Theoretical analysis of size distributions determined with screens and filters

Nets, screens, and filters are used to sieve out particles larger than the pore sizes. However, during separation a significant fraction of particles smaller than the pore diameters will also be removed by collision and adhesion to the mesh fibers. Two filtration models were used to predict the size and mass distributions of particles during size separation. A capillary tube model was used to calculate particle removal by different screens, and a fibrous filter model was used for glass-fiber filters. The extent of particle removal was modeled assuming a size distribution of 3,100 particles evenly distributed at logarithmic intervals over 3 1 size classes ranging from 0.1 to 100 pm. As many as 8% of the particles and 50% of the total particle mass could be retained by a 210-pm (pore diam) mesh even though all particles were < 100 pm. This high retention of particles implies that when size distributions are prepared with screens and filters, the mass concentration in smaller size fractions will be considerably underestimated. Screens and filters are routinely used by aquatic scientists to separate and concentrate material (e.g. Mullin 1965; Malone 197 1). By passing water samples through a series of screens, samples can be separated into different size distributions and further analyzed. For example, Back et al. (199 1) size-fractionated lake-water samples through 5 3and 1 O-pm Nitex nets and determined the distribution of Chl a and other photosynthetic parameters. Similarly, Azam and Hodson ( 1977) used different pore-diameter Nuclepore and Millipore filters to separate marine bacteria according to size. A body of literature on filtration has been developed to predict removal of colloidal aerosol particles by filters, but these models have not been applied to size fractionation of aquatic samples with screens and filters. Two different approaches have been used to construct these models. The first considers the filter to be a bundle of fibers (e.g. models by Fuchs 1964; Hinds 1983). The second model is based on flow through small capillary tubes or pores (Pith 1966; Spurney et al. 1969). Of these two models, only the fibrous models have been used to predict aquasol removal in marine systems Acknowledgments I thank Mark Gross for performing the filtration experiments and calculating microsphere sticking coefficients. Funding was provided by ONR grant NO001 4-9 l-J1249, National Institutes of Environmental Health Sciences grant P42ES04940, and NSF equipment grant BCS 90-08145. by a variety of organisms including echinoderm larvae, brittle stars, bdelloid rotifers, sea anemones, and zoanthids (Rubenstein and Koehl 1977). Recently, fiber models have been applied to grazing by cladocerans and the removal of particles on nets spun by caddisfly larvae (Fuller et al. 1983; Loudon 1990; Brendelberger 199 I). A comparison of the fiber and capillary pore models by Rubow and Liu (1986) showed that the fibrous models could be used to predict aerosol removal by fibrous filters and that only the pore models successfully predicted aerosol removal by Nuclepore filters. Further experiments have shown that fibrous models are also applicable to aquasol removal by glass-fiber and other types of large-pore fibrous filters (Logan et al. 1993). Both filtration models predict that some particles smaller than the mesh pore diameter will be retained by the net, in agreement with experimental observations. For example, studies on particle capture by the rectangular nets made by caddisfly larvae have shown that a substantial fraction of particles smaller than the net pores are retained, and the net removal efficiency is a function of the size and nature of the particle captured (Fuller et al. 1983; Loudon 1990). Malone et al. (1979) determined that 50% of material with a mean diameter of 16 pm was removed by mesh with a 22-pm more diameter. The removal of particles smaller than mesh pores can result in a distribution of particle sizes retained on the net that does not reflect