In the wake of a major hurricane: Differential effects on early vs. late successional seagrass species

At least 18 major storms have struck the Gulf of Mexico and Caribbean in the past 50 yr including Hurricane Harvey, a Category 4 storm that passed over extensive seagrass beds in the western Gulf of Mexico and became the second‐most expensive U.S. hurricane. We sought to identify the effects of an extreme hurricane on sediment physicochemical characteristics and seagrass species with contrasting life histories and morphologies. Surprisingly, Harvey's intense wind speeds resulted in decreases in blade length, vegetative cover, and greater overall loss of Thalassia, a persistent climax species relative to Halodule, a prolific pioneer species. Sediment ammonium and grain size changed, but not organic carbon. Our results indicate that effects of wind intensity are not only restricted to the differential impacts on seagrasses, but on the physicochemical characteristics of the sediments. These changes, coupled with the slow colonization abilities of Thalassia, may prolong recovery of disturbed seagrass meadows.

variety of coastal ecosystems including seagrasses, salt marshes, mangroves, coral reefs, and upland communities (Tilmant et al. 1994;Preen et al. 1995). These severe storms can defoliate and uproot vegetated communities, pummel underwater reef structures, and transport massive amounts of sediment that can affect habitat function. Seagrass meadows provide nursery habitat for recreationally and commercially important species, improve water quality, and sequester "blue" carbon (McLeod et al. 2011). Therefore, disturbances can disrupt the ecological functions of these foundation species by reducing carbon sequestration, releasing buried carbon (Macreadie et al. 2015), and altering habitat use by local fauna (Paperno et al. 2006).
The effects of severe weather phenomena on coastal habitats are variable and often differ in response to the individual characteristics of the storm. Tropical cyclones are typically largescale disturbances but the environmental consequences can vary in magnitude and spatial extent (Smith et al. 1994;Paerl et al. 2001;Anton et al. 2009). Wind speed and direction, intensity, and storm proximity are all important determinants of impacts on the surrounding ecosystems. Post-hurricane investigations have documented catastrophic effects on mangroves located near the eyewall and storm track (Milbrandt et al. 2006). These high-energy wind and wave disturbances can also redistribute the substrate, causing substantial removal and deposition of sediments (Fourqurean and Rutten 2004).
In subtropical and tropical regions, ecological succession in seagrass beds typically culminate with Thalassia testudinum, where disturbed areas are first colonized by early successional macrophytes including seaweeds, Halodule wrightii, or Syringodium filiforme (Patriquin 1975). Differences in life history strategies determine species succession and ultimately their response to a disturbance. Colonizing or opportunistic species have lower resistance but the ability to recover quickly because of shorter turnover times, faster sexual maturation, and a high investment in the production of dormant seeds (Kilminster et al. 2015). The cited effects of hurricanes on seagrasses within the Gulf of Mexico and Caribbean illustrate the ephemeral nature of these acute disturbances and imply that seagrass ecosystems are resilient (Tilmant et al. 1994;Byron and Heck 2006;Steward et al. 2006). Hurricane-impact studies documented a range of responses, from no effect to species-specific losses (Fourqurean and Rutten 2004;Byron and Heck 2006;van Tussenbroek et al. 2008;Anton et al. 2009). Fourqurean and Rutten (2004) documented substantial thinning of S. filiforme (19%) and H. wrightii (11%) beds following a Category 2 hurricane. Interestingly, this same study only recorded a 3% reduction in the density of T. testudinum, which was attributed to species differences in plant architecture. To test this, Cruz-Palacios and van Tussenbroek (2005) simulated hurricane-like disturbances and concluded that differential growth forms such as a deeply anchored rootrhizome system, coupled with flexible leaf structure, buffered the effects of hurricanes. Field surveys conducted by van Tussenbroek et al. (2008) following a Category 4 hurricane documented significant declines in S. filiforme but minimal changes in T. testudinum, corroborating the findings of Fourqurean and Rutten (2004) and Cruz-Palacios and van Tussenbroek (2005). But the hearty architecture of long-lived seagrasses like Thalassia may not be advantageous across all environments. Individual storm characteristics, drag forces, and wave heights dictated by the proximity of the bed to physical barriers may influence distinct successional species differentially, particularly those (like Thalassia) that exhibit significant morphological plasticity.
In this study, we examine the susceptibility of seagrasses to a major storm event by asking two research questions: (1) what is the effect of a Category 4 hurricane on two common seagrass species characterized by very different life histories (e.g., turnover rate, age, resource allocation) and morphologies; and (2) how are the sediment characteristics altered in seagrass beds? We addressed these questions at 126 stations on the central Texas coast in the aftermath of Hurricane Harvey, a catastrophic storm that dropped over 60 in. of rainfall, caused approximately 68 deaths, and resulted in an estimated $125 billion in damages in late August 2017 (Blake and Zelinsky 2018). We used pre-and post-hurricane survey data from the Texas Seagrass Monitoring Program (http://www.texasseagrass. org) to document and compare the immediate impacts of this major disturbance on a pioneer, shoalgrass (H. wrightii = Halodule), and climax, turtlegrass (T. testudinum = Thalassia), species. We show that a major hurricane substantially affected established beds of the late successional species, Thalassia. Since seagrass recovery is a function of species-specific growth rates and the size of the disturbance, a slow-growing species such as Thalassia could require a decade or more to recover through vegetative regrowth or from new seedling propagation (Dawes et al. 1997).

Storm and site description
Hurricane Harvey rapidly intensified into a Category 4 hurricane before striking the Coastal Bend on 25 August 2017 ( Fig. 1). Maximum sustained winds were 213 km h −1 with highest peak wind gusts of 241 km h −1 . Extreme winds and disastrous flooding made Harvey the second largest natural disaster in terms of economic losses in U.S. history (Blake and Zelinsky 2018).

Seagrass monitoring
As part of the Texas Seagrass Monitoring Program, a small team of trained observers has annually sampled 525 permanent stations along the Texas coast. We employed a restricted random sampling design where we generated one random, fixed station within a tessellated hexagon (Coastal Bend: 500-m edge; Laguna Madre 750-m edge) (Dunton et al. 2011;Neckles et al. 2012). Hexagons were created using the National Oceanic and Atmospheric Administration's 2004/2007 Benthic Habitat Assessment (https://coast.noaa.gov/digitalcoast/data/) with a minimum threshold of 50% cover. In 2015 and 2017, we visited these same stations (AE 10 m) along the same north-tosouth gradient (Coastal Bend to Laguna Madre; Fig. 1) during peak seagrass growth (Coastal Bend: July/August; Laguna Madre: August/September/October/November). Prior to the landfall of Hurricane Harvey in 2017, we sampled 126 stations from 26 July to 22 August within the Coastal Bend.
Following the storm, the field team managed to resurvey the same 126 stations in the Coastal Bend from 18 September to 08 October. The remaining 399 Laguna Madre stations were surveyed from 18 October to 17 November 2017. Since we had not surveyed the Laguna Madre prior to Harvey's landfall in 2017, we used pre-Harvey samples collected in 2015 for our pre-and post-storm comparison. For the Coastal Bend, our pre-and post-Harvey data were based entirely on samples collected prior to and immediately following landfall (126 stations). To determine seagrass composition at each of the 525 stations, we visually estimated percent cover by species using a 0.25 m 2 quadrat subdivided into 100 cells placed at four ordinal points around a shallow-draft vessel. Cover of seagrass species present in each quadrat was defined as the proportion of the frame obscured by vegetation. Immediately following each of the four cover observations, we collected five random shoots for all species present within the quadrat for measurements of blade length. Blade lengths were determined as the photosynthetic portion of the longest blade from each shoot. For subsequent analyses, we focused on the two dominant seagrass species, Halodule and Thalassia.

Sediment physicochemistry
In August 2011 and May 2012, sediment cores were collected from three fixed 50-m transects at a reference site within the Coastal Bend (27 54 0 N, 97 05 0 W) ranging in depth from 0.3 to 1.4 m. One core was collected at 10 permanent points along each transect and processed for sediment ammonium. Additionally, at each transect, two cores were collected at the 0-and 50-m point for total organic carbon (TOC) and grain size analysis. Following Hurricane Harvey, we obtained two sediment cores from 16 stations (n = 32), 0.5-0.75 km from the 2011/2012 reference site. From each of the 16 stations, we processed one of the cores for pore-water ammonium and the second core was subsampled for grain size and TOC. All sediment samples were collected within colonized or previously colonized sediments and extracted using a 10 cm (60 mL) syringe corer, placed on ice, and frozen prior to analyses. For pore-water ammonium, we thawed then centrifuged the sediment core, and processed the supernatant using the colorimetric techniques in Parsons et al. (1984). We determined sediment grain size fractions by weight using sieving and settling velocity to classify sediment particles (Folk 1961). Shell/sand fractions were classified as particles > 63 μm. For the evaluation of organic matter, we used loss-on-ignition; samples were dried at 60 C to a constant weight, weighed, combusted at 550 C for 4 h, and reweighed.

Spatial and statistical analyses
To illustrate continuous layers of changes in seagrass cover and wind speed, we used inverse-distance weighted (IDW) interpolation to assign values to areas between sampling points, weighted by distance. We used 12 sampling stations identified from a variable search radius (100 m 2 ) to generate a predicted value for each unknown point. We interpolated changes in seagrass cover using data from 525 stations, which were spatially restricted to the geographic limits of the submerged vegetation map (2004/2007. Changes in seagrass cover (ΔC) caused by the hurricane were quantified as follows: ΔC = [(C post − C pre )/ C pre ] × 100%, where C pre values were measured in 2015 (Laguna Madre: August-November) and 2017 (Coastal Bend: July-August). Since the Coastal Bend was exposed to major hurricane force winds, we focused on changes in cover at 126 stations within this region. For the interpolation of maximum sustained winds, we used 10-m maximum sustained wind gusts obtained from 70 weather locations reported by the National Weather Service (NWS; M. Buchanan) in the NOAA National Hurricane Center Hurricane Harvey tropical cyclone report (Blake and Zelinsky 2018). We used the tool "Extract Values to Points" to extract cell values from the interpolated wind raster to obtain an estimate of wind speed for each of the 525 seagrass sampling stations. All spatial analyses were performed in ArcMap 10.3 (Environmental Systems Research Institute).
We used maximum sustained winds derived from the IDW interpolations as a metric of wind intensity. We calculated the difference (C post − C pre ) for cover and blade length for each of the 525 stations. Pre-Harvey data used in these calculations were obtained in 2017 and 2015 for the Coastal Bend and Laguna Madre, respectively. All seagrass post-Harvey data were from 2017. We then modeled the effect of wind intensity on differences in blade length and cover using linear regression for all 525 stations. We assessed whether species responded differently (i.e., change in slope) using analysis of covariance (ANCOVA) to test if models with an interaction between species and wind speed differed from those without an interaction term. For the 126 stations in the Coastal Bend, we tested for differences in cover and blade length of monospecific or mixed (species co-occur) beds using either Student's t-test or Welch's t-test (for unequal sample size). We tested differences in sediment grain size between pre-Harvey (2011/2012) and post-Harvey (2017) measurements using Student's t-test for each size fraction after testing for normality. We examined for differences in sediment organic carbon and sediment ammonium via Wilcoxon rank sum test (p < 0.05) after testing for homogeneity of variance using Levene's test (cutoff value of p = 0.05). The data presented in this article are archived in the repository https://data.nodc.noaa.gov/cgi-bin/iso?id=gov.noaa. nodc:0187104 and https://doi.org/10.25921/9qkw-3w27.

Results
Hurricane Harvey battered coastal habitats in Texas with copious rain and winds before weakening as it crossed toward the interior of the state; Category 3 and 4 winds associated with the storm's eyewall were most prevalent in the Coastal Bend (Fig. 1, inset). There were significant albeit weak reductions in cover of Thalassia and Halodule at stations with higher winds but with greater explanatory power for Thalassia ( Fig. 2A;
In the Coastal Bend, monospecific and mixed beds experienced different maximum winds (Kruskall-Wallis rank sum test, Χ 2 (2) = 42.39, df = 2.0, followed by Bonferroni corrected pairwise Wilcoxon rank sum test, p < 0.05). Winds were significantly lower (x AE SD: 169.0 AE 34.1 km h −1 ) at stations with Halodule compared to Thalassia (x AE SD: 211.0 AE 9.1 km h −1 ) and mixed (x AE SD: 209.0 AE 9.7 km h −1 ) assemblages, but winds were not different between Thalassia and mixed beds (Bonferronicorrected α = 0.05). The chance occurrence of Thalassia beds in areas with the highest wind velocities may have contributed to the differential response by species. Regardless, the change in cover (ΔC) shows a greater loss of Thalassia (x AE SD; C pre : 29.2% AE 38.1%; C post : 19.6% AE 28.9%) relative to Halodule (x AE SD; C pre : 35.6% AE 39.0%; C post : 37.0% AE 37.7%; Fig. 4A). The negative effects of a hurricane on a climax species compared to other systems are notable and the largest reported for the Gulf of Mexico and Caribbean (Fig. 4). Belowground tissues of Thalassia were stripped in areas that experienced erosion, leaving behind exposed roots and rhizomes, or broken and decayed material (Fig. 4B). In areas with belowground biomass intact, blades displayed clean cuts (Fig. 4C) and were cropped (~90-120 mm). We observed~1.4% loss in seagrass cover in south Texas, but the Coastal Bend, which received the brunt of Hurricane Harvey, lost~20% of its seagrass area (based on complete losses of Thalassia and Halodule noted above).

Discussion
Our results stand in contrast to a large proportion of posthurricane assessments in the Gulf of Mexico and Caribbean that documented no change to seagrass communities (Tilmant et al. 1994;Steward et al. 2006;Anton et al. 2009) or only substantial declines in the abundance of pioneer species (Fourqurean and Rutten 2004;van Tussenbroek et al. 2008). Hurricane Harvey damaged seagrass habitats along the Coastal Bend as reflected by decreases in cover and blade length directly related to wind intensity. The most notable difference between our observations and previous studies is that a Category 4 hurricane measurably affected the abundance of a late successional species despite their higher reported resistance to disturbances as noted in earlier studies (Fourqurean and Rutten 2004;Cruz-Palacios and van Tussenbroek 2005;van Tussenbroek et al. 2008).
Shifts in sediment physicochemical properties, coupled with our field observations noting erosion and burial, point to the transport of sediments during the hurricane. Enhanced microbial activity following Harvey may be responsible for the difference between pre-and post-storm ammonium concentrations. Post-storm ammonium was higher and displayed greater variability, which could indicate increased microbial remineralization of organic carbon (e.g., dead roots/rhizomes) in disturbed seagrass meadows (Macreadie et al. 2015). Although we did not detect a significant difference in sediment coarseness, there was a shift in the fraction of particles > 63 μm as sand increased from 42% to 58% and shell decreased from 45% to 24%. Since our pre-Harvey sediment samples were acquired in 2011/2012, the increase in silt/clay and organic carbon likely reflect "muddification." Seagrasses trap sediments and accrete organic matter, and over time, sediments become rich in organic content and can shift from coarse to fine grains (Katwijk et al. 2010).
Typically, sediments along the northern part of the Gulf of Mexico are finer than their carbonate-rich counterpart, which could make it easier to dislodge seagrasses. Interestingly, sediments within the Coastal Bend were sandier (~82-93% shell/ sand, 7-17% silt/clay) than tropical areas like Florida Bay (~36% shell/sand, 64% silt/clay) (Jensen et al. 2009). Since the sediment fractions of our study area were comparable to compositions in the Caribbean (~90% shell/sand, 10% silt/clay; Cruz-Palacios and van Tussenbroek 2005), additional factors other than grain size may be at play in the removal of seagrasses during Harvey.
Thalassia exhibits significant morphological plasticity across the Caribbean and the Gulf of Mexico, with leaf width displaying the most significant variability in morphological traits (McDonald et al. 2016). An earlier study (May-Ku Cool colors represent an increase in cover, but the difference between pre-and post-Harvey measurements were within the minimal detectable range of a Braun-Blanquet cover abundance scale (25%; with the majority of these differences less than 10%). et al. 2010) also noted that Thalassia blade widths were significantly greater in exposed vs. sheltered locations. Such traits are consistent with reports that seagrasses with long/ wide blades have higher leaf breaking forces (Puijalon et al. 2011;de los Santos et al. 2016). The relatively narrow blade widths of Thalassia in our study (5.7 mm) compared to other regions (8 mm; McDonald et al. 2016), which is nearly half the 10.7 mm reported by May-Ku et al. (2010), may have contributed to the unexpected damage suffered by Thalassia in Texas.
The differential effects of Harvey on Thalassia and Halodule was also notable and may be related to the disjunct distribution of pure stands of both species as well as the protective barrier to shear forces provided by Thalassia in mixed stands. In comparison to Thalassia, Halodule often grows in sheltered locations or under the protection of larger species that are more resistant to wave exposure (de los Santos et al. 2016). In our system, it is not uncommon for Halodule to colonize shoals while Thalassia grows along the outer and deeper edges of the bed, so it is possible that Thalassia may have buffered the effects of Harvey in mixed beds.
The low damage to Halodule populations may also be related to the rapid regrowth potential of this species. We noted that Thalassia shoots exhibited horizontal cuts across the tissue (Fig. 4C), presumably severed at the sediment surface. These cuts were not apparent on the much narrower Halodule shoots, which may have responded very rapidly (peak growth rates range from 4-7 mm d −1 in early fall; Dunton 1994) to higher nutrient loads from freshwater run-off following Harvey. In contrast, reported growth rates for Thalassia in the Gulf of Mexico are considerably lower at 1-3 mm d −1 in the Coastal Bend of Texas (Czerny and Dunton 1995) and 2-4 mm d −1 in Florida (Zieman 1975). Slower growth rates are life history traits typically associated with climax species like Thalassia relative to "weedy" pioneer species such as Halodule.
Here, we report substantial declines in the abundance of a climax species that is widely distributed throughout the Gulf of Mexico and Caribbean (Fig. 4). We acknowledge the temporal limitations of our study but fortunately, our samples from the Coastal Bend were collected 3-31 d before the storm and likely reflect damage wrought by Harvey. The fragmented and decayed belowground tissues in well-established seagrass beds following Harvey suggest that these impacts are more than a preseason defoliation event. Because Thalassia plants have a slow growth rate and reproduce almost exclusively by vegetative propagation, recovery rates of disturbed belowground biomass are slower than aboveground tissues (Di Carlo and Kenworthy 2008). At locations that sustained severe damage to belowground tissues, recovery had not occurred over a year later (pers. obs.). Ultimately, plant architecture (e.g., blade morphometrics) influences a plant's response to major physical events like hurricanes, but differences in life history traits (e.g., slower growth and longer turnover times of shoots/ramets) will affect the long-term recovery of late successional species like Thalassia to large disturbance events.

Conclusion
In summary, the patchy losses of seagrass habitat illustrate the differences in the resistance of plant foundation species to a major hurricane. Thalassia populations experienced higher winds, and greater wind intensity corresponded with substantial but localized decreases in the cover and blade length of a climax relative to pioneer species, and altered sediment composition (Fig. 5). The successional paradigm predicts that the poor recolonization abilities of Thalassia results in a significantly slower recovery than either Halodule or Syringodium. Therefore, the loss of Thalassia may create space for the colonization of these opportunistic species. Consequently, this disturbance provides an opportunity to document changes in species abundances and compositions across spatial and temporal scales. Long-term monitoring efforts should capture the duration and extent of the recovery of seagrass beds impacted by a major hurricane and lead to better predictions of the effects of future large-scale disturbances. prior to, and (B) following a Category 4 hurricane. Higher wind gusts (up to 241 km h −1 ) defoliated and uprooted localized patches of the climax species T. testudinum; however, H. wrightii was relatively unaffected except for beds in the direct path of the hurricane. The high energy storm also stripped foliage from individual black mangroves on the periphery of mangrove forests (A. Armitage pers. comm.), and removed organic-rich surface sediments from the bay floor, shifting the substrate composition from fine to coarse grain sizes (A. Hardison and Z. Liu pers. comm.). The presence of shoals vegetated with black mangroves may have served to reduce wind and wave fetch on adjacent Halodule beds.