The Rate of Air-Sea CO2 Exchange: Chemical Enhancement and Catalysis by Marine Microalgae.

Chpater 8: CO2 exchange rates with algae: results and discussion

8.1 Overview

To investigate whether marine algae can catalyse the exchange of CO2 between air and seawater, eight blooms of five different species of algae were cultured in the gas exchange tank: four blooms of a green alga Dunaliella, one of a diatom Phaeodactylum, one of a dinoflagellate Amphidinium, one of another diatom Skeletonema and finally a bloom of the coccolithophore Emiliana Huxleyi. The choice of species was determined by a combination of factors: availability, diversity, ease of culturing, and whether they were known to produce carbonic anhydrase. Dunaliella was used four times as the results from the first bloom seemed particularly promising. I also took samples of seawater directly from the North Sea spring bloom in May and June 1996, which was dominated by Phaeocystis.

Each bloom lasted about one month, during which CO2 gas exchange was measured daily and other parameters occasionally. The temperature and motor stirring speed were kept constant at 15C and 13 rpm respectively (the temperature and stirring motor control systems were improved during this period so variation was greatest at the beginning). The water pCO2 fell during the initial growth stage of each bloom due to photosynthetic uptake by the algae. In most cases the pCO2 fell much lower and the algal density rose much higher than would be expected at sea, due to the abundance of nitrate and phosphate. N and P concentrations were the same as f / 2 culture medium - except for the North Sea samples (to which nutrients were not added initially), and the Emiliana Huxleyi bloom (for which lower nutrient levels were used to encourage calcification). The extreme conditions created by these high nutrient levels were welcome, because they were expected to amplify any catalysis of the CO2 transfer velocity by carbonic anhydrase produced in situ, which was the focus of this study.

The results from each bloom are discussed here in chronological order following a few general points about the presentation of the data. The transfer velocities from all the blooms will then be compared in composite plots which include control measurements from the previous chapter. A summary of all the blooms is given in table 8-1 below.

At the end of four of the blooms an algal film developed on the surface of some headspaces, and comparison of the headspaces suggested that the algae were taking up CO2 directly from the air. The gas exchange data during the periods with films has been excluded from the composite plots, and this unexpected but intriguing phenomenon is discussed in a separate section (8.5) at the end of this chapter.

8.2 General features of the graphs describing blooms

For each bloom, the measurements are summarised by two graphs, which should be considered together. The x-axis for both shows the time in days (starting from the beginning of the relevant month).

8.2.1 Gas exchange plot

The first graph shows the measured CO2 transfer velocities from both influx and efflux headspaces. It also shows the diffusion-only transfer velocity predicted from temperature and stirring speed based on the results of control experiments using acidified seawater data (as presented in chapter 7), and the enhanced transfer velocity predicted using the Hoover and Berkshire equation and carbonate chemistry based on measured pCO2. The "influx" measurements (high air pCO2) are generally more reliable than the "efflux" measurements (low air pCO2) for reasons already discussed in Section 6.5 . Note that the error bars are intended to give an indication of the maximum range of experimental error, which is a much wider range than a statistical standard error (see Section 6.5 ). The variability among the diffusion-only and Hoover and Berkshire predictions illustrates the effect of varying temperature, stirring rate and pH / pCO2. However since the Hoover and Berkshire prediction underestimates the effect of the uncatalysed reaction (see Section 7.7 ) it does not serve as a "control". For a comparison of CO2 transfer velocity with and without algae as a function of pCO2, refer to the composite plots figure 8-11 , figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 figure 8-16 , and the discussion in Section 8.4 . Transfer velocities for SF6 and oxygen are also shown when they were measured, but these are not very reliable and have already been discussed in

8.2.2 Carbon budget plot

The second graph shows the "carbon budget" for each bloom. Of the connected data points, the water pCO2 is the only quantity which has been measured directly, whilst the TCO2 and pH have been calculated assuming a constant Total Alkalinity. This was measured occasionally as shown on the graphs, and was found to be almost constant except during the Phaeodactylum bloom probably due to spontaneous calcification (see Section 8.3.3 ). There was also a slight variation during the Emiliana Huxleyi bloom, which was accounted for in the carbonate system calculations. Note that the Total Alkalinity during these experiments is slightly higher than typical oceanic levels for two reasons: the added phosphate (concentration as f/2 medium) adds about 0.05mM, and the seawater just off Lowestoft is influenced by the input of high-alkalinity water from the river Yare. Section 6.3 and figure 6-2 shown earlier demonstrate the importance of the phosphate contribution.

Measured values of TCO2 and pH are also shown where available, for comparison with the calculated values. These give an indication of the internal consistency of the carbonate system measurements. For a comparison of all measured and calculated values refer to figure 6-1 . Note that the pCO2 and TCO2 measurements are much more accurate than the crude pH and alkalinity measurements (see Section 5.5 , Section 5.9 , Section 5.10 ).

The observed pCO2 changes are in part due to uptake by the algae, but also due to deliberate changes in the pCO2 of the air flowing into the headspaces, which could be adjusted to gradually control the pCO2 in the water. Whether there was a net gain or loss of CO2 by the water is indicated by the "Total Carbon" curve. "Total" and "Biological" Carbon have been calculated by mass balance, since the CO2 fluxes in and out of the tank headspaces are known and the total CO2 stored in the headspaces is insignificant compared to the total CO2 in the water.

Total Carbon = Initial TCO2 + integral over time of net CO2 influx into headspaces

Biological Carbon (i.e. organic + particulate) = Total Carbon - TCO2

This calculation assumes that there is no "Biological Carbon" at the beginning of each bloom, which ignores the initial dissolved organic carbon and the carbon in the algae of the starter culture. However the offset due to this assumption should be small and constant.

The "biological" carbon can be compared with the chlorophyll values where these were measured. The scale on the chlorophyll axis was chosen in each case to bring the measurements to coincide with the "biological carbon" curve. Figure 8-1 shows all the chlorophyll data together, plotted as a function of calculated "biological carbon". This shows that the ratio of chlorophyll to biological carbon is fairly constant for all the Dunaliella measurements, and similar for two of the Phaeodactylum measurements. This ratio is lower but constant for the three Amphidinium (Dinoflagellate) measurements, and much lower for the other species. The particularly low chlorophyll measurements for Skeletonema, and the single measurements for Emiliana Huxleyi and Phaeodactylum at the right hand side of the plot, were taken after most of the algae had settled out at the end of the bloom, thus explaining the low chlorophyll /biological carbon ratio in these cases.

Temperature and stirring paddle speed were checked during each set of measurements and these values were used in the subsequent calculations, but this data is not shown on the graphs as the small variation is not of great interest (except during the second Dunaliella bloom). The range for each bloom is given in table 8-1 . Note that for the North Sea samples, some measurements were made at the sea temperature, not the normal tank temperature, hence the difference in pCO2.

For more details of the calculation methods, explanation of the "error-range" bars, and discussion of other uncertainties, refer to chapter 6.

8.3 Results from each algal bloom

8.3.1 First Dunaliella bloom

(This should be read with reference to figure 8-2 a and b)

The first species investigated was Dunaliella, a green alga. Although it is not especially common in the open sea, it is easy to culture and is known to produce carbonic anhydrase (see references in table 2-1 ). The starter culture (800ml) was added to the tank on 16th November 1995. The rapid fall in pCO2 was monitored but CO2 gas exchange measurements were not begun until 6 days later (for technical reasons), although oxygen and SF6 exchange rates were measured and found to be within the normal range (see Section 7.4.1 ). By 23rd November the pCO2 had by then already dropped to about 40ppm, and the CO2 transfer velocity was 5 - 6 cm hr-1 , slightly higher than expected from the uncatalysed chemical enhancement, suggesting an enzyme effect. The pCO2 continued to fall to a minimum of 9ppm on 28th November, and the transfer velocity rose correspondingly to over 11 cm hr-1, much higher than expected without catalysis. Subsequently the pCO2 rose slightly, due to a deliberate increase in the pCO2 of the air flowing into the"influx" headspace, and the transfer velocity began to fall again.

I then had to go away for four days, after which (5th December) most of the algae had settled out and a film had appeared on the surface. This gave rise to an even higher transfer velocity (about 13 cm hr-1) but the interpretation of the CO2 flux may be confused by the film, which is discussed further in Section 8.5.2 . The SF6 and oxygen transfer velocities were then measured again, but there was little change (a slight drop for SF6). For several days following this, the pCO2 remained about 90 ppm and the CO2 transfer velocity remained in the range 5-7 cm hr-1.

On 13th December (after the CO2 transfer velocity measurement), 1mg l-1 dansylamide - the same fluorescent sulphonamide as used for the carbonic anhydrase measurement technique investigated earlier - (see Section 3.5 ) was added as a potential enzyme inhibitor. Although some other sulphonamides are stronger inhibitors, they are also more toxic and a safe handling procedure for these had not yet been approved. In any case, the addition of the dansylamide did not seem to have any effect, probably because by this stage the surface film influenced the transfer velocity more than any enzyme.

On 16th-18th December the water pCO2 was raised again back to about 300ppm and on 19th-20th December it was lowered again to about 90 ppm, by adjusting the pCO2 of the air flowing into the influx headspace. Both influx and efflux transfer velocities fell and rose again correspondingly.

Considered as a whole, the plots of transfer velocity (influx measurements) and water pCO2 show an almost perfect mirror image. Since at this time I believed that the Hoover and Berkshire adequately predicted uncatalysed chemical enhancement, and as can be seen on the figure this predicts a much smaller response to the pCO2 variation than the measured transfer velocities. I therefore attributed the inverse relationship between pCO2 and CO2 transfer velocity almost entirely to algal physiology as described earlier in Section 2.3.4 and Section 3.2 - i.e. the high gas exchange rate was assumed to be due to the production of carbonic anhydrase by algae in response to the low pCO2.

Subsequently low pCO2 "blank" measurements were made as a control (the data reported in Section 7.5.1 ), and these suggest that some of the inverse relationship is due instead to the OH- reaction pathway. Nevertheless the peak transfer velocities measured during this bloom are significantly higher than the corresponding blank measurements without algae, as shown by the composite plots ( figure 8-11 , figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 , figure 8-16 ) and discussed in

8.3.2. Second Dunaliella bloom

(This should be read with reference to figure 8-3 a and b)

Since the first bloom had been so promising, it seemed wise to try to repeat this experiment. A similar starter culture was added on 13th January, this time to filtered seawater from the Lowestoft fisheries laboratory supply. During the second Dunaliella bloom, however, the algal growth was much less intense and seemed to be limited by some unknown factor. After 26th January both headspaces were used for efflux measurements in an attempt to lower the water pCO2, but still it did not drop below about 120 ppm and the highest transfer velocity was about 5.5 cm hr-1. Nevertheless the relationship between the measured transfer velocities and the pCO2 was consistent with that observed during the first bloom (as can be seen from the composite plots figure 8-11 , figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 , figure 8-16 ).

During this bloom the laboratory temperature fluctuated dramatically (due to cold frosty nights!). This affected not only the water temperature, since the temperature control system at that time was crude, but also the stirring motor speed. These physical factors caused the noticeable fluctuation in the predictions of the diffusion-only and enhanced (Hoover and Berkshire) transfer velocities, so in this case the temperature and stirring speed variation is shown below the main graph in figure 8-3 a. It is satisfying to note that the day-to-day fluctuations in the measured transfer velocities coincide well with the predictions, although as before the measurements rise gradually above the predictions as the pCO2 falls.

On 28th January the algae began to settle and soon after this the pCO2 began to rise again. As it seemed that there was little further prospect of intense growth as during the first bloom, I decided to investigate whether addition of acetazolamide (a more effective sulphonamide inhibitor of carbonic anhydrase) could bring the transfer velocity back down to coincide again with the predictions. 1.2 mg l-1 acetazolamide was added after measurements on the evening of 31st January. The next day the transfer velocity had dropped significantly, almost back to the Hoover and Berkshire prediction, suggesting that catalysis by carbonic anhydrase was indeed the reason for the observed effects. The composite plots ( figure 8-11 , figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 , figure 8-16 ) show that the transfer velocities measured during this bloom is consistent with those from the first Dunaliella bloom, and lie above the equivalent no-algae transfer velocities.

8.3.3 Phaeodactylum bloom

(This should be read with reference to figure 8-4 a and b)

We then decided to try a species of diatom Phaeodactylum. The starter culture, together with nitrate, phosphate and silicate (which is required by diatoms -see Section 5.2.3 ) was added to the tank on 19th February. Over the next few days the pCO2 fell very rapidly. Although both headspaces were initially used for efflux measurements (as at the end of the previous bloom), the carbon budget curves show that photosynthetic uptake dominates over the gas exchange flux. However the CO2 gas exchange rate did not rise until 27th February, when the pCO2 was already less than 10ppm. Influx measurements were used subsequently as at such low pCO2 the error in efflux measurements would be very large. Over the next two days the CO2 transfer velocity shot up dramatically to a maximum of almost 20 cm hr-1 on 29th February, possibly indicating a sudden onset of enzyme production.

Unfortunately the CO2 transfer velocity then fell again even more suddenly. This coincides with the development of a thin algal film on the surface, which may have inhibited the gas exchange. At this time too one air-stirring paddle fell into the water, possibly providing a sudden input of trace metals which might have interfered with algal physiology or inhibited carbonic anhydrase.

However the most likely explanation for the sudden drop in the CO2 transfer velocity is a spontaneous calcification in the water causing a sudden rise in pCO2. When the tank was opened at the end of the bloom 25 days later the total alkalinity was measured and found to have dropped to about 0.7 mM (from a typical initial seawater value of 2.4 mM). Calcification is the only plausible explanation for this large change in alkalinity, which was also confirmed when the layer of sediment on the base of the tank, initially assumed to be remains of algae, dried out to become pale crystals.

Such large changes in total alkalinity had not been anticipated during the blooms in the tank since gas exchange and biological uptake of CO2 do not affect the alkalinity, so the alkalinity was not measured at the time of the drop in transfer velocity. Therefore, since the exact time of the calcification event is not known, the carbonate system speciation, pH, TCO2, and "biological carbon" cannot be calculated during this period, hence the gap in the carbon budget plot. It is likely that there was not a gradual change but a discontinuity caused by a sudden spontaneous crystallisation induced by the very low pCO2. Seawater at such low pCO2 (the lowest measured was 5 ppm!) has a high pH (9.3) and a high concentration of CO32- ions (0.9 millimolar), and is therefore is extremely supersaturated with respect to Calcium Carbonate. Although the kinetic process of calcification is slow, once a few seed crystals had formed, they might have induced a sudden crystallisation. It is unlikely that the diatoms themselves played a role in this process as they are silaceous not calcareous algae.

Removing CaCO3 from seawater causes the pH to fall and the pCO2 to rise. Although there is not a sudden change in the measured pCO2 in the "equilibrium" headspace, the slow response may be due to to the surface algal film. This film dominated the CO2 fluxes in and out of the tank for the remainder of the bloom, and will be discussed separately later ( Section 8.5.3 ).

8.3.4 Amphidinium bloom

(This should be read with reference to figure 8-5 a and b)

The next culture was a dinoflagellate species, Amphidinium. The starter culture was added to the tank on 28th March. CO2 gas exchange and other measurements were not begun until 4th April. To avoid repeating the uncertainty of the previous Phaeodactylum bloom, TCO2 and Alkalinity were measured at regular intervals. The carbon budget shows that the Alkalinity did not change and the measured TCO2 matches well with that calculated from pCO2. Chlorophyll was also measured three times. The ratio of Chlorophyll to "biological carbon" seems to be higher for the middle of these measurements than at the beginning and the end.

The pCO2 fell as expected but not so rapidly as for the previous blooms. It eventually reached a minimum of 9ppm. As the pCO2 fell, so the CO2 gas exchange rate rose, and the inverse "mirror" relationship noted in the first bloom can be seen again here. This suggested the same physiological response of the algae to limiting pCO2. However the transfer velocities were not so high as for Dunaliella. On the 20th April 0.2 mg l-1 zinc sulphate was added in case lack of zinc was limiting the production of carbonic anhydrase, but this seemed to make no difference. On 26th April 4 mg l-1 acetazolamide was added with the hope of repeating the sudden drop observed at the end of the second Dunaliella bloom, but this time it had no effect. Although the transfer velocity / pCO2 relationship looks encouraging, in the composite plots ( figure 8-11 , figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 , figure 8-16 ) it can be seen that there is little difference between the transfer velocity / pCO2 relationship for this bloom, and for the low pCO2 blank measurements carried out later. Possibly this species possesses a carbon concentrating mechanism which does not involve carbonic anhydrase.

8.3.5 Phaeocystis samples from the North Sea spring bloom

(This should be read with reference to figure 8-6 a and b)

Between 30th April and 3rd July four samples were taken from the North Sea during the spring bloom, which was dominated by the colony-forming algae Phaeocystis. The water was collected from a small fishing vessel 5 miles SE of Great Yarmouth harbour entrance in 28 to 30m depth (depending on the state of tide), by submerging two 30 litre barrels just below the surface and opening at both ends (thereby avoiding CO2 exchange with the air). Table 8-2 shows the state of the sea at the time of sampling: Note that the flood tide flows from the North.

The pH and air and water temperatures were measured in situ. On each occasion water was collected about 7.30 am and was brought back immediately to the lab and siphoned into the tank (without filtering) within 3 hours of collection. The barrels were kept in the shade to minimise heating during this period, which was typically about 1 degree. The in situ pCO2 in the sea was later estimated from the air equilibrated in the tank, correcting for the temperature change and the original pCO2 in the headspace. On the 30th April gas exchange was initially measured at 8.6C (the sea temperature) but the other three samples were heated to 15C (as for the other blooms) before beginning gas exchange measurements, which were made as soon as the headspace was expected to have reached a steady state, i.e. about 4-5 hours after setting up the gas flows. TCO2, Alkalinity and Chlorophyll were also measured on the same day, as shown on the Carbon Budget graph. The Alkalinity was always 2.33mM.

The pCO2 and TCO2 dropped significantly during the course of the North Sea spring bloom. The estimated sea pCO2 on 30th April was 500ppm, whereas by 19th June it had fallen to 245ppm, after which it rose again. Note that during this period the sea temperature rose from 8C to 14C, which by itself would have caused a substantial pCO2 increase (due to lower solubility at higher temperature). The measured chlorophyll concentrations were much lower than during the previous blooms in f / 2 nutrient culture medium, but for the sea these values are fairly high. Note that the chlorophyll drops almost to zero by the time of the last sample although characteristic Phaeocystis foam was still present on the sea surface. On the 19th June Phaeocystis colonies a few millimetres across could be clearly seen in the sea.

The water was kept in the tank for a few days after each sample was collected, to see whether there would be any subsequent change due to further growth of the algae. During the longest such period three days after the third sample (22nd June) nitrate and phosphate were added, at the same concentration as for previous cultures (as f / 2 medium). Apart from changing the alkalinity and hence causing a sudden change in pCO2, this did not seem to affect the algae, which did not bloom intensely as did the previous cultures. Zinc sulphate (0.08mM) was also added on 28th June but again this seemed to have no significant effect.

The CO2 transfer velocities were significantly higher during the third sample when the pCO2 was lowest, but they did not rise above 4 cm hr-1. The difference between these measurements and the blank measurements without algae is small, compared to the variability in the data, as shown in the composite plot ( figure 8-12 ), so it is hard to distinguish any catalysis effect.

Note that a leak was detected in the efflux headspace (A) on day 10, following 3 anomalously high gas exchange measurements.

8.3.6 Third Dunaliella bloom

(This should be read with reference to figure 8-7 a and b)

Since the first Dunaliella bloom had shown the most promising sign of an enzyme effect, but the second Dunaliella bloom had not grown so intensely, this species was investigated a third time. Unfortunately, however the algal growth seemed to be even more limited during this third bloom that it was during the second. The pCO2 eventually fell to about 80ppm, lower than during the second bloom, but the carbon budget graph shows that this was more due to a fall in the total carbon lost by gas exchange (both headspaces were used for efflux measurements after 19th August) than due to biological uptake. Correspondingly, there was not a significant rise in the transfer velocity. The biological carbon was so low that these transfer velocities should not be directly compared with the results from the other three Dunaliella blooms (as is clear from figure 8-16 which will be explained later).

8.3.7 Fourth Dunaliella bloom

(This should be read with reference to figure 8-8 a and b)

We still wondered why algal growth during the second and third Dunaliella blooms was much less intense than during the first bloom. The chief difference between the initial set-up of the three blooms was that the first used unfiltered seawater, whereas since then it had always filtered (except for the North Sea Phaeocystis samples). As all cultures had excess nitrate and phosphate, we considered that perhaps the limiting factor was a lack of vitamins, which had not been added because we had assumed they were present in sufficient quantities in the original seawater. Perhaps the first, intense, Dunaliella bloom, contained bacteria which produced vitamins in situ, whereas these were not present in other blooms due to filtration.

Therefore for the fourth Dunaliella bloom vitamins (Thiamine, B12, Biotin) were added as for "f / 2" culture. Although trace metals and iron were not expected to be limiting due to the metal components of the tank, these were also added to the fourth Dunaliella bloom.

The culture and all nutrients were added on 18th September. This time the algae grew intensely as during the first bloom, so it seems that a lack of vitamins may well have been the limiting factor. Again, the pCO2 fell very low (to 7ppm) while the CO2 transfer velocity rose correspondingly, this time peaking a day later than the lowest pCO2, at about 13 cm hr-1. Although the other transfer velocities measured during this period were slightly lower than during the first bloom at equivalent pCO2, they still lie significantly above the low pCO2 "blank" measurements without algae (see composite plots figure 8-11 , figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 , figure 8-16 ). Therefore this to some extent confirms the suggestion of an enzyme effect observed in the first and second blooms.

On 30th September 1 mg l-1 acetazolamide inhibitor was added at 2pm after the gas exchange measurement. No immediate effect was observed in the pCO2 flowing out of the tank, so at 6.30pm 2 mg l-1 of ethoxyzolamide - a more potent inhibitor for some carbonic anhydrases (see Section 2.7.3 ) - was also added. The following day the transfer velocity was about 1 cm hr-1 lower but since by this time the pCO2 was already rising and the transfer velocity already dropping before the addition of inhibitor, this was not conclusive.

Following this an algal film developed on the surface which gave rise to some very high calculated "transfer velocities" (beyond the scale of the graph) but as explained in Section 8.5.4 , these are probably not due to normal gas exchange processes but due to direct uptake of CO2 by the surface algae.

TCO2 , Alkalinity and Chlorophyll were measured regularly as shown in the carbon budget. The Alkalinity was constant at 2.55 mM, and the TCO2 measurements match the predictions well except on 10th October when the pCO2 measurement was suspect due to the presence of a surface film. The SF6 transfer velocity was also measured once and was within the normal range but there was a large uncertainty in the endpoint (see Section 7.4.1 ).

8.3.8 Skeletonema bloom

(This should be read with reference to figure 8-9 a and b)

During the winter and spring of 1997-98 the tank was used for the experiments without algae reported in chapter 7. After this, there was sufficient time to try two further algal blooms in the tank. Since the results from the Phaeodactylum bloom had been confused by the spontaneous calcification and surface film, it seemed sensible to try another species of diatom. Professor Michael Merret from the University of Swansea kindly supplied a starter culture of a strain of Skeletonema which, in his laboratory, had been found to produce carbonic anhydrase. This culture was grown in a flask in f/2 medium with added silicate in the normal way, and when it had reached a sufficient density it was added to the tank, on the evening of 14th August 1998 (note that the first pair of gas exchange measurements in figure 8-9 (a) were made before the nutrients and algal culture were added). Vitamins, trace metals, and silicate (all as f/2 medium) were added the following day.

Growth seemed quite rapid, but by the 18th algae were already beginning to settle onto the base of the tank. On 20th the base of the tank was clear again and tiny specks could be seen in the water. On 24th settled algae could be seen again on the base of the tank, and by 26th this layer was quite thick. The settled algae were resuspended by a jet of air squirted from a syringe around the base of the tank, but within a few hours most had settled again. This procedure was repeated on 29th August and 4th September. TCO2, Alkalinity, pH and Chlorophyll measurements were made on 17th August and 4th September. On 4th September Chlorophyll measurements were made both before and after resuspending the settled algae, but the chlorophyll before resuspending was below the detection limit -the measurement after resuspending the algae is shown on the graphs. Later, on opening the tank, it was apparent that the algae had flocculated into clumps, but these were not attached to the tank itself.

Although the measured transfer velocities lie well above the Hoover and Berkshire prediction and there is an inverse relationship between transfer velocity and pCO2, the composite plots ( figure 8-11 , figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 , figure 8-16 ) show that the data from this bloom match those of the "control" experiments without algae at the same water pCO2. So, as for the Amphidinium bloom, the rise in transfer velocity at lower pCO2 can be attributed to the OH- reaction, and there is no evidence of any significant catalysis. This is not surprising, since the concentration of algae suspended in the water was low due to the flocculation, although there was sufficient carbon uptake to lower the water pCO2 to 65ppm.

8.3.9 Emiliana Huxleyi bloom

(This should be read with reference to figure 8-10 a and b. The different coloured lines correspond to different values for water pCO2 as explained below)

The final bloom in the tank was of the coccolithophorid Emiliana Huxleyi. Professor Merret kindly supplied us with a starter culture of a calcifying strain of Emiliana Huxleyi which had been found to produce carbonic anhydrase in his laboratory, although only under certain conditions (see references in Section 8.4.3 ). He advised us that, from their experience, to encourage calcification we should use much lower nutrient concentrations than usual: 5 m M phosphate (1/7th of the concentration in the usual f / 2 medium) and 20m M nitrate (1/20th of f / 2). These, together with the normal f / 2 amounts of trace metals and vitamins, were added to the tank on 15th September 1998. CO2 gas exchange, TCO2, pH and alkalinity were measured before the algal culture was added to the tank late that evening.

The strangely high efflux and low influx transfer velocities on 18th September may be explained by a slight leak in the gas flow system detected later that day. Therefore the CO2 transfer velocity did not really begin to rise until the 23rd September. On the 26th, 27th and 28th September the CO2 transfer velocity rose to a maximum of about 6 cm hr-1. Although the pCO2 had fallen during this period to a minimum of about 260ppm, this in itself is not sufficient to account for the higher CO2 transfer velocities, as can be seen from the composite plots figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 , figure 8-16 , where the Emiliana Huxleyi data from this period lie well above the control measurements without algae. This was very encouraging as there seemed to be a significant enzyme catalysis effect within the "normal" water pCO2 range.

However, during this period, there was also an unusual discrepancy of 10-20ppm between the pCO2 measured in the two equilibrium headspaces "B" and "C". Usually, there is only a difference of only 2ppm or less, which is due to headspace "B" being used to flush out the LiCOR sample cell and tubing before a more accurate measurement of the pCO2 in equilibrium with the water is made using headspace "C". If the higher pCO2 values from "B" rather than "C" are used to calculate the transfer velocities, the "influx" transfer velocity is raised and the "efflux" transfer velocity is lowered. The difference can be seen in figure 8-10 (a) where both sets of transfer velocities are shown, in blue when calculated using "C" pCO2 and in green when calculated using "B" pCO2. The same colours are used to show the different pCO2 measurements, and the TCO2, pH and biological carbon calculated from these, in the carbon budget figure 8-10 (b).

A week later, on 5th October, a film had clearly developed on the surface of the water in headspaces "C" and "D" ("D" was being used for the influx gas exchange measurements). The measurements after this date will be discussed later in Section 8.5.5 and have been excluded from the composite plots. However even if there was a slight film effect in headspaces "C" and "D" before this period, the composite plots from the early stages of the Emiliana bloom comparing both sets of transfer velocities up to 5th October ( figure 8-13 ) suggest that a film alone would not be sufficient to explain the high transfer velocities during this period, and there is probably still some enzyme catalysis. This question is considered in more detail later ( Section 8.4.2 ).

The water pCO2 was maintained at a reasonably high level after 23rd September by the set-up of the steady-state gas flow system, such that the carbon uptake by algae (increase in "biological carbon) was balanced by a net influx of CO2 through gas exchange (increase in "total carbon), with little change in the water TCO2, as can be seen from the carbon budget figure 8-9 (b). TCO2 and alkalinity measurements were made on both filtered and unfiltered samples on 18th, 23rd and 28th September, in the hope of detecting any calcification by formation of coccoliths: any solid CaCO3 present in the unfiltered samples should be dissolved by the acid used to drive off the CO2, and result in higher measurements of both TCO2 and Alkalinity. Actually, all the TCO2 and Alkalinity measurements using filtered samples were slightly lower than the corresponding measurements using unfiltered samples, but the difference (greatest on 23rd September) was only about 2%. Both sets of measurements are shown on the carbon budget graph -the lower ones are with filtered samples. So it is possible that there was some calcification but, if so, the amount of CaCO3 produced was small compared to the organic carbon uptake.

The chlorophyll measurements show that there was an increase between the 18th and 23rd September, but a subsequent drop in chlorophyll by the 28th September, although this is the time of maximum "biological carbon" uptake and highest transfer velocities. This suggests that, as for previous blooms, the algae had already started to settle out of the water column. In any case, by 30th September the pCO2 had begun to rise, the transfer velocity had fallen, and the peak of the bloom had clearly passed. In case the low concentrations of nitrate and phosphate were limiting growth, more of these nutrients was added to the tank on 30th September -to bring the nitrate concentration up to about 1/4 of that in f /2 medium and the phosphate concentration up to about 1/2 of that in f /2.

After 8th October the biological carbon, chlorophyll and CO2 transfer velocities rose again, but by this time there was clearly a film on the surface and later analysis of a sample from the tank showed that other rogue species of phytoplankton and even zooplankton had taken over. So this period after 5th October should be considered as a different, unintentional, bloom, which will be discussed in

8.4 Comparison of CO2 transfer velocities with and without algae.

8.4.1 Explanation of the composite plots

To compare the CO2 transfer velocities measured during all the algal blooms, both with each other and with the "control" measurements made without algae (which were discussed in detail in Section 7.5.1 ), all the data were combined together in one graph ( figure 8-11 ) showing log transfer velocity as a function of log pCO2. The log scales are used to evenly spread the datapoints for greater clarity. The region where pCO2 is higher than 60ppm is also shown separately (with transfer velocity on a linear scale) in figure 8-12 . This covers the full pCO2 range normally encountered in the real sea.

Each bloom is represented by a different colour, different species by different shaped symbols, and the influx measurements have closed symbols whereas the efflux ones have open symbols. The "maximum error" bars are shown as before: it can be seen that the efflux measurements at the low pCO2 end of the range have a very large error and should not be considered too seriously. The later periods of the blooms during which surface algal films developed, or after the addition of enzyme inhibitor, are not included in the composite plots to avoid confusion.

The predictions of enhanced transfer velocity calculated with the formula of Hoover and Berkshire (1969), and also the diffusion- only predictions (see Section 6.4 ), are also shown. The scatter in these is due the temperature and stirring motor speed variation, and indicates the variability expected due to these physical factors.

A black line also shows an arbitrary formula

Transfer Velocity = 2.7 + 12 / Ö ( pCO2 )

This formula happens to match the no-algae influx measurements better than any standard parameterisation for a statistical best fit, and is shown as a reference against which to compare the algal bloom measurements. The value 2.7 is taken from the acidified seawater blanks (i.e. for diffusion only), but there is no physical justification for the remainder of this formula. A higher dashed line fits approximately through the no-algae influx data but these points show a large scatter.

To make the picture less confusing, I have made another pair of composite graphs ( figure 8-14 and figure 8-15 ) using weighted average transfer velocities. Whenever two headspaces were used simultaneously, usually one for "influx" and one for "efflux", the formula for the weighted average transfer velocity was:

(flow / area)1 (pCO2 in - pCO2 out)1 - (flow / area)2 (pCO2 in - pCO2 out)2
dimensionless solubility * (pCO2 out1 - pCO2 out2 )

Where the subscripts indicate the two different headspaces, and the units are as in Section 6.6 ). Where only one headspace was used the transfer velocity was calculated in the normal way (consequently a few efflux transfer velocities are included in these plots, where there was no corresponding influx measurement, and these tend to lie above the other data).

It should be borne in mind that the "concentration" of the algae varied between the blooms for any given pCO2, so they are not strictly comparable. In particular, the pCO2 drop during the third Dunaliella "bloom" was caused more by gas exchange than by algal growth. To show the importance of the biological carbon whilst removing as much as possible the effect of the water pCO2, another graph ( figure 8-16 ) was made in which the ratio

measured transfer velocity (weighted average)
no-algae line (2.7 + 12 * pCO2 ^ [-0.5])

is plotted as a function of calculated biological carbon. Where this ratio is significantly greater than one, there is probably an enzyme effect.

8.4.2 Discussion of composite plots and identification of catalysis effect

At a first glance at either of the composite graphs, it can be seen that most of the datapoints for the Dunaliella blooms 1,2 and 4, and for the Emiliana Huxleyi bloom, lie significantly above the "no algae" control data. For Dunaliella, this difference is only observed below an apparent threshold of pCO2 = 250 ppm. On the other hand the data from the Amphidinium and Skeletonema blooms lie roughly along the same line as the "no algae" control data. So although for all species the transfer velocity still rises as pCO2 falls, this composite graphic suggests that there is an enzyme effect for Dunaliella and E. Huxleyi, and not for Amphidinium and Skeletonema. This also agrees with the results from adding enzyme inhibitor, which reduced the transfer velocity for Dunaliella but not for Amphidinium.

Note that the last four transfer velocities of the fourth Dunaliella bloom - the ones at water pCO2 lower than 9ppm in figure 8-11 and figure 8-14 seem anomalous -this is because by this point most of the algae had settled out onto the base of the tank so the concentration in the water column was much lower than before.

The difference between the catalysis observed during the Emiliana Huxleyi blooms, and the lack of catalysis during the Amphidinium and Skeletonema blooms, is clearer in figure 8-16 . Here the ratio of the measured weighted average transfer velocity, divided by the equivalent transfer velocity calculated from the line through the no-algae data, is shown as a function of biological carbon (as explained in the previous section). The data for Dunaliella and E. Huxleyi clearly lie above the line where the ratio is 1, whereas the data for Amphidinium and Skeletonema do not. It is also obvious that although the second and third Dunaliella blooms cover the same pCO2 range, they are not comparable because the concentration of algae was much lower during the third bloom. Note that the anomalous group of eight data points from the fourth Dunaliella bloom and from the Emiliana Huxleyi bloom, which are ringed on the figure, were measured after most of the algae had settled out. Therefore most of the calculated "biological" carbon was actually resting on the base of the tank not suspended in the water column.

The transfer velocities for Emiliana Huxleyi shown in figure 8-11 , figure 8-12 , are calculated as usual using water pCO2 from headspace "C", which was about 10-20ppm lower than headspace "B" during the period of the highest transfer velocities, as discussed earlier. If the lower pCO2 in "C" is due to a slight film (recalling that a visible film developed in "C" a week later), we should use the "B" values for water pCO2 instead (since there was not any such film in "B"). The two sets of transfer velocities are compared in figure 8-13 . When "C" values are used for water pCO2 the influx transfer velocities become higher, and the "efflux" transfer velocities become lower. The "influx" gas exchange was measured using headspace "D", in which a week later a visible film also developed, so this might explain the high transfer velocities here. However the "efflux" gas exchange was measured in headspace "A" which never developed a film during this bloom, and these values still lie well above the no-algae control measurements and even above the Dunaliella data at the same water pCO2. So even if there is a slight film effect in "C" and "D", there is probably also an enzyme catalysis effect. Note that the uncertainty about the equilibrium water pCO2 does not affect the "weighted average" transfer velocities used in figure 8-14 , figure 8-15 and figure 8-16 because they are calculated only from the CO2 flux data from "A" and "D", but these average figures would include any film effect in "D".

For Phaeodactylum most points lie significantly below the no-algae data, except at the lowest pCO2 where the transfer velocity suddenly increased to much higher values than the no-algae data. This might be explained by a slight inhibitory film effect even before the film was noticeable to the eye (note that after 1st March when the film first appeared, the data has been excluded from the composite plots), followed by a sudden onset of enzyme catalysis at very low pCO2- however more experiments would be needed to confirm this. The high transfer velocities lie beyond the range of the "no-algae" data so it is difficult to be certain that they are due to an enzyme effect.

For Phaeocystis the points do lie on average a little above the "no-algae" line, but the difference is smaller than the scatter between the points. Note that the three anomalous "efflux" measurements may be due to a leak.

8.4.3 Comparison with literature reports of CA activity in these species

There have been reports of carbonic anhydrase activity in all six of the species of algae, which were investigated during blooms in the tank, (for a summary, see table 2-1 ).

Dionose-Sese and Miyachi (1992) observed extracellular CA activity in several species of Dunaliella, and noted that in some cases this activity was stimulated by chloride ions. Amoroso (1996) and Ramazanov et al (1995a) showed that the activity of internal CA increased greatly in low-CO2 conditions. This is consistent with the catalysis by Dunaliella observed in the tank.

CA activity in the diatom Phaeodactylum appears to be more dependent on experimental conditions. Although Colman and Rotatore (1995) detected CA activity in Phaeodactylum, Rotatore et al (1995) report a lack of external CA in this species, and Dionosio-Sese and Miyachi (1992) found that chloride ions inhibited the CA in Phaeodactylum. Iglesias-Rodriguez and Merret (1996), Nimer et al (1997), and JohnMcKay and Colman (1997) all report evidence of extracellular CA activity in Phaeodactylum, but only in very low CO2 conditions. Iglesias-Rodriguez and Merret (1996) showed that CA activity was not detected until CO2 and HCO3- were both limiting, when TCO2 was lower than 1 mM, less than half its usual value in seawater. In the steady-state tank, although pCO2 fell as low as 5ppm, this still corresponds to a calculated TCO2 value of 1.2mM (see carbon budget, figure 8-4 (b)), due to the high buffering capacity of seawater. Later, a spontaneous calcification event in the water removed CaCO3 and lowered the alkalinity, and thereafter the TCO2 was much lower (about 0.25mM), but the exact time of the calcification was not known, and the interpretation of the transfer velocities was confused by a surface film. Generally, though, the lack of any catalysis in the early stages of the bloom is consistent with these varied literature reports.

The story is similar for the other diatom investigated here, Skeletonema. Nimer et al (1998), and also Nimer et al (1997), report that extracellular CA activity was only detected when [CO2] was less than 5m m, which corresponds to a water pCO2 of approximately 130ppm at 15C. Although pCO2 in the tank dropped below this level after 26th August, to a minimum of about 65ppm, by this time most of the Skeletonema had in the tank flocculated and settled, so the concentration suspended in the water was very low (as indicated by chlorophyll measurements). Alldredge et al (1995) report a similar mass aggregation of diatom blooms in a mescosom study, noting that flocculation began only 8 days after inoculation, and before the nutrient levels were depleted or the cell concentration became particularly high.

The tank experiments with Phaeocystis from the North Sea were inconclusive. Ulf Riebesell (personal communication) reported detection of CA activity in Phaeocystis in culture studies. However, Gideon Middleton (personal communication) found no evidence for CA activity in Phaeocystis, based on measurements of 13C:12C isotope ratios in a mesocosm study. Although there was a large drawdown of CO2, to 150ppm, he suggested this was due to direct uptake of HCO3-. A similar study on a diatom bloom showed that this, by contrast, was dependent on CO2 uptake alone.

Nimer et al (1997) report evidence of extracellular CA activity in Amphidinium Carterae, under all conditions investigated, of which the highest [CO2] level was 10m m (pCO2=260ppm, pH8.3). The same was true of various other dinoflagellate species which they investigated (see also Nimer et al 1999), and the activity increased only slightly at lower pCO2. Berman Frank et al (1994,1995) also detected CA activity in the dinoflagellate Peridinium Gatunense in Lake Kinneret, with much greater activity at low pCO2. In the light of these reports, it is perhaps surprising that no catalytic activity was detected during the relatively intense bloom of Amphidinium in the steady-state tank.

There have been many investigations of CA activity in Emiliana Huxleyi, but again, the conclusions vary depending on the experimental conditions, and the interaction between CO2 uptake by photosynthesis, and CO2 release by calcification (which can crudely be represented as Ca2+ + 2HCO3- =>CO2 +CaCO3 + H2O). Nimer et al (1994) and Nimer and Merret (1996) reported that calcification (coccolith production) could only be induced at fairly low nitrate concentrations. Calcification almost ceased as the bloom passed from the exponential to the stationary phase, and only at this stage was CA activity detected. This would suggest that CA production is switched on, as an alternative to CO2 supply from calcification. Sekino and Shiraiwa (1994) and Nielsen (1995) also suggest that calcification plays an important role in supplying CO2 for photosynthesis. However, other authors have suggested that CA may also catalyse the calcification process itself, as discussed earlier in Section 2.3.2 . In any case, significant catalysis was detected in the steady-state tank, despite the lack of any calcification (as determined by alkalinity measurements).

8.5 Surface Algal Films

8.5.1 General features of films

During the later stages of four of the blooms in the tank, algal films formed on the surface of the water in the tank. Later analysis showed that these films were composed of a mixture of species of algae, bacteria and zooplankton, unrelated to the original cultures which had been set up several weeks earlier. It is not surprising that a few rogue cells had been able to enter the tank and multiply during this long period. The conditions leading to the formation of these films (very high algal density and nutrients, low turbulence, and usually low pCO2) are perhaps unlikely to occur frequently at sea. Nevertheless this unexpected phenomenon is interesting in its own right and is therefore discussed further in this section.

The thickness and surface coverage of the films varied greatly between the headspaces at any one time, suggesting that the organisms attached to the surface could not easily transfer through the water, and under the dividing walls between the headspaces. Consequently, simultaneous measurements were made, of CO2 fluxes in and out of different headspaces with varying amounts of film coverage. Interpreting this data is complicated, but suggests that on some occasions the algae in the film were taking up CO2 directly from the air (evidence of this will be discussed in more detail below). In such circumstances, the transfer velocities calculated in the normal way could be very strange, -ranging from extremely high values, when the direct uptake complements the normal gas exchange process, to negative values when the direct CO2 uptake was in the opposite direction to the D pCO2 gradient. Clearly in this situation, when the flux is not related to D pCO2, the calculated transfer velocity can be meaningless, although the flux is real. The transfer velocities calculated during periods with surface films were therefore excluded in the composite plots discussed in the previous section. To explain this data, it is necessary to present it in a different way, reporting the measured changes in pCO2 in each headspace, and the calculated fluxes, as well as the calculated transfer velocities. Plots showing all this data are given for two of the blooms, and discussed in detail below. Fluxes are reported in mol m-2 yr-1, which are the units usually used in studies of global or regional air-sea carbon fluxes (see introduction Section 1.2.2 ). This helps to gives an indication of the potential impact of these films on the global air-sea CO2 flux, if they occurred naturally at sea.

Before considering the measurements during each bloom in more detail, the reader may find it helpful to consider figure 8-17 , which shows an example of pCO2 in the headspaces and water of the tank, and the fluxes between them, while there is a film on the surface of two of the headspaces. The pCO2 values are similar to those observed after the end of the fourth Dunaliella bloom. Note that this is only a schematic diagram: the tank is actually annular as shown earlier in figure 4-4 .

8.5.2 End of the first Dunaliella Bloom

During the second half of the first Dunaliella bloom, a film began to develop which was composed not only of a mat of green Dunaliella, but also of patches of brown filaments and later brown "blobs" which were later kindly identified by Keith Clarke of the Fisheries Laboratory in Lowestoft, as "Ectocarpus"(probably Terminalis), well known as a fouler of ships' hulls. The blobs are sporangia, supported by air bubbles and surrounded by a network of filaments. Since the water used for this first bloom experiment was not filtered and was collected from a MAFF ship, it is not surprising that a few rogue cells entered.

The film was first observed on 5th December after I had been away for four days, during which time most of the algae in the bulk of the water had also settled out onto the base of the tank. The two "equilibrium" headspaces seemed to contain different pCO2 levels (41.8 and 33.2 ppm), perhaps due to different amounts of film coverage taking CO2 directly from the air. The lower of these "equilibrium" pCO2 measurements could not be reconciled with the output from the efflux headspace (pCO2 34.6 ppm). So if the higher of these is used as the water pCO2, then the calculated influx and efflux transfer velocities are similar but unusually high (12 and 11 cm hr-1 as shown in figure 8-2 (a)).

This film was then disrupted in the efflux headspace by bubbling to create an oxygen disequilibrium for the oxygen and SF6 transfer velocity measurements. The SF6 measurement transfer velocity (measured using the other headspaces) was distinctly lower than the previous one earlier during this bloom (see table 7-2 and figure 8-2 (a)), suggesting that the presence of the film inhibited gas transfer by diffusion. After resuming normal steady-state CO2 flux measurements, transfer velocities were closer to those expected without algae at such low pCO2, and dipped as pCO2 rose, as noted earlier. The ectocarpus film was much more extensive in headspace used for influx measurements, than in the headspace used for efflux measurements. However, even after the gas flows were swapped (between days 46 and 47) such that headspace with the greater film was used for the efflux measurements, the efflux transfer velocities continued to be higher. This is consistent with the explanation, that the value taken for the equilibrium water pCO2 was too high.

8.5.3 End of Phaeodactylum bloom

A wrinkly film, similar in all headspaces, first appeared after 1st March and gradually thickened. It began to break up on the 11th and was mostly gone by the 18th (when enzyme inhibitor was added). During this period only "influx" transfer velocities are shown on the figure 8-4 (a) because the efflux transfer velocities would have been negative, since the pCO2 of the air flowing out of the efflux headspace was greater than that of the "equilibrium" headspaces. This was not of great concern since the differences were very small (typically: into efflux headspace 6 ppm, in equilibrium headspaces 8ppm, out of efflux headspace 10 ppm) so at the time it was assumed that this was due to a systematic experimental error. However, an alternative explanation might be that the film in the equilibrium headspaces was keeping the pCO2 in the air below that of the water, since, if the water pCO2 were higher, both the influx and the efflux transfer velocities would converge to a reasonable positive value. The spontaneous calcification event which probably occurred early during this period (as discussed in Section 8.3.3 above) might have raised the pCO2 in the water sufficiently to reconcile the transfer velocity calculations, whilst the presence of the film might have prevented such a rise being observed in the equilibrium headspaces.

8.5.4 End of fourth Dunaliella bloom

The film effects during the fourth Dunaliella bloom were particularly intriguing. No distinct film could be seen directly through the perspex lid, and when the tank was finally opened there was no sign of any rogue species, and only small patches of film remained. However by mid October the "murkiness" of the water below each headspace (noted by observing objects within the tank such as the steel cooling pipe) seemed to be different. This difference must have been due to a surface film since chemical and chlorophyll measurements of samples of the bulk water showed that it was still homogenous. Headspace B became the murkiest, followed by C, then D, and A was the clearest. Note also that the stirring paddle rotated the bulk water clockwise in this direction: B-C-D-A.

Figure 8-18 a summarises the pCO2 measurements in all four headspaces, and also the pCO2 of air entering the influx and efflux headspaces, during the period 7th - 19th October. The reader may be surprised by the variability of the pCO2 flowing into the headspaces. During the experiments themselves, the process of direct CO2 uptake by surface films was not understood, as it was only revealed later by careful analysis of the data from all four blooms. The strange transfer velocities were therefore attributed to an unknown experimental error, and in order to investigate the source of this "problem", the CO2 flow system was adjusted frequently. The "influx" headspace was initially A, but after the 14th October the high pCO2 gas flow was transferred to B. Note that the gas flows for both influx and efflux were about 140 ml min-1 and the transfer velocities between the 8th and 11th October (early stages of the film) were in the range 7-10 cm hr-1, at which rates the headspace CO2 should reach steady state within about 4 hours, which is much less than the gap between measurements.

Between the 11th and 14th October all four headspaces were left to equilibrate with the water, yet on the 14th the pCO2 in A (which looked the clearest) was about 70 ppm, in D it was about 50ppm, and in B and C (the murkiest) it was about 15ppm. Clearly the effect of this film was to lower the "equilibrium" CO2 in the headspace, whilst over the same period it had probably risen in the water. Therefore it seems that the algae might be taking CO2 directly from the air and transferring it to the water.

The continuous line on the graph shows my best guess of the true water pCO2 during this period. Up to the 11th October it rises slightly above the "equilibrium" headspaces B and C as their films develop, and from the 14th October onwards it is the same as measured in the relatively clear "equilibrium" headspace A. This "best guess" sequence of water pCO2 values was also used in the bloom plots presented earlier, figure 8-8 a and b).

The fluxes calculated from these pCO2 measurements are shown in figure 8-18 b (blue symbols), and the corresponding transfer velocities are shown by the numbers on the plot (these were also shown graphically in figure 8-8 a earlier). The fluxes which would be expected due to reaction and diffusion alone are also shown (pink symbols), estimated using the formula of the best-fit line through the no-algae transfer velocities in the composite plots (see Section 8.4.1 ). The difference between these two fluxes (green symbols) shows the effect of the algae, either due to the film or due to enzyme catalysis. There is probably some catalysis during the early period shown in this plot (8th to 11th October), which might explain the high fluxes and transfer velocities measured in headspace A (where D pCO2 was about 300ppm). However, catalysis would not be able to explain the very high fluxes in headspace B during the period 15th to 18th October, when D pCO2 was very low, and sometimes even in the opposite direction to the observed flux. During this period, these fluxes, which are in the range 5-7 mol m-2 yr-1, must be attributed to direct uptake of CO2 by the film.

Using the "best guess" values for water pCO2, the influx (A) and efflux (D) transfer velocities up to 11th October agree quite well and are within the "normal" range with algal catalysis (about 7 cm hr-1). However on 15th October the "influx" transfer velocity calculated using headspace B became negative, because there was a large flux of CO2 entering the water, although the pCO2 in the headspace was lower than it is in the water. During the following three days the film gradually cleared, and as it did so the influx "transfer velocity" gradually returned to more normal values. Note that as the headspace pCO2 rises past the water pCO2, the calculated transfer velocity does not pass through zero but crosses an asymptote from negative to positive infinity.

An alternative hypothesis might be that algae within the film were simply lowering pCO2 locally within the microlayer, and therefore increasing the air-water CO2 flux, rather than extracting CO2 directly from the air as suggested above. However, for this hypothesis to explain the magnitude of the observed fluxes, pCO2 in the microlayer would have had to be much lower than pCO2 in the bulk water. Since the latter was low already, and pCO2 cannot fall below zero, a reduction in the microlayer pCO2 would not in itself be sufficient to account for the observed fluxes.

Note that if the microlayer pCO2 was indeed very low in headspace B ("influx") on 15th October, then the calculated transfer velocity would be positive but still very high (about 40 cm hr-1). Meanwhile, if this low microlayer pCO2 also applied to headspace D ("efflux"), this transfer velocity would become extremely high or even negative. So the strange transfer velocities cannot be explained by uncertainty in the D pCO2.

8.5.5 End of Emiliana Huxleyi bloom

At the end of the Emiliana Huxleyi bloom a similar phenomenon was observed. Figure 8-19 a and b show the measurements of pCO2 in all of the headspaces, together with calculated CO2 fluxes and transfer velocities, for the whole of this bloom. The symbols are the same as for figure 8-18 a and b above, and in this case headspace "D" was used for "influx" measurements, and "A" for efflux, throughout the bloom. Note that the anomalous pCO2 flow into D on day 40, was due to the CO2 cylinder running out of gas overnight. Up to the 5th of October (day 35), the pCO2 measurements from headspace "C" have been used for the water pCO2 used to calculate the fluxes and transfer velocities shown in figure 8-19 b. After 5th October there was a film in "C", but no significant film in "B", so the values from "B" were used for the water pCO2.

To see both sets of transfer velocities calculated using either "B" or "C" values for water pCO2 consistently, refer instead to figure 8-10 a shown earlier. The transfer velocities calculated using "C" pCO2 are probably meaningless after 5th October as a real CO2 flux has been divided by a false D pCO2. They were included in figure 8-10 a (the blue circles) just for completeness.

The small (10-20ppm) difference between the pCO2 measured in these two equilibrium headspaces during the early period of this bloom, between the 26th and 30th of September, and the impact of this pCO2 difference on the calculated transfer velocity, has already been discussed in Section 8.4.2 , and was illustrated by the difference between the blue and the green lines and symbols in figure 8-10 a and b. This analysis suggested that the high transfer velocities during this period, must be attributed to catalysis by carbonic anhydrase produced by the Emiliana Huxleyi, even if there is also a slight film effect in "B".

This slight pCO2 difference was reduced to zero on 2nd October as the pCO2 rose again after the peak of the first stage of the Emiliana bloom. No film was observable by eye, until 5th October 1998, and then only in headspaces C and D. After this, the pCO2 difference between "B" and "C" then increased dramatically. Between 9th and 12th October (days 39 - 42 on the figure) the pCO2 in headspace "B" dropped from 298ppm to 118ppm but the corresponding values in "C" were 100ppm lower - 192ppm and 20ppm respectively!

However there was still no film observable by eye in either headspace "A" or "B". On 9th October black specks appeared on the surface of the water -about one per 5 cm2 in headspace "B", but over ten such specks per cm2 in "C" and about half that number in "D". In "C" and "D" these specks did not move freely on the surface but in concert together as if joined by invisible elastic. On 13th October a sample of the film was taken with a syringe, and analysed under the microscope by Mike Steinke. In this sample he observed many cyanobacteria -probably Synechococcus, and even several zooplankton grazing on the film, but little sign of the original Emiliana Huxleyi! Clearly rogue species had entered and taken over, which is not so surprising since 40 days had now elapsed since the original filtered water had been put into the tank. The coexistence of these diverse organisms within the microlayer is also consistent with reports of observations in samples collected from the surface microlayer at sea (see discussion in Section 2.4.3 and references therein).

After 12th October the pCO2 difference between headspaces "B" and "C" began to get smaller, as the pCO2 in "B" continued to fall while in "C" it reached a minimum of 12ppm and then started to rise again. By 19th October the pCO2 was higher in "C" than in "B", where the pCO2 dropped to a minimum of only 8ppm before it also began to rise. This reversal in the difference may be due to a slight film developing late in "B", or alternatively due to greater respiration than photosynthesis by the various organisms which made up the film in "C". By 19th the visible film had begun to break up and the black specks in C and D were swirling freely. By 1st December (day 62 on the graphs) the pCO2 difference between headspaces B and C was only 6ppm. Alkalinity measurements made on this day showed no significant change since the start of the bloom, and TCO2 measurements showed that that the air pCO2 measured in B and C was now in equilibrium again with the water samples drawn from beneath the surface.

During the early period before the end of September, the high fluxes and transfer velocities (both influx and efflux) can be attributed mainly to catalysis by carbonic anhydrase produced by Emiliana Huxleyi, as discussed earlier, although the slight discrepancy between the two headspaces suggests a slight film effect. During the later period, after 5th October, the interpretation is more complicated and requires close examination of figure 8-19 b.

Since there was no film visible in either "B" or "A" (which was used for the efflux measurements), the fluxes and transfer velocities measured in "A" (calculated using "B" values for water pCO2 as noted above) should be expected to be represent the "normal" air-water gas exchange process. As the water pCO2 falls to very low levels, D pCO2 also falls, and the flux in A drops to virtually zero by 17th October, while the transfer velocity in "A" stays fairly constant around 4 cm hr-1. The difference between the observed fluxes in "A" and those expected without algae gradually falls, indicating a declining effect of catalysis.

In headspace "D", the measured fluxes remain high and the calculated transfer velocities increase. Meanwhile the flux in "D" expected without algae drops after 5th October due to decreasing D pCO2, so the flux attributed to algae rises. By itself, this might possibly be attributed to an increasing effect of catalysis, although this would not be consistent with declining catalysis in "A". The highest transfer velocity in "D", 13.8 cm hr-1 measured on 12th October, is 3.3 times higher than the corresponding measurement in "A".

However, the clearest indication that there must be direct uptake by a surface film, is shown by the flux due to algae calculated for the closed equilibrium headspace "C". Here, there must have been a small air-water flux to account for the fall of pCO2 in the headspace, but this would be much less than observed, as the total amount of CO2 stored in any of the headspaces is small, compared to the fluxes discussed here integrated over time. If there were no direct uptake by the film, the pCO2 would be the same as measured in the other equilibrium headspace "B", but between 9th-13nd October, the pCO2 actually measured in "C" was up to 100ppm lower than in "B". This implies a large D pCO2 gradient, which would cause a normal diffusion-reaction flux from the water to the air, as illustrated by the pink symbols in figure 8-19 b. Since headspace "C" was closed, this must have been balanced by a direct uptake of CO2 by the film (green symbols). This flux, up to 1.5 mol m-2 yr-1, is of a similar magnitude to that attributed to algae in "D", at the same time. The simplest explanation, is that these are both caused by direct uptake of CO2 by the surface film. No amount of catalysis could explain such a steady-state pCO2 disequilibium, particularly on 12nd-13rd October when the change in pCO2 in "C" was minimal, but still much lower than in "B".

Note that on 10th October both headspaces "A" and "D" were used for efflux measurements -because the CO2 cylinder supplying the high pCO2 air to "C" had emptied. The transfer velocity calculated from "D" is lower (2 cm hr-1) -which is to be expected if an algal film was taking up CO2 from the air, in the reverse direction to the water-to-air flux due to normal gas exchange.

8.5.6 Conclusions regarding the effect of surface films

These observations while the films were present do not tell us anything about catalysis of chemical enhancement by carbonic anhydrase, but suggest that the algae are taking CO2 directly from the air. Although the conditions in the tank were extreme it is possible that in lakes or coastal waters such surface films of algae may develop (see also introduction Section 1.4.5 and Section 2.4 ). Rather than inhibiting gas exchange as is usually suggested, they may instead enhance it or even pump the CO2 in the opposite direction to that expected from the DpCO2 gradient. It was only possible to observe this intriguing phenomenon for the first time during these experiments, due to the unique design of the steady-state gas exchange tank with multiple headspaces. This deserves further investigation and ideas for future experiments will be discussed in section 10.5.1

It is worth remembering that during the period 9th-12th October of the Emiliana bloom, while the pCO2 in "B" was still within the normal range at sea (200-500ppm), active uptake by the film in headspace "C" reduced the effective "equilibrium" pCO2 at the top of the sea-surface microlayer by over 100ppm. If this phenomenon were also occurring during algal blooms at sea, it could make a large difference to the air-sea CO2 flux, since the D pCO2 difference between air and sea is normally much less than 100ppm. This effect would not be detected at all by standard shipboard pCO2 measurements of sub-surface water samples (which are typically collected at a depth of about 1m). Such direct uptake of CO2 by a surface film might even be large enough to explain some of the very high transfer velocities calculated from CO2 eddy correlation measurements (see introduction Section 1.2.10 ). Even the more modest differences of 10-20ppm observed during the early stages of the Emiliana bloom could make a very big impact on the net global air-sea CO2 flux, since the global average D pCO2 is only about 8ppm (see Section 1.3.3 ).

Using this value for D pCO2, and a global average transfer velocity of 12 cm hr-1 from the parameterisation of Liss and Merlivat (1986), the net global average air-sea CO2 flux is only about 3.5 mol m-2 yr-1, although locally the flux is sometimes much higher. By comparison, the fluxes attributed to films in the detailed analyses above, were between 5 and 7 mol m-2 yr-1, in headspace "B" after the end the 4th Dunaliella bloom, and between 1 and 3 mol m-2 yr-1, for headspaces "C" and "D" after the end of the Emiliana bloom. This indicates, that if such films occurred frequently over wide areas of the ocean, they could potentially take up as much CO2 from the atmosphere, as the normal air-sea gas exchange process.

The impact of such films on the global average 14CO2 flux would, however, be much smaller than on the 12C flux, because the one-way 14C flux is much less dependent on water pCO2 (as discussed in Section 3.3.4 ). Consequently, the potential scale of 12CO2 uptake by films is not constrained by the global 14C budget.

Assuming that some of the CO2 taken up the algae in the surface film later leaves the microlayer, either as dissolved or particulate organic carbon in the seawater, or assimilated into other organisms at a higher trophic level, then the film effectively provides a direct pathway for a flux of CO2 between the atmosphere and the pool of organic carbon in the ocean. The extra vertical flux of particulate carbon due to film uptake might be constrained by measurements using sediment traps, but these are not yet sufficiently accurate to predict the potential impact on atmospheric CO2. Vertical fluxes of dissolved organic carbon are also poorly quantified, although they are known to be large. So it is possible that a significant flux atmosphere-ocean flux of carbon, via algal films on the sea-surface, has been missed out entirely from the global carbon budget, due to insufficient consideration of biological processes within the sea-surface microlayer.

8.6 Conclusions regarding algal catalysis

The original aim of these experiments was to investigate whether it was possible for the algae to release enough external carbonic anhydrase into the surface microlayer to significantly increase the CO2 transfer velocity, and if so, to investigate whether the effect was greater at lower pCO2 as predicted by algal physiology. From a comparison of the transfer velocities with and without algae, as shown by the composite plots ( figure 8-11 , figure 8-12 , figure 8-13 , figure 8-14 , figure 8-15 , figure 8-16 ), it seems that for Dunaliella at least, and probably also for Emiliana Huxleyi, the answer is that there is a detectable effect. It is also interesting that the different species of algae produced quite different responses: the Amphidinium and Skeletonema clearly did not show any effect of enzyme catalysis, whilst for Phaeodactylum and Phaeocystis the result is ambiguous.

However it should be borne in mind that the conditions in these blooms of algal cultures were very extreme, with the pCO2 dropping as low as 5ppm, far below the levels encountered in the real sea (200-500ppm). Looking at figure 8-12 , only the high transfer velocities during the Emiliana Huxleyi bloom stand out well above the scatter of the "control" measurements within the normal range of pCO2 in the sea, and it is possible that these should be slightly lower, if they are calculated using headspace "B" rather than "C" for the water pCO2 (see figure 8-13 ). On the other hand the apparent lack of any enzyme catalysis on the CO2 transfer velocity for the Skeletonema bloom may be due to the algae settling out of the water column rather than not producing carbonic anhydrase, and measurements from just one bloom of each species of algae are not sufficient to rule out enzyme catalysis of CO2 gas exchange by this species under different experimental conditions.

For Dunaliella there seems to be a threshold for the onset of enzyme catalysis below about 250ppm, but this apparent threshold might possibly be an artefact of the sequence of the measurements, since the concentration of algae ("biological carbon" and chlorophyll) also rose while the pCO2 fell. However, generally the enzyme catalysis effect was clearly greater at lower water pCO2, confirming the prediction from the physiological calculations presented in chapter 3.

If any parameterisation of the enzyme effect were to be derived for use in hypothetical global CO2 flux calculations, it would be important to separate the two parameters: pCO2 and concentration of algae. The latter may be estimated crudely from satellite chlorophyll data. However during the blooms reported here, these two factors are too closely intercorrelated to be easily separated. The pCO2 in the tank fell mainly due to uptake by the algae, so where there seems to be an increase in the "enzyme" effect it is not possible to say whether this is due to the low pCO2 or to the higher concentration of algae. Of course, during any one algal bloom at sea the same intercorrelation will occur, but to calculate the net global air-sea flux of CO2, different areas of the ocean must be compared, in which case different physical factors also influence the pCO2 independently of the algae (see introduction Section 1.3 ). Further investigation would be necessary, therefore, before even a speculative parameterisation could be derived directly from experimental results. Measurements with algae should also be done over a range of different temperatures.

So to summarise, these results tell us that it is possible to detect catalysis of air-sea CO2 exchange by marine algae, but this effect varies between species and is only large in extreme conditions -low pCO2 and high chlorophyll.

Continue to Chapter 9:
The Effect of Catalysed and Uncatalysed Enhancement on the Net Global Air-Sea CO2 Flux.

If you jumped into this page from elsewhere, you may find it more convenient to go to the "frames" version of this thesis, alternatively you can use the links below to jump straight to a particular chapter:

Title Page Abstract Contents Figures Overview Chapter1 Chapter2 Chapter3 Chapter4 Chapter5 Chapter6 Chapter7 Chapter8 Chapter9 Chapter10 Appendix References Acknowledgements Links Download

Hits since 21st April 2000