Coastal wetlands provide critical habitat for aquatic organisms and important ecosystem services for the terrestrial and aquatic landscapes they bridge, but increasingly common invasive macrophytes disrupt plant communities, food webs, habitat structure and littoral-pelagic linkages. In Laurentian Great Lakes coastal wetlands, invasive cattails (Typha ×glauca and T. angustifolia, hereafter Typha) homogenize ecosystem structure and reduce nearshore dissolved oxygen, and plant, fish, and macroinvertebrate diversity. We hypothesize that management treatments that reduce Typha and its abundant litter promote structural heterogeneity and mitigate physiochemical and biodiversity impacts.

Study sites

We selected four Great Lakes coastal wetland complexes in northern Michigan (USA) for our study, one in the northern Lower Peninsula and three in the eastern Upper Peninsula. Two of the wetlands (Cheboygan Marsh and St. Ignace Marsh) are in the Straits of Mackinac, near the confluence of Lakes Michigan and Huron, and two (Munuscong Marsh and Sand Island Marsh) are in the St. Marys River, the connecting channel between Lakes Superior and Huron (Figure 1). All wetlands shared similar connectivity to the open lake, but hydrogeomorphic type varied among the wetlands; Cheboygan Marsh and St. Ignace Marsh are lacustrine open embayment types, Munuscong Marsh is a connecting channel river delta, and Sand Island is a connecting channel protected embayment (Albert et al., 2005).

Study design

We conducted our study within invasive Typha-dominated portions of the four wetland complexes. Due to the abundance of advanced generation hybrids in upper Midwest Typha populations and the high frequency of disagreement between morphological traits and molecular identification (Geddes et al., 2021), we did not attempt to differentiate between hybrid cattail (Typha ×glauca) and narrow-leaved cattail (Typha angustifolia), but clumped invasive cattails into one taxon for this study. Invasive cattails had been widespread in all study wetlands for over a decade (Lishawa et al. 2010, Lawrence et al. 2016) and for more than 50 years in Cheboygan Marsh (Lishawa et al. 2013) and were the dominant emergent plant taxon (> 50% relative dominance) of > 25 ha in each focal wetland.

To test biodiversity responses to invasive Typha harvest and aquatic connectivity channel treatments, we used a block design with four, 32 × 64 m plots nested within each block. Each treatment block consisted of four randomly assigned treatments: harvest, harvest-channel (a channel nested within a harvested plot), control-channel (a channel nested within an unharvested plot), and control (unmanipulated Typha­-dominated plot). We created one block in each wetland except St. Ignace, which had two blocks (Figure 1c; Figure 2). 

In July and August 2016, we randomly assigned treatments, delineated all blocks and plots, and implemented treatments. Plots were oriented with their length (64 m) perpendicular to the shoreline so that they crossed the primary water level gradient at each site. Each plot was 32 m wide. We harvested and removed all biomass (i.e., standing litter and living vegetation) at 20 cm above the water level from the harvest and harvest-channel plots using a low ground-pressure Softrak wetland plant harvester (Loglogic, Cullompton, United Kingdom). Within the harvest-channel and control-channel plots, we manually cut and removed all biomass at the sediment surface using aquatic weed whackers (RedMax, Lawrenceville, GA, USA) from 3-m wide by 50-m long channels, which bisected the center of each plot along the water depth gradient (Figure 2). The aquatic weed whackers have flat reciprocating cutting blades, which slide along the surface of the sediment, allowing us to cut stems approximately 2 cm above the surface with minimal substrate disturbance.

Field data collection

We collected water quality and biodiversity data (plant, fish, macroinvertebrate, and larval amphibian) for two years following treatment implementation, in summer (June – August) 2017 and 2018 from subplots within each plot; treatments were implemented at the plot level. Subplot distribution and density differed among the taxa sampled (see below).

Plant Sampling

Within each plot we established eight, 1 × 1-m subplots for plant community data collection. Four subplots were located in plot corners equidistant from the plot margins (8 m) and four subplots were located 12 m apart along the plot’s meridian to align with the channels (Figure 1c). In mid-July through mid-August of each study year (2017 and 2018), we estimated percent cover for every species and Typha stem density and height from each subplot. Invasive cattail height data were used to calculate biomass, using a height-to-biomass allometric equation (Lishawa et al., 2015).

Fish and Larval Amphibian Sampling

We established either four or six subplots for fish sampling within each plot: four plots within the harvest and control treatment plots, which corresponded with the corner vegetation subplots (i.e., 8 m from each plot margin), and two additional subplots located 15 and 30 m into the channels within the harvest-channel and control-channel plots (Figure 1c). We sampled fishes three times throughout the growing season (June, July, and August) in each year (2017 and 2018) from each subplot using three Gee minnow traps (mesh size 0.64 cm) set over one to three nights. We used minnow traps in lieu of fyke nets or electrofishing because minnow traps are effective at sampling small-bodied fishes in dense vegetation (Henning et al., 2014; Jacobus and Webb, 2005; Webb, 2008), and vegetation density in most subplots precluded efficient sampling by these other methods. Minnow traps were set 5 m apart, baited with dry dog food, and checked every 24 h. Fishes were identified as species (or lowest taxonomic group possible) in the field and enumerated. Any adults of difficult-to-identify species (e.g., shiner species) were euthanized with 250 mg/L tricaine methanesulfonate (MS-222), fixed in 10% formalin, and transferred to 70% ethanol after 24 h for later identification in the laboratory. Fish abundance was quantified for each subplot as catch-per-unit-effort (CPUE) as the number of fishes captured in the set of three minnow traps in each subplot in 24 h. We did not include fish data from the Munuscong Marsh treatment block in our analysis because fish abundance across all plots and treatments was extremely low, indicating that minnow trap sampling at this site was ineffective. Larval amphibians were collected in minnow traps during fish sampling and were enumerated and released in the field (Skelly and Richardson, 2010). For each subplot the CPUE for larval amphibians was defined as the number of larvae captured in the set of three minnow traps in a 24 h period.

Macroinvertebrate sampling

We collected macroinvertebrate samples in late July through early-August in 2017, to correspond with the period when emergent plants typically reach maximum annual biomass and late instars of most aquatic insects are present and easy to identify (Kashian and Burton, 2000). We collected at least three replicate samples per subplot, each replicate consisting of 1-m dip net (0.5 mm mesh) sweeps at the water surface, mid-depth, and at the substrate surface. In the field, we emptied nets into white pans and picked macroinvertebrates (Uzarski et al., 2004). In the laboratory, organisms were sorted into the finest taxonomic group possible, and richness and abundance metrics were estimated. Crayfish were sampled using the same method as fishes. In the field, we observed dramatically different patterns of invertebrate abundance by treatment within the Munuscong block, where large numbers of Asellidae were present in the control and harvest treatments. Therefore we statistically evaluated this metric (invertebrate abundance [number / m2]) as a full model (all blocks), with Munuscong block excluded, and with the Munuscong block alone. 

Water quality sampling

We measured standard water quality parameters to relate taxonomic data with abiotic conditions. These parameters included water depth, temperature, dissolved oxygen, turbidity, pH, specific conductivity (determined in situ), total N, total P, and chlorophyll a (determined in the lab). In situ water quality parameters were measured at each subplot and sampling date using a YSI Professional Plus Multiparameter Instrument (Xylem Inc., Yellow Springs OH, USA) and a Hach 2100Q Portable Turbidimeter (Hach, Loveland CO, USA). Nutrients (total N and P) were field-collected during fish sampling and analyzed using standard methods on a SEAL AutoAnalyzer 3 spectrophotometer (SEAL Analytical, Mequon, WI, USA). Chloride, fluoride, and sulfur were also collected during fish sampling, lab-filtered through a MF-Millipore 0.45 um MCE membrane filter, and analyzed on a Dionex Integrion High-Performance Ion Chromatography system (Thermo Fisher Scientific Inc., Waltham MA, USA). Chlorophyll a samples were collected by filtering a known volume of water through a 45 µm filter in the field. Filters were stored on ice and chlorophyll a was analyzed using a Turner Designs TD-700 Fluorometer (Turner Designs, San Jose, CA, USA). All analyses were completed at the University of Michigan Biological Station, in Pellston Michigan, USA.

Data is saved in Excel file (.xlsx) format and can be opened with Microsoft Excel or Google Sheets.