Peer Reviewed Title: Shallow-Water Piscivore-Prey Dynamics in California's Sacramento-San Joaquin Delta Journal Issue:

Predation is one mechanism that could lead to low native fish abundance in macrophyte dominated shallow-water habitats in the Sacramento-San Joaquin Delta. We used beach seine and gill net sampling to identify and compare the distribution and feeding ecology of three piscivores (striped bass, Morone saxatilis , largemouth bass, Micropterus salmoides , and Sacramento pikeminnow, Ptychocheilus grandis ) at five nearshore sites in the Sacramento-San Joaquin Delta. Sampling was conducted March-October 2001 and 2003. We addressed the following questions. What are the spatial and temporal distributions of age-1 and older striped bass, largemouth bass, and Sacramento pikeminnow? What prey are eaten by these predators? What is the relative importance of predator size versus seasonal prey availability on incidence of piscivory for these predators? What is the likely per capita impact of each piscivore on prey fishes, particularly native fishes? All 76 of our individual station visits yielded at least one of the three species, suggesting that piscivorous fishes frequently occur in Delta shallow-water habitats. All three piscivores had diverse diets comprised of numerous invertebrate and fish taxa. There were noticeable seasonal shifts in prey fish for each the three piscivores. In general, most native fish were during spring (March-May) and the highest prey species richness occurred during summer (June-August). Largemouth bass likely have the highest per capita impact on nearshore fishes, including native fishes. Largemouth bass preyed on a greater diversity of native fishes than the other two piscivores and consumed native fishes farther into the season versus May). Based binomial models, incidence of piscivory a of size for largemouth bass and Sacramento pikeminnow.

comprised of numerous invertebrate and fish taxa. There were noticeable seasonal shifts in prey fish for each of the three piscivores. In general, most native fish were consumed during spring (March-May) and the highest prey species richness occurred during summer (June-August). Largemouth bass likely have the highest per capita impact on nearshore fishes, including native fishes. Largemouth bass preyed on a greater diversity of native fishes than the other two piscivores and consumed native fishes farther into the season (July versus May). Based on binomial generalized additive models, incidence of piscivory was predominantly a function of size for largemouth bass and Sacramento pikeminnow. Largemouth bass became predominantly piscivorous at smaller sizes than Sacramento pikeminnow; about 115 mm versus about 190 mm respectively. In contrast, incidence of piscivory was predominantly a function of season for striped bass. Striped bass were typically most piscivorous during summer and fall regardless of size. We conclude that shallowwater piscivores are widespread in the Delta and generally respond in a density-dependent manner to seasonal changes in prey availability.

INTRODUCTION
The relative effects of predation in tidal rivers and their estuaries are controlled by a complex interaction of factors that directly and indirectly influence prey encounter and capture probabilities. These probabilities can be highly species-specific, location-specific, and strongly influenced by ecosystem change. Piscivorous fishes typically pass through a series of well-described diet shifts as they grow (Mittelbach and Persson 1998) and as prey availability changes (Hartman and Margraf 1992;Buckel et al. 1999). At large spatial scales, piscivore-prey dynamics can be regulated by flow patterns (Meffe 1984) and turbidity (Rodríguez and Lewis 1994;Gregory and Levings 1998), both of which are often changed substantially by dam construction (de Merona et al. 2001;Osmundson et al. 2002;Quist et al. 2004). In tidal river estuaries, river flow affects the overlap of suitable water quality and suitable habitat structures (Manderson et al. 2002;Peterson 2003), which can influence piscivore-prey dynamics (Manderson et al. 1999) that ultimately influence fish nursery habitat quality and recruitment.
At smaller spatial scales, piscivore-prey interactions also can be affected by local habitat features (Werner et al. 1983;Greenberg et al. 1995). One important structural aspect of estuarine habitat is water depth. One of many functions that shallow estuarine habitats may serve is that of predation refuge for small fishes; in effect they may reduce predator encounter and capture probabilities. Paterson and Whitfield (2000) concluded the shallow marsh creek system in a South African estuary was a predation refuge for young fishes because large predatory species were rarely collected. McIvor and Odum (1988) found there were fewer attacks on tethered fish in depositional tidal creek habitats than in nearby deeper erosional habitats.
However, the hypothesis that shallow estuarine habitats can be generalized as predation refuges has been questioned (Sheaves 2001). First, not all small estuarine fishes are adapted to use shallow habitats like marsh creeks; many species or early life stages of species are pelagic and thus frequent the deeper habitats often used by large piscivores (Laprise and Dodson 1989;Hagan and Able 2003). These pelagic species/life stages may use turbidity, rather than water depth, as a form of cover from predators (Abrahams and Kattenfeld 1997;Utne-Palm 2002). Second, not all estuarine piscivores are large. Piscivore-prey dynamics typically are strongly sizestructured (Reimchen 1990), and many juvenile predators (i.e., < 100 mm or so) can be both abundant and fairly piscivorous in estuarine shallows (Baker and Sheaves 2005). Small fishes recruiting to these nurseries may be readily preyed upon by small predators, so some shallow-water piscivores may have higher per capita impacts on shallow-water fishes than others. For instance, age-0 bluefish Pomatomus saltatrix, impart substantial predatory mortality on age-0 striped bass Morone saxatilis, in nearshore habitats of the U.S. Atlantic coast (Buckel et al. 1999).
The restoration of shallow-water habitats (SWH) has been considered an important ecosystem rehabilitation tool in California's San Francisco Estuary (Brown 2003). Because the Estuary has lost many of its historical marsh habitats to various reclamation activities, resource managers have expressed hope that SWH could contribute to increased production of declining native and sport fishes. Breaching levees of agricultural 'islands' in the Sacramento-San Joaquin Delta was conceived as a relatively inexpensive way to restore shallow areas to tidal inundation, However, subsequent studies of several breachedlevee flooded islands showed they were variably productive (Lucas et al. 2002), structurally simple, and overwhelmingly colonized by Brazilian waterweed Egeria densa and other submerged macrophytes (Brown 2003). Macrophyte-dominated shorelines are primarily inhabited by non-native fishes adapted to use these littoral habitats (Feyrer and Healey 2003;Grimaldo et al. 2004;Nobriga et al. 2005). We recently hypothesized that the interaction of hydrologic regulation of the San Francisco Estuary watershed and local habitat features facilitated the proliferation of exotic macrophytes (Nobriga et al. 2005). The macrophytes influence fish community composition, possibly including the relative importance of biotic structuring mechanisms like competition and predation.
Here, we describe and compare the distribution and feeding ecology of three piscivores that frequent nearshore habitats in the Sacramento-San Joaquin Delta: striped bass, Morone saxatilis, largemouth bass, Micropterus salmoides, and Sacramento pikeminnow, Ptychocheilus grandis. There is considerable management interest in the feeding ecology of piscivorous fishes in the Delta. Elevated predation rates are considered a potential 'indirect effect' of water diversion operations (Brown et al. 1996) and a potential hinderance to shallow-water habitat restoration (Brown 2003). We addressed the following questions. What are the spatial and temporal distributions of age-1 and older striped bass, largemouth bass, and Sacramento pikeminnow? What prey are eaten by these predators? What is the relative importance of predator size versus seasonal prey availability on incidence of piscivory for these predators? For the latter two questions, we hypothesized that native fish prey use would peak during spring, and total incidence of piscivory would peak during summer. Our primary goal was to integrate empirical piscivore-prey data into emerging conceptual models of ecological processes operating in shallow habitats of the Sacramento-San Joaquin Delta (e.g., Brown 2003;Grimaldo et al. 2004;Nobriga et al. 2005). This is a necessary precursor to the development of quantitative predator-prey or ecosystem models, and the design and implementation of successful habitat restoration projects. We also sought to provide insight into the impacts of these piscivores on prey fishes, particularly native fishes.

Study Area
The Sacramento-San Joaquin Delta is the landward edge of the San Francisco Estuary and a water supply nexus for much of California's population (Figure 1; Arthur et al. 1996;Kimmerer 2002). The Delta receives freshwater runoff from approximately 100,000 km 2 (40%) of California's surface area. Most natural runoff occurs during winter and spring (December-May), but significant proportions of natural runoff are captured in numerous reservoirs located throughout the watershed. Managed reservoir releases support regional agriculture and freshwater exports for agriculture and urban users in comparatively arid regions to the south and west. Primarily during the latter half of the nineteenth century and the first half of the twentieth century, the Sacramento-San Joaquin Delta was converted from a seasonally brackish marsh into a network of leveed channels conveying freshwater year-around; the Delta channels surround tracts of land drained to support agriculture (Nichols et al. 1986).

Species Background
Striped bass were introduced to San Francisco Estuary in 1879 (Dill and Cordone 1997). They thrived and supported a commercial fishery until 1935. They still support a recreational fishery despite a steady decline in juvenile production since the 1970s (Kimmerer et al. 2000). The striped bass is the San Francisco Estuary's most broadly distributed and abundant large piscivorous fish (Moyle 2002). Striped bass spawn in large, non-tidal tributaries to the Estuary. Its eggs and larvae are transported to the Estuary low-salinity zone by river currents. Age-1 and older striped bass occur throughout the Estuary and in adjacent freshwater and marine habitats.
Unlike striped bass, largemouth bass are primarily a freshwater fish that cannot successfully reproduce in brackish water (Moyle 2002). Largemouth bass also were introduced to the San Francisco Estuary watershed in the latter 19th Century (Dill and Cordone 1997), although their numbers in the Delta have increased only recently (Brown and Michniuk 2007). This increase was associated with increasing water clarity and submerged macrophyte abundance in the Delta. The increase in abundance has also been apparent based on fishing patterns; largemouth bass abundance has increased sufficiently to support a significant sport fishery (Lee 2000).
The native Sacramento pikeminnow is also a freshwater fish, most often associated with lotic habitats (Moyle 2002). Long-term trends in Sacramento pikeminnow abundance are unknown, but it is common in the Sacramento River basin (May and Brown 2002). Sacramento pikeminnow spawn in non-tidal streams and rivers; they often complete their life cycle within these habitats (Brown 1990). However, some age-1 and older individuals are transported into the Delta by winter-spring flow pulses and likely remain resident until maturity (Nobriga et al. 2006). Sacramento pikeminnow is not targeted by a sport fishery in the Delta, although it has been harvested for bounty in the upper Sacramento River to reduce presumed predatory effects on emigrating salmonid fishes (Moyle 2002).

Field Methods
We sampled monthly for fishes at five sites using beach seines in wadable-depth habitats, and using gill nets in deeper habitats adjacent to the beach seine sites (Nobriga et al. 2005; Figure 1). The primary purpose of our sampling design was to capture potentially piscivorous fishes from a variety of nearshore habitats for stomach content analysis. Sampling occurred during two piscivore growing seasons, March-October of 2001 and 2003. Two sites (Decker Island and Medford Island) were low velocity areas at the edges of channels. The other sites (Sherman Island, Liberty Island and Mildred Island) were shallow habitats along the internal remnant levees of flooded islands, a local term for former agricultural tracts that were flooded and not reclaimed. We deployed a 30 m x 1.8 m, 3.2 mm mesh beach seine via small, shallow-draft boats. We conducted 2-8 beach seine hauls per site per month (mean = 4). One site was sampled per day, so five days were needed to complete each month's sampling. Samples were collected afternoon through dusk, usually during a flood tide. Water depth at initiation of seine hauls averaged about 1 m at all sites, but varied from 0.4-1.5 m in individual hauls. All fishes collected in the beach seine were identified to species, and subsets of abundant species were measured for total length (TL) or fork length (FL) to the nearest mm. We conducted 158 beach seine hauls in 2001 and 141 beach seine hauls in 2003.
Deeper water (2-4 m) adjacent to the beach seine sites was sampled using gill nets (60 m x 2.4 m; randomized panels of 51-102 mm stretch mesh). The gill nets were set parallel to shore for 20-40 minutes, 2-6 times per visit (mean = 0.9 hr sampled per visit). Gill net effort was concentrated around sunset to target actively foraging fishes moving toward or from the wadable habitats. All fishes collected in the gill net were identified to species and measured for TL or FL to the nearest mm.
We examined the stomach contents of age-1 and older striped bass (n = 570; 63-776 mm) and largemouth bass (n = 320; 58-569 mm), and the foregut contents of Sacramento pikeminnow (n = 282; 65-504 mm). For each species, up to ten individuals per gear per day were retained for stomach contents evaluation. Fishes were chosen haphazardly from the available size ranges. Whole fish up to 306 mm were preserved in the field in 10% formaldehyde. In the laboratory, these fish were re-measured and weighed to the nearest 0.01 g using an electronic balance. Their stomachs/foreguts were then removed and all contents were identified to the lowest practicable taxon and summarized as frequencies of occurrence. For fish larger than 306 mm, stomachs /foreguts were removed in the field and preserved as above. The field dissection of stomachs from fishes > 306 mm precluded laboratory weight measurements. However, the stomach /foregut contents of field-dissected stomachs also were examined in the laboratory using the methods described above.

Data Analysis
For the first study question (piscivore distribution), we evaluated nearshore piscivore persistence at our sampling sites by combining the beach seine and gill net catches for use in a semi-quantitative, presenceabsence analysis because both beach seine and gill net capture efficiencies changed during the sampling seasons. In general, beach seine capture efficiency decreased with increasing piscivore length, whereas gill net capture efficiency increased with increasing piscivore length (Nobriga et al. 2006), impairing the accuracy of among-site and among-month comparisons of catch per unit effort. However, our gear choices reliably detected the presence of nearshore piscivores, hence our use of them for collecting fish for stomach contents. For the analysis, we calculated the percentage of sampling days that each predator species was collected by either beach seine or gill net at each site, each year. Then, we summed these percentages to develop 'persistence indices. ' These indices had a potential maximum value of 1000 (100% x 5 sites x 2 years). We assume that differences in the indices reflect the breadth of nearshore piscivore distribution, both spatially and seasonally.
We used stomach content data to answer study questions two and three. In the upper estuary there are persistent seasonal trends in the relative densities of young fish that we predicted would influence prey fish occurrence in piscivore diets. Most native fishes spawn from February-June, prior to most established nonnative fishes, which tend to spawn during May-August (Matern et al. 2002;Feyrer 2004;Grimaldo et al. 2004). The exact timing of reproduction and the relative densities of native and nonnnative fishes are affected by interannual differences in water temperature and flow magnitude, but native species relative abundance is always highest during spring before most of the nonnative species have spawned. Further, total young-of-year fish densities reach their annual peak during summer after most nonnative species also have spawned. We hypothesized that: 1) native fish prey use would peak during spring because this is the time of year that native fishes have their highest relative abundance in the Delta; and 2) total incidence of piscivory would peak during summer because that is the time of year that young fish densities are highest.
We tested these hypotheses using a two-step approach. To evaluate whether native species were most commonly eaten during spring, we summarized the monthly mixes of native and non-native prey fishes and qualitatively evaluated the mixes. We combined data for both years for this analysis. To test the hypothesis that total piscivory was highest during summer, we used a binomial generalized additive modeling (GAM) approach (logit link function) to test the relative importance of size and season on incidence of piscivory. For the purpose of the analysis, we defined "piscivory" as the presence of fish prey in the stomach/foregut and used this as our response variable. We used Julian day (January 1 = day 1) as a surrogate for prey fish relative density based on the seasonal trends we described above. By basing the analysis on the Julian day of each season, both years could be evaluated together. Each piscivore was modeled separately. We used predator fork length, Julian day, and their interaction term as explanatory variables. Explanatory variables were considered statistically significant predictors of incidence of piscivory if the P-value of their chi-square statistic was ≤ 0.05. The model fits were evaluated by comparing the difference between the null and residual deviance in the models. The null deviance is akin to the total variance in linear regression. The residual deviance is the amount of variability left unexplained after the model is fit. The difference between null and residual deviance is analogous to the r-squared statistic in linear regression.

RESULTS
Age-1 and older piscivores were consistently collected across stations and sampling days. All 76 of our individual station visits yielded at least one of the three species, and usually more than one (Figure 2). Striped bass were collected at every site in both years, and were collected from Liberty and Sherman islands during every visit. Sacramento pikeminnow also were collected from all five sites, though they were not collected from Mildred Island in 2001. The only site that always yielded Sacramento pikeminnow was Decker Island. Largemouth bass were always collected from Medford and Mildred islands, but were not collected at Liberty Island in either year. The persistence index was highest for striped bass (845), followed by Sacramento pikeminnow (757), and by largemouth bass (622). This suggests striped bass had the broadest spatio-temporal distribution and largemouth bass the narrowest.
All three piscivores had diverse diet compositions comprised of numerous invertebrate and fish taxa (Table 1). Corophiid and gammarid amphipods were frequently consumed by all three predators. All three predators also consumed annelids, but more frequently in 2003 than 2001. Striped bass were the most frequent consumer of mysids and decapod shrimp. Largemouth bass were the most frequent consumer of most insect taxa and the only preda-   tor found to occasionally prey on vertebrates other than fish.

SAN FRANCISCO ESTUARY & WATERSHED SCIENCE
We identified 15 prey fish species, and two additional genera (Lepomis and Tridentiger); hence, a total of at least 17 species were consumed during our study ( Table  1). Sixteen of the 17 taxa were eaten by largemouth bass, 12 of 17 by striped bass, and 10 of 17 by Sacramento pikeminnow. Two native fishes (prickly sculpin, Cottus asper, and tule perch, Hysterocarpus traski) and five nonnative fish taxa (threadfin shad, Dorosoma petenense, inland silverside, Menidia beryllina, striped bass, yellowfin goby, Acanthogobius flavimanus, and Tridentiger spp.) were preyed on by all three piscivores. Note that unidentified fish remains were common, so all of these results should be considered minimum estimates.
There were noticeable seasonal shifts in prey fish consumed by all three piscivores (Table 2). Collectively, most native fish use occurred during spring (March-May) and the highest prey species richness occurred during summer (June-August). Largemouth bass preyed on a greater number of native fish than the other two piscivores and consumed native fish farther into the season (July) than the other two piscivores (May).
Largemouth bass piscivory was significantly influenced by fork length (chi-square = 12.3; P = 0.005), but not season (chi-square = 2.10; P = 0.52). The fork length X season interaction also was not significant (chi-square = 3.46; P = 0.30). The percentage of null deviance explained by largemouth bass fork length was 18%. Sacramento pikeminnow piscivory also was affected by fork length (chi-square = 13.3; P = 0.003), but not season (chi-square = 4.02; P = 0.25), and with no significant interaction (chi-square = 4.57; P = 0.19). The percentage of the null deviance explained by Sacramento pikeminnow fork length was 33%. Largemouth bass became piscivorous at smaller sizes than Sacramento pikeminnow ( Figure 3). The GAMs predicted that 50% of largemouth bass were piscivorous at about 115 mm, whereas Sacramento pikeminnow were about 190 mm before 50% of individuals were piscivorous. In contrast to largemouth bass and Sacramento pikeminnow, striped bass piscivory was significantly affected by season (chi-square = 24.6; P = 0.00002), but not fork length (chisquare = 7.37; P = 0.06). The fork length x season interaction also was not significant (chisquare = 4.42; P = 0.22). The seasononly model explained 31% of the null deviance in striped bass piscivory. Striped bass typically only exceeded the 50% piscivory threshold during summer and fall regardless of size (Figure 4).

DISCUSSION
Striped bass, largemouth bass, and Sacramento pikeminnow are three of the major predators of juvenile and small adult fishes in the Delta. Moreover, they appear to be the dominant piscivores in nearshore habitats, where juvenile striped bass and Sacramento pikeminnow and juvenile-adult largemouth bass can be considered resident fishes (Stevens 1966;Feyrer and Healy 2003;Nobriga et al. 2005;2006;Brown and Michniuk 2007). Our study indicates that collectively these predators frequently occur in Delta shallow-water habitats (Figure 2). However, we acknowledge that having only five sampling sites limited our ability to generalize about piscivore distributions across the entire Delta. The study also yielded new information about predator-prey dynamics involving these fishes, and provided insight into the relative importance of piscivory on native fishes. These results have management implications for native fishes and the design of shallow water restoration projects.

Does Piscivory in Shallow Water Habitat Follow Predictable Patterns?
There is a large body of literature on the prey choices of piscivorous fishes. Piscivore prey choices are functions of encounter and capture probabilities. Both encounter and capture probabilities are probably affected by prey relative abundance (Buckel et al. 1999). Encounter probabilities also are influenced by environmental factors such as turbidity (Gregory and Levings 1998) and vegetation density (Greenberg et al. 1995;Buckel and Stoner 2000). Capture probabilities also are influenced by prey behavior (Scharf et al. 2003), and morphological features, the most important of which is prey to predator size ratios (Juanes and Conover 1994;Hartman 2000). Differences in piscivore habitat use and foraging strategy also influence prey use. Some piscivores forage most efficiently in open-water (unstructured) environments, whereas others forage most efficiently in structured environments (Greenberg et al. 1995;Buckel and Stoner 2000). In general, striped bass and Sacramento pikeminnow are examples of the former; largemouth bass is an example of the latter (Moyle 2002). These differences were reflected in diet compositions (Table  1). For instance, largemouth bass were the most frequent consumer of Lepomis spp. and prickly sculpin, fishes that are common in submerged macrophyte habitats (Grimaldo et al. 2004), and vegetation-oriented invertebrates like insect larvae. In contrast, and as reported previously (Stevens 1966;,  Each of the piscivores we examined had diverse and dynamic diets (Tables 1-2), which is consistent with studies from other locations with diverse prey assemblages (Hartman and Brandt 1995;de Almeida et al. 1997). Field observations of changes in piscivore stomach contents through time have indicated that piscivorous fishes exhibit prey switching behavior as predicted by a type-III (density-dependent functional response; Buckel et al. 1999). However, laboratory research does not support this (Buckel and Stoner 2000). Rather, the field-observed pattern may arise as an artifact of predators aggregating in microhabitats that support a transiently available abundant prey, thus 'ignoring' rarer prey occupying different microhabitats. In our study, native fish occurrence in piscivore stomachs was highest in spring, and total piscivory was highest in summer (Table 2; Figures 3-4). This matches the prey switching pattern reported by Buckel et al. (1999). We think that shallow-water piscivores in the Delta generally respond to seasonal changes in prey availability in the same way piscivorous fishes do in other locations, and are thus very predictable. Though piscivores may not invoke density-dependent prey switching, predation mortality is likely density-dependent because high relative abundance translates into more encounters with predators (Buckel and Stoner 2000). In the upper San Francisco Estuary, the intra-and interannual variation in young-of-year fish abundance (and relative abundance) is affected by water temperatures and for some species, river flow variation (Matern et al. 2002;Feyrer 2004;Grimaldo et al. 2004). Therefore, it is likely that these abiotic variables also fundamentally mediate predation mortality.

Are Some Predators Worse Than Others For Native Fishes?
Largemouth bass had the highest per capita predatory influence on nearshore fishes, followed by striped bass, and then Sacramento pikeminnow. This conclusion is supported by two lines of evidence. First, the number of both native fish species and total fish species observed in the stomach/foreguts of each pred-ator varied in the above-mentioned order (Table 2). It is interesting that largemouth bass consumed a greater number of native fish species than striped bass and Sacramento pikeminnow because largemouth bass had lower spatial overlap with most native fishes than the other two predators (Nobriga et al. 2005). Second, largemouth bass switched to piscivory at a smaller size than the other two predators (Figures 3-4). This means juvenile largemouth bass are a particularly effective small, shallow-water piscivore. Thus, largemouth bass predation may be particularly detrimental to native fishes where largemouth bass become established. In the Delta, largemouth bass have become established mainly where submerged macrophytes have proliferated (Nobriga et al. 2005; Michniuk in press).
The population-level impact of largemouth bass on pelagic native fishes and/or their pelagic early life stages may be mitigated by its limited use of openwater (Nobriga et al. 2005) and brackish habitats (Matern et al. 2002). Thus, our finding of a generally high per capita predatory impact may not reflect the cumulative predatory impact of largemouth bass on all potential prey species. For instance, striped bass likely remains the most significant predator of Chinook salmon, Oncorhyncus tschawytscha (Lindley and Mohr 2003), and threatened Delta smelt, Hypomesus transpacificus (Stevens 1966), due to its ubiquitous distribution in the Estuary and its tendency to aggregate around water diversion structures where these fishes are frequently entrained (Brown et al. 1996).

Are Shallow Water Habitats Predation Refuges?
Our study was not specifically designed to evaluate whether shallow-water habitats in the Delta provide predation refuges because we did not attempt to estimate predation rates across a depth gradient. However, the frequent collection of piscivores from the shallowwater habitats we sampled ( Figure 2) suggests these habitats are easily accessed by large and small predators alike. The Delta channels and flooded islands are generally simple habitats lacking extensive marsh development (Brown 2003). Most historical marsh habitats were lost to dredging and levee reinforcement. In much of the Delta, abrupt depth changes occur where narrow shoals abut dredged channels. Several large tidal lakes have formed where levee breaches have flooded islands because the land subsided during the period it was reclaimed for agriculture (Lucas et al. 2002). The sites we sampled for the present study had wadable shoal widths ranging only from 9-39 meters depending on specific location and tidal stage (Nobriga et al. 2005). Thus, we think our results provide some support for the conclusion of Sheaves (2001) that shallow estuarine habitats cannot be generalized as predation refuges, if only because of limited shallow-water habitat complexity.

How Do Our Results Affect Our Understanding of the San Francisco Estuary?
We found that the dynamics of striped bass piscivory differed from largemouth bass and Sacramento pikeminnow because age-1 and older striped bass piscivory was more strongly affected by season than fork length. This is consistent with the findings of Stevens (1966) who showed an intra-annual waxing and waning of striped bass piscivory superimposed on the gradual increase in piscivory between young-of-year and adults. Thus, it appears that striped bass respond strongly to the seasonal cycle of juvenile fish production in the upper Estuary. This may be relevant to hypothesized juvenile striped bass density-dependence (Kimmerer et al. 2000) given long-term declines in upper estuary pelagic fish abundance (Kimmerer 2006).
The upper San Francisco Estuary is an abioticallydriven ecosystem (Jassby et al. 1995;Matern et al. 2002). Abiotic variation (mainly stemming from climate and river flow variation) can strongly influence when and where piscivory substantially affects fish community structure (e.g., Meffe 1984;Rodríguez and Lewis 1994;Henderson and Corps 1997). Presumably, the relative and increasing success of non-native, summer-spawning fishes reflects the abiotic ecosystem changes that have occurred in the Estuary and its watershed (Matern et al. 2002;Feyrer 2004;Grimaldo et al. 2004;Nobriga et al. 2005). During summer, age-1 largemouth bass and Sacramento pikeminnow reach sizes that allow them to forage more efficiently on fishes (Figure 3), and striped bass of multiple ages also increase their use of fish prey (Figure 4; . Thus, the production of piscivore biomass in the upper estuary is linked to this changing fish fauna. This suggests that watershed-scale factors and species invasions that have changed the upper Estuary fish communities probably also control piscivore production. It would be interesting to explore whether piscivore populations based on nonnative fish production can exert significant long-term, top-down impacts on some or all declining native fishes due to decoupling of native fish production from piscivore production.

What Are The Implications For Habitat Restoration?
There is a growing body of evidence that shallow habitats dominated by submerged macrophytes are generally unsuitable for the Delta's remnant native fish fauna. Grimaldo et al. (2004) found very few native fish larvae in macrophyte-dominated habitats, suggesting they are either rarely used as spawning habitats or native fish larvae suffer high mortality when they use submerged macrophytes for spawning. Brown and Michniuk (2007) reported a long-term decline in native fish abundance relative to nonnative fish abundance in the Delta based on shoreline electrofishing. The native fish decline occurred coincident with the range expansion of submerged macrophytes (principally Brazilian waterweed, Egeria densa) and nonnative centrarchid fishes, including largemouth bass. Previously, we reported evidence of comparatively low native fish relative abundance and stronger biotic control of fish community composition in nearshore habitats dominated by submerged macrophytes (Nobriga et al. 2005). Predation by largemouth bass is one mechanism previously hypothesized to result in low native fish abundance where submerged macrophytes proliferate (Brown 2003;Nobriga et al. 2005). In the present study we have presented evidence that largemouth bass, the dominant piscivore of submerged macrophyte-dominated habitats, has a high per capita predatory influence. This is consistent with our previous finding of biotic structuring in this habitat type (Nobriga et al. 2005). We strongly suggest that restoration projects in the Delta need to discourage submerged macrophyte domination regardless of what mechanisms are primarily responsible for low native fish use.