Abstract

Observations of the turbulent exchange between a river surface and the atmosphere in a mountainous area in southern Brazil are presented and discussed. A micrometeorological tower was installed directly above the surface of a 60-m-wide river. This paper describes the observed turbulent fluxes over 12 days of observations at this site. Eddy correlation sensible and latent heat fluxes are directed toward the river during daytime and from the river at night, and they are controlled by differences between water and air temperatures. The magnitude of the vertical fluxes between the river and the atmosphere increases during daytime with increasing temperature gradient up to a threshold, beyond which the increasing stability starts to dampen the fluxes. Water and air temperatures show very little variations across the width of the river, indicating that the measurements taken at one margin may be representative of the mean river exchange. Local scalar budgets show that daytime warming and moistening rates above the river are controlled by local transport from the riverbanks. The main vertical fluxes have a very small magnitude: 0.8 W m−2 for sensible heat and 1.1 W m−2 for latent heat. Events of very large sensible heat fluxes from the river to the atmosphere and very large latent heat fluxes from the atmosphere to the river happened on 3 days, following nights with a very deep fog layer in the valley. These events represented the passage of a warm and dry air mass down the river. A process to explain the occurrence of these large fluxes is suggested that is associated with differential fog dissipation over the valley.

1. Introduction

The proper determination of surface fluxes over heterogeneous landscapes is currently an important subject of meteorological research (Mahrt 1987; Beyrich et al. 2002; Strunin et al. 2004; Leuning et al. 2004). Often, fluxes are observed over the different surface types within an area, with the purpose of determining the regional average. At the same time, regions of complex topography have also received great interest (Kossman et al. 1998; Whiteman 2000; Weigel and Rotach 2004), especially in terms of the local circulations and their impact on air quality and local microclimate. The purpose of this study is to look at a very specific surface type in a complex terrain environment (the river), and the exchanges between the river surface and the atmosphere.

Rivers have an important role in the atmospheric environment of their surroundings, in many cases. They may induce local circulations (Zhong and Takle 1992; Oliveira and Fitzjarrald 1994; Silva Dias et al. 2004); river basins are responsible for an appreciable input of moisture to the atmosphere, and these fluxes may affect regional flux averages, with important impacts for the surface parameterization of mesoscale and even global numerical models. Water bodies, in general, represent an important type of surface heterogeneity. Sun et al. (1998) showed that a lake is able to anchor the surface fluxes, being the preferential location for scalar exchange between the surface and the atmosphere over a region.

All of these studies refer in general to major rivers, with widths on the order of kilometers. The reason for that is obvious: these are the cases in which the presence of the river has the largest impact to the surface fluxes, as well as to the local circulations or average regional fluxes. Smaller rivers, with a typical width on the order of tens to a few hundred meters have been much less studied. Evans et al. (1998) and Webb and Zhang (2004) estimated sensible and latent heat fluxes over the water–air interface on a small stream, with the purpose of understanding the water energy budget and determining the factors that control water temperature. Because their main objective was not directly related to the atmospheric properties, the fluxes were not directly measured, but were estimated using the Penman–Monteith and Bowen ratio techniques. Papers that focus on the atmospheric processes over these environments are usually not directly interested in the transfers between the river surface and the atmosphere but, rather, in the main circulations that are developed in the river valleys (Whiteman 1982; Weigel and Rotach 2004) or in the surface heterogeneity and its impact on average regional fluxes (Strunin et al. 2004). This is not unexpected, because the river-to-atmosphere exchange, in this case, represents only a small portion of the total regional exchange. Moreover, this exchange may very likely be a consequence, rather than a cause, of the main circulations observed at the region, which are, in general, topographically driven.

Nevertheless, the proper characterization of the surface-to-atmosphere exchange over the river surface represents an important scientific question. It has implications on water temperature, as studied by Evans et al. (1998) and Webb and Zhang (2004). Another direct application of this knowledge regards the characterization of the microclimate in a narrow valley dominated by a river. Local circulations and mean atmospheric properties, such as temperature and relative humidity, may be affected by the presence of the river, and the extent of this dependence needs to be addressed.

In this study, we show observations of the turbulent exchange of heat and moisture between a small river within a narrow valley and the atmosphere, at a site in southern Brazil (Fig. 1). The paper is structured as follows: in section 2, the region, experiment sites, and measurements taken during the campaign are presented. In section 3, the surface fluxes are discussed. The local surface budget and its relationship to the vertical fluxes are analyzed in section 4, and in section 5 nonlocal events that are responsible for large transfers of heat and moisture between the river and the atmosphere are examined. Last, in section 6, the conclusions are presented.

Fig. 1.

Topographical map of the region, with the heights given by the legend. The inset shows the location of the river tower, indicated by the dot and triangle, in detail.

Fig. 1.

Topographical map of the region, with the heights given by the legend. The inset shows the location of the river tower, indicated by the dot and triangle, in detail.

2. Description of the site and measurements

The observations described in this study were conducted near the town of Nova Roma do Sul, in the mountainous region of Serra Geral, in the southern portion of Brazil (Fig. 1). Typical topographical relief ranges from 400 to 600 m MSL. Two main rivers run in the area, the 80-m-wide Antas and the 60-m-wide Prata (passing by the river tower station in Fig. 1). The original vegetation is forest (Mata Atlântica), which can still be found on most of the slopes. The higher regions are covered by agricultural fields: typically vineyards and corn and oats crops.

Three intensive observational campaigns have been held in the region. In the first one, in November–December 2001, a single micrometeorological tower was located near the edge of a 450-m-high hill, next to a very steep slope. The main results of this campaign have been discussed by Acevedo et al. (2002). The second intensive campaign occurred from June to July 2002. In this case, an extra micrometeorological tower was installed at the bottom end of the slope.

This study focuses entirely on the third observational campaign, held from 12 May to 9 June 2005. This time, a flux tower was assembled directly above the river surface. The tower was assembled at a rock adjoining the riverbank at the southern margin of the 60-m-wide Prata River at a location where it runs from west to east (Fig. 1, inlet). At the tower location, the valley has a width of 150–200 m. Average river depth is 8 m, and the water typically flows at a speed of 1.5 m s−1.

The instruments were located above the water surface. In calm conditions, with a normal river level, the fast-response observations (wind and temperature taken by a Campbell Scientific, Inc., CSAT 3D sonic anemometer; water vapor concentrations taken by a Licor, Inc., ICOR 7500 gas analyzer) were located at 4 m from the river surface. Slow-response measurements included air temperature and specific humidity at 4 and 6 m, wind speed and direction at 6 m, and shortwave incident solar radiation. The observations started 12 May 2005 and continued normally until 18 May 2005, when a mesoscale convective system caused intensive rains and a sudden elevation of the river level above the tower base. Observations restarted on 28 May 2005 and continued until 6 June 2005. Water temperature was measured from 30 May 2005 to 6 June 2005, and the sensor was located at a depth of 10 cm, attached to a buoy at 1 m from the rock where the tower was assembled. Given that the period of observations was short and there were instrumental failures, a total of 12 complete days of good-quality flux data were collected. The results discussed in this paper refer to those days.

On 3 June 2005, cross-river transects were taken from 0830 to 1630 LST, measuring air temperature and specific humidity at 4 and 6 m, wind speed and direction at 6 m, and water temperature. The transects were performed 100 m down the river from the flux tower. The purpose of these observations was to determine the representativeness of the flux measurements taken at one margin (see section 4). The instruments were assembled on a manually operated ferry (10 m long and 4 m wide), which is used to transport vehicles across the river. A total of 65 transects were performed during the day.

A careful examination is required for measurements of turbulent fluxes at such a heterogeneous environment. The only direction from which there is a reasonably good fetch is the westerly one, from which the river extends for near 1 km until it curves. Prevailing winds at all times are northwesterlies (Fig. 2). This overall flow comprises flow down the river (westerly component) and flow from the river to the bank (northerly component). Fetch requirements may be violated at times. This fact, together with the local surface heterogeneity, may lead to appreciable flux-sampling errors and nonstationary conditions at the site. Therefore, we applied the criteria suggested by Vickers and Mahrt (1997) to estimate systematic and random errors of the flux measurements, as well as the nonstationarity of the horizontal wind. The turbulent fluxes were determined for 5-min time intervals by the covariance between the wind component in the flux direction and the scalar series. These values were subsequently averaged over 1-h periods. No linear detrending or recursive filters were applied to the data.

Fig. 2.

Mean wind vector by hour of the day. Numbers next to arrowheads indicate the hour.

Fig. 2.

Mean wind vector by hour of the day. Numbers next to arrowheads indicate the hour.

By following the method of Vickers and Mahrt (1997), for each record we determined 1) the relative systematic error (RSE), which is an estimate of the portion of the total flux captured within a given time interval applied for flux determination, 2) the random flux error (RFE), which is an estimate of the random flux variability within a given 1-h period, and 3) the relative nonstationarity (RN) of the flux, which is a measure of the flux trend within the 1-h record.

We assume that a given record is contaminated when RSE, RFE, or RN exceeded 0.25. The portion of the records affected by the sampling problems is, on average, large (Table 1). This situation is to be expected, given the large local surface heterogeneity and also the fact that average fluxes have a small magnitude (section 3). RSE is especially large, affecting the majority of records of any flux. This situation is also expected because, for comparison, Vickers and Mahrt (1997) found that RSE or RN exceeded 0.25 for 100% of their flux records over a lake. The nonstationarity of the horizontal wind was also evaluated: 67% of the total records were considered to be nonstationary. One of the reasons for this large number is the very weak mean wind magnitude at the site. When only those records for which the average wind magnitude is larger than 1 m s−1 are chosen, the proportion of nonstationary records drops appreciably to 26%.

Table 1.

Percentage of records with each flux error.

Percentage of records with each flux error.
Percentage of records with each flux error.

In most of the analysis described in the subsequent sections, only those records for which no systematic flux errors were found were considered. This is important for the determination of the average fluxes, for which transient events should be neglected. The mean transfer is not significantly affected, however, by this record selection. In section 5, the analysis focuses on localized events of large scalar transport both along the river and vertically, which are suggested to be caused by the motion of dry and warm air masses along the valley. For these temporally localized events, the condition of only considering records without systematic flux errors has not been applied.

3. Surface fluxes

Two distinct processes were found to control the turbulent fluxes of latent and sensible heat fluxes in the region. In the first case, the fluxes are controlled by the water–air vertical gradients of temperature and specific humidity, and we refer to them as surface fluxes. The second process happens when there are air masses with substantially different characteristics than those existent at the site flow down the river. Even though this process is not frequent, it may induce very large fluxes and, therefore, it may be very important to the net exchange between the river and the atmosphere. These events will be examined in detail in section 5.

A typical evolution of the vertical surface fluxes shows that both sensible and latent heat fluxes are positive (from the river to the atmosphere) at night and negative during the day (Fig. 3). The flux directions are controlled by the temperature and specific humidity differences between the water and the air, and they change sign when the air at the sensor level becomes warmer and moister than the air just above the water surface. This process happens about 3 h after local sunrise (at 0700 LST), when there is the dissipation of the deep fog layer that forms at nighttime. The fog dissipation is evident from the shortwave radiation evolution (Fig. 3, upper-right panel).

Fig. 3.

(top left) Temporal evolution of air temperature (solid line), water temperature (dotted line), and sensible heat flux (circles) for 3 Jun 2005 at Nova Roma do Sul; (top right) incoming shortwave radiation; (bottom left) specific humidity (solid line), saturation specific humidity at water temperature (dotted line), and latent heat flux (circles); (bottom right) wind magnitude (solid line) and direction (arrows).

Fig. 3.

(top left) Temporal evolution of air temperature (solid line), water temperature (dotted line), and sensible heat flux (circles) for 3 Jun 2005 at Nova Roma do Sul; (top right) incoming shortwave radiation; (bottom left) specific humidity (solid line), saturation specific humidity at water temperature (dotted line), and latent heat flux (circles); (bottom right) wind magnitude (solid line) and direction (arrows).

Negative daytime sensible heat fluxes have been observed over lakes by Heikinheimo et al. (1999) and Batchvarova et al. (2001) and over a paddy field by Hiyama et al. (1995). Strunin et al. (2004) observed that both sensible and latent heat fluxes were downward at daytime over the Lena River when weak wind conditions prevailed. These results suggest that the river may be a sink of both heat and humidity, as Evans et al. (1998) estimated with simpler measurements. Furthermore, they indicate that the fluxes over the river surface are nonlocal, determined by the exchange between the water surface and the air, which in turn has its characteristics determined by the local exchange at the slopes and by the valley circulations.

In general, positive sensible heat fluxes happen when the water is warmer than the air and negative fluxes occur when the air is warmer than the water (Fig. 4). For positive fluxes, larger values occur as the gradient increases. In the case of negative fluxes, they start increasing in magnitude as the temperature difference increases, but the fluxes are damped when the temperature difference between the air and the water becomes larger than 2°C (which means a vertical temperature gradient larger than 0.3°C m−1). This flux decrease is expected, because large stability exists in these cases, removing energy from the turbulent motion. Similar results have been observed by Mahrt et al. (1998) and Moraes et al. (2004), who showed that the magnitude of nighttime sensible heat fluxes increases with increasing stability until the stability parameter z/L reaches a threshold of 0.1, decreasing thereafter. An important distinction between those results and the ones presented here, however, is that over the river the damped fluxes occur when air temperature is maximum, that is, in the middle of the day.

Fig. 4.

(top) Mean sensible heat fluxes, as a function of air-to-water temperature difference, and (bottom) mean latent heat fluxes as a function of the difference between air specific humidity and the saturation specific humidity at water temperature. In both panels, vertical lines indicate the standard deviations.

Fig. 4.

(top) Mean sensible heat fluxes, as a function of air-to-water temperature difference, and (bottom) mean latent heat fluxes as a function of the difference between air specific humidity and the saturation specific humidity at water temperature. In both panels, vertical lines indicate the standard deviations.

The overall behavior for the eddy correlation latent heat fluxes is similar to that for sensible heat fluxes, and the reason is that the difference between air specific humidity and saturation specific humidity at water temperature follows the difference between air temperature and water temperature. Therefore, the magnitudes of negative latent heat fluxes also decrease for larger gradients. An important difference regarding the latent heat fluxes is that they do not change sign when the specific humidity vertical gradient crosses zero, but do so, rather, at a negative value. This result suggests that the proper value to be used at the lower level is not exactly the saturation specific humidity at water temperature, but some smaller value.

An important question related to the flux observations discussed here is about the degree to which measurements taken near one riverbank represent the mean exchange between the river surface and the atmosphere. To answer that question, across-river transects were performed 100 m down the river from the flux tower during one entire day, measuring two levels of air temperature and the water temperature. This implies the assumption that the flux–gradient relationship observed at the tower site is preserved above the river. The day chosen had clear skies for most of the time after the fog dissipated. Results show that the maximum difference in air temperature between the two margins occurred just after the fog layer dissipated, around 1040 LST (Fig. 5a). At this time, the sun directly hits the southern margin, but the air temperature at this side never got more than 0.4°C warmer than at the northern margin. As the day proceeds, the temperature gradient across the river smoothed out, probably as a consequence of some mixing activity that existed despite the stable stratification above the river. A remarkable feature is a reversal of the across-river temperature gradient from 1200 to 1300 LST, when the southern margin became colder than the northern one.

Fig. 5.

(a) The 6-m air temperature, shown as the perturbation to transect (across the river) average, as a function of time and distance across the river. The thick solid line represents the 0°C contour. Dashed lines are negative contours, and solid lines are positive contours. The contour interval is 0.05°C. The shading scale is indicated at the right side. (b) As in (a), but for water temperature. The contour interval is 0.015°C.

Fig. 5.

(a) The 6-m air temperature, shown as the perturbation to transect (across the river) average, as a function of time and distance across the river. The thick solid line represents the 0°C contour. Dashed lines are negative contours, and solid lines are positive contours. The contour interval is 0.05°C. The shading scale is indicated at the right side. (b) As in (a), but for water temperature. The contour interval is 0.015°C.

Water temperature, on the other hand, shows simpler behavior (Fig. 5b). The across-river water temperature gradients also start to occur after fog dissipation, but the maximum occurs near 1230 LST, reaching a temperature difference between the margins of 0.1°C. After that, this value decreases, until the river temperature becomes homogeneous, when its surface is entirely shaded. Even though interesting and clear patterns were noticeable on the across-river temperature evolution, it is clear that there are no large differences of the vertical gradients as one moves from one margin to the other. The maximum difference occurs from 1200 to 1300 LST, when the difference between the 6-m air temperature and the water temperature is 0.2°C larger at the southern margin than at the northern one. According to the results presented in Fig. 4, such a difference in temperature vertical gradient is expected to have a very small impact on surface fluxes. Therefore, we believe that the observations at one margin are an acceptable approximation to the mean fluxes between the river and the atmosphere.

4. Local scalar budgets

The budget equation for a scalar S above the river surface can be written as

 
formula

Here, we are assuming the y axis is directed northward, that is, it is perpendicular to the river, and fluxes from land are assumed to be positive. Therefore, the x-axis points to the east and is positive for fluxes down the river. In Eq. (1), the first term in the RHS represents all forms of transport resulting from the mean components of the wind, the second term accounts for the horizontal turbulent exchange between the air above the river and the margins, and the third term is the vertical turbulent flux between the river and the atmosphere. The turbulent flux divergence along the river (∂/∂x) has been neglected by assuming that there is horizontal homogeneity in that direction. The last term in the RHS refers to local sinks or sources of the scalar, such as radiative flux divergence, for temperature.

The mean vertical fluxes have two important characteristics (Fig. 6, upper panel)—they have a small magnitude and they are reversed from most of the surface types, being negative during the day and positive at night. It indicates that the river is a small sink of both heat and moisture during daytime and that, therefore, the air just next to the river surface is colder and has a smaller moisture content than the air above it. This result suggests that nonlocal processes control the local temperature and specific humidity budget. In fact, the mean turbulent exchange between the margins and the air above the river [second term in the RHS of Eq. (1)] confirms a strong convergence of heat and moisture above the river (Fig. 3, lower panel). Around noon, the mean temperature and moisture lateral fluxes reach values around 10 W m−2. These are appreciable values if one considers that they are converging over the width of a narrow river. Therefore, assuming similar fluxes of opposite sign occur at the other margin, the horizontal turbulence flux convergence alone would lead to a heating rate near 1.2°C h−1 and to a moistening rate of 0.5 g kg−1 h−1. These values represent roughly one-half of the average heating and moistening rate in this period. The remaining one-half is probably performed by advective transport by the mean winds [first term in the RHS of Eq. (1)]. A large peak of horizontal turbulent temperature flux is observed later in the afternoon, from 1300 to 1800 LST, a time during which the cooling rate is largely reduced. We hypothesize that these larger fluxes may be due to strong heating of the north-facing slope in the afternoon. Advective transport from the margins should have the same sign and is unlikely to compensate for the fact that not enough heating occurs in the period. Turbulent flux divergence and advective transport along the river may be responsible for the difference.

Fig. 6.

Daily evolution of the turbulent temperature (solid lines) and moisture (dot–dashed lines) fluxes, averaged for the 12 days of observations.

Fig. 6.

Daily evolution of the turbulent temperature (solid lines) and moisture (dot–dashed lines) fluxes, averaged for the 12 days of observations.

5. Airmass motion along the river

Fog occurs every night at the bottom of the valley. However, the depth of the fog layer varies from night to night, and this situation is confirmed by the fact that saturation only occurs on some occasions at a nearby weather station located 400 m above the river surface. Fitzjarrald and Lala (1989) showed that the fog depth depends on surface conditions at sunset and the nighttime evolution of the valley winds. In moist environments, the depth of the fog layer can often be used as an approximation to the thickness of the surface inversion (Whiteman 2000, p. 176). Inversion breakup, therefore, can be assumed to occur when the fog layer dissipates, and it is often accompanied by a subtle decrease in surface specific humidity at the valley bottom (indicated by arrows in Fig. 7). The specific humidity decrease is a consequence of the mixing between the inversion air mass with the drier air located above. Similar processes have been reported by LeMone et al. (2002) and by Whiteman (1989) and Prevot et al. (2000), who looked at air pollution records. Acevedo and Fitzjarrald (2001) discussed the opposite effect, a moisture jump in the early evening, when the mixing layer depth collapses as the surface inversion develops. In all occasions when the humidity drop occurred (13 May, 17 May, 1 June, 2, June and 4 June), the fog layer reached the 400-m site. On the other hand, on the nights of 14 May, 28 May, 31 May, and 3 June, no saturation could be observed at the upper weather station, and on neither of these occasions did the humidity drop occur after fog dissipation (no observations were available in the 400-m site in the nights of 29 May, 30 May, and 5 June). This pattern can be related to the fact that when the fog layer is deeper there is a larger contrast between the dry air mass above it and that inside the valley. This contrast leads to the humidity drop when mixing takes place.

Fig. 7.

Specific humidity (g kg−1) evolution for the 12 days studied, at the river site. Arrows indicate the specific humidity drop discussed in the text.

Fig. 7.

Specific humidity (g kg−1) evolution for the 12 days studied, at the river site. Arrows indicate the specific humidity drop discussed in the text.

An important distinction exists among the 5 days on which there was the humidity drop, and it is related to the time of its occurrence. On 13 and 17 May, the drop happened shortly after the total dissipation of the fog layer (as indicated by shortwave radiation records), whereas on 1, 2, and 4 June it happened as much as 2 h after the inversion breakup.

The late occurrence of this transition is a consequence of the motion of a warm and dry air mass down the river, rather than from above. Vertical turbulent fluxes show the arrival of such air masses as strong events, with positive sensible heat flux and negative latent heat flux peaks that can reach hundreds of watts per meter squared (Fig. 8), which is an order of magnitude higher than typical observed fluxes. Horizontal turbulent fluxes support this conclusion, showing fluxes as large as 1000 W m−2 (Fig. 9), indicating that the characteristics of such an air mass are largely distinct from the local atmosphere. Along-river fluxes show a positive peak of temperature flux and negative peak of specific humidity flux. Because the along-river wind is assumed to be positive when it blows from west to east (down the river), and this is the direction of the mean winds at the period, such events represent the transport of warm and dry air down the river. Across-river flux peaks are not as large but indicate a warm and dry transport from the river to the margin at the sensor location (from north to south, that is, a negative across-river wind component), showing that as the air mass flows down the river it escapes across the margins, affecting the local environment.

Fig. 8.

Sensible (circles) and latent (triangles) heat fluxes for three days at the river site.

Fig. 8.

Sensible (circles) and latent (triangles) heat fluxes for three days at the river site.

Fig. 9.

Same as Fig. 8, but for horizontal fluxes. Fluxes along the river are assumed to be positive when they indicate transport down the river (from west to east). Fluxes across the river are positive when they indicate transport from the banks to the river (from south to north).

Fig. 9.

Same as Fig. 8, but for horizontal fluxes. Fluxes along the river are assumed to be positive when they indicate transport down the river (from west to east). Fluxes across the river are positive when they indicate transport from the banks to the river (from south to north).

The proper characterization of this warm and dry air mass that flows down the river on days that follow the existence of a deep nocturnal fog layer is still unclear. We suggest that it is related to differential dissipation of the fog layer within the valley. A sequence of satellite pictures from the morning of 26 May 2005 (Fig. 10) shows that fog dissipation is nonuniform along the valley and that there is a tendency for fog to persist longer in those places where the valley has a north–south orientation. Therefore, the following mechanism is suggested: Fog forms at the bottom of the valley at nighttime, as a consequence of cold-air pooling and the local moisture content. On clear nights, longwave radiative loss leads to the formation of a thick fog layer that completely fills the valley. As the sun rises, the fog layer starts to dissipate. Valley portions with a longer extension in the east–west orientation are subject to less shading and receive a larger amount of radiation, leading to a quicker complete fog dissipation. A local convective boundary layer starts to develop at those locations, causing the air to warm up as a consequence of local sensible heat flux and downward entrainment and to dry as a consequence of the entrainment of the air that was originally located above the inversion. When fog dissipates in the entire valley, the local air at those locations where dissipation occurred first is distinctively warmer and drier than in the other portions. Down-valley circulations, which are typical in the region (see section 3), cause the transport of this air mass down the river.

Fig. 10.

Sequence of Geostationary Operational Environmental Satellite images for 26 May 2005. An enhanced composite of bands 1, 2, and 4 was utilized. Solid lines indicate the rivers.

Fig. 10.

Sequence of Geostationary Operational Environmental Satellite images for 26 May 2005. An enhanced composite of bands 1, 2, and 4 was utilized. Solid lines indicate the rivers.

These strong events have an appreciable impact on the mean vertical fluxes observed at the river site. It is important, however, to characterize the origin of these fluxes properly. In the case of the suggested mechanism, the properties of the air that is transferred from the river to the atmosphere during these events are not originated from the physical interactions at the river surface but, rather, from a downward transport from the top of the inversion and then down the river. These processes happen at a larger scale that is determined by the valley geometry. These events should, therefore, not be accounted for when estimating the exchange between the river and the surface on a local scale.

6. Conclusions

On average, at a local scale, the river represents a very small source of moisture and heat to the environment. Mean sensible heat flux during the period (neglecting the fluxes that occurred in the strong events described in section 5) was 0.8 W m−2, and the mean latent heat flux was 1.1 W m−2.

It is crucial, therefore, to understand under what conditions the transferences between the river surface and the atmosphere remain small. A number of connected physical processes happen simultaneously, so that a linear extrapolation of these results may lead to wrong conclusions. The analysis presented here showed that the local exchange above the river surface is highly nonlocal. The negative fluxes during daytime above the river indicate that positive latent and sensible heat fluxes on the slopes are able to cause the air above the river to be warmer and moister than at the river surface. It is reasonable to suppose that such behavior only holds in the cases of a narrow valley or a small enough water body. Therefore, river elevation above a threshold for which this process no longer happens tends to have a stronger impact on the local climate. For that purpose, modeling studies are important to test on which conditions of river size, valley geometry, and large-scale weather the small negative surface fluxes above the river still persist during daytime. Seasonality may also be important, as has been shown by Webb and Zhang (2004), who estimated the mean sensible heat flux to be positive only from May to November. In that case, campaigns in other periods of the year are necessary.

Another important finding of this study is about the breakup of the valley nighttime temperature inversion and fog dissipation. It has been shown that on some occasions this breakup is not uniform throughout the valley. Therefore, in this case, the warm and dry air that entrains from above is distributed along the valley as intense turbulent events. Such occurrences may be very important to air-quality issues and for the proper characterization of the valley microclimate. More detailed field campaigns are necessary to characterize this event and its impacts clearly. In that case, measurements along the river would be important to show the evolution of this airmass motion along its path.

The river surface represents a unique environment. The exchange between the river and the atmosphere had never been measured before, partially because of technical difficulties and partially because the river represents, in general, a small surface type with a reduced impact on average regional fluxes that affect mesoscale and weather forecast regional models. The issue of climate modification resulting from river flooding, however, is an important one and needs to be solved. We believe this study is an initial contribution for that purpose.

Acknowledgments

This work was supported by Brazilian Research Agency Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) under PAMPA project Research Grant 471584/2003-7 and by Companhia Energética do Rio das Antas (CERAN). The Comissão de Aperfeiçoamento de Pessoal de Ensino Superior (CAPES) supports graduate students involved in the project. We also thank Ralf Staebler for reading the manuscript and two anonymous reviewers for their valuable suggestions.

REFERENCES

REFERENCES
Acevedo
,
O. C.
, and
D. R.
Fitzjarrald
,
2001
:
The early evening surface layer transition: Temporal and spatial variability.
J. Atmos. Sci.
,
58
,
2650
2667
.
Acevedo
,
O. C.
,
O. L. L.
Moraes
, and
R.
Silva
,
2002
:
Turbulence observations at the edge of a cliff. Proc. 15th Symp. on Boundary Layers and Turbulence, Wageningen, Netherlands, Amer. Meteor. Soc., 592–595
.
Batchvarova
,
E.
,
S-E.
Gryning
, and
C. B.
Hasager
,
2001
:
Regional fluxes of momentum and sensible heat over a sub-Arctic landscape during late summer.
Bound.-Layer Meteor.
,
99
,
489
507
.
Beyrich
,
F.
, and
Coauthors
,
2002
:
Experimental determination of turbulent fluxes over the heterogeneous LITFASS area: Selected results from the LITFASS-98 experiment.
Theor. Appl. Climatol.
,
73
,
19
34
.
Evans
,
E. C.
,
G. R.
McGregor
, and
G. E.
Petts
,
1998
:
River energy budgets with special reference to river bed processes.
Hydrol. Process.
,
12
,
575
595
.
Fitzjarrald
,
D. R.
, and
G. G.
Lala
,
1989
:
Hudson Valley fog environments.
J. Appl. Meteor.
,
28
,
1303
1328
.
Heikinheimo
,
M.
,
M.
Kangas
,
T.
Tourula
,
A.
Venäläinen
, and
S.
Tattari
,
1999
:
Momentum and heat fluxes over Lakes Tämnaren and Raksjö determined by the bulk aerodynamic and eddy-correlation methods.
Agric. For. Meteor.
,
98–99
,
521
534
.
Hiyama
,
T.
,
M.
Sugita
, and
I.
Kayane
,
1995
:
Variability of surface fluxes within a complex area observed during TABLE 92.
Agric. For. Meteor.
,
73
,
189
207
.
Kossman
,
M.
,
R.
Vögtlin
,
U.
Corsmeier
,
B.
Vogel
,
F.
Fiedler
,
H-J.
Binder
,
N.
Kalthoff
, and
F.
Beyrich
,
1998
:
Aspects of the convective boundary layer structure over complex terrain.
Atmos. Environ.
,
32
,
1323
1348
.
LeMone
,
M. A.
, and
Coauthors
,
2002
:
CASES 97: Late-morning warming and moistening of the convective boundary layer over the Walnut River watershed.
Bound.-Layer Meteor.
,
104
,
1
52
.
Leuning
,
R.
, and
Coauthors
,
2004
:
Spatial and temporal variations in fluxes of energy, water vapour and carbon dioxide during OASIS 1994 and 1995.
Bound.-Layer Meteor.
,
110
,
13
38
.
Mahrt
,
L.
,
1987
:
Grid-averaged surface fluxes.
Mon. Wea. Rev.
,
115
,
1550
1560
.
Mahrt
,
L.
,
J.
Sun
,
W.
Blumen
,
T.
Delany
, and
S.
Oncley
,
1998
:
Nocturnal boundary-layer regimes.
Bound.-Layer Meteor.
,
88
,
255
278
.
Moraes
,
O. L. L.
,
O. C.
Acevedo
,
R.
Silva
,
R. O.
Magnago
, and
A. C.
Siqueira
,
2004
:
Nocturnal surface-layer characteristics at the bottom of a valley.
Bound.-Layer Meteor.
,
112
,
159
177
.
Oliveira
,
A. P.
, and
D. R.
Fitzjarrald
,
1994
:
The Amazon River breeze and the local boundary layer: I. Observations.
Bound.-Layer Meteor.
,
63
,
141
162
.
Prevot
,
A. S. H.
,
J.
Dommen
,
M.
Bäumle
, and
M.
Furger
,
2000
:
Diurnal variations of volatile organic compounds and local circulation systems in an Alpine valley.
Atmos. Environ.
,
34
,
1413
1423
.
Silva Dias
,
M. A. F.
,
P. L.
Silva Dias
,
M.
Longo
,
D. R.
Fitzjarrald
, and
A. S.
Denning
,
2004
:
River breeze circulation in eastern Amazonia: Observations and modelling results.
Theor. Appl. Climatol.
,
78
,
111
121
.
Strunin
,
M. A.
,
T.
Hiyama
,
J.
Asanuma
, and
T.
Ohata
,
2004
:
Aircraft observations of the development of thermal internal boundary layers and scaling of the convective boundary layer over non-homogeneous land surfaces.
Bound.-Layer Meteor.
,
111
,
491
522
.
Sun
,
J.
,
R.
Desjardins
,
L.
Mahrt
, and
I.
MacPherson
,
1998
:
Transport of carbon dioxide, water vapor and ozone by turbulence and local circulations.
J. Geophys. Res.
,
103
,
25873
25885
.
Vickers
,
D.
, and
L.
Mahrt
,
1997
:
Quality control and flux sampling problems for tower and aircraft data.
J. Atmos. Oceanic Technol.
,
14
,
512
526
.
Webb
,
B. W.
, and
Y.
Zhang
,
2004
:
Intra-annual variability in the non-advective heat energy budget of Devon streams and rivers.
Hydrol. Process.
,
18
,
2117
2146
.
Weigel
,
A. P.
, and
M. W.
Rotach
,
2004
:
Flow structure and turbulence characteristics of the daytime atmosphere in a steep and narrow Alpine valley.
Quart. J. Roy. Meteor. Soc.
,
130
,
2605
2627
.
Whiteman
,
C. D.
,
1982
:
Breakup of temperature inversions in deep mountain valleys: Part I. Observations.
J. Appl. Meteor.
,
21
,
270
289
.
Whiteman
,
C. D.
,
1989
:
Morning transition tracer experiments in a deep narrow valley.
J. Appl. Meteor.
,
28
,
626
635
.
Whiteman
,
C. D.
,
2000
:
Mountain Meteorology: Fundamentals and Applications.
Oxford University Press, 355 pp
.
Zhong
,
S.
, and
E. S.
Takle
,
1992
:
An observational study of sea- and land-breeze circulation in an area of complex coastal heating.
J. Appl. Meteor.
,
31
,
1426
1438
.

Footnotes

Corresponding author address: Otávio C. Acevedo, Departamento de Física, Universidade Federal de Santa Maria, Santa Maria, RS 97105-900, Brazil. Email: otavio@smail.ufsm.br