the entrained water throughout the lake. L Formerly at Center for Water Research, University of Western Australia, Nedlands, WA 6009.

Transcription

1 Limnol. Oceanogr., 40(2), 1995, , by the American Society of Limnology and Oceanography, Inc. The physical response of temperate lakes to artificial destratification S. G. Schladow Department of Civil and Environmental Engineering, University of California, Davis I. H. Fisher AWT Science and Environment, Sydney Water Board, West Ryde, NSW 2114 Australia Abstract Prolonged periods of thermal stratification in a lake suppress convective motions capable of vertical transport of dissolved oxygen and other dissolved and suspended matter. When this leads to anoxia in the hypolimnion, it can give rise to profound water quality implications. Artificial destratification is one means whereby low oxygen levels in the hypolimnion may be averted. The operating principles of bubble plume destratification and its effect on the physical transport and mixing that occurs in a lake are described. Monitoring of temperature and dissolved oxygen was conducted at Lake Nepean, Australia, prior to the destratification, and then continued for two summers while the destratification system was operating. The results showed that artificial destratification had an immediate effect on lake motions across a range of scales. The actual process of destratification occurred by the erosion of the base of the thermocline. Deep-water temperatures were elevated well above historical values, the heat storage in the lake increased, and the time of turnover was brought forward 2.5 months. An integral plume model coupled with a one-dimensional lake model is shown to predict the effect of the bubble plume entrainment and the subsequent redistribution of the entrained water throughout the lake. The natural stratification that exists in most temperate lakes for a large part of the year effectively cuts off the hypolimnetic water from contact with the atmosphere and, hence, from a major source of oxygen. When this is coupled with generally increasing levels of nutrient input and the resulting increase in primary productivity (OECD 1982) there is a growing tendency for hypolimnetic water to become anoxic for significant periods of time. Aside from the direct effects of anoxia on aquatic life, anoxic conditions at the sediment-water interface often give rise to the release of nutrients and metal compounds from the sediments (Mortimcr 1941, 1942). This internal loading can in turn exacerbate eutrophication, lead to the formation of tastes and odors, and even render the water unsafe for consumption. Artificial destratification is one means whereby low oxygen levels in the hypolimnion can be averted. In its more common mode of operation, it is used during spring and L Formerly at Center for Water Research, University of Western Australia, Nedlands, WA Acknowledgments Funding for this research was provided by the Sydney Water Board, the New South Wales Department of Water Resources, and the Land and Water Resources Research and Development Corporation under the Centers of Concentration in the Water Research Program. George Kastl and Rod Frenda were responsible for the collection of field data. Processing and verifying the data, together with running the models, were due to the efforts of Kerry Greenwood. Discussions with J&g Imberger, John Patterson, Dale Robertson, Kresmir Zic, and Bob Banens were of value in clarifying the issues. 359 summer to prevent significant stratification from occurring or, where a strong seasonal thermocline has been established, to break it down long before the lake s natural turnover time. When used at the time of fall turnover, it may assist in ensuring that reoxygenation of the hypolimnion is complete. The methods used for destratifying lakes have varied and include mechanical stirrers, water pumps, and bubble plume devices. Burns and Powling (unpubl.) give many instances of the use of all these techniques. This paper deals only with the latter technique, although many of the features discussed are common to all forms of artificial destratification. Reviews of the literature concerning air bubble plumes in water have been presented previously (e.g. see Schladow 1992a; Zic et al. 1992). Bubble plume systems for destratifying lakes have been used for > 70 yr as a means of in situ water quality improvement (Scott and Foley 192 1). The results have supported the notion that artificial destratification is on balance beneficial in terms of improving a range of water quality parameters, though examples describing deterioration in water quality exist (Fast unpubl.; Steinberg and Tille-Backhaus 1990). Despite the obvious potential of this technique, there has been little direct monitoring of the mixing processes and of the changes in lake physics that arise as a consequence of artificial destratification. As these drive the changes in lake water quality, this absence of data has limited the understanding of how and when artificial destratification should be used. Our paper seeks to contribute to the understanding of the effects of a bubble plume destratification system on the physical mixing in a lake. A temporal and spatial description of the progress of thermal destratification is

2 360 Schladow and Fisher Fig. 1. Schematic of the process of entrainment and dctrainment with a bubble plume in a stably stratified water column. presented for Lake Nepean, Australia. The interpretation of field observations is aided by a bubble plume simulation model, which had previously been used as the basis of the design of the system. The bubble plume model was also included as part of an overall model of lake mixing, so that the interaction between entrainment by the bubble plumes, discharge of the entrained water, and natural lake processes could be studied. Bubble plume model The upward passage of a stream of air bubbles released at depth in a lake is accompanied bv the entrainment and mixing of ambient water to form a combined bubble and water plume (hereafter referred to as the bubble plume). In the presence of a density stratification there is potentially a terminal height to which the aqueous portion of such a plume will rise before falling back to its level of neutral buoyancy (Morton et al. 1956). The localized mixing that accompanies the entrainment results in formation of horizontal density gradients, which then sustain nearhorizontal intrusions away from the vicinity of the plumes (internal detrainments). The bubbles continue to rise to the water surface and not be shed with the entrained water and in fact cause the entrainmen,t of a further plume immediately above the level of the initial one. This process may be repeated several times before the bubbles reach the surface, culminating in a final detrainment that produces an intrusion a slight disl.ance below the water surface. Such a process has been demonstrated experimentally by McDougall ( 1978) and Asaeda and Imberger (1993) and is shown schematically in Fig. 1. The model for describing the rise, detrainment, and reformation of a bubble plume has been detailed previously (Schladow 19923). It is based on integration of the equations of mass, momentum, and buoyancy conser- vation for a single-phase buoyant plume, but where the buoyancy effects of stratification, bubble slip, and bubble expansion are included and the effects of bubble dissolution are neglected (McDougall 1978). The dynamics of bubble plumes can be characterized by two parameters: M (-QO/W) and C (-WH /Q,). Here Q. is the volumetric gas flow rate from the bubble source (at atmospheric pressure), His the total pressure head experienced by the bubbles (including atmospheric pressure head), N is the buoyancy frequency defined as N = [ -(g/p,)dpo/ dz] 2, g is gravitational acceleration, pr is a reference density, and dp,/dz is the equivalent (in terms of potential energy) linear density gradient. For a fixed water depth, C represents the effect of the stratification compared to the source strength. A small value of C implies weak stratification compared to the source strength and favors the plume reaching the surface without any internal detrainments. Mrepresents the source strength compared to the total pressure head. For the same value of C, a large value of M suggests a larger plume with a greater entrainment capacity. If we know the values of M and C, the properties of a bubble plume can be uniquely described. The mechanical efficiency of destratifying a water column can be defined as?imech = (PEJ - PEi)/ wise (1) where the isothermal work of compression, JVi,,, is defined as Wso = PlQo At W2h0 (2) Here, PE is the potential energy of the water column, subscripts i andfrefer to initial and final, At is the time over which the bubble plume operates, and p1 and p2 are the absolute pressures at the water surface and the compressor outlet. The relationship between the plume characteristics and mechanical efficiency has been demonstrated by Schladow (1992b) to take the form shown in Fig. 2. The curve for each value of M corresponds to the same water depth but different airflow rates. The peaks correspond to conditions where the uppermost plume in a cascade has just enough momentum to reach the surface. Thus, the peaks furthest to the left (low C values) correspond to a condition that would produce a single plume with a surface detrainment. The second peak in each curve corresponds to a condition where there is a cascade of two plumes with two detrainment levels, and so on. For a fixed airflow rate, A4 remains constant and C decreases during destratification. Thus, the process of destratification may be idealized as the progression along an M curve from right to left. In common with all integral models, the model described here greatly simplifies the true complexity of a bubble plume rising in a density-stratified fluid (e.g. Lemckert and Imberger 1993b). However, from the point of view of predicting the changes in density that result, the key is prediction of the entrainment flux; in this respect the model performs well over an extremely large range of values. Estimates of entrainment fluxes, based

3 Response to des ratijication Low Airflow, M = (Long diffuser) - High Airflow, M = 0.2 (Short diffuser) Kobus 1968 an Houten Fig. 2. Theoretical efficiency curves for two plume airflow rates as a function of stratification parameter, C. The peaks occur when an integer number of plumes fill the distance between the bubble source and the water surface. The symbols relate to thermal stratification at Lake Nepcan. C values with a strong thermal stratification (0.22 C m- I)-0; C values with a weak thermal stratification (0.03OC m-*)-o. on measured velocity profiles and plume widths, are compared to model-derived values in Fig. 3. Data compiled by Milgram (1983) for the experiments of Kobus (1968), Fannelop and Sjoen (unpubl.), Milgram and Van Houten (1982), and his own have been used for the comparison. The data cover a range of water depths from 3.66 to 50 m and a range of airflow rates from to 118 x 10m3 m3 s-l. At higher airflow rates (above 283 x 1 Om3 m3 s-l from a single source), the model underestimated the entrainment flux. However, such high airflow rates are one to two orders of magnitude larger than necessary or are normally used for destratification systems (Schladow 1992a). Lake model with artificial destratification The one-dimensional lake and reservoir model DY- RESM (Imberger and Patterson 198 1) is a process-based model that has been used to successfully simulate the vertical density structure of a large number of lakes under differing conditions. Since the model is based on parameterizations of the individual mixing processes, it is not necessary to perform site-specific model calibration. The processes included in the model are surface layer mixing, river inflows and outflows, and hypolimnetic transport. More recently, the model has been extended to include mixing produced by a bubble plume destratification system (Patterson and Imberger 1989; Schladow 1992a). The processes which are most important for the present discussion are those controlling the surface layer and those controlling the effects of the bubble plume. Surface layer dynamics are based on an integral turbulent kinetic energy model (Sherman et al. 1978). The turbulent kinetic energy budget is partitioned into four discrete processes-wind stirring, convective overturn, interfacial shear production, and Kelvin-Helmholtz billowing. The energy avail- C 1o-3 1o-2 10-l loo 10 lo2 Modeled Volume Flux (m3 8) Fig. 3. Comparison of experimental and modeled entrainment volumes. able through each of these processes is a function of the stratification and strength of the forcing. The total available energy is compared with the potential energy required to deepen the mixed layer by one computational layer. If sufficient energy is available, the mixed layer is deepened and the available energy decremented by this potential energy gain. The process is repeated until insufficient energy remains within the time step to continue the deepening process. The bubble plume algorithm used in the model is described by Schladow (19923). Each air outlet is treated as a point-source and the point-source equations for mass, momentum, and buoyancy conservation are integrated from the bottom up. As a consequence, there is entrainment into the rising plume for each computational layer. If at some height the calculated plume width exceeds the spacing of the sources, then from that level the pointsource equations are replaced by the analogous line-source equations and the integration continued. The details of this transition from a series of independent point-sources to a line-source are described by Robertson et al. (199 1). The stratification through which the plume rises is not confined to being linear, but rather it is the simulated density profile from the previous timestep. The insertion of the detrained fluid into the lake is treated as a linesource and is based on the description by Imberger et al. (1976). The insertion thickness is governed by a Grashof number and Froude number criterion, and a cosine velocity distribution applied across this thickness. Thus, the detrained fluid only affects some of the computational layers. By subtracting the entrainment profile produced by the rising plume from the insertion profile, an estimate of the profile of net horizontal flux is obtained. This profile only applies sufficiently far from the plume so that the flow field is primarily horizontal (i.e. beyond the region where the discharged fluid is falling to its level of neutral buoyancy). Given the somewhat arbitrary choice of the vertical velocity distribution within the intrusion, the calculated

4 362 Schladow and Fisher 0 I 500m, lkm I Scale I Long diffuser To Water quali?, and CTDO profihng sites Fig. 4. Monitoring stations at Lake Nepean. Inset shows the arrangement of the aeration system in relation to the adjacent monitoring stations. net profiles should be treated as only indicative of the types of flows produced. Thus, thd: number of intrusions, the centerlines of intrusions, and lhe heights of rise of the plumes are predictable with gre;ater certainty than the exact shape of the intrusion. Lake Nepean destratification system Lake Nepean is a water stora,ge reservoir km southwest of Sydney, Australia. It was completed in 1935, has a maximum depth of 65 m, a maximum volume of 70 x lo6 m3, and a surface area of 360 ha. It has a catchment area of 3 18 km*, with the main inflowing rivers being the Nepean, Burke, and Little Burke Rivers (Fig. 4). The bubble plume destratification system at Lake Nepean was designed following the method proposed by Schladow (1992a). The calculations suggested that a total airflow rate of 0.3 m3 s - l (at atmospheric pressure) through 10 sources (Q, = 0.03 m3 s-l) was required to destratify the strongest summer stratification in a 3-week period. Lake simulations supported this but also revealed that if the system was operated intermittently (so as to prevent strong stratification from being established), a configuration with 400 sources was equally economical. As a consequence, a composite system consisting of two separate diffusers connected to two 0.15 m3 s-l, 800 kpa compressors was installed at Lake Nepean at the end of A valve enables the total airflow to be directed to either of the diffusers. Both diffusers are made of poly-

5 Response to destrat$cation 363 ethylene tubing.with an internal diameter of 89.4 mm. The long diffuser is 400 m long and has 1.4-mm-diameter holes spaced every 1 m along its length. The short diffuser is 120 m long and has 10 clusters of 40 holes (1.4-mm diam) at 12-m intervals along its length. Figure 4 shows the location of the aeration system in relation to the adjacent monitoring stations. In terms of the nondimensional parameters for the design water depth of 55 m, the short diffuser (high airflow rate per plume) corresponds to it4 = 0.2 and the long diffuser (low airflow rate per plume) corresponds to M = (Fig. 2). The open circles shown are at M and C values corresponding to a strong stratification at Lake Nepean (an idealized 0.22 C m-i over the entire water column) as would be expected in midsummer, while the closed circles correspond to very weak stratification (0.03OC m-l over the entire water column). As C is dependent on both airflow and stratification, each diffuser has a different value of C for the same stratification. If destratification had commenced in midsummer using the short diffuser, then destratification would have proceeded along the A4 = 0.2 curve from point 1 toward point 2, with efficiency peaking at 8%. With the long diffuser, destratification would have followed the path from 1 to 2 on the A4 = curve, with the peak efficiency being only 4%. Where the bubble plumes are used to maintain a basically destratified state, the system would then begin operating between 1 and 2 (or 1 and 2 ). In general, higher efficiencies over the operating range were expected from the short diffuser. Measured data Regular monitoring of the lake includes measurement of temperature and dissolved oxygen at six fixed stations (DNEl-DNE6) along its length (Fig. 4). Measurements are taken manually at 2-4-week intervals. Typical vertical spacing of these data is 3 m. Although such monitoring is suitable for showing overall changes in the water column, it sheds little light on the dynamics of mixing. As part of the present study, two intermediate stations were added on the longer stretches (Sta. 4A and 5A), and profiling of the water column at all the stations using a Seabird SBE19 CTDO was initiated. This instrument provides electrical conductivity, temperature, and oxygen data at lo-20-cm intervals and therefore allows for more detailed description of mixing. A raft at station DNE3 houses a weather station and logger, with meteorological parameters from 2 m above the water surface logged at 15-min intervals. It also holds an Aanderaa Instruments thermistor chain, with 15 thermistors at depths from the water surface of 0, 2, 4, 5.5, 7, 8, 9, 11, 14, 17, 19, 24, 29, 34, and 44 m. The thermistors are accurate to O.O3 C, and the averages of 30-s readings for each thermistor are logged at 15-min intervals. Only temperature data are discussed in detail here. The gauge height and inflow temperature of the Nepean and Burke Rivers are logged at 15-min intervals. Daily measurements of water level, spillway flow, outflow to the Nepean River, and pumping station flow to the local arca are recorded manually. Since March the temperature, absolute pressure, and differential pressure of the compressors used for the destratification system have been recorded at 15-min intervals, enabling the precise airflow rate to bc calculated. Previously, airflow rate was estimated from the logged hours of daily operation. Energetics It is instructive to consider the magnitude of the energy fluxes associated with the processes of stratification and destratification of a water body. Taking December 1990 at Lake Nepean as representative of summer conditions, the net solar radiation flux was 276 W m-*, the net longwave flux was -85 W m-*, the latent heat flux was -63 W m-*, and the sensible heat flux was - 19 W m-*. Therefore the net energy flux into the lake was 109 W m-*. This net energy input, Q, is transferred to internal energy and volumetric expansion (work), such that Q=pAv+ Au. (3) p is the pressure (assumed constant), Av the change in specific mass, and Au the change in specific internal energy. From the definition of the specific heat at constant pressure, Cp, we can modify this equation such that Q - p(av/at)at + C,AT (4) where AT is the produced change in temperature. Thus, the fraction of energy input that goes to temperature increase and to volumetric expansion is in the ratio of Cp to p(av/at). Evaluating Av/AT at 20 C at a pressure of lo5 Pa yields a value of 2.03 x 1O-7 m3 kg-l K-* (Handbook of chemistry and physics 1980). Thus the ratio of energy partitioning is -2 x lo5 : 1. As it is only the expansion energy that needs to be overcome to achieve destratification, the true energy flux associated with strat- ification is 5.4 X 10e4 W m-*. The major destratifying agent (under natural forcing) is the wind. Following Philips (1976), if we assume that the surface-water velocity scale, w,, and the air velocity scale, u,, are both related to the surface shear stress, T, such that r = w,*p, and r = u,*p,, then the power input from the wind to the water can be expressed as P = rw,. Here 7 is defined by the bulk aerodynamic relationship r = C,p,u2, where C, is a drag coefficient and U is the mean stream air velocity. Combining these equations yields the expression for the wind energy flux, p = ~~312 l/2 Pa Pa u3. 0 PW (5) Assuming values for C, of 1 x 10e3, pa of kg m-3, p,+, of 998 kg rne3, and U of 3 m s-* (the mean velocity for Lake Nepean) yields an energy flux of 3.7 x 1O-5 W m-*. Th 1s value is an order of magnitude smaller than the stratifying energy flux. Extreme storm events with velocities higher than the mean would obviously do

6 364 Schladow and Fisher L 0 I I DAYS SINCE (13 September 1991) Fig. 5. Isotherms fbr 13 September 199 l-30 April A. Based on measured thermistor chain data. Solid bars indicate the periods during which the aeration system was operating. B. Based on numerical model results with artificial destratification. significantly more mixing (due to the velocity appearing to the power 3 in Eq. 5); however, the effect cannot be sustained. Little wonder that lakes stratify over summer. The energy input by artificial destratification systems can be estimated by Eq. 2. For the c:ase of Lake Nepean, free airflow rate is 0.3 m3 s-l (with p, = lo5 Pa and p2/ Pl = 8) and the lake surface area IS 360 ha, giving an energy flux of 1.7 x 10Y2 W m-2. With a mechanical efficiency of 6% (see Fig. 2) the effective energy flux is 1 x 10e3 W mf2, a factor of 2 greater than the stratifying flux. Despite the artificial destratification energy flux being an order of magnitude larger than the wind energy flux, significant interaction between the artificial and natural mixing processes is possible. Running the model for a 230-d period during which the destratification system was used intermittently at Lake Nepean allows the interaction to be quantified. Two model runs were conducted. In the first, the measured daily airflow rate for the destratification system was one of the inputs. In the second, no destratification was used. Model inputs common to both runs were daily averaged windspeed, vapor pressure and air temperature, daily total shortwave and longwave radiation, daily total rainfall, and daily inflow and outflow volumes. The isotherms based on the measured thermistor chain data for the simulation period are shown in Fig. 5A. The solid bars indicate when the destratification system was operating. Figure 5B shows isotherms produced by the first model run, using all the measured inputs, including destratification. The model result closely parallels the field observations, particularly for the surface layer depth. The average, daily total turbulent kinetic energy (TKE) production by wind stirring, convective overturn, and

7 Response to destrat&ation 365 shear for both model runs is shown in Table 1. For the period in question, the average daily windspeed was m s-l with a standard deviation of 0.84 m s-l. The TKE from wind stirring was unaffected by artificial destratification. This result is to be expected because this mechanism is independent of stratification. Convective overturn decreased by 30% with the use of artificial destratification, because the lower surface temperature was less effective at losing heat by evaporative and sensible transfer. By far the largest component of the TKE budget was shear production due to interaction of Reynolds stresses and mean shear. When using artificial destratification, shear production of TKE almost doubled, consistent with the observation that destratification using bubble plumes occurred from the bottom up (i.e. by the hypolimnion growing at the expense of the surface layer-see below). This process yielded a thinner surface layer than would otherwise be expected. As the velocity shear, AU, is inversely proportional to the surface layer depth (Fischer et al. 1979), greater shear velocities can be generated across the thermocline, leading to an increase in production of TKE by this mechanism. Although this is only a model result, the fact that the depth of the mixed layer is quite accurately predicted in the results offig. 5B supports it. The decrease in thickness of the surface layer is also in general agreement with laboratory results. There, experiments with the analogous setup of asymmetrical stirring on either side of a density interface resulted in the interface moving away from the region of more vigorous stirring until the entrainment rate on both sides of the interface balanced (Turner 1979). In our case the greater TKE comes from artificial destratification below the thermocline. Therefore, erosion is greater on the underside of the thermocline, causing the thermocline position to rise in response, yielding a thinner surface 1 ayer. Thermal response Temporal distribution -Isotherms interpolated from data at station DNE3 are shown as functions of depth and time in Fig. 6A. For the period 1 July 1989 to 19 February 1990, the isotherms are constructed from the temperature profiles collected every 2 weeks, while from 20 February 1990 daily means of the 15-min thermistor chain data are used. Figure 6B shows daily mean temperature traces at depths of 2 and 34 m below the water surface over the corresponding period, measured airflow rate and smoothed (2 1 -d running mean) air temperature. The degree of thermal stratification is best indicated by the average deviation of the vertical temperature distribution, defined as h IGW - dl&) dz s= s 0 s h (6) A(z) dz 0 Table 1. Average daily components of the turbulent kinetic energy (TKE) budget for the period 13 September 199 l-l 6 February 1992 (days ). Wind stirring Convective overturn Shear production Artificial destrat. 4.6 x x 1O x 10-5 TKE (W m-2) No artificial destrat. 4.6 x 1O x 1o x 1o-5 The mean, volume weighted temperature, P, is given by P= s 0 h s 0 T(z)A(z) h A(z) dz dz. (7) Here T(z) is the temperature and A(z) is the lake area at height z from the deepest point of the lake of total depth h. For the period July 1989 to July 1990 (when no artificial destratification device was used), the lake s natural cycle of summer stratification and winter overturn is seen in Fig. 6A. Hypolimnetic temperatures did not exceed 11 C throughout the year. Epilimnetic temperatures reached a maximum of 26 C in late February, after which time they generally decreased. There was a general deepening of the epilimnion over the stratified period. This pattern of surface layer deepening and cooling characterizes the lead up to overturn and is typical of temperate lakes. Overturn itself occurs when the surface layer has cooled to the hypolimnetic temperature in late June. Figure 6B shows that on a seasonal basis surface water (characterized by the 2-m temperature) heats up and cools down at a rate of 0. 1 C d-l, although day-to-day variability is greater. This parallels the rate of change of daily average air temperature. The hypolimnion in contrast (characterized by the 34-m temperature) has a more gradual, but consistent, warming over almost the whole year at a rate of C d-l, with cooling only occurring over l-2 months after overturn. Episodic events such as large inflows or outflows would change this rate, but for this period the lake level was relatively stable and large inflow volumes were confined to winter months. In the second year, July 1990 to July 199 1, significant stratification again commenced in early September, with a similar pattern of gradual epilimnion deepening. Artificial destratification did not begin until 16 December, initially with a low (0.08 m3 s-l) airflow rate through the long diffuser. The response to it was almost immediate, with the temperature at 34 m starting to rise at a rate of 0. 1 C d-l and the temperature at 2 m ceasing to rise at its natural rate. Destratification was stopped on 14 January and the hypolimnion temperature quit rising, but on recommencement 3 weeks later at the design airflow rate of 0.3 m3 s-l, the hypolimnion temperature again rose. The 2-m temperature initially fell at 0.27 C d-l

8 366 Schladow and Fisher 0 Time -- Mean -- Standard Deviation *r A A? Ic\ ET 0.0 l- 1989T-- L* Time Fig. 6. A. Isotherms for 1 July April From 2 1 February 1990, daily mean of thermistor chain data used. B. Temperature traces 2 and 34 m below the water surface, 2 1 -d running mean air temperature, and free airflow rate. Short diffuser shown as a bold line. C. Volume-averaged mean temperature and average deviation of vertical temperature distribution for the upper 44 m of the water column from 1 July 1989 to 30 April when aeration recommenced at the higher flow rate. De- and destratification operations were ended for the season. stratification, defined as occurring when the average de- The lake remained destratified thereafter. At that time viation of the vertical temperature distribution falls below the temperature was 17.5 C- about 7 C warmer than at 0.5 C for two successive days, was achieved on 3 April, the same time in The thermal destratification of

9 Response to destratljkation 367 the water column had occurred almost 3 months earlier than in the previous year and it had taken place by the hypolimnion growing at the expense of the epilimnion. Such growth was opposite to the manner in which natural destratification had been observed to occur (i.e. by the epilimnion growing at the expense of the hypolimnion). For the next 3 months the water column cooled by 0.06 C d-l, until by the end of July it was approaching its usual austral-winter temperature. This cooling and the continued near-isothermal status of the lake were completely due to natural processes. In the third year, July to July 1992, artificial destratification started on 30 September, soon after stratification first formed, and was used for discrete periods throughout austral summer. Both the short and long diffuser were used at different times. Due to the earlier commencement, hypolimnetic warming occurred very early in summer. As before, the hypolimnion temperature rose sharply only during periods of aeration, and the surface temperature rose and fell as aeration stopped and started. When sustained destratification was achieved on 3 1 March (by the earlier definition), the average temperature was 20.5 C, with the surface being -4OC cooler and the hypolimnion almost 11 C warmer than for the same time in 1990 (no artificial destratification). Although the previous year s result suggested that the near-isothermal conditions could be maintained at this time without intervention, artificial destratification was continued until 7 May. These data suggest that maintenance of a destratified state by natural mixing processes is possible only under certain conditions. For example, in mid-december and mid-february 1992, the water column was near-isothermal; yet, when the airflow was stopped, it restratified. At other times, for example mid-april 199 1, a destratified state was sustainable. The key to what ensues is the strength of the stratifying flux, as discussed above. In Fig. 6B, the average daily air temperature at station DNE3 is used as a measure of the stratifying flux (shortwave radiation shows the same trend). For the austral summer, the air temperature (and radiation input) had fallen to about midway between the maximum summer and minimum winter values when destratification was achieved in April. Sufficient reduction of the stratification flux had occurred so that the natural destratification processes could maintain thermal homogeneity. At the earlier times when aeration was stopped, this was not the case. Thus from a thermal standpoint, artificial destratification may need to be practiced for only part of the traditional stratification season. However, if the flux of oxygen (or any other chemical) due to internal sources and sinks is larger than that maintained by natural mixing, a chemocline will still develop. In such a case, artificial destratification will need to be maintained for a longer period, as occurred in the summer. Thermal destratification was achieved at the same time as the previous year; however, a heightened oxygen depletion rate due to a large, turbid inflow required that the destratification system be kept in operation for a further 4 weeks. The overall effect of artificial destratification on the water column is best seen with means and average deviations of the vertical temperature distributions plotted against time, as in Fig. 6C. Mean lake temperature is higher as a result of artificial destratification. The peak meantemperaturein was 17.4 C--4.l Chigher than the peak for the previous year. Over both these summers, the lake volume was about the same. In 199 l the peak mean temperature was 19.8 C. This further increase was in large part due to a 10-m drop in water level the preceding winter, which reduced the total lake volume by a third. The average deviation was markedly smaller during periods when artificial destratification was used. Whereas in the average deviation was over 75% of the mean (indicating the high surface and low hypolimnetic temperatures), it was < 10% of the mean during periods of aeration in the two subsequent years. The average deviation is in fact an excellent descriptor of the stratification, and comparison of it with the airflow records of Fig. 6B shows a strong inverse correlation. A more detailed picture of the temporal processes is obtained with the 15-min thermistor chain data, as shown in Fig. 7. For clarity, only the temperature traces at 2 and 34 m, and the daily average airflow rate are plotted. In Fig. 7A, the 15-min data began on 21 February (day 90052). There is little deviation from a constant temperature in the hypolimnion except for the period at over- turn when it falls - 1 C. There is more subdiurnal variability in the 2-m signal, but it is generally < 1 C. In Fig. 7B, the subdiurnal variability at 2 m appears to be somewhat larger during the heating cycle of the lake surface. However, once artificial destratification begins on 15 December (day 90349), there is a pronounced increase in the intensity of the surface temperature variability. On occasion, temperature fluctuates by as much as 6.2 C in 3 h. These fluctuations are indicative of the highly unsteady nature of the advection and mixing as- sociated with the bubble plumes. Similar fluctuations have been seen in laboratory studies (Zic et al. 1992). During this period, only the long diffuser was used. At its closest point it was still 500 m from the thermistor chain. As the water column approaches a homogeneous condition, the intensity of the temperature fluctuations diminishes. At this scale, the rate of change of hypolimnion temperature from 9 February (day ) to 10 April (day ) can clearly be seen to be rising at a gradually diminishing rate until it asymptotically approaches the temperature of the surface layer. By contrast, the epilimnion first experiences a rapid drop in temperature at the beginning of this period (on average, 0.3 C d-l), but by 24 February (day ) its mean temperature has fallen very little until complete destratification is achieved. The abrupt change in temperature on 11 June (day ) is the result of a large inflow. Because the inflow is cooler than any part of the water column, it underflows and has the effect of temporarily restratifying the water column. Note that if artificial destratification had not been taking place, the temperature of this inflow would have caused it to be inserted within the thermocline of the lake. In Fig. 7C the intensity of the temperature fluctuations is again seen to increase when the aerator is in operation. The hypolimnion temperature rises immediately after ar-

10 368 Schladow and Fisher 30.0 _ :A p V $ % g ls.o{ g m m -3 LtJ!f!s& 5: Eit 2 OS3 x I ( 1 1 I, I I I, I Ii 0.0 Jl Day Number Lnng Diffuser 2m -- Short DifTuser 25.0 s ; k a P lo.o,a, 1, I, I, 1, I -Airflow, I, I - o* Day Number _ _ _ g - 17 ii :34m * g lo.o- / 1, 03;. b Airflow b.l ,I ( * I, A!!? Day Number 10.0 jb,w.,,,.,,,!. -Airflow f o.o,,,,,,,,,,,,,,,,,,,,,,,,,,,,, a Day Number Fig. 8. A. Fifteen-minute thermistor chain and airflow data for 24 November-2 December 199 1, showing the impact of the commencement of artificial destratification on the temporal behavior. The depths below the surface of the temperature traces shown are (from the top) 0, 2, 4, 5.5, 7, 8, 9, 11, 17, and 34 m. B. Enlargement of the same data for November, showing the disturbance of temperature traces by the arrival of intrusion fronts. Day Number Fig. 7. Fifteen-minute thermistor chain traces at 2 and 34 m below the water surface and measured airflow data. A. 1 January-26 October B. 27 October August C. 24 August 199 l-23 June tificial destratification begins, gradually tapers off as it continues, and then abruptly quits rising when it is stopped. This pattern is repeated 4 times, corresponding to the four sustained periods of bubble plume operation. The temperature of the surface layer falls, although subdiurnal fluctuations increase. However, as in the previous year it eventually reaches a plateau. Presumably this represents a balance between cooling from below and warming by atmospheric transfers from above When artificial destratification is stopped, the surface temperature again rises sharply due to the atm0spher:i.c heat transfers. The sharp drop in temperature on 10 February (day ) is due to a large inflow which resullted in a 5-m rise in water level. As in the previous year, this underflowed and period, but almost 180 out of phase with the diurnal surface forcing. The 2-m temperature trace appears to be equally influenced by both. When artificial destratification began at hours on 27 November (day l), it took < 3 h for the changes to be noticeable at the thermistor chain. Similar time delays were observed during other start-up periods. The immediate response is for the seiche period to be lengthened and then for the seiche to disappear altogether as the temperature difference through the water column decreases. This response is well illustrated by following the 4-m thermistor trace. Spatial distribution -There are two types of changes in the spatial distribution of the temperature (and hence density) field in response to artificial destratification. The first is the changing of the basically one-dimensional density distribution to a two-dimensional distribution through homogenization of the water column in the vicinity of the aerator (the near field). The second is the response of the stratified water body to this two-dimensional structure and takes the form of longitudinal motions (the far field). restratified the water column. Changes right at the onset of artificial destratification The near-field response can be identified in longitudinal are evident in an enlargement of Fig. 7C. Figure 8A shows temperature sections of Figs. 9 and 10. The sections are all the thermistor traces and airflow data for 24 Novem- constructed from the CTDO profiles taken at the moniber-2 December (days ). Before commence- toring stations. Their locations and the vertical depth of ment of the airflow, pronounced diurnal heating and cool- each profile are indicated on the figures. With the exccping is present at the surface. Between 4 and 11 m all the tion of DNE3, all stations are nominally located in the thermistor traces are dominated by a seiche with a 24-h central channel of the lake and decrease in depth away

11 Response to destrat&ation A g 300 e l6 v 4% w s A A A A A A I I I I I I I I I DISTANCE FROM DAM WALL (m) B I I I I I I I I I I DISTANCE FROM DAM WALL (m) Fig. 9. Longitudinal temperature section based on CTDO data. DNEl is at 0 m. Arrowheads show location of each profile and maximum profile depth. A. 19 November 199 1, no artificial destratification occurring, B. 24 October 199 1, artificial destratification in progress. from DNE 1. The steep sides of the lake caused the profiler to intersect the base at a much shallower depth if it was deployed slightly off station. The undisturbed state is evident on 19 November (day 91323) in Fig. 9A. Dcstratification has not taken place for 18 d, and there is virtually no horizontal deviation in the temperature field. The section in Fig. 9B is from profiles taken between 1156 and 1341 on 24 October (day ). Destratification began - 16 d earlier through the short diffuser. The temperature section shows distinct cooling of surface water and warming of the bottom water between DNEl and DNE3. The distinct deviation of the isotherms from the horizontal in this region is maintained by the strong upward entrainment flow that is produced by the bubble plumes. The coarse horizontal spacing of the sampling stations smoothes the real temperature gradients, masking the sharp fronts that exist close to a bubble plume (Lemckert and Imberger 19933). However, the important result from the standpoint of destratification is that in the near field, the water column becomes more homogeneous. This structure is not associated with seiching as evidenced by the continuous temperature record at DNE3. It is also relatively long lived, being present at the previous profile survey on 13 October (day ). Beyond DNE3 (the far field), horizontal differences in the temperature field are small enough to be almost masked by the interpolation used in constructing the isotherms. However, close examination of the data shows consistent longitudinal gradients on the order of 1 OM20C km-l between stations DNE4 and DNE6 over much of the water column below 5 m. Such gradients must exist for there to be a baroclinically driven, global redistribution of the local effects of destratification away from the entrainment region, although the magnitude may vary due to the superimposed effect of seiching. The slope of the isotherms, tilting upward away from the bubble plumes, is consistent with a general circulation pattern that is away from the bubble plumes in the upper part of the water column and toward them in the lower part. Figure 10 shows dissolved oxygen sections at the same

12 370 Schladow and Fisher > =--- A A ~ DISTANCE FROM DAM WALL (m) 320 t B d (I I I I I I -_I DISTANCE! FROM DAM WALL (m) Filg. 10. As Fig. 9, but of dissolved oxygen section. times as the above temperature sections. When destratification is not taking place, dissolved oxygen isopleths are free to deviate from the horizomal, as they have no bearing on the density distribution. However, with destratification, they too show a far greater degree of twodimensionality. Between DNE 1 and DNE3 there is a similar homogenized region of oxygen increase in bottom waters and oxygen depletion in surface waters. Away from this region, the increase in slope is partly due to the natural slope of the isopleths parallel to the sediment and partly due to baroclinic circulation. Intrusions Earlier, the rise of a bubble plume in a stratified environment was said to produce one or more intrusions associated with the discharge of entrained water. Yet, Fig. 9B shows no clear suggestion of any such intrusions. Tntrusions are in fact very difficult to observe directly in the temperature signal, as they intrude into water of similar temperature (i.e. density). The large-scale temperature fluctuations noted previously suggest that the mixing generated by the rising plume is incomplete and that considerable density segregation of the entrained fluid occurs. Thus the intrusions are likely to be stratified, in a manner similar to the ambient fluid. The dynamics of the density front at the head of an intrusion, by contrast, permit unambiguous identification. Beyond an initial axisymmetric region (limited to 100 m at Lake Nepean by the shore), intrusions would be expected to propagate down the axis of the lake as two-dimensional intrusions. Certainly by the time the intrusion had reached station DNE3, it would have the attributes of a two-dimensional intrusion. Thus, the description provided by Imberger et al. (1976) is likely to be more apt than that for axisymmetric plumes (Lemckert and Imberger 1993a). Two-dimensional intrusions are dependent on the Grashof and Froude numbers, defined as Gr = WL4/v2 and Fr = q/nl2, where q is the volume flux per unit width of the insertion, L the length of the lake at the insertion height, and v the kinematic viscosity. Specifically, when FrGr Ii3 > 1 a buoyancy-inertia balance applies and when FrGr l/3 < 1, a buoyancy-viscosity balance applies. Conditions encountered at Lake Nepean

13 Response to destrat$cation 371 / entrainment - - detrainment - net flux Temperature ( C) Volume flux ( 106m3m- d-l) Fig. 11. A. Measured temperature profile for 27 November B. Model output of entrainment flux, detrainment flux, and net volumetric flux for 27 November. were always conducive to the buoyancy-inertia balance. Imberger et al. (1976) give the order of magnitude of propagation velocity of such intrusions as U = O(LNFr1 2). (8) The front or nose of such intrusions is always sharply defined and accompanied by breaking, behind which there is a turbulent wake (Benjamin 1968). Such abrupt, but coherent, changes in temperature are identifiable in Fig. 8B, which is an enlargement of part of Fig. 8A. The bubble plumes commenced at 1500 hours on 27 November (day ). Between and 1730 hours there is a sharp rise in temperature at the 7-, 8-, 9-, and (to a lesser extent) 1 l-m thermistors and a fall in temperature at the 4- and 5.5-m thermistors. Between 1730 and 1800 hours the temperature change at all these thermistors reverses by an even greater amount. These events are taken to indicate the passage of the fronts of two intrusions in the region between 2 and 17 m below the surface (as thermistors at these depths showed no abrupt changes). The relative stability of the 2- and 7-m thermistor traces suggest regions of return flow above each of these intrusions. The 2-m trace does not change appreciably from its preaeration pattern until 20 h after aeration has commenced, whereas the 4-m trace drops almost 4 C in this period. Similarly, the 7-m thermistor trace fluctuates about its preaeration value for almost 2 d. Calculations support these conclusions. The measured temperature profile just before commencement of the bubble plumes on 27 November 1991 (day ) is shown in Fig. 1 IA. The calculated volume fluxes as a function of depth for entrainment by the bubble plume and detrainment by two-dimensional intrusions are shown in Fig. 11B. The bubble plume model predicts that two plumes are formed, with the lower one rising to a height of -52 m and the upper one reaching the surface. The maximum height of rise of the lower plume can be distinguished by the drop in magnitude of the entrainment at that depth. The centers of the intrusions correspond to the two peaks in the detrainment. These arc at 3 and 14 m below the water surface and correspond to the levels of neutral buoyancy. These both fall within the region of observed intrusive flow. The fact that the center of volume of the entrainment distribution is always higher than the center of volume of the intrusion accounts for the net flux of epilimnetic water into the hypolimnion and the resulting increase in hypolimnion depth. The positive and negative values of the net flows are interpreted as flows away from and toward the bubble plume respectively. There are three regions of flow toward the aerator: at the surface, between the intrusions, and at the bottom. The large jetlike return flow between the two intrusions arises because the level of neutral buoyancy of the intrusion is considerably lower than the plume s maximum height of rise. It is likely to bc a feature of the flow whenever more than one plume is present. The surface return flow arises for a similar reason. These calculations indicate that it is always present to some degree, regardless of the number of plumes. The bottom return flow is simply the result of there being no detrainment at the bottom. The two upper return flows occur at a level consistent with the 2- and 7-m thermistor traces. From Fig. 11 the temperature gradients at the upper and lower intrusions are -0.2 and 0.06 C m-l respectively, corresponding to N values of 0.02 and Tntegrating the net intrusion profiles over depth in Fig. 11B yields flow rates of 1.43 and 5.18 x 1 O6 m3 d-l respectively. If we take the width of the lake as lo2 m and the length as 1 O4 m and substitute in Eq. 8, the upper intrusion has an estimated arrival time at DNE3 of 3.1 h, while the lower intrusion has an arrival time of 2.3 h. This estimate is in close agreement with the observed arrival time of the disturbance of h. Thus the theory suggests that there are two discrete intrusions that lie within the observed band characterized by the wavelike changes and that the theoretical time of