4 | DISCUSSION
Our study demonstrates differences in our ability to detect invasive and native larval fish presence accounting for various environmental conditions. Detection probabilities for bigheaded carp were lower than native taxa and required more samples to detect their presence. Detection probability estimates increased with water volume filtered for all species, but detection also varied among habitats for carp and freshwater drum, indicating spatial differences in sampling efficiency. Similar to detection, bigheaded carp occupancy was generally lower than native species, suggesting they have a narrower range of environmental conditions suitable for reproduction. Temporal and thermal environmental covariates improved occupancy probabilities among all species, while hydrological covariates were only beneficial to bigheaded carp occupancy.
Occupancy and detection models indicate some habitat separation among taxa. On average, bigheaded carp and drum detection probabilities were higher in thalweg and channel border habitats, whereas detection probability of shad and freshwater drum was similar among habitats. Pelagophil fishes, such as bigheaded carp and drum, reproduce in open water conditions (Balon, 1975; Welcomme et al., 2006) associated with the higher velocity thalweg and channel border river channel habitats. Larval fishes in these habitats are likely younger and passively drifting with the current (Holland, 1986), potentially making them easier to capture. Alternatively, larval bigheaded carp can migrate out of the main channel to areas with lower velocity at 5 to 6 days of age and larger, more developed larvae that are better swimmers may be more likely to avoid the ichthyoplankton net (Chapman and George, 2011; Roth et al., 2023). In contrast to detection, occupancy of carp and freshwater drum was similar among habitats, suggesting that all habitats are used by larval carp and drum. Thus, if the objective is to simply document bigheaded carp reproductive success, focusing sampling efforts in the thalweg would be more efficient. Alternatively, shad and percid detection was similar among habitats but occupancy was higher in backwater habitats than thalweg and channel borders, consistent with conventional reproductive guild classification for both shad (litho-pelagophil) and percids (lithophil; Simon, 1998). While not directly considered in this study, larval habitat selection can change during later developmental stages due to gas bladder inflation and increased horizontal mobility (Chapman and George, 2011). While we limited our occupancy and detection models to sub-juvenile stages, there may exist occupancy and detection habitat variation among sizes, ages, and developmental stages due to changes in mobility.
Native and invasive larval occupancy both varied temporally and were influenced by environmental conditions. Bigheaded carp reproduce in rivers during May through June when water temperatures exceed 17°C, with peak reproduction occurring at 22 to 26°C (Schrank et al., 2001), though protracted spawning documented into the fall can occur in some systems (Coulter et al., 2013; Papoulias et al., 2006). Freshwater drum typically spawn between May and June in the Upper Mississippi River (Butler, 1965) when water temperatures are between 18 to 25°C (Swedberg & Walburg, 1970). Shad spawn over a broad temporal window with reproduction occurring from late April into early August with water temperatures between 10 to 21°C (Becker, 1983), whereas percids,such as walleye, spawn the earliest from late April to early May with water temperatures ranging from 5 to 10°C (Becker, 1983; Bozek et al., 2011). Similarly, we found maximum occupancy estimates corresponded with Julian dates for bigheaded carp (June 19th), freshwater drum (July 3rd), and shad and percids (May 24th) that were typically at the upper limits of their documented spawning periods. This trend was mirrored in maximum occupancy occurring from freshwater drum when water temperatures was 23.9°C, whereas the water temperature for carp (20.7°C) was well within their recorded temperature limits and even lower than the cited temperature range for peak reproduction (22-26°C; Schrank et al., 2001). Maximum bigheaded carp occupancy in the Upper Mississippi River occurring below the peak reproduction threshold is likely due to stability of water temperatures in this region throughout our sampling period (mean water temperature of 22°C from April to August). In addition to mean temperature, increasing temperature variation led to lower occupancy probabilities for carp and drum, but an increase in occupancy probability for shad/percids. Fluctuations in water temperature can negatively influence bigheaded carp reproduction due to disruption in oogenesis (Majdoubi et al., 2022) and can reduce egg viability of other fishes (Van Der Kraak & Pankhurst, 2011), indicating post-spawning temperature stability could promote larval survival.
Larval occupancy among taxonomic groups was primarily affected by temporal and thermal variation; however, bigheaded carp occupancy also increased with mean river discharge and may have been negatively affected by variability in discharge. Discharge and water temperature are central catalysts of bigheaded carp reproduction (Camacho et al., 2023; Kolar et al., 2007; Lohmeyer & Garvey, 2009; Schrank et al., 2001). Adult bigheaded carp move upstream in spring with increasing discharge and spawn during peak flows when water temperature exceeds 17°C (DeGrandchamp et al., 2007; Kocovsky et al., 2012; Schrank et al., 2001). Effects of variation in discharge on bigheaded carp reproduction are less understood. With increasing variation in discharge, our models suggested a slight negative effect on bigheaded carp larval occupancy, although the slope overlapped zero. Schaick et al., (2023) observed a similar relationship between larval bigheaded carp densities and variable discharge and theorized sustained, high magnitude discharged events were preferred for bigheaded carp reproduction. While discharge is also often cited as a major driver of reproduction for lotic fishes (Dudley & Platania, 2007; King et al., 2016; Humphries et al., 2002), we did not find an effect of discharge on native larval fish occupancy, potentially because peaks in discharge were not synchronized with appropriate spawning temperatures or because native taxa are more adapted to flow regimes (Lytle & Poff, 2004). Alternatively, discharge may not affect the occurrence of reproduction (e.g., occupancy), but could affect the magnitude of reproduction (e.g., larval densities) and still be responsible for large versus small year-classes (Weber et al., 2021).
Understanding when, where, and under what conditions larval fishes are present is strongly dependent on successfully capturing them given they occur at a site. Drum, shad, and percids had higher detection probabilities than bigheaded carp. Adult bigheaded carp are difficult to capture (Bouska et al., 2017; Collins et al., 2015), but our results are the first to document challenges associated with lower capture success of invasive larvae compared to native taxa. Low detection of bigheaded carp larvae can make it difficult to document reproductive events, particularly along invasion fronts, resulting in misinformed population status assessments with implications for management decisions. For instance, efforts are underway in the Upper Mississippi River to install barriers to slow or stop adult upstream movements into areas where reproduction has not yet been documented. Sampling effort strongly influenced our ability to detect larval fishes, but effects varied among taxa. Biologists can improve larval detection probabilities by 1) focusing sampling in the thalweg, 2) increasing the water volume filtered per tow, and 3) increasing the number of tows collected per site visit. While we collected three tows per site visit, cumulative detection curves indicated we would need to collect 14 thalweg/channel border samples to achieve 90% detection probability of bigheaded carp larvae. Alternatively, increasing the water volume filtered by 33% would reduce the sample size to 9 thalweg/channel border tows. These results provide an adaptable and flexible framework to determine how many samples to collect and how much water to filter based on river conditions (e.g., debris load) and acceptable uncertainty in bigheaded carp presence. Further, sampling could also be adjusted to determine successful reproduction through targeted sampling based on habitat, temporal, thermal, and hydrological effects on occupancy to optimize future sampling to assess reproduction of cryptic invaders.