Terrestrial-Aquatic Linkages: Riparian Arthropod Inputs Alter Trophic Cascades In A Stream Food Web

INTRODUCTION

Ecosystems are rarely bounded by the area selected for study, and external factors have the potential to substantially affect the patterns and dynamics of a focal system. The effects of these across-habitat influences can often exceed internal factors (Polis and Hurd 1996, Polis et al. 1996, 1997). Food web dynamics occurring in a focal habitat, in particular, are often influenced strongly by trophic linkages, the movements of nutrient, detritus, and prey organisms across contiguous habitats (Polis et al. 1996, 1997). For example, allochthonous inputs of prey species can allow consumer populations in recipient habitats to increase, and these subsidized consumers in turn can depress in situ prey items (Polis and Hurd 1996). Few experimental studies, however, have tested for the effects of allochthonous prey inputs on trophic cascading interactions in natural habitats with the direct manipulation of prey flux at the habitat interface.

Dynamics of headwater stream ecosystems are generally regarded as occurring at the interface of aquatic and terrestrial ecosystems (i.e., the stream ecotone) where food web dynamics are influenced in complex ways by both in situ primary productivity and allochthonous inputs from the terrestrial landscape (Vannote et al. 1980, Naiman and Decamps 1997). In particular, inputs of particulate organic matter from terrestrial ecosystems represent an important energy source of production in most headwater streams of temperate deciduous forests. Most studies, however, have focused on the input of terrestrial plant matter, which is only indirectly available to top consumers such as predatory fish via invertebrate secondary production (Cummins et al. 1995, Wallace et al. 1997). Terrestrial arthropods that accidentally fall into stream channels represent a high-quality (i.e., low C:N ratio) food resource that is directly available to fishes (Mason and MacDonald 1982, Garman 1991, Garman and Moring 1992, Edwards and Huryn 1996, Wipfli 1997). This allochthonous input, which often exceeds total in situ aquatic secondary productivity (Cloe and Garman 1996), can provide a tremendous energy subsidy and can increase the abundance of predatory fish; the ensuing effects on consumers can cascade through the food web and ultimately affect resident primary producers. Nevertheless, the community-based impact of such allochthonous inputs via a cascading trophic interaction are poorly understood. We present experimental evidence that terrestrial arthropod inputs have an indirect but prominent effect on a stream benthic community by altering the intensity of fish predation in the stream's food web.

MATERIALS AND METHODS

A manipulative field experiment was conducted in the Horonai Stream in the Tomakomai Experimental Forest of Hokkaido University (TOEF; 42[degrees]43[minute] N, 141[degrees]36[minutes] E) in Hokkaido, the northernmost island of the Japanese archipelago. The riparian zone of this small, cold, spring-fed stream (15.4 k[m.sup.2] in drainage area, 14 km in total length, 2-5 m width, gradient [less than]1%) is covered with a secondary-growth deciduous forest dominated by oak (Quercus crispula), ash (Fraxinus mandshurica), and maple (Acer mono). Approximately 95% of the entire width of the stream is covered by forest canopy. In the Horonai Stream, discharge usually remains stable throughout the year (annual average discharge, 0.24 [m.sup.3]/s), with major disturbances rarely occurring. Three common fish species inhabiting the water column are Dolly Varden (Salvelinus malma), white-spotted char (S. leucomaenis), and rainbow trout (Oncorhynchus mykiss), all of which consume both terrestrial and aquatic arthropods. Terrestrial arthropod inputs to the stream measured from early July 1995 to late July 1996, varied considerably among seasons, peaking during summer (June to August; S. Nakano, unpublished data). These fishes consumed, on average, 60% of the total terrestrial arthropod input during the summer, which contributed [approximately]56% of their annual energy (S. Nakano, unpublished data).

Terrestrial arthropod inputs and the presence of predatory fish were experimentally manipulated at the reach scale in the stream during four weeks of early summer (June-July) in 1995. The four experimental treatments (reduced arthropod inputs with fish present, natural arthropod inputs with fish present, reduced arthropod inputs with fish absent, and natural arthropod inputs with fish absent) with two replications of each, were randomly assigned to eight 25 m long reaches (100-[m.sup.2] surface area) over a 1.3-km stretch of the stream. Each experimental reach was separated from those immediately adjacent by at least 100 m long unmanipulated reaches. The greenhouse-type covers (5 m wide, 2.5 m high, and 50 m long), which were made of transparent plastic sheets supported by aluminum frames, were constructed so as to cover the entire experimental reach plus 25 m upstream [ILLUSTRATION FOR FIGURE 1 OMITTED]. Both ends of the cover were blocked to the water surface with 1-mm mesh linen to prevent flying insects from entering the greenhouse. Two roof windows (1 m long and 4 m wide; [ILLUSTRATION FOR FIGURE 1 OMITTED]) at both ends of the cover allowed emerging aquatic insects to escape. The average air temperatures under the covers were [approximately]1 [degrees] C higher than outside, but water temperatures did not differ. Although light attenuation by the covers averaged 13.2%, such effects were within the natural range of variation in natural lighting conditions due to canopy cover differences among the reaches. Consequently, relative photon flux density above the water surface did not differ among the four treatments (Table 1). Other physical characteristics, depth, current velocity, and discharge, of the stream habitat did not differ among the four treatments (Table 1).