Hydro-physical processes at the plunge point : an analysis using satellite and in situ data

The plunge point is the main mixing point between river and epilimnetic reservoir water. Plunge point monitoring is essential for understanding the behavior of density currents and their implications for reservoir. The use of satellite imagery products from different sensors (Landsat TM band 6 thermal signatures and visible channels) for the characterization of the river-reservoir transition zone is presented in this study. It is demonstrated the feasibility of using Landsat TM band imagery to discern the subsurface river plumes and the plunge point. The spatial variability of the plunge point evident in the hydrologic data illustrates the advantages of synoptic satellite measurements over in situ point measurements alone to detect the river-reservoir transition zone. During the dry season, when the river-reservoir water temperature differences vanish and the river circulation is characterized by interflow-overflow, the river water inserts into the upper layers of the reservoir, affecting water quality. The results indicate a good agreement between hydrologic and satellite data and that the joint use of thermal and visible channel data for the operational monitoring of a plunge point is feasible. The deduced information about the density current from this study could potentially be assimilated into numerical models and hence be of significant interest for environmental and climatological research.


Introduction
According to the report published by the International Committee of Large Dams (ICOLD; www.icold-cigb.net) in 2000, over 40 000 large dams exist in the world, with a total storage capacity of 7000 billion m 3 .An average of 0.5 to Correspondence to: A. T. Assireu (assireu@gmail.com) 1 % of their capacity is lost each year due to sedimentation.Changes due dams can include increased depth, changes in temperature, the possible development of density stratification, the retention of nitrate and phosphate, the growth of plankton and algae, contaminant trends in sediment, changes in the aquatic ecosystem from lotic to lentic, and the release of carbon-based greenhouse gases (Baxter, 1997;Poff and Hart, 2002;Ramos et al., 2006).
Because the density of the river water reaching a reservoir usually differs from the density of the water at the reservoir surface, inflows enter and move through the reservoir as density currents (Ford, 1990).Temperature, total dissolved solids, and suspended solids can cause density differences.Depending on the density difference between the inflows and ambient waters, the density currents in a reservoir can flow into the downstream area as overflow, underflow, or interflow types (Martin and McCutheon, 1999).
Thus, after the inflow plunges the river, it can follow the thalweg as an underflow.It is frequently assumed that once an inflow has plunged into a reservoir and formed an underflow or interflow, its constituent load is isolated from the surface waters.Although this situation has been shown to be true in many cases, recent studies (Chen et al., 2006) have indicated that mixing can entrain inflow constituents into surface waters.
It is not uncommon for environmental fluids to be subjected simultaneously to the destabilizing effects of a velocity shear stress (induced by underflow) and to the stabilizing effects of density stratification.The outcome of such competition is often the Kelvin-Helmholtz (KH) instability (e.g.DeSilva et al., 1996;Thorpe and Jiang, 1998;Özgökmen and Chassignet, 2002;Umlauf and Lemmin, 2005).The resulting entrainment has been extensively studied elsewhere (e.g.Chung and Gu, 1988;Gu and Chung, 2003).
Improved knowledge of the behavior of density currents is essential to determine the spatiotemporal variability of  water characteristics.Density currents may transport undesirable materials, such as suspended solids, organic chemicals, or nutrients from the watershed, to reservoirs downstream (Chen et al., 2006) or may cause sedimentation (Annandale, 1987;De Cesare et al., 2001).Thus, appropriate reservoir operation to avoid sedimentation or water quality deterioration requires knowledge of density current behavior.
Some studies regarding density currents have been carried out in the Feitsui Reservoir (Chen et al., 2006), Lake Ogawara (Dallimore et al., 2001) and the Katsyrasawa Reservoir (Chikita, 1989) in Japan, Lake Geneva in Switzerland (Lambert and Giovanoli, 1988), Lake Onachota and Fayetteville Green Lake in the United States (Ford, 1990), and the Alpine Luzzone Reservoir in Switzerland (De Cesare et al., 2001).To our knowledge, there have been no studies reporting the use of remote sensing to study the plunge point location in reservoirs.
The available literature uses visible channel data to look at changes in the clarity and color of the water (Chipman et al., 2004) associated with changes in sediment input (Mertes et al., 1993) or the amount of chlorophyll (Novo et al., 2006) due to algal blooms (Duan et al., 2009) and uses thermal infrared data to look at changes in the surface temperature (Alcântara et al., 2010) associated with upwelling (Schladow et al., 2004;Steissberg et al., 2005) or changes in circulation (Ikeda and Emery, 1984).
Some authors (e.g.Anderson et al., 1995) deployed a hand-held thermal radiometer to obtain indirect measure-ments of the water surface temperatures to study the influx of cold water in reservoirs.However, this kind of data is limited in time and space.The knowledge of the physical, chemical, and biological mechanisms controlling water quality is limited by a lack of observations at the spatial density and temporal frequency needed to infer the controlling processes.The ability when using satellite images to combine synoptic information with temporal recovering makes remote sensing a very useful tool to study plunge point locations.
The plunge point location indicates where the riverine zone stops and the lake zone begins.If the riverine zone extends a significant distance into the reservoir, the reservoir is dominated by advective forces and cannot be considered to be vertically one-dimensional (Arneborg et al., 2004).Because the processes occurring in a reservoir are dependent on the river behavior along the reservoir, which is strongly determined by the contrast between incoming river water and reservoir water, remote sensing can be used to study some important aspects of reservoir functioning.It is important because, depending on the location, in situ measurements of temperature, such as those provided by the use of thermistor chains, might not capture the plunge point spatial variability and, consequently, might not be useful to determine if the inflow will behave as an under, inter or overflow.
A number of models have been proposed to predict plunge depth (Dallimore et al., 2004).These models are generally valid only for channels with a constant slope and width.The failure to accurately predict the plunge depth will result in an inaccurate representation of the initial underflow layer thickness and therefore cause a distortion of the underflow dynamics beyond the plunge point (Dallimore et al., 2004).
Based on this, the purpose of this paper is to jointly apply thermal and visible satellite instruments, in situ data and satellite-tracked drifters to monitor the density current between a water body and the inflowing water (at the plunge point location).

Field site and measurements
The Manso Reservoir (14 • 50 S; 55 • 45 W), located in a Brazilian savanna biome, has a dendritic shape with 427 km 2 of flooded area and a maximum depth of 60 m (Fig. 1).
The annual mean temperature is 26 • C, and the total annual precipitation averages 1750 mm, concentrated during the summer (the wet season).Furnas Centrais Elétricas, responsible for operating the reservoir, completed filling the reservoir in 1999.In the Manso Reservoir, the summer stratification is characterized by temperature differences of up to 4 • C between the epilimnion and hypolimnion.measured for each sample station according to Eaton and Franson (2005).

Environmental time series
The temporal variability of the environmental parameters used here was measured using data collected with an anchored buoy system, the Integrated System for Environmental Monitoring (SIMA).This system is a set of hardware and software designed for data acquisition and the real-time monitoring of hydrological systems (Stech et al., 2006), with fixed data storage systems, sensors (air temperature, wind direction and intensity, pressure, incoming and reflected radiation and a thermistor chain), a solar panel, a battery and a transmission antenna.The data were collected in preprogrammed time intervals and were transmitted by satellite in quasi-real time for any user within a range of 2500 km from the acquisition point.The selection of the environmental parameters measured using the SIMA (moored near the dam, indicated by Station-13 in Fig. 1) took into account aspects including the relevance as an environmental index (i.e. the variables that respond consistently to alterations in the functioning of an aquatic system), the importance of the greenhouse gas-emission process in aquatic systems, and the technical suitability for data acquisition and transmission from automatic platforms (Stech et al., 2006).

Environmental profiles
Vertical profiles were measured during the wet season (Fig. 1) as follows: (1) eleven stations were oriented in the longitudinal direction from the river (stations 1 to 8) toward the dam (stations 10, 12 and 13), following the thalweg, and (2) three stations were oriented in the transverse direction.This design allowed for the construction of one longitudi-nal transect composed of 11 profiles and two transverse transects, one near the dam (stations 13 and 14) and the other 20 km from the dam (stations 9, 10 and 11) (Fig. 1).
The stations were spaced 3 km apart in the reservoir and 0.5 km apart in the river-reservoir transition zone.The vertical profiles of pressure, water temperature (with a resolution of 0.01 • C and 5 % accuracy), dissolved oxygen concentration (resolution: 0.01 mg l −1 and accuracy: 2 %), and conductivity (resolution: 1.0 µS cm −1 and accuracy: 0.5 %) were made with a multi-parameter probe (Yellow Springs Instr., model 6600) and a submersible fluorimeter (Satlantic) with a resolution of 0.05 µg l −1 and range of 0-200 µg l −1 .
A total of 56 contained Hobo temperature loggers were deployed in five thermistor chains at 1-m intervals from depths 0 to 4 m and at 3-m intervals from depths 4 to 20 m.The sampling period of the thermistor was 15 min.

Circulation measurements
The water circulation measurements in the river-reservoir transition zone were carried out using drifters.These drifters were built following the design developed by Sybrandy and Niiler (1991) and consisted of a GPS data logger installed in a spherical fiberglass float and connected to a drogue to reduce the wind-slip effect.The drifters were drogued at 6 m because the focus in the present experiment was on the inflow velocity.This depth was selected based on the temperature profiles obtained with the multi-parameter probe.
The sampling frequency was 0.01 Hz, and the positioning precision was 10-m horizontal root mean square.Three drifters were released simultaneously along the riverreservoir transition zone at different positions (stations 1, 3 and 5 in Fig. 1) and collected successive position (latitude and longitude) data continuously for approximately 23 h.The field velocity was estimated from these time series of the drifter positions.To remove the noise, the resulting time A. T. Assireu et al.: Hydro-physical processes at the plunge point series were filtered using a low-pass filter with a cutoff period of 15 min.

Computing water surface temperature from
Landsat-5-TM data Generally, the grey level in images from the Thematic Mapper (TM) sensor of the Landsat satellites is given as a digital number (DN) ranging from 0 to 255 (8 bits).Thus, the computation of the brightness temperature from TM6 (thermal band: 10.5-12.5 µm) data includes the estimation of radiance from its DN value and the conversion of the radiance into brightness temperature.
For this purpose, we utilized the equation developed by the National Aeronautics and Space Administration (NASA),

+1
, where T is the effective at-satellite temperature in kelvin, K2 and K1 is the calibration constant 2 and 1 in kelvin (W m −2 sr −1 µm −1 ), respectively (Chander et al., 2009).The water vapor content and atmospheric temperature in each layer were approximated using local meteorological observation data (Qin et al., 2001), measured at the Weather Forecast and Climatic Studies Center (CPTEC) of the Brazilian National Institute for Space Research (INPE).

Wedderburn and Lake numbers
The Wedderburn number, W = c 2 h (u 2 * L) , was calculated following the method of Imberger and Patterson (1990).W is the ratio of the maximum baroclinic pressure force (before the upwind surfacing of the thermocline) and the surface wind force.The relevant internal gravity-wave speed was calculated as c = g h.The dynamic balance defined by W was extended with the vertically integrated Lake number (L N , Imberger and Patterson, 1990;Stevens and Imberger, 1996).This quantity is defined as , where g is gravity, ρ is the density of water, z T is the height of the center of the metalimnion, z g is the height of the center of volume of the lake, A is the lake area, H is the depth of the lake, u * is the friction velocity in water and S t (g cm −1 cm −2 ) is an estimate of the stability of the reservoir calculated as S t = zm 0 (z − z g )A(z)ρ(z)dz (Hutchinson, 1957).As Stevens and Imberger (1996) have shown, S t can be considered as an integral equivalent to W .

The entrainment velocity, the depth of the plunge point, and the Kelvin-Helmholtz instability
Some authors (Britter and Linden, 1980;Monaghan et al., 1999) have found from theory and laboratory experiments that the head of density currents travels at a constant speed, U F , which is proportional to a third power of the buoyancy flux per unit width, defined as B F = g Q, where g (the reduced gravity) is equal to g ρ p /ρ o (where ρ p is the density difference across the pycnocline, ρ o is mean density, and g is gravity) and Q = h r U r is the volume flux per unit width, with h r and U r being the average thickness and speed, respectively, of the current at the input location.Studying gravity and density currents, Özgökmen and Chassignet (2002) showed that numerical simulations compare well with laboratory measurements and that laboratory results are valid on geophysical scales.
The depth (m) of the plunge point was estimated with the following equation (Ford, 1990): where h is the hydraulic plunge point depth (m), ρ is the difference between the densities of the inflow and reservoir water (kg m −3 ), and L is the width of the convergence zone (m).
The density function, developed by Gill (1982) and Ford (1990), is presented in the following equations: where ρ T w is the density of water and is a function of water temperature (T w ), ρ s is the density increment owing to the concentration of suspended solids (C ss ), and "sg" is the specific gravity of suspended solids, assumed as reported by Chen et al. (2006).
The entrainment coefficient, E, and the normal depth of the underflow along the centerline, h c , were calculated as reported by Chen et al. (2006) using the following equations: where x is the distance downstream from the plunge point and F is the Froude number at a plunge point, defined as follows: The slope, together with the density difference, provides the forcing of the flow against friction.The mean slope of the reservoir bed (s) is approximately 30/40 000 (approximately 0.00075), and the drag coefficient (c d ) was assumed to be approximately 0.003, as reported by Arneborg et al. (2004).with that measured using the drifters.

Meteorological conditions and thermal structure
The temporal patterns of rainfall, wind and air temperature (Fig. 2a) and short-and long-wave radiation (Fig. 2b) show that the daily precipitation can exceed 80 mm during the wet season in the studied region.As detailed below, the resulting increases in the inflow have a strong effect on the thermal structure of the reservoir during the wet season.In the dry season the long-term decreasing trend in the short-wave radiation and increase in the wind speed are easily seen.Southwesterly winds were predominant in dry season (winter), when speeds reached up to 8.0 m s −1 , with an average 3.5 ± 1.8 m s −1 .The predominant southwesterly direction coincides with the direction of main Manso Reservoir fetch (Fig. 1).
At the end of April and July, seven cold-front events reached Manso Reservoir (arrows in Fig. 2a).Further details about the action of cold fronts in Manso Reservoir were reported by Lorenzzetti et al. (2005).Each peak of the wind speed was associated with a steep decrease of the air temperature during the cold-front passage.The air temperatures declined by approximately 6 • C during these events, and the daily average wind speed reached 6 m s −1 .
Because the latent and sensible flux terms are almost linearly dependent on the wind and the short-wave drops during the cold-front passage, the net surface radiation becomes negative, and the reservoir loses (latent) heat to the atmosphere (data not shown).Thus, the deepening of the surface mixed layer due to the stirring from the wind stress will be enhanced by downward convective mixing during dry seasons, as observed in Lake Victoria, East Africa, by MacIntyre et al. (2002).
The thermal stratification in Manso Reservoir is moderate and is spread over a thick metalimnion during the first half of the analyzed period (2 February-30 April) (Figs. 3 and  4).The metalimnion extends from 7-25 m, with temperatures ranging from 28-30 • C.During this period, the wind speed averaged 2.3±0.5 m s −1 , without any predominant direction.The stability of the water column, given in Hz by N = (gρ −1 ∂ρ/∂z) 1/2 (g, acceleration due to gravity; ρ is water density), was solely temperature-dependent (salinity ∼0.02 ‰ calculated from conductivity) and showed values within 5-18 cph [N(cph) = (3,600/2π)×N(Hz)], reflecting a moderate stratification (Fig. 1b) Values around 20 cph correspond to strongly stratified oceanic waters (MacIntyre et al., 2002).During the stratified period of the wet season, the thick metalimnion and moderate stratification could make Manso Reservoir susceptible to partial upwelling (W > 2 or LN < 10), i.e. metalimnetic water reaching the surface.However, an analysis based on the Wedderburn and Lake Number (Fig. 4) indicates that the wind-driven upwelling/downwelling was unimportant during the wet season.Some enhanced mixing events are visible during 16-19 February, 18-21 March, and 6-9 May, when water of 28 • C was transported to the surface.
In the next section, we show evidence regarding the contribution of the river to enhancing the mixing process, indicated by arrows in Fig. 3.The reservoir thermal structure (Fig. 4) suggests that in the wet season, the river stays at the hypolimnion level and contributes to the thermal stability of the water column.This situation persists during the wet season.In the dry season, the river circulation is characterized by interflow-overflow, as discussed below.

River effects on the thermal structure and Kelvin-Helmholtz instability
In situ measurements during the wet season showed a high contrast between the suspended solid concentrations in the reservoir (4.2 mg l −1 ) and river water (15.0 mg l −1 ).This contrast can be easily observed in the satellite images (Fig. 5a).This contrast is also observed in the surface water temperatures (Fig. 5b).Note that in Fig. 5, there is a difference of approximately 5.40 km in the plunging point location measured using the satellite image acquired in the visible spectrum (Fig. 5a) and that in the thermal region (Fig. 5b) during the wet season.This difference can be explained by the fact that the solar radiation interacts with the optically active components in the  uppermost portion of the water column (such as chlorophyll, inorganic particles and dissolved organic matter), whereas the thermal radiation detects skin temperatures.
The turbidity currents can be observed in Fig. 5a to travel underwater until they reach a depth at which the sensor can no longer detect them.Thus, the true plunge point position is given by the thermal image.In the dry season, the plunging point location in the visible-spectrum satellite image (Fig. 5c) is consistent with that of the thermal image (Fig. 5d).
As noted by Ford (1990), these water density differences are mainly caused by differences in temperature and dissolved-or suspended-solid loads throughout the riverreservoir transition.According to Akiyama and Stefan (1984), these physicochemical patterns cause density differences between the reservoir and river waters, which are responsible for the plunging inflows.
Manso Reservoir experiences large seasonal and short time runoff variations.As a result, the inflow variations present an intermittent pattern, with large inflows during the summer and autumn months (Fig. 6a).This fact suggests that the inflow controls the residence time for Manso Reservoir, as verified by Rueda et al. (2006) for the Sau Reservoir.The estimated power spectra at the thermocline displacement and the buoyancy flux due to inflow show similar variabilities, with frequencies between 0.3 and 0.6 cpd (Fig. 6b).
If the peaks in the range of 0.3-0.6 cpd were indicative of an interchange between thermocline oscillation and buoyancy flux, the peaks would be coherent but out of phase because the signal of inflow must be propagated 23 km downstream to the thermistor chain.If the signal had a propa-  gation speed u of approximately 0.1 m s −1 (as estimated using a model and confirmed using the drifter data) over a distance of 23 km, the propagation time would be approximately 2.5 days.The coherence spectra (Fig. 6c) indicate that motions in the range of 0.3-0.6 cpd are coherent and are 120 • out of phase (Fig. 6d).This result suggests that the inflow is an important mechanism driving the thermocline displacement.
The river water progressively sinks down to the level of the thermocline and contributes to the thermal stability of the water column during the wet season (Fig. 7).After the inflow plunges, it can follow the thalweg as an underflow partially enhanced by downstream turbine and spillway outflows (outflow at Manso ∼15 m).
The river water had a similar temperature, conductivity, and pH as the hypolimnion (Figs. 7, 8) at the dam area (39 km downstream of the river), indicating that the river followed the thalweg as an underflow.In this case, during the wet season, the following fluid behavior is expected: a denser fluid enters a domain and flows under a lighter fluid, waves develop along the interface (due to shear velocity), and these waves grow in the downstream direction, develop into billows and eventually roll up.This behavior is indicative of the Kelvin-Helmholtz instability, in which waves made up of fluid from the current (river) entrap the ambient fluids (Thorpe and Jiang, 1998;Corcos and Sherman, 2005).
The usual condition for the onset of Kelvin-Helmholtz instability is that the gradient Richardson number, Ri =  is Richardson-number dependent.Some authors (Kunze et al., 1990;Trowbridge, 1992;Peters et al., 1995;Baringer and Price, 1997) show that the Ri estimates are very sensitive to the length scale over which the vertical gradients are calculated.Thus, we apply in our analysis an alternative to Ri estimates for a two-layer flow.A similar bulk analysis was   performed by Cushman-Roisin (2005) and yielded a criterion for the vertical extent of the eventual mixing.If the following inequality holds true, (ρ 2 −ρ 1 )gH ρ o (U 1 −U 2 ) 2 < 1, then vertical mixing is energetically possible.
To evaluate the typical vertical velocity shear, the current measured using the drifters in the river-reservoir transition zone was analyzed.The mean difference between the currents at 6 m and 2 m was approximately 6.0 cm s −1 (Fig. 9b).This difference (U 1 − U 2 ), applied in the above inequality, indicates that the Kelvin-Helmholtz instability may play a strong role in the thermocline displacement during the wet season.Alongshore components of the drifter velocity (Fig. 9a) show the anisotropy in the mean kinetic energy.
During the dry season, the decrease in the river flow and the decrease in density differences (Fig. 5d) lead to river circulation characterized by inter-overflow.As discussed, many cold-front passages occurred during this period.Thus, important energetic mixing processes are driven by winds and by surface cooling during the dry season due to cold-front action, and hence, the closer the river water is to the surface, the higher is the probability that it experiences mixing with the reservoir waters.
An inter-overflow situation may have an important influence on reservoir water quality by introducing oxygendemanding material, nutrients, and bacteria directly into the surface waters or euphotic zone (Alavian et al., 1992).
Hydrol.Earth Syst.Sci., 15,[3689][3690][3691][3692][3693][3694][3695][3696][3697][3698][3699][3700]2011 www.hydrol-earth-syst-sci.net/15/3689/2011/   Available nutrients can be assimilated by phytoplankton and may stimulate plankton blooms.In the downstream direction, where an inter-overflow would hit the dam construction, an accumulation of phytoplankton could occur and may be seen from satellite imagery (Fig. 10).This process affected the water quality: the concentrations of suspended solids at the surface varied from 2.2 to 4.8 mg l −1 (dry season -interoverflow) and dropped to the range between 0 to 1.0 mg l −1 (rainy season -underflow) in the reservoir, while in the river, this concentration was 15 mg l −1 and 10 mg l −1 during the dry season and the rainy season, respectively.Therefore, the larger values observed in July (although a dry month) would be due to river-water insertion into the upper layers of the reservoir, as an inter-overflow.This process was also depicted in the water transparency (measured using  a Secchi disk) in Manso Reservoir, when the transparency dropped from 5.5 m to 4.5 m.Through these results, we will pass to discuss the plunging underflow in more detail.

Analysis of a plunging underflow
The important density current parameters (i.e.propagation speed, thickness, and the point of plunging) can be used to determine a change in water quality at different depths (Alavian et al., 1992).Despite the limited predictive ability of empirical models, due to the complex bathymetry of real reservoirs and continuous rather than layered stratification (Dallimore et al., 2004), which indicates the need for observations of density currents, we used a simplified onedimensional model (Eqs.1-7) to estimate and compare the model predictions with the hydrological data.The thermal structure, shown in Fig. 11, and the average inflow water temperature (26.5 • C) were utilized.
Based on the measured suspended solid concentrations in the river, the inflow suspended solid concentration was 15 mg l −1 .The estimated total density of the inflow, including the density increment due to suspended solids, was 996.53 kg m −3 .The reservoir mean water temperature was 28 • C, and the density difference between the inflow and reservoir water was 0.62 kg m −3 .The Froude number (Eq.7) of the flow was estimated to be 0.198, and the estimated depth of the plunge point was 10.5 m, which is close to the observed depth, 12 m (Fig. 10).The entrainment coefficient, E, was 0.00031.
Applying h in Eq. ( 6), the greatest depth (hc) that would have been reached by the density current was estimated to be 14.5 m at x = 20 km.This estimated value is very close to that observed from the thermistor chain positioned 20 km downstream (Fig. 12).The estimated result is thus consistent with the observed data.

Conclusions
With the objective of monitoring the plunge-point location in the Manso hydroelectric reservoir, the following conclusions were demonstrated: -The spatial variability of the plunge point, evident in the hydrologic data, illustrates the advantages of synoptic satellite measurements over in situ point measurements alone to detect the river-reservoir transition zone.
-During the wet season, the upward entrainment engendered by the Kelvin-Helmholtz instability was the main mixing mechanism in the reservoir waters.
-During the dry season, when the river and reservoir water temperatures were nearly homogeneous, the river circulation was characterized by inter-overflow.
-The joint use of thermal and visible channel data for the operational monitoring of the plunge point is feasible.The deduced information about the density current from this study could potentially be assimilated into numerical models and hence be of significant interest for environmental and climatological research.

Figure 1 :
Figure 1: Study area and bathymetry of Manso reservoir with contours at 20-m intervals.The grey circles (temperature, fluorescence and photosynthetically available radiation [PAR]) and filled circles (temperature, conductivity, dissolved oxygen, fluorescence and photosynthetically available radiation PAR) mark the sampling and transect stations.Position 13 also indicates the location of the limnolo-meteorological station (SIMA), MAN-13.The locations of the thermistor chain and drifter deployments are indicated by position 8.

Fig. 1 .
Fig. 1.Study area and bathymetry of Manso reservoir with contours at 20-m intervals.The grey circles (temperature, fluorescence and photosynthetically available radiation [PAR]) and filled circles (temperature, conductivity, dissolved oxygen, fluorescence and photosynthetically available radiation PAR) mark the sampling and transect stations.Position 13 also indicates the location of the limnolo-meteorological station (SIMA), MAN-13.The locations of the thermistor chain and drifter deployments are indicated by position 8.
by a factor of two, the speed and height change by 20-25 %.A model of a frictionally balanced gravity current (Stigebrandt, 1987) was used to compare the speed (u) estimated using the model U r = s c 21

Figure 2 :Figure 2 :
Figure 2: Daily means of air temperature, wind speed and precipitation (upper panel) and the daily means of solar short-wave radiation and long-wave fluxes (lower panel) from February to July 2007.

Fig. 2 .
Fig. 2. Daily means of air temperature, wind speed and precipitation (upper panel) and the daily means of solar short-wave radiation and long-wave fluxes (lower panel) from February to July 2007.

23Figure 4 :
Figure 4: Temporal evolution of the thermal structure (upper panel) and the Wedderburn and Lake numbers (lower panel) at Station 13.

Fig. 4 .
Fig. 4. Temporal evolution of the thermal structure (upper panel) and the Wedderburn and Lake numbers (lower panel) at Station 13.

,Figure 4 :
Figure 4: Temporal evolution of the thermal structure (upper panel) and the Wedderburn and Lake numbers (lower panel) at Station 13.

Fig. 5 .Figure 5 :
Fig. 5. Satellite images (in true-color composition) acquired on 26 March (a) and 16 July (c), showing the seasonal plunge point variation (red circles), and the water surface temperature (b, d) estimated from the thermal band (band 6) of Landsat 5 TM.

Figure 6 :
Figure 6: (a) Inflow and outflow, (b) power spectra of the time evolution of the thermocline displacements and buoyancy flux smoothed by three Hanning passes (the 95% confidence interval is indicated), (c) coherence (the 5% level for zero coherence is indicated) and (d) the relative phase for the thermocline and buoyancy flux.

Figure 7 :
Figure 7: Isopleths of the (a) water temperature, (b) density, (c) dissolved oxygen and (d) conductivity in Manso Reservoir during March (wet season).The dots at top of temperature panel correspond to longitudinal transects.

Fig. 7 .
Fig. 7. Isopleths of the (a) water temperature, (b) density, (c) dissolved oxygen and (d) conductivity in Manso Reservoir during March (wet season).The dots at top of temperature panel correspond to longitudinal transects.

Figure 8 :
Figure 8: Temperature, conductivity, and pH vertical profiles at Manso Reservoir measured in the river and at the MAN-13 station during the wet season.

Fig. 8 .
Fig. 8. Temperature, conductivity, and pH vertical profiles at Manso Reservoir measured in the river and at the MAN-13 station during the wet season.

28Figure 10 :
Figure 10: (a) Satellite imagery in a normal composition (RGB-321) showing the area with a chlorophyll bloom (from the left up to the dashed contour).(b) The same satellite imagery with a land mask and enhanced contrast to better discriminate the phytoplankton bloom (16 July 2007).

Fig. 10 .
Fig. 10.(a) Satellite imagery in a normal composition (RGB-321) showing the area with a chlorophyll bloom (from the left up to the dashed contour).(b) The same satellite imagery with a land mask and enhanced contrast to better discriminate the phytoplankton bloom (16 July 2007).

Figure 11 :
Figure 11: Temperature contours from the thermistor-chain data measured at the riverreservoir transition zone (January) (Station 06).Fig.11.Temperature contours from the thermistor-chain data measured at the river-reservoir transition zone (January) (Station 06).

Figure 11 :
Figure 11: Temperature contours from the thermistor-chain data measured at the riverreservoir transition zone (January) (Station 06).

Table 1 .
Mean general characteristics of Manso Reservoir.