Analysis of coastal storm damage resistance in successional mangrove species

Use of mangrove ecosystems for coastal flood protection requires reliable predictions of mangrove wave attenuation, especially if this capacity lessens due to storm‐induced forest damage. Quantifying and understanding the variation in drag forces and mechanical properties of mangrove vegetation can improve assessment of mangrove protective capacity. We studied five mangrove species common in the subtropical Pearl River Delta, south China. The studied species range from typically landward‐occurring to more seaward‐occurring pioneer species. We sampled across seven sites in the delta to study the impact of salinity on mechanical properties. We quantified strength and flexibility of branches (branch strength and flexibility related to branch diameter, modulus of rupture and modulus of elasticity), leaf strength (leaf attachment strength related to leaf size, and leaf mass per area) and drag properties (drag force related to surface area and drag coefficient). For all tested species, larger branch diameters resulted in higher mechanical strength. Larger leaf size resulted in larger peak pulling forces and larger branch surface area resulted in stronger drag forces. Notably, species that generally occur lower in the intertidal zone, where exposure to wind and waves is higher, had relatively stronger branches but more easily detachable leaves. This may be regarded as a damage‐avoiding strategy. Across the seven field sites, we found no clear effect of salinity on mangrove mechanical properties. This study provides a mechanistic insight in the storm damage process for individual mangrove trees and a solid base for modeling storm (surge) damage at the forest scale.


Reducing cost of flood protection with coastal ecosystems
The urgency for effective and affordable coastal flood defense will increase if sea levels and storm frequency rise as projected over the coming decades (Knutson et al. 2019;Nicholls et al. 2019). This is especially the case in highly urbanized deltas, such as the Pearl River Delta in south China, that harbors megacities like Guangzhou and Shenzhen (De Dominicis et al. 2020). The cost of maintaining coastal safety increases as larger barrier structures are needed. Coastal vegetation can attenuate waves, reducing the required height of barrier structures and resulting in lower construction and maintenance costs (Borsje et al. 2011;Temmerman et al. 2013). Consequently, coastal ecosystems are increasingly considered as an addition to conventional coastal safety structures (Temmerman et al. 2013;van Wesenbeeck et al. 2017; Morris et al. 2019). In subtropical and tropical areas, mangrove forests are known for the large extent of ecosystem services they can provide, and their dense vegetation and high elevation in the intertidal zone makes them effective natural wave attenuating structures (Bouma et al. 2014;Lee *Correspondence: huzh9@mail.sysu.edu.cn This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Additional Supporting Information may be found in the online version of this article. et al. 2014). Under the right circumstances, they may even attenuate extreme storm waves, provided that vegetation stretches wide and matches the height and length of the waves (Bao 2011;Horstman et al. 2014;Menéndez et al. 2020). However, mangroves are dynamic, living structures that do not always have the same vegetation width and density. Supplementing conventional coastal safety structures with mangrove ecosystems will require a strong understanding of how mangrove width and density fluctuate, to enable careful prediction and testing of the long-term structural integrity of these ecosystems (Bouma et al. 2014).

Variability in flood protection by mangrove forests
The coastal wave attenuating capacity of mangrove ecosystems is related to surge properties and to the size, height and density of the vegetated area: a larger, denser forest leads to better protection, and vegetation height relative to the storm surge determines what part of the vegetation (pneumatophores, tree trunks, canopy) can attenuate waves (Mazda et al. 1997;Bouma et al. 2014;McIvor et al. 2015). For example, height of native species in the Pearl River Delta in south China is comparable to the seawalls behind them and can thus experience waves over the full height of the tree during storms (Fig. S1). Naturally, these vegetation properties are variable across species, age and space inside and across forests (i.e., canopy height is globally related to precipitation, temperature and cyclone frequency; Koch et al. 2009;Simard et al. 2019). Furthermore, large irregular disturbances can strongly impact forest structure and consequently wave attenuation capacity as well as long-term presence. If disturbance regimes such as storm frequency or intensity are altered under climate change (Hoegh-Guldberg et al. 2018;Knutson et al. 2019), this could drive forests that are already vulnerable from previous disturbances over an ecological tipping point, resulting in substantial narrowing or loss of the ecosystem (Scheffer et al. 2001;Bouma et al. 2014). Even if forest size is relatively stable, recovering will take some time, during which local protection capacity is lowered (Johnstone et al. 2016;Krauss and Osland 2020). Conventional flood protection structures need to adhere to rigid safety standards, often measured to withstand a storm of particular intensity, that has an estimated return period (e.g., 100 years; CIRIA et al. 2013). To safely integrate mangroves in flood protection schemes, it is essential to understand how mangrove vegetation structure changes under disturbance regimes.

Mechanistic models for storm damage predictions
Storms and accompanying storm surges can cause mangrove tree damage through direct mechanical impact such as branch and trunk damage and complete defoliation, and indirect effects like extensive flooding and displacement of large volumes of sediment (Jimenez et al. 1985;Smith et al. 1994;Ouyang et al. 2021). Predicting direct mechanical storm damage in forests can be done with models like HWIND, FOR-EOLE, and GALES (Gardiner et al. 2008). Such models can predict tree uprooting and trunk breakage by comparing tree strength with drag force generated in local (storm) wind climates (Peltola et al. 1999). Mangroves can experience damage from both wind and waves during coastal storms. For example, Tanaka (2008) observed tree damage similar to that of damage by tsunamis or river floods after Cyclone Sidr hit Bangladesh in 2007. These existing storm damage models have been developed for trees in terrestrial forests and do not incorporate the impact of water motion that may impose much larger drag forces (a drag force imposed by water moving at a velocity of 2 m s À1 is roughly equivalent to wind speeds of 130 mph or 58.3 m s À1 ; Denny and Gaylord 2002). Still, the basic principle remains: a force is acting on a tree. Thus, the modeling principle may also be applied in the case of storm surge damage on mangroves, where wind is replaced with waves and terrestrial wood properties are replaced by mangrove wood properties.

Variability in storm damage across species and space
When making mechanistic predictions of mangrove tree damage, it should be considered that not all trees are damaged in the same way. For instance, it is known that tree species vary in their mechanical strength (Chave et al. 2009;Santini et al. 2013) and mechanical flexibility (where more flexible wood can reduce storm impact; Kauffman and Cole 2010). Furthermore, leaf mechanical properties such as leaf size, which increases drag force acting on the tree, and the potential for leaf reorientation and defoliation under influence of wind and waves (which reduce surface area and resulting drag force), vary across plant species (Vollsinger et al. 2005;Onoda et al. 2011;Butler et al. 2012). These differences often follow drag avoidance or drag tolerance strategies, such that some species, often pioneers, are better equipped to deal with the force of incoming surge waves (e.g., in saltmarshes, seaward pioneers have more flexible stems, which helps to reduce drag; Schoutens et al. 2020;Puijalon et al. 2011). Beyond species differences, environmental factors may also affect the mechanical properties of mangrove trees. Wood densitywhich correlates positively with mechanical strength (Chave et al. 2009)-differs across intertidal position, countries (Santini et al. 2012), and possibly salinity (Table S1). Given these potential sources of variation, predicting storm impact on mangroves requires sufficient knowledge of mechanical and drag properties across mangrove species and environmental variables.

Aims of this research
Safely integrating mangrove forests in coastal protection schemes requires accurate predictions of forest size and structure under the influence of storms. Here, we focus on identifying storm surge resistance of mangrove trees by quantifying the mechanical and drag properties of small branches (0.5-van Hespen et al. Damage resistance in mangrove species 1.75 cm diameter) and leaves of five species. We situate the study in the Pearl River Delta, one of the largest urban deltas in the world that could benefit from nature-based flood protection with mangroves (De Dominicis et al. 2020;Menéndez et al. 2020). The five species studied here are common in the Pearl River Delta and range from typically seaward occurring with pioneer traits to more landward growing species with late-successional traits. We quantify the species' mechanical and drag properties across a salinity gradient along seven sites in the larger delta area. We quantify the following properties: The collected data are analyzed to assess: (1) potential damage avoidance strategies for different species and (2) possible salinity impact on mechanical properties. We do this to provide mechanistic insight in storm damages for individual mangrove trees across species and salinities, ultimately supporting the aim to improve predictions of mangrove-based flood safety.

Materials
We collected data on mechanical and drag properties of five mangrove species to estimate potential storm damage on individual trees (Table 1). These data comprise: 1. Strength and flexibility of small mangrove branches (diameters ranging from 0.5 to 1.75 cm, allowing us to harvest and transport branches from field to the lab without causing unacceptably large damage to trees). We measured strength as the absolute peak force (F max , N) branches can withstand and flexibility as the amount of force needed to bend a branch a certain amount (e.g., 1 mm; F=x, N mm À1 ). From this, we derived relative material properties independent of branch diameter: MOR (N mm À2 ), a relative measure of strength, and MOE (N mm À2 ), a relative measure of flexibility. 2. Leaf attachment strength (F pull , N), where we measured the peak force required to detach a leaf under a static load and compared this to leaf size (leaf surface area A leaf , cm 2 ), petiole diameter (cm), and cost of leaf production (LMA, g cm À2 ). 3. Drag force on mangrove branches (F D , N), where we measured drag force linked to branch surface area (A proj , m 2 ) and derived a drag coefficient (C D , dimensionless).
The mangrove species studied are typical for the subtropical Pearl River Delta. To identify potential differences in storm damage avoidance or resistance traits, we selected species that range from generally more sheltered, landward occurring to generally more exposed, seaward occurring pioneer species: Acanthus ilicifolius, Kandelia obovata, Aegiceras corniculatum, Avicennia marina, and Sonneratia apetala, listed here from landward to seaward, respectively. We selected seven field sites to cover salinities that range from freshwater to seawater (0-15 psu) to study the potential effect of salinity on mangrove mechanical properties.

Species and site selection
We collected data at seven locations in the Pearl River Delta, Guangdong province, south China. Between 1954 and 2008, 181 typhoons have landed in this province (Jie et al. 2012). During a typhoon, wave height can reach up to 2 m surge height can reach up to 4 m at the coast inside the Pearl River Delta (Yin et al. 2017;De Dominicis et al. 2020). The seven sites have increasing salinity ranging from 0 to 15 psu ( Fig. 1; for salinity estimation see Supporting Information, Appendix S1). We sampled five species commonly found in the subtropical Guangdong province, south China: A. ilicifolius, K. obovata, A. corniculatum, A. marina, and S. apetala listed here as ranging from respectively high intertidal (landward species) to low intertidal (seaward species; based on Peng et al. [2016] and our own field observations). Particularly, A. marina and S. apetala are considered pioneer species in the region, K. obovata and A. corniculatum less so and A. ilicifolius is not considered a pioneer species (Ren et al. 2008;Chen et al. 2015). Note that A. ilicifolius is not a woody species, unlike the other four. In the Pearl River Delta, south China, these species do not reach tall heights ($2-4 m max.), except for the non-native species S. apetala, that reaches heights of $5-12 m. The native species have heights comparable to the seawalls behind the mangrove forests (i.e., built to resist expected storm water levels), such that they can experience waves over the length of the full tree (Fig. S1). The non-native species S. apetala can be significantly higher than adjacent seawalls and may therefore be less effective in wave attenuation and more sensitive to wind forces. S. apetala is an exotic species in China and was introduced from Bangladesh for mangrove afforestation in the mid-1980s (Xin et al. 2013). Although some of the selected species have low salinity tolerance (A. ilicifolius: low, K. obovata: mid, A. corniculatum: mid, A. marina: high and S. apetala: low; Ye et al. 2005;Reef and Lovelock 2015), the 0-15 psu range is relatively low for mangroves in general and likely tolerable for all species-indeed, they grow at most sites ( Fig. 1; but note that S. apetala trees are often nursery-raised and then planted; Ren et al. 2009).

Mechanical properties of mangrove branches
We collected branches at each site for one to five species, depending on availability (Fig. 1). Sampling was carried out from 9 to 24 January 2019. One or two branches were taken van Hespen et al.
Damage resistance in mangrove species from a tree, with diameters ranging between roughly 0.50 cm and about 1.75 cm (see Fig. S2 for branch diameters per hierarchy level). We selected this size range and excluded larger sizes for three reasons: (i) in several cases, taking larger branches would have resulted in destruction of a major part of or even the whole tree, (ii) large parts of tree biomass can be found in smaller branches (Fig. S2), as trees likely have more smaller than larger branches, and (iii) the MOR and MOE are sizeindependent measures for which diameter is not relevant (Fig. S3). Each tree was selected randomly but opportunistically, considering limited accessibility due to deep creeks or dense vegetation. For each required diameter, the straightest part of a branch was selected and stripped of any side branches such that it approximated a straight cylinder. This need not be the point where a branch will actually break during a storm, but it offered the best place to get reproducible measurements on location independent tissue properties such as MOR and MOE (Gere and Goodno 2009). These properties can then be used to calculate the actual strength at any place of a branch, assuming there are no weaknesses in the branch. Collected branches were then rolled in moist paper towels and stored in airtight bags in cool boxes with ice and transferred to a fridge upon arrival at the lab before analysis. Analysis using a three-point bending test happened within 48 h after collection of the branches. The three-point bending test was carried out to obtain both absolute tissue properties-maximum load F max (N) and force to bend F=x (N mm À1 )-and relative tissue properties-MOR (N mm À2 ) and MOE (N mm À2 ; Gere and Goodno 2009). The test was carried out with a universal testing machine (SUST CMT5105, Zhuhai SUST Electrical Equipment Co., Ltd, Zhuhai, China) in which we placed a branch in the load frame, with the cross head set to move at a speed of 25 mm min À1 . Maximum load F max (N) was determined at the point where the applied force was highest before the branch started to weaken and irreversible damage occurred. The sizeindependent mechanical property MOR (N mm À2 ) was obtained with: Cost of leaf production: LMA, in gram biomass invested to produce 1 m 2 leaf, considered a measure of cost of leaf production Modulus of leaf loss, area based: Force required to detach a leaf for a given leaf size Modulus of leaf loss, mass based: Force required to detach a leaf for a given leaf cost (using LMA as measure of cost) (m) undergoing loading F (N). Radius R was measured in the middle of each branch (note that branches tend to tapes), where the cross head met the branch (Fig. S4). Branches were cut so that they fit an arm of L ¼ ; Â 20 þ 2 (to fit inside the universal testing machine and following Onoda et al. 2010 who used lengths >20 times longer than the diameter in comparable measurements), but no larger than the load frame maximum width of 38 cm. This means that branches thicker than 1.9 cm were tested with a relatively shorter arm L (68 in total). As such, we excluded these data from absolute force measurements (F max ) as they could not be compared to other branches and only included them in relative, size-independent mechanical property measurements (MOR, MOE). The force to bend F=x (N mm À1 ) was determined during the initial bending process where the cross head pushes down the branch before breaking. We used the initial slope of the stress-strain curve to obtain this measure (see Fig. S4 for an example). To obtain the size-independent mechanical property MOE (N mm À2 ), we used: Mechanical properties of mangrove leaves Leaf mechanical properties were measured in the field during the January 2019 field campaign using a static pulling approach. For each tree of which a branch was sampled, at least five healthy, intact leaves were selected semi-randomly (i.e., with a bias for accessible leaf heights). We developed a method where we closed a thin steel wire loop around the petiole or base of the leaf (in case of a very short petiole). The other end of the wire was attached to a dynamometer with a minimum load of 0.2 kg, readability at 0.02 kg intervals, and an accuracy of AE0.08 kg (PCE-HS 50N, PCE Brookhuis, Enschede, the Netherlands). The dynamometer was then pulled at a constant speed until the leaf broke off and the peak load F pull was noted. Pulled leaves were stored in plastic bags van Hespen et al.
Damage resistance in mangrove species in a cool box with ice and transferred to a fridge upon arrival at the lab, and photos and size measurements (petiole diameter, leaf width, height, and surface area) were taken within 48 h. The leaf surface area A leaf (cm 2 ) was analyzed with photo analysis software ImageJ (Schindelin et al. 2012). Leaves were oven-dried at 60 C for 24 h. Dry weight M leaf (g) of pulled leaves, averaged per tree, was used to calculate the LMA-considered a measure of investment of biomass per cm 2 leaf (Onoda et al. 2011)-as the leaf dry mass M leaf per unit leaf area A leaf (LMA ¼ M leaf =A leaf , g cm À2 ). A measure of relative leaf attachment strength was obtained by standardizing the attachment strength against the leaf surface area A leaf , where A leaf can be considered a proxy for the drag forces a leaf may be experiencing (i.e., a larger leaf will experience higher drag forces; Albayrak et al. 2014): A second measure of relative leaf attachment strength was obtained by standardizing the attachment strength against the LMA, to observe if leaves that require a higher investment of biomass per cm 2 (LMA) have more strongly attached leaves (higher F pull ): Using relative measures of leaf attachment strength gives insight in whether a species invests more in leaf attachment strength F pull if they have larger leaves A leaf , that will inherently experience more drag, or if they have more expensive leaves, as reflected by a higher LMA. Petiole diameter was not used for calculating a relative attachment measure, as many leaves broke elsewhere; in a number of pulling attempts the whole twig with multiple leaves broke off (13% of pulling attempts for A. marina, 1% K. obovata and S. apetala), and in successful pulling attempts on average 72% of leaves broke off at the leaf base, such that the petiole remained attached to the tree (98% in A. ilicifolius, 84% in A. corniculatum, 81% in A. marina, 53% in K. obovata, and 42% in S. apetala). This might be an artifact of the pulling method, where force is concentrated in the point where the steel wire meets the leaf-under real wave loading, the interaction will be quite different, though detaching will likely require forces of a similar magnitude.

Drag force on mangrove branches
We obtained absolute (peak drag force F D , N) and relative (drag coefficient C D ) measurements of the forces experienced by mangrove branches, the latter of which can be used to make comparisons across species. Branches of each mangrove species were sampled at the PR3 field site and tested in June 2019. For each branch, with and without leaves, we cut the branch to fit inside the flume (40 cm high Â 60 cm wide). Branches were cut such that the basis of the branch matched the desired diameter; anything sticking out from the 40 Â 60 cm frame was cut off. We then took a photo to measure the projected frontal surface area A proj (m 2 ) with ImageJ (Schindelin et al. 2012). Maximum drag force F D on branches, with and without leaves, was measured by placing the branch in a flume (located at the School of Marine Sciences, Sun Yat-Sen University, Zhuhai campus) and attaching the base of the branch to a force transducer (Load cell M140, UTILCELL, s.r. o., Ostrovačice, Czech Republic). Drag force F D was measured for three scenarios: (1) waves without a current, (2) a current of 15 cm s À1 without waves, and (3) waves and a 15 cm s À1 current, with water height of 33 cm, wave height of 11 cm and a wave period of 1.5 s (achieving the highest possible conditions in this flume, resulting in around max. Orbital velocities of 0.25 m s À1 for waves and currents combined with Reynolds number of about 2500). We measured current velocity at half of the water depth, which approximates the depth-averaged velocity (Hu et al. 2014;Chen et al. 2018). Following basic fluid dynamics (Morison et al. 1950), a wave-averaged drag coefficient C D was derived from the peak drag measurements for each scenario: with u = current velocity (m s À1 ) measured with acoustic doppler velocimeters, ρ = fluid density of freshwater (1000 kg m À3 ), and A proj = projected frontal surface area of vegetation (m 2 ). We used peak drag force instead of average drag force to derive the drag coefficient, as it is the maximum force that a branch experiences that may cause breakage.

Results
Generic patterns in mangrove mechanical properties Branch mechanical properties followed generic patterns across species, where thicker branches can withstand larger forces before breaking (F max , Fig. 2a, adj. R 2 = 0.78, p < 0.05), as the volume to be broken increases cubically with branch diameter. Thicker branches also tended to be more rigid (F=x, Fig. 2b, adj. R 2 = 0.56, p < 0.05). Leaf mechanical properties also followed a generic pattern across species where leaves with a larger surface area A proj or thicker petiole could withstand larger pulling forces (F pull , Fig. 3c, adj. R 2 = 0.50; Fig. 3d, adj. R 2 = 0.31, p < 0.05). Contrarily, we did not find a generic pattern between cost of leaf production (i.e., estimated as LMA) and required pulling force to detach the leaf F pull (Fig. 3g).

Variability in mechanical properties across species
We identified a trend where species that occur more seawards had significantly stronger branches but more weakly attached leaves. Firstly, this was observed for branch strength.  Fig. 3a,b). Thirdly, we observed a significantly higher mass-based Modulus of Leaf Loss MOL M for the landward species A. ilicifolius (1341 AE 505 N g À1 cm 2 ) compared to the four other species (478 AE 146 N g À1 cm 2 for K. obovata, 609 AE 159 N g À1 cm 2 for A. corniculatum, 555 AE 103 N g À1 cm 2 for A. marina, 453 AE 758 N g À1 cm 2 for S. apetala), so that A. ilicifolius required relatively much higher forces to detach given the investment made to produce the leaf; Fig. 3h). Finally, the trend where species that occur more seawards had significantly stronger branches but more weakly attached leaves can be observed in Fig. 4. The relative flexibility, MOE, followed a less distinct pattern of increase as with distance from the coast (Fig. 2e), as A. corniculatum had slightly lower MOE (1350 AE 534 N mm À2 ), as can also be observed in Fig. 2f. Furthermore, the relative leaf attachment strength MOL A did not follow the seaward to landward species pattern observed above. Rather, when comparing leaf size A leaf to pulling force F pull (Fig. 3e) (Fig. 3f).

Variability in mechanical properties across salinity
We identified some significant differences of mechanical properties of branches MOR and leaves F pull between sites for K. obovata, A. corniculatum, and S. apetala and for leaves of K. obovata, A. corniculatum, and A. marina (Fig. 5), but found no clear correlation between salinity and mechanical properties (MOR, F pull ) for any species (Table S2). The observed differences between the sites were in general smaller compared to species differences (see Supporting Information, Appendix S1 for detailed results).

Drag force on mangrove branches
The absolute drag forces as measured on mangrove branches followed generic patterns, where branches with a larger projected surface area (A proj ) experienced larger drag forces F D than smaller ones (Fig. 6), as was expected following the general drag force equation (Eq. 5). Absence of leaves drastically reduced a branch's projected surface area A proj (Fig. 6), resulting in much lower drag forces F D (Fig. 6a). Our data show a significantly higher drag coefficient C D for A. marina branches (averaged across tests: 16 AE 5.99 with leaves, 5.37 AE 2.3 without leaves; Fig. 6b; Table S4), that also required relatively strong forces MOL A to detach a leaf (Fig. 3e).

Discussion
By systematically assessing mangrove mechanical properties, we aim to contribute to the safe integration of mangrove forests in coastal flood protection structures. We measured how much force mangrove branches can experience and withstand and how this differs between mangrove species and across environments with different salinities. We observed a generic pattern across species, where thicker branches are less flexible and can withstand larger forces. We also found that larger leaves withstand larger forces before detaching from the tree. We identified a trend where species that occur more seawards have significantly stronger branches but weaker leaf attachment compared to more landward species. Across sites, we found no clear correlation between salinity and mechanical properties. Finally, drag force experienced by mangrove branches followed a generic pattern, where branches with a larger projected surface area experienced larger drag forces than smaller ones. Leaf removal drastically reduced drag force and might protect branches from breakage during a storm, making easy leaf detachment a good adaptation for pioneer species living near the forest edge.

Mechanical strategies in successional mangrove species
Coastal mangrove forests are disturbance-driven, but not all trees are damaged equally when a storm hits (Krauss and Osland 2020). Trees that grow at the seaward edge of the forest will likely be hit much harder, as the wave is still in full height. Present results suggest that the seaward dwelling pioneer species are adjusted to these harsh conditions. That is, we identified a pattern where species that generally occur more on the seaward forest edge (S. apetala, A. marina) have stronger branches yet smaller and more weakly attached leaves than landward species (A. ilicifolius, with A. corniculatum and K. obovata residing in between; Fig. 4). Such strong wood can resist larger forces, reducing the chance of breakage, while the weaker leaves detach more readily and thus reducing the surface area and drag force on the tree. This pattern is similar to  van Hespen et al.
Damage resistance in mangrove species that of intertidal marsh vegetation in temperate zones, where pioneer species also exhibit clear wave-resistance or waveavoidance strategies (Bouma et al. 2005). We propose two possible explanations for this pattern. The first is that various species have evolved strategies to avoid damage, either through resistance or reduction of drag forces. This was studied and confirmed for 28 freshwater species, with the actual strategy (avoidance vs. tolerance) depending on the type of plant (Puijalon et al. 2011). A second explanation for the seaward vs. landward differences could be thigmomorphogenesis. Thigmomorphogenesis is the process where a plant grows smaller and more compact in response to mechanical stress such as wind or water flow (Jaffe 1973;Schoelynck et al. 2015;Gardiner et al. 2016). It is likely that plants growing at the exposed seaward edge of a mangrove forest might grow phenotypes to resist or avoid mechanical stresses. Observations on salt marsh vegetation support this hypothesis. For example, Cao et al. (2020) observed that seedlings-if they were able to survive the impact of wave exposure-developed shorter and stronger phenotypes under wave exposure. Here, we will not confirm or falsify either the evolutionary or the thigmomorphogenesis hypothesis as a driving factor behind branch and leaf strength, as (1) evolutionary studies require a different approach from our mechanistic angle (see Puijalon et al. 2011), and (2) the species sampling in this study was restricted by accessibility and availability within and across sites (e.g., A. ilicifolius was never found at exposed seaward locations). Given the findings of others described here we suspect a combination of the two will be at play.

Salinity impact on mechanical properties
Although salinity can impact wood properties (Table S1), we found no clear effect of salinity as an environmental driver of mechanical properties in mangroves across the seven sites  Drag properties of mangrove branches Larger branches are stronger, but also bigger and less flexible and experience more drag force. Furthermore, branches with a higher drag coefficient can experience larger drag forces. A. marina branches experienced largest drag forces and had largest drag coefficients. Note that the drag coefficients were obtained at relatively low flow velocities (around 0.25 m s À1 for waves and currents combined), typical to that of mangroves under normal conditions; typical flow velocities in mangrove reach 0.15-0.35 m s À1 during a flood tide , whereas (modeled) storm surge conditions can produce velocities around 0.5-0.7 m s À1 (Fig. 7; Roeber and Bricker 2015;Dasgupta et al. 2019). Despite the low flow conditions, the tests are still informative about the drag coefficient, which is known to be reasonably constant when the Reynolds number is above 1000 (2500 in our test; Hu et al. 2014;Chen et al. 2018). However, higher flow conditions would likely result in strong branch realignment, reducing the frontal surface area and thus drag force on the branch (Vollsinger et al. 2005). Thus, the derived drag coefficients can be used as a most conservative estimate for single branches. Furthermore, we looked at drag on single branches, while surrounding vegetation causes turbulence and can increase drag forces more than twofold (Paul et al. 2016;Norris et al. 2017). Ideally, a much larger flume that can encapsulate entire tree canopies would be used to generate realistic currents and waves under variable storm surge heights or up to the point of branch or tree breakage. However, such flumes are expensive to run (Möller et al. 2014;van Wesenbeeck et al. 2021). Regardless of these limitations, A. marina's larger drag coefficients are in line with their branch architecture, that is rather irregular with leaves directly attached to sturdy, inflexible branches (Fig. S6). Contrarily, S. apetala branches experienced much lower drag forces and have more flexible terminal branches that allow for realignment in the water stream. Aside from flexibility of branches, leaf removal also resulted in lower drag forces and drag coefficients for all species. Possibly, leaf loss can mitigate the effects of storm wind and waves, but it is not certain when this occurs. Our static measurements are useful for species comparisons but may be improved upon-to estimate leaf loss, it is necessary to know what drag forces individual leaves experience and how much they are impacted by dynamic loading from turbulent flow, which may be different from static pulling concentrated in a small point (Vogel 2009).
Mechanical tree damage during storm surges Following methods used in mechanical storm wind damage models (Gardiner et al. 2008), our measurements of mechanical and drag properties can be used to estimate storm surge damage on idealized branches. Here, we can compare the drag force generated under different wave orbital velocities to the maximum force a branch can withstand. A branch will break if the drag force (F D ) it experiences becomes larger than the maximum force (F max ) it can withstand. This maximum strength will depend on the relative strength of the wood (MOR ) and the branch arm the drag force is acting on (L ) (Eq. S1, Appendix S1). The drag force the branch experiences will depend on the projected frontal surface area of the branch (A proj ), the drag coefficient (C D ), and the flow velocity (u ) (Eq. S2, Appendix S1). We show an example of such an estimation for idealized branches that are rigid and completely submerged in Fig. 7 (see Supporting Information, Appendix S1 for the full calculation). When scaling up to full trees, a similar approach may be used where the diameter of the idealized branch becomes the diameter of the idealized trunk. There are a number of steps that can be taken to refine this simple model. First, one can validate that the MOR for branches (0.5-1.75 cm diameter) remains the same for tree trunks (Fig. S1) by measuring the MOR of tree trunks directly-but carefully, so that no vulnerable mangrove forests are harmed. Second, one can link drag coefficients more closely to natural conditions by including surrounding vegetation in drag measurements, Dashed lines indicated maximum force F max a branch can withstand. Red dotted lines indicate estimated peak orbital velocity during a typhoon (based on Hato, August 2017). See Supporting information, Appendix S1 for calculation.
van Hespen et al.
Damage resistance in mangrove species as discussed in the previous Section, or by measuring drag in the field. Third, one can include how the realignment of branches and leaves impacts the frontal surface area A proj or arm L. In three North American hardwood species placed in a wind tunnel under wind velocities of 20 m s À1 , branch realignment resulted in a 50% reduction in projected frontal surface area (Vollsinger et al. 2005). Fourth, one may consider the tree location and average height distribution of branches and leaves per species relative to (variable) storm surge heights. For example, late-successional species are generally located higher in the intertidal zone, where they may be less submerged by a storm surge, and more sheltered from storm waves. However, in the Pearl River Delta, the average height of native trees is comparable to that of the seawalls behind them, and can thus experience waves over the full height of the tree during storms. Depending on whether it is wind or waves that will reach branches during a storm, drag properties in wind will also need to be included in the model. Fifth, trees can experience much larger loads from dynamic loading than static loading (James et al. 2013). Measuring dynamic loads is currently not possible, given the lack of knowledge on motion. Thus, measuring the difference between static vs. dynamic loading on mangrove trees and branches should be addressed in future research. Furthermore, aside from breaking mechanisms, uprooting mechanisms should also be included, particularly if storms are accompanied by large erosion events that can result in the loss of a tree's mechanical stability (Gardiner et al. 2016). Last, measuring what drag forces leaves experience under real wind and wave loads and when leaf loss occurs will help us understand if trees will experience complete defoliation, which can be lethal, particularly in species that cannot resprout epicormically (e.g., Rhizophoraceae, Gill and Tomlinson 1969;Saenger 2002).
Toward long-term coastal flood safety with mangrove forests This research provides a basis for understanding the mechanistic processes that take place in mangrove forests during major storms. It provides a comprehensive dataset of mechanical and drag properties of mangrove species in one of the most highly urbanized, at-risk urban deltas in the world. This is a key input to model assessment and an important step in the direction of analytical storm damage modeling in mangroves. This research also produces new insights in damageresistance strategies across mangrove species. These results may provide guidance on mangrove afforestation efforts for flood protection, suggesting that pioneer species are more suitable to resist the influence of the exposed seaward environment. This amplifies knowledge that pioneer species are generally more suitable to grow in the lower intertidal zone (Lewis 2005;Primavera et al. 2016). In the case of the Pearl River Delta, we emphasize that S. apetala is an introduced and potentially invasive species and its long-term impact on natural ecosystem functioning is not fully understood (Ren et al. 2009). Thus, care should be taken in species selection when afforesting for long-term flood defense.
Ultimately, long-term flood protection will require scaling up individual tree damage to whole-ecosystem damage and taking into account the indirect impacts of storms such as prolonged flooding and sedimentation (Krauss and Osland 2020). It should also consider other drivers of variability across species and space, such as regeneration speeds between species (Gill and Tomlinson 1969;Saenger 2002) or changes in tidal regime or salinity, which could alter species zonation in the coastal zone and consequently change the storm resistance of the present ecosystem (Zhu et al. 2019). As more knowledge is gained about mechanical tree damage, it will become easier to predict forest structure and size over time and provide reliable long-term flood safety predictions. This will advance the incorporation of nature in flood defense schemes and enable affordable and durable coastal flood safety while preserving one of the most precious ecosystems on earth.