HomeMy WebLinkAboutCOM 0372.102 2024-2026From: Maki Morinoue
Sent: Sunday, August 17, 2025 9A8 PM
To: Council Testimony
Cc: Kimball, Heather, Kagiwada, Jennifer, Onishi, Dennis; Kierkiewicz, Ashley; Kanealii-
Kleinfelder, Matt; Galimba, Michelle M.; Villegas, Rebecca; Inaba, Holeka; Hustace,
James
Subject: Please SUPPORT Bill 66 o c�
Attachments: Coral reef benefit from reduced land -sea impacts under ocean warming47df oc n
z
cD
CD C-\
Aloha Chair Ashley Kierkiewicz, Vice Chair Michelle Galimba, and members of the P-glicyYc-,
Committee on Planning, Land Use, and Economic Development, `�
..
1 stand in strong support of Bill 66, which enshrines climate resilience as a foundational ark
enforceable principle in our General Plan —an essential compass for safeguarding clean water, soil,
and air for all life.
Why this matters:
A 20-year study in Hawai'i demonstrates that coral reefs fared significantly better when both land -
based pollution and fishing pressure were reduced, compared to either measure alone. These
reefs demonstrated stronger recovery after heatwaves, highlighting that integrated land -sea
protection significantly enhances ecosystem resilience.
This clear scientific evidence confirms that our actions on land have a direct impact on ocean
health —and our broader ecosystem services. When pollutants like urban runoff, wastewater, and
sediment compromise reef vitality, the consequences ripple through our water quality, livelihoods,
food security, and identity.
Bill 66 isn't a policy —it's a lifeline.
• It would formalize climate resilience —protecting our environment from short-term development
greed.
• It would demand interconnected, science -backed action across planning sectors.
• It would safeguard our natural heritage and future generations from preventable harm.
Please make climate resilience a non-negotiable standard in our General Plan.
Attached is the scientific article Coral Reefs Benefit From Reduced Land -Sea Impacts Under
Ocean Warming
Mahalo for this opportunity,
Maki Morinoue
Holualoa
Comm. P
'ef. To:
Ref. Date
Article
Coral reefs benefit from reduced land -sea
impacts under ocean warming
https://doi.org/10.1038/S4l5B6-023-06394-w
Received: 21 July 2022
Accepted: 30 June 2023
Openaccess
®Check for updates
Jamison M. Govet"a, Gareth J. Williams"'m, Joey Lecky', Eric Brown", Eric Conklin',
Chelsie Counsell', Gerald Davis', Mary K. Donovan''', Kim Falinski', Lindsey Kramer°,
Kelly Kozar10, Ning Li", Jeffrey A. Maynard", Amanda McCutcheon1e, Sheila A. McKenna1e,
Brian J. Neilson'', Aryan Safaie1°, Christopher Teague", Robert Whittier` &
Gregory P. Asner''"
Coral reef ecosystems are being fundamentally restructured by local human impacts
and climate -driven marine heatwaves that trigger mass coral bleaching and mortality'.
Reducing local impacts can increase reef resistance to and recovery from bleaching'.
However, resource managers lack clear advice on targeted actions that best support
coral reefs under climate change' and sector -based governance means most land- and
sea -based management efforts remain siloed°. Here we combine surveys of reef change
with a unique 20-year time series of land -sea human impacts that encompassed an
unprecedented marine heatwave in Hawaii. Reefs with increased herbivorous fish
populations and reduced land -based impacts, such as wastewater pollution and
urban runoff, had positive coral cover trajectories predisturbance. These reefs also
experienced a modest reduction in coral mortality following severe heat stress
compared to reefs with reduced fish populations and enhanced land -based impacts.
Scenario modelling indicated that simultaneously reducing land -sea human impacts
results in a three- to sixfold greater probability of a reef having high reef -builder cover
four years postdisturbance than If either occurred in Isolation. International efforts to
protect 30%of Earth's land and ocean ecosystems by203O are underway'. our results
reveal that integrated land -sea management could help achieve coastal ocean
conservation goals and provide coral reefs with the best opportunity to persist in our
changing climate.
Coastal areas contain some of the most biologically diverse and pro-
ductive marine ecosystems on Earth'. But with four times the popula-
tion density living within 20 km of the ocean compared to the rest of
the world', direct human impacts on local scales are fundamentally
restructuring these important marine communitiesa. Coastal areas are
also affected by stronger and more frequent disturbances fuelled by
human -induced climate change'.These human stressors are especially
acute on tropical coral reefswhere up to 90%ofthe local population live
alongtheshorelinet'. Land -based stressors, such as wastewater pollu-
tion, combine with sea -based stressors, such as overfishing, to disrupt
natural ecological feedbacks on reefs". Corals are further stressed by
prolonged periods of anomalously warm ocean temperatures, known as
marineheatwavesu, thatcan cause mass coral bleachingnand mortality
and fundamentally transform reef assemblages14'`.
Reducing human impacts on local scales to maintain ecosystem
Integrity has been the guiding model of coral reef conservation for
decades'. Its importance was established in the indigenous steward-
ship of island ecosystems, which used a decentralized and integrated
resource management strategy that extended from the mountains to
theses 16.17 '
hesea"•". By contrast, contemporary centralized governance means
most terrestrial and ocean management efforts remain siloed"A". As a
result,whereas local resource managers have aspired to an integrated
land-seaapproach",evidence of its efficacy above either approach in
Isolation remains wantingand difficult to test. Detectingconservation
benefits in highly dynamic ecosystems is challenging", but recent
studies have identified salient connections between local conditions
and coral reef resistance to and recovery potential following mass
bleaching''u,u-n Managers therefore require unambiguoustargets for
the combination of land -sea human impacts they should mitigate to
supportcoral reef persistence underclimatechange, Hamperingthese
efforts area lack of spatially resolved data on local drivers of coral reef
ecosystems overtime. Researchers are often forced to use proxies
'Pacific Islands Fisheries Science Center, National Oceanic and Atmospheric Administration (NOAA), Honolulu, HI, USA. 2schoolof Ocean Sciences, Bangor University, Menai Bridge, Anglesey,
UK. 'Pacific Islands Regional Office, National Oceanic and Atmospheric Administration, Honolulu, HI, USA. 'National Park of American Samoa, Pago Pago, American Samoa, USA. 'The Nature
Conservancy, Honolulu, HI, USA. °Cooperative Institute for Marine and Atmospheric Research Honolulu, HI, USA. 'Center for Global Discovery and Conservation Science, Arizona State
University; Hilo, HI, USA. 'School of Geographical Sciences and Urban Planning, Arizona State University, Tempe, AZ, USA.'Hawsi'i Wildlife Fund Kealakekua, HI, USA. 'sNational Park Service,
Pacific Island Network Inventory and Monitoring, Hawaii National Park, HI, USA. "Department of Ocean and Resources Engineering, University of Hawarl at MAnoa, Honolulu, HI, USA.
13Symbio5eas, Carolina Beach, NC, USA. "Hawarl Division of Aquatic Resources, Honolulu, Hl, USA. °Graduate School of Oceanography, University of Rhode Island, Narragansett, RI, USA.
sHami'i Department of Health, Honolulu, HI, USA. "School of Ocean Futures, Arizona State University, Hilo, HI, USA. "These authors contributed equally. Jamison M. Gove, Gareth I. Williams.
%.mail: lamison.gove@noaogov; g.l.wiBiams@bangor.ac.uk
Nature I www.nature.com I 1
Article
a
wwaia,l9ares
o
II--'
Paa3c -
O m •�
Y
S
reg an
Human
POPUatbn
I
m
a
0
0 51
LL
U,ban
mnoff
Wastewater
pollution
A
Nutrient
loading
sediment
A Input A
rainfall A
We"
exposure A
y
i
L
tP
Number of people Area Total. luent Total nhmgen
(103 within 15 km) p03m%ha`) (103I hare) (kg he-)
permanent reef survey locations
Predaturbance
o ResponseW2015marineheatwave
2 Few years postdisturbance
Permanent reef survey data avallability
Sediment
(10a kg ha-1)
Rainfall
p03mn ha^)
Wave power
(kW .-I)
Year I I I I' I I' I I I I I Tempmalchange d
ryCl" �^ ryCpry ry�A ryO�b ryO�y ry86 ryCQ� �� ��,�1,� ry0,, ryO,ry ry0,� ryO,ti �0,� ry0,6 ^0,� ry0�� ry0,9 DaemaSe ®I�y�® IncreasB
Predisturbance Disturbance Pandkturbance
Fig.I I Select local land -sea human Impactsandenvironmental factorson
coral reefs In our studyreglon In Howell]. a, Geographic location of the
Hawaiian Islands. b,Studyregion with reef surveysshown for the following:
reef trajectories predisturbancein-23; Fig. 2).cars I response to the2015
marine heatwave In = 80; Fig. 3) and coral reefs four years postdisturbance
(n=55; Fig.4). c, Spatialdlstribution In annual, high -resolution (100 m) data on
local humanimpactsand environmental factorsfrom 2000 to 2019 (coloured
lines).Theyaxis represents distance alongthecoastline in kilometres from
north tosouth alongthe study region in b.Verticai bar represents thechange
overtime (A) for each 100 msection alongthe coast.Achange overtime is high
(H,d 250%), moderate (M, 0>d <50%) orthere is nochange (NC,grey),with
such as population density24-u and reef accessibility26, or composite
indices such as'water quality"' that can be affected by anything from
deforestation toaquaculture".Suchproxiesdonotidentifythepolicy
levers local resource managers can pull and are less likely to result in
management actions or successful conservation outcomes.
Here we present a unique 20-year time series of land -sea human
impacts and environmental factors known to affect coral reef ecosys-
tem processes across ourstudyregion in the Hawaiian Islands (Fig.la).
Human factors include urban runoff, wastewater pollution, nutrient
loading, sediment input and local restrictions on types of fishinggear.
Environmental factors Include peakand annual rainfall,wave exposure,
variability in oceantemperatures and heatstress, irradiance and phyto-
plankton biomass. We also incorporate multiple fish biomass metrics
that represent the critical role reef fish play in maintaining coral reef
ecosystem function39-u (see Extended Data Table 1 for a full list of
Factors). Wecombined this datasetwith recurring, permanentlymarked
andsite-specific underwater survey data on coral reef benthic communi-
ties (Fig.1b). Our study reefs spanned large spatiotemporal gradients
in land -sea human impactsand environmental factors (Fig.1c) chat are
compambletocoral reef ecosystemsglobally (Extended Data Fig.1), and
which experienced the most severe marine heatwave on record in the
blue hues indicatingdecreasesand red hues indicatingincreases. Change is
based on the mean difference between the firsts years(2000-2004) and the
mostrecent5 years(2015-2019)In thetimeseries. Thisaccounted foryear-to-year
variability in the episodic nature of factorssuchas wave exposure, rainfall and,
sedimentinput.Asubsetof factors isshown in cowing to space constraints.
Additional factors (notshown) include annual rainfall, phytoplankton biomass,
ocean temperature (meanandvariability), heatstress, irradiance, fishinggear
restrictions, depth and metrics offish biomass. The distribution, change
over time and variabilityof all factors areshown In Supplementary Fig. 1.
See Extended Data Table 1 and Supplementary Information for detailed
information on local land -sea human impacts and environmental factors.
Hawaiian Islands (Extended Data Fig. 2). We quantified drivers of coral
reef benthic changeatthe scale of individual reefs over 12 years before
disturbance (2003-2014), duringand immediately followingthe marine
heatwave (2014-2016) and fouryears postdisturbance (2016-2019). Our
findings show that simultaneously mitigating local human impacts on
bothlandandseasupports positive coral covertrajectories in theabsence
of periodic acute disturbance, reduces coral loss during a marine heat -
wave and promotes coral reef persistence followingsevere heat stress.
Reef trajectories predisturbance
Coral coveramong reefs surveyed in 2003 was 36.9 ±2.3%(mean ±s.e.;
n = 23) and changed byless than 3% In the subsequentyears leadingup
to the 2015 marine heatwave (Fig.2a). However,coral cover trajectories
on individual reefs varied considerably over this time period: 44%of
reefs showed a positive trajectory (that is, increased coral cover), 35%
of reefs showed a negative trajectory (that is, decreased coral cover)
and the remaining reefs showed no change (Fig. 2b).To the best of our
knowledge, no acute disturbance occurred that can explain these diver-
genttrajectories. Yet,wedid find distinct differences in local conditions
between positive and negativetrajectory reefs in the years before and
2 1 Nature I www.nature.com
b
0.30 —2003 20 •Positive trajectory
—2007 15 • Negative trajectory
0.25 —2011
—2014 to
0.20
Id 5
0.10
Range in-10 `".. ..1
0.05 mean cover -15
\ _r
-20 p
0 10 20 Coral cover(%J 8070 80 'L M1�d 9ry$0'L�^e'Le ,o ,, ,ry p Year
a 00• 0 es W 0 •Go
i;
F, -0.3 -0.2 -0.1 0 0.1 0.2 0.3
squared canonical correlation (83.2%)
50
50
100
8060
�ki�IIir�j �J%�'i yr 51 L� fIIIG',
40
20
o
20
su
40
60
v� .s
37d "e Q1'+sP`cPL°��Q\3yd`�
z so
Q
Fig. 21 Reef trajectories predlsturbanceand associated local land -sea
human Impacts and environmental factors. a. Coral cover distributions
amongsurveyed reefs between 2003 and 2014 in= 23). b, Coral cover
trajectories orind IvIdual reefs.A reef was considered on a positive trajectory
(blue; n=10)or negative trajectory(red;n= 8) ifcoralcoverbetween 2003 and
2014changed by morethan 3%.This cut-off was based on mean coral cover
rangeamongail 23 reefs forthel2-yearpredisturbance period (range 2.8%; min
34.1%; max 36.9%). Reefs with no coral cover cha nge (within t3%) are not shown.
C. Difference in local conditionsbetween positiveversus negative trajectory
reefs (PERMANOVA, pseudo-Ptat= 3.38,P= 0.001) visualized alonga single
multivariateaxis (capturingthe multidimensional and correlated nature ofthe
data, Supplementary Fig.2) usinga canonical analysisofprincipalcoordinates
(n = sameas in b). Allocation success equalled 90 and 87.5%for positive
and negativetrajectory reefs, respectively (more than50%indicates an
Inclusiveof this time frame (Fig.2c). For example, the average biomass
ofall fishes, all herbivorous Fishes andgroupsof herbivorous fishesthat
fill important ecological roles such as scrapers,grazers and browsers'0
were 24-113%(29-214 kg ha-') greater on reefswith positivetrajectories
compared tothosewith negativetrajectories (Fig.2d and Extended Data
Fig.3). These patterns probably reflect positive feedbacks, whereby
increasing coral cover promotes habitat suitability for reef fishes, with
herbivorous fishes then facilitating coral growth by reducing com-
petitive exclusion by Fleshyalgae". By contrast,wastewater pollution,
nutrient loading and urban runoff were 46-80%greater on reefs with
negative trajectories compared to those with positive trajectories.
Despitethese land -based human stressors beingcomparatively higher
on reefs with negative trajectories, reefswith positivetrajectories had
63%greater human population density (the number of people within
Increasingly more distinct set of conditions than expected by chance alone).
d. Mean difference (dots) in drop-onejackknifevalues with upper and lower
bars representing the respective maximum and minimum differences In local
human impacts and environmental factors between positive and negative
trajectory reefs (n =same as in b). Blue and red shaded regions indicate factors
that were greater on reefs that had positive and negative trajectories,
respectively. Zero line represents equal values. See Extended Data Fig.3 for
the percentage difference In local conditions between positive and negative
trajectory reefs. We included all local human impacts and environmental
factors Ind to provide a general comparison of local conditions between reefs
with divergent trajectories. See Fig. lb for reeflocations and Supplementary
Fig.3 for predictor variable distributions. See Methods, Extended Data Tablet
and Supplementary Information for detailed information on local land -sea
human impacts and environmental factors.
a 15 km radius). This finding supports the notion thathuman popula-
tion density is a poor indicator of human -driven land -sea impacts at
local scales". We observed minimal differences between positive and
negative trajectory reefs relative to fishing gear restrictions, depth,
sediment input, ocean temperatures, phytoplankton biomass and
rainfall. Wave exposure was slightly higher (8.6 kW m ') on reefs with
positive trajectories, but the difference is minor because the entire
study region is generally protected from large wave events".
Coral responseto the marine heatwave
In 2015, the Hawaiian Islands experienced the strongest marine heat -
wave on record over the past 120 years (Extended Data Fig. 2).Ocean
temperatures across our study region were 2.2 °C above normal and
Nature I www.nature.com 1 3
Article
i
3
5
E
9
d
Year
11 0.3
b
0.2
0
ii 0.1
a
,.p solo ryo �o o „o
Coral rover change(%)
DHW CC -weeks)
•4
Phytoplankton (mg ml
Urban runoff (me ha i)
-25
(201412015) (2016) `
Before marine Aflermarine 0 39 625 3,16410,0
heatwave heatwave Sediment input(kg hart)
peaked at29.4 eC (Fig.3a). Degree heatingweeks (DH Ws), a widely used
heat stress metric for coral reefs, averaged 12 DHWs among surveyed
reefs (Fig. 3b), far exceeding the eight DHW threshold expected to
cause severeandwidespread coral bleaching and mortality35. Reef sur-
veys performed one year following the marine heatwave showed that
nearly one -quarter of reefs (19 outof 80) lost more than 20%coral cover
whereas the hardest -hit reef lost 49%(Fig.3c). But not all reefs expe-
rienced such catastrophic change. Coral cover remained unchanged
or increased on 18% (14 out of SO) of reefs surveyed. This divergent
ecological responsewas unexpectedgiven that all reefswere exposed
to similarly extreme levels of heat stress (Fig.3b).
Interactions between heat stress and local conditions such as a high
abundanceof competitive macroalgae can exacerbate coral bleaching
and mortality22. However, we lack a detailed understanding of the land -
and sea -based factorsthat mediatecoral responseto marine heatwaves.
Using a generalized additive mixed -modelling framework, we identi-
fied the land -sea factors that best explained variations in coral cover
change (accountingfor starting cover) among reefs one yearafter the
201S marine heatwave in Hawai'i (Fig.3d and Extended Data Table 2).
Fig. 31 Local land -sea human Impacts and environmental factors that
modified coral response to the 2015 marine heatwave. a." istorlcal
(1986-2019) SSTs during the seasonal peak (July -December) averaged across
thestudy region;2015 marine heatwaveshown in red.b, Maximum DHW
exposure in 2015, a common heat stress metric, amongsurveyed reefs. All
reefs exceeded the eight DHW threshold expected to produce severe and
widespread coral bleaching and mortality. c, Coral cover before(2014-2015)
and one yearfollowing (2016) the marine heatwave among surveyed reefs
In = 80, Fig.lb). The Inset represents the distribution of ab solute coral cover
change. d,TheGAMM results (Rt= 0.79) showingkey factorsexplaining coral
response to the marine heatwave. Change accounts for starting condition,
defined as: percentage difference= ((A,,,-Ab1)lAk,) x 100, where & and A, are
the mean coral cover values at each reef in 2014 or 2015, and 2016, respectively
(Methods and Supplementary Fig.4). Positive and negative relationships
reduce or increase coral loss,respectively. Shaded regions represent80%
confidence Intervals. Factors with the strongest model averaged slopes are
shown. Total fish biomass and scraper biomass were also important factors in
our models but had weak slopes (representing less than 5%change; Extended
Data Fig.4). Relative Importance offactors among all models (that is, sum of
AICc model welghtsacross all modelscontainingeach factor) were: sediment
Input (0.99), scraper biomass (0.99), total fish biomass (0.90), urban runoff
(0.60), phytoplankton b In mass (0.38), wastewater pollution (0.28), peak
rainfall (0.20), nutrient loading(0.19), grazer biomass (0.16), DHW (0.08), wave
power (0.07), depth (0.06) and fishing gear restrictions(0.05). See Extended
Data Table lfor full listof factors Included In the analysis, lncludIngthose
removed that were highly correlated (r> 0.7,see Methods and Supplementary
Fig.5). See Supplementary Fig.6 for predictor variable distributions.
Coral bleaching involves the breakdown of the mutualistic rela-
tionship between the coral animal and its algal endosymbionts36
A prolonged breakdown in this relationship often results in coral star-
vation and death, as much of the energetic demands of corals are met
by the photosynthetic activity of its endosymbiontsJ6. We found that
reefs with the highestlevels of water column phytoplankton biomass
(that is, chlorophyll -a) during the marine heatwave showed reduced
coral mortality (Fig.3d). Productivity increases nearshore to tropical
islands such as Hawa i'i31 and is further concentrated by small-scale
ocean processes that attract dense aggregations of plankton3s. The
increase in nutritional subsidies to the coral animal may have helped
to reduce coral starvation during the heatwave or provided higher
energetic reserves that promoted their recovery39 In other regions (for
example, Great Barrier Reef), high levels ofchlorophyll-aare an indica-
tor of poor water quality that drives negative outcomes for corals4o
Here, chlorophyll -a was uncorrelated to land -based human impacts
(Supplementary Fig. 6) and probably reflective of natural gradients
in energetic subsidies that facilitated coral survival. Working towards
locally relevant management strategies requires understanding how
human impacts superimpose on natural biophysical drivers, such as
phytoplankton biomass26, to influence reef ecosystem response to
acute disturbance.
Coastal runoff can deliver a broad spectrum of land -based contami-
nants that degrade nearshorewater quality, with cascading effects on
coral health61. We found that reefs exposed to the lowestlevels of urban
runoff, and to a lesser extent sediment input, experienced a modest
reduction in coral mortality from themarine heatwave (Fig.3d). Urban
runoff often contains heavy metals and petrochemicals that cause
coral tissue death43 and sediment input can impede the photosyn-
thetic capacity of corals and reducegrowth byburyingcoral colonies41.
Together, these stressors can undermine the natural defence abilities
of corals and increase the likelihood of mortality from heat stress40
Although turbid waters mayshade corals from excessive sunlight that
can exacerbate coral bleaching, high levels of heat stress can override
any protective benefits decreased light may provide 43. Existing but
underused local and national policies such as the Clean Water Act in
the United Statesprovideactionable pathways for marine management
4 1 Nature I www.nature.com
810
640
= 490
0
360
0
6
250
a
'6 160
A
V 90
3
40
�10
1
0 5 10 15 20 25 30 35 40 45 50
Cover of reef -building organisms (%)
Resource
Probability
management scenario
Low
Moderate
High
Initial condition
O
0.83
0.17
0.02
Sea -based only
Q
0.30
0.70
0.14
Land -based only
Q
0.17
0.83
0.26
Integrated land -sea
Q
0.02
0.98
0.80
25 50 75 100 125 150 175 200 225 250 275
» Scraper biomass (kg ha') 4111111111(
Fig.4111-ocal management scenariosthatsupportcoral reef persistence
four yearspostdisturbance.a, Percentage coverofreefbuilding organisms
(hard coral+ crustosecorallinealgae)amongreefssurveyed in =55) in 2019,
rouryearsfollowingthe marine heatwave. Colours representlow (s25th
percentile), moderate (>25th and <75th percentile) or high (z75th percentile)
cover. b, Probabl Illy a Flow, moderate or high cover of reepbuilders shown in
relation to variations 1nscraper biomass and wastewater pollution. Example
scenarios show that simultaneously decreasing wa stewater pollution
and increasingscraper biomass results in a far greater probabilityoFhigh
reef -builder cover (scenario'C') than achieving elthe I management scenario In
isolation (scenarios'Nand'B'). The up per (250 kg he-) and lower (30 kg ha')
management scenarios for scraper biomass represented the 92nd and 36th
percentiles, respectively. Wespeclficallychose 250 kg he-' as it approximates
the long-term mean (2003-2019; n=17) scraper biomass in Kealakekua Bay, a
marine protected area In our study region where no fishing has been allowed
since 1969 (Supplementary Fig. 11). Similarly, the upper (600,0001 ha') and
lower (2,5001 W) management scenarios chosen for wastewater pollution
represented the95th and 36th percentiles of the 2019distribution,
respectively (Sup pie mentary Fig. 12). Probability va lues and lines were derived
from the top model from ordinal logistic regression modelling (Extended Data
Table3, Methods and Supplementary Information). Colours for low, moderate
and high in bare the same as those in a. See Extended Data Tablet for full list of
local land -sea human impacts and environmental factors Included in the
analysis, includingthose removed that were highly correlated (r> 0.7, Methods
and Supplementary Fig.8). See Su pp lementary fig. 9 for predictor variable
distributions.
interventions of land -based stressors44. Management strategies that
leverage such policies to help mitigate coastal runoff, particularly in
urban areas, may support increased coral survival during severe marine
heatwaves.
We also found thattotal fish biomass and scraper biomass were impor-
tant factors in our models (Extended Data Table 2). Nealthyfish popu-
lations provide numerous reef -scale ecosystem functfonss', including
some species releasingbeneficial nutrientsubsidies that increasecoral
thermal tolerance45. Scrapers remove fast-growing algal turfs that could
otherwise outcompete and overgrow stress -compromised corals30. By
comparison to phytoplankton biomass and coastal runoff, the slopes
of the relationships between total fish biomass and scraper biomass
with heat -driven coral loss were weak (Extended Data Fig.4). Intense
marine heatwaves can cause severe coral mortality even on highlypro-
tected, uninhabited reefswith intact fish populations46,suggestingthat
extreme heat stress may simply overwhelm the functional roles of reef
fish over short timescales. However, abundant fish populations, in par-
ticular herbivores, can support coral reef recovery potential following
disturbances. Understandingwhetherthis positive relationship holds
across gradients in land -based impacts is key forsupportingtargeted
fisheries management in coastal marine ecosystems.
Coral reefs fouryearspostdisturbance
The dominant reef -builders in tropical coral reef ecosystems are
hard corals and crustose coralline algae". Crustose coralline algae
are encrusting calcifying algae that fuse the reef framework together
and promote coral growth byserving as a successional prerequisite for
coral recruitment and suppressing competitive Fleshy algae". Given
that coral cover can take a decade or more to recover to prebleaching
IevelS41, assessing the total cover of reef -building organisms (hard
coral +crustosecorallinealgae) is moreindicativeofcoral reefrecovery
potential following disturbance. Our surveys fouryears followingthe
2015 marine heatwave found that reef -builder cover ranged from 3.4
to 51.9%(mean of 24.3%± 1.7 s.e.; n = 55; Fig.4a). Critically, there were
different reefs with high (morethan orequal to the75th percentile) and
low (lessthan or equal to the 25th percentile) reef -builder cover before
and after the marine heatwave. Nearly two-thirds of reefs with high
reef -builder cover in 2019 did not support such levels of cover before
the marine heatwave. Similarly, we observed a morethan 40%change
in the location of reefs with low reef -builder cover between 2015 and
2019. This reshuffling of reefs in terms of relative reef -builder cover
suggested differential coral reef persistence in the years following
severe heat stress.
We used an ordinal logistic regression framework to identify the
local land -sea human impacts and environmental factors that best
supported coral reef persistence in theyears followingthe201.5 marine
heatwave. Decreased wastewater pollution and increased scraper
biomass were the most important and significant (P <0.05) in pre-
dicting whether a reef had relatively higher reef -builder cover four
years postdisturbance (Extended Data Table3). Pollution from human
waste affects coastal marine ecosystems globally" and is especially
harmful to corals from untreated sources, such as septic tanks and
cesspools, which are both common in Hawai'i49. Consequently, high
concentrations of toxins and pathogens leach into coastal waters that
Increase coral disease, reduce coral growth and reproduction, and
increase coral susceptibility to bleaching42. These negative impacts
on coral persistence are therefore much reduced in areas of decreased
wastewater pollution. Scrapers reveal bare substrate as they feed and
facilitate the settlement, growth and survival of crustose'coralline
algae and corals following acute disturbance30. Beyond thesetop-down
effects on benthic condition, bottom -up effects of improved habitat
quality could be contributingto the positive relatfonshipwe observed
between scraper biomass and higher reef-buildercover. Parrotffsh are
the dominant scrapers in Hawai'i, and typically have home ranges of
Nature I www.nature.com 1 5
Article
less than 1 km (ref. 50). Furthermore, our scraper biomass estimates
were derived from multiple observations across several time points
followingthe marine heatwave, rather than a single snapshotestimate.
Such strong site -based fidelity, combined with our recurring surveys,
suggests that resident scrapers played a key role in promoting higher
reef -builder cover ratherthan the association driven purely byan influx
of individuals seeking more favourable habitat postdisturbance.
Sea -based management efforts are often disconnected from those
occurring on land1718. We generated management scenarios of how
varyingscraperbiomass (sea -based management) andwastewater pollu-
tion (land -based management) influenced the probability of being in
a low, moderate (more than the25th and less than the 75th percentile)
or high reef -builder cover category. Our findings indicate that an inte-
grated management approach can result in a positive synergistic
outcome for coral reefs (Fig. 4b). For example, four years following
the marine heatwave, a reef across our study region with low scraper
biomass (forexample,30kgha 1)andrelativelyhighwastewaterpollu-
tion (forexample, 600,000 I ha -I) is most likelyto have low reef -builder
cover (83% probability) (Fig. 4b,'initial condition'). Where scraper
biomass is higher (forexample, 250 kg ha-') but wastewater pollution
remains high,there is a 70%probability of moderate reef -builder cover
(scenario A). Conversely, where wastewater pollution is lower (for
example, 2,5001 ha '), but scraper biomass remains low, there is an
83% probabilityof moderate reef -builder cover (scenario B). However,
if both land and sea managementscenarios occur, there is an 80%prob-
abilityof high reef -builder cover (scenario Q. Combining land and sea
management resulted in a three- to sixfold increase in the probability
of high reef -builder coverfour years following severe heat stress than
if land or sea were managed in isolation.
Conclusion
Hereweshowthat simultaneously mitigating local land -and sea -based
human impacts promotes coral reef persistence before, during and in
the years following a historically unprecedented marine heatwave in
Hawai'i.Our uniquespatially and temporally resolved data highlighted
thespecific impacts that best correlated with coral reef persistence in
each of thesetempoml periods. For example, the biomass of all reef -Fish
groups wasassociated with positive reef trajectoriesover the 12 years
leadingup tothe marine heatwave. By contrast,scraper biomasswas the
onlyfishgroupassociated with positive outcomes for reefs four years
following severe heat stress. This suggests that reef fish play essential
functional roles at different points in time and that particular feeding
and behaviours are probablycritical for reef persistence following acute
distrubance30. Similarly, land -based impacts consistently emerged as
driving negative coral reef outcomes, but the combination of stress -
ors changed depending on the observational time window in ques-
tion. Highly resolved data on the local human impacts that drive reef
ecosystem trajectories over time are unlikely to be available in most
regions. However, our overarching finding that integrated land -sea
management benefits coral reefs under ocean warming, is applicable
to populated reefs globally.
The local human impacts we identify here represent the direct or
proximate drivers of reef condition in our study. These in turn are dic-
tated by an array of distal socioeconomic and cultural factors such
as human migration and urbanization, finance, trade and tourism
that indirectly affect how people interact with coral reefs'•51. Distal
human drivers also underpin climate change that is driving severe
marine heatwaves that trigger mass coral bleaching at global scales.
Increases in future ocean temperatures and the frequency and sever-
ity of coral bleaching events52 could simply overwhelm the positive
effects of local management actions on coral reefs. However, there is
substantial variation in the projected rates of ocean warming within
and among countries under reduced emissions scenarios". Actions
that support coral reef persistence locallyalongside global reductions
in greenhouse gas emissions may buy reefs more time to adapt and
persist into the future. Contemporary governance must therefore
shift cowards an integrated approach to align management strategies
with reef ecosystem processes and thecoincident multiscale human
drivers that affect them'.
An ambitious effort is underway to protect 30%of Earth's land and
sea areas by 2030 as part of the recently adopted Kunming -Montreal
Global Biodiversity Framework'. The motivation behind the'30 by 30'
is to support ecological resilience, conserve biodiversityand preserve
ecosystem services that underpin human well-being53. The 30 by 30
has broad participation and is being incorporated into conservation
efforts by nations globally. However, our results reveal that sea -based
managementalone is insufficient to mitigatethe full spectrum of local
human effectson coastal ecosystems such as coral reefs. These efforts
must therefore explicitly couple the respective 30% land -sea targets
to realize coastal ocean conservation goals. But in most coastal geo-
graphies, 30% protection is impractical and unethical given the high
proportion of peoplethat live nearand depend on these ecosystems54.
Instead, mitigating land -based impacts such as wastewater pollution
must occur together with fisheries governance for successful con-
servation outcomes, akin to long-standing indigenous stewardship
practices of island ecosystems". Only by adopting coupled land -sea
policy measures, alongsideglobal emissions reductions, will coral reef
ecosystems and the human communities they support have the best
opportunity for persistence in our changing climate.
Online content
Any methods,additional references, Nature Portfolio reportingsumma-
ries, source data, extended data, supplementary information, acknowl-
edgements, peer review information; details of author contributions
and competing interests; and statements of data and code availability
are available at https://doi.org/10.1038/S4l5i6-023-06394-w.
1. Hughes. T.P. at aL Coral reefs In the Anthropocene. Nature 546,82-90(2017).
2. Graham, N.A.J.,Jennings,S.,MacNeil. M.A.,Mouillot, D.& Wilson, S. K. Predicting
cBmate-driven regime shifts versus rebound potential in coral reefs. Nature618, 94-97
(2015).
3. McLeod, E. at at. The future of resilience -based management in coral reef ecosystems.
J. Environ. Manage. 233, 291-301(2019).
4. Tallaard,S.at aL Implementing integrated coastal management in a sector -based
govemance system. Ocean Coast. Manage. 87, 39-53 (2012).
5. CBD. Kumning-Montreal Global Biodiversity Framework. In Proc. Conference cfParties to
the Convention on Biological Diversity Fifteenth Meeting CBD/COP/15/L25 (2022).
6. Tittensor, D.P. et aL Global patterns and predictorsof marine biodiversityscrew laxa.
Nature 466,1098-1101(2010).
7. Kummu,M. at it, Over the hills and further away from coast global geospatial patterns of
human and erwhonmentover the 20th-21st centuries. Envimm Re& Lett. 11, 034010
(2016).
8. He, Q.&Slttiman.B.R. Climate charge, human impacts, and coastal ecosystems in the
Anthropocene. Corr. Bid. 29, R1 021-R1035 (2019).
9. Dory, S.C,et al. Climate change impacts on marine ecosystems. Annu. Rev. Mar. Sci. 4,
11-37(2012).
10. Andrew.N. L., Bright, P., de is Rua, I., Took S. J.&Vickers, M. Coastal proximity of
populations in 22 Pacific Island countries and territories. PLOS ONE 14, eO223249(2019).
11. MacNeil, M.A. at aL Water quality mediates resilience on the Great Barrier Reef. Nat Ecol.
Evo/. 3,620-627(2019).
12. Oliver, E. C. J. et al. Longer and more frequent marine heatwaves over the past century.
Nat. Common. 9,1324(2018).
13. Hughes, T.F.. at aL Global warming and mcurrent mass bleaching of corals. Nature 543,
373-377(2017).
14. Hughes, T.P. at aL Global warming transforms coral reef assemblages. Nature 556,
492496(2018).
15. Edgar, a J. at al. Continent -wide declines in shallow reef liteovar a decade of ocean
warming. Nature 615, 858-865(2023).
16. Wiinter,K.B. at Indigenous stewardship through novel approaches to collaborative
management in Hawaii. EcoL Soc.28, 26(2023).
17. Sandin, S. A. eteL Harnessing island -ocean connections to maximize marine benefits of
fsland conservation. Proc. Nad Aced. Sci. USA 119, e2122354119 (2022).
18. Halpern, B. S., Lester, S. E. & McLeod, K. L Placing marine protected areas onto the
ecosystem based management seascape. Proc. Nat Aced. Sci. USA 107, 1B312-18317
(2010).
19. Marshall. P. A., Schlbtenberg, H. Z.&West, 1. M. A Reef Manager's Guide to Coral
Bleaching (Great Barrier Reef Marine Park Authority. 2006).
6 1 Nature I www.nature.com
20. Mumby,P.J., Chaloupka,M., Swec, Y:M„Steneck, R. S. & Montem-Serra, L Revisiting the
evidentiary basis for ecologIcalcascadeswith conservation impacts. Comerv. Lett. 15,
e12847(2022).
21. Asne,GR, et al. Mapped carat mortality and refugia in an archipelago -scale marine heat
wave. Prot. Nall Aced. Sci. USA 119. e2123331119 (2022).
22. Donavan. M. K. at at. Local conditions magnify coral loss after marine heatwaves. Science
37Z 977-980 (2021).
23. Baum, J.K. et al Transformation of coral communities subjected to an unprecedented
heatwave is modulated by local disturbance. Sci. Adv. 9, eabg5615 (2023).
24. Williams, G.J.,Gave, J.M., Eynaud,Y., Zg9czynski, B. 1.& Sandin, S. A. Local human
Impacts demuple natural biophysical relationships on Pacific coral reefs. Ecography38,
751-761(2015).
25. Smith 1.E.at at. Reevaluating the health of coral reef communities: baselines and
evidence for human impacts across the central Pacific. Proc. R. Sot. & 283, 20151985
(2016).
26. Make. E. at al How accessible are carat reefs to people? A global assessment based on
travel time. EmL Lett. 19, 351-360 (2016).
27. Maine, J.et at. Human deforestation outweighs future climate change impacts of
sedimentation on coral reefs. Nat. Commun. 4,1986 (2013).
28. Hozumi,A., Hong, P. Y., Kaamedt S., Rostad, A.&Jones, B. H. Water quality, seasonally,
and trajectory of an aquaclllturewastewater plume in the Red Sea. Aquaculr. Environ.
Inter. 10, 61-77 (2018).
29. Brandt S. J. etal. Carat reef ecosystem functioning: eight core processes and the rate of
biodiversity. Front. Ecof. Environ. 17, 445-454 (2019).
30. Bellwood, D. R., Hughes, T. R. Folke, C.&Nystrom, M. Confronting the coral reef crisis.
Nature 429, 827-833 (2004).
31. CinnetJ.E. et at Meeting fisheries, ecosystem function, and biodiverftygoals ina
hummdominetedworld. Science 368, 307,311(2020).
32. Same, Y.-M., Yakob,L., Balearic, S.IS Mumby,RJ. Reciprocal facilitation and non -Linearity
maintain habitat engineering on coral reefs. Oikos 122, 428-440 (2013).
33. Cinner, J. E., Graham, N. A. J., Huchery, C.&MacNeil M. A. Global effects of local human
population density and distance to markets on the condition of coral reef fisheries.
Comerv. Sio1.27,45345B(2013).
34. Storm,J.E.at at. Wave energy resources along the Hawaiian Island chain. Renew. Energy
55,305-321(2013).
35. Skirving,W. at al ComMemp and the Coral Reef Watch Coral Bleaching Heat Stress
Product Suite version 3.1. Remote Sens. 12,3856 (2020).
36. Given, P.W. Corabreef bleaching- ecological perspectives. Coral Reefs 12, 1-17(1993).
37. Gave, 1. M. at at. Near -Island biological hotspots In barren ocean basins. Nat. Commun. 7,
10581(2016).
38. Whitney, 1. L. at at. Surface slicks are pelagic nurseries for diverse ocean fauna. Sci. Rep.
11, 3197 (2021).
39. Grottoll,A.G.,Rodrigues, L.1.&Palardy,J.E. Heterotrophic plasticity and resilience in
bleached corals. Nature 440, 1186-1189 (2006).
40. Wooldridge, S. A. Water quality and carat bleaching thresholds: formalising the
linkage for the inshore reefs of the Great Barrier Reef, Australia. Mar. Pollut. Butt. 58,
745-751(2009).
41. Fabricius,K.E. Effectsof terrestrial runoff on the ecology of corals and coral reefs: review
and synthesis. Mar. Pollut. Bull. 5O, 125-146 (2005).
42. Nafley,E.M. at al Water quality thresholds for coastal contaminant impacts on teals:
a systematic review and meta -analysis. Sci. Total Environ. 794,148632 (2021).
43. Carlson, R. R., U. J.. Crowder, L. B."net, G. P. Large -wale effects of turbidity on carat
bleaching in the Hawaiian islands. Front. Mar. Be!. 9, 969472 (2022).
44. Carlson, R.R.,Foo,S.A..Burns, J.H.R.&Asnet G. P. Untapped policy avenues to protect
coral reef ecosystems. Prec. Natl Aced. Sci. USA 119,.2117562119 (2022).
45. Shantz A.A. at al. Positive interactions between corals and damselfish Increase coral
resistance to temperature stress. Glob. Change BioL 29, 417-431(2023).
46. Vargas-Anget, B. et aL El Nillo-associated catastrophic coral mortally at Jarvis Island,
central Equatorial Pacific. Coral Reefs 38, 731-741(2019).
47. Gilmour, J. R, Smith, L. D„Heyward, A. J.,Baird, A. H.& Pratchett, M. a. Recovery of an
isolated coral reef system following severe disturbance. Science 340, 69-71(2013).
48. Tuhatske,C. at at. Mapping global inputs and impacts from of human sewage in coastal
ecosystem. PLGS ONE 16, e0258898 (2021).
49. M ... capo, M. at at. Review article: Hawail's cesspool problem: review and
recommendations for water resources and human health. J. Cont. War. Res. Ed. 170,
35-75(2020).
50. Meyer, C.G., Papastamatioµ Y. P. & Clark, T. 6. Differential movement patterns and site
fidelity among trophic groups of reef fishes in a Hawaiian marine protected area. Mar.
BioL 157,1499-1511(2010).
51. Cinne4l.E.&Kittinger,1.N.in EcalogyofFishes an Coral Rests (ad. Mora, C.) 215-220
(Cambridge Univ. Press, 2015).
52. van Hooidonk,R. et al Locabscale projections ofcoral reef fueeesand implications of
the Paris Agreement. Sci. Rep 6,39666 (2016).
53. Dinerateiq E. at al. A global deal for nature: guiding principles, milestones, and targets.
Sci. Adv. S, eaaw2869 (2019).
54. Obura, D.O. et at. Achieving a nature -and people -positive future. One Earth 6,105-117
(2023).
Publishefs note Springer Nature remains neutral with regard to jurisdictional claim in
published maps and Institutional affilations.
OpetrAecess This article is licensed under a Creative Comments Attribution
4.0 International License, which permits use, sharing, adaptation, distribution
and reproduction in any medium or fam:at as tong as you give appropriate
credit to the original author(s) and the source, provide link to the Creative Commons licence,
and indicate if changes were made. The images or other third party material in this article are
Included in the article's Creative Commons licence, unless indicated otherwise in credit tine
to the material. If material is not included in the article's Creative Commons licence and your
Intended use is not permitted by statutory regulation or exceeds the permitted use, you wil
need to obtain permission directly from the copyright holder. Toview a copy of this licence,
visit hdp-.//creagvecommonszrg/licenses/by/4.0/.
® This Is U.S. Government work and not under copyright protection in the US; foreign
copyright protection may apply 2023
Nature I www.nature.com 1 7
Article
Study site
Hawaii Island (19.550 N,155.660 W) is the southeastern most island of
the Hawaiian Archipelago, located in the northern central Pacific (Fig.1).
The western section has roughly200 km of coastline predominantly
oriented north to south. Th e coastline contains the longest contiguous
reef ecosystem in the main Hawaiian Islands" and large gradients in
human population, local land -sea impacts and environmental factors
that are comparable to reef ecosystemsglobally (Extended Data Fig.1).
The region represents an ideal stu dy I ocation for resolving the land -sea
human impacts driving reef ecosystem change and coral trajectories
following acute climate -driven disturbance.
Reefsurveys
Full details related to sampling design, site selection and survey
frequency for benthic and reef -fish data collection across our study
region are in the Supplementary Information. In brief, underwater
visual surveys of benthic assemblages were collated fromthreemoni-
toring programmes for the following years (number of reefs surveyed
are in parentheses): 2003 (23),2007 (23),2011(23),2014 (40),2015
(40), 2016 (80), 2017 (80), 2018 (15) and 2019 (55). All benthicsurveys
used permanently marked pins to ensure the same area of reef was
surveyed overtime. High -resolution photographs were collected by
using photoquadrats atl m intervals along 25 m belt-transects (n = 26
photographs per transect). Between 30 and 50 random points were
overlaid on each photograph and the benthic component under each
point was identified to the lowest possible taxonomic level. Percent-
age cover of the major functional groups at each reef were used in this
analysis, namely hard coral and crustose coaalline algae. Surveys of
reef -fish assemblages were performed along the same permanently
marked 25 m transects concurrently with benthic surveys. However,
reef -Fish surveys were performed more frequently (one to six times
per year from 2003 to 2019) than benthic surveys, depending on the
reef location and monitoring programme performing the surveys. In
all surveys, fishes were identified to species, sized and enumerated.
To account for differences among programmes in how researchers
surveyed reef fish, counts were calibrated using species and method
specific adjustments6.
Local land -sea human impacts and environmental factors
Fish biomass. The biomass of fishes at a given reef was measured as
total fish biomass, herbivore fish biomass and thebiomass of browsers,
grazers and scrapers'`. Total fish biomass is an indicatorof the overall
stateof the fish assemblage"and is reduced in areasthat have increased
fishing pressure'a".In Hawaii, non-commercial nearshore fisheries
dominate, with people fishingfor recreational, subsistenceand cultural
purposes60•61. However,the dominant harvesting modes and magnitude
of fishing activities are largely unknown at spatial or temporal scales
relevant to this stud yA2. As such, we include total fish biomass in part
to represent fishing effort on reefs but recognize its shortcomings in
capturing reef- and species -specific differences in fishing pressure
across our study region. We also include herbivores and subdivisions
by feeding guilds that represent important indicators of resilience
on coral reefs30,61'6° Browsers are defined as herbivores that feed on
macroalgae and associated epiphytic material, and are important for
reducing the cover of larger, more established macroalgae. Grazers
are herbivores that feed largely on small algal turfs, helpingto prevent
their succession into larger macroalgae, and scrapers are herbivores
thatcloselycrop the substrate and open up new space to promote the
settlement,growth and survival of crustose corallinealgae and corals'°.
We followed established methods for calculating fish biomass56
The biomass of individual fishes was estimated using the allometric
length -weight conversion: W = aTl_ , where parameters a and b are
species -specific constants, TL Is total length (cm) and Wisweight(g).
Length -weight fitting parameters were obtained from a comprehen-
sive assessment of Hawai'i specific parameters` and FishBase6'. Fish
specieswere excluded from fish biomass calculations accordingto life
history characteristics that are not well captured with visual surveys,
including cryptic benthic species, nocturnal species, pelagicschooling
species and manta rays.
Human population. We quantified human population density using
NASA Gridded Population of the World v.4 (ref.66). The dataset is avail-
ableatl kmresolutlon at5-year intervals. Linear interpolation was used
to fill in the missing years and produce annual time steps of human
population withinl5 kmofeach100 mgrid cell acrossourstudy region
(Supplementary Fig.12).
Wastewater pollution. Wecalculatedwastewatereffluent(Iha 'yr')
and nitrogen input (kg ha-' yr ') from onsite sewage disposal systems
(for example, cesspools and septic tanks) and injection wells (collec-
tively OSDS) in coastal watersat 100 m resolution. OnlyOSDS located
within a modelled one-year groundwater travel time of the coastwere
included in the analysis and nutrients from OSDS were assumed to
Row to the nearest point on the shoreline. Wastewater effluent and
nutrient inputwere estimated on the basis of ref.67and discharge rates
and nutrient loading according to ref. 68. A Gaussian decay function
was used to estimate dispersal offshore, approaching zero at 2 km
(Supplementary Figs.13-15).This samedispersal function wasalsoused
for nutrient input, urban runoff, sediment input and rainfall, which are
each described below.
v
Nutrient Input. We calculated nutrient input (kg ha ' yr ') at 100 m
resolution as the combination of total nitrogen from OSDS (Waste-
water pollution section above) and golf courses. The total golf course
area per watershed was derived from NOAA Coastal Change Analysis
Program (CCAP) land -use and land -cover data and Landsat cloud -free
composite Images created with Google Earth Engine. The golf course
area was multiplied byanannual nitrogen application rateof585 kg ha`
(refs. 69,70)and then by a leaching rate of 32%"-" to estimate nitrogen
that either runs off or reaches the groundwater. We also imposed a
reduction in nitrogen that reached the ocean on the basis of distance
inland and used subwatershed catchment data" to estimate nutrient
transport from golf courses to the coastline (Supplementary Figs.16-18).
Urban runoff. We quantified the total area of impervious surfaces
(that is, paved roads, parking lots, sidewalks and roofs) within 10 km
of the coastlineat 100 m resolution foreach year from 2000 to 2017
(Supplementary Figs.19 and 20). Data were extracted from NOAA CCAP
land -use land -cover data from 1992, 2001,2005and 2010. We also digi-
tized 2017 impervious surface cover from a single cloud -free Landsat
8 image (courtesyof the United States Geological Survey, USGS) (15 m
resolution pan -sharpened). Years in between data availability were
filled in by linear interpolation.
Rainfall. We quantified annual rainfall (mt ha') and peak rainfall (maxi-
mum 3-day rainfall total, m'ha ') at100 m resolution. Daily rainfall data
were generated following refs. 75,76. Rainfall from each rain station
was used to derive interpolated surfaces at annual time steps using
Empirical Bayesian Kriging inArcGIS. subwatershed catchment data'4
were clipped to 0-10 km from the coast and used to calculate rainfall
per drainage area (Supplementary Figs. 21 and 22).
Sediment input. The Integrated Valuation of Ecosystem Services
and Tradeoffs sediment delivery model was used to derive long-term
annual averagesedimentinput(kg ha') reaching the coast""at lA0 m
resolution. We then modulated the long-term annual average sedi-
ment overtimeby watershed on thebasis of discharge calculated from
peak rainfall data (Rainfall section above). Discharge bywatershed was
calculated following ref.81. Sediment load was assumed to scale with
discharge according to a approximate ratings curve following ref.82
(Supplementary Figs. 23 and 24).
Fishing gear restrictions. We created a categorical value for local fish-
ing gear restrictions us! ng regulation information and marine managed
area bou ndary designatio ns updated from ref. 80. All regul atio ns were
evaluated for pr6hibition of gear categories in relation to fishing for
reef finfish species over time: line fishing, lay nets, spear fishing and
aquarium collection. Ranked fishing gear categories are as follows:
(1) full no -take, (2) no lay net, spear or aquarium, (3) no lay net oraquarium,
(4) no lay net, (5) no aquarium and (6) open to all gear types (Su pple-
mentaryTable 1and Supplementary Fig.25).
Sea surface temperature and heat stress. The mean and variability
(that is, standard deviation) In summertime sea surface temperature
(SST) was calculated over a 90-day window centred on the maximum
value of a 7-day moving window average For each SST pixel (Supple-
mentary Fig.26). Mean regional temperature (Fig.3a) wascalculated by
takingthe 7-day runningmean ofdailyvaluesand then averaging across
all coastal pixels within our study region. Heat stress on reefs during
the 2015 marine heatwave was assessed using DH W35, a widely used
metric by coral reef scientists across the world. All data were NOAKs
Coral Reef Watch v.3.1, available daily at5 km resolution15.
Phytoplankton biomass and irradiance. We used satellite derived
chlorophyll-a(mg m'; a proxy for p hytoplankto n biomass) and irradi-
ance (E m ' d-') from two sources. The long-term mean (2002-2013) in
8-day, 4 km data were obtained from ref. 80 and shown in Fig. 2d and
Extended Data Fig.3. All subsequent analysis used the visible -infrared
imaging radiometer suite, which has high spatial (750 m) and tempo-
ral (daily) resolution data starting in 2014 (provided by NOAXs Coral
Reef Watch). All data were quality controlled and masked to account
for cloud cover (Supplementary Information) and optically shallow
waters following ref.83 (Supplementary Fig.27).
Wave exposure. Wave power (kW m ') combines wave height and
period and provides a more representative metric of wave exposure
than wave heightaloneB4. A series of nestedgrids (fromglobal to50 m)
using WAVE WATCH III65 and Simulating Waves Nearshore86 were used
to quantify wave transformation over the reef environment at 50 m,
at hourly intervals across our study region From ref. 87 and updated
for this study. Annual data were then generated for each 50 m grid
cell by taking the mean of the top 97.5%in daily maximum wave power
(Supplementary Fig. 28).
Depth. Depth of the reef Floor (m) was measured using diver depth
gauges during the in -water reef surveys.
Statistical analyses
Coral reef trajectories predisturbance. We quantified the change
in coral cover at 23 reefs from 2003 to 2014. A reef was considered to
have a positive trajectory or negative trajectory if coral cover from
the 2003 survey to the 2014 survey increased or decreased by greater
than 3%, respectively (Fig.2b). This cut-off was based on the range in
mean coral cover among all 23 reefs across the 12-year period (range
2.8%; minimum 34.1%; maximum of 36.9%). We then quantified local
human impacts and environmental factors at each reef as follows:
fish biomass metrics were from the mean of all annual surveys for
each year from 2003 to 2014; human population, wastewater pol-
lution, nutrient loading, urban runoff, annual rainfall, peak rainfall,
SST mean and SST variability from the mean of all data from 2000 to
2014. Phytoplankton biomass and irradiance were from the maximum
monthly climatology from2002 to2013. Sediment and wave exposure
came from the mean of the top five events from each year spanning
2000-2014. Fishing gear restrictions were from marine managed
area designation at theonset of reef surveys and the depth came from
in -water diver -assessed values.
The difference in local human impacts and environmental factors
between positive and negative trajectory reefswere then calculated as
the difference in the mean drop -one Jackknife values for each impact
or factor. Upperand lower bars in Fig.2d represent the respective
maximum and minimum differences in drop -one jackknife values
between positive and negative trajectory reefs. Before calculating
the drop -one jackknife values, we identified and removed outliers
that fell outside a threshold of±2 standard deviations of the median.
We formally tested for a difference in the local conditions of positive
versus negative trajectory reefs using a multivariate permutational
analysis of variance (PERMANOVA)89 based on a Euclidean distance
similarity matrix, type III (partial) sums -of -squares and unrestricted
permutationsof the normalized data. Wevisualized theresults in Fig.2c
using a constrained analysis of principal coordinates90 and calculated
the cross -validation allocation success (a measure of group distinct-
ness) from the leave -one -out procedure of the constrained analysis of
principal coordinates analysis.
Coral response to the 2015 marine heatwave. Our goal was to as-
sess the local land -sea human impacts and environmental factors
that best explained changes in coral cover as a consequence ofthe
2015 marine heatwave. Any potential to observe change, however,
could be influenced by variations in starting condition. Reefs with
higher initial cover (such as those on positive coral cover trajectories
predisturbance, Fig. 2b) had greater scope for loss and vice versa91
(Extended Data Fig.5). To account for this and ensure comparability
across reefs (Supplementary Fig.4) we calculated coral cover change
following ref. 92 as:
%differenceA=[(A,.1—AbJ)/Ab,1] x 100
whereAb andA, are the mean coral cover values at each reef in 2014 or
2015, and 2016, respectively.
We then calculated the following predictors based on current litera-
ture and our hypotheses of the principal factors that drive changes in
coral cover owing to severe heat stress (Extended Data Table 1). Fish
biomass metrics included the mean of fish data that were coupled
with benthic surveys: 2014 in = 40) or 2015 (n =40) and 2016 (n = 80);
human population, wastewater pollution, nutrient loading,urban run-
off, annual rainfall, peak rainfall and wave exposure were taken from
the mean of all data From 2012 to 2016, sediment was measured from
the mean of the top three events from 2006 to 2016; SST mean and
SST variability were taken from the mean from 2000 to 2014; DH W
was the maximum value for 2015; phytoplankton biomass and irra-
diance was the mean from June to November 2015, representing the
time inclusive of the marine heatwave; Fishing gear restrictions was
the marine managed area designation before the marine heatwave
(2014 or 2015, depending on the reef surveyed) and depth came from
in -water diver -assessed values.
We then tested for correlations between coral loss and our suite of
predictor variables using a generalized additive mixed -effects mod-
elling (GAMM) framework' with the gamm4 (ref. 93) package for R
(www.r-project.org) v.4.0.2. Before model fitting, we identified the
presence of outliers in our predictor variables as any point that fell
outside a threshold of±2 standard deviations of the median. We then
applied an additional step to retain any point above this threshold that
was within 25%of the maximum predictorvalue below thethreshold.
This ensured that no data points were unnecessarily discarded from
our formal model -fitting process because of applying an arbitrary
threshold cut-off for data inclusion. The following predictors were
square -root transformed to down -weight the influenceof valuesatthe
extremeends oftheir distributions: all fish biomass metrics, wastewater
Article
pollution, urban runoff, nutrient loading, phytoplankton biomass and
peak rainfall. A fourth -root transformation was applied to sediment.
To reduce model overfitting, Pearson's correlation coefficientswere
calculated among all predictors (Supplementary Fig.5), removing
one of each pair of highly correlated (r> 0.7) predictors. To further
strive for model parsimony, we a priori excluded human population
density from the model -fitting process as it was a poor indicator of
human -driven land -to -sea impacts on local scales (Figs.lc and 2d and
Extended Data Fig. 3). We also excluded browser biomass as they rep-
resented less than 10%on average oftotal herbivore biomassacross all
reefsbefore, during and postdisturbance.This resulted isthe following
predictors included in the models (correlated predictors in parentheses
were removed): total fish biomass, biomass of scrapers, biomass of
grazers (total herbivore biomass), DHW (SST mean and variability),
wastewater pollution, nutrient input, urban runoff,sediment input and
peak rainfall (annual rainfall correlated with both), wave power, phy-
toplankton biomass (irradiance), fishing gear restrictions and depth.
The decision of which correlated predictors to retain was based on a
hypothesis -driven approach, In part whether the given predictor had
the potential to directly (for example, sediment input) rather than
indirectly (for example, annual rainfall drivingsediment input) affect
heat -driven coral loss.
We incorporated a random spatial factor to account for the possible
influence of a change in an underlyingvariable alongthe coastline not
quantified in this study. This was done by breaking the coastline up
into discrete SO km sections running north to south. Section sizewas
determined using hierarchical clustering based on pairwise Euclid-
ean distances between reefs and identifying an inflection point in the
intragroup variance' (Supplementary Fig. 7). We fitted GAMMs for
all possible candidate models (unique combinations of the predictor
variables) using the UGamm wrapper function, in combination with
the dredge function in the MuMIn package". Nonlinear smoothness in
the models was determined using penalized cubic regression splines,
with the number of knots (limited to fourto reduce overfitting) spread
evenly throughout each covariate. All possible candidate models were
computed (unique combinations of the predictor variables) but limit-
ing the total number of predictors in anygiven candidate model to five
to reduce overfitting. We used Akaike's information criterionwith a bias
correction for small sample sizes95 (AICc) for model comparison and
all models within AAICc s 2 of the top model (AAICc = 0) are presented
In Extended Data Table 2. To visualize the effect of predictor terms on
coral cover change,we averaged the coefficients from the top models
(that is, AAICc < 2) to generate a predicted dataset and set all other
predictor terms to their median value. Finally, we calculated a meas-
ure of predictor variable relative importance within each candidate
model by calculatingthe sum of AICc model weights for each predictor
(that is, the sum of model weights across all models containing each
predictor; Fig. 3).
Coral reefs four years postdisturbance. Our goal was to assess the
local land -sea human impacts and environmental factors that best
explained variations in thecover of reef-buildingorganisms fouryears
following the marine heatwave. The cover of reef -building organisms
for reefs surveyed in 2019 in =55) were parsed into three categories
on the basis of the following percentiles: low, less than or equal to the
25th; moderate, morethan 25th and less than 75th; and high, more than
or equal to the 75th. We then performed ordinal logistic regression96
to determine the probability of a given reef having high, moderate or
low coverof reef-buildingorganisms on the basis of the prevailing local
human impacts and environmental factors (that is, predictorvariables;
Extended Data Table 1). Logit models are multivariate extensions of
generalized linear regression models that provide parameterestimates
by means of maximum likelihood estimation (MLE) to model the rela-
tive log odds of observing a reef -builder cover category or less versus
observing the remaining higher categories:
(PO', P
In l P(Yt 5>j)=C.l+Btzit+... +Bkzit
Here,1 indexes each of observations, with categoriesy„ and the
left-hand side of the equation is the logit of the probability of a
reef -builder category ofj or lower, fo rj=1(high) or 2 (moderate). Reefs
with low reef -builder covercontributed to the regression through cal-
culation of the log odds. Each Cjis an MLE-computed model intercept,
and each B,, is the MLE coefficientcorresponding to the standardized
independent variable za, for k =1 through n, where n is the va riable
number of predictors used in a given candidate model, hence the
ellipsis( ... ). A fundamental component of this model isthe assumption
of proportional odds, or parallel regression, which indicates that B,
values are independent of the logit level j. The validity of this parallel
regression assumption was ascertained using Brant's Wald tes07, as
well as a likelihood ratio test (a= 0.05).
We then calculated the following predictors based on current litera-
ture and our hypotheses of the principal factors that drive changes in
reef -builder cover across space and time following a major thermal
disturbance: fish biomass metrics, wastewater pollution, nutrient
loading, urban runoff, annual rainfall, peak rainfall, wave exposure,
phytoplankton biomass and irradiance: the mean of all data from 2016
to 2019; sediment was measured as the mean of top three events over
the 2006-2019 time period; SST mean and SST variability: mean of all
data from 2000 to 2018. Notethat2019 was excluded in SST mean and
SSTvariabilityowingtothemarine heatwavethataffectedHawai'i21 but
occurred after our 2019 fish and benthic surveys; fishing gear restric-
tions involved the marine managed area designation in 2016 and depth
was assessed by in -water diver -assessed values.
We used thesame process as in theGAMM analysis to remove outliers
in our predictor variables (above). We then square -root transformed
the following predictors to down -weight the influence of values at the
extreme ends of their distributions: total fish biomass, wastewater
pollution, sediment input and nutrient loading. Pearson's correlation
coefficients were calculated among all predictors (Supplementary
Fig. 8), removing highly correlated (r> 0.7) predictors. For the rea-
sons outlined in our GAMM analysis and for continuity, we a priori
excluded human population density and the biomass ofbrowsers from
the model -fitting process. This resulted in the following predictors
included in the models (correlated predictors in parentheses were
removed): total fish biomass, biomass ofscrapers, biomass of grazers
(total herbivore biomass), wastewater pollution, nutrient input, sedi-
ment input, urban runoff (phytoplankton biomass), wave exposure,
fishing gear restrictions and depth. The decision of which correlated
predictors to retain followed the same logic as our GAMM analysis. The
mean and variability in SST were excluded given the negligible range
of values among reefs (0.1 and 0.025 °C, respectively). All possible
candidate models were computed while limiting the total number of
predictors in anygiven candidate model to four (to reduce overfitting
and to account for the lower response variable replication compared
to our GAMM analysis). Models were computed using the multino-
mial logistic regression function mnrfit in MATLAB. We again used
AICc for model comparison and all models within AAICc s 2 of the top
model (AAICc= 0) arepresented in Extended Data Table 3. McFadden's
pseudo-Rrwas computed for the highest ranked models and ranged
from 0.21to 0.22. Unlike trad itional Revalues, McFadden's pseudo-R'
of more than 0.2 represents an excellent fit98. Modelswithin AAICc 52
ofmodel I in Extended Data Table3 demonstrated comparable levels of
goodness of fit and parsimony""'. Many of the parameter coefficients
within these models were sensitive to the underlyingvariabllity in the
data and their estimates did notdiffersignificantly from zero (P <0.05).
The top model contained parameterswith covariate estimates signifi-
cantly different from zero, namely scraper biomass and wastewater
pollution. Using model1, we examined changes in the probability of a
given reef having high (morethan orequal to the75th percentile), mod-
erate (morethan the 25th and less than the75th percentile) or low(less
than orequal to the 25th percentile) reef-buildercover (Fig.4a) on the
basis of variations in these two land -sea predictors (Fig.4b). Probabil-
ity curves for high, moderate and low were calculated on the basis of
changing scraper biomass and wastewater pollution and holding all
other predictors at their mean.
Resource management scenarios. The resource management sce-
narios presented in Fig.4b were selected on the basis of the Following
rationale. We chose 250 kg ha' as the management target for scraper
biomass as this value approximates the long-term mean (2003-2019;
n =17) biomassofscraperswithin Kealakekua Bay, a marine protected
area where no Fishing has been allowed since 1969 (Supplementary
Fig.10). Kealakekua Bay isalso exposed to numerousland-based stress -
ors, includinghigh levels ofwastewaterpollution (258,0001 h-' in 2019).
As such, our value of250 kg ha 'represents an estimate of scraper bio-
mass on a reef with strong fisheries protection but with land -based
stressors present. In addition, we compared our upper (250 kg ha-')
and lower (30 kg ha') scraper biomass values to the distribution of
scraper biomass among all reefs (n = 80) in 2019, the mostrecent time
point In which all reefs were surveyed within the same year (Supple-
mentary Fig. 10). The upper and lower limits represent the 92nd and
36th percentiles, respectively. For wastewater pollution, we used our
2019,100 m grid cell values that fell along the 10 m isobath (same as
Fig. lc) but constrained the latitudinal extent to be consistent with the
northern- and southern -most locations of the 2019 reef surveys. This
approach provided Far greater replication and a more representative
assessment of wastewater pollution along the coastline for which to
assess our management scenarios. The upper (600,0001 ha') and
lower(2,500I h I) values chosen forwastewater pollution represented
the 95th and 36th percentiles of the 2019 distribution, respectively
(Supplementary Fig.11).
Reporting summary
Further information on research design is available in the Nature
Portfolio Reporting Summary linked to this article.
Data availability
All data that support the findings of this study are available at https://
github.com/jamisongove/Coral-Reef-Persistence. Reef fish length -
weight parameters were obtained from FishBase (https://fishbase.
org) and ref.56, human population data from NASA Gridded Popu-
lation of the World v.4 (https://sedac.ciesin.columbJa.edu/data/
set/gpw-v4-population-count-revll), land -use and land -cover data
from the NOAA Coastal Change Analysis Program (https://www.
coast.noaa.gov/htdata/rasterl/landcover/bulkdownload/), soils
data from USDA Gridded Soil Survey Geographic Database (gSSURGO;
https://www.nres.0 sda.gov/resources/data-and-reports/gridded-sol1-
survey-geographic-gssurgo-database),subwatershed catchment
data from USGS Stream Stats (https://water.usgs.gov/GIS/metadata/
usgswrd/XML/ds680_archydrohucs.xml) ",watershedanddigitaleleva-
tion model data from USGS National Hydrography Dataset (https://
www.0 sgs.gov/national-hydrography/nati onal-hydrography-dataset),
rainfall data from refs. 75,76, Landsat 8 satellite Image from USGS
(h[tps://earthexplorer.usgs.gov/), Landsat 7 and 8 cloud -free com-
posites derived usingGoogle Earth Engine (https://earthengine.google.
com/), individual wastewater systems for Hawaii from refs.101,102,
marine managed area designation from ref. 80 and downloadable
from the State of Hawai'i (https://planning.hawall.gov/gis), fish-
ing regulations from the State of Hawaii (https://dlnr.hawaii.gov/
dar/fishing/fishing-regulations/), SST and DHW data from NOAA
Coral Reef Watch (https://coralreefwatch.noaa.gov/product/Skm),
ocean colour (chlorophyll -a and irradiance) data from NOAA Coral
Reef Watch (https://coraireefwatch.noaa.gov/product/oc/index.
php) and ref. 80. See Methods and Supplementary Information for
more detailed information on the data used to support the findings of
this study.
Code availability
Statistical analyses were performed using the software packages R
(www.r-project.org) v.4.0.2 (using libraries gamm4, MuMin, foreach,
doMC, ggplot2, gmt, tidyverse, zoo and lubridate)103, MATLAB (www.
mathworks.com) using v.2021a (using Statistics and Machine Learn-
ing toolbox), ArcGIS Desktop (www.esri.com) v.10.6 with Advanced
licensing and extensions Spatial Analyst and Geostatistical Analyst,
Integrated Valuation of Ecosystem Services and Tradeoffs Sediment
Delivery Ratio model (https://naturaleapitalproject.stanford.edu/
software/invest) and the PERMANOVA+(ref.89) add -on for Primerv.7
(ref.104). Code is available for download at https://github.com/
ja mi songove/Coral-Reef-Persistence.
55. Jokiel, P. L,Brown, E. K., Friedlander, A., Rodgers, S. K. U. & Smith, W. R. Hawaii coral at
assessment and monitoring program: spatial pattemsand temporal dynamics in reef
coral communities. Pas. Sci. 58, 159-174 (2004).
56. Donavan. M. K. at at. Combining fish and benthic communities into multiple regimes
reveals complex reef dynamics. Sci. Rep. 8,16943 (2018).
57. McClanahan, T. R. at at, Prioritizing key resilience lndcalors to support coral reef
management in a changing climate. PLoS ONE 7, e42884(2012).
58. ClnnegJ.E.at at. Bright spots among the world's coral reefs. Nature 535,416-419(2016).
59. Cinner,J.E.atat. Gravity of human impacts mediates coral reef conservation gains. Proc.
NatfAcad. Sci. USA 115, E6116-E6125 (2018). '
60. Kittinger, J. N. at al. From reef to table: social and ecological factors affecting coral reef
fisheries, artissnal seafood supply chains, and seafood security. Kos ONE 10, e0123856
(2015).
61. Grafeld, S„Oleson,K.L.L., Teneva, L. & Kittinger, 1. N. Fallow that fish: uncovering the
hidden blue economy in coral reef fisheries. PLcS ONE 12. eOl 62104 (2017).
62. Delaney, D.G.at at. Patterns in ardsanalcoral reef fisheries revealed through loeal
monitoring efforts. Peerl 5, e40B9 (2017).
63. Heenan, A. & Williams. 1. D. Monitoring herbivorous fishes as indicators of coral reef
resilience in American Samoa. PLoS ONE 8, e79604 (2013).
64. Green, A. L. & Bellwood, D. R. Monitoring Functional Croups of Herbivorous Reef Fishes as
Indicators of Coral Reef Resilience: A Practical Guide for Coral Reef Managers in the Asia
Pacific Region (IUCN, 2009).
65. Freese, R. & Pauly. D. FishBase. a global information system on fishes. FishBase www.
fishbwe.og(2002).
66. CIESIN. Gridded population el the World, version 4(GPWv4). NASA Socioeconomic Data
andApplicedons Center (SEDAC) https.1/sedac.clesin.mtumbia&du/data/cogectiw/
gpww4(2016).
67. Whitimr, R. B. & Et-Kadi, A. 1. Human Health and Environmental Risk ofOnsife Sewage
Disposal Systems for the Hawaiian /stands oJ`K ..L Maui. Molokai, andHawaff(Hawaii
Department of Health, 2014).
68. Delevaux, J. M. S. at at A linked land -sea modeling framework to inform ridge -to -reef
management in high oceanic Islands. PLoS ONE 13. e0193230 (2018).
69. GCSAA. Nutrient Use and Management Practices on U.S. Gaff Courses. Golf Course
Environmental Profile Volume It (GCSAA, 2016).
70. Brosnan,).T.&Deputy, J. Bermudagrass. Turf Management https://sclmlarspacemanca.
hawaff edulserver/api/corelbitsU"e /a09dflab.e6O442b6 bd814197cad6d060/
content(2006).
71. Kunimamu,L,Bud..M.&Kawachi,T. Loading rates of nutrients discharging from a golf
course and neighboring forested basin. WaterSci. Tachnol. 39, 99-107 (1999).
72. Wong, J. W. C.Chan, C.W.Y.&Cheung,K.C. Nitrogen and phosphorus leeching from
fertilizer applied on golf course: lysimeter study. Water, Air, Soft Pollut. 107, 335-345
(1998).
73. Shuman, L.M. Phosphate and nitrate movement through simulated golf greens.Water,
Air, Soil Pollut. 129, 305-31B (2G01).
74. Res, A. & Skinner, K. D. Geospatialdatasets for watershed delineation and
characterization used in the Hawari StreamStats web application. US Gael. Sum.. Data
Sec 680,12 (2012).
75. Longman, R.J. at al. Compilation of climate data from heterogeneous networks across
the Hawaiian Islands. Sci Data 5,180012 (2018).
76. Longman, R. J., Newman, A. I., Giambelluca, T. W.& Lucas, M. Characterizing the
uncertainty and assessing the value of gap filled daily rainfall data in Hawayi. J.APPL Met.
CUM. 59, 1261-1276(2020).
77. Sharp, R. at aL In VEST Version 3.2.0 User's Guide (The Natural Capital PmiecL The Nature
Conservancy and Wald Wildlife Fund, 2015).
78 Hamet. P., Chaplin -Kramer, R., Sim, S.&Mueller, C. A new approach to modeling the
sediment retention service(InVEST 3.0): case study of the Cape Fear catchment. North
Carolina,USA.Sol. TotalEnviron.524-525, 166-177(2015).
79. Fallnski,K.A. Predicting Sediment Export into Tropical Coastal Ecosystems to Support
Ridge to Reef Management. Thesis, Univ. Hawaii at Marva (2016).
80. Wedding, L. M. at al. Advancing the integration of spatial dam to map human and natural
drivers on coral reefs. PLoS ONE 13, e0189792 (2018).
Article
81. Natural Resources ConservationService. National Engineering Handbook -Part 630
HydrologyCh. 10 (US Department of Agriculture, 2004).
82. Glysson,G.D.Sediment.Tamporl Curves Report No. 87-218(USG5, 1987).
83. Gov.. 1. M. etaL Quantifying otmat logicatranges and anomalies for Pacific coral reef
ecosystems. PLoS ONES, e61914(2013).
84. Gove,J.M. at at Coral reef benthlc regimes exhibit nom8near threshold responses to
natural physicalddvers.Mar. Ecol. Prog. Ser. 522,33-48(2015).
85. Tolman, H.L. A mosaic approach to wind wave modeling. Ocean Model Onitne25,35-47
(2008).
86. Booil,N.,Ris,R.C.&Holthuijsen,L.H.Athird-generation wave model for coastalregions:
1. model description and vatidation.l. Geophys. Res. (Oceans) 104,7649-7666 (1999).
87. U. N. etal. Thirty-four years of Hawail wave hindcastfrom downscaling of climate
forecastsystem reanalysis. Ocean Model. Online 100, 78-95 (2016).
88. Gmham,N.A.).at at. changing role of coral reef marine reserves in a warming climate.
Nat Commun. 11, 2000 (2020).
89. Anderson, M., Gutsy, R. N.& Clanke, R. K. Permanova. for Primer: Guide to Software end
Statistical Methods (Pr1mer-E Limited. 2008).
90. And .... n,M.J.&WiNs,T.1. Carwnicat analysis of principal coordinates: a useful method
of constrained ordination for ecology. Ecology B4, 511-525 (2003).
91. Cote,I.M., Gilt 1.A.,Gardner, IA.&Watkinson, A. R. Measuring carat reef decline
through meta -analyses. Philos. Trans. R. Soc. Ser. B: Blot. Sci. 360, 385-395 (2005).
92. Graham, N. A. J. at at. Climate warming, marine protected areas and the ocean -scale
integrity of coral reef ecosystems. PLOS ONE 3, e3039(2008).
93. Wood, S., Scheipl F.&Wood, M. S.Package'gamm4'. Am. Stat. 45,0.2-5(2017).
94. R Core Team. MUMIn:Multi-Model Inference R package v.1.13.4(R Foundation for Statistical
Computing. 2015).
95. Hurvich, C. M.&Tsai C.-L. Regression and time series modelselection in smalisamples.
Siometrika 76, 297-307 (1989).
96. Safaie,A.et at. High hequencytemperature variability reduces the riskofcoralbleaching.
Nat Commun. 9,1671(2018).
97. Brant R. Assessing proportionality in the proponianalodds model for ordinal logistic
regression. S/ometrics 46,1171-1178 (1990).
98. McFadden, D. Conditional Login Analysis of Qualitative Choice Behavior(Academic Press,
1974).
99. Burnham, K. P.& Anderson, D. R. (ads) Model selection and multimodel inference.
A practical informatiomtheoredc approach (Springer, 1998).
100. Richards, S. A. Testing ecological theory using the informatiomiheorstiwl approach:
examples and cautionary results. Ecology 86, 2805-2814 (2005).
101. Individual Wastewater System Database(Hawai'i Department of Health, 2017).
102. Underground Injection Control Permit Application Files(Hawail Department of Health,
2017).
103. RCore Team. R; A Language and Environment for Statistical Computing. R Foundation for
Statistical Computing https.1/www.R.projec rg/(2021).
104. Clarke, K. & Gorley, R. Getting started with PRIMER 0. PRIMER.E: Plymouth (Plymouth
Marine Laboratory, 2015).
105. Beyer. H. L. at at. Risk -sensitive planning for conserving coral reefs under rapid climate
change. Conserv. Lett 11, e12587 (2018).
106. Darting,E.S.etaLSocial-environmentaldriverainformstrategicmanagementofcoral
reefs in the Anthropocene. Nat. Scot. Evol. 3,1341-1350 (2019).
107. Andrelto,M. at at A global map of human pressures on tropical coral reefs. Conserv. Lett.
15, e12858(2022).
108. Adam, T.C.,Burkepite,D.E.,Ruttenberg,B.1.&Paddack, M. J. Herbivary and the resilience
of Caribbean cmalmefs: knowledge gaps and impticationsfor management. Mar. Ecol.
Prog. Ser. 52Q 1-20 (2015).
109. William% 1. D. at at. Human, oceanographic and habitat drivers of Central and Western
Pacific coral reef fish assemblages. PI cS ONE 10, e0120516 (2015).
110. Johnson, J. V., Dick, J. T. A. & Pincheira Donosa,D. Local anthropogenic stress does not
exacerbate coral bleaching under global climate change. Global Ecol. Biogeogr. 31,.
1228-1236(2022).
111. Wear, S. L.& Thurber, R. V. Sewage Pollution: mitigation is key for carat reef stewardship.
Ann. N. Y. Aced. Sol. 1355,15-30 (2015).
112. JoMet P.L., Hunter. C.L..Taguchi, S.&Watorai, L. Ecological impact of a freshwater'reef
kftr in Kaneohe Bay, Oahu, Hawaii. Coral Reefs 12, 177-184 (1993).
113. Rodgers, Ku. S. at al. Impact to coral reef populations at Hasna and Pile's, KauaL following
a record 2018 freshwater flood event. Diversity t3, 66 (2021).
114. Dollar, S.J. Wave stress and coral community structure in Hawaii. Coral Reefs 1, 71-BI
(1982).
115. Start ..I,C.D., Brown, E.K.,Paid, M.E..Rodgers, K.& Jokiet P. L. A model for wave
control on coral breakage and species distribution in the Hawaiian Islands. Coral Reefs
24,43-55(2005).
116. Chassm,E. at at Global marine primary production constrains fisheries catches. Ecol.
Left. 13.495-505 (2010).
117. Duane, C. & Cebrian, 1. The fate of marineautotrophic production. Nmnol. Oceanogr. 41,
1758-1766 (1996).
118. Weis, V.M. Cellular mechanisms of Crvdarian bleaching: stress causes the collapse of
symbiosis. J. Exp. Biol. 211, 3059-3065 (2008).
119. Gonaalez Espinosa, P. C. & Donner, S. D. Cloudiness reduces the bleaching response of
carat reefs exposed to heat stress. Glob. Change Blot 27, 3474,3486(2021).
120. MacNeil, M. A. at at Recovery potential of the world's coral reef fishes. Nature 520,
341-344 (2015).
Acknowledgements We thank A. Dillon of A5ce Designs for graphics support, the numerous
divers, boat drivers and support staff from the Hawaii Division of Aquatic Resources, National
Park Service and The Nature Conservancy for logistical and data collection support, individuals
from the Calf Course Superintendents Association of America, M. Johnson and R. Doog for their
input In determining the golf course nitrogen application rates, R. Longman for contributing
updated rainfall data, E. Darling for providing data layers from ref. 57, P. Neubauer and Y. Eynaud
for analytical advice, W. Walsh for being an catty supporter of this effort and 1. Link and
1. Samhouri forlheir review of an early version of the manuscript This research was supported
by NOAAS Integrated Ecosystem Assessment Program (contribution no. 2022 7), NOANs
Fisheries and The Environment Program, NOAA'e Pacific Islands Fisheries Science Centel and
grants from the NOAA Coral Reef Conservation Program (grant no. NAI6NOS4820059) and
National Marine Fisheries Service, Office of Habitat Conservation (grant nos. EA133F17SE1203
and NA17NMF4630301) and the Harold K.L. Castle Foundation.
Authorcontributional.M.G., GJ.W. and J.L. conceived the study, I.M.G., G.J.W. and J.L. developed
and implemented the analyses. J.M.G. and G.J.W.led the writing of the manuscript with GP.A.
and I.L. All other authors made substantive contributions to the manuscript and contributed
data that were central to this effort.
Competing Interests The authors declare no competing Interests.
Additi onat Informatlon
Supplementary Information The online version contains supplementary material available at
hops://dci.org/10.1038/s41586.023-06394 w.
Corresmridenceand raqumtsfor materials should be addressed to Jamison M. Gove or
Gareth J. Williams.
Peerreviewlnfarmation Nature thanks Joshua Cinneg Nicholas Graham, Peter Mumbyand the
other, anonymous, reviewer(.) fortheirconlribution tothe peer review of this work. Peer
reviewerreponsare available.
Reprints and Permissions information is available w hnp://www.nature.com/repdms.
a
b
.a
0 10 20 30 40 50
60
0 0.1 0.20.30.40.50.60.70.90.9 1
Coral Cover (% cover)
Management(cinmso OReswossio uis)
C
d
nm
w,
ge
o to 4o 60 80
100
0 1000 2000 3000 4000
Human Population (to'People)
Small -Scale Fisheries Market Gravity
0
f (number of people/(hours or tmveP))
0 25 50 75 1000
0 1000 2000 3000 4000 5000
Nutrient Input from Wastewater Pollution Sediment Input (103 kg km
9 (103 kg perwatershed)
In
an
/A
YN
0 5000 10000 150M 20000 25000 30WO 0 0.2 0.4 0.6 0.8 1
Tourism (tourist trips per year)
Cumulative Pressure
0 1a to 3o 40
50 0 sou two 1500 2000
Wave Exposure (kW m ')
Primary Production (mg C m ' day')
It
I
-1.5 -1 IS 0 0.5 1
1.5 -1.5 .1 -0.5 0 0.5 1 1.5
Historical Thermal Stress
Recent Thermal Stress
R& Global Mean :2SD
•- Hawal'I Mean . 2SD
Extended Data Fig.I I Comparison ofhuman, environmental,and climate
factors forreefsin Hawai'lWith coralreefecosystemsglobally.Dots represent
global (light grey) and Hawai'1(darkgrey) mean values. Error bars represent
the mean 12 standard deviation (SD). Factors presented are: a, Coral cover
(percent hard coral;global n= 2,584, Hawal'1 n =137); b, Proportional reef area
by country that is open (fished), gear restricted (restricted), or fully restricted
(no -rake) to frshing(sample number sameas Ina); c, Human population within
5 km in 2018 (global n=54,596; Hawai'I n =199); d, Small-scale Fisheries market
gravity (numberofpeople/(hours of travel)' represents human use and fishing
pressure related to the size and accessibility of coral reefs to nearby human
settlements and markets" (n=same as in c); e, Annual input of nitrogen (10'kg)
per watershed from wastewater pollution on coral reefs (global n=38,033;
Hawaii n= 324); f Sediment input (103kg km') tocoral reefs (n=sameasln c);
g, Annual number of tourist visits driven by coral reefs combining on -reef
(e.g., recreational diving and snorkeling) and reef -adjacent (e.g., provision of
calmwaters,sand beaches,views,and seafood) aspects(samplenumbersame
as Inc): h, Cumulative pressure score from stressors to coral reefsper5 km reef
containing pixel (unitless); 1, Mean wave energy,or wave power(kW m'), from
1979-2009 (n=same as in a); j, Mean primary productivity (mgC m'day')
between 2003-2013 in=sameas Ina); k-1, Unitlessmetric of (k) historical
(1985-2017) and (1) recent (2014-2017) thermal stress on coral reefs, whereby
positive values represent more desirable (i.e., less thermal stress) over the
respective time framesvo (n=same as Inc). The mean for reefs in Hawaii falls
within 2SD of the global mean For all factors. Data are from the following
sources: a,bJ J from'"a From"; c,d,f g,h,k, fram1oJ Factors presented here
were obtained from global drawers and will differ from those presented with in
our present study owing to the methodological differences as well as differences
in their spatiotemporal extent and resolution.
Article
6
1900 1905 1910 1915 1920 1925 19M 1935 1940 1945 1950 1955 1960 1965 1970 1975 1990 1985 1990 1995 2000 2005 2010 2015 20M
Year
Extended Data Fig.2I Long-term ocean temperature record for Hawai'I. time series. Data are from NOAA's Extended Reconstructed SST v5(https:H
Monthly sea surface temperature (SST) for the main Hawaiian Islands from www.ncel.noza.gov/products/land-based-station/noaa-globa1-temp)and
1900-2020. Dashed lines represent± 2standard deviations (SD) above and values shown are the 90'"percentile of monthly SST from within the vicinity of
belowthe longterm mean. Red line isthe 12•month moving average. The 2015 the main Hawaiian Islands(18.5 to 22.5°N;-160.5 to-154.5°W).
mari ne heatwave is represented by the highest SST values over the 120-year
ai y Qmd rm h6O �+ 3m^y Qm c ¢r Ile y
Extended Data Fig. 3 l Percentdifference in mean drop-onejackknife greater on reefs that had positive and negative trajectories, respectively.
valuesof local human lmpactsandenvironmental factorsbetween positive Zero line representsequal values.Outliers that fell outside athreshold oft2
and negative trajectory reefs. The percentdifference((VI-V2)/[(Vl+V2)/2]; standard deviations ofthe medlanwere removed priorto analysis. SeeFig.Ib
dots)wasquantified bytaking the ratioofthe mean indrop-onejackknife for reef locations and Fig. 2d formean absolute differences in factor values.
values between positive in = 10) and negative(n- 8) trajectory reefs(sensu)88. See Methods, Extended Data Tableland Supplementary information for
Upperand lowerbars represent the respective maximum and minimum per detailed information on local land•sea human impacts and environmental
cent differences. 6lueand red shaded regions indicate factors that were factors.
Article
Phytoplankton (mg m-')
Sediment Input (kg ha-')
Urban Runoff (m? he") Total Fish Biomass (kg he"
Extended Data Fig.41 Generalized AdditiveMixed Model (GAMM) results
(R2= 0.79) showing key local land -sea human impacts and environmental
factors that modif led coral response to the 2015 marine heatwave. Positive
and negative relationships reduce or Increase coral loss, respectively. Because
changes in coral cover following disturbance can be affected byvariations in
starting condition (reefswlth higher initial cover havegreater scope forloss,
and vice versa)91we modelled relative coral cover change following rer.92 to
ensurecomparabliltyacross reefs (see Methods). Median valueswithshaded
region representingthe 80%confidence interval. The relative importanceof
25
0 T
2,
so V
-75
0 40 160 31
Scrapers Biomass (kg ha")
factorsamongall models(i.e.,sum ofAICc model weights across all models
containingeach factor) was as follows: sediment input (0.99), scraper
biomass (0.99), total fish biomass (0.90), urban runoff (0.60), phytoplankton
biomass (0.38),wastewater pollution (0.28), peak rainfall (0.20), nutrient
loading(0.19),grazerbiomass(0.16), DHW (0.08),wave power(0.07), depth
(0.06), and fishinggear restrictions (0.05). See Extended Data Table lfor full list
oflocal land -sea human impacts and environmental factors included in the
analysis, Including those removed thatwere highly correlated (Fig. S5). See
Fig. S6 for predictor variable distributions.
G
a
2014 2016
b Mean Reef
Traleatorles s95% CI
Posit}tNe Negative
1 f
50 1
45 11
40
35
U`m 1f
30
0 25
U
20
1s
10
zola zols
C
.10
-15
?-20 Change
a -25
a
m -30
U .35
d
a -40
U L
oge
U -50fference)
-55
-60
d
0
.5
.10
.15
v
-20
U -25
A
o -30
U
-35
-40
e
c
m -10
i
v -20
0
m .30
a
C
U -40
o -50
U
� -so
U
5
Coral Cover 2015
• 0
•
15 20 25 30 35 40 45 50 55 Be
Coral Cover 2015
Extended DataFig.5 l Coralcoverchangefollowingthe2O15 marineheatwave both prior to, and to a lesser extent, followingthe marine heatwavecompared
on positive versus negativecoral covertrajectory reefs.a, Comlcover on to negative trajectory reefs.c, Positive trajectory reefsexperience increased
positive (blue;n=10)and negative (red;n = 8) trajectory reefs surveyed absolutecoralcover lossfollowingthe marine heatwave(underlyingrelationship
(see Fig. 2b in main manuscript)prior to(2014) and I -year following(2016) the shown in panel d), but this difference is largely absent once starting coral cover
marineheatwave.b, Positive trajectoryreefshavea hlghermean coral cover condition is accounted for (underlying relationship shown in panel e).
0
Article
Extended Data Table 1 I Local land -sea human impacts and environmental factors considered for our analyses
Impactor
Metric
Units
Spatial
Temporal
Temporal
Data Source
Justification
Factor
Resolution
Resolution
Range
Abundant reef fish populations support
Total Biomass,
lei surveys,
See
reef -scale ecosystem functions such as
Fish
Herbivores,
kg ha''
25 in
per site, per
2003-2019
Supplemental
predation and nutrient release-.
Herbivorous fishes support ecosystem
Herbivorous
Biomass
Grazers, Browsers,
Scrapers
year
Information
and mitigate the negative effects
fleshy algae have on coral survival'-.
Human
Population Density
people/15
1 km
Annual
2000-2019
NASA GPWv4
Human population density is a widely used
for local human Impacts','°°•"^.
Population
km
proxy
High concentrations of toxins (e.g..
endocrine disruptors, pathogenic bacteria
and viruses, pharmaceuticals, and heavy
Mod�ad from
metals) are found in wastewater
Wastewater
Total Effluent
Lha"
loom
Annual
2000-2019
rercr
pollution"'. These toxins can drive
Pollution
higher incidence of coral disease, reduced
coral growth and reproduction, increased
cover of fleshy algae, and increased coral
bleaching and subsequent modalkyo.
See
Human -derived nutrient input can promote
Nutrient Input
Total Nitrogen
kg ha-'
100 in
Annual
2000-2019
Supplemental
rapid algal growth, (utcompeting corals
Information
and disrupting ecosystem functions.
Runoff can deliver a broad spectrum of
land -based contaminants (e.g., heavy
metals and household chemicals)that
Urban Runoff
Impervious
ma ha'
100 in
Annual
2000-2019
See
Supplemental
degrades nearshore water quality, Win
cascading effects on coral health',
Surfaces
Information
including the natural defence abilities of
corals and Increase the Ikelihood of
See
Sediment input can impede the
Sediment
Sediment
kg ha'
Igo m
Annual
200o-2019
Supplemental
photosynthetic capacity of corals and
Input
Information
reduce growth by burying coral colonies".
Large pulses of freshwater from storm
events can cause localised die -off of
nearshore corals, fish, and other reef -
Annual & Peak
ha'
100 m
Annual
2000-2g19
Modified from
m
associated organiss"•. Rain events can
Rainfall
Rainfall
ms
refs"-
also mobilise high levels d nutrients,
sediment, and land -based debris that
Impact nearshore water quality and coral
reef health",.
Gradients In wave exposure and
See
associated flow produce varying levels of
Wave
Wave Power
kW m''
50m
Hourly
2000-2019
Supplemental
disturbance that can play a major role In
Exposure
Information
determining coral reef community
patlemsxk' ells
Chlorophyll -a Is a widely used indicator for
changes In phytoplanktan production"s
that propagates through the food -web'^.
NOAA's Coral
Corals can supplement nutritional
Phytoplankto
Chlorophyll -a
mg m's
750 in&
Daily &
2002-2014
Reef Watch),
requirements through heeterchophic
n Biomass
4 km
9-day
and ref
feeding on zooplankton High levels of
chlorophyll -a are also indicative of poor
water quality that can have negative
outcomes for coraWb
Sea-surtace
Summattime Mean
Coral
The mean and variability In summertime
Temperature
& Standard
°C
5 km
Daily
200o-2019
R
Reef Welch
Reef
ocean temperature is a widely used metric
(SST)
yi
Deviation
for coral real reslliences'.
Degree heating week is the accumulation
of heat stress above the coral bleaching
Heat Stress
Degree Heating
°C-weeks
5km
Daily
200o-2018
NOAA's Coral
Reef Watch
threshold over a 12-week period as is the
Week
dominant metric in coral reef research to
guanfify heat stress on corals le.n.,-).
Irradiance
Photosynthetically
Einstein
750 m
Daily
2015-2019
NOAA's Coral
Excessive Ioadlance can cause light stress
Active Radiation
m'ad'
Reef Watch
that exacerbates coral bleaching"ane.
See
We use fishing gear restrictions as
Fishing Gear
Gear Rank
Categorical
NA
NA
2000-2019
Supplemental
metric to represent spatial fisheries
Restrictions
Infnrmatinn
manaoement'-.
Deeper reefs are often less Impacted by
Depth Depth Metres NA NA 2003-2019 Reef Surveys heal stress compared to these located in
shallower depths•-r%
See'Local land-seahuman impacts andenvironmenml factors'section in Supplementalinfwmadon fordetailed information oncalcutatirg each impact orfactoc including data collection
methods, date sources and ancillary date sets, and specific tools or software uugzecf0 l-.
Extended Data Table 2 I Summary of generalized additive mixed effects models (GAMM) relating coral response to the 2015
marine heatwave with local land -sea human impacts and environmental factors
Significant Land -Sea and Environmental Factors Log Likelihood AICc AAICc Adjusted Rz
Model 1. Phytoplankton Biomass, Urban Runoff,
Sediment Input, Scraper Biomass, Total Fish-302.98 638.34 0 0.79
Model 2. Urban Runoff, Sediment Input, Peak-303.93 640.25 1.9t 0.78
Rainfall, Scraper Biomass, Total Fish Biomass
The top twocxndidate models are shown. AICc, Akaike'a information criterion corrected for small sample size; AAICc, chance in Al Cc across the candidate models (AAICcs2 of the top model
are shown); Adjusted R', proportion of variation in the response variable explained by the candidate model.
Article
Extended Data Table 3 1 Summary of ordinal logistic regression (OLR) models relating the per cent cover of reef -builders
(hard coral + crustose ccraLline algae) four years following the 2015 marine heatwave to Local Land -sea human impacts and
environmental factors
Model Output
P
AICC AAICc McFadden'
pseudo-R
Predictors
Scraper
Wastewater
Sediment Input
Peak Rainfall '
Biomass
Pollution
v
Coefficients
0.689
-0.867'
0.159
0.547 103.28 0 0.22
0
p value
0.031
0.019
0.647
0.130
Predictors
Scraper
Wastewater
Urban Runoff
Sediment Input
Biomass
Pollution
N
a
Coefficients
0.705
-0.567
-0.348
0.330 104.72 1.44 0.21
0
p value
0.025
0.115
0.282
0.303
Predictors
Depth
Scraper
Wastewater
Peak Rainfall
Biomass
Pollution
n -
Coefficients
-0.508
0.540
-0.802
0.728 104.73 1.45 0.21
0
g
p value
0.109
0.088
0.022
0.038
Predictors
Scraper
Urban Runoff
Sediment Input
Peak Rainfall
Biomass
a
m
v
0
Coefficients
0.673
-0.655
-0.076
0.493 105.19 1.91 0.21
p value
0.033
0.039
0.813
0.162
AICc, Akaike's information criterion corrected for smalisample size; AAICc, change in AICc across the candidate modets(AAICc s2 of the top model are shown); McFaddenb pseudo-W
proportion of variation explained by candidate model(02-0.4 represent an exceifentfit)sa. Significant predictors at p<0.05 are in bold.
natureportfolio
Reporting Summa
Corresponding author(s): Jamison Gave, Gareth Williams
Last updated by author(s): Aug 7, 2023
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.
Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
® The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
® A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one -or two-sided
El ® Only common tests shouldbe describedsolely by name; describe more complex techniques in the Methods section.
® A description of all covariates tested
® A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
El ® AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
® Give P values as exact values whenever suitable.
®E] For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
® For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
® Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on storistics for bloloaists contains articles on many of the points above.
Software and code
Policy information about availability of computer code
Data collection All data and code that support the findings of this study are available at https://github.com/jamisongove/Coral-Reef-Persistence.
Data analysis Statistical analyses were performed using the software packages R (www.r-project.org) version 4.0.2 (using libraries gamm4, MUMIn, foreach,
doMC, ggplot2, gmt, tidyverse, zoo, lubricate) (ref. 1), Matlab (www.matthworks.com) using v2021a (using Statistics and Machine Learning
toolbox), ArcGIS Desktop (www.esri.com) v10.6 with Advanced licensing and extensions Spatial Analyst and Geostatistical Analyst, InVEST
Sediment Delivery Ratio model (https,.//naturaicapitalproject.stanford.edu/software/invest), and the PERMANOVA+ (ref. 2) add -on for Primer
version 7 (ref. 3).
1 Team, R. C. in R Roundation for Statistical Computing (2021).
2 Anderson, M., Gorley, R. N. & Clarke, R. K. Permanova+for primer: Guide to software and statistical methods (2008).
,3 Clarke, K. & Gorley, R. Getting started with PRIMER v7. PRIMER-E: Plymouth, Plymouth Marine Laboratory (2015).
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described In published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further Information
Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
-Accession codes, unique Identifiers, or web links for publicly available datasets
- A description of any restrictions on data availability
- For clinical datasets or third party data, please ensure that the statement adheres to our oo licy
All data that support the findings of this study are available at https-.//github.com/jamisongove/Coral-Reef-Persistence. Reef fish length -weight parameters were
obtained from FishBase (https://fishbase.org) and ref. 1, human population data from NASA Gridded Population of the World v4(https:Hsedac.clesin.columbia.edu/
data/set/gpw-v4-population-count-rev1l), land use and land cover data from the NOAA Coastal Change Analysis Program (https://www.coast.noaa.gov/htdata/
rasterl/landcover/bulkdownloadn, soils data from USDA Gridded Soil Survey Geographic Database (gSSURGO; https://www.nres.usda.gov/resources/data-and-
reports/gridded-soil-survey-geographic-gssurgo-database), sub -watershed catchment data from USGS Stream Stats (https://water.usgs.gov/GIS/metadata/usgswrd/
XML/ds680_archydrohucs.xml) (ref. 2), watershed and digital elevation model data from USGS National Hydrography Dataset (https://www.usgs.gov/national-
hydrography/national-hydrography-dataset); rainfall data from refs. 3,4, Landsat 8 satellite Image from USGS (https.,//earthexplorer.usgs.gov/), Landsat 7 and 8
cloud -free composites derived using Google Earth Engine (https://earthengine.google.com/), individual wastewater systems for Hawai'i from refs. 5,6, marine
managed area designation from ref. 80 and downloadable from the State of Hawaii(https:HPlanning.hawaii.gov/gis), fishing regulations from the State of Hawayi
(https://dinr.hawaii.gov/dar/fishing/fishing-regulations/), sea surface temperature and degree heating week data from NOAA Coral Reef Watch (https://
coraireefwatch.noaa.gov/product/5km), ocean color (chlorophyll -a and Irradiance) data from NOAA Coral Reef Watch (https://coralreefwatch.noaa.gov/product/oc/
index.php) and ref. 7. See Methods and Supplemental Information for more detailed Information on the data used to supportthe findings of this study.
1 Donovan, M. K. et al. Combining fish and benthic communities into multiple regimes reveals complex reef dynamics. Sol. Rep. 8, 16943 (2018).
2 Rea, A. & Skinner, K. D. Geospatial datasets for watershed delineation and characterization used in the Hawai StreamStats web application. US Geol. Surv. Data
Ser. 680, 12 (2012).
3 Longman, R. 1., Newman, A. J., Giambelluca, T. W. & Lucas, M. Characterizing the Uncertainty and Assessing the Value of Gap -Filled Daily Rainfall Data in Hawaii. 1.
Appl. Met. Clim. 59, 1261-1276 (2020).
4 Longman, R. J. et al. Compilation of climate data from heterogeneous networks across the Hawaiian Islands. Scientific Data 5, 180012 (2018).
5-DOH. Individual Wastewater System Database. Hawaii Dept. of Health (2017).
6 DOH. Underground Injection Control Permit application files. Hawaii Dept. of Health (2017).
7 Wedding, L. M. et al. Advancing the integration of spatial data to map human and natural drivers on coral reefs. PLoS ONE 13 (2018).
Human research participants
Policy information about studies involving human research oarticioants and Sex and Gender in Research
Reporting on sex and gender NOT APPLICABLE
Population characteristics NOT APPLICABLE
Recruitment
NOT APPLICABLE
Ethics oversight NOT APPLICABLE
Note that full information on the approval of the study protocol must also be provided in the manuscript.
Field -specific reporti
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences ❑ Behavioural & social sclences ® Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature com/documents/nr-reoortine-summary-fiat.odf
Ecological, evolutionary & environmental sciences study design
All studies must disclose on these points even when the disclosure is negative.
Study description The study tested the hypothesis that mitigating local human impacts facilitates coral reef persistence in the face of climate change -
induced disturbance, specifically mass coral bleaching. Our goal was to move beyond commonly used proxies of local human impacts
and generate spatially resolved data on specific land -sea human activities to identify actionable outcomes. This further allowed us to
quantify the effects mitigating either land- or sea -based human impacts in isolation or simultaneously had on the ability of key reef -
building organisms to recover post -disturbance. We achieved this by combining recurring in -water SCUBA surveys of coral reef
benthic and fish communities with a 20-year time series of land -sea human impacts and other environmental factors thought to drive
-coral reef ecosystem processes. Our�y Inc u e ree s across a broad range o eco1W1caI states, large spatlotemporal gradients In
land -sea human impacts and environmental factors, and which experienced the most severe marine heatwave on record in the
Hawaiian Islands.
Research sample We quantified changes in the per cent cover of major reef -building benthic groups (hard coral, crustose coralline algae) and related
these to concurrent changes in numerous land -sea human Impacts, including urban runoff, wastewater pollution, nutrient loading,
sediment input, and local restrictions on fishing gear types. Environmental factors Included peak and annual rainfall, wave exposure,
variability In ocean temperatures and heat stress, irradiance, and phytoplankton biomass. We also incorporated multiple fish biomass
metrics that represent the critical role reef fish play in maintaining coral reef ecosystem dynamics. All human impacts and
environmental factors were chosen based on prior evidence in the literature that they represent key drivers of reef ecosystem
processes and were quantified using a variety of modelled and satellite -derived data sources.
Sampling strategy Underwater visual surveys of shallow -water benthic and reef -fish assemblages were collated from the following three coral reef
ecosystem monitoring agencies to maximise spatial and temporal replication across the study region: State of Hawaii Division of
Aquatic Resources, National Park Service, and The Nature Conservancy. Each program conducted surveys using similar data collection
methods (see below) in shallow -water (<30 m) depths over hard -bottom substrate.
Data collection All coral reef surveys used a traditional 25 m belt-transect method. Benthic surveys used permanently marked pins to ensure the
same area of reef was surveyed overtime. High resolution photographs were collected via photoquadrats at 1 m intervals along 25 m
belt-transects (N = 26 photographs per transect). Thirty to fifty random points were overlaid on each photograph and the benthic
component under each point was identified to the lowest possible taxonomic level. Per cent cover of the major functional groups
were used in this analysis, namely hard coral, crustose coralline algae, macroalgae, and turf algae. All data were averaged among
each transect and then among all transects for each site (1-4 transects per site, per year, depending on the monitoring program).
Surveys of reef -fish assemblages were performed along the same permanently marked 25 m transects concurrently with benthic
surveys. In all surveys, fishes were identified to species, sized, and enumerated. To account for differences among programs in how
researchers surveyed reef fish, counts were calibrated using species and method specific adjustments previously developed for the
region.
Timing and spatial scale Underwater visual surveys of benthic assemblages were collated from three monitoring programs for the following years (number of
reefs surveyed are in parentheses): 2003 (23), 2007 (23), 2011(23), 2014 (40), 2015 (40), 2016 (80), 2017 (80), 2018 (15), 2019 (55).
All benthic surveys used permanently marked pins to ensure the same area of reef was surveyed over time. High resolution
photographs were collected via photoquadrats at 1 m intervals along 25 m belt-transects (N = 26 photographs per transect). Thirty to
fifty random points were overlaid on each photograph and the benthic component under each point was identified to the lowest
possible taxonomic level. Per cent cover of the major functional groups at each reef were used in this analysis, namely hard coral and
crustose coralline algae. Surveys of reef -fish assemblages were performed along the same permanently marked 25 m transects
concurrently with benthic surveys. However, reef fish surveys were performed more frequently (1-6 times per year from 2003 —
2019) than benthic surveys, depending on the reef location and monitoring program performing the surveys. In all surveys, fishes
were identified to species, sized, and enumerated. To account for differences among programs In how researchers surveyed reef fish,
counts were calibrated using species and method specific adjustments. The survey region spanned roughly 200 km of coastline on
the Island of Hawaii.
Data exclusions Fish species were excluded from fish biomass calculations according to life history characteristics that are not well captured with
visual surveys, including cryptic benthic species, nocturnal species, pelagic schooling species, and manta rays. We also accounted for
extreme observations of schooling species, which were defined by calculating the upper 99.9%of all individual observations,
resulting in 26 observations out of over 0.5 million, comprised of 11 species. The distribution of individual counts in the entire
database for those 11 species was then used to identify observations that fell above the 99.0%quantile of counts for each species
individually. These observations were adjusted to the 99.0%quantile for analysis.
Other data exclusions Include outliers in predictor variables (the local human Impacts and environmental factors). Within the section
"Coral reef trajectories pre -disturbance", prior to calculating per cent difference, we identified and removed outliers that fell outside
a threshold of t 2 standard deviations of the median. Within the section "Coral response to the 2015 Marine Heatwave", prior to
model fitting, we identified the presence of outliers in our predictor variables as any point that fell outside a threshold of i 2
standard deviations of the median. We then applied an additional step to retain any point above this threshold that was within 25%
of the maximum predictor value below the threshold. This ensured that no data points were unnecessarily discarded from our formal
model -fitting process because of applying an arbitrary threshold cutoff for data inclusion. We used the exact same process to identify
and remove outliers within the section "Coral reefs four years post -disturbance" prior to formal model fitting.
Reproducibility A description of the methodologies used is provided in the Methods and expanded on substantially for several of the human impact
and environmental factors In the Supplementary Information. The data and full code necessary to reproduce the findings are
available at https://github.com/jamisongove/Coral-Reef-Persistence
Randomization Survey sites were either randomly or haphazardly chosen by the various monitoring agencies Involved In data collection. Sites were
separated by a minimum distance (250 m) and transects within sites were also separated by a minimum distance (5 -10 m). To
minimise observer bias of fish counts, sizing calibration dives were conducted using fish models of known size at the beginning of
each field season. Observer crossover training was done using two observers side by side when possible. Benthic cover estimates
were quantified by randomly assigning 20 points to each image using post -hoc image analysis programs (Photogrid or Coral Point
Count with Excel Extensions) and identifying the benthic group to the lowest taxonomic rank under each point.
Blinding All in situ benthic and reef fish surveys were conducted prior to this research question being conceived. The divers carried out the
surveys for the most part without prior knowledge of the local human impacts and environmental factors for their respective survey
locations —we later quantified these for each reef location and time of survey, thus blinding in this respect was achieved. In some
cases, divers were aware of any local fishing restrictions in effect, but this was unavoidable as many of them specifically survey Inside
and outside of these zones
Did the study involve field work? ® yes n No
Field work, collection and transport
Field conditions Because of the nature of collecting underwater benthic information, field conditions must be relatively calm (i.e., low wind and wave
activity) with relatively good underwater visibility (i.e., > 5 m)._-
Location Our study site was Hawaii Island (19.55°N, 155.66°W), USA, which is the southeastern most island of the Hawaiian Archipelago,
located in the northern central Pacific. The western section has roughly 200 km of coastline that Is predominately oriented north to
south. The coastline contains the longest contiguous reef ecosystem In the main Hawaiian Islands and large gradients in human
population, local land -sea impacts, and environmental factors. The region represents an ideal study location for resolving the
interacting land -sea human impacts driving reef ecosystem change and coral trajectories following acute climate -driven disturbance.
All reefs included In this study were in shallow -water (depth < 30 m).
Access & Import/export All survey data were collected with the knowledge and consent of the State of Hawayi, which has legal jurisdiction of all waters from
0-3 nm of the shoreline. The director of the State of HawaYfs Division of Aquatic Resources, which is the managing agency of State
waters, contributed survey data and both a collaborator and coauthor on this manuscript.
Disturbance All surveys were performed by professional scientific divers that aim to minimise contact and disturbance of the reef. No coral reef
benthic or fish species were removed from their habitat as part of this effort
ReDortinl; for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental syst
n/a Involved in the study
®F-1 Antibodies
® Q Eukaryotic cell lines
®❑ Palaeontology and archaeology
®❑ Animals and other organisms
®Q Clinical data
®� Dual use research of concern
ms Methods
n/a Involved In the study
M ❑ ChIP-seq
®F] Flow cytometry
®F1 MRI-based neuroimaging