Environmental Science

jessicalawrence30

riegl__purkis_-_2015.pdf

Home >Environmental science homework help >Environmental Science

Coral population dynamics across consecutive mass mortality events B E R N H A R D R I E G L and SAM PURKIS

National Coral Reef Institute, Nova Southeastern University, 8000 N Ocean Drive, Dania, FL 33004, USA

Abstract

Annual coral mortality events due to increased atmospheric heat may occur regularly from the middle of the century

and are considered apocalyptic for coral reefs. In the Arabian/Persian Gulf, this situation has already occurred and

population dynamics of four widespread corals (Acropora downingi, Porites harrisoni, Dipsastrea pallida, Cyphastrea

micropthalma) were examined across the first-ever occurrence of four back-to-back mass mortality events (2009–2012). Mortality was driven by diseases in 2009, bleaching and subsequent diseases in 2010/2011/2012. 2009 reduced P. har-

risoni cover and size, the other events increasingly reduced overall cover (2009: �10%; 2010: �20%; 2011: �20%; 2012: �15%) and affected all examined species. Regeneration was only observed after the first disturbance. P. harrisoni and A. downingi severely declined from 2010 due to bleaching and subsequent white syndromes, while D. pallida and

P. daedalea declined from 2011 due to bleaching and black-band disease. C. microphthalma cover was not affected. In

all species, most large corals were lost while fission due to partial tissue mortality bolstered small size classes. This

general shrinkage led to a decrease of coral cover and a dramatic reduction of fecundity. Transition matrices for dis-

turbed and undisturbed conditions were evaluated as Life Table Response Experiment and showed that C. microph-

thalma changed the least in size-class dynamics and fecundity, suggesting they were ‘winners’. In an ordered

‘degradation cascade’, impacts decreased from the most common to the least common species, leading to step-wise

removal of previously dominant species. A potentially permanent shift from high- to low-coral cover with different

coral community and size structure can be expected due to the demographic dynamics resultant from the distur-

bances. Similarities to degradation of other Caribbean and Pacific reefs are discussed. As comparable environmental

conditions and mortality patterns must be expected worldwide, demographic collapse of many other coral popula-

tions may soon be widespread.

Keywords: climate change, coral reef, demographics, mass mortality, population dynamics

Received 13 March 2015; revised version received 26 May 2015 and accepted 29 May 2015

Introduction

Coral reefs are among the most sensitive ecosystems to

excursions from mean environmental conditions, such

as temperature, UV irradiation, and nutrient levels and

suffer heavy mortality from consequent bleaching and

diseases (Hoegh-Guldberg et al., 2007; Selig et al., 2010;

Wiedenmann et al., 2013; D’Angelo & Wiedenmann,

2014; Riegl et al., 2015). Rising global temperatures and

alterations in nutrient dynamics are predicted to fur-

ther increase the frequency of coral bleaching and mor-

tality events that have already increased dramatically

since the 1980s (Donner et al., 2005; Baker et al., 2008).

Forecasts predict worst-case scenarios of annual bleach-

ing variably between 2040 and 2100 as well as increased

severity of disease outbreaks with associated concerns

for the future persistence of coral reef ecosystems

(Sheppard, 2003; Donner, 2009; Van Hooidonk et al.,

2013; Logan et al., 2014; Maynard et al., 2015).

Precise description of population-level effects of such

annual bleaching and mass mortality on corals reefs is

already possible today. The SE Arabian/Persian Gulf,

no stranger to extreme temperatures and frequent coral

bleaching (Coles & Fadlallah, 1991; Burt et al., 2011;

Purkis et al., 2011; Riegl et al., 2011; Bauman et al.,

2013a), recently experienced four significant coral mor-

tality events back-to-back. Three were caused by

bleaching and subsequent diseases (2010, 2011, and

2012), one only by a disease outbreak (2009). Thus,

annual coral mortality associated with bleaching and

diseases, as forecast by Van Hooidonk et al. (2013) and

Maynard et al. (2015), seems to already be a reality in

the world’s warmest sea with coral reefs (Feary et al.,

2010; Bauman et al., 2011; Burt et al., 2011; Riegl & Pur-

kis, 2012), and the first recorded four consecutive mass

mortalities may allow a glimpse into the future of coral

population dynamics in a highly disturbed world.

We present here the results of a detailed monitoring

program of coral populations across said four distur-

bances with special reference to flow among size

classes, because size distributions in the Gulf have been Correspondence: Bernhard Riegl, tel. +1 954-262-3671,

fax +1 954-262-3647, e-mail: [email protected]

Global Change Biology (2015) 21, 3995–4005, doi: 10.1111/gcb.13014

shown to be shaped by environmental extremes (Bau-

man et al., 2013b). This provides baseline information

about dynamical processes required for further model-

ing. We use matrix models (Caswell, 2001) to explore

aspects of population dynamics and persistence

(Hughes & Tanner, 2000). As the Gulf harbors a depau-

perate, yet typical, Indo-Pacific fauna (Coles, 2003),

these results can serve as direct analog to the future of

reefs across the Indo-Pacific.

Material and methods

Coral communities and population dynamics were monitored

in the area of Ras Ghanada (UAE). From 2007 onward, 40 pho-

totransects were obtained at two sites, three times per year. A

probabilistic sampling approach (Smith et al., 2010) was

employed, and transects were placed haphazardly near sam-

pling areas defined by GPS points, within a radius of about

50 m. Overlapping photographs created 0.75 9 5 m photocor-

ridors. Images of each transect were merged and gridded to

unit pixel size, where 1 pixel = 1 mm2. All corals were identi- fied to species, their outlines digitized and assigned a unique

color code. Individual colonies were defined as any autono-

mous, freestanding coral colonies, or parts thereof with living

tissue (Bauman et al., 2013a), a definition allowing colonies to

undergo fission by partial tissue mortality. Coral areas, mea-

sured as projected surface area, were assigned to five size bins

that corresponded to five coral life stages, defined by radius:

size class 1 (=SC1, <5 cm), SC2 (<10 cm), corals entering sex- ual maturity above a puberty size (=SC3, 10–20 cm; but vari- ability can exist in minimum size at puberty; Soong, 1993),

SC4 (20–30 cm), and SC5 (>30 cm). These size classes are justi- fied by life-history traits (citations in Riegl et al., 2012). SC3

corals truncated at edges of the transect image were included

if they measured >30 cm diameter at any point or had areas equivalent to 15 cm radius, automatically including them in

SC3. Only very few SC4 corals were encountered in transects

and were treated identical to SC3 corals. Otherwise, most large

corals would have had to be excluded, leading to bias against

the large size classes. SC5 corals were encountered so rarely

that SC4 and SC5 were amalgamated for all analyses. Small

truncated corals (i.e., that only partially fell within the photo-

transect) were ignored because inclusion might have artifi-

cially bolstered smaller size classes than the truncated corals

actually belonged to (Riegl et al., 2012).

Recruits, colonies of <2 cm diameter, were counted when- ever possible and concentrated efforts were made in 2009,

2011, and 2012. Recruitment was highly variable in space and

time (see also Bauman et al., 2011, 2014). Most recruitment was

observed in areas of caprock or artificial substrata (see also

Burt et al., 2009) outside the dense coral area, within which

recruits were frequently found in close proximity to larger cor-

als – making unequivocal distinction from asexual propagules difficult. As consistent and defensible recruit counts proved

elusive, we excluded this size class from quantitative analysis.

Over time, the size-class distributions of the measured cor-

als represented a sequence of population vectors from which

transition matrices (T) containing survival, growth, and

shrinkage probabilities could be inverted using Wood’s

method as implemented in Caswell (2001). The resulting

size-class transition models were verified by hindcasting (see

Supplemental Information). Test model runs were begun by

multiplying the inversely solved size-class transition matrices

by the first (oldest) monitoring population vector, running for as

many time steps as available from the monitoring effort (max.

N = 19), and comparing results (Figs S1, S3, S6, S9, S12, S15). As recruitment was not explicitly measured (i.e., number of

newly settled recruits), the solved matrices (T) contained pri-

marily juvenile and adult life-history events and the full life

cycle was best considered as expressed by two matrices where

A ¼ T þ F ð1Þ The full transition matrix A is the sum of the known life-his-

tory event matrix (T), which expresses the flow among the

measured size classes, and a matrix (F) of size-class-specific

fertilities. F was unknown due to spatial and temporal vari-

ability of coral recruitment and difficulty of assigning parent-

age (whether from within or outside the populations) to

observed coral recruits. T was known as it could be solved

inversely by Wood’s method.

Thus, fertility was only indirectly assessed via the fre-

quency of small corals. Recruits, which would be required for

reliable fertility estimates, were not visible in phototransects.

As the dynamics of the evaluated coral populations depend

on a combination of shrinkage and largely unquantified

recruitment, and the true value of matrix F (eq. 1) was unob-

tainable, the derivation of an asymptotic population growth

rate from the matrices (dominant eigenvalue) could not be

expected to yield all information required to understand

potential future trajectories. Nonetheless, the matrices solved

from the undisturbed period could still be treated against

those from the disturbed period as a Life Table Response

Experiment (LTRE; Hughes & Tanner, 2000; Caswell, 2001) to

explore what changes in life stages (i.e., survival/growth/

shrinkage probabilities of individual SCs) had the greatest

influence on the dominant eigenvalue of matrix T. The domi-

nant eigenvalue k of a transition matrix is the population’s asymptotic growth rate (if >1 the population will grow), and its associated eigenvector, if scaled to proportional contribu-

tion of each element, is the stable size-class distribution (SsD;

Fig. S1) at which the population will settle, if an equilibrium is

attained. Whether an equilibrium, and thus SsD, can be

reached at all depends on patterns of recruitment. Strong local

retention of larvae (i.e., recruitment of the population into

itself) will favor SsD, strong connectivity (import and export)

can make attaining SsD unlikely. Thus, the distance of

observed population vectors from theoretical SsD can give

insight into levels of connectivity vs. local recruitment (see

also Supplemental Information).

Not all results of population analyses are shown here.

Results from analyses for different species are repetitive and

therefore are mostly shown in the Supplemental Information

and described in detail there.

As phototransects were of a fixed, relatively small size,

any growth occurring in the enclosed coral populations was

3996 B. RIEGL & S. PURKIS

faithfully reproduced. Sizes of corals were evaluated as their

projected area and if considered circular, then a 10% increase

in radius would lead to 21% increase in area at each step. For

example, an arbitrary population of 100 corals consisting of

70% SC 1, assuming that all are at 5 cm diameter, will increase

in cover in 1 year from 14% of a 500 9 75 cm phototransect to

21%. All massive corals grow in the range of 1–2 cm radius per

year, Acropora grows up to 10 cm per year (Coles & Fadlallah,

1991; Riegl, 2002; Burt et al., 2010; Vajed Samiei et al., 2012;

Bauman et al., 2013a,b). Thus, the observed rapid changes in

cover were indeed realistic (Fig. 1), and they also coincided

with short-term variability as observed by Burt et al. (2010).

To calculate fecundity, we used data from Bauman et al.

(2011) and Shlesinger et al. (1998) for number of oocytes per

polyp or mesentery in Gulf or Red Sea corals, where informa-

tion from the study area was not available. Using generaliza-

tions based on measurements of skeletons following

numbers cm�2 were assigned: 12 polyps*14 oocytes cm�2 in Acropora downingi, 60 polyps (30 female)*10 oocytes cm�2

in Porites harrisoni (as this coral is gonochoric and sex ratios

in the examined population were unknown, we assumed

an even distribution with a ratio of 1; Soong, 1991), 1

polyp*700 oocytes cm�2 in Dipsastrea pallida, 15 polyps*78 oocytes cm�2 in Cyphastrea microphthalma, and 20 mesenter- ies*14 oocytes cm�2 in Platygyra daedalea. Fecundity was calculated as number of polyps/mesenteries cm�2*number of oocytes per polyp/mesentery * projected surface area of colony. It was then expressed as ‘standard fecundity’, the

summed oocyte output of 100 colonies apportioned into

observed fecund size classes (i.e., SC3, SC4 which included

SC5), thus accounting for size dependencies of life pro-

cesses. These 100 corals were assigned a space cover of ‘1’.

As disturbances decreased coral space cover and therefore

per-spatial-unit reproductive output, ‘per unit’ fecundity

was projected by multiplying ‘standard fecundity’ by the

decline in space cover (Table S2). Thus, a given fecundity

value (e.g., 1000 oocytes in 100 corals with a cover value 1)

would be decreased if fewer individuals and thus less

reproductive surface area were available (e.g., 1000 oocytes

in 100 corals*0.5*0.5, if cover had declined by half and of that cover, that of the fertile SCs had also decreased by

half).

Results

Monitored coral system

In the monitored coral system near Ras Ghanada in

eastern Abu Dhabi, 29 of the southern Gulf’s regularly

occurring 35 species existed (Table S1). Five species

accounted for >50% space cover and are known (Riegl, 1999; Riegl & Purkis, 2009, 2012) as keystone species in

southern Gulf reefs due to their common occurrence,

and/or importance as frame builders (Purkis et al.,

2005). Following analyses concentrate on these species:

(a) (b)

(g)

(a-f)

(c)

(d) (e) (f)

Fig. 1 (a-f) Trajectory of coral cover. Coral cover trajectories are linear regressions over coverage values P. harrisoni shows three

episodes of cover declines (September 2009/2010/2011), A. downingi crash after two events, D. pallida and P. daedalea cover declined

after September 2011, C. microphthalma cover remained essentially unchanged. Note that y-axis is proportional substrate cover, not

share in total coral cover. (g) Changes in the % contribution to the community, as share in the space covered by all corals. A. downingi

virtually disappeared, and C. micropthalma gained in importance, changes in the other species are not significant. Total cover by these

five species fell to about a quarter of predisturbance levels.

CORAL POPULATIONS ACROSS MASS MORTALITIES 3997

Acropora downingi (2007: 15% of coral cover; prior to

dieback in 1996/1998 cover was >70%, almost disap- peared in many areas but strong regeneration until

2010; a fast-growing and aggressive species that rapidly

monopolizes space and builds framework). Porites har-

risoni (2007: 39% of coral cover; dominant frame builder

since decline of Acropora in 1996/1998), Dipsastrea (for-

merly: Favia; Budd et al., 2012); pallida (2007: 9% of coral

cover; ubiquitous and key colonizer of new substrates);

Platygyra daedalea (2007: 28% of coral cover; common

everywhere); Cyphastrea microphthalma (2007; 9% of

coral cover; most heat-tolerant of all Gulf corals; com-

mon in most habitats and often the last coral to persist

after mass mortalities).

Diseases and temperatures as drivers of mortality events

Mortality events are shown in Table 1 and responded

to temperatures reaching or exceeding the bleaching

threshold of 35.7 °C (Riegl et al., 2011) during peak temperatures at the end of August. Except in 2009, coral

bleaching preceded diseases. Not all corals were

equally affected (Table 1). Except in 2009, onset of

events was sharp, in the last week of August or first of

September, and epidemic mortality lasted until Novem-

ber. The 2009 event was poorly defined with elevated,

but nonepidemic, mortality only in P. harrisoni from

June to November. During winter, coral diseases

always recess (Riegl, 2002). While some nearby areas

also bleached in 2013, the study area escaped renewed

bleaching apparently due to strong winds that caused

cooling of the shallow water column.

Over the study period, coral cover declined in

response to the mass mortalities, but effects and subse-

quent regeneration differed among species and events

(Fig. 1). From September 2007 to September 2010, there

was either increase in cover (P. harrisoni, D. pallida,

A. downingi) or stasis (P. daedalea, C. microphthalma).

The 2010 event reduced overall coral cover by about

20%. While some species experienced cover reduction

(P. harrisoni and A. downingi, with steepest decline; and

D. pallida), others were not affected (C. microphthalma,

P. daedalea). Regeneration of space cover or continuous

growth was observed in all species, but in A. downingi,

it was slower than predisturbance growth (Fig. 1).

Regeneration was interrupted again in 2011 with a

comparable decline (�20%) of total cover. Also cover by P. harrisoni declined as much as in the previous

year. P. daedalea also declined strongly, with less drastic

decline in D. pallida and A. downingi (already at ~5% cover). Following this event, cover per species as well

as total cover showed a persistent decline, which

reflected the continuous negative effects of coral dis-

eases. The 2011 event caused significant cover reduc-

tions across all species (Table 1) and decline in cover

continued over the entire next year.

The 2012 event caused the least decline (�15%), but cover was already low (Fig. 1). Except C. microphthalma,

which remained at essentially steady cover, and

A. downingi, which had all but disappeared, all other

species continued to decline. Over the following year,

declines in cover slowed, C. microphthalma remained

steady, while P. harrisoni increased slightly. A. downingi

remained at zero level.

In sum, cover in these four species declined over the

four events and subsequent years by 72%. This was

more than the 65% decline of coral cover in total

(Table 1), which suggests that some of the rarer species,

not studied in detail here, suffered relatively less mor-

tality (Fig. 1g). Importance (relative contribution) of

C. micropthalma quintupled (the ‘winner’), while

A. downingi disappeared from the dataset (the ‘loser’;

Table 1 Characterization of mortality events at Ras Ghanada, Abu Dhabi. Reduction in coral cover refers to all species present,

not only the four studied in detail in this study. These suffered higher losses than shown in this table (72%, see text). Bleaching level

refers to percentage of bleached corals observed across the entire community: B, bleached; PB, partially bleached; NB, not bleached

Year of event Months Cause Bleaching level Species most affected

Overall coral

cover reduction, %

2009 May–September White Syndrome, Cyanobacteria

B = 0 PB = 0 NB = 0

Porites harrisoni 10

2010 September–November Moderate bleaching,

White Syndrome,

Cyanobacteria

B = 0.1 PB = 0.11 NB = 0.78

Acropora downingi 20

2011 August–November Strong bleaching,

Black-band disease,

Cyanobacteria

B = 0.39 PB = 0.35 NB = 0.15

Platygyra daedalea,

Dipsastrea pallida,

Porites harrisoni

2012 August–November Strong bleaching,

Black-band disease

B = 0.82 PB = 0.17 NB = 0.01

Platygyra daedalea,

Dipsastrea pallida,

Porites harrisoni

3998 B. RIEGL & S. PURKIS

however, A. downingi did not disappear from the

region, colonies remained at nearby sites that fell with-

out the sampled area). While the other three species

shrank, their relative contribution to the community

did not change significantly (t-test, P = 0.7).

Size-class distribution across three subsequent mortality events

Size-class distribution clearly reflected the disturbance

events by an overall shrinkage of average coral size

(Figs 2b and S2). At the beginning of monitoring, ~70% of corals were in the smallest SC, which increased to

85% by September 2013 (Fig. S2). Most drastic changes

were seen in the major framework-building species.

A. downingi changed from 72% in the smallest size-class

predisturbance to 95% after two disturbances and then

completely disappeared. P. harrisoni increased from

55% to 86% in the smallest SC. D. pallida lost primarily

small colonies across the first two events, while the

third event reduced large colonies. P. daedalea changed

little across the first event (smallest SC changed from 50

to 52%; Fig. S2), but colonies clearly shrank across the

next two events. C. microphthalma suffered the smallest

changes in size distribution and cover across all three

mortality events (Figs 1, 2 and S2), but also suffered

loss in the big SCs.

Fecundity reduction due to tissue loss

The reductions in coral size and cover had repercus-

sions on fecundity, and significant reproductive poten-

tial was lost in all species, with smallest changes in

C. micropthalma (Table 2). The severely reduced rem-

nant population of A. downingi lost all fecundity prior

to complete disappearance from the sampling area.

This made recovery from a local population highly

unlikely and suggested that this species can only persist

if connected to another, still fertile, population. Also

P. harrisoni, D. pallida, and P. daedalea fecundity had

declined to a level that, coupled with the shrinkage in

live cover, would suggest a declining trajectory

(Table 2). C. microphthalma was a clear ‘winner’ in the

situation: not only had it suffered little cover reduction

and therefore increased its total share of cover in the

overall community (Fig. 1g), it also had suffered the

least size-based fecundity reductions. This would sug-

gest increasing importance of C. microphthalma in future

community composition (Fig. S18).

Dynamics based on projection matrices

Availability of sufficient time series of size distribu-

tions allowed inverse solution of transition matrices

containing survival, growth, and shrinkage probabili-

ties of size classes for each of the investigated species

(details in Supplemental Information). This allows

explicit definition of altered growth/mortality dynam-

ics of the survivors of the bleaching/disease out-

breaks. For all species, at least one ‘undisturbed’ and

one ‘disturbed’ matrix could be solved prior to and

across the consecutive mass mortalities (see also

Supplemental Information) and compared as a Life

Table Response Experiment (LTRE; Caswell, 2001).

Undisturbed scenarios identified all species except

A. downingi and D. pallida near theoretical stable-stage

distribution (Fig. 2). Undisturbed A. downingi showed

strong flow into the large SCs and strong retention

there, as would be expected from a fast-growing spe-

cies. This is concordant with populations tending

toward forming closed canopies of large tables in

undisturbed systems (Riegl, 1999; Baird & Hughes,

2000; Purkis & Riegl, 2004). Under disturbance, there

was a strong tendency toward shrinkage into the small-

est SC and retention there. The largest SC disappeared

(Fig. 2c).

All other species showed strong flow from the large

SCs into the smallest SC, even in the undisturbed sce-

nario. This partial mortality is indicative of constant

low-level tissue loss (predation, diseases, competition,

etc.) in a generally stressful environment. This dynam-

ics, despite strong growth into larger SCs, maintain a

large proportion of small corals, even in the expected

SsD (Fig. 2b). The net effect of disturbance across all

species was a stronger retention in the smallest two size

classes (=absence of growth) and loss of retention in the largest size classes. This was caused by shrinkage due

to partial tissue mortality. Prior to disturbance, all spe-

cies had loops in either SC3 or 4. Due to the distur-

bances, none ended with any loops in larger than SC2

and severe shrinkage, predominantly from SC3 and

SC4 into SC1, was observed. Prior to disturbances,

shrinkage was in many cases only by a single size class

(Fig. 2a,b). The overall tendency toward shrinkage due

to partial tissue loss was also clearly reflected by less/

more deviation of observed size distributions from the

theoretical SsD before/after disturbance (Fig. 2b; see

also Supplemental Information).

The LTRE confirmed stronger retention in SC1 and

SC2 and the various shrinkages into smaller SCs as the

major differences in life-history characteristics prior to

and after the disturbances in all species (see Supple-

mental Information). Among the most important contri-

butions to changes in k in all species were modifications in SC1 dynamics (greater retention). All k were below 1, which is due to matrices emphasizing

asexual (=fission due to partial tissue mortality) over sexual reproduction (see Supplemental Information)

CORAL POPULATIONS ACROSS MASS MORTALITIES 3999

(a) (b) (c)

Fig. 2 Species-specific dynamics and expected stable size-class distributions (SsD) under undisturbed (columns a and b) and disturbed

(column c) conditions. Life cycle graphs show transition probabilities among size classes (bubble size shows frequency at SsD, arrows

scale 0–1, matrices shown in Supplemental Information). Columns (b) and (c) show deviation of observed size distributions (period in

legends) from the expected SsD in undisturbed conditions (for further quantification see Supplemental Information). Note that values

in b are means over different periods than in Fig. S2

4000 B. RIEGL & S. PURKIS

and reflect the importance of shrinkage into small

(<5cm) but presumably infertile tissue remnants (Fig. S18a). In A. downingi and P. harrisoni, k were greater in the disturbed matrices, a result of SC1 having

been boosted by shrinkage and colony fission, which

increases the number of individuals, but not the

expected space cover (Fig. S18b). In the LTRE for P. har-

risoni, changes in SC1 and SC2 dynamics had the great-

est absolute effect on k (Figs S3 and S6). In A. downingi, it was the loss of SC3 and SC4. Further evaluation of

ramet survival (asexually produced, genetically identi-

cal propagules, i.e., tissue fragments) is needed to

assess whether shrinkage events can indeed be consid-

ered successful asexual recruitment events, potentially

mitigating the effects of greatly decreased space cover

and fertility and thus truly raising k. In D. pallida and P. daedalea, k decreased after disturbance as did expected cover at SsD (Fig. S18b). The severe reductions

after 2012 may be evidence of lost resilience in most

species due to decimation of the bigger SCs and, conse-

quently, fertility (Fig. S2). The LTRE suggested that

C. micropthalma was the only species not suffering

major life-dynamic changes. C. micropthalma was the

only species that additionally did not even decline in

cover and was expected to decline the least at SsD

(Fig. S18b). Except for C. microphthalma, population

resilience appears severely challenged in all species, lat-

est after three back-to-back disturbances.

Under the assumption that populations would again

approach SsD, the coral community composition

obtained from the undisturbed vs. disturbed matrix

models (Fig. 2) would differ and be dominated by

C. microphthalma and P. daedalea (Fig. S18).

Discussion

This study provides a glimpse into future coral popula-

tion dynamics. Annual bleaching events coupled with

disease outbreaks, such as described herein, are

expected as a potentially lethal endpoint of present cli-

mate warming between 2040 and 2100 (Donner, 2009;

Van Hooidonk et al., 2013; Maynard et al., 2015). Gulf

corals already exist at temperatures predicted for that

interval (Riegl et al., 2011), and disturbance frequencies

are apparently also attaining the predicted values (Van

Hooidonk et al., 2013).

One main effect of these disturbances was shrinkage

into smaller size classes and increased retention there,

causing significant loss of coral cover. Thus, as pre-

dicted for fish communities (Cheung et al., 2012), future

corals will be smaller and therefore less fecund, with

consequent lower regeneration potential of popula-

tions. Also, the linkage between bleaching and disease

outbreaks is again clearly demonstrated (Miller et al.,

2009; Selig et al., 2010; Maynard et al., 2015). Degrada-

tion dynamics, which led to ~65% reduction of overall coral cover was species-specific. The previously domi-

nant species A. downingi was a clear ‘loser’ (Loya et al.,

2001) and declined primarily by whole-colony mortal-

ity. A previously not dominant but nonetheless com-

mon species, C. microphthalma, suffered the least

changes due to partial mortality and ended as a ‘win-

ner’. The other studied species showed comparable

declines (Table S2). In all species except C. microph-

thalma, disturbances changed primarily growth-

oriented population dynamics (growth transitions

stronger than shrinkage) into such dominated by

shrinkage. This mirrors predictions by Darling et al.

(2013) that coral decline is linked to life histories.

A. downingi had dominated the undisturbed commu-

nity prior to 1996/1998 (Riegl, 1999), can be classified

as ‘competitive’ (Darling et al., 2013; McClanahan,

2014) and had proven highly susceptible to distur-

bances but capable of strong recovery (Burt et al., 2008).

After this disturbances quadruplet, these corals disap-

peared from the study site, albeit not from the entire

Table 2 Standard fecundity (expressed as oocytes per 100 colonies) prior to, and after, the significant coral cover reduction of 2010–

2013. The changed size distributions reduce fecundity (column 4), which is acerbated by reduced overall coral cover (column 5).

‘standard fecundity’ is relative to 100 fertile (SC3-5) colonies, standard fecundity was multiplied by the proportion of space cover

(planar coral area) remaining after the disturbances and the proportion of corals in SC3-5 (see Material and Methods and Table S2)

Standard fecundity

of 100 corals

2007–2010 prior

to demographic

change

Standard fecundity

of 100 corals 2013

after demographic

change

Remainder after

demographic

change only, %

Remainder after

demographic plus

cover changes, %

Acropora downingi 562 096 0 0 0

Porites harrisoni 1 658 101 241 313 15 4

Dipsastrea pallida 683 966 55 270 8 3

Platygyra daedalea 1 276 738 331 620 26 8

Cyphastrea microphthalma 4 896 132 3 187 107 65 65

CORAL POPULATIONS ACROSS MASS MORTALITIES 4001

region, where variably sized populations persist (Riegl

B, Purkis S, Burt J, Bauman A, Howells E, personal

observation). A. downingi has previously disappeared

regionally and required restocking (larvae influx) from

connected populations (Riegl, 2002; Burt et al., 2008).

However, even if some corals survived, shrinkage

would have significant negative implications for fertil-

ity, especially because Acropora also have peripheral

sterile zones (Wallace, 1999). Thus, it remains to be seen

whether surviving connected populations will be

fecund enough to sustain a comeback in this species.

P. harrisoni had dominated space cover in the study

area since the original decline of A. downingi in 1996/

1998 (Riegl & Purkis, 2012). It suffered significant par-

tial mortality and vestigial ramets, not connected by tis-

sue bridges, were the sad remnants of most colonies.

As in A. downingi, this caused a life cycle shift (Fig. 1)

favoring shrinkage transition over growth with signifi-

cantly depressed fecundity (Table 2), creating an obsta-

cle to recovery also in this species.

Within the merulinids, a hierarchy in resistance to

damage existed from D. pallida to P. daedalea and C. mi-

crophthalma. The latter was the only coral that main-

tained almost undisturbed coral cover and only minor

changes in growth dynamics across all events. D. pall-

ida and P. daedalea lost much tissue area, partial mortal-

ity was widespread and many corals shrank. Due to the

smaller size of colonies, more whole-colony mortality

was observed than in P. harrisoni. But populations

maintained higher fertility, especially P. daedalea

(Table 2) and growth dynamics (Figs 1 and S9) suggest

that the species may be able to recover. P. daedalea may

grow into larger colonies (stronger growth transition

from SC2 into SC3 in disturbed model, Fig. 2) and thus

yet again obtain higher relative cover. While also C. mi-

crophthalma suffered elevated partial mortality, this had

no clear effect on population dynamics or cover (Figs 1

and S15) and it is expected to be increasingly dominant

in the future (Fig. S18). It is therefore clearly a ‘stress-

tolerant’ species (Darling et al., 2013).

Thus, a ‘degradation cascade’ existed with declining

impacts from the originally dominant species (A. down-

ingi, the ‘loser’) to an intermediate community that may

persist until disturbances become too frequent (alter-

nate dominance by P. harrisoni, or the merulinids

D. pallida and P. daedalea), until finally a ‘stress-toler-

ant’ community (Darling et al., 2013; McClanahan,

2014) is characterized by a ‘winner’ that was previously

a less predominant species, C. microphthalma. This

observation is consistent with previous descriptions of

mortality and degradation events in the region (Riegl,

1999; Burt et al., 2008; Riegl et al., 2011; Bauman et al.,

2013a,b) and other Caribbean, Indo-Pacific, and Pacific

reefs (Berumen & Pratchett, 2006; Perry et al., 2014;

McClanahan, 2014; Glynn et al., 2015; and many

others).

The existence of such a degradation cascade is not

unique to the Gulf and has been observed in other

regions as well, most famously on Caribbean reefs

(Perry et al., 2014). The formerly dominant Acropora

(A. palmata, A. cervicornis) were depleted by bleaching

and diseases to give way to dominance by Orbicella,

especially O. annularis, ecologically equivalent to the

Gulf’s P. harrisoni. Later, Orbicella also declined to give

way to the stress-tolerant P. astraeoides and Undaria

agaricites (Green et al., 2008; Vermeij et al., 2011; Perry

et al., 2014; Edmunds, 2015). The Gulf equivalent is

dominance by small merulinids, and a degradation ser-

ies from Acropora to Platygyra and Dipsastrea to Cyphas-

trea dominance. In the most degraded and most

frequently disturbed Gulf coral systems, only C. mi-

crophthalma remains. Thus, wordwide, Acropora suffer

the greatest losses and are considered to be in wide-

spread decline (Carpenter et al., 2008).

The proximity to stable-stage distribution (SsD) may

be a useful indicator for connectivity dynamics. SsD is

best reached by closed populations or such with little

or infrequent import of recruits. A. downingi was fur-

thest from SsD, and its dynamics under disturbance

(mortality of large parts, or of entire local populations;

Riegl, 2002; Burt et al., 2008) suggest a strong connectiv-

ity requirement. Merulinids had converged toward SsD

at comparable rates and distance (as expressed by Key-

fitz’s D, see Supplemental Information), a potential indication of strong local recruitment dynamics. P. har-

risoni, the only gonochoric species, was closest to SsD,

suggesting highest reliance on local recruitment. These

findings suggest disadvantages for A. downingi and

P. harrisoni. A. downingi disadvantages stem from

increasingly few connected populations, with shrinking

fecundity due to more small, less fecund, corals

(Table 2, Fig. 2). Also local recruitment, apparently

characteristic for P. harrisoni, will decrease for the same

reasons (Table 2). Frequent disturbances should thus

increasingly favor merulinid dominance, ending in

communities of mainly C. micropthalma, which seem to

suffer less shrinkage than the other species and retain

the most fecundity (Table 2).

The observation of a ‘degradation cascade’ suggests a

pathway and potential basins of stability. Such dynam-

ics can be visualized as a weighted digraph, and by

applying the Quirk-Ruppert criteria of qualitative sta-

bility (Edelstein-Keshet, 2005). Drawn as a strongly

simplified Acropora-Porites-Merulinidae (i.e., Dipsastrea,

Platygyra, and Cyphastrea) system (that can be compli-

cated and expanded at will), we find stability in the

undisturbed system (trace < 0), instability in the mod- erately disturbed system (trace > 0), and stability again

4002 B. RIEGL & S. PURKIS

(trace < 0) in the disturbed system (Fig. 3). The instabil- ity of the moderately disturbed system is caused by

positive (i.e., amplifying) feedbacks to the death of the

previously aggressive competitor A. downingi, the

denuded skeletons of which form preferred settling

substrate for recruits of all, but most particularly their

own, species (hence ‘+’ in Fig. 3). Also, dead Porites form preferred settlement substrate for both merulinids

and Porites. This mechanism allows return to the undis-

turbed state, in which live A. downingi colonies disad-

vantage their own and recruits of others, while

aggressing on other corals, thus maintaining the

A. downingi dominated state. If, however, no A. down-

ingi recruits are available, denuded skeletons will

rapidly disintegrate (Purkis & Riegl, 2004) and the spe-

cies will play no more role in the competitive network

(hence ‘0’ in Fig. 3). While merulinids and poritids

maintain their own competitive interactions (with

merulinids disadvantaging Porites, hence ‘�’), the sys- tem is yet again stable (trace < 0). Further degradation to a merulinid-only (Cyphastrea dominated) system is

possible and also stable, as merulinid larvae will not

settle on their own live tissue (thus, trace < 0).

The described mechanism is, of course, akin to a shift

from a high-coral to a low-coral stable state. Progressive

loss of species can be quantified by changes in popula-

tion dynamics described above: corals suffer significant

tissue loss which shrinks colonies but frees up dead

skeleton as settlement substrate, provided enough graz-

ers keep algae at bay. But shrinkage also reduces fecun-

dity, and there exists a minimum size for colonies to

remain fertile. Therefore, a sequence of disturbances,

each leading to increased shrinkage of remnant coral

tissue, can lead to the eventual disappearance of corals

from the system due to loss of fertility (Hughes & Tan-

ner, 2000; Edmunds, 2007; Vermeij et al., 2011). Stability

of the degraded state can be achieved by purely demo-

graphic means (shrinkage into infertility, which, espe-

cially in case of dependence on local recruitment,

hinders regeneration), even without superior competi-

tors. In the Caribbean, this situation was famously exac-

erbated by herbivore loss that favored a capable,

unchecked, competitor (macro-algae; Hughes, 1994). In

the Gulf, macroalgae are held in check by high urchin

densities (John & George, 2001), thus, demographic

parameters and settlement dynamics (Bauman et al.,

Fig. 3 Strongly simplified community dynamics (A, Acropora; P, Porites; M, Merulinidae) in Arabian/Persian Gulf expressed as a

weighted digraph. ‘+’ signifies an interaction benefitting the species at the head of the arrow, ‘�’ disadvantages the species at the head of the arrow, and ‘0’ signifies absence of interaction, because a species is dead and most skeletons gone. The graph can be expressed as

an adjacency matrix (middle row). Signs in the main diagonal can be summed to obtain the trace.

CORAL POPULATIONS ACROSS MASS MORTALITIES 4003

2014) are the key determinants of coral regeneration

after disturbances. However, in the Eastern Pacific

(southern Gal�apagos), corals were so severely reduced

by the 1982 ENSO (El Ni~no Southern Oscillation) that

skeletons as well as reef frameworks were removed by

urchin grazing, which subsequently also disadvantaged

coral settlement (Glynn et al., 2015). Thus, these reefs

never recovered from a state of sparse, widely spaced

coral heads (mainly Pavona clavus). Depensatory (Allee)

effects clearly have come into play in Gal�apagos and

maybe also Gulf coral dynamics, where shrinkage into

the infertile SCs has depressed relative population

growth rate, thus maintaining corals in a low-cover/

small colony size state (Knowlton, 1992).

Another few disturbances and the remaining vigor-

ous Gulf reefs may also reach such a permanently

degraded ecological state. The same fate awaits all reefs

that harbor species with comparable population

dynamics. Many reefs across the Indo-Pacific are in

immediate peril.

References

Baird AH, Hughes TP (2000) Competitive dominance by tabular corals: an experi-

mental analysis of recruitment and survival of understorey assemblages. Journal of

Experimental Marine Biology and Ecology, 251, 117–132.

Baker AC, Glynn PW, Riegl B (2008) Climate change and coral reef bleaching: an eco-

logical assessment of long-term impacts, recovery trends and future outlook. Estu-

arine Coastal Shelf Science, 80, 435–471.

Bauman AG, Baird AH, Cavalcante GH (2011) Coral reproduction in the world’s

warmest reefs: southern Persian Gulf (Dubai, United Arab Emirates). Coral Reefs,

30, 405–430.

Bauman AG, Pratchett MS, Baird AH, Riegl B, Heron SF, Feary DA (2013a) Variations

in size structure is related to environmental extremes in the Persian Gulf. Marine

Environmental Research, 84, 43–50.

Bauman AG, Feary DA, Heron SF, Pratchett MS, Burt JA (2013b) Multiple environ-

mental factors influence the spatial distribution of reef communities in the north-

eastern Arabian Peninsula. Marine Pollution Bulletin, 72, 302–312.

Bauman AG, Baird AH, Burt JA, Pratchett MS, Feary DA (2014) Patterns of coral set-

tlement in an extreme environment: the southern Persian Gulf (Dubai, United

Arab Emirates). Marine Ecology Progress Series, 499, 115–126.

Berumen M, Pratchett M (2006) Recovery without resilience: persistent disturbance

and long-term shifts in the structure of fish and coral communities at Tiahura Reef,

Moorea. Coral Reefs, 25, 647–652.

Budd AF, Fukami H, Smith ND, Knowlton N (2012) Taxonomic classification of the

reef coral family Mussidae (Cnidaria: Anthozoa: Scleractinia). Zoological Journal of

the Linnean Society, 166, 465–529.

Burt J, Bartholomew A, Ussgelio P (2008) Recovery of corals a decade after a bleach-

ing event in Dubai, United Arab Emirates. Marine Biology, 154, 27–36.

Burt J, Bartholomew A, Bauman A, Saif A, Sale PF (2009) Coral recruitment and early

benthic community development on several materials used in the construction of

artificial reefs and breakwaters. Journal of Experimental Marine Biology and Ecology,

373, 72–78.

Burt J, Feary D, Usseglio P, Bauman A, Sale PF (2010) The influence of wave exposure

on coral community development on man-made breakwater reefs, with a compar-

ison to a natural reef. Bulletin of Marine Science, 86, 839–859.

Burt J, Al-Harthi S, Al-Cibahy A (2011) Long-term impacts of coral bleaching

events on the world’s warmest reefs. Marine Environmental Research, 72, 225–229.

Caswell H (2001) Matrix Population Models. Sinauer, Sunderland.

Carpenter KE, Abrar M, Aeby G et al. (2008) One-third of reef-building corals face ele-

vated extinction risk from climate change and local impacts. Science, 321, 560–563.

Cheung WWL, Sarmiento JL, Dunne J et al. (2012) Shrinking of fishes exacerbates

impacts of global ocean changes on marine ecosystems. Nature Climate Change, 3,

254–258.

Coles SL (2003) Coral species diversity and environmental factors in the Arabian Gulf

and the Gulf of Oman: a comparison to the Indo-Pacific region. Atoll Research Bul-

letin, 507, 1–21.

Coles SL, Fadlallah YH (1991) Reef coral survival and mortality at low temperatures

in the Arabian Gulf: new species-specific lower temperature limits. Coral Reefs, 9,

231–237.

D’Angelo C, Wiedenmann J (2014) Impacts of nutrient enrichment on coral reefs: new

perspectives and implications for coastal management and reef survival. Current

Opinions in Environmental Sustainability, 7, 82–93.

Darling ES, McClanahan TR, Cote IM (2013) Life histories predict coral community

disassembly under multiple stressors. Global Change Biology, 19, 1930–1940.

Donner S (2009) Coping with commitment: projected thermal stress on coral reefs

under projected future scenarios. PLoS ONE, 4e, 5712.

Donner S, Skirving W, Little C, Oppenheimer M, Hoegh-Guldberg O (2005) Global

assessment of coral bleaching and required rates of adaptation under climate

change. Global Change Biology, 11, 2251–2265.

Edelstein-Keshet L(2005) Mathematical Models in Biology, 2nd edn. SIAM Classics in

Applied Mathematics 46, 1–575.

Edmunds PJ (2007) Evidence for decadal-scale decline in the growth rates of juvenile

scleractinian coral. Marine Ecology Progress Series, 341, 1–13.

Edmunds PJ (2015) A quarter-century demographic analysis of the Caribbean coral,

Orbicella annularis, and projections of population size over the next century. Lim-

nology and Oceanography, 60, 840–855.

Feary DA, Burt JA, Bauman AG, Usseglio P, Sale PF, Cavalcante GH (2010) Fish com-

munities on the world’s warmest reefs: what can they tell us about the effects of

climate change in the future? Journal of Fish Biology, 77, 1931–1947.

Glynn PW, Riegl B, Purkis S, Kerr J, Smith T (2015) Coral reef recovery in the Gala-

pagos Islands: the northernmost islands (Darwin and Wenman). Coral Reefs, 34,

421–436.

Green DH, Edmunds PJ, Carpenter RC (2008) Increasing relative abundance of Porites

astreoides on Caribbean reefs mediated by an overall decline in coral cover. Marine

Ecology Progress Series, 359, 1–10.

Hoegh-Guldberg O, Mumby PJ, Hooten AJ et al. (2007) Coral reefs under rapid cli-

mate change and ocean acidification. Science, 318, 1737–1742.

Hughes TP (1994) Catastrophes, phase-shifts and large-scale degradation of a Carib-

bean coral reef. Science, 265, 1547–1551.

Hughes TP, Tanner JE (2000) Recruitment failure, life histories, and long-term decline

of Caribbean corals. Ecology, 81, 2250–2261.

John DM, George JD (2001) Coral death and seasonal seawater temperature regime:

their influence on the marine algae of Abu Dhabi (UAE) in the Arabian Gulf. Pro-

ceedings of the 17th International Seaweed Symposium, Capetown, 341–348.

Knowlton N (1992) Thresholds and multiple stable states in coral community dynam-

ics. American Zoologist, 32, 674–682.

Logan CA, Dunne JP, Eakin CM, Donner SD (2014) Incorporating adaptive

responses into future projections of coral bleaching. Global Change Biology, 20, 125–

139.

Loya Y, Sakai K, Yamazato K, Nakano Y, Sambali H, van Woesik R (2001) Coral

bleaching: the winners and losers. Ecology Letters, 4, 122–131.

Maynard J, van Hooidonk R, Eakin CM et al. (2015) Projections of climate conditions

that increase coral disease susceptibility and pathogen abundance and virulence.

Nature Climate Change, 5, 688–694.

McClanahan T (2014) Decadal coral community reassembly on an African fringing

reef. Coral Reefs, 33, 939–950.

Miller J, Muller E, Rogers C et al. (2009) Coral disease following massive bleaching in

2005 causes 60% decline in coral cover on reefs in the US Virgin Islands. Coral

Reefs, 28, 925–937.

Perry CT, Steneck RS, Murphy GN, Kench PS, Edinger EN, Smithers SG,

Mumby PJ (2014) Regional-scale dominance of non-framework building corals

on Caribbean reefs affects carbonate production and future reef growth. Global

Change Biology, 21, 1153–1164.

Purkis SJ, Riegl BM (2004) Spatial and temporal dynamics of Arabian Gulf coral

assemblages quantified from remote-sensing and in situ monitoring data. Marine

Ecology Progress Series, 287, 99–113.

Purkis SJ, Riegl BM, Andrefouet S (2005) Remote sensing of geomorphology and

facies patterns on a modern carbonate ramp (Arabian Gulf, Dubai, UAE). Journal

of Sedimentary Research, 75, 861–875.

Purkis SJ, Renegar DA, Riegl BM (2011) The most temperature-adapted corals have

an Achilles’ Heel. Marine Pollution Bulletin, 62, 246–250.

Riegl B (1999) Coral communities in a non-reef setting in the southern Arabian Gulf

(Dubai, UAE): fauna and community structure in response to recurrent mass mor-

tality. Coral Reefs, 18, 63–73.

4004 B. RIEGL & S. PURKIS

Riegl B (2002) Effects of the 1996 and 1998 SST anomalies on corals, coral

diseases and fish in the Arabian Gulf (Dubai, UAE). Marine Biology, 140,

29–40.

Riegl B, Purkis SJ (2009) Model of coral population response to accelerated bleaching

and mass mortality in a changing climate. Ecological Modelling, 220, 192–208.

Riegl B, Purkis SJ (2012) Coral Reefs of the Gulf: Adaptation to Environmental Extremes.

Springer, Dordrecht, 1–379.

Riegl BM, Purkis SJ, Al-Cibahy AS, Abdel-Moati MA, Hoegh-Guldberg O (2011) Pre-

sent Limits to Heat-Adaptability in Corals and Population-Level Responses to Cli-

mate Extremes. PLoS ONE, 6, e24802.

Riegl BM, Sheppard CRC, Purkis SJ (2012) Human impact on atolls leads to coral loss

and community homogenization: a modeling study. PLoS ONE, 7, e36921.

Riegl BM, Glynn PW, Wieters EW, Purkis SJ, D’Angelo CW, Wiedenmann J (2015)

Water column productivity and temperature predict coral reef regeneration across

the Indo-Pacific. Scientific Reports, 5, 8273.

Selig ER, Casey KS, Bruno JF (2010) New insights into global patterns of ocean tem-

perature anomalies: implications for coral reef health and management. Global

Ecology and Biogeography, 19, 397–411.

Sheppard CRC (2003) Predicted recurrences of mass coral mortality in the Indian

Ocean. Nature, 425, 294–297.

Shlesinger Y, Goulet TL, Loya Y (1998) Reproductive patterns of scleractinian corals

in the northern Red Sea. Coral Reefs, 132, 691–701.

Smith SG, Swanson DW, Chiappone M, Miller SL, Ault JS (2010) Probability sampling

of stony coral populations in the Florida Keys. Environmental Monitoring and

Assessment, 183, 121–138.

Soong K (1991) Sexual reproductive patterns of shallow-water reef corals in Panama.

Bulletin of Marine Science, 49, 832–846.

Soong K (1993) Colony size as a species character in massive corals. Coral Reefs, 12,

77–83.

Vajed Samiei J, Dab K, Abtahi B (2012) Short term growth rate of Acropora downingi

in the coral reef of Hengam Island, the Persian Gulf. Journal of the Persian Gulf

(Marine Science), 3, 49–54.

Van Hooidonk R, Maynard JA, Planes S (2013) Temporary refugia for coral reefs in a

warming world. Nature Climate Change, 3, 508–511.

Vermeij M, Bakker J, van der Hal M, Bak RPM (2011) Juvenile coral abundance has

decreased by over 50% in only three decades on a small Caribbean island. Diver-

sity, 3, 296–307.

Wallace CC (1999) Staghorn Coral of the World. CSIRO Publications, Canberra, ACT, 1–

421.

Wiedenmann J, D’Angelo C, Smith EG, Hunt AN, Legiret F-E, Postle AD, Achterberg

EP (2013) Nutrient enrichment can increase the susceptibility of reef corals to

bleaching. Nature Climate Change, 3, 160–164.

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Figure S1. Flow of analyses for results presented in the following text. Figure S2. Size distribution of corals at Ras Ghanada in the intervals surrounding the three subsequent bleaching and mortality events that influenced all corals as shown in Fig. 1 and also indicated by fat arrows here. Figure S3. Growth model for A. downingi. Figure S4. (a) Sensitivity and elasticity of dominant eigenvalue (k) to changes in matrix entries of disturbed and undisturbed mod- els in A. downingi. Figure S5. (a) Stable size-class distribution from undisturbed model and proximity of observed size-distributions prior to (grey col- umns) and after (white colums) disturbances, after which the deviation is increasing. Figure S6. Growth model for P. harrisoni. Figure S7. (a) Sensitivity and elasticity of dominant eigenvalue (k) to changes in matrix entries of disturbed and undisturbed mod- els in P. harrisoni. Figure S8. (a) Stable size-class distribution from undisturbed model and proximity of observed distribution prior to and after dis- turbances. Figure S9. Growth model for D. pallida. Figure S10. (a) Sensitivity and elasticity of dominant eigenvalue (k) to changes in matrix entries of disturbed and undisturbed mod- els of D. pallida. Figure S11. (a) Stable size-class distribution from undisturbed model and proximity of distribution prior to (grey columns) and after (white colums) disturbances, when deviation increased. Figure S12. Growth models for P. daedalea. Figure S13. (a) Sensitivity and elasticity of the dominant eigenvalue (k) to changes in matrix entries of disturbed and undisturbed models of P. daedalea. Figure S14. (a) Stable size-class distribution from undisturbed model and the proximity of the observed size-class distribution prior to (grey columns) and after (white colums) disturbances, when deviation is increasing. Figure S15. Growth models for C. micropthalma. Figure S16. (a) Sensitivity and elasticity of the dominant eigenvalue (k) to changes in matrix entries of disturbed and undisturbed models of C. microphthalma. Figure S17. (a) Stable size-class distribution from undisturbed model and the proximity of the observed size-class distribution prior to (grey columns) and after (white colums) disturbances, when deviation is increasing. Figure S18. Differences in (a) dominant Eigenvalue k (=asymptotic population growth-rate) prior to and after disturbances, corre- sponding to life-cycle graphs in Fig. 1 in the main text. Table S1. Coral species encountered at Ras Ghanada. Table S2. Reduction of space cover per species.

CORAL POPULATIONS ACROSS MASS MORTALITIES 4005