KAISER Joseph E.， SMITH Shannon C. F.， SCHRAMM Jr.， HAROLD L. and EGGLETON Michael A.

(1. Department of Aquaculture and Fisheries， University of Arkansas at Pine Bluff， 1200 N. University， Mail Slot 4912， Pine Bluff，Arkansas 71601， U.S.A; 2. U.S. Geological Survey， Mississippi Cooperative Fish and Wildlife Research Unit， Mississippi State University， Mail Stop 9691， Mississippi State， Mississippi 39762， U.S.A.)

Abstract: The lower Mississippi River (LMR) has been heavily modified for multiple human purposes such as navigation， flood control， and bank stabilization. However， the LMR simultaneously supports a diverse fish fauna that includes recreational and commercial fisheries. Due to river training and diversion structures constructed during the past 80 years， the historic characteristics of the LMR have been drastically altered and have likely influenced fishes and fisheries in the system. One common restoration measure used throughout the LMR has been to “notch” wing-dike structures that close secondary (side) river channels. Dike notching allows year-round flows through secondary channels， which enhances habitat diversity and promotes biological productivity at the ecosystem scale. Although notching is presumed good for LMR fishes and other biota， few studies have examined its effects on fish assemblages. In this study， fish assemblages were sampled at seven LMR secondary channels spanning from river kilometer (rkm) 628 (Louisiana-Mississippi， U.S.A.) upstream to rkm 1504 (Missouri-Kentucky， U.S.A.). Four secondary channels were termed “permanent” (i.e.，with notched dikes) while three secondary channels were termed “temporary” (i.e.， without notched dikes).Fishes were sampled by boat-mounted electrofishing conducted during falling and low stages from 1995—1997. Fish assemblages differed between permanent and temporary secondary channels， and varied somewhat between falling and low stages. Gizzard shad (Dorosoma cepedianum)， threadfin shad (D. petenense)， and white bass (Morone chrysops) demonstrated consistent preferences for low-current conditions associated with temporary secondary channels. Conversely， blue catfish (Ictalurus furcatus)， flathead catfish (Pylodictis olivaris)， and freshwater drum (Aplodinotus grunniens) were more associated with permanent secondary channels. Future restoration strategies in the LMR should consider dike notching and resultant maintenance of permanent secondary channels in selected river reaches. However， temporary secondary channels also contain unique fish species， and also appear to be important sites of riverine primary production. Restoration strategies should consider a balance of both secondary channel types， which should support the greatest biodiversity for the LMR ecosystem.

Key words: Mississippi River; Secondary channels; Dikes; Fish assemblages; River restoration; River rehabilitation

The lower Mississippi River (LMR) consists of the Mississippi River segment running from Cairo，Illinois downstream to the Gulf of Mexico， a reach that is entirely within the U.S.A. The LMR is approximately 1535 km in length， and is the longest freeflowing segment of the Mississippi River main-stem.Although the LMR contains no impoundments， it is still a regulated river wherein the main channel is permanently aligned with dike fields and bank revetments， and isolated from most of its historical floodplain by an extensive levee network[1]. Levee construction in both the LMR main-stem and its tributaries has separated the river from approximately 93%of its historical floodplain[2]. These modifications of the LMR for purposes of navigation， flood control，and bank stabilization began almost three centuries ago and have persisted to the present day[3，4]. Additionally， channel neck cutoffs constructed in the 1930s and 1940s have initiated and augmented channel incision processes that have occurred since over many decades[5]. Channel incision impacts include decreases in annual peak river stage height and longevity， which greatly diminishes the ability of the LMR to inundate its remaining floodplain[6，7]. Over longterm time scales， channel incision is purported to have affected seasonal hydrographs[5]and thermographs[7]，while increasing current velocities[5，8]. These combined changes have likely affected the natural thermal-hydrologic synchrony throughout the LMR[7，9]，which has had unknown， though possible negative， effects on river fisheries.

Although completely restoring the LMR to its natural， unaltered state is not economically or socially feasible， rehabilitation of selected reaches is a successful restorative tool that returns functionality to habitats affected by anthropogenic activities. Ecosystem rehabilitation efforts throughout the Mississippi River have increased notably within the last 40 years[10].Due to the collaborative efforts of multiple state and federal agencies， cost-effective rehabilitation projects have allowed for improvements to habitats important to native and endangered species in the LMR[11，12].Coordinated， multi-agency rehabilitation efforts have targeted increased quality and quantity of available fish， invertebrate， and bird habitats， improved water quality， increased economic opportunities for riverside communities， and increased recreational use and awareness in the LMR[12]. Many of these projects use river training structures engineered for habitat rehabilitation purposes， which include wing-dikes， bank revetments， and bendway weirs[1，10，13].

Large-river secondary channels are referred to by various terms， including chutes， abandoned channels，side channels， and sloughs[1，10]. Herein， they will be referred to as secondary channels. The closure of secondary channels by fields of wing-dikes (herein referred to as dikes) has been a common training practice in large navigable rivers throughout the world[10，13].In the LMR， dike fields on inside channel bends in combination with bank revetments on the outside banks are used for concentrating and redirecting river flow to increase scouring and channel depth， and reduce erosion of outside banks[13，14]. Dike fields also are constructed just upstream and within secondary river channels to prevent channel meandering while maintaining alignment of the navigation channel and reducing the need for dredging. In the LMR， there are approximately 125 separate dike fields， with each consisting of 2—12 dikes in succession along the inside banks of channel bends[4]. Of the 774 dikes in the LMR， approximately 30% (230) have been constructed or retro-fitted with notches[10，11]. Dike notches function to increase channel border diversity， promote scour holes， and allow seasonal reconnection of some off-channel habitats to the main river channel.Notching also allows permanent flow to secondary channels during the lowest stages[12]， which creates what are termed “permanent” secondary channels(Fig. 1)[15]. Conversely， secondary channels containing traditional dikes without notches are termed “temporary” secondary channels. These channels may seasonally disconnect from the main river channel or completely dewater， and often develop more lacustrine characteristics (e.g.， phytoplankton blooms and greater clarities) during low stages in late summer and fall (Fig. 2)[15]. However， by notching selected LMR dikes and creating permanent secondary channels，more diverse habitats are created for LMR fishes[12].The increased habitat heterogeneity provided by secondary channels has the potential to increase species diversity at the ecosystem scale[11，12]. This species diversity is not limited to fishes， but also includes invertebrates， mollusks， birds， and plants[11，13].

Despite that dike notching in secondary channels has been done for many decades throughout the LMR， the effects on fish assemblages have not been extensively studied and formal evaluations (e.g.，[15])have been few[10，11]. The difficulties in quantifying these effects stem from the LMR being a large-river system with many logistical challenges. Furthermore，the LMR is a multi-jurisdictional resource that has obstacles for long-term data collection with respect to funding availability and research priorities that differ across states and management agencies. In any event，given the purported value of dike notching in the context of river rehabilitation， comprehensive studies of the responses of fish assemblages and other biota are warranted.

The primary objective of this study was to assess fish assemblage differences between LMR secondary channels containing notched dikes (i.e.， “per-manent” secondary channels) and those containing unnotched dikes (i.e.， “temporary” secondary channels). Fish assemblages were assessed in three common macrohabitats within each secondary channel type – sandbars， steep natural banks， and the dikes themselves[15]. Sampling was conducted during two different river stages， i.e.， falling stages (usually July-August) and low stages (usually September-October)over a 3-year period from 1995—1997.

1 Materials and Methods

1.1 Study areas

Seven secondary channels located in the LMR between river kilometer (rkm) 1504 and rkm 628 were used for this study. From north to south， these secondary channels were located along 876 kilometers of LMR main channel， with the uppermost and lowermost locations encompassing a 5-degree latitudinal gradient. Four secondary channels were permanent secondary channels， which included Wolf Island(Missouri-Kentucky， U.S.A.， rkm 1504—1497)， Island 8 (Missouri-Kentucky， U.S.A.， rkm 1475—1465)，Chicot Landing (Arkansas-Mississippi， U.S.A.， rkm 908—898)， and Bondurant Towhead (Louisiana-Mississippi， U.S.A.， rkm 634—628) (Fig. 3). Permanent secondary channels were located throughout the entire study reach of the LMR. The other three secondary channels were temporary secondary channels，which included Below Prentiss (Arkansas-Mississippi， U.S.A.， rkm 934—926)， Cracraft Landing(Arkansas-Mississippi， U.S.A.， rkm 821—815)， and Middle Ground Island (Louisiana-Mississippi，U.S.A.， rkm 662—655) (Fig. 3). Temporary secondary channel locations were less widely distributed，with all located along a 279 rkm reach downstream of Rosedale， Mississippi， U.S.A. Regardless of whether permanent or temporary， all seven secondary channels were of similar sizes. Secondary channel widths ranged from 0.4—0.8 km and averaged 0.5 km; channel lengths ranged from 3.6—12.0 km， and averaged 6.4 km[15].

Fig. 1 Wolf Island secondary channel， Missouri-Kentucky， U.S.A.， RKm 1501， near Columbus， KentuckyRiver flows from upper right downstream to the lower center， with the secondary channel on the right and navigation channel on the left.This is a characteristic lower Mississippi River permanent secondary channel， with multiple notches in the main closure dike (denoted by white arrows). Note that the dike notches permit year-round flow to the secondary channel at most stages. These flows maintain greater currents and more depth， and prevent sedimentation and eventual terrestrialization. Image was obtained from Google Earth

Fig. 2 Montezuma Towhead secondary channel， Arkansas-Mississippi， U.S.A.， RKm 1056， near Helena-West Helena， ArkansasThe navigation channel runs downstream from the upper right to the lower center， with the secondary channel to the left. This is a characteristic lower Mississippi River temporary secondary channel. Flow is completely cut off to the secondary channel by sand deposits that accumulated around the upstream dikes (denoted by white arrows)， which prohibit flow to the downstream channel at moderate to low stages.Note the patches of more permanent vegetation that are growing at the upstream end of the sandbar. The habitats remaining in the secondary channel become entirely lentic and have many lacustrine features. Image was obtained from Google Earth

Fig. 3 Sampling locations of lower Mississippi River secondary channels used in this studyAll sampling occurred during falling (July-August) and low-water (September-October) stages during 1995—1997

Within each secondary channel， three macrohabitats were assessed， including dikes， sandbars， and steep natural banks. Dike macrohabitats were characteristic rock-wing structures that extended from the shoreline towards the main river channel (i.e.， thalweg). Dikes were composed of large rock rip-rap with areas of sand substrate; current velocities ranged from 0—1 m/s， with water depths of 1—4 m[15]. In all cases， the primary closure dike， which was usually the largest， was sampled within each secondary channel.Sandbar macrohabitats were representative of gradually sloping areas with gradients of less than 15%. This macrohabitat encompassed from the shore-water interface on the island created by the secondary channel to the deepest portions of the thalweg of the secondary channel. This macrohabitat experienced proportional increases in current velocity from inshore regions towards the secondary channel thalweg (from no flow to 1 m/s). Maximum depths in sandbars ranged from 1—3 m; substrates consisted largely of sand， with occasional large woody debris[15]. Steep natural bank macrohabitats extended from the shoreline to the thalweg of the secondary channel and contained gradients greater than 15%， though gradients exceeding 30%—40% were occasionally encountered when locations contained high amounts of bank sloughing. In this macrohabitat， depths generally ranged from 1—8 m， with current velocities from 0.0—1.5 m/s. Steep natural banks contained heterogeneous substrates consisting of silt， sand， and gravel.Whirlpools， eddies， and segments of upstream flow were common in this macrohabitat due to concentrations of large woody debris and erratic changes in topography due to adjacent sloughing banks[15].

1.2 Boat-mounted electrofishing

All sampling in each macrohabitat and secondary channel was conducted using boat-mounted electrofishing. For the years of study， scientific collection permits were obtained from all LMR border states，which included Missouri， Kentucky， Arkansas， Mississippi， and Louisiana. The electrofishing equipment used included a Smith-Root Model 7.5 GPP (Smith-Root， Inc.， Vancouver， Washington， U.S.A.) used in conjunction with 7500-W Briggs and Stratton generator (Briggs and Stratton， Inc.， Wauwatosa， Wisconsin， U.S.A.). In sandbars and steep natural bank macrohabitats， five 10-min. electrofishing samples were taken. Occasional exceptions occurred when a particular macrohabitat was too small to permit five samples， in which case， four 10-min. samples were taken. In the cases of dikes， three samples ranging from 2—10min. in length were taken dependent on the dike size and other logistics. Over the 3 years of study， this design equated to 535 electrofishing samples being completed. During the timed samples，sample lengths were mostly dependent on local current velocities， but 10min. of electrofishing in swift current could require up to 1 km of shoreline to complete.

For each individual 10-min. sample， two different pulsed-DC electrofishing outputs (500 V-60 Hz and 1000 V-15 Hz) were used[16]. Each output was used for one-half of the 10-min. sample， with the starting output randomly selected upon arrival. This approach insured that each electrofishing sample received equal pedal-time (i.e.， time duration that electric current was flowing in the water) of each output.Equal sampling effort for both outputs on all transects facilitated collection of the most species possible while minimizing selectivity bias[16]. All macrohabitats within each secondary channel were sampled during two different river stages: +1.5—3.0 m above low-water reference plane (LWRP) (usually July-August， i.e.， falling stage)， and +0.0—1.5 m above LWRP (usually September-October， i.e.， low-water stage) during 3 years (1995—1997). Low-water reference plane is defined as the river stage equivalent to a discharge exceeded 97% of the time in the LMR[15].The U.S. Army Corps of Engineers sets and periodically updates LWRP stages for all LMR gages.

All electrofishing samples were taken in a downstream direction. Dike microhabitats were sampled by operating the boat electrofisher approximately 3—5 m from the downstream side of the main closure dike across the entire extent of the secondary channel that could be safely accessed. In temporary secondary channels， this usually included the entire length of the dike; in permanent secondary channels， sampling was restricted to dike areas away from the notch， which tended to be turbulent and difficult to sample safely and effectively. Sandbars were sampled by meandering the boat electrofisher downstream in an S-shaped pattern for the specified sample time in water 1—3 m deep. Steep natural banks were sampled by operating the boat parallel to the shoreline in water 1—6 m deep[15].

1.3 Data analysis

Data analysis was divided into two distinct approaches. The first approach involved statistical comparisons of traditional measures of catch-per-unit-effort (as number of fish/10-min. sample; CPUE) for what were termed to be “prevalent” fishes. Prevalent fish species (n=12) in this study were defined as species， or species groups (e.g.， gars [Lepisosteusspp.])，that represented at least 2% of the combined catch throughout the study. CPUE of prevalent fishes were compared between secondary channel types (permanent and temporary) with repeated-measures analysis of variance (ANOVA) using the MIXED procedure in SAS V.9.2 (SAS Institute， Inc.， Cary， North Carolina，U.S.A.). This analysis tested for the effects of secondary channel type (n=2)， year (n=3)， and year × secondary channel-type interaction (n=6). In these analyses，secondary channel type (i.e.， permanent and temporary) was the main effect， year (1995， 1996， and 1997) was the repeated variable， and individual sampling locations nested within secondary channel type served as subjects (i.e.， a given secondary channel is either permanent or temporary， and cannot be randomly assigned). Location nested within secondary channel type (i.e.， subject) also was specified as a random effect in the model. If the year × secondary channel-type interaction was significant， it was interpreted as a difference between secondary channel types， with the effect not consistent across years. In these cases， a least-square means post-hoc test was used to determine if mean CPUE differed between secondary channel types when adjusted for year (i.e.，marginal means). In cases where the year × secondary channel-type interaction was not significant， then the secondary channel main effect was interpreted directly. If this term was significant， the analysis was interpreted as secondary channel types differed for the measure in question， and that the difference was consistent across years. A significant year effect was not of primary interest， but indicated there was a consistent upward or downward trend in mean CPUE of the prevalent species being tested for both secondary channel types. In light of suspected power issues and concern for Type II error rates， significance for all analyses was set at an alpha level of 0.10. These analyses were conducted separately for falling and lowwater stages for each prevalent species.

The second approach used multivariate ordination to examine overall fish assemblage compositional differences between permanent and temporary secondary channels. Non-metric multidimensional scaling (NMS) using a Bray-Curtis dissimilarity measure was conducted on boat-mounted electrofishing data collected at falling and low-water stages during 1995—1997. Because of the relatively high incidence of uncommon species in the dataset， a species exclusion rule was applied to reduce the frequency of zeros in analyses. Specifically， a species was excluded from these analyses if catches during a given stage represented less than 0.1% of the total catch during the study. This criterion allowed for 17—23 species to be included in each NMS analysis. Six separate NMS analyses were performed – one for each macrohabitat (i.e.， dike， sandbar， and steep natural bank) within each stage (i.e.， falling and low-water).For each analysis， dimensionality was evaluated and chosen to minimize stress. Additionally， if a species was present in a given stage but not present in a given macrohabitat， that species was excluded from the NMS analysis of said macrohabitat. The influence of secondary channel type on fish assemblage structure was assessed using the ENVFIT function within NMS for all six analyses. The ENVFIT function fits a linear regression model that reflects the amount of within-NMS variation explained by a given variable;models were generated from 999 permutations of the data matrix using a Markov-Chain Monte Carlo algorithm[17]. Statistical significance for all NMS analyses also was set at an alpha level of 0.10. All NMS analyses and significance testing were performed using the program R， Version 3.3.2 (R Foundation for Statistical Computing， Vienna， Austria)， using the“VEGAN” package[17].

2 Results

Three years of electrofishing sampling in LMR secondary channels collected 8934 fishes from 39 species representing 15 families (Tab. 1). Supplemental sampling conducted for other purposes outside of LMR secondary channels and at other stages yielded nine additional species， though all were collected in very low abundances[15].

2.1 Dike macrohabitat

A total of 1122 fishes from 25 species representing 11 families were collected at dike macrohabitats during 1995—1997 sampling (Tab. 1). Total fish abundance at dikes was generally the least of the three macrohabitats sampled， usually ranging from 1.4—2.4 fish/min. Total fish abundance at dikes varied little between secondary channel types， and CPUE of prevalent species did not vary greatly between secondary channel types or through time. Although gizzard shad (Dorosoma cepedianum) were consistently three- to four-fold more abundant in temporary secondary channels (Fig. 4)， adjusted mean CPUE was significantly greater only during low-water stages (P=0.022) (Tab. 2). Threadfin shad (D.petenense) exhibited similar trends to gizzard shad，and was up to 10-fold more abundant in temporary secondary channels (Fig. 4)， though differences were not significant due to the high incidence of zero catches in permanent secondary channels (Tab. 2).White bass (Morone chrysops) were on average six-

fold more abundant in temporary secondary channels，though the increase was significant only during the falling stage (P=0.001) (Tab. 2). Similarly， river carpsucker (Carpiodes carpio) CPUE was greater in temporary secondary channels during the low-water stage (P=0.095) (Tab. 2). No river carpsucker were collected at dikes during falling stages.

Tab. 1 Master list of all fish species captured from lower Mississippi River secondary channels and their percent composition in each macrohabitat and in total， 1995—1997

In permanent secondary channels， blue catfish(Ictalurus furcatus) were consistently more abundant，especially at low-water stages (Fig. 4). However， blue catfish CPUE in temporary secondary channels was so highly variable that differences could not be detected (P=0.251—0.413) (Tab. 2). Flathead catfish(Pylodictis olivaris) were consistently more abundant in permanent secondary channels， but increases were only detectable during low-water stages (P=0.072)(Tab. 2; Fig. 4). Additionally， freshwater drum(Aplodinotus grunniens) were significantly more abundant in permanent secondary channels， though only during falling stages (P=0.048) (Tab. 2). The lack of consistent statistical findings across both stages was a common observation at dikes， underscoring the high variation in CPUE estimates.

Fig. 4 Mean catch-per-unit-effort (CPUE) of prevalent fish species from dike macrohabitats in the lower Mississippi River， 1995—1997Dashed lines represent temporary secondary channels， whereas solid lines represent permanent secondary channels. Sampling during falling stage is shown on left; sampling during low-water stage is on right. Standard errors not shown to avoid cluttering in the graphs

After applying the species exclusion rules for the multivariate NMS analyses， 17 and 21 species were included for dike analyses in falling river and low-water stages， respectively. NMS analyses indicated that fish assemblages at dike macrohabitats were the most similar between secondary channel types of the three macrohabitats sampled. Similarities were stronger during falling as opposed to low-water stages， where there was a notable difference in assemblage composition (ENVFITP=0.047) (Fig. 5). This observation appeared to be driven by the stronger associations of gizzard and threadfin shads， white bass， common carp(Cyprinus carpio)， and to a lesser extent， striped mullet (Mugil cephalus)， with temporary secondary channels (Fig. 5). During low-water stages， assemblages became more distinct between secondary channel types (ENVFITP=0.002). In particular， gizzard and threadfin shads， white bass， river carpsucker， and to a lesser extent， largemouth bass (Micropterus salmoides)， were more associated with temporary secondary channels (Fig. 5).

During falling stages， NMS analyses indicated that flathead and blue catfishes， freshwater drum， and to a lesser extent， smallmouth buffalo (Ictiobus bubalus)， were more associated with permanent secondary channels (Fig. 5). At low-water stages， permanent secondary channel assemblages continued to be dominated by blue and flathead， catfishes， and freshwater drum， but also included more skipjack herring (Alosa chrysochloris). There also were several species weakly associated with permanent secondary channels – these included bigmouth buffalo (I. cyprinellus)， smallmouth buffalo， and goldeye (Hiodon alosoides) (Fig. 5).

2.2 Sandbar macrohabitat

Tab. 2 Mean catch-per-unit-effort (CPUE) and standard errors， and P-values for repeated-measures ANOVA effects for prevalent fish species from dike macrohabitats of permanent and temporary secondary channels in the lower Mississippi River

A total of 2160 fishes from 30 species representing 13 families were collected from sandbar macrohabitats during 1995—1997 sampling (Tab. 1). Total fish abundance in sandbars was intermediate of the three macrohabitats sampled. Total fish abundance typically ranged from 3.3—4.9 fish/min， with consistently greater values found in temporary secondary channels. Although more species were collected at sandbars compared to dikes， mean CPUE for most species was often much less at sandbars compared to other macrohabitats. One clear finding was that gizzard shad were significantly more abundant in temporary secondary channels during both falling(P=0.064) and low-water (P=0.009) stages (Tab. 3;Fig. 6). As with dikes， threadfin shad were far more abundant in temporary secondary channels， though increases were never significant due to high variation in permanent secondary channel CPUE (Tab. 3; Fig. 6).Flathead catfish were significantly more abundant in permanent secondary channels during low-water stages (P=0.096); trends were similar during falling stages， though not significant (P=0.150; Tab. 3). Despite the difference during low-water stages， flathead catfish CPUE was never high in sandbars regardless of secondary channel type (Fig. 6). Although differences were not significant (P=0.112—0.189) due to high CPUE variation in temporary secondary channels， gars (Lepisosteusspp.) were consistently more abundant in permanent secondary channels.

Fig. 5 Nonmetric multidimensional scaling (NMS) ordination plots for individual macrohabitats: dikes (a， b)， sandbars (c， d)， and steep natural banks (e， f) during falling (left) and low-river (right) stagesDashed lines are groupings based on mean levels of factors， with solid lines indicating 95% confidence intervals for those groupings generated from the ENVFIT algorithm in R. Gray lines denote temporary secondary channels (i.e.， with unnotched dikes) while black denotes permanent secondary channels (i.e.， notched dikes)

NMS ordinations of sandbars included 19 and 22 species during the falling stage and low-water stages，respectively. Fish assemblage composition was distinct between the two secondary channel types at both stages (falling: ENVFITP=0.011; low-water: ENVFITP=0.008). NMS results were consistent with repeated-measures analyses and showed several consistent species-channel type associations regardless of stage. Fish assemblages in temporary secondary channels were dominated largely by gizzard and threadfin shads (Fig. 5). In contrast， permanent secondary chan-nel assemblages contained greater abundances of blue， flathead， and channel catfishL.Punctatus， freshwater drum， goldeye， gars， skipjack herring， and river carpsucker (Fig. 5).

2.3 Steep natural bank macrohabitat

A total of 5652 fishes from 36 species representing 14 families were collected at steep natural bank macrohabitats during 1995—1997 sampling (Tab. 1).Total fish abundance at steep natural banks was consistently the greatest of the three macrohabitats sampled. Total fish abundance ranged from 5.0—9.4 fish/min， with consistently greater abundances recorded in temporary secondary channels. Similar to sandbars， gizzard shad were significantly more abundant in temporary secondary channels during both falling (P=0.010) and low-water (P= 0.027)stages (Tab. 4; Fig. 7). Threadfin shad exhibited identical trends， being far more abundant in temporary secondary channels regardless of stage (P=0.008—0.057) (Tab. 4; Fig. 7). All three catfishes tended to be more abundant in permanent secondary channels.However， significance was limited only to flathead catfish during low-water stage (P=0.017) (Tab. 4)，with blue catfish and channel catfish during low-water stage and flathead catfish during falling stage only near significant (P=0.112—0.176). Otherwise， smallmouth buffalo were significantly more abundant in temporary secondary channels， but only during lowwater stages (P=0.094) (Tab. 4).

Tab. 3 Mean catch-per-unit-effort (CPUE) and standard errors， and P-values for repeated-measures ANOVA effects for prevalent fish species from sandbar macrohabitats of permanent and temporary secondary channels in the lower Mississippi River

After exclusion of uncommon species， 19 and 23 species were included in NMS ordinations for the falling and low-water stages， respectively. Like sandbars，fish assemblages at steep natural banks contained several strong species-channel type associations regardless of stage (falling: ENVFITP=0.002; low-water:ENVFITP=0.001). Most species in temporary secondary channels had relatively low abundances as assemblages were dominated by gizzard and threadfin shads (Fig. 5). Although the association was weaker， largemouth bass showed some affinity for steep natural banks in temporary secondary channels，probably due the consistent structure located within.Conversely， assemblages in permanent secondary channel assemblages were dominated by blue， flathead， and channel catfishes， freshwater drum，goldeye， gars， and skipjack herring (Fig. 5).

3 Discussion

3.1 Temporary secondary channels

Temporary secondary channels in the LMR contained fewer species (31) overall compared to permanent secondary channels (36). However， temporary secondary channels often contained fish abundances of two-fold or more greater than those found in permanent secondary channels. Greater abundances in temporary secondary channels were consistently driven by the high abundances of gizzard and threadfin shads that existed in the lower-current habitats. At lower water stages， upstream inflows are eliminated and many of these channels become more lentic in that they contain no or little flow， develop phytoplankton blooms， have lower turbidities， and are conducive to detrital accumulation[12，18]. Thus， at lower stages occurring seasonally， these habitats become major centers of primary and secondary production for the LMR[2，4，15]. Because permanent secondary channels contained notched dikes that allowed continuous year-round flow， their environment was not optimal for trophic requirements of LMR clupeids， which dominated assemblages in temporary secondary channels.

Fig. 6 Mean catch-per-unit-effort (CPUE) of prevalent fish species from sandbar macrohabitats in the lower Mississippi River， 1995—1997Dashed lines represent temporary secondary channels， whereas solid lines represent permanent secondary channels. Sampling during falling stage is shown on left; sampling during low-water stage is on right. Standard errors not shown to avoid cluttering in the graphs

The dominance of clupeids in these types of LMR habitats has been previously documented. Gizzard shad flourish in temporary secondary channels because of enhanced feeding opportunities due to their high planktivory and additional use of detritus as a food source[15，19]. Previous studies have reported that gizzard shad (both juveniles and adults) were more abundant in temporary secondary channels， particularly in steep bank habitat[20]. Similarly， Baker，et al.[1]categorized gizzard shad as common in nearly all lentic habitats throughout the LMR. Temporary secondary channels appeared to be important sites of gizzard shad production within the LMR main-stem，which likely constitutes the largest proportion of the fisheries forage base. Although high catches were observed in temporary secondary channels， threadfin shad catches were extremely low or zero in permanent secondary channels. This was especially true during lower stages when threadfin shad abundances were 5% or less than that of gizzard shad. Like gizzard shad， threadfin shad are highly planktivorous，and prefer low-current lentic habitats associated with temporary secondary channels[19，21]. Pennington，et al.categorized threadfin shad as a “frequent” species in LMR temporary secondary channels， but “infrequent”in permanent channels[20]. Baker，et al. also categorized threadfin shad as common in a variety of lentic environments throughout the LMR[1]. Although temporary secondary channels also appeared to be important sites of threadfin shad production， this species appears to be a lesser component of the LMR fisheries forage base than gizzard shad.

Tab. 4 Mean catch-per-unit-effort (CPUE) and standard errors， and P-values for repeated-measures ANOVA effects for prevalent fish species from steep bank macrohabitats of permanent and temporary secondary channels in the lower Mississippi River

Despite often being considered a lotic species[22，23]，white bass were more frequently associated with LMR temporary secondary channels. Although white bass need flowing-water habitats for spawning， they have also previously been characterized as fairly plastic， utilizing both lentic and lotic habitats[24]. Summer and fall sampling that occurred in this study may have missed seasonal white bass use of more lotic habitats. Using a similar sampling design to this study， Driscoll reported no detectable patterns in white bass abundances among habitat types or between secondary channel types in the LMR[25].Baker，et al. reported white bass to be common in both steep natural bank and sandbar habitats regardless of secondary channel type[1]. However， they also reported white bass to be uncommon to absent in lentic habitats such as oxbow lakes， sloughs， and borrow pits. Interestingly， Chadwick，et al. suggested that the reduced availability of some low-flow habitats (e.g.，oxbow lakes) in the LMR might be related to extirpations of white bass in some parts of Louisiana，U.S.A.[24]. Results with white bass were particularly important in light of its status as a popular sport fishery in many U.S.A. river systems.

Several other species exhibited some association for temporary secondary channels， though these preferences were weaker and not always consistent. NMS analyses suggested that smallmouth buffalo， river carpsucker， striped mullet， common carp， and largemouth bass ordinated in multivariate space as species more abundant in temporary secondary channels. The observation with striped mullet may be spurious， as the species is only seasonally present in the LMR and usually in the main river channel or adjacent border habitats[15，26].

3.2 Permanent secondary channels

Fig. 7 Mean catch-per-unit-effort (CPUE) of prevalent fish species from steep natural bank macrohabitats in the lower Mississippi River，1995—1997Dashed lines represent temporary secondary channels， whereas solid lines represent permanent secondary channels. Sampling during falling stage is shown on left; sampling during low-water stage is on right. Standard errors not shown to avoid cluttering in the graphs

Despite exhibiting greater overall species richness (36)， permanent secondary channels in the LMR contained fish abundances that were typically only 50%—70% of those found in temporary secondary channels. Permanent secondary channels consistently contained lower abundances of gizzard and threadfin shads and greater abundances of fluvial species，ictalurids in particular[27]. Additionally， permanent secondary channels supported most of the more signi-ficant sport and commercial fisheries in the LMR[28].Based on field observations， these habitats also were frequented more often by recreational and commercial fishermen.

In general， all three ictalurid catfishes exhibited strong preferences for flowing water habitats in LMR permanent secondary channels， though findings were not always supported with statistical significance[15].Blue catfish prefer a variety of lotic habitats， including deep channels or large pools with swift flows[23，26，27].However， Jackson also described blue catfish as preferring open-water habitats such as those associated with main river channels and large reservoirs[29]. Prior to this study， blue catfish were found in moderate to high abundances within the main channel， secondary channels， large floodplain lakes， and dike habitats throughout the LMR[15，30]. Similarly， flathead catfish were the most abundant species in this study that was strictly piscivorous. This species is considered an integral component of main-stem fisheries throughout the LMR[7，26，30]. Similar to blue catfish， flathead catfish indicated clear preferences for flowing waters associated with permanent secondary channels， especially along steep natural banks[31]. Although flathead catfish have been reported as inhabiting both lentic and lotic habitats， most studies indicate that fluvial habitats are preferred[26]， especially when containing high amounts of physical structure[15，31].

Channel catfish are a more generalist species and have been reported from a wide variety of riverine，floodplain， and reservoir habitats[19，28，32—34]However，several studies have reported preferences for more lotic conditions[20，25，35]. Findings from this study supported that contention， as CPUE of channel catfish was consistently greater in LMR permanent secondary channels， especially at sandbar and steep bank macrohabitats during low-water stages. Nearly all analyses indicated that these three ictalurids were dominant components of the LMR food web[15，36].

Freshwater drum are ubiquitous throughout the LMR[37]， and tolerant of many different flow and turbidity conditions[22]. Although freshwater drum exhibited a weaker affinity for permanent secondary channels than the three ictalurids， this preference was not always consistent， as the species generally occurred in all macrohabitats regardless of secondary channel type. Given their wide distributions across locations and habitats， there was a high likelihood that this species uses multiple LMR habitats seasonally. However，other studies also have reported freshwater drum to be predominant in flowing-water type habitats similar to those found in the permanent secondary channels of this study[20，38].

Several other LMR species associated with permanent secondary channels， though patterns were weaker and inconsistent since these species also were found in temporary secondary channels. NMS analyses suggested that skipjack herring， goldeye， river carpsucker， smallmouth buffalo， and common carp associated with species typical of permanent secondary channels. Although skipjack herring and goldeye are commonly considered lotic species， common carp，river carpsucker， and smallmouth buffalo also are fairly plastic species that are collected from a wide variety of large-river habitats[15]. Although relatively rare throughout LMR secondary channel sampling，blue sucker (Cycleptus elongatus) were collected exclusively from permanent secondary channels in this study.

3.3 Template for river rehabilitation

The LMR historically meandered across the alluvial floodplain forming secondary channels through natural channel cut-off processes[39]. Although these secondary channels were diverse in size and complexity， they were gained and lost as the LMR formed new channels to the Gulf of Mexico through time[40].In modern times， levee systems， bank revetments， and dike fields have been constructed to stabilize the river channel and its floodplain. Dikes and revetments combined have reduced hydraulic connections between the main river channel and floodplain[41]， and caused large-scale sediment aggradation within dike fields that has resulted in loss of aquatic habitat[10].These channel alignment strategies have greatly reduced， and in some cases eliminated， the formation of new secondary channels[13].

The total number of functioning secondary channels in the LMR is dependent on stage. At greater discharges common from late-winter through early-summer， lateral flow reconnects many secondary channels that otherwise completely dewater at lower stages. Williams and Clouse reported from hydrographic surveys that the LMR contained up to 145 secondary channels during the period from 1960 to 2000[40]. By the 1990s， fewer than 100 secondary channels remained， as channels filled with sediment over time， with some becoming terrestrialized[11].Consequently， LMR secondary channels have become a finite resource， and have been decreasing in number due to sedimentation and loss of connectivity with the main channel[12].

Killgore，et al. recommended use of a Priority Index to rate the relative benefit and cost-effectiveness of selected secondary channels for rehabilitation[11].This index combined a video-based habitat quality index with estimated costs of rehabilitation. This approach would be a valuable tool for ranking LMR secondary channels for potential rehabilitation， particularly with respect to dike notching to create permanent secondary channels. Permanent and temporary secondary channels in this study had distinct fish assemblage structures， with some species exhibiting associations for both channel types. In general， permanent secondary channels with notched dikes tended to benefit more traditionally riverine or lotic species， whereas temporary secondary channels with unnotched dikes created conditions conducive to planktivorous species that were more lentic and comprised the majority of the fisheries forage base. Additionally，temporary secondary channels support species that would be abundant in floodplain habitats， which have decreased significantly in the LMR compared to historical times[2]. Thus， those groups interested in restoration of selected LMR secondary channels should be mindful of these characteristics.

Not all previous research corroborates the findings in this study. Working with naturally occurring sandbars in the Missouri River， Schloesser，et al.could not detect differences in species richness， diversity， or occupancy by native fishes between notched and unnotched dikes[29]. As with this study，their research indicated that， although notched dikes created more habitat for fluvial species， abundances and occupancies for most Missouri River fishes was similar in sandbars at unnotched dikes. The findings in this study also contradicted previous research conducted on LMR secondary channel types where no differences in fish assemblages were detected[42]. A similar study in the upper Mississippi River demonstrated no differences in relative abundances of fish between riverine channel (high current)， lacustrine channel (no current)， and intermediate successional channel (low current) types from June-October[43].Results of the current study were actually more similar to those of Braun，et al.[27]， who reported that species abundances differed among notched， unnotched，and L-head dikes in the middle Mississippi River for a few species， with multivariate ordination suggesting some separation between fish assemblages at notched and unnotched dikes.

Similarly， we documented differences in species abundance and composition among macrohabitats and between secondary channel types. Differences may be attributable to seasonal flow gradients experienced with the macrohabitats. Each macrohabitat contained an apparent water-flow gradient from near shore outward that provided heterogeneous habitats for the prevalent fish species， particularly during low stages[15].Flow interacts with many physical properties in large rivers besides current velocities. These factors include sedimentation， substrate type and stability， turbidity， dissolved oxygen， pH， detrital accumulation，and the diversity and abundance of macrobenthic organisms[22，26，44]. Changes in these factors can be observed seasonally in which flow regimes include both high and low water levels， and can either deter or attract species to different macrohabitats at certain times of the year. Madejczyk，et al. observed that certain fish species exhibited local preferences for certain habitat types[45]. They also reported that the overall fish assemblage within channel border habitats was greatly influenced by the dominant types of artificial habitat (e.g.， dikes) provided. Examining secondary channels at the macrohabitat scale， as done in this study， allowed better inference of the effects that dike notching may have on fish assemblages， compared to studies that only examined secondary channel differences with a single habitat type (e.g.，[42，44]).

4 Conclusions

The LMR has experienced many alterations that have affected its form and function over decadal time scales. However， notching of wing-dike structures could be implemented in conjunction with routine maintenance. This practice provides a cost-effective manner to recover natural flows through sections of the LMR containing secondary channels， with the overall aim of improving habitat diversity for many aquatic taxa. Prioritizing secondary channels that need rehabilitation (e.g.，[11， 12]) could further reduce cost and maximize the benefit of dike notching.Selection of certain dike structures within a dike field also will assist in maintaining the effectiveness of a particular dike field as a river training structure.

This research emphasizes differences in permanent and temporary secondary channels for multiple fish species. For some species， use of permanent or temporary secondary channels transcended macrohabitat type and was consistent during both summer and fall flows. Other species preferred certain macrohabitats during a certain discharge， whereas many others showed little or no preference for either secondary channel type. This latter category of species would be consistent with most large-river fishes having fairly plastic and generalized life-history requirements[19]，which is a common observation in rivers around the world. Examination of permanent and temporary secondary channels at the macrohabitat scale provided clearer insights into how certain species may utilize the two channel types. We would encourage such an approach in future studies in the LMR and other rivers. It is also important to note that forage fishes integral to the LMR fisheries exhibited a greater preference for temporary secondary channels containing unnotched dikes. Thus， maintaining habitat heterogeneity throughout the LMR will likely require a variety of dike structures creating a diversity of secondary channel characteristics.

Thorough evaluation of the effects of dike notching requires assessment of fish communities during all seasons. Future research should focus on seasonal monitoring of fish movements in order to assess secondary channel use by fishes during all seasons，though additional sampling gears may need to be deployed. Additional research would ideally improve understanding of fisheries dynamics in the LMR (including movement/migration studies)， and enhance species-specific knowledge of LMR fishes with respect to population studies， habitat use and suitability，and bioenergetics.

Acknowledgements:

Special thanks are extended to Larry Pugh， Rob Mayo， Todd Driscoll， and Richard Davis for assistance with fish sampling.