Habitat Use by Telemetered Alabama Shad During the Spawning Migration in the Lower Flint River, Georgia

Abstract

The Alabama Shad Alosa alabamae is an anadromous clupeid that lives in the northern Gulf of Mexico and ascends freshwater rivers in spring to spawn. Populations have experienced substantial range-wide declines due to habitat alteration. The largest known population of Alabama Shad is found in the Apalachicola River in northwest Florida. To assess their movement during the spawning migration, 250 Alabama Shad were fitted with acoustic or radio transmitters, depending on year, and transported upstream from Jim Woodruff Lock and Dam during 2010–2014. The 153 relocations from 126 individual fish revealed congregation areas that were suspected to be spawning locations. Alabama Shad selected limerock boulder substrate, avoided limerock fine and rocky substrates and used sandy substrate in the same proportion as its availability. Alabama Shad upstream movement was greatest during April through mid-May. Movements ≥ 20 km were generally clustered together over a period of a few days, with ~90% of such movements occurring following periods of increased river discharge. All limerock boulder substrate areas on the lower Flint River were identified to focus future efforts to determine exact spawning locations.

Received July 11, 2016; accepted April 27, 2017

Six species of Alosa are found in North America, and all but Skipjack Herring A. chyrsochloris have an anadromous reproductive strategy, whereby adults ascend freshwater rivers to spawn after reaching sexual maturity in marine environments (Laurence and Yerger 1967; Kissil 1974; Loesch and Lund 1977; Mettee and O’Neil 2003; Harris et al. 2007). The Alabama Shad A. alabamae is one of these anadromous clupeids that lives in the northern Gulf of Mexico (Ross 2001; Mettee and O’Neil 2003).

The historic distribution of Alabama Shad ranged from the Suwannee River in northwest Florida westward to the Mississippi River. Alabama Shad were reported to have traveled as far inland as the lower Ohio and Missouri rivers and the Mississippi River in central Iowa (Coker 1930; Lee et al. 1980; Pattillo et al. 1997). There are no known estimates of historical abundance, although anecdotal observations indicate that Alabama Shad populations were once large enough to support commercial and recreational fisheries (Coker 1930). However, Alabama Shad have experienced substantial range-wide declines over the past century. Collection data over the past 30 years have revealed that this species has become restricted to small portions of its former range, and most records consist of a few individuals collected at sporadic intervals (Smith et al. 2011). The National Marine Fisheries Service has recently determined that listing Alabama Shad as threatened or endangered under the U.S. Endangered Species Act is not warranted (NMFS 2017). Dam construction that limits access to historical spawning grounds is likely the primary reason behind the population declines of Alabama Shad (Boschung and Mayden 2004). The largest known population spawns in the Apalachicola River in northwestern Florida (Mettee et al. 1996; Ross 2001; Boschung and Mayden 2004).

In 2005 a multifaceted study was undertaken to better understand the Alabama Shad population in the Apalachicola–Chattahoochee–Flint (ACF) river basin in southwestern Georgia and northwestern Florida (Young et al. 2012). Alabama Shad were found to successfully pass through the lock chamber at Jim Woodruff Lock and Dam (JWLD) during their spawning run, accessing spawning habitat upstream from Lake Seminole. Schaffler et al. (2015) used otolith microchemistry to determine that 86% of adult Alabama Shad returning to JWLD were spawned in the lower Flint River, illustrating the importance of the lower Flint River to this population. However, little is known about spawning habitat of this species upstream of JWLD or elsewhere.

The closely related American Shad A. sapidissima (Nolan et al. 2003) broadcasts eggs over benthic substrates (Walburg and Nichols 1967; Jenkins and Burkhead 1994) composed of sand, gravel, and/or limestone in areas with moderate currents (Laurence and Yerger 1967; Mills 1972; Fox et al. 2000; Mettee and O’Neil 2003). Methods used to characterize substrate use by American Shad have typically involved point sampling (Beasley and Hightower 2000) or transect sampling in the vicinity of suspected spawning activity (Bowman 2001; Hightower and Sparks 2003). However, advances in side-scan sonar technology over the past decade have led to this technology now being used for fast, affordable, and effective substrate characterization in lotic habitats (Kaeser and Litts 2010). This method shows promise to allow identification and quantification of alosine spawning habitat at the reach scale, but currently this has not been attempted.

Alabama Shad may share characteristics of spawning biology with American Shad, but the lack of information regarding substrate use and movement of Alabama Shad during the spawning migration represents a substantial knowledge gap in the life history of this species. Therefore, objectives of this study were to (1) combine telemetry results with a detailed substrate map to describe substrate use by Alabama Shad during the spawning migration and (2) describe movement patterns of Alabama Shad during the spawning migration. Results from this study will help guide future management efforts, which are expected to include more precise identification of Alabama Shad spawning sites.

STUDY SITE

The ACF river basin has its headwaters in northern Georgia. Both the Chattahoochee and Flint rivers flow south and west and join at Lake Seminole, forming the Apalachicola River downstream of JWLD near the Florida–Georgia–Alabama state lines (Figure 1). While the Chattahoochee River has been heavily impounded and fragmented with 10 dams upstream from JWLD, the Flint River has only two dams upstream from JWLD, both near Albany, Georgia. This dam is the only one in the ACF river basin that currently operates the lock chamber to provide seasonal passage to anadromous fishes (Young et al. 2012). The approximately 145-km reach of the Flint River from Albany Dam downstream to JWLD (Figure 1) is commonly referred to as the lower Flint River. Geomorphology of this reach of river is characterized by limestone outcroppings, a deep channel, and a hydraulic connection to the Upper Floridan aquifer resulting in large groundwater inputs (Opsahl et al. 2007). Land use within this reach is dominated by agriculture; however, the river corridor remains mostly forested and other than the area around Albany, Georgia (2013 population, ~76,200), there is little urban development within the watershed. The numerous springs found throughout this reach add to stream flow and provide summer thermal refugia for species such as the Gulf of Mexico strain Striped Bass Morone saxatilis (Van Den Avyle and Evans 1990; Opsahl et al. 2007; Ingram and Kilpatrick 2015).

FIGURE 1. The lower Apalachicola–Chattahoochee–Flint River basin, including the lower Flint River from Jim Woodruff Lock and Dam to Albany, Georgia. Only those locations from Albanay Dam downstream to Bainbridge, Georgia, that corresponded to the lower Flint River substrate map were used for analysis.

METHODS

Telemetry

From 2010 to 2014, 288 Alabama Shad were tagged with either acoustic transmitters (2010; Vemco, Halifax, Nova Scotia; model V7-2L; Young et al. 2012) or radio transmitters (2011–2014; Advanced Telemetry Systems, Isanti, Minnesota; models F1030 in 2011 and F1440 in 2012–2014; Sammons and Young 2012; Young et al. 2012; Sammons 2013, 2014). During the 5-year study, the numbers of fish outfitted with transmitters varied from 30 to 86 (Table 1). Fish were collected from the tailwaters of JWLD using boat electrofishing and angling. Tags were inserted into the esophagus following the methods of Young et al. (2012), then fish were transported to and immediately released from one of two locations upstream from JWLD (Figure 1). Across the years, 38 fish were considered to be tagging and/or transport mortalities, so that of the original 288 fish, 250 were stocked. Stocking locations were altered in 2013 and 2014 to allow tagged fish a more equitable choice between ascending either the Flint or Chattahoochee rivers. The entire study area, from Albany Dam downstream to Lake Seminole, was traversed by boat once a week to locate tagged shad, which required up to 3 d per week to accomplish. Telemetry was conducted regardless of river discharge, including during extremely low and high discharges. Spring Creek and the Chattahoochee River arm of Lake Seminole up to George W. Andrews Lock and Dam were also frequently searched for tagged fish, but few fish were ever located outside the Flint River (fewer than five locations throughout the study). In 2013 and 2014, a fixed array system was located at George W. Andrews Lock and Dam to detect any tagged fish in the tailrace, but no detections were recorded in either year. Therefore, only locations from the Flint River were used in these analyses.

TABLE 1. Number, release dates, tag type, and release location for Alabama Shad outfitted with transmitters during 2010–2014.

Number	Release dates	Tag type	Release location
30	April 14–15, 2010	Acoustic	Bainbridge, Georgia
50	March 17–20, 2011	Radio	Bainbridge, Georgia
48	March 10–13, 2012	Radio	Bainbridge, Georgia
74	April 2, 3, 11, and 19, 2013	Radio	4.5 km upstream from JWLD
86	April 8–10, 2014	Radio	4.5 km upstream from JWLD

In 2010, fish were detected by boat using a Vemco VR-100 general purpose acoustic receiver (Young et al. 2012). During radio-tracking activities in 2011–2014, fish were located by boat using an Advanced Telemetry Systems R-200 receiver. In all years, fish were located by decreasing the receiver gain as the boat neared the fish location to provide directionality to fish location. Once the general location of a fish was determined, the boat was then driven over the fish’s location in a manner such that the tag could only be heard when the YAGI antenna (or hydrophone) was pointed directly beneath the boat. Signal bearings were used to generate a best estimate of fish location when turbulent flow, presence of complex bedrock shoal habitat, or other abiotic variables reduced the precision of telemetry efforts. A waypoint was taken once the best location for a telemetered fish was determined. All tracking was conducted by experienced personnel who had been trained by the investigators prior to the study using known-location tags (Brenden et al. 2004). Tracking began immediately following stocking and ceased when the spawning migration was over, usually in late May. Of the 250 tagged fish, 126 were eventually relocated, with the vast majority (~97%) of these relocations occurring in the lower Flint River and Lake Seminole. There were 153 total locations during 2010–2014 after removing mortality and same-day repeat locations. Multiple relocations of the same fish within the estimated position error of GPS receivers (typically 1–3 m) over a multiple-week period were considered to be mortalities. Only those relocations in which a fish was not found in the same place during subsequent relocations were considered to be from live fish. While those relocations attributed to fish mortalities (n = 30) could be attributed, in part, to tag regurgitations, it would be impossible to discern a difference between those two transmitter fates.

Substrate map

Kaeser et al. (2013) developed a high-resolution, landscape-level inventory of aquatic substrates in the lower Flint River using images collected with side-scan sonar during high river discharge. Images were georeferenced and substrate types were delineated in ArcGIS software (ESRI 2013) as individual polygons. Site visits and underwater cameras were used to ground-truth substrate types after classification. The resulting product was a detailed mosaic of the lower Flint River broken into substrate types.

Analysis

Mapped substrate was divided into categories following Kaeser et al. (2013) with the exception of unsure sandy and unsure rocky, which were converted to sandy and mixed rocky substrate types, respectively, due to the likely composition of these areas despite the difficulty in classification from sonar images (Table 2). This conversion was based upon the suspected composition of unsure substrate types drawn by Kaeser et al. (2013), which was based on ground-truthing of sandy and rocky substrate sonar images and knowledge of local geomorphic processes. While unsure sandy and mixed rocky substrate types on sonar images typically had characteristics of known sandy and mixed rocky substrates, complicating factors such as shallow water depth, sonar shadows, and boat speed and orientation during data collection reduced certainty in classification of images into particular substrate groups. Similarly, the substrate types, no data and sonar shadow in Kaeser et al. (2013), were included in a single missing data substrate type. To further increase the number of fish locations per substrate type, the substrate types rocky boulder, rocky fine, and mixed rocky were combined into a single “rocky” substrate due to the low numbers of observations in each of those substrate categories (McDonald 2014).

TABLE 2. Substrate class percent composition and associated definitions developed for the lower Flint River substrate map (Kaeser et al. 2013) and adapted for this study.

Substrate class	Composition	Definition
Sandy	46%	>75% of area of composed of particles < 2 mm in diameter and unsure sandy.
Limerock fine	17%	≥75% of the area composed of limestone as bedrock,or an outcropping (not fractured in blocks > 500 mm in diameter).
Rocky boulder	11%	An area ≥ 314 m^2 that includes three or more boulders, each> 500 mm across longest axis, and each boulder within 1.5 mof the next adjacent boulder, regardless of underlying substrate.
Rocky	6%	>25% of area composed of rocks 2–500 mm across the longest axis.
Missing data	6%	No data (out of range but within river channel) and sonar shadow(within range but behind reflective objects).
Mixed rocky	9%	Two or more substrates (at least one being rocky) arranged such that no single unit > 314 m^2.
Limerock boulder	4%	≥75% of area composed of limestone fractured into blocks> 500 mm in diameter across longest axis.
Island	1%	Wholly contained within the river channel and surrounded by water during typical winter or spring discharge.

Analysis of Alabama Shad telemetry data and substrate use using GIS was based upon recently developed techniques used in the upper Flint River for three species of black bass Micropterus (Goclowski et al. 2013), where fish locations were overlain with georeferenced substrate data. This analysis is a distance-based approach that is robust to positional error inherent in data collected with GPS equipment (Conner et al. 2003; Kaeser and Litts 2010). Only those waypoints that fell within the boundaries of the lower Flint River substrate map (Bainbridge, Georgia, to Albany Dam; Figure 1) from 2010 to 2014 were used, resulting in 110 fish location waypoints from 70 different fish.

An index of substrate complexity in the vicinity of each fish location was generated by placing a 15-m circular buffer around each waypoint, and then extracting the substrate information contained within the buffers using the geoprocessing tools within ArcGIS (Goclowski et al. 2013). A smaller radius would begin to approach the estimated position error of waypoints (typically 1–3 m), while a larger radius would provide unnecessary information on substrate use given our ability to obtain precise locations for radio-tagged fish. Total perimeter (m) of the substrate types (shapes) contained within the buffer was then calculated. Those buffers that contained more substrate types (higher subsrate complexity) had higher perimeter values, while those buffers that contained fewer substrate types (lower substrate complexity) had lower perimeter values. An index of substrate complexity throughout the entire study area was then generated by placing a regular grid of 15-m circular buffers spaced 45 m apart over the substrate map (Goclowski et al. 2013). Perimeters for substrate complexity within study area buffers were calculated as for individual fish location buffers.

Next, Euclidian distance (m) from fish locations to the nearest polygon of each substrate type was calculated to describe the average distance of Alabama Shad to different substrate types (Goclowski et al. 2013). Distances were derived using the model builder function and NEAR tool in ArcGIS. The substrate polygon containing the fish location was assigned a distance value of 0. A regular grid of points spaced 20 m apart was then placed over the substrate map, and distances from each point to the nearest polygon of each substrate type was calculated in a similar manner as for fish locations.

Chi-square tests are commonly used for quantifying habitat associations of telemetered fish when the amount of available habitat is known or can be readily estimated. Chi-square log-likelihood statistics, while favored over the simpler chi-square test statistic, are dependent upon the number of observations being large enough to use the fish as the sampling unit for analysis (Rogers and White 2007). However, Rogers and White (2007) advocated pooling observations such that they may be treated as the sampling unit when few observations come from many fish, as was the case in this study (109 observations from 70 individual fish). In instances where overall sample size comprises fewer than 1,000 observations, exact goodness-of-fit tests are recommended (McDonald 2014) to avoid artificial deflation of P-values. Exact goodness-of-fit tests examine the likelihood of the observed distribution against all possible combinations of distributions (~200,000 with this data set) while being robust to small data sets.

An exact multinomial goodness-of-fit test was performed using the EMT package (McDonald 2014; Menzel 2015) in the statistical software R (R Core Team 2014). Exact binomial post hoc testing was used to determine which substrates were being used differently in the event of a significant overall EMT test. A Bonferroni correction for multiple comparisons was computed for the P-values resulting from post hoc testing.

An additional multinomial goodness-of-fit test using the XNomial package was performed to obtain a P-value for both the likelihood ratio and the multinomial probability (Engels 2013). The likelihood ratio (LLR) is the probability of the observed result over its probability given the alternative and is given by the following equation:

where m_i is the number of objects in category i and p_i is the hypothesized probability of that category. The multinomial probability is the probability of the observed outcome under the null hypothesis and is given by the following equation:

where m is the number of objects in category i and P is the hypothesized probability of that category. Finally, Manley’s selection ratio was calculated in the R package adehabitatHS (Calenge 2015) to see whether Alabama Shad were selecting for or against certain substrates (Manly et al. 1993; Rogers and White 2007). Manley’s selection ratios (w_i) are given by the following equation:

where u_i is the amount of habitat type i used by all fish and u₊₊ is the total number of habitat units used by all fish. Test values > 1.0 indicate selection for a particular habitat, while test values < 1.0 indicate avoidance of a particular habitat.

Temporal movements by Alabama Shad were investigated by measuring displacement of fish relocations from their stocking locations. Displacement was calculated using ArcGIS, pooled across years, and grouped into 20-km bins from −40 km (i.e., downstream from stocking location) to 160 km. These data were then placed into five different temporal categories (March 15–31, April 1–15, April 16–30, May 1–15, and May 16–31) and plotted as a percentage of all movement per 20-km movement bin for a given temporal period. Similarly, Alabama Shad movements were plotted as discussed above to investigate variation in upstream movement by year. Alabama Shad movement in relation to river discharge variation was investigated by obtaining daily flow data for the period March 15–May 31 from the U.S. Geological Survey stream gauge 02353000 near Newton, Georgia, (Figure 1) for 2010–2014. Alabama Shad locations that represented an upstream movement ≥ 20 km from the initial release location or previous telemetry location were then plotted, by day and year, against the percentage of mean discharge for the period from March 15 to May 31 of each year during 2011–2014. Data from 2010 were omitted due to the low number of relocations ≥ 20 km that year resulting from detection problems associated with the use of ultrasonic tags. Previous efforts using acoustic tagging equipment were abandoned in favor of radio tags beginning in 2011 due to the extremely low detection of tagged fish, even by stationary, data-logging, ultrasonic hydrophones.

RESULTS

Of the 153 total locations considered to have live fish, 110 live-fish locations were used for analyses after removing those locations that fell outside the extent of the substrate map; these were mostly fish that remained in the Flint River arm of Lake Seminole. Of these 110 locations, 41 fish were located once, 21 fish were located twice, five fish were located three times, and three fish were located four times. Telemetry revealed several distinct river reaches from Bainbridge upstream to Albany, Georgia, where Alabama Shad were located across multiple years during the spring migration, suggesting possible spawning locations within the lower Flint River.

Average substrate complexity index values within a 15-m circle of fish locations were comparable (169.0 m) to the study area point grid (162.4 m; Table 3). Complexity index values for fish locations ranged from 94.1 to 395.7 m. On average, Alabama Shad were farther away from all substrate classes compared with the point grid, with the exception of limerock boulder; Alabama Shad were more than twice as close to this substrate compared with the point grid. Sandy was the closest substrate to Alabama Shad locations; whereas, island and rocky fine substrates were the farthest from Alabama Shad locations. Alabama Shad locations and the point grid were similar distances from sandy, mixed rocky, rocky boulder, and limerock fine substrates. A plot of observed versus expected values of Alabama Shad substrate use showed that Alabama Shad used sandy substrates in accordance with its availability, limerock fine and all rocky substrates less than their availability, and limerock boulder greater than its availabilty (Figure 2). The overall multinomial test in the EMT package was highly significant (P < 0.0001). Exact binomial post hoc testing gave significant results for all rocky (P = 0.0052) and limerock boulder (P < 0.0001) substrates. Exact binomial post hoc testing for sandy and limerock fine substrates were not significant (P > 0.05). The likelihood ratio in the multinomial package XNomial was highly significant (P < 0.0001), while the multinomial probability was also highly significant (P < 0.0001).

TABLE 3. Average distance (m) to nearest polygon of each substrate type from fish locations and regular grid of points placed over study area. Average perimeter of substrates contained within 15-m buffers around fish locations and regular grid of points placed over study area are: 15-m-diameter circle, 94.13 m; fish locations, 169.0 m; study area grid points, 162.0 m.

Substrate type	Distance from fish (m)	Distance from grid (m)
Sandy	29.1	24.5
Limerock fine	255.4	171.1
Rocky boulder	339.0	208.0
Rocky fine	5,810.0	1,204.3
Mixed rocky	161.9	101.6
Limerock boulder	150.2	367.8
Island	17,258.5	6,641.1

FIGURE 2. Observed and expected frequencies of substrate use by Alabama Shad in the lower Flint River, Georgia, 2010–2014.

Manley’s selection ratios indicated selection against limerock fine substrate (w_i = 0.87, 95% CI = 0.33–1.42), selection against all rocky substrate (w_i = 0.52, 95% CI = 0.21–0.85), and selection for limerock boulder substrate (w_i = 4.86, 95% CI = 2.29–7.44). No selection for or against sandy substrate was evident (w_i = 1.0, 95% CI = 0.96–1.11). X–Y coordinate centroids were created for all limerock boulder substrate polygons within the study area (N = 195), and emphasis was given to those limerock boulder polygons that either contained a fish location or were within 500 m of a fish location.

Peak upstream movement of Alabama Shad during 2010–2014 generally occurred in April (Figure 3). Lesser upstream movement was observed during May, while no upstream movement was observed during March. Timing of large (≥20 km) upstream movement typically occurred within 7–10 d following periods of river discharge ≥ 200% of mean discharge after April 1 (Figure 4). The lowest number of relocations and least amount of observed upstream movement occurred during years with relatively high sustained river discharges (2010 and 2014; Figures 4, 5). The one exception to this was 2013, when river discharge never exceeded 150% of mean discharge after April 1 and was generally more stable than in other years. In this year, upstream movement ≥ 20 km was more spread out and occurred over a 4-week period.

FIGURE 3. Temporal displacement from stocking sites for Alabama Shad in the lower Flint River, Georgia, 2010–2014.

FIGURE 4. Alabama Shad movements ≥ 20 km in relation to percentage of mean daily discharge for the period from March 15 to May 31 during 2011–2014. Hollow circles on the x-axis represent stocking date(s). Data from 2010 are omitted due to low number of relocations. Dates on the x-axis are given for reference only and do not necessarily represent days that telemetry efforts occurred.

FIGURE 5. Total relocations (percent of number stocked) and displacement (km), by year, from stocking sites for Alabama Shad in the lower Flint River, Georgia, 2010–2014.

DISCUSSION

Results from this study indicated that Alabama Shad in the lower Flint River selected against rocky substrates and selected for limerock boulder substrates, and sandy and limerock fine substrates were generally used in accordance to their availability. The closely related American Shad is known to select for larger, coarser substrates such as gravel, cobble, and boulder–bedrock during spawning activities, and use finer substrates such as silt and sand less (Hightower et al. 2012). Other members of the genus Alosa also select larger substrates (Caswell and Aprahamian 2001; Harris and Hightower 2011), although Alewives A. pseudoharengus are known to use a wide variety of habitats for spawning (Loesch and Lund 1977).

Spawning habits, including diel timing, of Alabama Shad are unknown. Many alosine species, such as American Shad and Alewife, are known to spawn at night but others have been observed to spawn during the day (Irwin and Bettoli 1995; Harris and Hightower 2010). In the absence of data, we assumed that these daytime locations of Alabama Shad in the lower Flint River were associated with, and may even actually have been, spawning areas. A study is currently underway to further define temporal and diel spawning timing of this species in the lower Flint River.

Upstream movement occurred during the general spawning period of March–May reported for Alabama Shad (Laurence and Yerger 1967; Mills 1972), although the period of greatest upstream movement in the lower Flint River occurred from April to mid-May. While the low number of relocations could have affected temporal movement data (i.e., upstream movement increased with elapsed time from stocking date), the extent of upstream movement observed decreased as the general spawning period progressed from a movement peak during April into late May. River discharge appeared to cue large upstream Alabama Shad movements, as almost all observed movements ≥ 20 km followed periods of river discharge ≥ 200% of mean discharge after April 1, usually within 7–10 d following peak discharge. The overall extent of upstream movement also appeared to be influenced by river discharge, and fewer fish moved shorter distances upstream during high-discharge years than in lower-discharge years.

Dutterer et al. (2011) found that the migration extent of American Shad was affected by river discharge in the St. Johns River, Florida, where shorter upstream migrations occurred during low-discharge years than in high-discharge years. Alabama Shad in the lower Flint River seemed to exhibit the opposite behavior, and fewer movements ≥ 20 km occurred during high-discharge years (2010 and 2014). These movements also occurred in mid to late May, which was generally later than during low-discharge years. This may reflect a reduced ability of the shad to orient in the swirling, turbulent currents present in the lower Flint River during higher flows. American Shad can experience disorientation and reduced swimming performance in the presence of high or turbulent flows (Leggett 1976; Katz 1986). Temperature may also play a role, whereby cold spring rains delay the timing and extent of spawning activity, including upstream movement (Ellis and Vokoun 2009). Unfortunately, temperature data were unavailable for most years of this study; thus, we were unable to investigate how temperature influenced spawning activity. Given the importance of water temperature in regulating alosine activity during the spawning migration, future studies should give consideration to the use of temperature-logging equipment installed within the lower Flint River.

Large migrations of Alabama Shad in the lower Flint River were less associated with discharge in 2013, when the river hydrograph was more stable and few large spates of water occurred. Other factors such as water temperature may also play a role in mediating upstream spawning migrations, especially in years lacking high-discharge events (Richkus 1974; Ellis and Vokoun 2009). Water temperature appeared to be related to the timing of the overall spawning migration of Alewives in a Rhode Island river, but fish numbers arriving at a fishway from a downstream estuary appeared to be governed more by river discharge (Richkus 1974). This also appears to be the case with Alabama Shad, as Young et al. (2012) found that passage through the lock chamber at JWLD in 3 years generally occurred following periods of high discharge; whereas, water temperatures were variable and less associated with fish passage. Thus, water temperature and discharge likely interact in a complex manner to drive timing of Alabama Shad spawning migrations.

Hightower et al. (2012) identified three primary strategies for confirming American Shad spawning areas: spawning splashes, ichthyoplankton sampling, and telemetry. Spawning splashes are probably the best direct indicator of where spawning activity is occurring, but there have been no nighttime surveys to confirm this behavior in Alabama Shad. Ichthyoplankton sampling for eggs and larvae is an efficient and simple method to sample many locations within a river reach; staging of eggs combined with river discharge data can allow for close approximation of spawning areas (Harris and Hightower 2011). We considered telemetry to be the most reliable method to delineate habitats within the lower Flint River for future studies directed at identifying specific Alabama Shad spawning sites, but we experienced unexpected difficulties in locating fish. We observed low relocation rates throughout the study, even in low discharge years, which hindered our ability to assess habitat use and movement of this species. Many factors have been found to affect detection probability of radio-tagged fish, including tag-to-receiver distance, environmental variables, and fish behavior (Melnychuk 2012). Turbulence and depth are both known to affect detectability of radio tags (Freund and Hartman 2002; Melnychuk 2012); thus, high, fast water that is typical during spring in the lower Flint River may have reduced our ability to detect telemetered Alabama Shad. The numerous “blue-hole” springs present in the limestone bedrock geology found in the lower Flint River (Opsahl et al. 2007) may have complicated telemetry efforts as well by providing deep-water refuge to fish, which could have resulted in greatly attenuated transmitter signals.

“Fallback” is a term used to describe anadromous fishes that abandon their spawning migration as a result of stress from handling, transmitter implantation, or environmental conditions (Moser and Ross 1993). It is typically manifested as tagged fish that are not subsequently detected or are detected downstream from the tagging location and do not return upstream. Reported fallback rates for Alabama Shad tagged downstream from JWLD ranged from negligible to 35% (Sammons and Young 2012) and certainly affected our ability to locate telemetered fish. Additionally, this species can make large-scale movements of over 90 km in a single day (Sammons and Kern 2015). Thus, the possibility of missing fish despite a systematic telemetry searching protocol appears to be high. Typical for most telemetry studies, movement data presented herein are likely minimums. A fish as mobile as this species is capable of traversing the entire study reach in a day, and as with all mobile telemetry studies, there is no reasonable way to account for movement outside of tracking periods. Fixed acoustic telemetry arrays were originally used in an attempt to address this issue but proved to be unreliable for this species in this system (Young et al. 2012).

Despite the difficulty in relocating Alabama Shad and the complications arising from a mobile telemetry design, efforts on the lower Flint River were successful in that the multiyear telemetry design (Winter 2012) allowed us to identify patterns of substrate use and movement during the spawning migration that will be useful. This study is the first attempt at quantifying Alabama Shad behavior in the lower Flint River during the spawning migration. While substrate selection by actively spawning Alabama Shad remains unknown, the apparent selection of limerock boulder substrate provides a starting point for future investigations, whereas there were previously no data upon which to base such studies.

The latitude–longitude coordinates generated for limerock boulder substrate locations in the lower Flint River are the focus of current conservation efforts directed at identifying Alabama Shad spawning locations in the lower Flint River. By conducting ichthyoplankton sampling in the vicinity of locations given for limerock boulder, large reaches of unproductive water may possibly be eliminated (Harris and Hightower 2011). Secondary sampling for adult Alabama Shad by means of boat electrofishing in reaches with positive ichthyoplankton collections will reduce complications associated with downstream drift. Sampling that begins within a few days after the arrival of adult female Alabama Shad at JWLD, coupled with a sufficient return of adult Alabama Shad to JWLD, will focus efforts aimed at identifying more precise spawning locations for Alabama Shad in the lower Flint River.

ACKNOWLEDGMENTS

This project was funded by Georgia Department of Natural Resources, and A. Kern was supported by a grant from Alabama Department of Conservation and Natural Resources. Additional logistic support was provided by Auburn University, U.S. Fish and Wildlife Service, The Nature Conservancy, U.S. Army Corps of Engineers, and Florida Fish and Wildlife Commission. David Belkoski, Jeff Buckingham, Alex Christopher, Chase Katechis, Craig Robbins, Josh Tannehill, Rob Weller, and Shawn Young assisted in the field work for this project. Finally, the authors thank the two anonymous reviews whose thoughtful comments and suggestions proved valuable to the development of this work.