The Savaia coastal marine protected area (MPA) is located in Lefaga Bay in the south coast of Upolu (Figure 1). It is characterized by a fringing reef, often split by freshwater discharge from streamflow, and supported by a low-lying coastal plain [40]. Savaia was selected for our study based on its present coral reef coverage, existing data related to coral and fishery resources, history of successful projects, and community management principles for safeguarding project equipment. Lefaga reefs comprise of 0.1% living coral, which is comparatively lower than other reefs considered to be doing well including Palolo Deep (0.48%), Amaile (0.23%) and Aga (0.3%) [40]. A most recent survey reported Lefaga coral cover decreased from 30% in 1996 to 1% - 5% in 2016 [13]. Savaia is located in an ecotourism area, which is important for subsistence economy in the respective community [40] while at the same time, presenting concerns for reef damage from tourism activities.

Figure 1. Map of Samoa showing the location of Lefaga District and the location of collection sites (red dot) in Savaia. (image source: NUS drone).

As noted previously, the goal of this study was to characterize the variability in carbon chemistry along the Savaia MPA using surface water quality measurements to identify potential trends and processes affecting pH. Four sites (Site1, Site 2, Site 3, Site 4) (Figure 2) have been defined as spatial zones corresponding to distances in km along the coastline and surface water longitudinal gradient of the Savaia MPA (Table 1).

Table 1. Description of sampling sites defined as spatial zones corresponding to distances in km along the coastline and surface water longitudinal gradient of the Savaia MPA.

Seawater samples collection followed the method described by Dickson et al. (2007) [41]. Bottle samples were collected at four designated sites (Site 1 - Site 4) within the subtidal zone of Savaia Marine Protected Area (MPA), each corresponding to specific distances (Figure 2, Table 1) along the fringing reef and coastal gradient. Sampling points were located just inside the MPA boundary, with additional comparative samples taken adjacent to the iSAMI deployment point situated outside the boundary. All bottle samples were collected at a consistent depth of approximately 1 m below the surface to minimize variability due to stratification. Seawater was collected in borosilicate glass bottles (or the Niskin bottle) underwater immediately after opening the bottles. This approach was necessary to minimize the exchange of CO₂ with the air space in the collection bottle which affects all carbon parameters except total alkalinity (A_T). Total alkalinity (A_T) is the measure of the capacity of seawater to neutralize acids, measured by titration and expressed in μmol∙kg⁻¹. The samples were immediately transferred into an ice cooler for transport from the collection site to the marine lab at the National University of Samoa ocean campus in Mulinuu. The time to travel from the collection site to the marine lab was 45 minutes. Upon arrival to the lab, the samples were stored in a cool, dark, location (preferably refrigerated but not frozen) until use. For quality assurance, some duplicate sampling was carried out, both from the same sampling bottle (or Niskin bottle) and, if possible, from two

Figure 2. Marine protected area (MPA) and the estimated location of the iSAMI and sampling sites (not to scale).

sampling containers tripped together at the same depth, to assess the quality of the sampling procedures. Samples collected for A_T and pH analysis in the lab were preserved with a mercuric chloride solution [41] to stop biological activity from altering the carbon distributions in the sample container before analysis. Monthly sampling was conducted between February 2019 and February 2020, with the exception of July 2019 due to equipment malfunction. To reduce tidal and diurnal variability, collections were made at approximately whenever possible. Duplicate samples were taken at selected sites to evaluate consistency, and extreme pH values were cross-checked against duplicates before inclusion in the dataset.

Preparation for sampling and handling samples was carried out according to the guidelines set out in the Standard Operating Procedures by Dickson et al. (2007) [41]. Samples for pH were also analyzed directly from the sample containers. The pH of seawater was measured in the lab by spectrophotometric means. The samples were collected directly into 10 cm path-length optical cells and sealed with polytetrafluoroethylene (Teflon®) caps to ensure that there was no headspace. Samples for DIC and A_T measurements were collected in high-quality borosilicate glass bottles such as Schott Duran (l.c.e. 32 × 10^–7 K^–1) [41], as they retain sample quality for both temporary and longer-term storage. The bottles were sealed using greased ground glass stoppers held in place with some form of positive closure, or in some alternate gas-tight fashion.

2.4.1. Determination of the pH of Sea Water Using the Indicator Dye m-Cresol Purple

Seawater pH (pH_T, pH measured on the total scale, or hereafter simply pH) is the negative logarithm of hydrogen ion concentration, measured spectrophotometrically and with the iSAMI. A glass Pasteur pipette was used to transfer the sample from the newly opened sample bottle into the spectrophotometric cuvette. This was done carefully and quickly to ensure minimal sample interaction with atmosphere. For work of highest sensitivity and precision, a double-beam spectrophotometer is desirable. However, good results can be obtained with a high-quality single-beam instrument. We used a Thermo Scientific^TM Orion^TM AquaMate 7000 Vis spectrophotometer provided with the Global Ocean Acidification Observing Network (GOA-ON) Kit for to carry out the pH measurements in this study. The absorbance of the seawater sample was measured and recorded at three wavelengths: 730 nm, 578 nm, and 434 nm. These wavelengths correspond to a non-absorbing wavelength (730 nm) for m-Cresol purple and the absorption maxima of the base (I²⁻) and acid (HI⁻¹) forms of the m-Cresol dye (respectively 578 and 434 nm).

2.4.2. Determination of the Total Alkalinity of Sea Water

Total alkalinity (A_T) is the measure of the capacity of seawater to neutralize acids, measured by titration and expressed in μmol∙kg⁻¹. A_T of the samples was determined by titration with standardised 0.1N HCl made up in a 0.6M NaCl solution to approximate the ionic strength of the samples being titrated. The results are expressed as micromoles per kilogram (µmol/kg) of sea water. This method is suitable for the lower total alkalinity range with a smaller initial acid addition [41][42].

2.4.3. InSitu pH Measurements by iSAMI Sensor

The iSAMI pH sensor (Sunburst Sensors, Missoula, MT, USA) provided in the GOA-ON Kit is autonomous compared to the spectrophotometry method. It uses a highly accurate colorimetric reagent method and does not suffer from the drift that plagues most electrode-based pH probes. The iSAMI provides valuable in-situ time series data near the surface at depths of up to 3 meters. The iSAMI pH sensor was programmed to take in-situ pH readings every 15 minutes for 3 weeks of each month following the operation manual iSAMI-pH [43], then deployed at about 3m depth in the location 13˚57.366′S and 171˚570744′W. In the fourth week since initial deployment, the instrument was retrieved from the sampling site to extract/download the data onto a laptop, reprogrammed, then deployed again to continue the measurements for the next four weeks. It is noted that the iSAMI experienced technical malfunction after three months (December 2018 and February-March, 2019) and was unavailable for further measurements in the duration of the study. The iSAMI data presented here are the measurements taken from December 2018 and February-March 2019.

The study collected quantitative data on pH, alkalinity, temperature and total dissolved CO₂ in seawater. The other seawater properties such as DIC, pCO₂ were calculated using the CO2SYS spreadsheet [44]. Temperature (˚C) was measured insitu at each sampling site. Here, temperature, pH and total alkalinity were measured and used to calculate the remaining carbon parameters. The calculated values include those for partial pressure of CO₂ (pCO₂, μatm) calculated from measured pH and A_T using CO2SYS, total dissolved inorganic carbon (DIC, μmol∙kg⁻¹) calculated from pH and A_T, and bicarbonate ion concentration ( H C O 3 − , μmol∙kg⁻¹) calculated from pH and A_T.

The differences in each response variable across sites and months were analysed by using a linear model and One-Way Analysis of Variance (ANOVA). pH, A_T, temperature, pCO₂, DIC, and HCO₃^- were modeled with Site as a fixed factor and Month as a repeated measure. Repeated monthly observations from same sites were handled by a One-Way Repeated Measures ANOVA including Month as a within-subject factor, thereby controlling for temporal autocorrelation. The use of ANOVA and linear models is justified by the balanced sampling design, with repeated monthly measurements across four sites providing sufficient replication to detect differences in mean values. This approach allows us to test site effects while accounting for temporal variation, ensuring that both spatial and seasonal variability are appropriately represented in the analysis. All the plots and anova tests were created and carried out using R [45].

All pH and total alkalinity (A_T) samples were subject to quality-control checks prior to analysis. Duplicate samples were collected at selected sites and compared to evaluate consistency in sampling and analytical procedures. Extreme pH values outside the expected seawater range (7.5 - 8.5) were flagged and re-checked against duplicate measurements before inclusion in the dataset. Measurements that did not meet qualitycontrol standards were excluded, resulting in a final dataset of 12 months of observations (February 2019-February 2020, excluding July 2019 due to equipment malfunction).

Seawater temperature (˚C) did not show any major differences between each site, showing close synchrony between the values (24˚C - 28˚C) (Figure 3). There was an indication of differences in temperature with time (month of year), where the highest mean temperature was in January (28.1˚C ± 0.1˚C) and the lowest mean temperature in April (20.5˚C ± 0.2˚C) across all four sites.

(a) Calculated summary of Mean temperature (˚C) over time (months)

(b) Mean temperature (˚C) showing distribution of data points across the four sites

Figure 3. Mean temperature (˚C) in the four sites with (a) calculated summary of mean temperature (˚C) and (b) mean temperature (˚C) showing distribution of data points across the four sites.

The mean pH (Figure 4), measured on the total scale and denoted pH_T, was not significantly impacted by the site location but rather by the time of year (p < 0.001). The differences in means of pH_T for four sites were non-significant (p = 0.5, Table 2). However, even though the mean pH_T was not statistically different between the four different sites, there was indication for a consistent pH range of 7.9 - 8.2 across all sites (Figure 4, Figure 5).

Table 2. Mean pH_T measured in the four sampling sites. Values are Means (±SE).

The iSAMI sensor successfully recorded continuous insitu pH measurements in December 2018, February and March 2019. However, despite multiple attempts to reprogram and deploy the sensor, it failed to yield insitu pH data for the remaining duration of the study. There was no statistical test done for iSAMI measurements due to negligible sample sizes that lack the power to detect true differences. Running a test would yield a high rate of false negatives.

(a) Calculated summary of Mean pH over time (months) per site

(b) Mean pH of seawater showing distribution of data points across the four sites

Figure 4. Mean pH (total pH or pH_T) for the four sites in Savaia showing (a) calculated mean pH over time across the four sites, and (b) mean pH trends by site.

Based on the parametric data collected by iSAMI, mean pH was 7.9 ± 0.01 and pH range between 7.6 and 8.4 (Figure 6). Temperature tended to increase from 17 February until 24 February, with a mean temperature 30.5˚C ± 0.5˚C, after which it decreased to a mean 29˚C ± 0.04˚C.

Figure 5. Distribution of measured pH values ranging between 7.1 and 9.7 in the four sites.

Figure 6. Real-time pH measurements by iSAMI for December 2018, and February-March 2019.

The mean total alkalinity (A_T µmol∙kg⁻¹) between the four sites (Figure 7, Figure 8) was 1955.1 µmol∙kg⁻¹ ± 25 and was not significantly different (p = 0.9, Table 3). However, there was a significant impact of time of year, where there was a significant decrease (red lines on Figure 8) in A_T of seawater between the months of April and September (p < 0.001).

The values for pCO₂ (µatm) (Table 4) were variable (150 - 550 µatm) between the four sites (Figure 9). There was no significant difference between the means of pCO₂.

Figure 7. Mean of total alkalinity (A_T) (μmol∙kg⁻¹) across the four sites in Savaia.

Figure 8. Mean Total alkalinity (A_T) (μmol∙kg⁻¹) by site.

Table 3. Total alkalinity (A_T) (µmol∙kg⁻¹) measured in the four sampling sites. Values are Means (±SE).

Table 4. Dissolved carbon dioxide (pCO₂) (μatm) measurements in the four sampling sites. Using the measured values of pH_T and A_T, pCO₂ was calculated using the calculator CO2SYS [44] and applying the constants K1, K2 from Mehrbach et al. (1973) [46]. Values are Means (±SE).

Figure 9. pCO₂ (μatm) measurements across the four sites.

Total CO₂ (µmol∙kg⁻¹) (Figure 10) was not significantly different between the four sites; however, there was a significant difference (p < 0.001) in the means between the months of year (Table 5).

Table 5. Total carbon dioxide (CO₂) measurements in the four sampling sites. Using the measured values of pH_T and A_T, pCO₂ and CO₂ were calculated using the calculator CO2SYS [44] and applying the constants K1, K2 from Mehrbach et al. (1973) [46]. Values are Means (±SE).

Figure 10. Calculated total CO₂ (μmol∙kg⁻¹) values across the four sites.

The concentration of carbonic acid H C O 3 − (µmol∙kg⁻¹) was not significantly different between the four sites; however, there was a significant impact of time of year (p = 0.04) (Table 6). The three sites Site 1, Site 2 and Site 3 showed a similar trend of increasing H C O 3 − concentration in the year and declined in November (Figure 11). Site 4 had missing values for February, June and October and may have impacted on the overall trend in the H C O 3 − variability in this location.

Figure 11. Calculated H C O 3 − (μmol∙kg⁻¹) values across the four sites.

Table 6. The concentration of carbonic acid ( H C O 3 − ) was measured in the four sampling sites. Using the measured values of pH_T and A_T, pCO₂ and CO₂ and H C O 3 − were calculated using the calculator CO2SYS [44] and applying the constants K1, K2 from Mehrbach et al. (1973) [46]. Values are Means (±SE).

The pH environment of the four sites (Site 1, Site 2, Site 3, Site 4) was typical of natural tropical shallow intertidal environments, with a mean pH of 8.06 ± 0.07 displaying regular monthly periodicity [47]-[49]. The monthly variation in pH was low for each site, indicating a potentially less dramatic impact of the combined influence of different factors, including temperature changes, freshwater input and biological activity. Studies have shown that seasonal or monthly variations are generally lower than daily variations in coastal areas. This is because daily pH fluctuations are driven by biological processes like photosynthesis and respiration, which can cause large and rapid changes in pH [50][51]. Monthly or seasonal variations in pH may be less pronounced because the shorter and more intense fluctuations in pH are averaged out by seasonal changes in other factors, including temperature and biological cycles [51][52]. In hindsight, daily pH measurements would have been an important analysis by itself in this study because of the large and rapid changes in pH within a 24-hour cycle. We have noted this crucial aspect for future research. Although the occurrence of the lowest mean pH 7.1 (Site 4) and highest pH 9.7 (Site 1) was not statistically significant, the large difference between the lowest and highest pH across the four sites is significant in itself, and could be an indicator of several potential factors. These may include seawater thermal stratification and high phytoplankton biomass [50][53] in site (Site 1) where the highest pH values occurred (Figure 4). Such disparity in pH values is unsurprising in a highly productive coastal zone in summer [54], with a summer pH of similar level [55] and higher elsewhere [56]. In such areas, CO₂ depletion due to net photosynthetic production exceeds its replenishment from atmospheric infiltration, respiration or from other sources [57]. It should be noted however, that sources of CO₂ for example calcification, can cause oversaturation of surface waters, even in the presence of high photosynthetic production such as reported on some coral reefs [58][59].

The sharp low seawater pH of 7.1 in Site 4 is likely to be due to benthic CO₂ sources such as respiration and calcification combined with its potential separation from surface waters by minor temperature stratification [60]. The site is part of a community marine reserve with aquaculture activities present such as the giant clam farm, seaweed beds and seagrass meadows here which likely contributed CO₂ via respiration and calcification. Another reason for the very sharp decrease in seawater pH in Site 4 site could be due to a freshwater source from rivers in the vicinity of the coastal area. It is well established that freshwater from rivers has a lower pH of 7 than seawater pH 8 [61][62] due to low salinity levels [63].

Total alkalinity (A_T µmol∙kg⁻¹) is an important property affecting pH due to its buffering capacity to resist large changes in pH levels in the natural seawater environment [41]. Coastal and estuarine ecosystems such as our sampling sites in Savaia are highly dynamic and therefore can be classified as weakly or strongly buffered depending on whether the A_T was below or above that of ocean water [64].

The spread of low A_T was consistent across the four sampling sites and corresponds with our pH_T values, which demonstrated lower mean pH_T from June to December with decreasing A_T (Figure 6, Figure 7). This relationship between pH_T and A_T was evident in the trends in the mean values (Figure 4, Figure 5). The reason for this decreasing pH and A_T is likely due to two reasons: first, the effect of marine biota via the buildup of nutrient and organic matter cause AT to decrease [65] and secondly the variation in sea surface temperature (SST) where higher temperatures (˚C) have been found to lower A_T in some tropical seawater environments [66]. Our data are in agreement with those findings, where in the Savaia sampling sites, all of the four sites demonstrated a high mean temperature 23.7˚C ± 0.4˚C, with the highest temperature of 28.9˚C in Site 3. There is a pattern of consistently higher temperatures in June-October. The other possible reason for lower A_T is potentially low salinity or salinity fluctuations in our sampling sites; however, this was not well documented in our measurements, and we have acknowledged this oversight for consideration in our continued monitoring work.

Studies have shown that the high pH variability of coastal ecosystems is primarily driven by inputs from land [67]. These include freshwater inputs from rivers, streams and estuaries, which typically dilute the alkalinity of seawater [68]. This results in reduced buffering, nutrients enhancing productivity and pH, as well as organic matter supporting excess respiration driving acidification [64]. In our study, we overlooked the existence of freshwater sources near the site and we assumed the level of salinity at 35 ppt in our calculated values of carbon parameters. We acknowledge this oversight and have addressed this for improvement in our ongoing seawater monitoring activities. In hindsight, our results have highlighted for the first time, the natural pH variability in Savaia where the geophysical profile is highly dynamic with potential significant influence of freshwater input. This is important given this site is a popular giant clam sanctuary. Additional factors that can affect pH variability in coastal regions include geochemical changes associated with land use changes, hydrodynamic circulations and coastal water table or aquifer discharges [68]. Our study site at Savaia is located within the village MPA and has integrated ecotourism to promote sustainable development. There have been developments in the area in terms of infrastructure [69] and although the developments may be small-scale, it can have potential impact on the coastal environment and carbonate chemistry of seawater.