscattering at 700nm from the ECO Triplet. Single periods featurea striking high level of chlorophyll for all depths. Especially in march a low chlorophyll contentbelow 30m and in mid April a high chlorophyll content can be depicted. Increases or decreasesin temperature may explain the peaks or lows in chlorophyll. Generally the temporal patterns inparticulate backscattering coefficient at 700nm were similar to chlorophyll, however the depth-specific relationships within each profile demonstrated considerable variability. For a better

Seite 7 von 10

Internship Report, J. Scholze

understanding of the relationships further analysis can be performed.

Figure 7: Various water parameter and constituents for 2018 over depth and time

4 Match Up Analysis

For the comparison of the Thetis data sets to remote sensing products, Sentinel-3 Ocean andLand Colour Instrument (OLCI) Level-2 Ocean Colour full resolution products were used. Spa-tial and temporal Thetis matching products were selected and evaluated. The following flagswere used to mask out pixels. This combination of flags is the result of several tests on thequality of the match ups.

Table 1: Used ocean quality and science flags for masking pixel

Flag Name Flag Description

INVALID Instrument data missing or invalidLAND Clear sky land

CLOUD Cloudy pixelCLOUD AMBIGUOUS Potentially cloudy pixel

CLOUD MARGIN Margin around cloud flags of 4 pixelsSUSPECT Transmission errors means measurements may be unreliable

HIGHGLINT Glint risk highRISKGLINT Glint correction not reliable

ACFAIL Atmospheric correction is suspectHIGHRW High RW at 560nm or CASE2F raised

T865 > 0.15 AOT at 865nm higher than 0.15

An average of 3x3 and 5x5 pixels around the was applied. For further analysis the 3x3average was used.

Seite 8 von 10

Internship Report, J. Scholze

4.1 Remote Sensing Reflectances

Wavelength dependent scatterplots including both years are illustrated (Figure 8). All scatter-plots show a underestimation of RRS in the OLCI products. The bands at 400nm and 412nmparticularly outline this effect. At 442nm and 490nm significant underestimations by the OLCIproducts is evident during 2018. As expected no trend was evident in the match ups at 510nm.

Figure 8: Temporal matching Thetis and OLCI Level 2 RRS

A recurring pattern in the blue the OLCI spectra was observed as shown by the two sampledates, illustrated in figure 9. An unusual inflection is evident in the blue bands around 412nm,and was more prominent in matchups from 2018. The cause of the inflection remains unde-termined. Closer analysis to this inflection were made focusing on Aerosol Optical Thickness(AOT) at 865nm, geometry of the sensor and different flagging of the products. Except for theTSM NN and CHL NN no correlation to the error between Thetis and OLCI was found. Un-common negative values at wavelength 778.75nm and 865nm for the OLCI bands were detectedand may be a lead to a not proper working atmospheric correction. Due to a very low aerosolconcentration in the atmosphere at the coast of Western Australia a failure of the atmospheric

Seite 9 von 10

Internship Report, J. Scholze

correction may be possible. Furthermore suspicious products were reviewed optically and somefeatured a masking of flags, which were neglected during the processing (e.g. OC4ME failure).

Figure 9: Daily RRS spectra of (a) 16-09-2017 and (b) 08-03-2018 for OLCI and Thetis data

5 Chlorophyll

For further analysis a chlorophyll match up would be interesting to investigate. The OLCILevel 2 match ups already include the selected pixel values and can be used for this analysis.For the Thetis data a optically weighted chlorophyll concentration has to be calculated over theprofiles. Single attempts were made but no results were produced.

6 Conclusion

Taking everything into consideration a glimpse of the possibilities with the Thetis data setswere shown. For 2017 and 2018 optical, physical and biological properties were analyzed.Technical failures in deployed instruments led to decreased data information within the Thetisbut common structures over time and depth were observed (Summer and Winter differences).A linear relationship between the ALH and the Fluorometer could be observed. Furthermore amatch up analysis to ocean color remotes sensing products was processed. The results showedgood approximations with a unexpected inflection in the blue bands. Possible errors for thisinflection were outlined. A deeper analysis for the Thetis data can be use full and interestingand a match up comparison between remote sensing chlorophyll and Thetis fluorescence is anapproach for further analysis.

Attached Plots

Used and further plots are resided at:R : \ThetisProfiler − SLIV KM − SE03729\Jorrit\PLOTSFinal poster and scripts are resided at:R : \ThetisProfiler − SLIV KM − SE03729\JorritFinal match ups are resided at:R : \ThetisProfiler − SLIV KM − SE03729\Jorrit\MATCHUPS

Seite 10 von 10