Effects of particle aggregation and disaggregation on their Effects of particle aggregation and disaggregation on their inherent optical properties inherent optical properties

: In many environments a large portion of particulate material is contained in aggregated particles; however, there is no validated framework to describe how aggregates in the ocean scatter light. Here we present the results of two experiments aiming to expose the role that aggregation plays in determining particle light scattering properties, especially in sediment-dominated coastal waters. First, in situ measurements of particle size distribution (PSD) and beam-attenuation were made with two laser particle sizing instruments (one equipped with a pump to subject the sample to aggregate-breaking shear), and measurements from the two treatments were compared. Second, clays were aggregated in the laboratory using salt, and observed over time by multiple instruments in order to examine the effects of aggregation and settling on spectral beam-attenuation and backscattering. Results indicate: (1) mass normalized attenuation and backscattering are only weakly sensitive to size changes due to aggregation in contrast to theory based on solid particles, (2) the spectral slope of beam-attenuation is indicative of changes in PSD but is complicated by instrument acceptance angle, and (3) the spectral shape of backscattering did not provide as clear a relationship with PSD as spectral beam attenuation, as is predicted by theory for solid spheres.

13. W. Hou, K. L. Carder, and D. K. Costello, "Scattering phase function of very large particles in the ocean," Proc.

Introduction
The scattering of light in aquatic environments is dominated by the effects of particulate material. The intensity and spectral characteristics of scattering depend strongly on the concentration, composition, and particle size distribution (PSD) of suspended matter. In many environments a large portion of suspended particulate material is packaged as aggregated particles [1], and the overall characteristics of the particulate matter pool are a result of multiple processes including resuspension, aggregation, and disaggregation [e.g., 2,3]. Aggregation and disaggregation affect changes in particle porosity and size, and the composition of an aggregate is remarkably dynamic, reflective of the heterogeneity of its physical, biological, and chemical environments, as well as to its role as a scavenger, gaining and losing material as it is transported throughout the water column [e.g., 4,5]. Components of marine aggregates include bacteria, organic and inorganic colloids, algal particles and associated detritus, mineral particles, as well as polymers, fibrils, and gels, originating biologically and abiotically. Despite the profound consequences of aggregation, there is no accepted framework to describe the effects of aggregation on the scattering properties of suspended particulate material, and the idealized model of the homogenous sphere remains dominant in the study of particle optical properties [6]. However, consideration of particle packaging is likely needed for the extension of optical methods into environments such as river plumes, bottom boundary layers, and phytoplankton blooms. The optical properties of aggregates have attracted substantial attention in disciplines other than oceanography, mostly relating to aerosols, interstellar dust, and colloids. Such studies are usually concerned with loose (diffusion limited) fractal aggregates constructed of submicron monomer particles smaller than the wavelength of incident light. Some of these algorithms invoke a (relatively) simple superposition of Rayleigh-Debye-Gans scattering for each monomer, ignoring internal scattering [e.g., 7], while others consider a rigorous Mie-based multiple scattering solution [e.g., 8]. Invariably, despite simplifying assumptions, these approaches are defeated by computational limitations for the particle sizes relevant to aquatic aggregates. In contrast, Latimer and Wamble [9] presented a model describing the scattering properties of aggregates whose component particles are somewhat larger than the wavelength of incident light. They hypothesize that light scattering due to a suspension of randomlyoriented aggregates caries only information about the overall size and porosity (void fraction) of the aggregates. Given this assumption, they then approximated the optical effects of aggregate structure on optical properties using models for randomly-oriented spheroids and coated spherical particles having equivalent gross volume and net mass as the aggregates. Results from their simple model and experimental data from suspensions of latex sphere aggregates agreed to first order, with some of the disagreement likely explained by inaccuracies inherent in the microscopic analysis of the aggregates [10].
There are few studies of the optical properties of marine aggregates in the laboratory or field. Early work by Carder and Costello [11] qualitatively considered the effects that aggregation could have on observational closure of optical properties by packaging mass into particles that are large and rare relative to the measurement sample volume. Costello et al. [12] examined variability of optical properties during a controlled diatom bloom mesocosm study and found beam-attenuation to be an excellent indicator of particulate organic carbon despite changes in PSD, and saw increase in the variance of optical properties as the diatom population aggregated. In a follow-up to their previous work, Hou et al.
[13] later used a specialized instrument to measure the scattering properties and PSD of marine snow particles greater than 280 µm throughout the water column and concluded that these large particles could contribute up to 20% of total scattering as well as enhance backscattering efficiency. More recently, Hatcher et al. [14] examined the optical backscattering of phytoplankton-drill mud aggregates created in the laboratory using an upwelling tank. During the course of the 37-day experiment, during which the aggregates formed and aged, the relationship between backscattering and projected cross-sectional area for particles greater than 10 µm in diameter remained constant. A subsequent experiment by Flory et al.
[15] observed backscattering and PSD (for particles greater than approximately 100 µm in diameter), and found evidence that the effect of large particles on backscattering has been underestimated. However, none of these studies independently measured particles less than 10 µm, which we would expect not only to be highly efficient scatterers, but also to be correlated with the concentration of aggregates because of the role of aggregates as particle scavengers.
The mass-specific optical properties of aggregates will differ from solid particles as a result of the fractal nature of an aggregate, where the large fluid fraction within the aggregate results in a cross-sectional area that is larger than that of a solid particle of the same mass. Previously [16], we examined the beam-attenuation of marine particles using Latimer's model [9,10] that approximates aggregate particle structure as an ensemble of hollow spheres and randomly-oriented ellipsoids, with aggregate porosity a function of size. Using a traditional homogenous sphere model, mass-specific beam-attenuation varied significantly as a function of changing PSD. However, with the aggregate model, we found mass-specific attenuation to be remarkably constant and consistent with observations of marine particles in the environment encompassing a wide range of particle sizes and composition [17].
Finally, two additional studies have considered potential effects of particle dynamic processes on bulk inherent optical properties. Boss et al. [18] examined the tight relationship between particulate beam-attenuation magnitude and spectral slope (an indicator of PSD slope), and found the two parameters to be consistent with resuspension and size-dependent settling in the bottom boundary layer for most of their data. Deviation from this tight relationship occurred on the sampling day following a passing hurricane, and the authors consider aggregation dynamics to be a possible explanation for their observations. More recently, Ackleson [19] used a simple model linking optical properties derived from Mie theory and changes in PSD expected from disaggregation and settling scenarios to examine Long Island Sound and Connecticut River plume data, finding that disaggregation was able to explain optical variability at the plume boundary. However, Ackelson also found that changes in spectral slope may also be explained by mixing between the two water masses, and concluded that the method of using spectral optical properties to examine particle dynamics requires additional research.
To further increase our understanding of the effects of particle aggregation on optical properties, we conducted an in situ manipulation experiment, measuring and comparing optical properties of the natural suspension and the natural suspension subjected to shear (in order to break aggregates). Using two Sequoia Scientific LISST-100 instruments (measuring near-forward scattering and beam-attenuation) [20], one open to the environment and the other employing a sample chamber and pump, this experiment allowed us to qualitatively examine the effect of aggregation on beam-attenuation.
A second experiment was conducted in the laboratory to further investigate the effects of packaging of particles into aggregates. In this experiment, clays were aggregated using salt and observed over time by a LISST-100X instrument, open-path WET Labs ac-9 [21] (measuring multi-spectral beam-attenuation), and a WET Labs ECO Triplet (measuring volume scattering function, VSF, at 117° at three wavelengths) [22], in order to examine the effects of increasing aggregate size on optical properties.

LISST PSD and beam-attenuation
The LISST-100 (Sequoia Scientific, Inc.) is an in-water instrument designed to measure PSD in the field. The LISST-100 infers PSD from the scattering of a red laser beam (670 nm) introduced into a sample volume (5-cm path-length). The beam is scattered by particulate material within the sample volume, and near-forward scattered light at angles ranging from approximately 0.075° to 14.9° is received by a Fourier lens and transformed onto a set of 32 logarithmically-spaced, co-planar, concentric photodetector rings. For large particles that scatter light more near-forward, the most inner of the concentric photodetectors respond. Conversely, as particle size increases, angular light scatter becomes less concentrated in the near-forward and response increases in photodetectors further from the center. In addition, photodetectors measure transmitted (0.0269° acceptance angle) and reference (beam-split) laser power in order to estimate beam-attenuation (670) pg c , where "pg" refers to combined particulate and dissolved components. Before each field or laboratory experiment, blank measurements (zscat) were made with the manufacturer-supplied software using Barnstead NANOpure water in the small volume chamber insert or by filling the laboratory sink with reverse osmosis (RO) water. In both cases the water was allowed to sit to reduce the effects of bubbles, and blank measurements were repeated (with instrument cleaning) until our zscat scattering patterns were comparable in shape but lower in amplitude than those supplied by the manufacturer. Raw scattering due to particulates could then be calculated by subtraction of the zscat from measurements made during the experiments [20,23]. Data processing and PSD inversion for the LISST was performed on un-binned data, and subsequently binned to five-minute intervals or burst-averaged.
Sequoia Scientific supplies an algorithm [20] to invert the angularly-resolved scattering pattern into a volume PSD ( ( ) i V D in units of μL/L) having 32 size classes with geometric mean diameters ( , {1,2 32} i D i  ) from approximately 1 μm to 184 μm. We used an updated version of the inversion kernel based on scattering by randomly-shaped natural particles [24,25]. In order to condense changes in PSD into a single parameter, we employ a weightedaverage particle size calculated as 32 32 where ( ) i A D is the areal PSD in suspended cross-sectional area per volume (m 2 m 3 ) for each LISST size class i, with mean diameter i D . Areal size distribution (m 2 /m 3 ) is calculated from volume size distribution (μL/L) by assuming spherical geometry:

ac-9 spectral attenuation measurements
The WET Labs, Inc. ac-9 [21] is a combination spectral beam transmissometer (0.93° acceptance angle) and reflecting-tube absorption meter, normally measuring absorption and beam-attenuation at nine illumination wavelengths in the visible spectrum (412-715 nm) by use of a rotating filter wheel in the light source. In the laboratory aggregation experiment we used a 10-cm path-length version and left the absorption tube sealed, but left the transmissometer-side open to the environment with no flow sleeve so that aggregates would fall through the illuminated sample volume undisturbed. The ac-9 was blanked in the laboratory sink using RO water, and particulate beam-attenuation was then calculated by difference of the experimental measurement and blank. Particulate beam-attenuation spectra from the ac-9 were fit to a power-law function of the form ( ) by unconstrained nonlinear optimization (MATLAB "fminsearch") [18,26] using all available wavelengths except 715 nm. For ac-9 data from the laboratory experiment, the percent difference of the fit residuals relative to measured data was in general less than 2%.

ECO Triplet volume scattering function measurements
A WET Labs, Inc. ECO Triplet BB-3 was used to measure the VSF at a fixed angle of 117°, (117 , )    , at three wavelengths (λ = 532, 660, 880 nm), with a sampling rate of ~1 Hz. The BB-3 was calibrated at the factory with 2-µm polystyrene microspheres in order to determine a scaling factor and dark offset, S and D, respectively. The calibration values S and D are used to determine ( Thus the effects of the dark offset, D, were subtracted out. Path-length attenuation correction was not performed since absorption measurements were not available. For typical environmental measurements with absorption below 1 m 1, error is expected to be small, 4% [22,29]. Assuming a single scattering albedo for particles of ~0.95 and a maximum particulate attenuation of ~10 m 1, we expect our particulate absorption was less than 1 m 1, however this remains a potential error in our estimates of (117 , ) The spectral shape of un-binned ( to have percent differences of greater than 25%, with an obvious trend across the wavelength channels, indicating that a power-law fit is not suitable to our measurements. Therefore, ratios of individual wavelength pairs were also considered in order to reduce the possible influence of calibration (slopes, S) errors; the channel ratios were transformed to an equivalent bb  ,

In situ disaggregation experiment
The qualitative effect of aggregation on beam-attenuation was observed by comparing the measurements of two similar LISST-100 (Type B) instruments deployed side-by-side with one having a mechanism to break aggregates prior to the sample being measured. The instruments were deployed in the same package at ~1 m above bottom in the Damariscotta River Estuary (~10 m mean water depth), Walpole, ME over approximately 24 hours. The first of the instruments (LISST A) was open to the environment while the second (LISST B) sampled water that was introduced to a sampling chamber via a pump (SeaBird SBE 5T, 3000 rpm) intended to break aggregates through increased turbulent shear, denoted by the superscripts "(open)" and "(shear)," respectively. Note that the shear is not quantified, nor do we know what percent of aggregates were broken; thus the comparison of measurements between the two treatments provides only a qualitative indication of the effect aggregation has on the optical properties as measured by the LISST. During the last two hours of the deployment, the sample chamber and pump were removed from the second instrument so that both instrument sample volumes were open to the environment. Both instruments were configured to sample in bursts, timed at five minute intervals. LISST measurements were processed using standard methods and then burst-averaged.

Laboratory aggregation experiment
The laboratory experiment was performed in order to examine how aggregation affects optical properties as a function of increasing aggregate size. Two beam transmissometers (a LISST-100 Type B, acceptance angle 0.0269° and an ac-9, acceptance angle 0.93°) were arranged side by side with their sampling volumes open to the environment in the bottom of a large 100×40×45-cm sink. The LISST measured both beam-attenuation and near forward scattering, which was inverted to PSD as described earlier. While both sample volumes were open to allow aggregates to sink through them, we assume the contribution of dissolved materials that might be released by the clay to be negligible during the experiment and refer to attenuation as p c rather than pg c . A WET Labs ECO-BB3 was used to measure backscattering at a single angle in the backwards direction, at three illumination wavelengths (532, 660, 880 nm). Care was taken to position all instruments to sample at the same depth.
The tank was also outfitted with sampling tubes having inlets at the instrument sampling depth. Samples (100 mL, in triplicate) were pumped gently at regular intervals throughout the experiment. Suspended particulate mass measurements (SPM) were made gravimetrically, using dried and pre-weighed 0.8-μm polycarbonate filter pads, and included a 100-mL deionized water rinse to remove accumulated salts. All data were captured by a single PC during the experiment, and later processed (time-stamping, calibration, inversion, and timebinning) in MATLAB. Calibration, correction, and data processing were performed using standard methods, and subsequently measurements were time-binned to five-minute intervals. avg D was calculated according to Eq. (1) for the individual LISST measurements within each time bin.
On the day of the experiment, the sink and all instruments were first cleaned thoroughly. The sink was then filled with particle-free reverse-osmosis water, which was allowed to degas and was used to blank all instruments in the sink. A slurry of bentonite clay was disaggregated by vigorous stirring for ~30 minutes and then added and mixed into the water (4 g dry weight in 120 L of water, producing an environmentally relevant mass concentration of approximately 33 g m 3 ). A calcium chloride solution (0.4 g CaCl L 1 ) was then mixed into the sink to initiate particle aggregation. Note that this procedure was repeated with differing instrumentation before the specific experiment discussed here, and in each case results were very similar, differing slightly in the timing of aggregation and sinking. Sampling protocol for the SPM measurements are described in more detail in Russo et al. [30].
Mass-specific optical properties ( , , where bar notation indicates bin or triplicate mean value. Uncertainty in the mass-specific optical properties is determined by standard propagation of uncertainty, for example, where  denotes standard deviation of the triplicate or binned measurements.

In situ disaggregation experiment
PSD inverted from the LISST scattering measurements reveal disappearance (destruction) of large particles by the pump and creation of smaller particles consistent with disaggregation ( Fig. 1A). During the control period, the size spectra for the two different instruments were very similar in shape (Fig. 1B). A time series of attenuation and avg D during the experiment ( Fig. 2A,B)  The beam-attenuation in the sheared treatment is ~30% higher (relative to control) compared to the open treatment (Fig. 2C), consistent with the idea that the smaller particles are more efficient attenuators per mass, though significantly less so than predicted by theory of solid particles [16]. These results also suggest that a large fraction of the particles contributing to the beam-attenuation are aggregates and that aggregation and disaggregation affect the beam-attenuation measured with the LISST. The observed effects on the beamattenuation (as opposed to PSD inversion) are due either directly from changes in attenuation efficiency between the two treatments, or indirectly by making less material scatter within the acceptance angle of the instruments [31,32]. For the latter to be important, particles greater than ~400 µm would have to be broken, based on the acceptance angle of the LISST [32].

Laboratory aggregation experiment
PSD derived from the LISST (Fig. 3A) show a suspension initially dominated by small particles (~5 μm, likely tightly-bound clay micro-aggregates). Formation of a population of aggregates is clear in the PSD, as it grows in both modal size and magnitude (first clearly visible as a bump at ~60 μm). The mode of the aggregate population grows and eventually appears to exceed the maximum size bin of our LISST instrument (large "snowy" aggregates were clearly visible to the naked eye after ~2 h). As particles aggregated in our tank, the effects of settling and scavenging eventually began to dominate: large aggregates settle out of suspension, and through their large capture cross-section, scavenge and incorporate smaller particles [e.g., 33-35]. The final (red wide-dashed) PSD shows a significant and unbiased decrease in fine particles less than ~20 μm, expected as a result of scavenging by the large settling aggregates. In a previous experiment (identical setup, but with different instrumentation), we examined samples from the start of the experiment and ~1.5 h later with a microscope equipped with digital camera (Fig. 4). The microphotographs provide additional confidence in the LISST measurements, as the initial population (Fig. 4A) appears to be single particles, and aggregates are clearly visible in the image from ~1.5 h (Fig. 4B). Furthermore, the size of aggregates in Fig. 4B (approximately 75-100 μm) agrees roughly with the modal diameter of the aggregate population seen at ~1.5 h in Fig. 3A. The evolution of aggregation during the experiment can be illustrated by dividing the PSD into three distinct pools: (1) primary particles smaller than 6 μm, (2) small aggregates from 6 to 60 μm, and (3) large aggregates greater than 60 μm [36]. Area concentration in primary particles decreases throughout the experiment as they themselves form aggregates and/or are scavenged by larger aggregates settling. Initially, small aggregates are formed, leading to formation of larger aggregates, sweeping both primary particles and smaller aggregates from the water. The rate of decrease in primary particles is highest as large aggregates dominate, due to scavenging. Evolution of the area-weighted average size, avg D , can also be seen in Fig.  3B and shows rapid aggregation (increase in avg D until ~2 h after the start of the experiment), followed by settling and scavenging (slow reduction in avg D ).  Deviation between beam-attenuation measured by the ac-9 and LISST is also evident in Fig. 6A. This deviation is consistent with differences in acceptance angles of the LISST and ac-9 since the larger acceptance angle of the ac-9 compared to the LISST collects more nearforward light scattered by large particles [31,32]. We find that the ratio of beam-attenuations from the two instruments correlates well with the average particle size (Fig. 7). This relationship is not due to the optical peculiarities of aggregates, but rather to the effects of particle size on near-forward scattering. Departure from the correlation for large avg D is likely due to the presence of large aggregates beyond the size range resolved by the LISST inversion.