This study uses a simulation method to explore how estuarine pH is affected by mixing between river water, anthropogenic CO2 enriched seawater, and by respiration. Three rivers with different levels of weathering products (Amazon, Mississippi, and St. Johns) are selected for this simulation. The results indicate that estuaries that receive low to moderate levels of weathering products (Amazon and St. Johns) exhibit a maximum pH decrease in the midsalinity region as a result of anthropogenic CO2 intrusion. This maximum pH decrease coincides with a previously unrecognized mid-salinity minimum buffer zone (MBZ). In addition, water column oxygen consumption can further depress pH for all simulated estuaries. We suggest that recognition of the estuarine MBZs may be important for studying estuarine calcifying organisms and pH-sensitive biogeochemical processes.
 In recent years, ocean acidification studies have expanded into coastal and estuarine waters [Cai et al., 2011; Sunda and Cai, 2012; Wootton et al., 2008]. These areas host many economically valuable animal species (e.g., shellfish) [Dickinson et al., 2012; Miller et al., 2009; Tomanek et al., 2011] and deteriorating vegetated habitats (e.g., seagrass meadows) [Waycott et al., 2009]. Sensitivity to acidification in these areas is thus more concerning than in the open ocean because acidification potentially has a more immediate effect on the health of estuarine and nearshore ecosystems as well as regional economies. For example, losses in shellfish harvesting from both natural fishing grounds and aquaculture have been reported in the literature recently [Feely et al., 2008, 2010], and high CO2 exposure has been demonstrated to affect shellfish species at different life stages [Barton et al., 2012; Dickinson et al., 2012; Gazeau et al., 2007; Hettinger et al., 2012; Kurihara et al., 2007; Parker et al., 2010; Waldbusser et al., 2013]. In addition to ocean acidification caused by rising atmospheric CO2, respiration of organic matter also produces CO2; together with high levels of dissolved inorganic carbon (DIC) from upwelled seawater, these factors can seriously acidify coastal and estuarine waters [Feely et al., 2010]. This type of acidification is especially worrisome in stratified estuarine bottom waters [Cai et al., 2011; Mucci et al., 2011] where air-sea exchange is restricted. In addition to CO2-induced acidification, rivers that flow through drainage basins that have acidic sulfate soils could also bring sulfuric acid into estuaries and cause serious acidification in estuarine waters [e.g., Sammut et al., 1995].
 The susceptibility of estuarine waters to acidification has been attributed to the lower buffer capacity from their alkalinity, which is lower than that in the ocean water [Miller et al., 2009]. However, not all estuaries are equal. It is unknown how estuarine buffer capacity along an estuarine salinity gradient changes, how it varies among rivers that carry different levels of bicarbonate as a result of drainage basin weathering (and to some extent due to land use changes such as in the Mississippi drainage basin) [e.g., Raymond et al., 2008], and how estuaries' buffer capacities would change in response to ocean acidification. Furthermore, in situ respiration will further affect the estuarine carbonate systems and also needs to be evaluated.
 In this work, we examine the pH and buffer behavior of estuarine waters using a series of simulations. Different freshwater end-members with different levels of weathering products are included in this discussion, together with the effects of respiration. Based on the result, we propose a new designation—the estuarine minimum buffer zone (MBZ).
 We simulated three river-ocean mixing scenarios that involve three representative rivers (Table 1): a tropical river (Amazon) with a drainage basin that has low concentrations of weathering products, a mid-latitude river (Mississippi) with a drainage basin that has high concentrations of weathering products, and a mid-latitude river with a drainage basin that has moderate concentrations of weathering products (St. Johns, ME, USA). The rationale for using these rivers is because of their distinct carbonate chemistries, ranging from the relatively well-buffered Mississippi River to the low to moderately buffered Amazon and St. Johns Rivers, and because these rivers are distributed across different climate zones.
Table 1. Carbonate Chemistry in Amazon, Mississippi, and St. Johns Rivers
aC-Alk represents carbonate alkalinity in rivers, and the value is calculated by subtracting organic alkalinity in both the Amazon (35 µmol kg−1) [Devol et al., 1995] and St. Johns Rivers (293 µmol kg−1) [Hunt et al., 2011] from total alkalinity (300 µmol kg−1 and 885 µmol kg−1, respectively). For the Mississippi River, we assume that the organic alkalinity contribution is negligible as pCO2 calculated using TA and DIC agrees well with measured values [Guo et al., 2012].
bTemperature is assumed to increase 2°C from the preindustrial time to year 2100.
 For simplicity of discussion and easier comparison, we use a single Atlantic Ocean surface water at salinity 36 and alkalinity 2364.7 µeq kg−1 [see Millero et al., 1998, Table 3] as the ocean end-member. Because we will be discussing the influence of increasing atmospheric CO2 on the estuarine water carbonate chemistry, we use two different atmospheric CO2 levels (280 ppmv during the preindustrial period and 800 ppmv in year 2100 according the “business-as-usual” scenario) [Intergovernmental Panel on Climate Change (IPCC), 2001] to illustrate the changes in estuarine pH due to anthropogenic CO2 intrusion. Because both river water and ocean water have been experiencing an increase in temperature [IPCC, 2001; Kaushal et al., 2010], in this work, we assume that all these estuaries will undergo a 2°C temperature increase from the preindustrial period to 2100. The average temperatures in Table 1 represent present conditions.
 To examine the respiration effect, for simplicity of discussion, we assume the remineralization of organic matter follows Redfield stoichiometry:
 Then we assume that respiration consumes 50% of the dissolved O2 with waters being initially O2 saturated. Thus, further pH changes caused by this hypothetical metabolic CO2 addition, along with the minor alkalinity reduction due to acid production (equation (1)), can be calculated.
 To examine the magnitude of pH change between the preindustrial time and 2100, we define ∆pH = pH2100 − pHpreindustrial. First, we contend that seawater is “saturated” with respect to the atmospheric CO2 due to air-sea equilibrium in both the preindustrial period and year 2100, whereas river water has no anthropogenic CO2 signal due to its generally much higher CO2 partial pressure (pCO2) compared to atmospheric values [e.g., Butman and Raymond, 2011]. Thus, the concentration of anthropogenic CO2 (Cant) in estuarine waters is proportional to the water salinity as a result of conservative river-ocean mixing:
Here Sestuary and Ssea are salinities of the estuarine water and the seawater end-members, Cant (estuary) and Cant (sea) are the anthropogenic CO2 levels of the estuary and the seawater end-members, respectively. Cant (sea) can be calculated using seawater alkalinity by assuming the seawater is equilibrated with the two different air CO2 levels at the chosen simulation temperatures (Table S1 in the supporting information).
 The MatLab® version of the program CO2SYS [Lewis and Wallace, 1998] was used to run the simulations. Carbonic acid dissociation constants are from Millero , the bisulfate dissociation constant is from Dickson and Riley , and the borate to salinity ratio is from Uppström .
 The pH (here we use total pH scale throughout this discussion) in both the Amazon and St. Johns estuaries starts from relatively low values (6.8–6.9) at the river end-members and increases toward the ocean end-members (Figures 1a and 1c). In comparison, pH in the Mississippi estuary shows a much smaller variation along the salinity gradient and the freshwater is slightly alkaline (pH >8) (Figure 1b).
 Intuitively, one may expect that ∆pH as a result of Cant intrusion would follow the proportion of Cant present in the estuarine water due to river-sea mixing (e.g., equation (2)), i.e., the magnitude of ∆pH approaches zero monotonically toward the river end-member as Cant decreases linearly with salinity. However, for both the Amazon and St. Johns estuaries, first, there is a maximum pH decrease (i.e., most negative ∆pH) as salinity decreases, then ∆pH diminishes as salinity approaches zero (Figures 1a and 1c); whereas the expected ∆pH trend based on the amount of Cant is only seen in the Mississippi estuary (Figure 1b). Furthermore, across the salinity gradients of the Amazon and St. Johns estuaries, under selected simulation temperature regimes (Table 1), the maximum pH decrease in the Amazon estuary occurs at salinity ~9, while for the St. Johns estuary, the same salinity is ~19 (Figure 1c).
 Finally, regardless of the river water chemistry, oxygen consumption in the water column can further decrease pH (∆pHoc, dashed red line in Figure 2) as was illustrated in earlier work [Cai et al., 2011; Sunda and Cai, 2012]. If we assume 50% oxygen consumption from the saturation values across the salinity gradient and that overlaying ∆pHoc onto Cant caused the pH decrease (∆pHant, solid black line), the maximum pH decreases (∆pHoverall, dashed green line) in all three estuaries (i.e., −1.2, −0.7, and −1.2 pH units in Amazon, Mississippi, and St. Johns, respectively) are much greater than the ∆pHant alone (~0.5 pH units, Figure 2). A greater pH decrease caused by oxygen consumption than by Cant intrusion across the salinity gradient in the St. Johns estuary can be attributed to higher oxygen solubility at lower temperature (Figure 2c). Consistent with Cant induced pH decrease, the most negative ∆pHoverall also occurs at a higher salinity in the St. Johns estuary (Figure 2c) than in the Amazon estuary (Figure 2a).
 The seemingly peculiar behavior of ∆pH along the salinity gradient can be explained by the relative changes in buffer behaviors in these three estuaries that have different carbonate chemistries. Typically, the Revelle factor [Revelle and Suess, 1957] is used when a quantitative examination of the responses of the carbonate system to CO2 addition is needed. However, pH responses as a result of CO2 addition cannot be determined explicitly using the Revelle factor. Therefore, we choose the ßDIC factor proposed by Egleston et al.  as this factor provides a more direct measure of the pH response due to DIC addition:
Here a decrease in ßDIC indicates a decrease in buffer capacity. The results (Figure 3) suggest that the buffer capacity of the Mississippi estuary increases monotonically with increasing salinity. However, for the Amazon and St. Johns estuaries, a minimum buffer zone (MBZ) appears in the midsalinity region that closely corresponds to the most negative ∆pH in each estuary due to Cant intrusion (Figure 3). With increasing Cant concentration in estuarine water, the MBZ is moving toward higher salinities (i.e., the dashed lines in Figure 3). Therefore, the maximum pH decrease in these two estuaries is a combination of the minimum buffer capacity and the amount of Cant present in estuarine waters (equation (2)). An alternative explanation for the minimum buffer zone is that it occurs at a salinity where the estuarine pH approximately equal (pK1 + pK2)/2, where pK1 and pK2 are dissociation constants of carbonic acid (Figure S1).
 It is worth noting that once the salinity is below the MBZ salinity, ßDIC increases with decreasing salinity in both the Amazon and St. Johns estuaries (Figure 3). This increase simply reflects that there are very low levels of CO32− (compared to those in the Mississippi estuary, Figure S2) to neutralize additional CO2; thus, further addition of DIC mostly leads to an increase of the dissolved aqueous CO2, or CO2(aq) [Egleston et al., 2010]. In other words, the buffering of the DIC addition is thus based on dilution of DIC into the existing CO2(aq) pool, and subsequently, the pH (or [H+]) change becomes less sensitive to further DIC additions when compared with that at the MBZ.
 In addition to the seawater Cant-caused MBZ shift (Figure 3), both river carbonate chemistry and temperature play an important role in determining the location of the MBZ. If we consider a range of river alkalinity of 300–2400 µmol kg−1, at two temperatures (28°C and 16°C) and two river pCO2 levels (1200 µatm and 4600 µatm), which correspond to the upper and lower limits of the river parameters considered in this simulation, we can calculate the distribution of βDIC as a function of river alkalinity and salinity by assuming that the river waters mix with preindustrial seawater. The results (Figure 4) suggest that the MBZ occurs in higher salinity waters when the river pCO2 is high (i.e., higher DIC) or temperature is low. Therefore, rivers enriched in DIC as a result of respiration (thus having a higher DIC/TA ratio) at higher latitude (i.e., colder climate) should have their MBZs occurring closer to the ocean than those at lower latitude with low DIC/TA ratios (Figure 4d versus 4a). On the other hand, for a river with high alkalinity (> ~ 1600 µmol kg−1) but moderate river pCO2 such as the Mississippi, the MBZ does not exist (Figures 4a and 4b) under our examined temperature range. However, if the river pCO2 was high enough (due to remineralization of riverine organic carbon), even estuaries that receive high alkalinity freshwater input would start to have an MBZ (Figure 4c).
 This study is the first that examines estuarine acidification caused by the increase in anthropogenic CO2 in the ocean, in situ respiration, and river carbonate chemistry fluctuations. Earlier, Sunda and Cai  simulated acidification caused by eutrophication-induced excess respiration in stratified estuarine bottom waters along a salinity gradient. However, they used a 1 mmol kg−1 alkalinity model river end-member without taking variability in river carbonate chemistry into account. Furthermore, they treated the Cant signal along the salinity gradient by assuming equilibration with the atmosphere under different air CO2 levels but not with the amount of Cant that would be introduced through mixing with the ocean water. On the other hand, Salisbury et al.  explored carbonate saturation states in various world river estuaries, although they did not consider the acidification effect in the estuaries caused by increases in seawater Cant.
 Based on the simulation results, the Mississippi estuary shows the most resistance to acidification by having the least overall pH reduction under both Cant influence alone and combined with DIC input from anthropogenic and respiration sources. Also, pH in this estuary remains the highest throughout most of the salinity gradient (Figure 2b). Therefore, given the high pH in warm estuaries such as the Mississippi estuary, these environments likely would remain habitable for calcifying organisms for the longest extent under current and predicted ocean acidification conditions. Moreover, compared with the estuaries on the east coast of the United States, the higher growth rate of bivalves in the Gulf of Mexico coastal estuaries [Eastern Oyster Biological Review Team, 2007], where river discharge of higher levels of weathering products occurs [Benoit et al., 1994], should at least partially contribute to a more favorable carbonate chemistry.
 The appearance of the MBZ in the midsalinity regions of estuaries with low to moderate levels of weathering products such as bicarbonate (e.g., Amazon and St. Johns) suggests that waters in the MBZs are extra sensitive to acidification stress, regardless of the acid source (CO2, acidic soil, and nitrogen deposition). Recently, a long-term (1985–2010) pH decline was reported in the polyhaline (S > 18) waters of the Chesapeake Bay [Waldbusser et al., 2011] with a pH decrease of −0.006 to −0.012 year−1, which is far greater than the rate in the open ocean during the same time frame (−0.0017 year−1) [Byrne et al., 2010; González-Dávila et al., 2007]. While the detailed carbonate system chemistry during this long-term observation is unknown, we propose that the MBZ may have played an important role in pH regulations in this estuary, and further study would be needed to examine this hypothesis. Furthermore, we hypothesize that reduced pH in the midsalinity waters probably has already taken a toll on both the settlement efficiency of juvenile calcifiers (by extending the duration of pelagic phase) and the well-being of benthic adult individuals and populations [Waldbusser et al., 2011, and references therein].
 Other than the effect of pH on calcification, many biogeochemical reactions are pH-sensitive. For example, ammonia oxidation strongly depends on the environmental pH although exactly how pH mediates such reactions is not without controversy [Fulweiler et al., 2011; Kitidis et al., 2011]. Given the fact that estuaries are important locations for removing fixed nitrogen [Seitzinger et al., 2006], further studies on changes in pH due to acidification within estuaries, and subsequently, how such pH change would affect the nitrogen cycle is probably an important research topic in these dynamic systems.
 Finally, this work discusses highly simplified scenarios that only consider the conservative mixing of river water and ocean water, and water pH is controlled mostly by the carbonate system (and borate to a much smaller extent, in the calculations of βDIC and pH using CO2SYS). In nature, however, estuarine biogeochemical reactions that involve proton production and consumption are rather complex (for example, the presence of organic acid and other acid-base species) and are often accompanied by nonconservative mixing behaviors [e.g., Cifuentes et al., 1990]; therefore, MBZs may not appear exactly at the salinities predicted using the conservative mixing model. Nevertheless, the MBZ should remain a fundamental estuarine property governed by acid-base chemistry. Under the context of ocean acidification, it would be of great interest to differentiate the long-term estuarine pH trends due to oceanic CO2 uptake from the short-term variations due to biogeochemical reactions, the latter probably being larger in magnitude, in many disparate estuarine systems.
 This study was supported by the startup fund to XH provided by the College of Science and Engineering at Texas A&M University-Corpus Christi. We thank the Associate Editor and two anonymous reviewers for their critical comments that helped to improve the quality of this work. We also thank S. Cooley for her thorough editorial assistance.
 The Editor thanks an anonymous reviewer for assisting in the evaluation of this paper.