In this study we investigated the relations between community calcification of an entire coral reef in the northern Red Sea and annual changes in temperature, aragonite saturation and nutrient loading over a two year period. Summer (April–October) and winter (November–March) average calcification rates varied between 60 ± 20 and 30 ± 20 mmol·m−2·d−1, respectively. In general, calcification increased with temperature and aragonite saturation state of reef water with an apparent effect of nutrients, which is in agreement with most laboratory studies and in situ measurements of single coral growth rates. The calcification rates we measured in the reef correlated remarkably well with precipitation rates of inorganic aragonite calculated for the same temperature and degree of saturation ranges using empirical equations from the literature. This is a very significant finding considering that only a minute portion of reef calcification is inorganic. Hence, these relations could be used to predict the response of coral reefs to ocean acidification and warming.
 Ocean acidification together with severe decrease in live coral cover in tropical coral reefs resulting from eutrophication, thermal bleaching, disease, crown of thorn starfish outbreaks and various other anthropogenic and natural causes [Wilkinson, 2000; Bellwood et al., 2004] has raised serious concern regarding the ability of coral reefs to continue maintaining their CaCO3 framework in the near future [Kleypas et al., 1999a]. Calcification in individual stony corals and in coral mesocosms that simulate a reef community decreases significantly with lowering of Ωarag (corals deposit the mineral aragonite [Langdon, 2000]). Similar relationships between Ωarag and rates of CaCO3 precipitation were reported for several tropical communities (Bahama Banks [Broecker and Takahashi, 1966] and a coral reef flat near Okinawa, Japan [Ohde and van Woesik, 1999]). Hence, it is expected that ocean acidification resulting in Ωarag decrease will lead to a decrease in coral reef calcification [Kleypas et al., 1999a].
 Obviously, Ωarag, temperature and nutrients influence calcification simultaneously and it is difficult to determine their individual effect on community calcification from in situ measurements [Smith, 1978; Barnes and Chalker, 1990; Langdon and Atkinson, 2005]. Furthermore, recent attempts to examine the interaction of temperature and Ωarag [Reynaud et al., 2003] and Ωarag and nutrients [Langdon and Atkinson, 2005] in laboratory experiments have displayed complex responses, which are difficult to interpret. The primary objective of this study was to identify and quantify the combined and individual influences of Ωarag, temperature and nutrients on community calcification of a natural coral reef. To this end we have studied the seasonal fluctuations in community calcification of a single coral reef in the northern Gulf of Eilat, Red Sea. During these studies we observed natural variations in temperature, nutrient levels, solar irradiance and carbonate chemistry, which are similar in their magnitude to those associated with the global geographic distribution of coral reefs (ca. ±30° north and south of the equator, Table 1). Hence, we assume that our findings will be applicable to coral reefs in general.
Table 1. Averages and Extremes of Environmental Parameters Associated With the Global Distribution of Coral Reefs and Their Corresponding Values in the Nature Reserve Reef (NRR), Northern Gulf of Aqaba, Red Sea
 The Nature Reserve Reef (NRR) in the northern Gulf of Eilat (Aqaba), Northern Red Sea, is a high latitude fringing reef (∼29.5°N) located along the western shore of the Gulf approximately 10 km south of the city of Eilat (Figure 1). The NRR is ∼1 km long and the live coral cover is between 20 and 40% [Zakai, 2000; Loya, 2004; Genin and Silverman, 2004]. It is a declared nature reserve and has a restricted and subject to well supervised public access to the Fore-Reef along its Northern half. The southern half of the reef, where these measurements were conducted, has been completely closed off to public access since 1996.
 The net surface flow along the western coast of the Gulf of Aqaba in the vicinity of the NRR is southward during most of the year except for a short period of northward flow between November and January [Genin and Paldor, 1998]. Near the reef the current velocity varies between 0–10 cm·s−1 and has an on-shore component of 0–2 cm·s−1. Both on-shore and long-shore components are modified by the wind and tide (Figure 2). During flood tide the current is weakened in comparison to the ebb tide because the current is generally southward. The current is also forced by the prevailing northerly wind, blowing from this direction ca. 90% of the time (Op. Cit.). The wind is usually strongest during the daytime hours reaching velocities of 10 m·s−1.
2.2. Water Sampling, Laboratory Measurements, and Data Reduction
 Diurnal cycles of AT, pH, nutrients and physical parameters (air and water pressure, air and water temperature, salinity, light intensity, wind velocity and relative humidity) were measured at the NRR on a ca. monthly basis between March 2000 and March 2002. During these studies samples of water were taken from the reef at the lagoon station (Figure 1) every 3–4 hours for a period of 24–48 hours. pH and AT samples were collected, stored, prepared and analyzed according to the methods outlined by DOE . pH was measured within 24 hours of sampling using a Radiometer PHM93 combined glass electrode calibrated with Radiometer commercial buffers (traceable to IUPAC and NIST, 7 and 10). Samples were measured at 25°C using a thermostated water bath. Precision of this measurement was ±0.003 pH units (3 measurements per sample). AT samples were analyzed within 24 hours of sampling with an automated Metler Toledo DL-67 Titrameter with a temperature probe and compensation employing a Gran type titration and computation as described by Sass and Ben-Yaakov . Precision of this measurement was ±2 μeq·kg−1 (2–4 measurements per sample). Nitrite (NO2−1) was measured with a colorimetric method described by Grassohoff et al. (1999) using a Flow Injection Autoanalyzer (Lachat Instruments Model QuickChem 8000). Nitrate (NO3−1) was measured by reducing it to nitrite using a copperized cadmium column. Precision of nitrite and nitrate measurements was ±0.02 μmol·L−1.
 Calculations for the carbonate system in seawater were made using the equations from Zeebe and Wolf-Gladrow (2001) and the ion activity scale (NBS) apparent thermodynamic dissociation constants from Mehrbach et al.  and CaCO3 solubility product from Mucci . In two earlier studies conducted in 1997 and 1998 measurements of total dissolved inorganic carbon (CT) were also made. CT samples were poisoned with HgCl2 after extraction from the Niskin bottle and sealed with no headspace in a brown glass 63 ml bottle with a screw cap and a Teflon ring as well as covering the bottle cap with masking tape. Samples were extracted after a week into an evacuated Erlenmeyer bottle with concentrated H3PO4 acid and connected to a vacuum line. The partial pressure of CO2 gas was measured after a sequence of 2 freezing and thawing cycles in liquid air and cooled alcohol (−40°C) cold traps with an IR manometer in a known volume and corrected for room temperature. Comparisons were made between activity and total hydrogen pH scale calculations of Ωarag using the apparent thermodynamic dissociation constants developed by Roy et al.  for the total hydrogen scale (Figure 3 and Table 2). Values of Ωarag were calculated using the dissociation constants of the total hydrogen and activity pH scales, and with different combinations of AT, CT and pH. These values agreed well with each other and in any case were within the range of error for Ωarag specified by Millero  (ΔΩarag = ±0.08), which used the activity scale dissociation constants and similar analytical errors in AT and pH. The average difference between all Ωarag calculated with AT, pH and AT, CT measurements and the NBS and total hydrogen apparent thermodynamic dissociation constants plotted in Figure 3 was 0.08.
Table 2. Daily Averages of pH (Activity Scale Measured at 25°C), Total Alkalinity (AT), Total Dissolved Inorganic Carbon (CT), Temperature, Salinity, and Calculated Values of Ωarag as a Function of Measurement Combinations (pH, AT; pH, CT; AT, CT) in the NRR for Studies Conducted in June 1997 and February 1998
Calculated with the apparent thermodynamic dissociation constants of the carbonate system for activity scale pH measurements [Mehrbach et al., 1973].
Calculated with the apparent thermodynamic dissociation constants of the carbonate system for the total hydrogen scale [Roy et al., 1993].
 Meteorological data was used to calculate rates of evaporation (E) according to a simplified version of Fairall et al.  TOGA/COARE code using the hfbulktc.m matlab function in the airsea toolbox (version 2.0, 1999), developed by R. Pawlowicz and A. Anis (available for download off R. Pawlowicz's website, http://www.eos.ubc.ca/about/faculty/R.Pawlowicz.html). The analytical error for evaporation used in calculations of residence time was 1 mm·d−1.
2.3. Community Rates of Calcification and Dissolution
 The net rate of community calcification (Gnet) in coral reefs is the difference between gross biogenic precipitation of CaCO3 (Ggross) and the biological and chemical dissolution (D) occurring within the reef framework (Gnet = Ggross − D [Tribble et al., 1988; Glynn, 1997; Yates and Halley, 2006]). These processes can be described by the equation [Ca+2] + [CO3−2] ↔ [CaCO3]s. Hence, the decrease of Total Alkalinity (AT), which is primarily a measure of [HCO3−1] and [CO3−2] in seawater, indicates CaCO3 precipitation (Ggross) and AT increase CaCO3 dissolution (D).
 In this study Gnet values were estimated using the AT anomaly approach [Smith and Key, 1975; Smith, 1978]. The deviations of AT measured at a single station in the NRR lagoon from its corresponding open sea values (Figure 4) reflect primarily the sum of precipitation and dissolution of CaCO3 and water exchange with the open sea. Equation (1) expresses the time dependent rate of AT change in the NRR lagoon. It is assumed that AT measured at a single station is representative of the whole community, i.e., the NRR is a well mixed reservoir (see discussion below), and that the open sea is an infinite reservoir with constant AT. These assumptions were tested and validated in a number of field studies (see below).
A is the horizontal surface area of the reef; Zt is the average depth of the entire NRR at any given t; LG is the subscript designating the value of AT measured in the NRR lagoon at any given t; OS is the subscript designating the value of AT measured in the open sea, which is assumed to be a constant value throughout the duration of the study; τt is the residence time of water in the NRR at any given t.
 The second term (on the right side of equation (1)) expresses the loss or gain of AT resulting from water exchange between the reef and the open sea. The exchange coefficient (A · Zt/τt) varies with time according to the ebb and flood of the tide, which affects the reef water volume, and causes variations in the surface current. The general southerly current was verified in five field studies with Inter-Ocean S4 current meters deployed in the fore reef, reef flat and the lagoon of the NRR. Using these data we were able to make a rough estimate of the daily average residence time (τ) of water in the reef, which ranged between 3 and 6 hours. Unfortunately, high quality current measurements in such an environment are nearly impossible because of the noise introduced by the complex reef topography. τ was estimated independently from a salt and water budget between the reef and the open sea (equation (2)).Continuous salinity records measured in the reef oscillated above a minimum value, which corresponded to its relatively constant open sea value. Hence, the daily average reef salinity, which retains a relatively constant positive deviation from its open sea value, could be considered to be at steady state for the duration of the study. This deviation is a function of the daily average evaporation rate and the residence time of water in the reef as demonstrated by a salinity record from the NRR lagoon and fore reef measured during March 2003 (Figure 5a). The relative isolation and flushing of the NRR is more clearly demonstrated by a salinity record measured during a rain event, which occurred during January 2002 (Figure 5b). Note, how reef salinity recovers from the perturbation attaining its quasi steady state value within a few hours. Using a first order box model approximation of salinity in the reef (dSLG/dt = 1/τ·(SOS − SLG) + E/ZLG·SLG), it is possible to calculate a residence time of water in the reef of 5.3 hours for this event. Using the same model and taking into account an evaporation rate of 5–8 mm·d−1, average water depth of 1.5–2 m and τ of 4–6 hours, we calculate a theoretical steady state reef salinity, which is ∼0.01–0.05 higher than open sea (Figure 6). The quasi steady state behavior of salinity in the NRR lagoon remained consistent over 4 years of observations between 2001 and 2004 (data not shown) and it is a result of the interaction between the diurnal changes in wind stress, evaporation and the tidal cycle.
 Following equation (1) again for a daily average steady state in reef water AT relative to the open sea (ΔAT/Δt for 24 hours = 0, considering the effect of salinity changes on AT as well) it is possible to calculate the diurnal average Gnet (equation (3)). The factor of 0.5 in equation (3) converts calcification rates to units of moles carbon per m2 per day.
The overbars in equations (2) and (3) indicate that the variable is assigned a 24 hour average value, S stands for salinity, and E is the evaporation rate.
 Typically AT decreases from early morning to a minimum value during the mid to late afternoon as a result of net CaCO3 precipitation (Figure 4). After attaining its minimum daytime value AT starts rising because Gnet (Ggross − D) is lower than the incoming AT flux from the open sea until a maximum value is reached during the night. This value is usually slightly lower (5–10 μeq·kg−1) than its corresponding open sea AT. However, in 7 out of 18 field studies conducted during the period 1997–2002 including that presented in Figure 3, nighttime AT in the reef rose above open sea levels (taking into account the conservation of AT with salinity) providing direct evidence for net CaCO3 dissolution (D) during this time (Gnet < 0). In these studies the rate of dissolution was estimated using equation (4), which is similar to equation (3). These values of D represent the high end estimates for this process in our system. Independently, it was also possible to estimate D values for periods where lagoon water AT did not exceed its corresponding open sea level as follows: We calculate a theoretical reef water AT increase from its minimum value as a result of water exchange alone (equation (5)). The difference between the measured and the calculated AT curves was often negative indicating CaCO3 dissolution (Figure 4). This was observed in 13 out of 18 studies (including the 6 studies discussed above). The time integrated AT difference was used to estimate CaCO3 dissolution (Drec, equation (6)). Both our estimates of D represent a minimum value because many organisms continue to calcify and therefore lower AT against the increase caused by D.
2.4. Justification for an Eulerian Approach for Assessing Community Level Processes (What do Measurements at a Single Station in the NRR Lagoon Represent?)
 The cross-shore changes of chemical and physical parameters are cumulative as water flows into the lagoon from the open sea reservoir (Figures 7a and 7b). In addition, north south variations in these parameters in the NRR lagoon, reef flat and fore reef were small and insignificant between the Middle and South profiles (Figure 8). Based on these observations, which were carried out in detail in a number of field campaigns, we conclude that the southern part of the NRR behaves as a well mixed reservoir. This condition is a prerequisite for using flushing time of reef water as the timescale for community metabolic rates [Monsen et al., 2002]. Hence, measurement of the diurnal cycle of chemical constituents at a single station represents the cumulative effects of biological (calcification) and physical (evaporation, water exchange) processes affecting the water as it flows in over the entire reef from the open sea.
 During the winter, surface cooling in the northern Gulf of Eilat lowers water temperature below 21°C, which drives a deep vertical mixing (>400 m) of the open sea water column, causing surface nitrate to increase markedly (maximum of 1.8 μmol·L−1, Tables 1 and 3). In spring, after the onset of stratification, surface nitrate concentrations drop below 0.05 μmol·L−1 and remain low throughout the summer. It is worth noting that nitrate concentrations vary by two orders of magnitude between summer low and winter high. Hence, changes in this parameter may have a significant effect on community calcification and dissolution of CaCO3. Reef water temperature varied between a summer high of ∼27.8°C and a winter low of 20.8°C. Temperature and open sea nitrate appear to be inversely related, however below 23°C nitrate is O(10−2) μmol·L−1 during March 2001/2002 and May 2001. Daily average values of reef water Ωarag varied between 3.7 and 4.4. These relatively high Gulf water values are associated with the high AT (∼2500 μeq·kg−1), which reflects its high salinity (∼40.7 PSU, Tables 1 and 3). Ωarag increases during summer mainly because of the temperature's influence on the CO3−2 ion concentration at relatively constant pCO2. However, contrary to the expected below 23°C Ωarag values increase with decreasing temperature to a maximum value of 4.18 in January 2001 when the temperature was 21.22°C (Table 3).
Table 3. Daily Average Values of the Physical and Chemical Parameters Measured in the NRR Lagoon and Open Sea (OS Subscript)
The pH value is the daily average of pH measurements at 25°C.
Daily average AT taking into account changes in lagoon relative to open sea salinity (specific AT = AT(measured) · SOS/SLG).
These values were calculated with daily average reef water pH (in the table) corrected for in situ temperature, AT and the NBS activity scale apparent thermodynamic dissociation constants for the carbonate system according to Mehrbach et al.  and the aragonite solubility product from Mucci , calculated from daily average temperature and salinity in the reef.
This value was measured one week before the field study.
 The daily average differences between reef and open sea salinities varied between 0.02 and 0.04 and are at the sensitivity limit of conductivity cells as specified by the manufacturer per measurement. However, the standard deviation of the 24 hour moving averages in reef and open sea salinity records was on the order of ∼10−3. Evaporation rates varied between a summer low of ∼5 mm·d−1 and winter high of ∼8 mm·d−1. Warmer than atmospheric sea surface temperatures during the winter cause a destabilization of the atmospheric boundary layer and positive buoyancy flux resulting in higher evaporation rates during the winter. The average reef water depth varied also between a summer low of 1.5 m and a winter high of 1.8 m. This seasonal range (0.3 m) is in good agreement with observations reported by Monismith and Genin . Residence times of water in the NRR calculated with equation (2) and using the daily average salinity differences, evaporation and reef water depth appearing in Table 2 vary between 3.6 and 6.4 hours. These values are in good agreement with the estimates made with the few direct current observations made in the NRR.
 Estimates of Gnet and D for field campaigns in the NRR during 2000–2002 were calculated with the daily average values of specific AT, Z, E, SLG and SOS presented in Table 3. In these studies Gnet varied between 60 ± 20 and 30 ± 20 mmol·m−2·d−1, and was higher during the summer than during the winter (Table 4). These values are significantly lower than those measured on reefs in the Indo-Pacific, where average Gnet = 165 mmol·m−2·d−1 [Kinsey, 1985]. The current NRR rates are lower by a factor of 2–4 compared to the rates measured for this reef in the early 1990s [Barnes and Lazar, 1993; Silverman et al., 2004]. This is mostly due to the substantial decrease in live coral cover caused by local anthropogenic stresses [Zakai, 2000; Loya, 2004]. The average D was 13 ± 7 mmol·m−2·d−1 but the values obtained by direct measurement were consistently higher (by a factor of ∼2) than those estimated by the AT repletion method (Table 3). Because actual D is most likely larger than the estimated (see above) and no seasonal variability was observed, we used the average D (20 ± 10 mmol·m−2·d−1) obtained from direct measurements for further calculations. This value is at most ∼30% of the annual average Ggross indicating that CaCO3 dissolution in coral reefs is a significant process, which should be taken into account in their CaCO3 budget.
Table 4. NRR Rates of Net Community Calcification (Gnet) and Community Dissolution of CaCO3 Calculated With Equation (4) for Dmax and Equations (5) and (6) for Drec
These field studies were conducted under extreme weather conditions of cloudiness, strong winds, and rain. Therefore, Gnet values from these studies were considered anomalous and were not taken into account in the discussion of the results.
 In the following section we analyze and discuss the relations of Ωarag, temperature and nutrients with the measured rates of community calcification in the NRR. Apparently Gnet correlates very well with temperature (Figure 9); however, while Gnet increases between 23°C and 27°C it seems to be inversely related to temperature between 20°C and 23°C. Gnet decreases in the upper temperature range at a rate of ca. −12% per °C (relative to the maximum Gnet, dashed trend lines in Figure 9). Measurements conducted in the NRR during the late 1990s representing the high temperature range display Gnet values, which are ca. 40% higher than those obtained during 2000–2002 and yet maintain the slope of −12% per °C. These relations suggest that other environmental parameters other than temperature may influence Gnet.
 Tagging each data point with its daily average Ωarag shows that above 23°C, Gnet increases with Ωarag as well as with temperature. Furthermore, sorting the data points in Figure 9 according to the open sea nitrate concentrations with squares and triangles above and below 0.2 μmol·L−1, respectively, shows that for the same temperatures and the same Ωarag values high nitrate data points show lower Gnet values compared to low nitrate ones. This is in accord with laboratory studies that have demonstrated decreased rates of coral growth when exposed to nutrient enriched water albeit with concentrations much greater than those measured in our study [Marubini and Davies, 1996; Marubini and Atkinson, 1999; Marubini and Thake, 1999; Ferrier-Pages et al., 2000; Langdon and Atkinson, 2005]. The specified NO3−1 threshold (0.2 μmol·L−1) merely represents the trophic condition of the open sea separating between eutrophic winter mixing and oligotrophic summer stratification and should not be taken as a biological threshold. As expected, the seasonal variations in Gnet and NO3−1 suggest an inverse relation between them. However, the increase in Gnet with decreasing temperature below 23°C, where many of the nitrate values are above 0.2 μmol·L−1 implies that Gnet is affected by an additional environmental parameter which overrides the effects of increased nutrients and low temperature.
 The increase in Ωarag is caused by the increase in CO3−2 concentration with temperature in an open system (constant pCO2). However, below 23°C Ωarag increases while temperature decreases, contrary to the expected. This is caused by excess production over respiration both in the reef [Silverman et al., 2004] and in the open sea [Iluz, 1997; Lazar and Erez, 2004], which causes an increase in pH and hence in CO3−2 concentration. This may be analogous to the findings of Ferrier-Pages et al. , who showed that corals exposed to higher nutrients showed lower growth rates compared to the low nutrient control. However, continued exposure to higher nutrient levels increased their production to respiration ratio and their growth rate also increased but was still lower than the low nutrient control. For the low nutrient data points note the increase in Gnet and Ωarag at similar temperatures in the same range for the data points measured in the late 1990s in comparison to those measured during 2000–2002 (Figure 9 and Table 3). Additionally, the two data points found between 21–22°C and Gnet ∼50 mmol·m−2·d−1 (Figure 9 and Table 3), which were measured at the end of March 2001 and 2002 and are not associated with high nutrients (Tables 3 and 4), have an Ωarag of ∼3.8. This is a result of the open water spring bloom, which occurred very near or during the time of these measurements. Finally, plotting lines of constant Ωarag for the NO3−1 < 0.2 μmol·L−1 data points alone displays roughly horizontal lines (Figure 9). Hence, the partial derivative of Gnet with temperature at constant Ωarag appears to be nearly 0 suggesting that Gnet is positively correlated with temperature mainly through its effect on Ωarag.
 In Figure 10 we plot Gnet vs. Ωarag, which reveals a general positive correlation between the two variables. Sorting the data points according to the open sea nitrate above and below 0.2 μmol·L−1, yields two improved regression lines (Figure 10). In Figure 9 data points below 23°C, which were associated with high nutrients and inversely correlated with temperature, suggest a positive effect of nutrients on calcification. Figure 10 shows that the dependence of Gnet on Ωarag under high nutrient conditions is reduced. For example, in the Ωarag range 3.7–4.4, under high NO3−1 Gnet is lower by 10–40% compared to low NO3−1 values in agreement with some experimental results of Marubini and Davies , who showed a 25% growth rate reduction in corals exposed to nutrient enriched waters with NO3−1 = 1 μmol·L−1. Furthermore, Langdon and Atkinson  showed in a recent study that exposure of corals placed in a flume to nutrient enriched water causes a decrease in the dependence of coral growth (measured by alkalinity anomaly) on Ωarag relative to nutrient depleted water.
 The relation of Gnet with Ωarag in Figure 10 is expressed by a power rate law similar to previously described laws for inorganic precipitation and dissolution of aragonite [Zhong and Mucci, 1989]:
where R is the precipitation rate, K is a constant which includes the kinetic rate constant and the reactive surface area of the CaCO3, and n is the order of the reaction. Typical values of n in inorganic precipitation experiments are between 2.3–2.6 (Op. Cit.). We modified this equation to include the fraction of reef area actively depositing CaCO3 (Cc, i.e., Coral cover). We also added a dissolution term, D as discussed above in order to get values for Gnet as follows:
The best fits of the data (assuming Cc = 0.3, i.e., 30% live coral cover), yield K = 8.5 and 22 mmol·m−2·d−1, and n = 3.2 and 2.1 for low and high nitrate conditions respectively (Figure 10). These values of n are similar to those obtained from experiments of inorganic aragonite precipitation (Op. Cit.). If we fit the rate equation for the entire data set, K and n are 12 mmol·m−2·d−1 and 2.8, respectively. We validated the low and high nutrient rate equations for the NRR by plotting (empty markers in Figure 10) the Gnet values (normalized for the twofold reduction in live coral cover [Zakai, 2000; Loya, 2004]) measured in 1991 during the summer (data from Korpal ) and winter (data from Barnes and Lazar ). Note the agreement between the early 1990s data points and both the fitted rate equations. From this analysis it seems that the dominating environmental parameter controlling community calcification in the NRR is Ωarag (Figures 9 and 10).
 A number of studies have discussed the dependence of coral growth on temperature. These experimental studies showed a typical optimal growth rate between 24°C and 27°C [Clausen and Roth, 1975; Coles and Jokiel, 1978; Marshall and Clode, 2004]. In these studies coral growth was measured as the rate of 45Ca+2 incorporation into the coral skeleton during short term (∼1 hour) incubations [Goreau, 1959]. In these studies the apparent growth increase with temperature (until the optimum) was also associated with a considerable increase in Ωarag (according to our rough calculation by 0.5–0.8 Ωarag units) under the aerated experimental conditions. Hence, the temperature effect cannot be separated from the Ωarag effect in these studies. In the temperature range above the optimum, the rate of calcification decreased despite an increase in Ωarag indicating a direct negative effect of temperature on the coral physiology. Coral reef communities are composed of many different species of corals, as well as calcareous algae and other benthic invertebrates, which deposit CaCO3. Each of these organisms may be affected differently by temperature. Hence, it is possible that an optimal calcification temperature may not be observed by community level Gnet measurements.
 On the other hand, coral growth rates measured in situ by linear extension and skeletal density, and buoyant weight changes showed a positive and linear relation with temperature over different timescales of days, weeks, months, years and different geographical locations [Shinn, 1966; Weber and White, 1974; Glynn, 1977; Lough and Barnes, 2000]. The majority of the data in the Lough and Barnes  study, who measured the extension rate of massive corals (mainly Porites from the Great Barrier Reef (GBR), Australia) in the last ∼100 years, showed a linear increase in growth rate with temperature between 23°C and 27°C. We also observed a linear increase in community calcification with temperature between 23°C and 27°C (Figure 9) and the percent change of growth rate per °C (relative to the maximum) was −12% similar to that of Shinn  (−9% per °C), Weber and White  (−8.4% per °C), Glynn  (−10% per °C), Lough and Barnes  (−13% per °C), and Bessat and Buigues  (−11% per °C). Unfortunately, carbonate system data are not available for these studies.
 The difference in K and n between high and low nutrient data points plotted in the field of Gnet vs. Ωarag (Figure 10), and the covariance of temperature with nutrients below 23°C (except for March 2001/2002 and May 2001, Table 3) suggests that K and n may be temperature dependent. Empirical relations between K, n and temperature for inorganic precipitation of aragonite were developed by Burton and Walter . It is possible to calculate the inorganic precipitation rate (GB&W) of aragonite for each study period in the NRR using equation (7), and the daily average temperature and Ωarag as input (K(T) = −0.0177·T2 + 1.4697·T + 14.893 in units of μmol·m−2·hour−1; n(T) = 0.0628·T + 0.0985, derived from the data in Burton and Walter  for the 5–40°C temperature range). Plotting GB&W vs. the actual Gnet for all of the NRR field studies including both high and low nutrient data points displays a good correlation between the two (Figure 11, n = 17, R2 = 0.92). GB&W values are lower than their corresponding Gnet values by a factor of ∼10 possibly reflecting the surface area used to calculate the rates (horizontal surface area (NRR), which is lower than the surface area of the actively precipitating CaCO3 used in the Burton and Walter study). Another likely possibility is the higher rate of CaCO3 precipitation expected in corals because it is biologically mediated (e.g., light-enhanced calcification [Barnes and Chalker, 1990]). Additional Gnet and GB&W values plotted in Figure 11 were obtained from a comprehensive study conducted by Ohde and van Woesik  in the Rukan-Sho Atoll in the south of Okinawa, Japan, in the early 1990s. This reef flat has a live coral cover (36%), which is slightly higher than the NRR's. Although there is a slight increase in the slope of the general trend line (NRR and Rukan-Sho data points), which may reflect the slightly higher coral cover, the linear trend and slope appear to remain consistent (n = 22, R2 = 0.94). The use of a chemical approach to describe the biologically mediated process of coral calcification is supported by experimental evidence, which demonstrates the unequivocal effect of CO32− or Ωarag on calcification in corals [e.g., Schneider and Erez, 2006]. This suggests that while various physiological factors may affect the biomineralization process (e.g., light-enhanced calcification) the organism is strongly affected by the carbonate chemistry of the water in which it calcifies. The present study with its detailed analysis demonstrates that the simultaneous effects of temperature and Ωarag on coral community calcification resemble their kinetic effects on inorganic precipitation of aragonite. Obviously very little of the CaCO3 precipitation in the reef is inorganic, and hence one must conclude that corals are behaving as an open system with respect to the seawater in which they calcify. Using this functional relationship between calcification, temperature and Ωarag it is possible to predict changes in calcification with temperature and pCO2 as displayed in Figure 12. In this chart lines of constant Ωarag and pCO2 are plotted in a field of calcification vs. temperature (calculated for a constant AT of 2250 μeq·kg−1 and salinity of 35). At low temperatures lines of constant pCO2 and Ωarag are nearly horizontal indicating that changes in pCO2 only can drive relatively large changes in calcification. In contrast, as temperature increases the slope of pCO2 and Ωarag contours increases significantly indicating a growing influence of temperature on calcification especially at lower pCO2 values. However, at high pCO2 (>600 μatm) although changes in calcification are relatively small they are influenced mostly by changes in pCO2 with a minor influence of temperature. A similar behavior was reported by Reynaud et al. , who studied the interacting effects of pCO2 and temperature on calcification and photosynthesis under laboratory conditions in single corals. They showed that at low temperature (25°C) changes in pCO2 from 450 to 734 μatm had no effect on calcification. While at high temperature (28°C) calcification decreased by almost 50% for pCO2 of 798 relative to 470 μatm. Using equation (7) and the Burton and Walter  rate constants, we calculate a 55% decrease in calcification at 25°C, and a 58% decrease at 28°C for the above changes in pCO2. It is encouraging that at least the experimental change in calcification at high temperature agrees with our calculation. Calculating the changes in calcification (GB&W) with temperature at pCO2 of 280 to 360 μatm, (preindustrial to 1990s level) in the temperature range 23–28°C we obtain an average value of 10.4 ± 0.1% per °C (relative to the maximum value). This value is very similar to those obtained from in situ measurements of coral growth vs. temperature cited above including the comprehensive data set of Lough and Barnes . Finally, assuming a 2°C temperature increase and pCO2 doubling (560 μatm), the change in GB&W from preindustrial conditions (T = 28°C, pCO2 = 280 μatm, AT = 2250 μeq·kg−1 and salinity 35), is a decrease of 55%. This value is in total disagreement with the ∼+20% proposed by McNeil et al. , who concluded that the contribution of temperature to calcification far outweighs the reduction due to acidification.
 It is most probable that Ωarag, temperature and nutrients may affect the dissolution rate of CaCO3 (D) in coral reefs. Yates and Halley  have shown that different type substrate enclosures in a coral reef exposed to increasing levels of pCO2 display decreasing rates of Gnet well below 0, i.e., net dissolution. This indicates that D is indeed affected by Ωarag. Glynn  has shown that eutrophication increases coral infestation with boring bivalves, thus increasing bioerosion. It can be speculated that with increasing nutrient levels respiration rates in coral reefs should increase due to higher plankton consumption and increased productivity within the reef ecosystem. It is also safe to speculate that higher temperatures will increase respiration of the entire community as well as that of the endolithic community and both should increase D. Obviously the decrease in Ωarag of the ambient surface water (due to atmospheric CO2 increase) should also increase D. Therefore, our estimate of D should be considered as minimal.
 Community calcification of the NRR varies annually with temperature, Ωarag and nutrients in a manner which agrees with most previous studies at the organism level both in the laboratory and in situ. Furthermore, we have shown that community dissolution has a significant influence on the CaCO3 budget of coral reefs and considering the potential affects of acidification, warming and eutrophication on this process it should be further investigated in future studies. Finally, our interpretation of the interacting affects of temperature and Ωarag on calcification suggests that the entire coral reef and perhaps large coral colonies as well may start to dissolve in the future when atmospheric CO2 doubles.
horizontal surface area of the reef, m2.
total alkalinity in seawater, μeq·kg−1.
fraction of the coral reef horizontal area precipitating CaCO3 or coral live coverage percentage.
rate of coral reef community CaCO3 dissolution, mmol·m−2·d−1.
rate of evaporation from the sea surface, mm·d−1.
inorganic precipitation rates of aragonite calculated with temperature, Ωarag and the empirical relations of Burton and Walter  for n and K, mmol·m−2·d−1.
gross rate of coral reef community CaCO3 deposition, mmol·m−2·d−1.
net rate of coral reef community CaCO3 deposition, mmol·m−2·d−1.
CaCO3 precipitation rate coefficient, which includes the kinetic rate constant and the reactive surface area, mmol·m−2·d−1.
order of the reaction resulting in CaCO3 precipitation.
partial pressure of atmospheric CO2, μatm.
daily average salinity in the reef, PSU.
salinity in the open sea surface water, PSU.
daily average water depth of the entire reef from the reef crest to the shore, m.
aragonite degree of saturation.
residence time of water in the reef, hours.
 Funding for this project was provided by the Israel Science Foundation and US-AID. We acknowledge the technical assistance in the field of M. Dray, T. Rivlin, A. Rivlin, M. Lazarovich, and R. Shem-Tov, the help of the staff at the Marine Biology Laboratory (IUI), Eilat, and the use of facilities as members (B.L. and J.E.) of The Moshe Shilo Center for Marine Biogeochemistry. We thank C. Gildor and W. Silverman for helpful comments on this manuscript.