Developing multimetric indices for monitoring ecological restoration progress in salt marshes

Marine Pollution Bulletin 64 (2012) 820–835

Contents lists available at SciVerse ScienceDirect

Marine Pollution Bulletin

journal homepage: www.elsevier .com/locate /marpolbul

Developing multimetric indices for monitoring ecological restoration progressin salt marshes

O.C. Langman a,⇑,1, J.A. Hale a,1, C.D. Cormack a,1, M.J. Risk b, S.P. Madon c,2

a Pandion Technology Ltd., Suite #601, 6th Floor, 28th October Street, Limassol 4065, Cyprusb PO Box 1195, Durham ON, Canada N0G 1R0c CH2M HILL, Water Resources and Ecosystems Management, Ecosystem Planning and Restoration, 402 W. Broadway, Suite 1450, San Diego, CA 92101, USA

a r t i c l e i n f o

Keywords:Multimetric indexSalt marshEcological restorationEnd states of restorationRestoration trajectoryAdaptive management

0025-326X/$ - see front matter � 2012 Elsevier Ltd.doi:10.1016/j.marpolbul.2012.01.030

⇑ Corresponding author. Tel.: +1 608 3207761; fax:E-mail address: [email protected] (O.C. L

1 Tel.: +1 803 5135649; fax: +1 803 2546445.2 Tel.: +1 619 6870120x37233.

a b s t r a c t

Effective tools for monitoring the status of ecological restoration projects are critical for the managementof restoration programs. Such tools must integrate disparate data comprised of multiple variables thatdescribe restoration status, including the condition of environmental stressors, landscape connectivity,ecosystem resilience, and ecological structure and function, while communicating these concepts effec-tively to a wide range of stakeholders. In this paper we describe the process of constructing multimetricindices (MMIs) for monitoring restoration status for restoration projects currently underway on the east-ern coast of Saudi Arabia. During this process, an initial suite of measurements is filtered for response andsensitivity to ecosystem stressors, eliminating measurements that provide little information and reduc-ing future monitoring efforts. The retained measurements are rescaled into comparable domain metricsand assembled into MMIs. The MMIs are presented in terms of established restoration theories, includingrestoration trajectory and restoration endpoint targets.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Degradation of aquatic ecosystems, primarily from anthropo-genic activities, has led to major efforts to regenerate, rehabilitate,or convert ecosystems towards a more desirable configuration (Na-tional Research Council, 1992). The motivation behind restorationprojects is the restoration of ecological services and functions,which impact a wide range of stakeholders beyond restoration sci-entists that will need to be informed of the progress of the project.Despite the large body of theory that supports the developmentand design of restoration projects, it has been pointed out (Jonesand Schmitz, 2009; Reeves et al., 1991;Roni et al., 2003) that mon-itoring efforts have often proven inadequate to quantify physicaland biological responses within the ecosystems being modified. Gi-ven this possibility of failure and the importance of communicatingecological information to stakeholders, monitoring programs needto play several roles, including: (1) integrating the scientificknowledge and theories behind the design of the project into themonitoring program to include the measurement of appropriatestressor and response variables, (2) developing and implementingan analytical framework that evaluates monitoring data to providethe pertinent information needed to adaptively manage the

All rights reserved.

+1 803 2546445.angman).

restoration to improve its chances of success, and (3) presentingthe progress and condition of the restored ecosystem to the stakeholders in a manner that is easily interpretable and understand-able, yet based on valid scientific assessments.

1.1. Endpoints of restoration

The endpoints of restoration have been described in terms ofcommunity structure as well as supporting chemical, biological,and physical processes (National Research Council, 1992).Descriptions matching this level of detail for the desired state ofremediation sites are rare, which has lead to the practice of havingreference ecosystems provide the basis for both developing remedi-ation methodology and evaluating the progress of an ecosystemrestoration (Society for Ecological Restoration, 2004). While a refer-ence system can be used as a model for a desirable outcome of res-toration, the restored site will at best approximate the condition ofthe reference site due to spatial variability, however, slight, in thephysical, chemical, and biological gradients forming the basis ofthe ecosystem processes. Further variability within a reference sys-tem emerges from the innate non-static nature of an ecosystem,across season variability, community-level evolution, or naturalprogression of the reference systems to new states (Duarte, 1991;Horne and Schneider, 1995; Palmer and Poff, 1997). Sadly, the desireto force an ecosystem into an overly specified state is common, andhas resulted in restoration ‘failures’ that are, for the most part, func-tional ecosystems in their own right (Simenstad and Thom, 1996).

http://dx.doi.org/10.1016/j.marpolbul.2012.01.030

mailto:[email protected]

http://dx.doi.org/10.1016/j.marpolbul.2012.01.030

http://www.sciencedirect.com/science/journal/0025326X

http://www.elsevier.com/locate/marpolbul

Degradedecosystem

Recla

mat

ion

(repl

acem

ent)

Rehab

ilitati

onDeg

radati

on

Natural

proc

esse

s

(prim

ary su

cces

sion)

Targetecosystem

Ecosystem structure

Ecos

yste

m F

unct

ion

Degradedecosystem

Targetecosystem

Alternatestates

Remains thesame

Ecos

yste

m C

ompl

exity

/ Fu

nctio

nTime

Degradedecosystem

Targetecosystems

Ecos

yste

m h

ealth

Ecosystem stressors

Stressors reduced viaremediation activities

Ecos

yste

m h

ealth

resp

onds srosserts

nisnoitcuder

ot

Less stressMore stress

A B C

Fig. 1. Theoretical spaces that describe the response of an ecosystem to remediation. (A) An original space that assesses the response of ecosystem health to the reduction ofstressors (B) A mapping of ecosystem complexity through time, redrawn from Hobbs and Mooney (1993) (C) Evaluating ecosystem function (biomass, nutrient content,cycling) against structure (species diversity, complexity), redrawn from Dobson et al. (1997).

O.C. Langman et al. / Marine Pollution Bulletin 64 (2012) 820–835 821

1.2. Trajectory

Equally as important as the restoration endpoint is the progres-sion from degraded to a restored ecosystem. Numerous conceptualmodels for the progression of a restoration site through time havebeen proposed (Dobson et al., 1997; Hobbs and Mooney, 1993;Hughes et al., 2005; Magnuson et al., 1980), with the term ‘‘trajec-tory’’ being used to describe the hypothetical pathway traversedduring the restoration progress (Fig. 1).The multiple interpreta-tions of restoration trajectory are based on which ecosystem attri-butes (e.g., ecosystem health, structure, and function) are beingtracked (Fig. 1) indicating the difficulty of consolidating therequirements of ecological restoration even at a conceptual level.As a reflection of this, the theoretical spaces leave trajectories sim-plified, indicating a general direction and approximate endpoint.Practical applications of the trajectory concept have largely in-volved developing multiple trajectories for individual parameters,often indicator species, used to represent restoration status. Theinherent variability of single parameters, for example, over stressorgradients and temporal/spatial scales, however, often results ininconclusive representation of restoration trajectory (Odum et al.,1995;Zedler and Callaway, 1999), and aggregating parameters intoa single trajectory has proven difficult (Society for Ecological Res-toration, 2004). An ideal trajectory would integrate disparate datathat describe site condition (and thus restoration status), and pro-vide information that may be used to adaptively manage the resto-ration project.

1.3. Adaptive management

Without a regular assessment of restoration status supportedby a well-developed monitoring program, a restoration site mayfollow a trajectory different from the desired outcome. While ourunderstanding of succession is continually improving, knowledgeof the current state of an ecosystem and the stresses that it willface during restoration will never be complete, leading to difficultyin making accurate predictions of site evolution over the durationof the restoration project. The proposed solution to this problem isto monitor the restoration status and to nudge the system towardthe desired trajectory and adaptively manage it if a significantdeviation is detected. According to Shreffler et al. (1995) , this con-cept is not new, but there are few examples in the literature thatindicate the principle is being used. Possible reasons for this in-clude insufficient funding for additional manipulation, lack of clearresolutions to the problems, or a lack of supporting data to drive

the management. In the latter case, the decision to re-engage inthe manipulation of the restoration site can be improved by pro-viding managers with a broader dataset that describes ecosystemstatus, and by extension, the range of problems that can occur dur-ing restoration.

Actually describing the concepts of trajectory, restoration end-points, and adaptive management within the context of a monitor-ing program remains a difficult task. Recent attempts at describingecosystem status have moved in two distinct directions: (1) iden-tifying organisms that can be used to integrate multiple signalsfrom the ecosystem (indicator species); or (2) by collecting largeamounts of data to produce community descriptors. Indicatorspecies have been developed as the corner stone of monitoringprograms (Metcalfe et al., 1984;Reynoldson, 1987). Monitoringprograms based on indicator species are particularly attractive be-cause acquiring the data is often time- and cost-effective. Severalcommon assumptions about the relationship of indicator specieswith the greater community have proven unreliable, however,such as the idea that high species richness or habitat diversity iscorrelated with the occurrence of rare species (Pearson andCassola, 1992), or that associations between species remain similaracross a given habitat (Niemi et al., 1997). When evaluating eco-system status, particularly in restoration projects where the suc-cessional trajectory can be short-cut through plantings, speciesintroductions, and other modifications, the lack of reliability ofthese relationships reduces the value of using just species–com-munity relationships for evaluating the condition of the ecosystem.

Complex systems need to be described using a framework ofmany parameters, although this task can overwhelm the research-er with data. Important parameters for ecosystems includeelements of structure, function, landscape connectivity, and resil-ience to perturbations, all of which must be addressed to evaluatethe status of a restoration. Karr (1981) introduced a multimetricindex (MMI) to represent elements of biological condition in a vari-ety of different systems. Karr’s work stemmed from the use ofwater quality data as a surrogate for biotic assessment, in caseswhen biological condition could not adequately be characterized.Until then, water quality was primarily monitored chemicallyand physically by the EPA and other monitoring institutions, butdespite extensive monitoring and management programs waterquality continued to deteriorate (Davis and Simon, 1989; EPA,1987; Karr, 1981). This resulted from not only having excess chem-ical data that were swamping managers, but from a lack of data onthe biological processes working on the ecosystems (Karr, 1981).Since its introduction as a method of monitoring biotic condition

822 O.C. Langman et al. / Marine Pollution Bulletin 64 (2012) 820–835

in streams, the use of MMIs has expanded to other ecosystems, buthas proven particularly useful in aquatic ecosystems (Emery et al.,2003; Stoddard et al., 2008).

The purpose of this paper is to use monitoring data to developan MMI that integrates disparate datasets and produces resultsthat can guide restoration efforts and provide useful informationto managers. The MMI will be based on preliminary monitoringdata from a salt marsh restoration project currently underway, lo-cated on the eastern coastline of Saudi Arabia, developed in re-sponse to persistent degradation that was a result of the largestoil spill in history. At the conclusion of the 1991 Gulf War, an esti-mated 6–11 million barrels of crude oil were intentionally releasedinto the Arabian Gulf, affecting approximately 800 km of the SaudiArabian coastline (Abuzinada and Krupp, 1994; Tawfiq and Olsen,1993). Both primary and secondary (faunal burrows) porosityallowed oil to penetrate the substrate of salt marshes andnearly two decades later, this oil persists. The United Nations

Fig. 2. Location of reference marshes (C1, C2, C3, C4, C5), restoration sites (R1,

Compensation Commission was formed to administer monetaryclaims against Iraq, and awarded $463 million USD toward the eco-logical restoration of degraded salt marsh along the Saudi Arabiancoast, creating one of the largest ecological restoration projects inhistory. We were tasked with developing and implementing amonitoring plan and an analytical framework for the monitoringdata to estimate the status of post-remediation recovery, a taskwhich included managing the restoration projects as well as incor-porating local and international stakeholders into the decision-making process.

The procedure used to generate the MMI is as follows:

� Identify attributes of the restoration site that may be areas ofconcern.� Create a list of potential measurements useful for evaluating

these attributes� Condense these measurements into a set of metrics.

R2, R3, R4, R5, R6, R7, R8, R9, R10), and the set-aside sites (S1, S2, S3, S4).


� Test these metrics for utility, based on initial monitoring.� Eliminate those metrics that show no difference between resto-

ration and reference sites, or that have high measurement ortemporal variability.� Evaluate reference sites to establish appropriate endpoints for

remediation.� Examine the current state of the restoration project in order to

provide managers with data that can be used to adjust the tra-jectory of the project.

2. Materials and methods

2.1. Study area

All field work was (and continues to be) conducted within inter-tidal salt marsh systems on the eastern coast of Saudi Arabia, be-tween Abu Ali and Batinah Islands in the south, and the Balbolembayment to the north (Fig. 2). This segment of shoreline repre-sented the area with the largest amount of stranded oil from the1991 Gulf War oil spill. In the period between January and Mayof 1991, an estimated 6–11 million barrels of Kuwaiti crude oilwere pumped or released into the Arabian Gulf, creating the largestoil spill in history (Abuzinada and Krupp, 1994; Tawfiq and Olsen,1993). Much of the oil was trapped in the south due to orientationof the islands and bays between Abu Ali Island and Tanajib. Withinthese bays and islands, large areas of sheltered mudflats and inter-tidal marsh systems were moderately to heavily oiled (Fig. 3) dur-ing the initial spill, with oil penetrating particularly deeply inmarshes due to an abundance of burrowing organisms (Getteret al., 2005; Pandion Technology Ltd., and Research Planning Inc.,2003; Michel et al., 2005). A follow-up survey performed in2002–2003 mapped the extent and toxicity of the remnant oil over

Fig. 3. Examples of various levels of oiling. (A) Heavy oiling located on the marsh surfacedue to oil trapped in relic burrows by laminate mat covering. (C) Moderate to light oil shein channel bank.

12 years after the initial oiling. Marsh areas that were described asmoderately oiled in the 2002–2003 survey report contain oil that,when flooded, produces sheen and small droplets of liquid oil.Areas that were described as heavily oiled produced thick patchesof liquid oil under similar conditions (Fig. 3). Salt marshes weredetermined to contain over 23% of the remnant oiled sediments,with total PAH (sum of 43 PAHs) of ranging from 3580 ng/g to126,900 ng/g in areas with visible remnant oil, which included allof the restoration marshes in this study. Bejarano and Michel(2010) found that heavily oiled salt marshes had total PAHs at con-centrations that were above sediment quality benchmarks andPAH distributions indicating slow weathering rates compared toother habitats.

These marshes were fine-grained, low-energy systems; tidalrange was 0.5 m, and inputs of fresh water were minimal. Produc-tivity was assumed to be low due to limitations imposed by ex-tremes of heat and salinity: during summer, air temperaturesregularly exceeded 45 �C, and salinity in the channels reached90ppt. Although no studies could be located that quantified pri-mary production for the marsh intertidal zones in the Arabian Gulf,it is assumed that algae were responsible for a higher fraction ofprimary productivity relative to marsh systems in other parts ofthe world (Al-Zaidan et al., 2006). The dominant vascular plantswere two succulent perennials, Arthrocnemum macrostachyumand Halocnemum strobilaceum, and two succulent annuals, Salicor-nia europea and Suaeda maritima. Benthic algal mats comprised ofdiatoms and cyanobacteria often covered much of the marsh sur-face, with diverse morphologies influenced by elevation, hydrol-ogy, and nutrient gradients (Kendall et al., 1968). The crustaceancommunity’s main constituent was Nasima dotilliforme, with Macr-ophthalmus sp. occupying some channel bottoms and low eleva-tions extending into seaward mudflats, and the predatory

due to oil trapped in non-burrow pore spaces. (B) Moderate oiling in channel banken in channel bank. (D) Moderate oiling from relic small (insect/polychaete) burrows

Fig. 4. Photograph of an excavation that shows a cross-section of laminated algalmat. Individual layers of sedimentation and algal growth are readily visible.Laminate mats grow as thick as 10 cm in some areas. An oil-lined burrow that waslocated below the laminate mat has been placed in the upper left of the photograph.


Metopograpsis messor, Metaplax indica, and Eurycarcinus orientalisappearing sporadically. Other invertebrates included the snails Pir-inella conica and Nodilittorina sp., as well as numerous burrowingpolychaetes and amphipods.

Secondary effects from the oil spill have further impaired natu-ral recovery of the salt marsh. In many areas of degraded saltmarsh, the lack of burrowing fauna and algal grazers has allowedthe algal mat to develop into a thick, laminated covering over themarsh surface (Fig. 4), impairing recovery by acting as a barrierto re-colonization, and preventing further weathering of theremaining oil (Al-Thukair and Al-Hinai, 1993; Barth, 2003, PandionTechnology Ltd., and Research Planning Inc., 2003). The thicknessof the laminated mats ranged between 5 and 25 cm, with the thick-est mat located in depressions on the surface of the marsh and de-graded tidal channels. In these cases, the secondary effects werecomplicated further by the algal mat impairing marsh hydrology,which created a positive feedback with the algal mat, whichthrived in areas with limited drainage.

The remediation and restoration activities planned for the res-toration sites include excavation of existing or impaired tidal chan-nels (blocked primarily by algal mat), creation of new tidalchannels, sediment tilling and direct treatment of algal mat, andtransplantation of local flora, primarily Avicennia marina. Whileremnant oil still exists within these marshes, the focus of the res-toration efforts will be on accelerating the natural recovery ofthese systems rather than the removal of oil. Most of the remnantoil in the marshes is contained within collapsed burrows of Nasimaand other burrowing organisms, with oil deposits occurring atdepths of up to 0.6 m (Pandion Technology Ltd., and ResearchPlanning Inc., 2003; Michel et al., 2005). No feasible method forlocating and extracting the oil from these burrows exists, andattempting to expose burrows through tilling or other invasivemethods would impair any natural restoration that has occurredand potentially expose adjacent areas to re-oiling. Since some ofthe areas have recovered naturally to a large degree, the restora-tion activities will only be employed in areas where they can rea-sonably be expected to improve the sites. Many of the marshestargeted for restoration within this study appear to have stabilizedin various states of natural recovery, but all are expected to benefitfurther from targeted restoration activities.

2.2. Site selection and sampling strategy

Following an initial coastline survey in late 2009, marshes wereprioritized for remediation activities based on a rapid assessmentprotocol that focused on evaluation of overall ecosystem health

(Hale et al., 2011). Salt marshes were placed along a gradient ofdisturbance that ranged from no impact to heavy impact, basedprimarily on an ecological evaluation. Sites identified as having re-ceived little to no impact during the initial oiling event, or sitesthat have since been described as sufficiently recovered, were usedas reference marshes. Sites that were initially heavily oiled andhave since had little natural recovery were divided into sites tar-geted for remediation, and set-aside sites, which will not be re-stored and serve as a baseline for monitoring natural recovery.This study incorporated data from five reference sites. Four of thereference sites were not impacted by the initial spill, the fifthwas deemed to have recovered sufficiently due to statistical simi-larity with the non-impacted sites, and has been classified asrecovered by other authors (Höpner and Al-Shaikh, 2008). The ini-tial set of five restoration sites was selected from a range of impactlevels and levels of ecological recovery between a medium level ofimpact to a high level of impact. Reference sites were located closeto impacted sites when possible, although the number of referencesites was limited due to the high extent of oiling that occurredalong the coastline.

Field sampling was divided into biannual efforts, with the firstyear’s monitoring efforts represented here. Fall sampling occurredbetween September and November of 2010, and spring samplingoccurred between February and April of 2011. Spring sampling oc-curred over 19 sites, including 10 impacted sites, 5 reference sites,and 4 set-aside sites. The sampling windows helped to reduce theeffect of temporal variability of the field measurements, which isoccasionally a concern for MMIs (Blocksom, 2003).

The field program was designed to detect changes in environ-mental attributes that differ between reference and restorationsites due to primary or secondary effects from the oiling, and to de-tect changes that were the direct result of, or affected by, the reme-diation activities performed during the salt marsh restoration. Saltmarsh restoration sites ranged from around 5 ha to larger than80 ha in size, making sub-sampling desirable for most aspects ofmonitoring, and particularly for determining species diversityand density. Within each site, natural recovery was most evidentalong channels and in the lower elevations of the marsh, possiblyreflecting differing rates of recovery relative to the degree of tidalflushing and initial vertical oil penetration.

To ascertain the relative importance of elevation and channelproximity to the rate of recovery, nine sub-sampling plots were lo-cated within each reference and restoration site. The nine plotswere placed at three relative elevations on the marsh surface, atlow, medium, and high elevations, which, based on the referencemarshes, corresponded to three distinct communities that differprimarily in the abundance and spatial distribution of the organ-isms present. For example, Salicornia europea is found abundantlyon the marsh surface in the low elevation, only along channels inthe medium elevation, and is largely absent in the high elevationplots. Despite the marked differences among the communities,the actual elevation range for the marsh surface rarely exceeded30 cm from the mudflats that marked the end of the low elevationto the toe of a narrow sand beach that typically marked the extentof the upper marsh.

At each elevation, two plots were associated with channels, andone plot was placed in the interfluvial space on the marsh surfacebetween channels. The channel-associated plots were placed adja-cent to two different channels within the same channel network ofdifferent order. Depending on the configuration of the marsh, thesechannels may either to be first and second order, or second andthird order, but the larger of the two channels was between 1 mand 3 m wide and the smaller was <1 m wide.

Data were collected and archived independently for each plotwithin a site for each season, and later combined during analysis.It was hypothesized that elevation and proximity to the channel

Table 1List of attributes of restored ecosystems (SER, 2004).

Attribute Description

1 The restored ecosystem contains a characteristic assemblage of the species that occur in the reference ecosystem and that provide appropriate communitystructure

2 The restored ecosystem consists of indigenous species to the greatest practicable extent.3 All functional groups necessary for the continued development and/or stability of the restored ecosystem are represented or, if they are not, the missing

groups have the potential to colonize by natural means.4 The physical environment of the restored ecosystem is capable of sustaining reproducing populations of the species necessary for its continued stability or

development along the desired trajectory.5 The restored ecosystem apparently functions normally for its ecological stage of development, and signs of dysfunction are absent.6 The restored ecosystem is suitably integrated into a larger ecological matrix or landscape, with which it interacts through abiotic and biotic flows and

exchanges.7 Potential threats to the health and integrity of the restored ecosystem from the surrounding landscape have been eliminated or reduced as much as possible.8 The restored ecosystem is sufficiently resilient to endure the normal periodic stress events in the local environment that serve to maintain the integrity of

the ecosystem.9 The restored ecosystem is self-sustaining to the same degree as its reference ecosystem, and has the potential to persist indefinitely under existing

environmental conditions.


networks may influence the rate of recovery, which required thatthe data be separable based on these criteria. Furthermore, sea-sonal data were kept separate to allow for analysis of marsh pro-gression through time.

The data collected at each plot included measurements that as-sessed elements of ecosystem structure, function, and remnantanthropogenic effects within each site. Using the SER’s nine attri-butes of restored ecosystems as a guide (Table 1), a wide rangeof measurements were taken to assess current ecosystem status(Society for Ecological Restoration, 2004). For each of the SER attri-butes that applied to this project, several measurements that willlater provide the information for developing ecosystem metrics,and that would represent each attribute in the overall restorationstatus, were identified (Table 2). The initial field samplings in-cluded as many of these measurements as could feasibly be mea-sured by the field crews, with the assumption that some of thesemeasurements would prove to be non-informative for the overallmonitoring program and would later be dropped from the fieldprogram.

2.3. Metric filtering

Once the initial monitoring was complete, the measurementswere further developed into metrics. A ‘metric’ was defined as a va-lue that could be derived from one or more discrete or continuousmeasurements of the marsh or surrounding landscape. Metrics lar-gely coincided with the planned measurements in Table 2, but of-ten the individual measurements could be utilized by one or moremetrics, which were further broken down into components wherepossible. A metric that was derived from three measurements butcould potentially be represented by a combination of two measure-ments or any individual measurement was developed into everypossible combination. For example, the health of benthic infaunacould be described as simple presence/absence of species, densitymeasurements of the species, or by measurements of a subset ofspecies present. A metric was developed for each possibility result-ing in several, often highly correlated, metrics derived from thesame source data. This process, combined with filtering at a laterstage, resembles a rarefaction analysis. Categorical measurementsrequired conversion before they can be interpreted as metrics. Anexample of this was the algal form observations, which were con-verted into a rank value arranged from least to most desirable. Al-gal forms that were regularly observed in reference marshes areassigned a value of 1, while polygonal laminate and flat laminatealgal forms were assigned values of 3 and 5, respectively, repre-senting their relative impact on hydrology and impairment to col-onization (Fig. 5).

2.3.1. RangeMetrics that have limited ranges were not useful for describing

differences between marsh systems. Discrete variables such as rareorganism abundances may lack the statistical power needed toprovide useful information. Continuous variables occasionally ex-hibit measurement error that exceeds the range of the measure-ment. To address these concerns, discrete metrics wereeliminated if their range was 2 or less. Continuous measurementswere eliminated if their range was within the estimate of error forthe measurement. For example, visual estimates of percent coverwere determined to be ±5% based on photographic post-analysisof field estimates, so if the range fell under the 10% threshold,the metric was discarded. The criteria for discarding percent covermetrics matches Klemm et al. (2003), although it was unclear howKlemm et al. arrived at their rejection threshold. Other continuousmetrics had rejection thresholds ranging from 7% to 35%, and werecreated using estimates of measurement variability derived fromrepeated samplings using different people and equipment. Finally,any metric where >90% of the values were 0 was discarded, such asmeasurements of abundance of snails of the genus Cerithium,which were rarely found within either the reference or restorationsites. Metrics that failed the range test were eliminated from fur-ther consideration.

2.3.2. Response to disturbanceWithin the context of a restoration due to anthropogenic dam-

age, a metric must have been be able to distinguish between refer-ence and impacted sites. Each site was classified according to theinitial survey as ‘‘impacted’’, ‘‘set-aside’’, or ‘‘reference.’’ Prior toremediation activities, the ‘‘set-aside’’ and ‘‘impacted’’ classifica-tions were treated as a single classification. A one-way analysisof variance (ANOVA) was used to test each metric for its abilityto classify set-aside/impacted and reference sites. Significant met-rics (F-value significance of p6 0.05) were preserved, while metricsthat was incapable of distinguishing between set-aside/impactedand reference sites were discarded.

2.3.3. Signal-to-noise ratioMetrics that exhibited a large amount of within-site variability

relative to the variability between sites did not contribute ade-quate information to the index. The signal-to-noise ratio was cal-culated as the variance over all of the sites (impacted, set-aside,and reference) as the signal, divided by the variance of the metricbetween seasonal visits at individual sites as the noise (Kaufmannet al., 1999). Various ratio thresholds for retaining metrics havebeen suggested during the development of other MMIs, rangingfrom >1.5 to >3 (Klemm et al., 2003; McCormick et al., 2001,

Table 2Target measurements and SER attribute associations.

SER attribute

Measurement References

Characteristic assemblage of species, intact community structureAmphipod abundance Gómez Gesteira and Dauvin (2000)Active ocypode burrow counts Adam (1990)Benthic infauna abundance Sacco et al. (1994)Species richness Jones and Richmond (1992), Jones et al. (1998)

Presence of invasive species*

Invasive richness Mooney and Hobbs (2000)Invasive diversity Mooney and Hobbs (2000)

Presence and condition of key functional groupsPerennial halophyte abundance Adam (1990)Perennial halophyte canopy cover Adam (1990)Gastropod abundance Peck et al. (1994)Ocypode abundance Adam (1990)

Physical environment which supports biotaChannel, porewater, and ponded salinity Broome et al. (1988), Portnoy (1999), Howard and Mendelssohn (1999)Substrate temperature in rhizosphere Callaway and King (1996), Lindig-Cisneros and Zedler (2002)Burrow abundance on channel banks Adam (1990)Burrow abundance in perpendicular transect extending onto marsh surface (Original)

Normal ecosystem function and developmentMarsh surface drainage/hydrology Montalto and Steenhuis (2004)Subsurface hydrology Montalto and Steenhuis (2004), Osgood and Zieman (1998)Perennial growth rates Howard and Mendelssohn (1999)

Integration into landscapeFish abundance Chamberlain and Barnhart (1993)Fish size distributions Madon et al. (2001), Weisburg and Lotrich (1982)Fish gut contents Allen et al. (1994)C,N,S stable isotope analysis Kwak and Zedler (1997), Peterson et al. (1986)Bird foraging Brawley et al. (1998)Bird diversity Warren et al. (2002)

No threats to adjacent systemsAnalyses of foraminifera Morvan et al. (2004), Sabean et al. (2009)

Resistance and resilienceShannon–Weaver diversity Tilman (1996), Yachi and Loreau (1999), Naeem and Li (1997)Euclidean distance index for functional attribute diversity Walker et al. (1999)Molecular biomarkers Downs et al. (2001a,b)

Self-organizing, self-sustaining ecosystemChannel morphology D’Alpaos et al. (2005), Phillips (1999)Annual halophyte abundance Adam (1990)

* No invasive species were detected in the study area.


respectively), but little support for the different thresholds wasapparent. After examining the response of the metrics that demon-strated significant response to the stressor gradient with signal tonoise ratios between 0 and 5, we determined that a threshold of 3was a tolerable compromise between eliminating variables withlimited information and ensuring that the SER attributes wereadequately defined by the measurements collected. Fig. 6 demon-strates the difference between a metric with a high signal-to-noiseratio (Nasima density) and a low signal-to-noise ratio (percentageof exposed sediment within a quadrat), with the latter serving asan example of a rejection based on the signal-to-noise criteria.Further consideration is given to address issues of metric scoringbelow.

2.4. Metric scoring

The diversity of values and types of measurements that com-prised the metrics required that they be rescaled to comparablevalues before they were integrated into a single multimetric value.Specifically, sites were scored continuously to the range 0–100,with the lower threshold of 0 representing the worst or most unde-sirable condition and the upper threshold of 100 representing themost desired condition. Several metrics required transformationsbefore they could be rescaled to address directionality and

distribution concerns (Table 3). Metrics that had higher valuesrepresenting worse condition needed to be inverted to match thedesired scheme. For example, algal mat cover, a percentage metric,was considered more degraded at higher degrees of cover, whichwas addressed by inverting the cover values for algal mat cover.Metrics that were recorded as percentages originally derived fromcount data needed to be arcsine transformed to convert from abinomial distribution to a normal distribution. All such transfor-mations were applied prior to the rescaling.

The low number of sites made performing a power analysis (asper Blocksom, 2003) impossible, but the guidance provided byBlocksom’s analyses aided in the choice of rescaling algorithms,resulting in the selection of a continuous scaling method. SeveralMMIs in the literature have used the 95th percentile value of a met-ric to determine the upper bound and the 5th percentile to deter-mine the lower bound used in the transformation for a specificmetric (Blocksom, 2003; Stoddard et al., 2008).The limited numberof sites available in this restoration project, however, allowed indi-vidual sites to alter these bound values drastically, meaning thatthe use of percentile-derived boundaries for rescaling may providelimited benefit. Instead, the minimum and maximum observed val-ues were used as the upper and lower bounds. In Eq. (1), S is the re-scaled metric. The mean of the set of measurements M is linearlyrescaled between 0 and 1 and then multiplied by 100.

Fig. 5. Photographs of three algal forms. (A) A typical ‘‘laminate’’ algal form, with an average thickness 52 mm. (B) A typical ‘‘folded’’ algal form, with an average thickness of1 mm. Note that this form is regularly seen in the reference marshes. (C) A typical ‘‘polygonal laminate’’ algal form, with an average thickness of 48 mm.

50

75

100

Met

ric V

alue

0

25

50

Nasima density(Reference)

Nasima density(Impacted Sites)

% Exposed Sediment (Reference)

% Exposed Sediment (Impacted Sites)

Fig. 6. Demonstration of signal-to-noise ratios. Both the Nasima density metric and the percentage of exposed sediment (i.e., not covered by algal mat or asphalt pavement)metric have significant values for responsiveness to the disturbance, but the percentage of exposed sediment metric fails the signal-to-noise ratio test. Signal-to-noise valuesfor Nasima density and percent exposed sediment are 3.14 and 1.21, respectively.


S ¼ M � lowerupper� lower

� 100 ð1Þ

The inability to correct for potential outliers without drasticallyincreasing the number of set-aside and reference sites indicatedthat reducing measurement and seasonal variability was

particularly important. Increasing the signal-to-noise thresholdfor metric inclusion reduced the potential for a variable metric tocause rescaling to affect interpretative capacity. For this reason,the signal-to-noise threshold of 3 was preferred over lower litera-ture values to stabilize the edge effects from variability.

Table 3List of metrics and filtering statistics.

Metric Range (pass/fail) Responsiveness (p-value) Signal-to-noise Transform

Algal form rank Pass 0.02 1.19 NoneAlgal mat% cover Pass 0.00 44.82 InvertAmphipod density Pass 0.04 5.26 NoneArthrocnemum% cover Pass 0.06 1.09 ArcsineArthrocnemum count Pass 0.14 1.10 NoneAsphalt pavement% cover Fail 0.00 1.80 InvertAvicennia% cover Pass 0.68 Undef ArcsinAvicennia count Pass 0.03 2.99 NoneCerithidea density Fail Undef Undef NoneChannel belt transect – Nasima burrows Pass 0.00 22.35 NoneChannel belt transect – total burrows Pass 0.00 27.03 NoneExposed sediment% cover Pass 0.00 1.46 NoneHalocnemum% cover Pass 0.03 4.61 ArcsinHalocnemum count Pass 0.98 126.28 NoneMacrophthalmus density Pass 0.19 1.60 NoneMetopograpsis density Pass 0.32 20.47 NoneMicro-channel% cover Pass 0.03 12.07 NoneNasima density Pass 0.01 3.78 NoneNodilittorina density Pass 0.00 1.80 NoneOcypode predator:prey ratio Pass 0.34 25.32 NoneOcypode richness Pass 0.00 4.61 NonePirinella density Pass 0.25 12.78 NonePonded water% cover Pass 0.08 3.65 InvertPneumatophore density Pass 0.68 1647.59 NoneSalicornia density Pass 0.65 3.66 NoneSalinity – channel Pass 0.11 4.30 NoneSalinity – ponded Water Pass 0.74 7.73 NoneSalinity – porewater Pass 0.35 1.19 NoneScopimera density Pass 0.48 0.89 NoneSediment temp in rhizosphere Pass 0.00 37.51 NoneShannon index – all organisms Pass 0.04 3.17 NoneSmall insect burrow density Pass 0.59 1.29 NoneSnail richness Fail 0.16 2.59 NoneSuaeda density Pass 0.78 3.29 NoneTotal perennial% cover Pass 0.04 126.78 ArcsineTotal perennials count Pass 0.08 1.41 NoneTotal snail density Pass 0.25 13.10 NoneTotal species richness Pass 0.03 4.35 NoneVisual oiling estimate Pass 0.83 3.73 NoneWhole plot ponding Pass 0.05 3.59 Invert


2.5. Categorization of datasets/treatment of redundancy

Once the metrics were filtered and rescaled, the remaining sur-vivors were grouped with the SER attribute that they most closelyrepresent. After the metrics were assigned, there were several pos-sibilities: (1) an SER attribute lacked any descriptive metrics; (2) anattribute was assigned exactly one metric; or (3) an attribute wasrepresented by several metrics. In the first case, the metric eithercould not be evaluated for the site and needed to have additionalmetrics considered to evaluate the attribute or, more desirably,the attribute was not a concern for the remediation project. Inthe second case, the metric would directly represent that SER attri-bute. In the third case however, the metrics needed to be examinedfor redundancy and combined into a single metric. Redundancywas evaluated using Pearson’s product-moment correlation coeffi-cient, with a rejection of a metric correlated with another atr P 0.75. If two metrics were redundant, the metric with a lowersignal to noise ratio was discarded. The remaining metrics wereadded into a single index. The resulting combined values wererescaled.

3. Results

This method of use and reuse of measurements produced an ex-cess of metrics. This initial suite of measurements contained moreinformation than strictly needed to evaluate marsh condition, withthe assumption that many of the metrics would prove redundantor uninformative for evaluating restoration status. As a

consequence, the reduction of the set of metrics, and by extensionthe measurements taken during the field program, was a processintegrated into the MMI development. Measurements that wererejected due to statistical insignificance or redundancy could bedropped from the sampling program altogether, which would al-low researchers to focus valuable field time (and efforts) on rele-vant metrics. To aid this process, a single metric could be splitinto a gradient of complexity; if a metric was derived from twocomponents and could potentially be represented by either oneof those components, the individual components were includedin case they were statistically sufficient to describe the desiredattribute. For species presence and density metrics, often the met-ric was not significantly improved by incorporating the density ofrare species, and so the effort needed to quantify the rare speciescan potentially be reduced. For example, the density of the preda-tory Metopograpsis was less informative (lower signal-to-noise ra-tio) due to high variability in densities than simply noting thepresence or absence of the organism at a site.

Metrics that had sufficient range, were repeatable (high signal-to-noise ratio), and exhibited a response to the disturbance wereintegrated into the field program and monitored for two seasons,spring and fall. Whole ecosystem development and recovery wereexpected to be slow, with perennial vegetation and soil organiccontent potentially taking decades to recover to reference status.The most marked differences in salt marsh status between seasonswill be visible during the spring recruitment period. Preliminarystudies have shown rapid re-colonization of excavated channelsby burrowing infauna, including amphipods and Nasima.


3.1. Metric selection

Metrics were selected for possible inclusion into the MMIs outof a total of 72 possible metrics. Of the 72 metrics, nearly half(32) were alternative forms of metrics and were discarded asunderperforming versions produced as part of the rarefaction anal-ysis used to generate more efficient metrics. While this rarefactionapproach produced an excess of metrics, the assumption was thatmost of these metrics were redundant, but occasionally a metriconly required a subset of the sampling effort to obtain adequateinformation. In practice, only the best performing metric in termsof range, responsiveness to disturbance, and signal-to-noise ratiowas preserved. Out of the remaining 40 metrics, 26 metrics wererejected (Table 3).Three of the metrics failed the basic range test,19 failed the response test, and 4 failed the signal-to-noise ratiotest.

The visual oiling estimate failed to distinguish between refer-ence and remediation sites. This was presumably a response tothe methodology used to initially classify the sites; the initial clas-sification considered the ecological condition, regardless of currentvisual oiling, allowing a recovered but oiled site to persist as a ref-erence site. In addition, areas that contained relatively little rem-nant oil occasionally exhibited a large degree of secondaryimpact in the form of extensive algal mats. As a result, the visualpresence of oil proved not to be a good metric. Instead, secondaryeffects that presumably developed in response to the oiling, includ-ing pervasive, laminated forms of algal mat and degraded hydrol-ogy, were used to represent the stressors acting on the marsh.

3.2. SER attribute axes

Three of the nine SER attribute MMIs are presented here, eachexhibiting a distinct difference in condition between the impactedand reference sites (Figs. 7–9). Attribute 4 (Fig. 9), which repre-sents aspects of the physical environment that support local biota,

Fig. 7. MMI describing the characteristic assemblage of species and community structurstressors. The y-axis is comprised of the following metrics: species richness, amphipod prrestoration target for this attribute. Significant difference between impacted and referen

is of particular note. In this case, the impacted sites had a largerange of variability. The differences were likely due to the densitiesof the tidal channel networks; algal mat has in-filled many of thechannels at the sites that performed the most poorly on this axis(R5, R7, R8, R9), while some of the sites that performed markedlybetter had more intact networks (R2, R3). This difference reflectsthe remediation and restoration plans for these sites; the sites withbetter condition will receive less focus on physical environment.The other attributes showed similarly degraded conditions acrossthe impacted sites, particularly across the attributes that repre-sented floral and faunal structure (Figs. 7 and 8).

From the pre-restoration plot of an individual attribute, it waspossible to tell whether the attribute was degraded in the marshsystem. An ANOVA with a significant f-value (see Figs. 7–9) indi-cated that the attribute in question was degraded relative to thereference marshes. If there was no detectible difference, the attri-bute was not considered degraded within the restoration sites.Comparison of magnitude across attribute MMIs was meaninglessdue to metric rescaling.

An integrated MMI (Fig. 10) was created by taking the entirecollection of metrics and incorporating all of them into a singleMMI using the same process used for individual attribute MMIs.Some metrics included in individual attribute MMIs were excludeddue to high correlation with metrics from other attributes, result-ing in a smaller total subset of all of the considered metrics. Likethe attribute MMIs, individual metrics were equally weighted.

3.3. Stressors axis

The stressors MMI combined metrics that represented elementsof marsh condition that were directly targeted by the restorationand remediation activities. In this case, the activities includedexpansion and repair of drainage networks, removal and breakupof the algal mats, and perennial transplants, which were designedto improve drainage, reduce laminate algal mat cover, and

e (SER attribute #1) plotted against a separate index describing the environmentalesence and abundance, Nasima presence and abundance. The grey oval indicates thece sites (ANOVA, p = 0.0010).

Fig. 8. MMI describing the presence and condition of the key functional groups of organisms (SER attribute #3) plotted against a separate index describing the environmentalstressors. The y-axis is comprised of the following metrics: Shannon–Weaver (Shannon–Weiner) diversity index, perennial cover, and the presence and abundance ofpredatory ocypodes. The grey oval indicates the restoration target for this attribute. Significant difference between impacted and reference sites (ANOVA, p = 0.0002).

Fig. 9. MMI describing the condition of the physical environment (SER attribute #4) plotted against a separate index describing the environmental stressors. The y-axis iscomprised of the following metrics: porewater salinity, relative sediment temperatures, and the abundance of evidence of burrowing crabs on channel banks. Significantdifference between impacted and reference sites (ANOVA, p = 0.0076).


potentially increase the rate of degradation of the remaining oil.These activities were selected to offer the greatest chance of recov-ery for the effort expended, and do not necessarily represent at-tempts to address all of the stressors observed at each site. For

example, the removal of remnant oil (itself a stressor, particularlyfor burrowing fauna) from sealed infaunal burrows was determinedto be infeasible due to the difficulty of locating burrows with pock-ets of remnant oil as well as actually extracting the oil. All of the re-

Fig. 10. MMI integrating all surviving metrics across all SER attributes. The y-axis is comprised of the following metrics: Nasima abundance, Amphipod abundance, perennialcover, integrated snail abundance, porewater salinity, relative sediment temperatures, and the abundance of evidence of burrowing crabs on channel banks. Significantdifference between impacted and reference sites (ANOVA, p = 0.0001).

100

90

80

70

60

50

40

30

20

10

0

Fig. 11. Time series plot of the Ecosystem health MMI (Fig. 10) for three reference marshes. Marsh C1 exhibits marked seasonal differences that do not appear in marshes C2or C3.


sponse axes (plotted on the vertical axis) were plotted against thesame suite of stressors (plotted on the horizontal axis), since theprimary concern of the monitoring program is to evaluate the effec-tiveness of the restoration and remediation techniques (Figs. 7–10).

3.4. Marsh restoration target

A marsh restoration target (Figs. 7–10) was developed bycalculating the standard deviation on the reference marshes

for each axis and creating an oval centered on the referencemarshes to serve as a target. While this was contrived, it wasuseful because it integrated all of the elements of the SER attri-butes that described a successfully restored ecosystem, providedthe attributes were represented and were given equal weight.Similar restoration targets may be developed for individualattributes, allowing for characterization of partial restorationsuccesses. The restoration target for all of the marshes was de-fined as the space occupied by the reference marshes with a


buffer of one standard deviation of the variability of the targetmarshes.

3.5. Natural gradients

No adjustments were made to address association between themetrics that may be correlated with the natural conditions thatdetermined the initial degree of oiling. The topography of individ-ual marshes varied greatly; several marshes contained dense den-dritic channel networks and shallow pools while others lackedpools or displayed a lower density of tidal channels. The marsheswere generally located across several different sheltered bays withsurprisingly distinct tidal regimes due to two competing amphi-dromic centers, and sediments that were primarily muddy butwith a varying sand component. Despite this surficial topography,sediment variability, and a 100 km distance between the northern-most and southernmost marshes involved in this study, the floraland faunal communities were similar across the referencemarshes. It appears that the most important factor in determiningthe degree of oiling at a site was the predominant (NW) winddirection and magnitude during the initial oiling event (PandionTechnology Ltd., and Research Planning Inc., 2003).Thus, directionof the marsh exposure was correlated with many of the metricswhich resulted in reference sites within the impacted area gener-ally having a southern exposure.

3.6. Seasonal changes

Due to seasonal rescaling and re-evaluation of metrics, it washypothesized that seasonal differences would not significantly af-fect the output of the MMI generation process. One of the referencesites, however, suggests that this expectation was unfounded(Fig. 11). Site C1 exhibited a significant amount of seasonal varia-tion derived from a robust infaunal community (primarily amphi-pods, Grandiderella sp.) that was delayed with respect to the otherreference sites in the spring but which had established itself by thefall season sampling period. As a result, the MMIs that consider theinfaunal community for the spring sampling periods rate site C1poorly. Since other reference sites lacked that particular dynamicor other unique local seasonal changes, the seasonal variationwas not as pronounced for other sites.

4. Discussion

The primary concern of a restoration project is to initiate oraccelerate the recovery of all attributes of a degraded ecosystemtoward a more desirable state, with respect to ecosystem health,function, sustainability, and ongoing impacts to adjacent landscape(Society for Ecological Restoration, 2004). In practice, this has pro-ven exceedingly difficult, which can be inferred based on the num-ber and frequency of failed restoration projects (Hobbs and Norton,1996; National Research Council, 1992). Failures happen for anynumber of reasons, including unexpected spatial and temporal var-iability in the physical setting, misunderstood relationships andinteractions between resident organisms, and changes in the hu-man desires driving the restoration. The process presented hereinis derived from the example of establishing a monitoring frame-work for the restoration program on the eastern coastline of theKingdom of Saudi Arabia. This methodology will be transferableto different habitats and different regions and will identify whenthe restoration as a whole is underperforming, and will presentthe relevant information in terms of the prevailing theories behindecosystem restoration, including trajectory and restoration targets.By further compartmentalizing the attributes of a restored

ecosystem, problems can be localized, and an adaptive adjustmentcan be implemented.

4.1. Marsh restoration target

Restoration targets have been defined in many ways, such asproviding habitat for a target species, meeting some minimum offunctional performance, or meeting a structural requirement(Ehrenfeld, 2000;Zedler, 2001). In each case, reference systems thatmeet the requirements define the target ecosystem and likely spana gradient of potential conditions that will support the target goals.For this restoration project, the reference marshes were selected toinclude a range of healthy marsh configurations, partially due to thelimited selection of reference sites due to the high degree of oilingalong the coastline, but also to avoid over-specifying the restorationendpoint. Variability with the resulting ecosystems is tolerable aslong as it resembles other systems in the area (Sacco et al., 1994).The configuration of the reference marshes varies in some measure-ments, but is remarkably similar in others. For example, the Arthro-cnemum density metric had a range of 81 (after rescaling) for thereference marshes alone, indicating an immense degree of variabil-ity within the reference marshes for that metric. Channel density asestimated by analysis of satellite imagery and porewater salinity,however, were quite similar across all of the reference marshes.Porewater salinity in healthy sites ranged from 60 to 95 ppt, witha mean value of 82. Salinities at impacted sites were even higher,regularly exceeding the detection limits of the equipment(100 ppt), likely due to poor drainage (flushing) and the retentionof water in the substrate. Variability of the traits is reflected inthe size of the individual attribute targets, with larger acceptableranges appearing for highly variable marsh characteristics.

Despite these differences, the integrated, whole ecosystem MMIscores were similar (Fig. 10), indicating that there were likely sev-eral configurations that had similar overall complexity and organ-ism densities represented within the reference marshes. Smallerrestoration projects often do not have the benefit of a monitoringprogram large enough to characterize variability across multiplereference marshes. This is one possible reason why some projectsfail to meet the specific criteria set out as restoration targets, de-spite the marsh forming into a viable and desirable ecosystem(Simenstad and Thom, 1996).

4.2. Trajectory

The restoration trajectory concept is particularly intuitive, whichmakes it useful for communicating restoration status, of particularuse when incorporating stakeholders into the decision making pro-cess. In most of the two-dimensional restoration spaces described inthe literature (Fig. 1), restoration trajectories were represented as asmooth line leading directly from the degraded state in the lowerleft corner to the desired system state in the upper right corner.The MMI attribute and stressors spaces were intentionally set upto mirror this configuration. The trajectory concept, however, waslikely oversimplified. In most of these spaces, ecological restorationwas portrayed as degradation in reverse (Bradshaw, 1987). Therecovery may involve feedbacks that are not reversible, however,such as salinization or nutrient shifts that impair recovery (Proberet al., 2002; Davis et al., 2003). Alternatively, the restoration activi-ties may not promote response across all aspects of an ecosystem,and may negatively impact the site during restoration activities.While these incongruities and sources of variability will make thetrajectory less clean, the overall movement from a degraded to arestored condition is preserved. Non-technical stakeholders willbenefit greatly from the intuitive representation of restorationprogress, potentially encouraging ownership in the project. If the‘‘ecosystem health’’ descriptor is too generic (Fig. 10) for the


intended audience, the individual SER attribute MMIs can be pre-sented using the same trajectory and restoration target concepts(for example, Figs. 7–9), further expanding depth of understandingwhile preserving the method and simplicity of interpretation.

Trajectory in the attribute-stressor spaces presented in Figs. 7–10 will not follow a strictly linear pathway from lower-left toupper-right due to non-linear response. The stressor axis repre-sents the measureable elements that will be directly affected byactivities undertaken to accelerate restoration at the site, includingprimarily metrics of hydrology and algal mat cover. After restora-tion activities are performed, which include direct manipulationof algal mat cover as well as increases in channel density and im-proved hydrology, the stressor index is expected to improve imme-diately, although not to the levels of the reference systems. Furtherdesiccation and natural reduction in algal mat is expected as a re-sult of increased drainage, which will only occur over an as yet un-known period of time, presumably at least several months,following the restoration of hydrology and the re-introduction ofgrazers. The relative rates of recovery present many possibilitiesfor recovery trajectory, all of which are desirable. A problem occursif the attribute stabilizes outside of the acceptable restoration tar-gets, or if it begins to decline toward a degraded state. For example,Harris et al. (1985) described a phenomenon in which carbonatecrusts formed in the intertidal zone as a result of repeated wet/dry cycles. Such crusts have been observed on tidal channel bankswithin the study region and appear to limit the habitat availablefor burrowing organisms and also provided hard substrate for theestablishment of sessile species (particularly Euraphia sp.) thatare foreign to a soft sediment environment. The development of ce-mented crusts post-restoration would drive the trajectory towardsan undesirable endpoint. The rapidity of detecting changes in sta-tus of a restoration site depends on the rate of natural regenera-tion. In our case, evidence from preliminary trials suggested thatinfauna rapidly recolonize, particularly along channels, but therewas little evidence for a rapid recovery of the perennial vegetation,possibly due to degraded hydrology or the establishment of lami-nated forms of algal mat. Similarly, attributes based largely oninfaunal evidence are expected to change more quickly than attri-butes based on perennial vegetation.

4.3. Natural recovery

As presented, these MMIs represent the components of ecosys-tem recovery that can be attributed to anthropogenic contribu-tions. The set-aside sites are particularly important, since theywill continue to move along their own, natural trajectory (notnecessarily toward the restoration target) independent of the sitesaffected by remediation and restoration activities. The seasonal re-scaling that is performed during the creating of the MMIs creates amoving bar against which progress is measured, however, as longas the set-asides and restoration sites were similar with respect totheir conditions prior to the restoration activities, this simply re-sults in the natural recovery (or decline) effectively being factoredout. This makes evaluating the anthropogenic contribution to themarsh restoration apparent and assists in answering whether therestoration activities have aided site recovery, since movementwithin the MMI space should be a response to restoration activi-ties. One potential issue with rescaling, however, is that the sea-sonal rescaling is particularly sensitive to measurementvariability within the set-aside and reference sites. Measurementvariability, particularly for the set-aside sites and reference sites,can result in shifting the MMI values several points in either direc-tion, depending on the magnitude of the measurement variability.The mean value for the observed shift for the set-aside and refer-ence sites in the Ecosystem Health MMI to date is 2.2, a measure-ment which can be used to track how much measurement

distortion is present in individual seasons. Increasing the numberof reference sites and set-asides and reducing the measurementvariability through any available means are both recommendedto help deal with this issue.

4.4. SER attribute MMIs

The development of individual MMIs for each restoration attri-bute is designed to provide additional information in the event thata restoration project is underperforming. Indeed, partial successesare the majority result for ecosystem restoration projects (Lock-wood and Pimm, 1999). While it is not feasible to monitor everymetric that could indicate underperformance in a restoration pro-ject, the SER attribute MMIs attempt to ensure that the subset ofmetrics that are monitored over the longterm at least track impor-tant and measurable differences between the impacted and refer-ence systems.

The stakeholders in a restoration project, usually composed ofpeople of disparate backgrounds and education, ultimately deter-mine when a restoration project will take place, the acceptableendpoints for the remediation, and when or if additional measuresneed to be taken to guide the natural development of a restorationsite. The MMI framework developed herein presents restorationstatus in a way that closely matches elements of restoration theorythat are particularly intuitive, including restoration targets and tra-jectories. The breakdown of the whole-ecosystem MMI into com-ponents along the lines of the SER attributes of a restoredecosystem further extends the interpretive capability by preserv-ing the assumptions made from the method of presentation, allow-ing a deeper level of understanding from the stakeholders.

The framework as presented is most applicable to large restora-tion projects due to the requirement of having several reference sitesas well as several set-aside sites to establish the boundaries of the res-toration space. With few set-asides and reference sites, seasonal andmeasurement variability may result in instability in the index values,lessening the usefulness of the index. In practice, it is recommendedthat there be at least four reference and set-aside sites. While projectsof this size are currently rare, the increasing importance of restora-tion and replacement of ecosystems suggests that this frameworkmay have increased opportunities for use in the future.

5. Conclusions

Measurements that are poorly correlated with the identifiedenvironmental stressors, exhibit high degrees of measurement var-iability, or are similarly represented by other metrics are eliminatedduring the filtering process, reducing future monitoring time andeffort.The filtering process identified visual oiling estimates, den-sity measurements of Pirinella, Cerithidea, Suaeda, and Salicornia ascandidates for elimination.The most descriptive measures to datein terms of responsiveness to the stressors are Amphipod density,Nasima density, and estimates of species richness.Undesirabledevelopments during the rehabilitation, such as the developmentof the carbonate crusts, may be considered as deviations from thepreferred trajectory and should trigger an adaptive managementresponseMMIs are particularly useful for their ability to integratemultiple parameters into a cohesive, useful index. Established con-cepts in restoration including trajectory, alternate states, and resto-ration targets can be represented with minimal effort for entirerestoration sites or for specific attributes of restored ecosystems.

Acknowledgements

This study was funded by United Nations Compensation Com-mission awards 5000451 ‘‘Remediation of damage to coastal


resources’’ and 5000456 ‘‘Remediation of damage to marine re-sources’’, and sponsored by the Presidency for Meteorology andEnvironment of the Kingdom of Saudi Arabia. We would like tothank S. Allen, M. Bice, J. Gabriel, A. Kopinski, and T. Minter fortheir assistance in the field program. We also would like to thankJ. Michel and T. Montello for their editing assistance, and D. Littleand M. Guard for their comments and suggestions during thedevelopment of this project.

References

Abuzinada, A.H., Krupp, F., 1994. The status of coastal and marine habitats two yearsafter the Gulf War oil spill. Courier Forschungsinstitut Senckenberg 166, 1–80.

Adam, P., 1990. Salt Marsh Ecology. Cambridge University Press, Cambridge, UK.Allen, E.A., Fell, P.E., Peck, M.A., Gieg, J.A., Guthke, C.R., Newkirk, M.D., 1994. Gut

contents of common mummichogs, Fundulus heterclitus, in a restoredimpounded marsh and in natural reference marshes. Estuaries 17, 462–471.

Al-Thukair, A.A., Al-Hinai, K., 1993. Preliminary damage assessment of algal matsites located in the western gulf following the 1991 oil spill. Mar. Pollut. Bull.27, 229–238.

Al-Zaidan, A.S.Y., Kennedy, H., Jones, D.A., Al-Mohanna, S.Y., 2006. Role of microbialmats in Sulaibikhat Bay (Kuwait) mudflat food webs: evidence from 13Canalysis. Mar. Ecol. Prog. Ser. 308, 27–36.

Barth, H., 2003. The influence of cyanobacteria on oil polluted intertidal soils at theSaudi Arabian Gulf shores. Mar. Pollut. Bull. 46, 1245–1252.

Bejarano, A.C., Michel, J., 2010. Large-scale risk assessment of polycyclic aromatichydrocarbons in shoreline sediments from Saudi Arabia: environmental legacyafter twelve years of the Gulf war oil spill. Environ. Pollut. 158, 1561–1569.

Blocksom, K.A., 2003. A performance comparison of metric scoring methods for amultimetric index for mid-Atlantic highland streams. Environ. Manage. 31,670–682.

Bradshaw, A.D., 1987. The reclamation of derelict land and the ecology ofecosystems. In: Jordan, W.R., Gilpin, M.E., Aber, J.D. (Eds.), RestorationEcology: A Synthetic Approach to Ecological Research. Cambridge UniversityPress, Cambridge, UK, pp. 53–74.

Brawley, A.H., Warren, R.S., Askins, R.A., 1998. Bird use of restoration and referencemarshes within the Barn Island Wildlife Management Area, Stonington,Connecticut. USA. Environ. Manage. 22, 625–633.

Broome, S.W., Seneca, E.D., Woodhouse Jr., W.W., 1988. Tidal salt marsh restoration.Aquat. Bot. 32, 1–22.

Callaway, R.M., King, L., 1996. Temperature-driven variation in substrateoxygenation and the balance of competition and facilitation. Ecology 77,1189–1195.

Chamberlain, R.H., Barnhart, R.A., 1993. Early use by fish of a mitigation salt marsh,Humboldt Bay, California. Estuaries 16, 769–783.

D’Alpaos, A., Lanzoni, S., Marani, M., 2005. Tidal network ontogeny: channelinitiation and early development. J. Geophys. Res. 110, 1–14.

Davis, W.S., Simon, T.P. (Eds.), 1989. Proceedings of the 1989 Midwest PollutionControl Biologists Meeting. United States Environmental Protection Agency,Chicago, Illinois, USA. EPA-905/9-89-007.

Davis, N.M., Weaver, V., Parks, K., Lydy, M.J., 2003. An assessment of water quality,physical habitat, and biological integrity of an urban stream in Wichita, Kansas,prior to restoration improvements (phase I). Arch. Environ. Con. Tox. 44, 351–359.

Dobson, A.P., Bradshaw, A.D., Baker, A.J., 1997. Hopes for the future: restorationecology and conservation biology. Science 277, 515–522.

Downs, C.A., Dillon Jr., R.T., Fauth, J.E., Woodley, C.M., 2001a. A molecular biomarkersystem for assessing the health of gastropods (Ilyanassa obsoleta) exposed tonatural and anthropogenic stressors. J. Exp. Mar. Biol. Ecol. 259, 189–214.

Downs, C.A., Fauth, J.E., Woodley, C.M., 2001b. Assessing the health of grass shrimp(Paleaomonetes pugio) exposed to natural and anthropogenic stressors: amolecular biomarker system. Mar. Biotechnol. 3, 380–397.

Duarte, C.M., 1991. Variance and the description of nature. In: Cole, J.J., Lovett, G.M.,Findlay, S.E.G. (Eds.), Comparative Analyses of Ecosystems. Springer–Verlag,New York, USA, pp. 301–318.

Ehrenfeld, J.G., 2000. Defining the limits of restoration: the need for realistic goals.Restor. Ecol. 8, 2–9.

Emery, E.B., Simon, T.P., McCormick, F.H., Angermeier, P.L., Deshon, J.E., Yoder, C.O.,Sanders, R.E., Pearson, W.D., Hickman, G.D., Reash, R.J., Thomas, J.A., 2003.Development of a multimetric index for assessing the biological condition of theOhio River. T. Am. Fish. Soc. 132, 791–808.

EPA, 1987. Surface water monitoring: a framework for change. Office of Water,Office of Policy Planning and Evaluation, United States EnvironmentalProtection Agency, Washington, D.C., USA.

Getter, C.D., Michel, J., Hayes, M.O., 2005. Natural recovery of salt marshes from the1991 oil spill on the Saudi Arabian gulf coast, in: Proceedings of the 2005International Oil Spill Conference, May 2005. Miami, Florida, USA.

Gómez Gesteira, J.L., Dauvin, J.C., 2000. Amphipods are good bioindicators of theimpact of oil spills on soft-bottom macrobenthic communities. Mar. Pollut. Bull.40, 1017–1027.

Hale, J.H., Cormack, C.D., Cotsapas, L., Montello, T.M., Langman, O.C., Gabriel, J.J.,Michel, J., 2011. Relationships between key indicators of environmental

condition and degrees of oiling in sediments in salt marsh habitats: a balancebetween contamination and ecological recovery by natural processes, in:Proceedings of the 2011 International Oil Spill Conference, May 23-26 2011,Portland, Oregon, USA.

Harris, P.M., Kendall, C.G.St.C., Lerche, I., 1985. Carbonate cementation —a briefreview. In: Schneidermann, N., Harris, P.M. (Eds.), Carbonate Cements, SEPMSpecial Publication, vol. 36. Tulsa, Oklahoma, USA, pp. 79–95.

Hobbs, R.J., Mooney, H.A., 1993. Restoration ecology and invasions. In: Saunders,D.A., Hobbs, R.J., Erlich, P.R. (Eds.), Nature conservation 3: reconstruction offragmented ecosystems, global and regional perspectives. Surrey Beatty andSons, Chipping Norton, New South Wales, Australia, pp. 127–133.

Hobbs, R.J., Norton, D.A., 1996. Towards a conceptual framework for restorationecology. Restor. Ecol. 4, 93–110.

Höpner, T., Al-Shaikh, K.A., 2008. Shoreline bioremediation after the 1991 Gulf waroil spill. In: Abuzinada, A.H., Barth, J., Krupp, F., Böer, B., Abdessalaam, T.Z. (Eds.),Protecting the Gulf’s Marine ecosystems from Pollution. Basel, Switzerland, pp.265–272.

Horne, J.K., Schneider, D.C., 1995. Spatial variance in ecology. Oikos 74, 18–26.Howard, R.J., Mendelssohn, I.A., 1999. Salinity as a constraint on growth of

oligohaline marsh macrophytes. Am. J. Bot. 86, 785–794.Hughes, F.M., Colston A., Mountford, J.O., 2005. Restoring riparian ecosystems: The

challenge of accommodation variability and designing restoration trajectories.Ecol. Soc. 10:12. (online) url: <http://www.ecologyandsociety.org/vol10/iss1/art12/>.

Jones, D.A., Richmond, M.D., 1992. Intertidal and subtidal marine habitat surveys. In:Establishment of a Marine Habitat and Wildlife Sanctuary for the Gulf Region.Final Report for Phase 1, Jubail and Frankfurt, CEC/NCWCD, pp. 134–160.

Jones, H.P., Schmitz, O.J., 2009. Rapid recovery of damaged ecosystems. PLoS ONE 4,e5653.

Jones, D.A., Plaza, J., Watt, I., Al-Sanei, M., 1998. Long-term (1991–1995) monitoringof the intertidal biota of Saudi Arabia after the 1991 Gulf War oil spill. Mar.Pollut. Bull. 36, 472–489.

Karr, J.R., 1981. Assessment of Biotic Integrity Using Fish Communities. Fisheries 6,pp. 21–27.

Kaufmann, P.R., Levine, P., Robison, E.G., Seeliger, C., Peck, D.V., 1999. QuantifyingPhysical Habitat in Wadeable Streams. EPA/620/R-99/003. U.S. EnvironmentalProtection Agency, Washington D.C.

Kendall, C.G.St.C., Skipwith, P.A., 1968. Recent algal mats of a Persian Gulf lagoon. J.Sed. Res. 38, 1040–1058.

Klemm, D.J., Blocksom, K.A., Fulk, F.A., Herlihy, A.T., Hughes, R.M., Kaufmann, P.R.,Peck, D.V., Stoddard, J.L., Thoeny, W.T., Griffith, M.B., Davis, W.S., 2003.Development and evaluation of a macroinvertebrate biotic integrity index(MBII) for regionally assessing mid-Atlantic highland streams. Environ. Manage.31, 656–669.

Kwak, T.J., Zedler, J.B., 1997. Food web analysis of southern California coastalwetlands using multiple stable isotopes. Oecologia 110, 262–277.

Lindig-Cisneros, R., Zedler, J.B., 2002. Halophyte recruitment in a salt marshrestoration site. Estuar. Coast. 25, 1174–1183.

Lockwood, J.L., Pimm, S.L., 1999. When does restoration succeed? In: Weiher, E.,Keddy, P. (Eds.), Ecological Assembly Rules: Perspectives, Advances, Retreats.Cambridge University Press, Cambridge, UK, pp. 363–392.

Madon, S.P., Williams, G.D., West, J.M., Zedler, J.B., 2001. The importance of marshaccess to growth of the California killifish, Fundulus parvipinnis, evaluatedthrough bioenergetics modeling. Ecol. Model 136, 149–165.

Magnuson, J.J., Regier, H.A., Christie, W.J., Sonzogni, W.C., 1980. To rehabilitate andrestore Great Lakes ecosystems. In: Cairns, J., Jr. (Ed.), The Recovery Process inDamaged Ecosystems. Ann Arbor Science, Ann Arbor, Michigan, USA, pp. 95–112.

McCormick, F.H., Hughes, R.M., Kaufmann, P.R., Peck, D.V., Stoddard, J.L., Herlihy,A.T., 2001. Development of an index of biotic integrity for the mid-Atlantichighlands region. T. Am. Fish. Soc. 130, 857–877.

Metcalfe, J.L., Fox, M.E., Carey, J.H., 1984. Aquatic leeches (Hirudenea) asbioindicators of organic chemical contaminants in freshwater ecosystems.Chemosphere 13, 143–150.

Michel, J., Hayes, M.O., Getter, C.D., Cotsapas, L., 2005. The Gulf War oil spill twelveyears later: consequences of eco-terrorism, in: Proceedings of the 2005International Oil Spill Conference, May 2005. Miami, Florida, USA.

Montalto, F.A., Steenhuis, T.S., 2004. The link between hydrology and restoration oftidal marshes in the New York/New Jersey Estuary. Wetlands 24, 414–425.

Mooney, H.A., Hobbs, R. J. (Eds), 2000. Invasive Species in A Changing World. IslandPress, Washington D.C., USA.

Morvan, J., Le Cadre, V., Jorissen, F., Debenay, J., 2004. Foraminifera as potential bio-indicators of the ‘‘Erika’’ oil spill in the Bay of Bourgneuf: field and experimentalstudies. Aquat. Living Resour. 17, 317–322.

Naeem, S., Li, S., 1997. Biodiversity enhances ecosystem reliability. Nature 390, 507–509.

National Research Council, 1992. Restoration of aquatic ecosystems: Science,technology, and public policy. National Academy Press, Washington D.C., USA.

Niemi, G.J., Hanowski, J.M., Lima, A.R., Nicholls, T., Weiland, N., 1997. A criticalanalysis on the use of indicator species in management. J. Wildlife Manage. 61,1240–1252.

Odum, W.E., Odum, E.P., Odum, H.T., 1995. Nature’s pulsing paradigm. Estuaries 18,547–555.

Osgood, D.T., Zieman, J.C., 1998. The influence of subsurface hydrology on nutrientsupply and smooth cordgrass (Spartina alterniflora) production in a developingbarrier island marsh. Estuar. Coast. 21, 767–783.

http://www.ecologyandsociety.org/vol10/iss1/art12/

http://www.ecologyandsociety.org/vol10/iss1/art12/


Palmer, M.A., Poff, N.L., 1997. The influence of environmental heterogeneity onpatterns and processes in streams. J. N. Am. Benthol. Soc. 16, 169–173.

Pandion Technology, Ltd., Research Planning, Inc. 2003. Oiled shoreline survey insupport of the marine and coastal damage assessment. Presidency ofMeteorology and the Environment, Jeddah, Kingdom of Saudi Arabia.

Pearson, D.L., Cassola, F., 1992. World-wide species richness patterns of TigerBeetles (Coleoptera: Cicindelidae): Indicator taxon for biodiversity andconservation studies. Conserv. Biol. 6, 376–391.

Peck, M.A., Fell, P.E., Allen, E.A., Gieg, J.A., Cuthke, C.R., Newkirk, M.D., 1994.Evaluation of tidal marsh restoration: comparison of selected macroinvertibratepopulations on a restored impounded valley marsh and an unimpounded valleymarsh within the same salt marsh system in connecticut. USA. Environ.Manage. 18, 283–293.

Peterson, B.J., Howarth, R.W., Garritt, R.H., 1986. Sulfur and carbon isotopes astracers of salt-marsh organic matter flow. Ecology 67, 865–874.

Phillips, J.D., 1999. Divergence, convergence, and self-organization in landscapes.Ann. Assoc. Am. Geogr. 89, 466–488.

Portnoy, J.W., 1999. Salt marsh diking and restoration: biogeochemical implicationsof altered wetland hydrology. Environ. Manage. 24, 111–120.

Prober, S.M., Thiele, K.R., Lunt, I.D., 2002. Identifying ecological barriers torestoration in temperate grassy woodlands: soil changes associated withdifferent degradation states. Aust. J. Bot. 50, 699–712.

Reeves, G.H., Hall, J.D., Roelofs, T.D., Hickman, T.L., Baker, C.O., 1991. Rehabilitatingand modifying stream habitats. In: Meehan, W.R. (Ed.), Influences of forest andrangeland management on salmonid fishes and their habitats, AmericanFisheries Society Special Publication 19. Bethesda, Maryland, USA, pp. 519–557.

Reynoldson, T.B., 1987. Interactions between sediment contaminants and benthicorganisms. Hydrobiologia 149, 53–66.

Roni, P., Liermann, M., Steel, A., 2003. Monitoring and evaluating responses ofsalmonids and other fishes to instream restoration. In: Montgomery, D.R.,Bolton, S., Booth, D.B. (Eds.), Restoration of Puget Sound rivers. University ofWashington Press, Seattle, Washington, USA, pp. 318–329.

Sabean, J.A.R., Scott, D.B., Lee, K., Venosa, A.D., 2009. Monitoring oil spillbioremediation using marsh foraminifera as indicators. Mar. Pollut. Bull. 59,352–361.

Sacco, J.N., Seneca, E.D., Wentworth, T., 1994. Infaunal community developmentof artificially established salt marshes in North Carolina. Estuaries 17, 489–500.

Shreffler, D.K., Thom, R.M., Scott, M.J., Wellman, K.F., Walters, M.A., Curran, M.,1995. National Review of Non-Corps Environmental Restoration Projects. PacificNorthwest National Laboratory, Richland, Washington. Final Report Preparedfor US Army Corps of Engineers, Institute for Water Resources, Alexandria, VA,IWR Report 95-R-12.

Simenstad, C.A., Thom, R.M., 1996. Functional equivalency trajectories of therestored Gog-Le-Hi-Te estuarine wetland. Ecol. Appl. 6, 38–56.

Society for Ecological Restoration, 2004.The SER international primer on ecologicalrestoration, version 2. Retrieved December 4, 2010, from http://www.ser.org.

Stoddard, J.L., Herlihy, A.T., Peck, D.V., Hughes, R.M., Whittier, T.R., Tarquinio, E.,2008. A process for creating multimetric indices for large-scale aquatic surveys.J. N. Am. Benthol. Soc. 27, 878–891.

Tawfiq, N.I., Olsen, D.A., 1993. Saudi Arabia’s response to the 1991 gulf oil spill. Mar.Pollut. Bull. 27, 333–345.

Tilman, D., 1996. Biodiversity: population versus ecosystem stability. Ecology 77,350–363.

Walker, B., Kinzig, A., Langridge, J., 1999. Plant attribute diversity, resilience, andecosystem function: the nature and significance of dominant and minor species.Ecosystems 2, 95–113.

Warren, R.S., Fell, P.E., Rozsa, R., Brawley, A.H., Orsted, A.C., Olson, E.T., Swamy, V.,Niering, W.A., 2002. Salt marsh restoration in connecticut: 20 years of scienceand management. Restor. Ecol. 10, 497–513.

Weisburg, S.B., Lotrich, V.A., 1982. The importance of an infrequently floodedintertidal marsh surface as an energy source for mummichog Fundulusheteroclitus: an experimental approach. Mar. Biol. 66, 307–310.

Yachi, S., Loreau, M., 1999. Biodiversity and ecosystem productivity in afluctuating environment: the insurance hypothesis. P. Natl. Acad. Sci. USA96, 1463–1468.

Zedler, J.B. (Ed.), 2001. Handbook for restoring tidal wetlands. CRC Press, BocaRaton, Florida, USA.

Zedler, J.B., Callaway, J.C., 1999. Tracking wetland restoration: do mitigation sitesfollow desired trajectories? Restor. Ecol. 7, 69–73.

http://www.ser.org

Developing multimetric indices for monitoring ecological restoration progress in salt marshes

Documents