Successes, failures, and opportunities in the practical application of drift-foraging models

Abstract

Accurately measuring productive capacity in streams is challenging, and field methods have generally focused on the limiting role of physical habitat attributes (e.g. channel gradient, depth, velocity, substrate). Because drift-foraging models uniquely integrate the effects of both physical habitat (velocity and depth) and prey abundance (invertebrate drift) on energy intake for drift-feeding fishes, they provide a coherent and transferable framework for modelling individual growth that includes the effects of both physical habitat and biological production. Despite this, drift-foraging models have been slow to realize their potential in an applied context. Practical applications have been hampered by difficulties in predicting growth (rather than habitat choice), and scaling predictions of individual growth to reach scale habitat capacity, which requires modelling the partitioning of resources among individuals and depletion of drift through predation. There has also been a general failure of stream ecologists to adequately characterize spatial and temporal variation in invertebrate drift within and among streams, so that sources of variation in this key component of drift-foraging models remain poorly understood. Validation of predictions of habitat capacity have been patchy or lacking, until recent studies demonstrating strong relationships between drift flux, modeled Net Energy Intake, and fish biomass. Further advances in the practical application of drift-foraging models will require i) a better understanding of the factors that cause variation in drift, better approaches for modelling drift, and more standardized methods for characterizing it; ii) identification of simple diagnostic metrics that correlate strongly with more precise but time-consuming bioenergetic assessments of habitat quality; and iii) a better understanding of how variation in drift-foraging strategies are associated with other suites of co-evolved traits that ecologically differentiate taxa of drift-feeding salmonids.

Appendix 1 Appropriate mesh sizes for invertebrate drift samples

Choosing the appropriate mesh size for sampling drift involves tradeoffs. An excessively coarse mesh will not capture all drifting invertebrates that might constitute prey for drift-feeding fishes, but a very fine mesh will clog quickly with detritus at high velocities, may include invertebrates that are too small for larger size-classes of fish, and may take much longer to process in the laboratory. Below we briefly consider the implications of mesh size for underestimation of prey size classes that may pass through a coarse mesh. To determine potential differences in sample catch and processing effort, researchers or managers should deploy pilot drift nets of different mesh sizes in their target stream(s), and directly evaluate and compare samples and processing time. Ideally, nesting coarse mesh nets inside fine mesh nets would also provide a direct assessment of the abundance and size of prey that would be discounted in drift-foraging models that are calibrated with coarse drift net sets.

Figure 7. illustrates how gill raker spacing increases with fish size (data for Atlantic salmon from Wankowski (1979)). If one assumes that gill raker spacing is broadly similar for other juvenile salmonids, and that fish are unlikely to forage effectively on prey that are substantially smaller than the spacing between their gill rakers, then gill raker spacing may serve as a coarse index of minimum prey size. However, it is the longest prey dimension that will likely determine both whether a prey taxon will pass through the gill rakers, and whether the prey is detected in the water column. In contrast, it is the smallest prey dimension that will likely determine whether prey can pass through a drift net. Because of their long, narrow body form and their relative abundance in both the drift and diet of juvenile salmonids (e.g. Keeley and Grant 1997), chironomids are likely the taxon of greatest concern for underestimation by coarse drift nets, although drift-feeding juveniles may also feed on small spherical prey (e.g. ostracods) that could also pass through coarser nets. For example, Keeley and Grant (1997) found that chironomids accounted for 30 % of the drift and 55 % of prey items in the stomachs of juvenile Atlantic salmon, although they likely constituted less in terms of biomass.

Fig. 7

Gill raker spacing as a function of Atlantic salmon fork length [data from Wankowski (1979)]

Over 95 % of chironomids in the drift and diet of juvenile salmon in Keekey and Grant (1997) had widths of 0.45 mm or less. Keeley and Grant (1997) used 300 μm drift nets, and their chironomid size distribution is truncated at head capsule widths below 0.2 mm (their Fig. 4). If one therefore assumes that a chironomid larva can pass through a mesh opening 0.15 mm wider than its’ width, then this information may roughly inform the proportion of chironomids that may be undersampled with different mesh sizes. Using an average chironomid length:weight ratio of 13 (Nolte 1990), Fig. 8 shows that chironomids with a head width less than 0.35 mm and a body length less than 4.5 mm would be likely on average to pass through a 500 μm drift net, which would include more than 80 % of the chironomids observed in juvenile Atlantic salmon stomachs by Keeley and Grant (1997).

Fig. 8

Chironomid width as a function of body length (after Nolte (1990)). Horizontal lines indicate the width of a chironomid that will pass through the respective mesh sizes, assuming that a chironomid will pass through a mesh 0.15 mm wider than its’ body width. Vertical lines indicate the corresponding average body length threshold for retention in the drift net

Although the small size of chironomids means that their contribution to prey biomass is likely less than their contribution to abundance (and Keeley and Grant (1997) note that invertebrate prey in their study are smaller than reported elsewhere), the above exercise suggests that mesh sizes larger than 250–300 μm may substantially undersample the availability of drifting prey items for juvenile salmonids. Researchers and managers should therefore carefully consider the relative costs and benefits of using larger drift net mesh sizes for sampling drift; coarser meshes produce cleaner samples that may be faster and easier to process because of less detritus, and may be more temporally representative because they clog more slowly, but this will come at the cost of undersampling smaller prey items.

One option for sample optimization may be to set a combination of coarse and fine drift nets in the study stream(s), and if the proportion of smaller prey in the fine nets appears relatively stable (e.g. there is relatively little variation in the size distribution of drifting invertebrates), then the abundance of smaller size classes could be extrapolated from the abundance of larger prey caught in the coarser drift nets. At the very least any discrepancy in prey abundance between coarse and fine mesh nets would inform whether underestimation of small prey size classes is a serious concern. Ideally, this should be complemented by at least cursory sampling of the prey size distribution in stomach contents of juvenile salmonids in the target stream(s).