[NB: The first entry in this series was published here, Introduction to Sediment Traps, which covers the motivation, use and issues of sediment traps.]

Although validation of the data was lacking, sediment traps (or sedimentation traps as they were once called) were being used by the 1970s in the study of lakes (Davison et al., 1982; Scott and Miner, 1936). Following shortly behind, oceanographers started using sediment traps to study shallow water columns around estuaries and inlets.

Some of the earliest sediment traps, such as those used by Hakanson (1976), consisted of large, open dishes anchored to the sediments and which must be manually covered by divers prior to retrieval (Hakanson, 1976; Scott and Miner, 1936). Members of the limnic community were more confident and quicker to utilize sediment traps (Davison et al., 1982) than the oceanography community due to methodological concerns of their applicability in anything but the calmest waters (Kirchner, 1975). Consider for the moment if we consider a sediment trap a “rain gauge of the ocean” in its collection of sinking particles. on land, rain falls in nearly straight lines from the sky to the ground. Sinking particles in the ocean are moving downward, but they are also moving sideways and swirling around in all such manners. This leads to clear problems in interpreting the results of sediment traps.

Studies quickly recognized the sensitivity of sediment trap collection efficiency to both turbulence (Murray, 1970) and horizontal flow (Jobson and Sayre, n.d.). Based on these early laboratory experiments, the aspect ratio of the trap was shown to decreased collection efficiency by up to 80% even within calm, steady flows (Hargrave and Burns, 1979).

Attempts to improve the design of sediment traps while reducing their methodological uncertainty lead to a multitude of designs. To reduce hydrodynamic biases in collection efficiencies lead to free-drifting sediment trap arrays (Staresinic et al., 1978), while the desire for replicate samples lead to a remote, replicate-sampling sediment trap (Kimmel et al., 1987). Systematic analysis and comparison of these traps were lacking until the 1980’s when both laboratory and field studies were conducted and reported on.

Gardner (1980) conducted a laboratory review of sediment trap shapes and the resulting impact of these shapes on collection efficiency and dominant flow patterns. He found that flow over the mouth of the container and the resulting eddies lead to resuspension of trapped particles and under-collection while points of stagnation would lead to enhanced settling and over-collection. Gardner found that the collection rate of cylinders whose height to width ratio was greater than 2.3 best matched the true sediment input rate over the range of conditions tested (i.e. velocities < 4.5 cm/s, particle size between 0.8 μm – 63 μm). The height to width ratio strongly impacts the collection of a sediment trap (Gardner, 1980) and recent cylindrical trap designs have had height to width ratio of ~6-8 (Haskell et al., 2015; Stukel et al., 2015).

The biased collection of certain size fractions of particles leaded to skewed estimates of true particle flux was found to depend on the turbulence over the trap (Gardner, 1980). While the introduction of baffles reduces the size of the eddies above sediment traps, their impact on reducing the size-dependent biasing was negligible even as the overall collection ratio changed (Gardner, 1980).

Collection efficiency was found to vary substantially even between identical traps due to tilting. Gardner (1985) found large variances in sediment trap collection efficiencies in cylindrical traps when tilted from vertical by as little as 5 degrees either up or downstream. The impact was enhanced with current velocity and would also be important in regions with high frequency, internal waves where the traps would experience non-perpendicular currents (Gardner, 1985).

$$v = \frac{2}{9}\frac{\Delta \rho}{\mu} R^2$$

According to the simplified Stokes Law (above), the sinking velocity of a particle is directly proportional to the square of its radius (Durkin et al., 2015); therefore aggregates sink faster than their constituents. Traditional sediment trap procedures are too disruptive to collect aggregates intact leading to large uncertainties in the number and relation of aggregates in the export of particles. An early attempt to reduce the disturbance of aggregates during collection used polyacrylamide gel to preserve particles for later analysis (Jannasch et al., 1980). By trapping particles within the gel matrix, microscopic analysis of aggregate post-collection was possible. Similar techniques have been in use recently (Durkin et al., 2015; Ebersbach and Trull, 2008; Guidi et al., 2008; Hung et al., 2012).

Although the early years of sediment trap studies were dominated by analytic studies of collection efficiency, studies of shape, and an interest in calibrating the technique (e.g. Butman, 1986; Gardner, 1980; Lee et al., 1992); the proportion of these analytical assessments has dropped to less than 10% of all ST publications. Part of this downward trend can be explained by the rapid increase in field studies reusing methods, which were established by previous reports. Indeed, the number of ST designs in common use is few (Table 1). As seen in Figure 2, the 90’s were a time of implementation of STs of experimental designs with <40% of articles fitting into any of the three paradigms setup in Table 1.

Additionally, the early papers identified major uncertainties in the ST methodology that have no clear solution due to nature of the problem. Swimmers degrade the composition of ST samples while also representing an unknown proportion of the “true” particle flux (some swimmers are either dead or sinking so as to count as POM while other swimmers are diel-vertical migrators and should not count in flux estimates). Some studies manually remove swimmers by use of a microscope while others make corrections for them analytically. Both methods had drawbacks and preclude direct comparison of many datasets.

With new computational, manufacturing and mechanical developments rendering previous research and development (R&D) antiquated, a fresh look into the design process and development of STs should be made. The stagnation in the annual number of ST publications over the past 15 years (Figure 1) suggest that novel tools and techniques may find applicability in this import field of research. Mechanical and computational design facilities have made rapid developments in recent decades, while ST design archetypes (Table 2) have primarily been based off of tried and true designs (Figure 4).