Coral reefs worldwide are affected by excessive sediment and nutrient delivery from adjacent watersheds. Land cover and land use changes contribute to reef ecosystem degradation, which in turn,diminish many ecosystem services, including coastal protection, recreation, and food provisioning. The objectives of this work were to understand the role of coastal oceanic and biophysical processes in mediating the effects of sedimentation in shallow reef environments, and to assess the efficacy of land-based sediment remediation in the coastal areas near Maunalei reef, Lāna’i Island, Hawai’i. To the best of our knowledge, this was the first study of sediment dynamics on an east-facing (i.e., facing the trade winds) reef in the Hawaiian Islands. We developed ridge-to-reef monitoring systems at two paired stream bed-to-reef sites, where one of the reef sites was adjacent to a community stream sediment remediation project. We found that the two reef sites were characterized by different processes that affected the sediment removal rates; the two sites were also exposed to different amounts of sediment runoff. The community stream sediment remediation project appeared to keep at least 77 tonnes of sediment off the reef flat in one wet season. We found that resuspension of sediments on this reef was similar to that on north-facing and south-facing reefs that also are exposed to the trade winds. We posit that sites with slower sediment removal rates due to slower current velocities or high resuspension rates will require more-robust sediment capture systems on land to reduce sediment input rates and maximize potential for reef health recovery. This suggests that interventions such as local sediment remediation and watershed restoration may mitigate sediment delivery to coral reefs, but these interventions are more likely to be effective if they account for how adjacent coastal oceanographic processes distribute, accumulate, or advect sediment away from reefs. Our results on the effectiveness of gabion dam sediment capture may help guide scalable solutions for erosion control on islands.

## Introduction

Coral reefs across the globe are affected by land-based sediment, which can increase turbidity and, in turn, smother corals and reduce light levels. Sedimentation may result in reduced coral growth [1,2] and recruitment [3,4,5], reduced photosynthesis rates in coral symbionts [6,7,8], and prevention of waste removal and gas exchange in corals [9]. Effects on reef fishes can include changes in physiological and ecological processes, including reduced foraging efficiency [10], prolonged larval development [11], and changes in use of coral reefs [12].

The likelihood of effects from sedimentation is higher for reefs in poorly flushed locations than in well-flushed locations [6]. Water circulation in coral reefs is highly variable in space and time, which leads to large variation in where land-sourced and reef-sourced sediments are deposited, how long sediment remains, how often it is resuspended, and its movement across the reef [13]. These spatial and temporal variations affect the distributions of coral species that have different levels of tolerance to sediment exposure [14,15,16].

In Hawai’i and other islands of steep, short montane watersheds, there is broad recognition that land and watershed management affects inputs of nutrients and sediments [17,18]. Requirements for mitigating effects of land-based sediments have been developed and implemented in such settings in Hawai’i through partnerships among state agencies, federal agencies, and local community watershed alliances[19]. These efforts identify major sources of terrestrial sediment and develop practices for land management to reduce sedimentation [20].

In north-facing Hanalei Bay, Kauai, for example, river plume sediments draining into the bay during the summer are displaced by wave driven resuspension, which intensifies and prolongs the potential effects of the plume itself. On the west-facing reefs of the western Maui, Hawai’i coast, waves and winds drive water flow in the shallow areas, and tides modulate currents in the deeper parts of the reef [21,22]. Along the south-facing coast of Molokai, Hawai’i, waves and tidally driven flows modulate the sediment advection across the reef flat; during rising tides, waves driven by trade winds blowing across the east-west trending reef flat lead to sediment resuspension, which sometimes can drain off the reef on ebb (falling) tides [23].

We focused on an east-facing reef and coastal area on Lāna’i Island, Hawai’i,where a local community organization (Maunalei Community Managed Makai Area) had developed a sediment remediation and stream restoration project in the ephemeral Maunalei stream that employed gabion dams. We also evaluated the efficiency of these mesquite check dams as sediment remediation mechanisms. This was a collaboration involving a non-governmental organization (Conservation International), community organization on Lāna’i Island (Maunalei Community Managed Makai Area), academic scientists (University of Hawaii-Manoa), and the Pacific Islands Ocean Observing System (a network of ocean monitoring systems for Pacific Islands). Our objectives were to assess spatial patterns of total suspended sediment (TSS) and investigate factors that affect the variability of TSS loads; to compare TSS and hydrodynamics at two reef flats in eastern Lāna’i, one of which receives substantial amounts of terrigenous sediment during the winter rainy season; and to estimate water and sediment residence time at those two sites. Such data are critical for informing local stream restoration and sediment remediation.

## Methods

### Study site

Lāna’i (20.8971°N, 156.8740°W) is the sixth largest of the main Hawaiian Islands, with an area of 364 km2 (Fig. 1A). Lāna’i is located in the rain shadow of eastern Molokai and western Maui and receives about 270–1,400 mm of rain per year, with rainfall increasing as elevation increases; this rainfall is roughly 10 times less than that observed at the highest elevations of Molokai and Maui(Supplementary Material [SM], Section 1, Fig. S1) [24]. Across the Hawaiian Islands, flash floods occur most frequently in the winter and spring (October through April) [25]. Lāna’i has a long history of land use, from its early occupation by native Hawaiian societies to extensive agriculture during the colonial era. Currently, about 3,100 people live on the island [26]. Watersheds on Lāna’i and in Hawai’i, overall, are very short (on average < 10 km) compared to continental watersheds and catchments. Therefore, streams in semi-arid areas of this region are ephemeral and are prone to flash flooding, as rainfall can cause stream flow to initiate and increase rapidly.

Figure 1

Field study location: (A) Lāna’i Island within the Hawaiian Islands (inset), (B)Maunalei in east Lāna’i and study sites. Black boxes on land indicate where pressure sensors were deployed. (C) Distances between pressure sensors. Larger black boxes on the reef flat indicate the locations where we monitored water quality and current velocity. White boxes indicate the locations for which we developed models of sediment removal. Imagery accessed through Google Earth TM.

Figure 1

Field study location: (A) Lāna’i Island within the Hawaiian Islands (inset), (B)Maunalei in east Lāna’i and study sites. Black boxes on land indicate where pressure sensors were deployed. (C) Distances between pressure sensors. Larger black boxes on the reef flat indicate the locations where we monitored water quality and current velocity. White boxes indicate the locations for which we developed models of sediment removal. Imagery accessed through Google Earth TM.

We conducted our study in the eastern portion of the island, which faces the trade winds. On land, we worked at two ephemeral streambeds (Site A-SS [Stream Site] and Site B-SS [Stream Site]),and in the ocean, we worked on two reef flats adjacent to the streambeds (Site A-RF [Reef Flat] and Site B-RF [Reef Flat]) (Fig. 1B).

Because no measurable sediment had been delivered by the stream into the reef flat at A-RF, we treated it and A-SS as controls. At B-SS, the community organization (Maunalei Community Managed Makai Area) constructed seven gabion dams of invasive mesquite (Prosopis pallida;kiawe [Hawaiian language]) logs and branches across the dry streambed (Fig. 2B). The gabions were staked in place from January–March 2014 to capture sediment that would otherwise flow into the coastal ocean.

Figure 2

(A) Gabion dams, made from invasive kiawe branches, in ephemeral stream site. Photo credit:Kolomona Kaho’ohalahala. (B) Example of HOBO Onset pressure sensor attached to streambed with pipe at Site B (between gabion dams). Photo credit: Lida Teneva.

Figure 2

(A) Gabion dams, made from invasive kiawe branches, in ephemeral stream site. Photo credit:Kolomona Kaho’ohalahala. (B) Example of HOBO Onset pressure sensor attached to streambed with pipe at Site B (between gabion dams). Photo credit: Lida Teneva.

### Instrument deployments

We deployed seven Onset HOBO pressure sensors (Bourne, MA, USA) from January 10–March 17,2014 between the gabions in Maunalei stream (Site B). We deployed three sensors in the streambed at Site A at the same elevations as the upstream, midstream, and downstream sensors in Maunalei stream(Fig. 1C; SM, Section 2, Table S1). These pressure sensors were deployed to monitor the resulting stream flows during rain events. Here, we define rain events as precipitation over the course of 24 hours, and any resulting flash floods that flowed through the streambed. We took temperature and pressure readings at 300-s (5-min) intervals. After recovering the sensors, we recorded the distance between the streambed measurement markings on the sensors (see SM, Section 2) to estimate the sediment accumulation at each sensor. We aimed to test whether sediment accumulation varied as a function of distance from the furthest-upstream gabion dam.

We deployed two types of water quality sensors (a Sea Bird Electronics [Bellevue, WA, USA] 19plus CTD meter with a WETlabs C-Star transmissometer and WETstar flourometer) at Site A[20.895136°N, 156.87099°W] and a SBE16 with a Wetlabs ECOFLNTUS at Site B[20.89880°N, 156.87485°W]) at Site B-RF (Fig. 1B)from 10 January–17 March 2014 (wet season) and again from 5 June–4 August 2014 (dry season) to measure temperature, salinity, and light transmittance at a 600-s (10-min) interval. These were the instruments available for the study, and purchase of new equipment to duplicate instrument packages was beyond our budget. The transmissometer reports transmittance, and the ECOFLNTUS yields outputs in nephelometric turbidity units (NTUs). We took 10 measurements per sample and delayed 5 s before sampling. Light transmittance can be used as a proxy for turbidity. The sensor at the site A was deployed in a channelized, coral-dominated region; the sensor at site B was deployed in a wide sandy area. We deployed the sensors to monitor whether the variables changed on the reef as a result of a rain event recorded by the pressure sensors on land. We also deployed Nortek Aquadopp current profilers (Rud, Norway) at Site A-RF and Site B-RF from 5 June–4 August 2014 and recorded a 1-min average of current velocity once every 10 min (Fig. 1B). We deployed a solar-powered Davis Weather Monitor II station(Hayward, CA, USA) with an anemometer and wind-vane on an adjacent ridge (20.8943°N,156.8763°W) (Fig. 1B). Wind was recorded as an hourly average of measurements taken at 30-s intervals.

We installed tubular sediment traps (6.8 cm diameter with 1 cm diameter baffles on the open end)next to the other instruments at Site A-RF and Site B-RF from June 5–August 4, 2014. After recovery of the traps, we processed the sediments and measured their inorganic carbon(CaCO3) and total carbon and nitrogen. To determine the potential sources of organic carbon, we measured δ13C and δ15N (refer to SM, Section 3, Fig. S6,Fig. S7).

### Sediment analysis

We corrected sediment data for drift from biofouling by adjusting values on the basis of a linear fit of light transmittance over time. After conversion to TSS, the results of the linear correction were closer to what was measured in the field, i.e., through the bottle sample data. Therefore, we proceeded with linear fit. Furthermore, the most conservative way to address fouling was to assume that the fouling was linear. We calibrated between the transmissometer measurements of light attenuation and the sediment concentration in the reef waters (i.e., TSS), which allowed us to report data in TSS units (for more details, refer to SM, Section 4).

### In-stream water volume and sediment volume

Because no velocity measurements were available, we used the Gauckler-Manning formula to estimate water velocity as $V=knRh2/3S1/2$

$V = \frac{k}{n}R_h^{2/3}{S^{1/2}}$
⁠, where V is velocity, k/n is a constant coefficient, k is a conversion factor between SI and English units, n is the Gauckler-Manning coefficient, Rh is the hydraulic radius(area: wet perimeter), and S is the slope. The constant k/n varies between 20-80 m1/3s–1 as a function of surface roughness; we used a mean value of 50 m1/3s–1 as our constant [27]. We used a slope of 0.1 (taken from a topographic map). Our initial calculation of water volume was based on data from sensor 13 and assumed a constant water height (21 cm), cross sectional area (0.38 m2), and velocity (1.06 m s–1) over a 2-hour period.

For sediment volume calculations, we calculated the cross sectional area at a sensor with the measured channel width and maximum water level, and assumed that the channel was triangular. This allowed for calculation of a slope angle (which we assumed was constant), and for the area to be expressed as a function of water level. This method also allowed us to determine how much of the total sediment volume was captured between each of the gabion dams.

### Statistical analyses

We calculated a spectral density for each dataset (i.e., salinity, temperature, current speed,pressure, and wind speed) with the Welch method [28]. We then calculated the correlation between the frequencies at which the peaks in each spectrum occurred in relation to the primary 12-hr tidal frequency. We tested the hypothesis that TSS varies as a function of temperature, pressure, chlorophyll a, salinity, current speed and direction, wind speed and direction. In addition, we tested whether TSS varied as a function of time (cumulative hours) to help capture temporal autocorrelation in the time-series, and as a function of site to represent any site-specific differences not captured in the other predictors. We fit a generalized linear model(GLM) with an inverse link function to allow for the positive, continuous response of the TSS variable. We also tested link functions of “log” and “identity” and assessed the behavior of the resulting residuals with reference to model assumptions. We used Pearson’s correlation and variance inflation factors (VIFs) to test whether predictors were colinear. We removed highly colinear predictors and refit the model until all VIF < 3 [29]. The final set of predictors was temperature, pressure, salinity, current direction, wind speed, wind direction, and time.

We fit models with all combinations of main effects before testing for 2-, 3-, and 4-way interactions. We selected models with Akaike’s Information Criterion adjusted for small sample sizes (AICc). We assessed error distributions and the linearity of relations in the best (lowest AICc) model and models with AICc values within 2 units of the model with the lowest AICc. We tested the assumption of linearity by comparing the best-fit GLM to a generalized additive model (GAM) that included the same main effects (interactions cannot be included in additive models). We used hierarchical partitioning to estimate the variance in all hierarchical models explained by each main effect. We conducted all analyses in R [30] with the“car” [31], “AICcmodavg” [32], “hier.part” [33], “dplyr” [34], “mgcv”[35], and “lmtest” [36] packages. Significance levels were 0.05.

We developed a box model to estimate sediment transport at each site. An imaginary 400 m ×400 m box (approximately the distance from the shore to the edge of the reef) was centered on each sensor location (Fig. 1B). We assumed an average water height of 1 m, no shear throughout the water column and homogeneous flow across the designated planar area,and that the sediment was carried across the box at the same velocity as the depth-integrated Lagrangian transport (i.e., the sum of the Eulerian and Stokes drift transports [37]). We assumed the dispersive flux was negligible and that the advective flux predominated [38]. For each site, we then calculated the flux across a 400-m2 section perpendicular to the shore by multiplying the 400-m2 area by the average sediment concentration and the average alongshore velocity component for that site. This is a first-order, rapid-assessment box model. The model could be improved if a greater number of current profilers were available. The data from these profilers could be used to quantify the cross-shore heterogeneity in along-shore flow speeds and the variation in suspended-sediment concentrations [39]. Hence, the simplistic box model developed here was based on the best data available for these sites. No other physical oceanography data have been published for Lāna’i Island reefs.

We used data generated by the Simulating Waves Nearshore (SWAN) model [40] and provided by the University of Hawaii to assess wave direction over the summer deployment.

## Results

### Maunalei stream and reef flat during the wet season (January–March 2014)

Stream depth at Site B-SS during a rain event on 3 February 2014 reached a depth of 0.3 m to 0.5 m in less than an hour; water flowed in the stream for a few hours during this event and during a 3 March event (Fig. 3). The water level at the peak of the rain event on 3 February was higher at the upstream sensors (1, 2, 3, 4), than the downstream sensors (5,6, 7, Fig. 3).

Figure 3

Time-series of pressure data from Site A-SS pressure sensors during rain events on 3 February 2014 (A) and 3 March 2014 (B). At each sensor location we measured both the absolute distance across the channel and the distance along the ground.

Figure 3

Time-series of pressure data from Site A-SS pressure sensors during rain events on 3 February 2014 (A) and 3 March 2014 (B). At each sensor location we measured both the absolute distance across the channel and the distance along the ground.

In-stream pressure data indicated that the rain events during February and March 2014 induced flash floods in Site B-SS (Fig. 4B), but no flow occurred in the control streambed, Site A-SS (Fig. 4A). The sediment deposition and maximum water depth during the rain event did not consistently decrease as distance downstream increased, but were highest in the furthest-upstream sensor (#1) at Site B-SS (Fig. 3; Table 1, SM, Section 5,Table S2). We estimated that 55.4 m3 (77.5 tonnes) of sediment was captured. Cumulatively, the first and second gabion dam trapped 88.5% of the total captured sediment (Table 2). Total water volume through the channel was estimated to have been ~350 m3 in under an hour for the 3 March 2014 event, and over 2800 m3 in less than 2h for the 3 February 2014 event.

Figure 4

(A) No flow through Site A-SS resulted in ambient air pressure recorded in the pressure sensors during the wet season. (B) Rain events appear as spikes above the ambient air pressure for Site B-SS as flash floods occur. The gradual increase in mean ambient air pressure is observed at both study sites and is interpreted as an influence from atmospheric variability, e.g., a high pressure system in the region.

Figure 4

(A) No flow through Site A-SS resulted in ambient air pressure recorded in the pressure sensors during the wet season. (B) Rain events appear as spikes above the ambient air pressure for Site B-SS as flash floods occur. The gradual increase in mean ambient air pressure is observed at both study sites and is interpreted as an influence from atmospheric variability, e.g., a high pressure system in the region.

Table 1

Sediment accumulation at gabion stream sensors on March 17, 2014.

 Sensor (upstream to downstream) #1 #2 #3 #4 #5 #6 #7 Sediment accumulation (cm) 21.6 10.5 1.27 2.22 2.54 1.27 0.32
 Sensor (upstream to downstream) #1 #2 #3 #4 #5 #6 #7 Sediment accumulation (cm) 21.6 10.5 1.27 2.22 2.54 1.27 0.32
Table 2

Total amount of sediment trapped by each mesquite gabion dam from January–March 2014.

SensorArea (m2)Sediment trapped (m3)

#1 166 35.4
#2 133 13.6
#3 63.7 0.605
#4 75.4 1.43
#5 140 3.12
#6 127 1.21
Total 55.4
Total #1 & #2 49.0 (88.5%)
SensorArea (m2)Sediment trapped (m3)

#1 166 35.4
#2 133 13.6
#3 63.7 0.605
#4 75.4 1.43
#5 140 3.12
#6 127 1.21
Total 55.4
Total #1 & #2 49.0 (88.5%)

During the rain events on 3 February and 3 March 2014, Site B-RF had higher levels of turbidity(NTU) than Site A-RF (SM, Section 6, Fig. S2, Fig. S3, Fig. S4). Site A-RF has higher current speeds, higher transmittance %, and lower TSS than Site B-RF, and similar salinity and temperature(Table 3).

Table 3

Maxima, minima, and means for measured properties during winter (10 January–17 March, 2014)and summer (June 5–August 4, 2014) in-water instrument package deployment at reef flat Site A and Site B.

Winter

Control (Site A-RF)Gabion (Site B-RF)

Temperature (°C) 22.6 28.4 25.1 21.7 30.6 25.2
Pressure (dB) 0.4 1.4 0.8 0.2 1.3 0.6
Transmittance (%) 94.3 69.7
Turbidity (NTU)    1.0 34.3 3.8
Salinity 32.7 35.4 35.0 32.1 35.5 34.8
Summer

Min Max Mean Min Max Mean

Temperature (°C) 24.0 29.4 26.2 23.8 29.6 26.3
Pressure (dB) 0.7 1.7 1.1 0.9 1.4
Transmittance (%) 28.0 93.4 78.6 5.2 56.3 43.9
TSS (mg L–12.6 12.0 4.6 4.7 31.8 6.5
Salinity 33.9 35.2 35.0 34.8 35.3 35.1
Current Speed (m s–10.02 0.54 0.30 0.10 0.03
Current Direction (°)   240   317
Wind Speed (m s–16.3 4.0
Wind Direction (°)   245
Winter

Control (Site A-RF)Gabion (Site B-RF)

Temperature (°C) 22.6 28.4 25.1 21.7 30.6 25.2
Pressure (dB) 0.4 1.4 0.8 0.2 1.3 0.6
Transmittance (%) 94.3 69.7
Turbidity (NTU)    1.0 34.3 3.8
Salinity 32.7 35.4 35.0 32.1 35.5 34.8
Summer

Min Max Mean Min Max Mean

Temperature (°C) 24.0 29.4 26.2 23.8 29.6 26.3
Pressure (dB) 0.7 1.7 1.1 0.9 1.4
Transmittance (%) 28.0 93.4 78.6 5.2 56.3 43.9
TSS (mg L–12.6 12.0 4.6 4.7 31.8 6.5
Salinity 33.9 35.2 35.0 34.8 35.3 35.1
Current Speed (m s–10.02 0.54 0.30 0.10 0.03
Current Direction (°)   240   317
Wind Speed (m s–16.3 4.0
Wind Direction (°)   245

### Hydrodynamics during the dry season (June–August 2014)

The dominant winds were onshore, from the northeast, with a peed of 4 m s–1± 1.1 (mean ± SD) (Table 3; SM, Section 7, Figure S5). On the eastern shore of Lāna’i, wind velocities during the day were higher than during the night, likely reflecting a solar insolation-driven pressure differential. The SWAN model indicated that wave direction in the study area from June–August 2014 was consistently from 40–80° (E-NE), generally perpendicular to the shore, and in line with trade winds. This wave directionality was consistent with the average current direction of 240° (towards shore)at Site A-RF with a speed of 0.3 m s–1 ± 0.1 (mean ± SD). The sensor at Site A-RF was deployed in a channelized, coral-dominated region, which explains the relative unidirectionality of the flow (Fig. 5A). At Site B-RF, current velocity of 0.3 m s–1 ± 0.02 (mean ± SD) in a 317° northwest direction (Table 3; Fig. 5B). The sensor at site B was deployed in a wide sandy area, unlike the sensor at Site A,and no single flow direction was dominant. The pressure record from the reef flat at Site A and Site B revealed patterns common to semidiurnal mixed tides. Depth and tidal ranges were 1.1 m(0.6–1.8 m) for Site A and 1.4 m (1–2 m) for Site B. Light attenuation peaked during spring tides (Fig. 6A,B).

Figure 5

Current vectors at Site A-RF (A) and Site B-RF (B) in m s–1. Red arrows represent the average current magnitude and direction over the sample time frame in summer 2014. Average current direction for Sites A-RF and B-RF were 240° and 317°, respectively.

Figure 5

Current vectors at Site A-RF (A) and Site B-RF (B) in m s–1. Red arrows represent the average current magnitude and direction over the sample time frame in summer 2014. Average current direction for Sites A-RF and B-RF were 240° and 317°, respectively.

Figure 6

Water depth [red] and light attenuation [blue], which we used as a proxy for TSS, at Site A-RF(A) and Site B-RF (B) between 5 June and 4 August 2014.

Figure 6

Water depth [red] and light attenuation [blue], which we used as a proxy for TSS, at Site A-RF(A) and Site B-RF (B) between 5 June and 4 August 2014.

### Statistical analyses

The best model of TSS (no other models were within 2 AICc) included pressure, salinity, wind speed, time, and all possible 2-, 3-, and 4-way interactions. The large number of interaction effects is likely due to the high correlation typical of environmental time-series. The AICc value of the GLM was 89 units less than that of the GAM. The GLM was statistically significant(p < 0.001) and explained 81% of the deviance in TSS (Table 4).

Table 4

Partitioning of deviance explained by each effect or interaction (*).

Effect or interactionExplained devianceHierarchical partitioning
(main effects only)

Pressure [dB] 0.63 0.81
Salinity 0.011 0.04
Wind speed [m s–1< 0.01 0.05
Time [hrs] 0.02 0.10
Pressure * Salinity 0.04
Pressure * Wind speed 0.01
Salinity * Wind speed < 0.01
Pressure * Time 0.03
Salinity * Time < 0.01
Wind speed * Time < 0.01
Pressure * Salinity * Wind speed 0.02
Pressure * Salinity * Time < 0.01
Pressure * Wind speed * Time 0.05
Salinity * Wind speed * Time < 0.01
Pressure * Salinity * Wind speed * Time < 0.01
Effect or interactionExplained devianceHierarchical partitioning
(main effects only)

Pressure [dB] 0.63 0.81
Salinity 0.011 0.04
Wind speed [m s–1< 0.01 0.05
Time [hrs] 0.02 0.10
Pressure * Salinity 0.04
Pressure * Wind speed 0.01
Salinity * Wind speed < 0.01
Pressure * Time 0.03
Salinity * Time < 0.01
Wind speed * Time < 0.01
Pressure * Salinity * Wind speed 0.02
Pressure * Salinity * Time < 0.01
Pressure * Wind speed * Time 0.05
Salinity * Wind speed * Time < 0.01
Pressure * Salinity * Wind speed * Time < 0.01

Because we retained outliers on the basis of their ecological relevance to our hypotheses,residuals were significantly different from normal regardless of the link function used (results given used an inverse link function). In addition, autocorrelation was present in the residual time-series, particularly at Site A-RF. The autocorrelation seemed to represent tidally influenced processes with a period of about 12hrs.

The diel range in temperature within each site during both seasons (6–9°C) exceeded the seasonal difference between the sites (1.1°C), which is typical of shallow tropical reefs(Fig. 7A; Table 3). Salinity at both Site A-RF and Site B-RF was close to oceanic with the exception of the two storm events (3 February and 3 March), during which salinity values declined by as much as 2.5 psu. (Fig. 7B). We recorded simultaneous high winds and high turbidity at both sites during the winter (SM, Section 6, Fig. S3, Fig. S4). Even in the summer dry season, TSS was higher at Site B, which received a higher sediment load during the wet season. Mean TSS at the Site A-RF was 4.59 mg L–1 (range 2.55–12.04 mg L–1), whereas the mean TSS at Site B-RF was 6.46 mg L–1 (range 4.69–31.84 mg L–1) (Fig. 7C).

Figure 7

(A) Temperature (°C), (B) Salinity (PSU), and (C) Total Suspended Solids (TSS) (mg L–1) for the reef flat study site A (blue) and site B (red) from 5 June–4 August 2014.

Figure 7

(A) Temperature (°C), (B) Salinity (PSU), and (C) Total Suspended Solids (TSS) (mg L–1) for the reef flat study site A (blue) and site B (red) from 5 June–4 August 2014.

We used the longer time series from the summer as input to the one-dimensional box model, which yielded an outward flux of 0.1 kg s–1 at Site A-RF and 0.03 kg s–1at Site B-RF. The amount of time it would take to move 77.5 t of sediment (the amount estimated to have been trapped by the gabions during a winter rain event) out of the box at the control site(Site A-RF) (i.e., the rate of sediment removal from the site) would be 8 d, whereas the removal rate for the same amount of sediment across the same distance at the gabion site (Site B-RF) would be 35 d (Table 5). Thus, the removal rate at Site B-RF was about 4.5 times slower than that at Site A-RF for the same theoretical input of sediment.

Table 5

Site, sediment concentration, current velocity, and flux.

SiteSediment concentration
(mg L–1)
Alongshore velocity
(m S–1)
Outward flux
(kg s–1)
Days to remove 77.5 t of sediment

4.59 0.06 0.115 7.8
6.46 0.01 0.026 34.5
SiteSediment concentration
(mg L–1)
Alongshore velocity
(m S–1)
Outward flux
(kg s–1)
Days to remove 77.5 t of sediment

4.59 0.06 0.115 7.8
6.46 0.01 0.026 34.5

Over the course of the 60-d deployment, Site B-RF collected 152.5 cm3 (wet volume) of sediment, and the control site, Site A-RF, collected 214.3 cm3. From the total amount of sediment collected at Site A-RF and Site B-RF, we estimated a gross annual accumulation rate of 1,286 cm3 (455 g) and 915 cm3 (324 g) per 36.3 cm2 (planar area of sediment traps) for Sites A-RF and B-RF, respectively. This translates to 126.4 kg m–2 y–1 (346.2 g m–2 d–1) for Site A-RF and 90 kg m–2 y–1 (246.6 g m–2d–1) for Site B-RF.

We estimated net accumulation rates from 11.5–17.3 g m–2d–1 for Site A-RF and 8.2-12.3 g m–2 d–1 for Site B-RF. Percent weight of calcium carbonate (CaCO3) in the sediments, which we assumed to be marine-derived, was 39.1% for Site A-RF and 39.3% for Site B-RF (for more details, SM, Section 3). The fractions of inorganic carbon and organic carbon, 4.7% and 1.8%, respectively, were the same at both sites. Terrigenous sediment was not deposited at either site during the dry summer season,which could explain their similar chemical compositions. Although we did not deploy sediment traps during the wet winter season, we expect the organic carbon content in the sediment to be higher at Site B-RF during that time of year due to substantial sediment influx.

## Discussion

Our research suggests that 1) the rate of sediment removal from the reef, i.e., reef exposure to sediment, depends on proximity to terrestrial sediment sources and coastal current dynamics;sediment resuspension can increase the time of local sediment retention from 35 days to 18 months,2) the rate of net sediment accumulation at Site B-RF is higher than that in some other locations in the Hawaiian Islands, 3) reef sites in east Lāna’i within 1 km can have very different sediment and current dynamics, and 4) streambeds adjacent to the east Lāna’i coastline and within 1 km can have different hydrological characteristics, including exposure to winter flash floods.

We found that a community-level effort to remediate sediment measurably reduced the amount of sediment reaching reefs. Our partnership among a community organization (Maunalei Community Managed Makai Area), environmental nonprofit organization (Conservation International), the University of Hawaii, and the Pacific Islands Ocean Observing System demonstrated that insights from studies of sediment movement on reefs close to sediment plume sources could inform future land-based sediment remediation activities on this island or other locations across the Hawaiian Islands.

### Physical forcing in coral reefs

Resuspension repeatedly exposes the same reef area to the same amount of sediment, amplifying the effect of a unit of sediment load by an order of magnitude in some areas [41]. Not only sediment concentration but also sediment duration (i.e., residence time) and frequency of sediment flux events affect responses of reefs to sedimentation [2]. Once land-based sediment reaches the reef, water circulation driven by the interaction among tides, currents, winds, coastline topography, and bottom relief control sediment removal and transport [22,39,43,44,45,46]. We found that sediment movements differed substantially at two locations on the Maunalei reef on Lāna’i Island and that these movements were affected by physical factors, including the combined effects of current flow, tides, and, likely, bottom relief of the reef (for a detailed discussion on the role of waves, see SM, Section 8).

The hydrodynamic regime on the reef flat modulates the sediment removal rate once terrigenous sediment reaches the reef and can strongly affect how rapidly a portion of the reef can recover from sediment influx. The relative difference between sediment removal rates at the different sites is likely to be an underestimate. Addition of a second dimension to the box model, such as constraints on sediment accumulation on the bottom, would improve understanding of the controls on sediment removal rates at these sites. True net accumulation rate can be 20–30 times smaller than the gross accumulation rates estimated by canonical sediment traps [41]. If the ratio of net accumulation rate to gross accumulation rate can be used as a coefficient for residence time of sediment in the water column on a reef that experiences resuspension, the removal time of sediment might be much longer than 35 d, perhaps as much as 30 times longer (or ~18 mo.).

Coral cover and coral assemblages are affected by physical processes and land use [46], and the latter can alter physical and ecological interactions typical of reef systems [47]. Reefs can be exposed to high turbidity and sediment loads even when adjacent to catchments with little current human use [45]. Some coral reefs can persist with high turbidity for weeks to months by increasing their non-photosynthesis-dependent feeding [48,49]. Because the high sediment load and turbidity conditions on Lāna’i are believed to be relatively recent (since the early 20th century rather than on geologic time scales), we suggest that the Lāna’i reefs are not adapted to such high levels of sediment load and likely will be negatively influenced. However, data are too scarce to identify whether sediment load thresholds exist for many reefs,especially in relation to sediment exposure, which is a function of the relations between the amount of sediment delivered to the reef area and the residence time, which is controlled by currents.

### Sedimentation on the reef

The sediment accumulation values we computed are much higher than the sediment trap gross accumulation rates estimated for a site in Hanalei Bay, Kauai (67–172 g m–2d–1) [41]. If the true net accumulation rate is 20–30 times smaller than gross accumulation rates estimated by canonical sediment traps, as suggested by novel sediment traps, SedPods [41], net accumulation rates would be 11.5–17.3 g m–2 d–1 for Site A-RF and 8.2–12.3 g m–2 d–1 for Site B-RF. These rates also are higher than the net accumulation rates from Hanalei Bay, Kauai (3.5–6 g m–2 d–1) [41]. Although these calculations provide information about exposure to sediment, grain size is also relevant to effects of sediment on reef health and function and needs to be considered [50].

### Community efforts and efficiency of gabion dams

Community-level watershed management and efforts to revegetate bare soils have previously improved coastal water quality [51]. On Lāna’i,the Maunalei Community Managed Makai Area, in the context of the sediment remediation project described here, used a volunteer collective to implement Hawaiian cultural resource management practices for sediment remediation by identifying drainage area in the ephemeral stream most prone to erosion and building the gabion dams perpendicular to the water-flow direction. The dams slowed water flow during flash floods, allowed for some water retention in the ground, and trapped about 77 tonnes of sediment during one wet season. The dams are simple in design, but labor-intensive. Community-level initiatives on Lāna’i and in watershed partnerships across the Hawaiian Islands continue to raise awareness about methods for reducing erosion that may maintain and restore Hawaiian ecosystems and food security. High levels of sediment capture in the dams we studied suggest that future efforts to remediate sediment with mesquite gabion dams can focus on constructing fewer dams which would still realize the same efficacy in trapping sediment.

On the basis of our data, we would recommend gabion-dam installation in watersheds adjacent to reef areas that have naturally slow sediment removal rates. Per unit of time, effort, and funds spent on sediment remediation, such a strategy would offer the highest return on investment in terms of increasing the speed of ecological recovery in reef areas prone to sedimentation. We focused on the relations between sediment capture efforts on land and actual removal rates on reefs. We found that gabion dam sediment capture can have a measurable effect and may help guide other small-scale solutions for sediment remediation and erosion control.

## Competing Interests

The authors declare that they have no competing interests.

## Acknowledgments

We thank the Pacific Islands Ocean Observing System (PacIOOS) program for support with field instrumentation. We acknowledge the substantial efforts of the Maunalei Community Managed Makai Area in building the gabion dams and in aiding with fieldwork for the field deployments of physical oceanography equipment. We thank Jay Carpio for assistance with the installation of the weather station and to Christina Comfort for field assistance in deploying and maintaining in-water instruments. This work was funded through grants from Conservation International Hawai’i to the Community Managed Makai Area and to the University of Hawai’i-Mānoa. We also thank an anonymous reviewer for helpful comments, which improved the manuscript.

1
Te

F
Responses of Hawaiian Scleractinian Corals to Different Levels of Terrestrial and Carbonate Sediment. Thesis (PhD)
2001
University of Hawaii-Manoa
2
Erftemeijer

P
Riegl

B
Hoeksema

B
Todd

P
Environmental impacts of dredging and other sediment disturbances on corals: A review
Mar Pollut Bull
2012
, vol.
64
(pg.
1737
-
65
)
PMid: 22682583
3
Rogers

C
Responses of coral reefs and reef organisms to sedimentation
Mar Ecol Prog Ser
1990
, vol.
62
(pg.
185
-
202
)
4
Dikou

A
van Woesik

R
Survival under chronic stress from sediment load: Spatial patterns of hard coral communities in the southern islands of Singapore
Mar Pollut Bull
2006
, vol.
51
(pg.
7
-
21
)
PMid: 16293267
5
Perez

K
Rodgers

K
Jokiel

P
Lager

C
Lager

D
Effects of terrigenous sediment on settlement and survival of the reef coral Pocillopora damicornis
PeerJ
2014
6
Fabricius

K
Dubinsky

Z
Stambler

N
Factors determining the resilience of coral reefs to eutrophication: A review and conceptual model
Coral Reefs An Ecosyst. Transit
2011
pg.
493

7
Junjie

R
Browne

N
Erfttemeijer

P
Todd

P
Impacts of sediments on coral energetics: Partitioning the effects of turbidity and settling particles
PLoS One
2014
, vol.
9
pg.
e107195

8
Piniak

G
Storlazzi

C
Diurnal variability in turbidity and coral fluorescence on a fringing reef flat:Southern Molokai, Hawaii
Estuar Coast Shelf Sci
2008
, vol.
77
(pg.
56
-
64
)
9
Stafford-Smith

M
Ormond

R
Sediment-rejection mechanisms of 42 species of Australian scleractinian corals
Aust J Mar Freshw Res
1992
, vol.
43
(pg.
683
-
705
)
10
Johansen

J
Jones

G
Sediment-induced turbidity impairs foraging performance and prey choice of planktivorous coral reef fishes
Ecol Appl
2013
pg.
23

11
Wegner

A
McCormick

M
Endo

G
McLeod

I
Kroon

F
Jones

G
Suspended sediment prolongs larval development in a coral reef fish
J Exp Biol
2014
, vol.
217
(pg.
1122
-
8
)
12
DeMartini

E
Jokiel

P
Beets

J
Stender

Y
Storlazzi

C
Minton

D
, et al.
Terrigenous sediment impact on coral recruitment and growth affects the use of coral habitat by recruit parrotfishes (F. Scaridae)
J Coast Conserv
2013
, vol.
17
(pg.
417
-
29
)
13
Storlazzi

C
Field

M
Bothner

M
Presto

M
Draut

A
Sedimentation processes in a coral reef embayment: Hanalei Bay, Kauai
Mar Geol
2009
, vol.
264
(pg.
140
-
51
)
14
Woolfe

K
Larcombe

P
Terrigenous sedimentation and coral reef growth: A conceptual framework
Mar Geol
1999
, vol.
155
(pg.
331
-
45
)
15
Crabbe

M
Smith

D
Sediment impacts on growth rates of Acropora and Porites corals from fringing reefs of Sulawesi, Indonesia
Coral Reefs
2005
16
Browne

N
Smithers

S
Perry

C
Spatial and temporal variation in turbidity on two inshore turbid reefs on the Great Barrier Reef, Australia
Coral Reefs
2013
, vol.
32
(pg.
195
-
210
)
17
Davidson

K
Hamnett

MCM
The First Four Years: Hawai’i Coral Reef Initiative Research Program,1998–2002,
2003
University of Hawaii at Manoa
Social Science Research Institute
18
Friedlander

A
Aeby

G
Brown

E
Clark

A
Coles

S
Dollar

S
, et al.
The State of Coral Reef Ecosystems of the Main Hawaiian Islands
2005
19
US EPA
USDA/NRCS
HI DOH
HI DBEDT
NOAA
USFWS
, et al.
Hawai“i”s Local Action Strategy to Address Land-Based Pollution Threats to Coral Reefs
2004
20
Risk

M
Assessing the effects of sediments and nutrients on coral reefs
Curr Opin Environ Sustain
2014
, vol.
7
(pg.
108
-
17
)
21
Storlazzi

C
McManus

M
Logan

J
McLaughlin

B
Cross-shore velocity shear, eddies and heterogeneity in water column properties over fringing coral reefs: West Maui, Hawaii
Cont Shelf Res
2006
, vol.
26
(pg.
401
-
21
)
22
Storlazzi

C
Jaffe

B
The relative contribution of processes driving variability in flow, shear, and turbidity over a fringing coral reef: West Maui, Hawaii
Estuar Coast Shelf Sci
2008
, vol.
77
(pg.
549
-
64
)
23
Storlazzi

C
Ogston

A
Bothner

M
Field

M
Presto

M
Wave- and tidally-driven flow and sediment flux across a fringing coral reef:Southern Molokai, Hawaii
Cont Shelf Res
2004
, vol.
24
(pg.
1397
-
419
)
24
Giambelluca

T
Chen

Q
Frazier

A
Price

J
Chen

Y-L
Chu

J
, et al.
Online Rainfall Atlas of Hawai’i
Bull Am Meteorol Soc
2013
, vol.
94
(pg.
313
-
6
)
25
National Weather Service
Flash Flood in Hawaii: Event Statistics 2015

Available at http://www.prh.noaa.gov/hnl/pages/weather_hazards_stats.pdf(accessed February 12, 2015)
26
HI DBEDT
Census Demographic Profile: Neighbor Islands 2010
2010

27
Chow

V
Open-channel hydraulics,
1959
New York
McGraw-Hill

PMCid: PMC2537784
28
Welch

P
The use of fast fourier transform for the estimation of power specta: A method based on time averaging over short, modified periodograms
IEEE Trans Audio Electroacoust
1967
, vol.
15
(pg.
70
-
3
)
29
Zuur

A
Ieno

E
Elphick

C
A protocol for data exploration to avoid common statistical problems
Methods Ecol Evol
2010
, vol.
1
(pg.
3
-
14
)
30
R: A language environment for statistical computing
2013
31
Fox

J
Weisberg

S
A {R} Companion to Applied Regression
2011
32
Mazerolle

M
AICcmodavg: Model selection and multimodel inference based on(Q)AIC(c)
2013
33
Walsh

C
MacNally

R
hier.part: Hierarchical Partitioning
2013
34
Wickham

H
Francois

R
dplyr:dplyr: a grammar of data manipulation
2014
35
Wood

S
Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models
J R Stat Soc
2011
, vol.
1
(pg.
3
-
36
)
36
Zeileis

A
Hothorn

T
Diagnostic Checking in Regression Relationships
R News
2002
(pg.
7
-
10
)
37
Monismith

S
Fong

D
A note on the potential transport of scalars and organisms by surface waves
Limnol Oceanogr
2004
, vol.
49
(pg.
1214
-
7
)
38
Falter

JL
Lowe

RJ
Atkinson

MJ
Monismith

SG
Schar

DW
Continuous measurements of net production over a shallow reef community using a modified Eulerian approach
J Geophys Res
2008
, vol.
113
(pg.
1
-
14
)
39
Presto

M
Ogston

A
Storlazzi

C
Field

M
Temporal and spatial variability in the flow and dispersal of suspended-sediment on a fringing reef flat, Molokai, Hawaii
Estuar Coast Shelf Sci
2006
, vol.
67
(pg.
67
-
81
)
40
Booij

N
Ris

R
Holthuijsen

L
A third-generation wave model for coastal regions
J Geophys Res
1999
, vol.
104
(pg.
7649
-
66
)
41
Field

M
Chezar

H
Storlazzi

C
SedPods: A low-cost coral proxy for measuring net sedimentation
Coral Reefs
2013
, vol.
32
(pg.
155
-
9
)
42
Hoeke

R
Storlazzi

C
Ridd

P
Hydrodynamics of a bathymetrically complex fringing coral reef embayment: Wave climate, in situ observations, and wave prediction
J Geophys Res – Ocean
2011
pg.
116

43
Prouty

N
Storlazzi

C
McCutcheon

A
Jenson

J
Historic impact of watershed change and sedimentation to reefs along west-central Guam
Coral Reefs
2014
44
Storlazzi

C
Field

M
Bothner

M
Presto

M
Draut

A
Sedimentation processes in a coral reef embayment: Hanalei Bay, Kauai
Mar Geol
2009
, vol.
264
(pg.
140
-
51
)
45
Albert

S
Fisher

P
Gibbes

B
Grinham

A
Corals persisting in naturally turbid waters adjacent to a pristine catchment in Solomon Islands
Mar Pollut Bull
2015
, vol.
94
(pg.
299
-
306
)
PMid: 25752531
46
Gove

J
Williams

G
McManus

M
Clark

S
Ehses

J
Wedding

L
Coral reef benthic regimes exhibit non-linear threshold responses to natural physical drivers
Mar Ecol Prog Ser
2015
, vol.
522
(pg.
33
-
48
)
47
Williams

GJ
Gove

JM
Eynaud

Y
Zgliczynski

BJ
Sandin

Sa
Local human impacts decouple natural biophysical relationships on Pacific coral reefs
Ecography (Cop)
2015
, vol.
38
(pg.
1
-
11
)
48
Anthony

K
Fabricius

K
Shifting roles of heterotrophy and autotrophy in coral energetics under varying turbidity
J Exp Mar Bio Ecol
2000
, vol.
252
(pg.
221
-
53
)
49
Smithers

S
Larcombe

P
Late Holocene initiation and growth of a nearshore turbid-zone coral reef: Paluma Shoals, central Great Barrier Reef, Australia
Coral Reefs
2003
, vol.
22
(pg.
499
-
505
)
50
Storlazzi

C
Norris

B
Rosenberger

K
The influence of grain size, grain color, and suspended-sediment concentration on light attenuation: Why fine-grained terrestrial sediment is bad for coral reef ecosystems
Coral Reefs
2015
, vol.
34
(pg.
967
-
75
)
51
Albert

S
Grinham

A
Gibbes

B
Tibbetts

I
Udy

J
Indicators of coral reef ecosystem recovery following reduction in logging and implementation of community-based management schemes in the Solomon Islands
Pacific Conserv Biol
2014
, vol.
20
(pg.
75
-
85
)
This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 Unported License (CC-BY 3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. See http://creativecommons.org/licenses/by/3.0/.