Applications of in situ HPLC

The first in-situ HPLC was developed under my dissertation research through collaboration between Georgia Tech and Analytical Instrument Systems, Inc. for the purpose of in situ measurement of marine and sediment interstitial waters (Beckler, 2014; Beckler et al., 2014).  Because traditional anion chromatography requires complex instrumentation, it is not very suitable for deployment in situ.  Therefore, we developed an HPLC technique using UV detection that has been employed by the in situ chromatograph configured both "sipping" from a sediment incubation chamber on a benthic lander to measure sediment fluxes, and also from a porewater profiling probe that is affixed to a micromanipulator to collect discrete sediment porewater samples as a function of depth alongside Au/Hg voltammetric microelectrodes.

The in situ HPLC operated on the benchtop, sipping interstitial samples from the sediment core and measuring concentrations of major marine anions.

Figure 1a) An overlay of chromatograms obtained using the Analytical Instrument Systems ISEA-IV-LC when sipping pore waters from sediments using the sample sipping probe; b) A corresponding sediment profile from a station 9 km upstream (salinity 0.63) from the mouth of the Altamaha River demonstrate the AIS ISEA-IV-LC can resolve all discrete redox layers  (Aerobic respiration, denitrification, Mn and Fe reduction) and salinity features (Bromide) using a combination of in situ voltammetry and HPLC. 

Brevetoxin production is increased under low nutrient regimes and in the presence of competing organisms, and toxins may persist long after blooms have dissipated. On the other hand, toxins are not always present when cells are present (Abraham et al., 2006; Twiner et al., 2007).  Public health decisions such as closing a shellfishery are often made at a threshold cell concentration of 5,000 cell/L, even though brevetoxin may be below detection. A need exists to measure brevetoxin concentration routinely, rapidly, and autonomously, to provide economical data for management, and to enhance our understanding of the coupling between K. brevis blooms and toxin production. We have begun work at MML to modify the in situ High Performance Liquid Chromatograph (HPLC) for brevetoxin measurement. Mote is currently funded by the State of Florida to develop an analytical method for brevetoxin detection appropriate for use on a similar in situ system. Briefly, HPLC will be used to physically separate the molecules of interest from the bulk of seawater constituents, and UV absorbance measurements used to quantify the separated molecules.  Brevetoxins are normally measured using LC/MS, allowing for both the analytes retention time on the LC column and its m/z ratio to be used for identification (Hua et al., 1995).  Instead, the new rapid screening method will use HPLC retention time and UV/VIS spectral fingerprinting (similar to how the Optical Phytoplankton Discriminator can identify phytoplankton) to identify and quantify brevetoxins. The instrument will autonomously collect water samples, filter, preconcentrate, quantify, and transmit data periodically throughout the deployment. Deployment at a fixed location in tandem with an OPD is scheduled in less than two years, and continuous brevetoxin and K. brevis cell concentration reporting to SO-COOL (coolcloud.mote.org) will become an additional feature of our HAB observatory.

Inherent Optical Property (IOP) determination

            While technologies to measure water clarity and CDOM in the field and from remote sensing have revolutionized in the past decades, coastal zones are under significant influences of tides, seasonal inflows, and sediment resuspension events.  As a result, optical properties in these Case II regions have high frequency and spatial variations not typically captured with daily remote sensing products (IOCCG, 2000).  There are few installations providing continuous IOP data, and even fewer providing a full suite of explicitly partitioned absorption and scattering properties, thus making it difficult to evaluate spatial and temporal heterogeneity, to determine long-term climatically-driven trends, and to evaluate the uncertainties due to assumptions inherent in standard remote sensing products when applied to these complex waters.
            Mote Marine Laboratory has designed a submersible, highly sensitive absorption instrument, the Optical Phytoplankton Discriminator (OPD; Figure 1) which has been in operation since 2005 to measure high-frequency CDOM (a_g) and particulate (phytoplankton and detrital; a_ph + a_d) absorption spectra 

Figure 1: Mote’s Optical Phytoplankton Discriminator allows for the autonomous collection of both raw and filtered absorption spectra in a long path-length liquid waveguide cell.

  at several coastal sites in Florida (Kirkpatrick and Hillier, 2007; Shapiro et al., 2014). Unattended OPD measure absorption spectra of raw (with the ability to concentrate particulates) and filtered (0.2 μm) samples using extended pathlengths and at user-selected frequencies.  The instrument provides a) full spectrum results (300-800nm); b) has an onboard calibration system; and c) has the ability to partition the particulate absorbance spectra (a_ph + a_d)  from the CDOM spectra (a_g) via a unique hollow fiber filtration scheme;  and d) using a 4^th derivative approach, separate the phytoplankton (a_ph) and the detrital (a_d) components (Figure 2).

Figure 2: Example spectra measured or derived from in situ OPD measurements of a coastal water. The total particulate spectrum was obtained by difference of the filtered measurements from the unfiltered measurements.  Then, the particulate spectrum was de-convoluted to obtain the particle backscattering (b_b) loss, and the detrital (a_d) and phytoplankton pigment (a_ph) absorption spectra using a novel 4^th order derivative and least squares regression technique.

These data have recently been statistically related to co-located measurements of relevant physical and biogeochemical parameters (e.g. salinity, temperature, Chl. a, FDOM, flow, and tide height from Sanibel Captiva Foundation “RECON” sensor suites), have the ability to detect high frequency events in both absorption and spectral slope (Figure 3), and are yielding novel insights into variation of and environmental controls on CDOM spectral absorbance. Covariate plots constructed from the time-series data in (Figure 4) demonstrate relationships of remarkable predictive utility permitting variable spectral slopes to be predicted given salinity and a single measurement of base spectral absorption.

Figure 3: Example time series of OPD CDOM absorption and spectral slope measurements (Mote) compared to salinity and FDOM measurements (SCCF) at the co-located OPD/RECON Sanibel site.  The lower time series is a zoomed section to demonstrate the ability for the instrument package to resolve high-frequency intertidal variations.

Figure 4: Covariate plots constructed from the time series in Figure 3 demonstrate that CDOM spectral slope evaluated between 350 and 400 nm can be better predicted from both a₄₄₀ and salinity than either measurement alone. This result suggests that for this particular site, remote sensing efforts can be improved by modeling CDOM spectral slope using the relationship derived from the bottom plot: S_350-440 = 0.779 / (salinity x a₄₄₀)

Remote estimation of CDOM and water clarity have evolved to employ sophisticated algorithms based on radiative transfer theory (Lee et al., 2002, 2005, 2009; Le and Hu, 2013). These algorithms generally work well in the open ocean where relationships of the optical properties of the in-water constituents are well understood (e.g., CDOM and phytoplankton co-vary, or CDOM spectral shapes and phytoplankton absorption shapes are well defined). Such relationships are assumed implicitly in the empirical algorithms and explicitly in the analytical or semi-analytical algorithms. However, such spectral relationships for CDOM (defined by the spectral slope) are highly variable in coastal environments in both space and time where CDOM is dominated by terrestrial sources, and can lead to significant errors when a fixed CDOM spectral slope is used. This situation is exacerbated across regions by the influence of land cover on CDOM spectral slope as demonstrated by Le et al. (2015) where several Gulf of Mexico coastal estuaries were compared. Adding to this complexity is the variable proportions between CDOM and particulate absorption and variable spectral slopes of particulate absorption, which all impact algorithm performance. Because water clarity is usually expressed by the diffuse light attenuation (K_d, m^-1; Lee et al., 2015) and K_d is dominated by total absorption (in coastal waters, total absorption is often dominated by CDOM), the accurate full spectrum estimation of CDOM absorption is critical in deriving overall water clarity from remote sensing.

Given the need for accurate CDOM and water clarity data for estuaries and coastal waters and difficulties in measuring these properties continuously using either typical field techniques or remote sensing, this project seeks to leverage existing infrastructure to enhance and automate high frequency IOP measurements, link in situ full spectrum CDOM and particulate characterization to more routine proxy measurements in Case 2 waters (e.g. salinity, FDOM, tide height, river flow), and to refine existing remote sensing algorithms incorporating the enhanced measures of particulate and dissolved absorption and spectral slopes to derive and provide CDOM and water clarity data products of enhanced accuracy.

Figure 5: Continuous measurements of filtered a₄₄₀ (left) and S_350-400 (right) obtained from an OPD integrated into a flow-through system of a ship demonstrate the ability of the instrument to spatially map variations of the inherent optical properties of water.