• Radiometers calibration & characterization
  • Calibration
  • Stability
  • Immersion coefficient
  • Bio-fouling
• HPLC analyses
• Filter pad absorption
• Buoy data pre-processing
• AOPs from SPMR profiles
• AOPs from buoy data
• Sun photometer data

The absolute calibration of the SPMR and SMSR with respect to NIST-traceable standards is performed every six months in the Satlantic optics calibration laboratory. The aim of the absolute calibrations is to find a coefficient for each sensor to relate the raw instrument counts measured in an environment to the actual quantity of the parameter being measured in that environment. The method for doing this for radiance and irradiance sensors varies slightly and has been documented by Hooker et al. (2002).

For a radiance sensor (identified by ), the calibration coefficient is computed using a plaque (identified by ) with a calibrated reflectance, along with a standard lamp (identified by ) that has a calibrated irradiance, . As a general procedure, the lamp is required to be positioned on axis and normal to the center of the plaque at a distance d. Dark digital voltage levels are recorded with the radiance sensor capped and an average dark level, , is taken from these dark samples.

The radiance sensor position allows a 45^o view of the plaque with respect to the lamp illumination axis. Once the lamp is powered on, the voltage levels of each of the individual sensor channels are recorded. From these, an average calibration voltage for each channel, , is obtained. The calibration coefficient is calculated as:

where, d is given in centimeters.

For an irradiance sensor (identified by SID), the calibration coefficients are computed using an FEL standard lamp () with a calibrated irradiance cal . As a general procedure the lamp is required to be on axis and normal to the face plate of the irradiance sensor at a distance, d. Similar to the radiance sensor, dark voltage levels (digital) are recorded with the radiance sensor capped and an average dark level, , is taken from these dark samples.

With the lamp powered on, the voltage levels of each of the individual sensor channels are recorded. From these, an average calibration voltage for each channel, , is obtained. The calibration coefficient is calculated as:

where, d is given in centimeters.

Back to top

The methodology and equipment used for tracking the stability of our main field radiometers are now sketched out in the following three subsections.

Radiometer stability: The SQM-II
The SQM-II is a portable stable light source originally designed under the name SQM, standing for SeaWIFS Quality Monitor, by NASA and NIST (Johnson et al. 1998). The SQM-II has been redesigned and built by Satlantic, Inc. The SQM-II consists of a lamp housing and a power box, which are connected by a 5m cable. A serial port provides the capability of monitoring and controlling the system with a PC. An internal memory, an LCD display, and several buttons on the back also allow manual control and monitoring. Several status LEDs are also used to provide a quick method for determining the state of the instrument.

The SQM-II has two sets of eight bulbs each. There is no individual bulb control, a set is either completely turned on or off. The light output of the second set, called the high power bank, is about three times brighter than the light output of the first set, named the low power bank. They are often referred to as HiBank and LoBank respectively. These two lamp banks provide three illumination levels: low illumination when only LoBank is on, medium illumination when only HiBank is on, and high illumination when both lamp banks are on.

A sensitive internal detector at 490 nm is used to monitor the stability of the lamps. The outputs from this detector and other internal sensors measuring temperatures, voltages and currents are monitored by a 20-bit analog-to-digital converter.

The SQM-II setup is precisely described in the SQM-II user's manual edited by Satlantic.

Back to top

Radiometer stability: Methodology
To check the stability of the different radiometers in the field and to monitor the performance of the SQM-II itself, a calibration evaluation and radiometric testing (CERT) session and a data acquisition sequence (DAS) have been defined following the procedure described in Mueller et al. (2002). A CERT session is a sequence of DAS events, which are executed following a prescribed methodology. Each DAS lasts for 3min to represent enough data to statistically establish the characteristics of the instrument involved referred to as DUT, standing for Device Under Test. This records about 1,000 DUT records and 450 SQM-II records. A typical sequence of procedures for each CERT session is as follows:

1. The electronics equipment (lamp power supplies, the SQM-II fan and internal heater power supplies, lamp timers, etc.) is turned on at least 1 hour before the CERT session begins. The total number of hours on each lamp set are tracked by recording the starting and the ending of hours on each lamp set. Each lamp bank acts as a separate state machine. The state of a given lamp bank is independent of the state of the other bank.
2. The SQM-II lamp bank will remain in this coarse stability state for 1h, allowing enough time for the lamp and the system to thermally stabilize. It is not recommended that radiometer measurements be taken while in this state since the lamps have not yet reached their highest stability. After one hour, the lamp banks will automatically enter the fine stability state.
3. If the mixture of radiometers used in the CERT session changes over time, at least one radiometer (preferably two of different types, i.e. radiance and irradiance) should be used in all CERT sessions. This would be practically the case when the SPMR system and the buoy radiometers are not calibrated during the same CERT.
4. S. Hooker advises to keep the banks in the fine stability state for another hour before starting the CERT, particularly in highly variable environments. This advice was followed so far even if the CERT took place in a laboratory. The warm-up period can be considered completed when the internal SQM-II monitor data are constant within 0.1%. The radiometric stability usually coincides with a thermal equilibrium as denoted by the internal thermistors.
5. The insertion of a DUT into the shadow collar of the SQM-II light chamber has a small loading effect on the SQM-II due to its reflectivity, i.e., some light reflected back into the light chamber. This in turn affects what the radiometer sees. However the reflectivity of a field radiometer is constantly changing over time due to the normal wear and tear of use, altering the loading effect on the SQM-II monitor detector and the effective light field seen by the radiometer.
6. Hence, upon the completion of the warm-up period, SQM-II monitor data are collected for the black glass (radiance) and white (irradiance) fiducials successively. A fiducial is a non-functional DUT, whose reflective surface must be carefully maintained over time so that its reflectivity remains essentially constant. It is also important that the position of the fiducial in the compartment is always the same and this is ensured by a fixed collar allowing only one position.
7. Once these control measurements are completed, the individual radiometers (DUTs) are then tested sequentially. This begins with an 1=OCR-2001 that is designated specifically for use on the SQM-II and nothing else. Its use is restricted solely to SQM-II sessions to preserve the performance of its original calibration and, hence, serve as a verification of SQM-II stability between sessions.
8. Firstly, the DUT sensor head is inserted and secured in the SQM-II compartment. Data is collected over 3min. The DUT is secured using the three fasteners on the SQM-II which clamp on a collar that has been fastened to the DUT. This collar is at a specific distance from the sensor head and of a particular rotational position, determined by the flat side of the collar. This positioning of the flat part of the collar ensures that the DUT always enters the SQM-II in exactly the same way.
9. The DUT is removed from the SQM-II slot and the caps are placed over the sensor heads to block out all possible light to the them. The black fiducial is then placed in the 1=SQM-II1 and data is again collected from both 1=SQM-II1 and DUT to provide a dark reading for the DUT and stability check for the 1=SQM-II1 . Each time any file is recorded, the voltage at the SQM-II internal detector is noted.
10. Once all instruments have been tested, SQM-II monitor data are collected a second time for black, radiance and irradiance fiducials successively.
11. As soon as all measurements with a lamp bank are complete and its use is no longer required, the lamps are switched off to minimize aging and possible deteriation. Where an alternative lamp bank combination is required for testing DUTs at a different light intensity, 2h warm-up time is once again required to reach optimum stability.
12. The use of all three light intensity levels (Lo, Medium and Hi-Bank) was discontinued in 2003 because changes in SQM-II performance causing saturation of the internal detector of the SQM-II during Hi-Bank operation have led to the omittance of Hi-Bank measurements.
13. Before the SQM-II is finally shut down and the CERT session completed, after the lamps are powered down, the ending number of hours on each lamp set is recorded.

The SQM-II operation is precisely described step by step in the SQM-II User's Manual edited by Satlantic, although some changes have been made to our method to reduce the duration of each SQM-II session and to accommodate the aforementioned internal detector saturation during Hi-Bank operation.

Back to top

Radiometer stability: Measurement schedule
The objective of the scheduling of the SQM-II operations is to arbitrarily monitor variability and drift of the optics sensors during the period of use between the near-biannual absolute calibrations which are performed by Satlantic at their Halifax, Nova Scotia facility. The data collected in these SQM-II sessions aids in the reduction of error in the field data caused by changes in the performance of the sensors between the times of absolute calibrations. When differences occur between these calibrations, it cannot be assumed that this change has been a gradual and linear process during the period of activity.

As soon as possible after the radiometers are received in Villefranche after shipment by Satlantic from the absolute calibration in Canada, they are tested on the SQM-II in the optics laboratory in Villefranche, using the method described in section 6.2.2.

Once the sensors have been active in the field, either fixed on the buoy or profiled from the ship, they are once again subjected to testing on the SQM-II. The exact scheduling of this testing depends upon the duration of their activity. For the buoy radiometer heads, which are intended to be exchanged on a monthly basis, the SQM-II session is performed on them as soon as possible after they have been taken off the buoy and returned to the lab, once they have been wiped down with soapy water and a soft cloth. For the SPMR and SMSR, the SQM session is performed after each cruise, assuming that the instrument has been used.

Back to top

Extraction procedure
Each filter is placed in 3 mL of methanol (HPLC grade) containing an internal standard^*. After 30 minutes at -20°C the filters are disrupted by ultrasonication using an ultrasonic probe and returned to the freezer. Another 30 minutes later, the sample is clarified through a 25 mm GF/C Whatman filter (1.2 µm porosity). The filtrate is finally stored at -20°C until analysis (within 24 hours).

HPLC analysis
All parts of the HPLC system at the LOV are Agilent Technologies products:

• A degasser (1100 model);
• A binary pump (1100 model);
• An autosampler6 (1100 model) with temperature control (4^oC) and automatic injection (Rheodyne valve) for mixing the sample with the ammonium acetate (1N) buffer;
• A diode array detector (1100 model) with measurements at 440nm (for carotenoids), 667nm (for chlorophylls and degradation products) and 222nm (for Vitamin E Acetate internal standard); and
• A fluorescence detector with excitation and emission wavelengths respectively at 417 and 670nm.

Column temperature is maintained at 25^oC and the injection volume is 200 μL.

The analytical method is based on a gradient separation between a methanol:ammonium acetate (70:30) mixture and a 100% methanol solution, comparable to that described by Vidussi et al. (2000), but with a few differences allowing for improvement of sensitivity and peak resolution. Modifications to this method to separate certain peaks and increase sensitivity included a) a flow rate of 0.5mLmin^-1, and b) a reversed phase chromatographic C₈ column with a 3mm internal diameter (Hypersil MOS 3μm). The gradient used is presented in Table 3.

The different pigments that are quantitated are presented in Table 4. An example of a chromatogram is shown in Fig. 14.

Back to top

Table 7.2: **Replace with new Table 4. The list of pigments detected by HPLC at the LOV, their detection wavelengths, and their possible coelution with another pigment. The variable forms, which are used to indicate the concentration of the pigment or pigment association, are patterned after the nomenclature established by the SCOR Working Group 78 (Jeffrey et al. 1997). Abbreviated forms for the pigments are shown in parentheses.

Example of a chromatogram containing all major pigments from a surface sample at the BOUSSOLE site in April 2003. Numbers refer to pigment names in Table 4.

Back to top

Calibration
Response factors for 8 pigment standards provided by DHI  (International Agency for C₁₄ determination, Denmark) are determined by spectrophotometry (Perkin Elmer) followed by HPLC analysis:

Quantification
The Agilent Technologies Chemstation software is used for conducting the analysis as well as for post-analysis processing. This includes peak integration and spectral identification. Peak identification is manually verified by retention time comparison and observation of the absorption spectra. Quantification is based on peak area related to the specific extinction coefficient and concentrations are given in milligrams per cubic meter. When two pigments tend to co-elute, their identification is first done spectrally then they are summed, e.g., chlorophyll c₁+c₂, or total chlorophyll b (chlorophyll b plus divinyl chlorophyll b).

Validation

A solution of methanol containing the internal standard is injected every 12 samples during the analytical sequence. The average of these control analyses provides the reference peak area for the internal standard. The standard deviation of these analyses provides information about the precision as well as the stability of the instrument during the analytical sequence. Total chlorophyll a values are compared to the fluorescence signal from CTD bottle data in order to detect possible inaccuracies. Total chlorophyll a values are also compared to particulate absorption measurements which have been carried out on the same filters just before pigment extraction (Sect. 7.2).

Method performance
This HPLC analytical method has proved to be particularly sensitive, with detection limits of approximately 0.001 mgm^-3 and a good resolution between divinyl chlorophyll a and chlorophyll a which allow it to be particularly adapted to the oligotrophic waters of the Mediterranean.

The precision of the method is generally characterized by a variation coefficient between 0.3-1.0%. Furthermore, two international intercomparison exercises (SeaHARRE-1 (Hooker et al. 2000) and SeaHARRE-2) have demonstrated that the LOV produces satisfactory results for the analysis of chlorophyll a and accessory pigments in seawater samples from different concentration regimes.

-----------------------------------------

^*Samples before March 2003 were analysed with trans-β-apo-8’-carotenal as internal standard then replaced with Vitamin E Acetate.

Back to top

Laboratory analyses
The protocol for the analysis of spectral absorption coefficients of particulates is based on those recommended in chapter 15 of the NASA Ocean Optics Protocols for Satellite Ocean Color Sensor Validation paper (Mitchell et al, 2002).

Sampling is performed in the same way as for the HPLC pigments as described in Sect. 8.1 where the seawater samples are collected from Niskin bottles and filtered through 25mm GF/F Whatman filters (0.7 porosity). In most cases 2.8L are filtered for each sample and the filters placed in Petri slides, wrapped in aluminium and stored first in liquid nitrogen on board then in a -80^oC freezer in the laboratory. The sample filters used for the analysis are also used afterwards for HPLC analysis. This method has been tested to ensure that using the same filter for the two methods does not compromise either measure.

The spectrophotometer used is the Perkin Elmer Lambda 19 (L19) Dual Path Spectrometer with an integrating sphere compartment attached. The instrument scan range is set from 750 to 350nm with a scan interval of 1nm. After a 1h warm-up period, the machine is ready to use.

The samples are taken out of the -80^oC freezer and stored temporarily in an ice-cooled cold box. Each sample, in its petri dish, is then placed in a dark box on the bench to defrost for 5min. This defrosting period is kept to a minimum because the pigment composition once thawed becomes susceptible to degradation which could affect the analyses, particularly the HPLC.

The L19, being a dual path spectrometer, has a slot for a reference material and another for the sample. These slots are actually the outer walls of the integrating sphere at the two positions where the two beams enter the sphere. For the reference path, rather than use a blank filter, which can create variations in the spectra attributable to differences in the level of moistness of the filter, the pathway is left open (without any filter blank) thus using an air blank. In the sample material slot, a blank GF/F filter is used. This blank has been soaked in distilled water for 12h and had excess water drained off before being placed in the sample slot.

An autozero is performed with the blank filter in the sample path and the reference path open, which sets the absorbency values measured at each wavelength to zero. As a verification of this and to provide a baseline, the blank is then scanned to measure absorbency. The spectra produced from this should be flat and close to zero (within 0.005). If this is not the case, the autozero is repeated and the blanks scanned again.

Once satisfactory baselines have been achieved, the sample filters can then be scanned. For baselines and samples, each filter, scanned twice, should result in two similar replicates. If replicates do not match well, the filter position should be checked and the sample scanned again. Drying of the filter between replicates may cause some vertical offsetting. This is still acceptable if the spectral shape for each replicate is similar.

The spectra for each sample are saved as a file in American Standard Code for Information Interchange (ASCII) format with two columns: wavelength and absorbency. The mean of the absorbency values of replicates for each sample is calculated and any offset from zero removed by subtracting the absorbency value at 750nm from the whole spectrum.

Data postprocessing
The optical densities are then corrected for any nonzero signal in the NIR (750nm), and transformed into absorption coefficients by accounting for the so-called β effect. The total particulate absorption spectra are then numerically decomposed into a phytoplankton and a detritus absorption spectra following Bricaud and Stramski (1990).

Back to top

The SPMR is a central element in our program. It is providing reference data against which the data collected from the buoy can be checked, and it is providing also the information on the vertical that is not provided by the buoy surface measurements.

Measurement suite
What is measured in the water column is the upwelling irradiance, E_u(z, λ), at 13 wavelengths from 412-865nm, plus the downward irradiance at the same wavelengths, E_d(z, λ). The latter is not used in the computation of the reflectance.

The above-water downward irradiance, E_d(0⁺, λ), 0⁺ indicates a height immediately above the surface), which is often referred to as E_s(λ), is recorded on deck (at the bow of the ship), again at the same 13 wavelengths.

Corrections and extrapolations
From the vertical profile of E_u(z, λ), the upwelling irradiance at null depth, i.e., immediately below the sea surface, is obtained as (λ is omitted for brevity) :

where z is depth (0^- indicates a height immediately below the surface amd is referred to as the null depth), is the shallowest value of for which the tilt is less than 2^o, and K_u is the attenuation coefficient for the upwelling irradiance computed from the measurements of collected at different depths between and m.

Several interpolation procedures designed to derive the E_u(0^-) from the vertical profile of have been tested against true values of E_u(0^-) (i.e., values directly measured below the sea surface by installing the radiometer on a floating frame), and the method that eventually provided the closest values to the true ones was selected. This experimental work, which is not further detailed here, is just mentioned to indicate that the contribution to the overall uncertainty budget of the extrapolation error has been minimized and is below 3% across the entire spectrum.

The above-water reference measurement, E_d(0⁺) is corrected to account for the loss of irradiance at the air-sea interface:

where the mean transmission of the sea surface for sky and sun irradiance, expressed by , is equal to 0.957 (3% according to atmospheric turbidity and sun elevation), and the internal reflectance (accounted for by , where is ), varies slightly with . With a mean value of 3% this term is equal to 0.985 (1.5% if varies between 0-6%). Assuming these two terms are constant,

where the value of E_d(0⁺) is obtained from the first 10s of recording starting after the release of the SPMR (this corresponds approximately to the upper 5m of the descent), to which a fit is adjusted in order to eliminate variations in E_d(0⁺) that are only due to the tilt of the SMSR (which is not installed on a gimbal). This technique provides similar results as compared to just picking the measurements taken for tilt angles less than 1^o.

Computed parameters
The reflectance R is then

Note that before the above ratio is formed, the E_u(0^-) is corrected for instrument self shading as per Gordon and Ding (1992). In this correction, the instrument radius, the total absorption coefficient, and the ratio between direct-sun and diffuse-sky irradiances (r_d) is computed following Gregg and Carder (1990).

Back to top

Numerous methods have been developed in the past years, which are dedicated to the retrieval of some kind of information about aerosol types, usually particle size distributions, refractive index, or phase functions from measurements of sky radiances collected either in the principal plane, following almucantars or in the solar aureole (e.g. Santer and Martiny 2001; Dubovik and King, 2000; Dubovik et al. 2000; Nakajima et al. 1983; 1996),

A similar method has been developed, which in addition uses the information provided by the degree of polarization, as measured at 870nm by the sky photometer. Using this additional piece of information in principle decreases the ambiguities.

In order to find the best candidate aerosol model that allows a reconstruction of the sky radiances distribution observed in the principal plane, the method either uses the trial-and-error principle or a more sophisticated inversion algorithm based on the use of a neural network. A radiative transfer code is used in these inversions, which is based on the successive orders of scattering method and uses the vector theory OSOA code (Chami et al. 2001).

As for the neural network, which has the advantage of allowing fast processing of time series, and which is able to properly deal with measurement errors, the important step is the training. This training has been performed by using the OSOA radiative transfer code (Deuze et al. 1989; Chami et al. 2001), and using aerosol models with varying index of refraction and a particle size distribution following a Jünge model. A realistic noise that accounts for instrumental errors is added on the synthetic data.