Tropicalization of the barrier islands of the northern Gulf of Mexico: A comparison of herbivory and decomposition rates between smooth cordgrass (Spartina alterniflora) and black mangrove (Avicennia germinans)

The expansion of black mangrove Avicennia germinans into historically smooth cordgrass Spartina alterniflora-dominated marshes with warming temperatures heralds the migration of the marsh-mangrove ecotone northward in the northern Gulf of Mexico. With this shift, A. germinans is expected to outcompete S. alterniflora where it is able to establish, offering another prevalent food source to first order consumers. In this study, we find A. germinans leaves to be preferable to chewing herbivores, but simultaneously, chewing herbivores cause more damage to S. alterniflora leaves. Despite higher nitrogen content, A. germinans leaves decomposed slower than S. alterniflora leaves, perhaps due to other leaf constituents or a different microbial community. Other studies have found the opposite in decomposition rates of the two species’ leaf tissue. This study provides insights into basic trophic process, herbivory and decomposition, at the initial stages of black mangrove colonization into S. alterniflora salt marsh.


Introduction
As climates warm, tropical species are expanding their population ranges into historically subtropical environments [1], and lack of severe freeze events as the climate warms is expected to allow for more permanent expansion of mangrove populations in the northern Gulf of Mexico (nGOM) [2] and around the world [3,4], rather than the alternating expansion/contraction cycles witnessed in the past. Coastal wetlands of the nGOM are largely dominated by herbaceous marsh comprised of smooth cordgrass Spartina alterniflora and/or black needlerush PLOS ONE | https://doi.org/10.1371/journal.pone.0210144 January 7, 2019 1 / 13 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Juncus roemerianus, though some small regions of warmer climate within the nGOM contain stunted populations of the black mangrove Avicennia germinans. The Intergovernmental Panel on Climate Change predicts fewer severe cold events in the coming decades [5], which would facilitate continued poleward expansion of A. germinans and a gradual conversion of coastal wetlands along the nGOM from herbaceous marsh to woody mangrove swamps [2]. Along with temperature, other factors also limit the establishment and growth of A. germinans, such as wave action [6], propagule dispersal [7], propagule viability [8], microclimate [9], and hydrology [10,11]. A. germinans often colonizes shorelines populated with S. alterniflora through herbaceous marsh entrapment of mangrove propagules [7], and this mangrove colonization comes at the expense of some salt marsh [12]. This shift from salt marsh to mangrove is associated with substantial shifts in ecosystems structure and function [13,14]. This mangrove succession over salt marsh may lead to shifts in energy transfer through consumer preference and detrital pathways. Despite the significance of energy transfer on ecosystems, it can be difficult to quantify in the field.
Studies of herbivory and decomposition rates on mangroves are limited to areas where mangroves are dominant, but these rates likely differ near the edges of their geographic extent due to colder climates and different faunal and microbial assemblages. Duarte & Cebrian [15] conducted an extensive review of carbon budgets in coastal systems, and concluded herbivory amounted to 7-11% of carbon transfer in mangroves and 24-38% carbon transfer in marshes, though more recent work suggests herbivory on mangroves has been underestimated [16,17]. Decomposition rates vary with environmental conditions, including temperature [18], decomposer community [19], redox potential [20], and nutrient availability [21]. No study to the authors' knowledge has compared herbivory and decomposition rates of marsh and mangrove at the poleward extent of A. germinans.
In this study, we focus on a barrier island located at the most northern extent of A. germinans occurrence in the nGOM [22]. Three A. germinans individuals were discovered in 2012 to have colonized Horn Island. Though the exact time of colonization is unknown, the trees were 0.72-1.08m tall in April 2012. We carried out a two-year study comparing leaf herbivory, leaf nutrient content, and decomposition rates between S. alterniflora and A. germinans at this early stage of A. germinans colonization. Our results demonstrate a dichotomy in the levels of leaf herbivory, based upon mass or areal removal analysis, and offer novel findings on differences in decomposition rates between A. germinans and S. alterniflora, illustrating important consideration of region-specificity in generalizations concerning expanding or migrating species ranges.

Study site
Horn Island is a barrier island off the coast of Mississippi (30˚14'27" N, 88˚40'43"W) that marks the northern limit of A. germinans in the Gulf of Mexico [22]. The entire known mangrove population on the island, composed of three individuals at the time of this study (2012-2013), resided along the shoreline of a bayside inlet to a lagoon on the interior of Horn Island (Fig 1). The shoreline was dominated by S. alterniflora interspersed with patches of black needlerush J. romerianus and other salt marsh species.

Study design
For the first year of the study (2012), we established three stations, each comprised of a single "mixed" plot centered on one of the three surveyed A. germinans trees, thus containing the tree and underlying marsh. Each station was expanded in the second year (2013) to include a second, marsh-only ("reference"), plot, 5m away from its paired mixed plot. S. alterniflora was the dominant species of marsh plant and referred to as "S. alterniflora-mixed" in mixed plots and "S. alterniflora-ref" in reference plots. Mixed and reference plot sizes were scaled to the size of the A. germinans tree (0.25-1.0m 2 ) in the corresponding station (Fig 1).
Floral communities, A. germinans and S. alterniflora morphometrics, and herbivory were measured bimonthly over the 2012 and 2013 growing seasons (June-October). Leaf decomposition rates using the conventional litterbag method [23,24] were measured in 2012; in 2013, many of the litterbags were washed away, precluding analysis of 2013 data, and leaf carbon and nitrogen content were measured bimonthly in 2013.

Measurements
Leaf herbivory. With the permission of the United States Fish and Wildlife Service, thirty green A. germinans leaves were collected haphazardly from each tree on each sampling date. Thirty live shoots of S. alterniflora were collected haphazardly from just outside each plot around or adjacent to the tree. Photos were taken of all A. germinans leaves. S. alterniflora leaves were peeled off the shoot and their basal terminus defined by a distinct line where the leaf diverges from the stem. Height and basal width were measured for all S. alterniflora leaves. Photos were taken of the leaf sections with bite marks. A. germinans leaf area, and the area of the bite marks on A. germinans and S. alterniflora leaves were measured with ImageJ software.
To estimate leaf area for S. alterniflora leaves with tip present, we derived an allometric equation with fifteen S. alterniflora shoots collected from a marsh on Dauphin Island (301 5'3.56"N, 88˚4'39.91"W). Many S. alterniflora leaves were missing the tip, particularly those with many bite marks. This was not the case for A. germinans leaves since grazed leaves maintained their tip and oval shape despite having grazing marks. Missing tips could have endured substantial grazing; thus, disregard of these tips could lead to severe grazing underestimates [25]. We used a two-step approach to derive a possible range of herbivory intensity for leaves where the tip was missing. Four possible situations were encountered for S. alterniflora leaves collected (Fig 2). To estimate area of missing tips, we took photos of leaves with tip missing and extended both edges of each leaf using ImageJ [26] until they intersected. We measured the taping angle (mean value ± standard error: 89.5˚± 0.2˚). Then, assuming an isosceles triangle shaped-leaf, we calculated leaf area as tan(Θ)�(0.5b) 2 , where θ is the mean tapering angle and b is the basal width. From this we derived the width of the top edge at which the leaf was cut off and the tip missing (i.e., width of the isosceles triangle at the height of the leaf), using the relationship: top width = 2Δh/tan(Θ), where Δh is height difference between the reconstructed isosceles triangle and leaf height. In turn, the width of the top leaf edge was taken as the basal width for the isosceles triangle representing the missing tip, and the area of the missing tip was calculated from this width measurement and the mean tapering angle value. These estimates were included in the S. alterniflora leaf area allometric regression equation, Leaf Area calculated = 0.3478 � Leaf Area measured + 0.1383, (r 2 = 0.91), to obtain final values of missing tip areas. The missing tip reconstruction approach was tested by severing 20 complete leaves at haphazardly-chosen cross sections along each leaf blade to determine an adjustment factor for our calculated missing tip areas: Missing Tip calculated = 2.6498 � Missing Tip measured + 0.2265, (r 2 = 0.92).
Upon estimation of the area of the missing tip, the second step involved derivation of herbivory values assuming that 25%, 50%, or 75% of the missing tip had been consumed to illustrate a wide range of possible consumption scenarios. These calculations were done separately for each missing tip and the area of the bite marks added to the grazing estimation for the missing tip. Herbivory estimates were derived as percent leaf area consumed (cm 2 grazed cm -2 leaf ), absolute leaf area consumed (cm 2 grazed leaf -1 ), and absolute leaf mass consumed (grams dry weight leaf -1 ).
Cumulative leaf damage from grazers is a common proxy for herbivory [27,28], but when compared between species, leaf lifespans should be taken into consideration. Leaves of A. germinans and S. alterniflora in the studied area have similar lifespans (100-150 days, [29,30]); thus, our estimates should represent similar time scales for cumulative leaf damage.
Leaf nutrient content. Between 10-20 of the green A. germians or S. alterniflora leaves used to measure herbivory were oven dried (60˚C to a constant weight) then ground into powder using a Thomas-Wiley plant tissue grinder and analyzed for percent carbon (%C) and nitrogen (%N) by mass using a CNS analyzer [31].
Leaf litter decomposition. The conventional litterbag method [23,24] was used to estimate leaf decomposition rates. Nylon mesh bags (1.0 mm mesh) containing a known mass (air-dried) of green leaf detritus of A. germinans or S. alterniflora were fastened to rebar at the sediment surface within the plot. Leaves of both species begin to senesce while still attached to the plant, such that plucked senescent leaves would have been at different stages of decomposition. To normalize the stage of decomposition, we used green leaves for comparison. For measuring litterbag mass remaining, a conversion between oven-and air-dried material was necessary: A. germinans (r 2 = 0.98): Air-dry weight = 1.8143 � Oven-Dry weight + 0.0477; S. alterniflora (r 2 = 0.92): Air-dry weight = 1.3959 � Oven-Dry weight + 0.0186.
The following equation was fit for each vegetation type in each of the three stations: where DM retrieval and DM deployment correspond to the detritus mass (air-dry weight) remaining at retrieval and deployed at the beginning of the incubations, respectively; C is a constant, k is the decomposition rate, in days -1 ; and Δt is the time elapsed from deployment to retrieval, in days. The fits were done with least-square regression in SigmaPlot 12.3.

Statistical analyses
We used the following model for the analysis of leaf nutrient content, herbivory, and decomposition: where Station is a random blocking factor corresponding to each of the three locations where the plots were situated, Subject is a random factor, nested within Station, accounting for repeated measurements on individual organisms, Plant Type is a fixed factor corresponding to A. germinans, S. alterniflora-mixed or S. alterniflora-ref, and Time is a random, within-subject factor corresponding to collection date. For nutrient content, the response variable was percent nitrogen; for herbivory, the response variable was amount grazed (percent (%), absolute, or mass); and for decomposition, the response variable corresponded to ln(DM retrieval DM deployment ). All values obtained from a given plot on each sampling date were averaged into one true replicate. We used R version 3.2.4 [32] with package "lme4" [33] for analysis. A likelihood ratio test was used to determine significance of each factor: (1) two models were constructed for each comparison, the full model and a model lacking the factor of interest; (2) a likelihood ratio test using the "anova ()" function provided a comparison between the models; and (3) any differences between the models are attributed to the factor of interest [34]. When comparing models for a fixed factor, maximum likelihood estimators were used in the full and partial models; for random factors, restricted maximum likelihood estimators were used for both models being compared [34,35]. Post-hoc comparisons were computed for specific factor levels following the same approach. Recent, robust statistical analysis developments have allowed for inclusion of several factors to be accounted for smaller sample sizes (e.g., n = 3), though the statistical power may still be reduced [33][34][35].

Leaf herbivory
Herbivores removed more leaf area of S. alterniflora but consumed more leaf mass of A. germinans, per leaf. . Significantly more area was removed from each S. alterniflora leaf than each A. germinans leaf. On a mass basis, even if 75% of missing S. alterniflora tips were consumed, more mass per leaf would be entering herbivores via A. germinans than via S. alterniflora (Fig 3 and Table 1). Post-hoc analysis did not reveal any significant differences in herbivory between S. alterniflora-mixed and S. alterniflora-ref. A. germinans leaves were more frequently damaged (63.1% of leaves intact; 36.9% of leaves grazed) than S. alterniflora leaves (See Fig 2; 76.1% intact, no grazing; 6.4% intact, grazed; 3.2% not intact, no grazing; 14.2% not intact, grazed).

Leaf nutrient content
Leaf nitrogen content was higher in A. germinans leaves than in S. alterniflora leaves (p < 0.05, Fig 4 and

Leaf decomposition
Leaf decomposition was faster for S. alterniflora than for A. germinans leaves (Plant Type x Time: p < 0.001, Fig 5 and Table 1). Similar leaf detritus weights for the two plant types remained in the bags over the first collection times, but 328 days after deployment only ca. 2% of the initial weight remained for S. alterniflora as opposed to ca. 15% for A. germinans leaf detritus. We used the single exponential decay model because other models (such as the linear or double exponential) did not improve the quality of the fit (S1 and S2 Figs).

Discussion
There was a greater mass consumed from A. germinans leaves than S. alterniflora leaves, even when assuming 75% of each missing S. alterniflora tip was consumed (A. germinans: 0.018 g leaf -1 , S. alterniflora-mixed: 0.007 g leaf -1 , S. alterniflora-ref: 0.004 g leaf -1 , Fig 3 and   (0.8% DW, [41]) but still more than S. alterniflora (0.4% DW, [42]). The most likely dominant grazers of A. germinans in our system are grasshoppers, based upon identification of nearby marsh grazing populations [43]. The most likely dominant grazer of S. alterniflora is the marsh periwinkle Littoraria irrorate [44], who incidentally consumes S. alterniflora as it consumes fungus that infects S. alterniflora [45]. The results demonstrate A. germinans may be a preferable option to certain grazers in the nGOM. Despite this apparent grazer preference of live leaf tissue, the morphology differences between A. germinans and S. alterniflora leaves lend to a greater impact on S. alterniflora, the less consumed plant.
Primary production is a function of leaf area, and leafy plants have exhibited several adaptations to exposure of leaf surface area to sunlight [46,47]. Grazing incidentally decreases leaf area, but to sustain the same level of secondary production, more leaf area of one species of plant may need to be consumed (or removed through sloppy feeding, wherein material is removed from the plant but not consumed by the grazer) than the leaf area of another plant Table 1 asserts that usually <10% of mangrove defoliation is attributed to insect herbivory. As such, most leaf biomass ends up as detrital matter. Carbon, nitrogen, and energy also make their way into the food chain via detrital pathways: detritivores and decomposers extract a portion of this sustenance before it is either exported out of the system or sequestered in the wetland peat. Higher leaf nutrient content may result in faster decomposition [49][50][51], yet we found slower decomposition of the nitrogen-rich mangrove leaves than the nitrogen-poor cordgrass leaves. This counter-intuitive result could be influenced by potentially stronger factors, such as higher lignin concentrations or waxier cuticle [50,52], tannins, alternative nutrient limitation (i.e., phosphorus), or different microbial communities [19]. While marine systems are often nitrogen-limited, Johnson et al. [53] documented phosphorus limitation in seagrass beds nearby, suggesting detritivores may have preferred phosphorus-rich material. Phosphorus levels in leaves may have varied significantly but were not measured. Studies in this region indicate leaf phosphorus concentrations are similar between A. germinans (0.11-0.15%, [54]) and S. alterniflora (0.08-0.15%, [55]). Perry and Mendelssohn [56] found A. germinans leaves decomposed faster than S. alterniflora leaves in an area with a long-established population of A. germinans, though microbial communities can vary substantially with relative mangrove influence [57]. Microclimate conditions between litterbags of each study were unlikely to have a significantly different effect, despite the variable dimensions and mesh sizes [58].

. Chi-squared (χ 2 ), degrees of freedom (df), and p-values of factors within Likelihood Ratio Test: Station, Plant Type (A. germinans, S. alterniflora-mixed, S. alterniflora-ref), Date, and interaction of Plant
In this study, we found contrasting perspectives by investigation of first order consumption. Between A. germinans and S. alterniflora leaves, chewing herbivores consumed more A. germinans leaf material while causing less damage to the leaf, perhaps enticed by the higher nitrogen content. Chewing herbivores removed more leaf area from S. alterniflora but ingested less plant material. S. alterniflora leaves' greater area to mass ratio likely also contributed to its more rapid decomposition. The introduction of new producers like A. germinans within a system presents herbivores (and detritivores) another food choice. The magnitude of subsequent shifts in trophic processes will align with the extent of the new producer's dominance within the native system; however, those magnitudes will also align with the relative preference of first order consumers to a producer growing in that location, which may differ from that producer's growth in its native environment. By understanding dichotomies of herbivory and decomposition at range extremes, we can better identify and explain the impending changes to shifts associated with economically and ecologically valuable systems, like S. alterniflora-dominated salt marsh ecosystems of the nGOM. We look forward to future research identifying changes in trophic processes with mangrove expansion and/or changes in the marsh communities alongside other climate-motivated species range expansions.