The effect of stream shading on the inflow characteristics in a downstream reservoir

In thermally stratified reservoirs, inflows form density currents according to the interplay between inflow temperature and reservoir stratification. The temperature of inflowing water is affected by catchment properties, including shading by riparian vegetation. We hypothesize that the degree of shading in the catchment can affect the inflow dynamics in downstream reservoirs by changing inflow temperature and consequently the nature of the density current. We test it for a subtropical drinking water reservoir by combining catchment‐scale hydrological and stream temperature modeling with observations of reservoir stratification. We analyze the formation of density currents, defined as under, inter and overflow, for scenarios with contrasting shading conditions in the catchment. Inflow temperatures were simulated with the distributed water‐balance model LARSIM‐WT, which integrates heat‐balance and water temperature. River temperature measurements and simulations are in good agreement with a RMSE of 0.58°C. In simulations using the present state of shading, underflows are the most frequent flow path, 63% of the annual period. During the remaining time, river intrusion form interflows. In a scenario without stream shading, average inflow temperature increased by 2.2°C. Thus, interflows were the most frequent flow path (51%), followed by underflows (34%) and overflows (15%). With this change, we would expect a degradation of reservoir water quality, as overflows promote longer periods of anoxia and nutrient loads would be delivered to the photic zone, a potential trigger for algae blooms. This study revealed a potentially important, yet unexplored aspect of catchment management for controlling reservoir water quality.

and water storage for drinking water and irrigation (Grill et al., 2019;Lehner et al., 2011). Dams disrupt natural hydrological, geological and biogeochemical cycles (Friedl & Wüest, 2002;Poff, Olden, Merritt, & Pepin, 2007;Vörösmarty et al., 2003). The consequences of river damming for water quality and biodiversity have been intensively assessed for downstream river reaches and river basins (Bunn & Arthington, 2002;Vörösmarty et al., 2010), but also water quality in the impounded water bodies is of great concern, as it potentially jeopardizes their economic and societal values.
The efficient trapping of nutrients, including phosphorous and nitrogen (Akbarzadeh, Maavara, Slowinski, & Van Cappellen, 2019;Maavara et al., 2015), in combination with prolonged water residence time and the potential development of thermal stratification, promote eutrophication and harmful algae blooms in the impoundments (Winton, Calamita, & Wehrli, 2019). Nutrient enrichment is the primary cause for eutrophication and the occurrence of harmful algae blooms, which nowadays are the main problems related to water quality (Paerl & Otten, 2013;Schindler, 2012;Smith & Schindler, 2009). The possible consequences are the killing of fish due to the depletion of oxygen or the release of toxins from algae and sediments, as well as increased concentration of suspended and dissolved substances that affect odor and color. Such degradation of water quality increases treatment costs (Dodds et al., 2009;Pretty et al., 2003;Walker Jr, 1983).
In addition to climatic and geographic boundary conditions, reservoir water quality is controlled by the inflowing nutrient load and reservoir hydrodynamics. In stratified reservoirs, inflows form density currents that are classified according to their depth of intrusion: underflows follow the reservoir bottom, overflows stay at the reservoir surface, and interflows enter at intermediate depths (Wetzel, 2001). The type of density current depends on inflow temperature and reservoir stratification and eventually controls the distribution of the nutrient load in the reservoir and its availability for algae growth (Ayala, Cortés, Fleenor, & Rueda, 2014;Rueda, Fleenor, & de Vicente, 2007).
Water temperature of the inflowing streams depends on meteorological and hydrological conditions (Caissie, 2006;B. W. Webb, Hannah, Moore, Brown, & Nobilis, 2008). However, stream temperature is also affected by catchment properties, including shading by riparian vegetation. In particular, stream water temperature has been observed to increase following deforestation, or to decrease in response to tree growth in many studies (see reviews in Beschta, Bilby, Brown, Holtby, & Hofstra, 1987; D. R. Moore, Spittlehouse, & Story, 2005). More recently, the impact of riparian vegetation on stream water temperature has been studied using different empirical and modeling approaches. These studies showed, that the magnitude of temperature reduction due to riparian vegetation depends on many different aspects, including vegetation density, vegetation height, stream width, stream orientation, contribution of net shortwave radiation to the overall energy budget, geographical latitude, solar angle and many others (e.g., Dugdale, Malcolm, Kantola, & Hannah, 2018;Garner, Malcolm, Sadler, & Hannah, 2014, 2017Garner, Malcolm, Sadler, Millar, & Hannah, 2015;Kalny et al., 2017 Trimmel et al., 2018). In general, the effect of dense riparian vegetation is most pronounced during times of high water temperature, when the energy budget is dominated by short wave radiation (e.g., Garner et al., 2014;Hannah, Malcolm, Soulsby, & Youngson, 2008). It is thus well established that riparian vegetation helps to reduce maximum stream water temperatures and thermal variability. Consequently, shading is often considered in catchment management and stream restoration efforts and is among the three most important environmental state variables in assessments of stream restoration success (Feld et al., 2011).
The effect of stream shading on inflow dynamics in downstream reservoirs has not been studied. We hypothesize that the degree of shading in the catchment can affect the inflow dynamics in downstream reservoirs by changing inflow temperature and consequently the nature of the density current. As riparian stream shading is closely related to land use in the catchment, this mechanism would constitute an unexplored influence of catchment management on reservoir water quality. Here we test this hypothesis for a tropical drinking water reservoir. We combine catchment-scale hydrological and stream temperature modeling with observations of reservoir stratification and analyze the formation of density currents for scenarios with contrasting shading conditions in the catchment. We further discuss the potential implications of stream shading for reservoir water quality and the broader relevance of the studied process.

| Study site
Passaúna reservoir is a drinking water reservoir in a tropical to subtropical region in south Brazil (latitude: À 25 30 0 and longitude: À 49 22 0 ). It produces around 1.8 m 3 s À1 of drinking water for parts of the city of Curitiba and three neighboring cities (SANEPAR, 2013).
The reservoir has a maximum depth of 15 m, a surface area of 9 km 2 and an approximate volume of 60 Â 10 6 m 3 (Carneiro, Kelderman, & Irvine, 2016). The catchment of the reservoir covers an area of 143 km 2 (Figure 1). Mean air temperature within the catchment was approximately 18.7 C and mean yearly precipitation approximately 1,650 mm (years 2009 through 2018). A relatively high proportion of the catchment (44% for catchment of gauge Campo Largo) was covered by broad-leaved mainly evergreen forest (Figure 1). Passaúna River is the dominant inflow to the reservoir. It drains an area of approx. 100 km 2 and delivers around 75% of total annual inflow to the reservoir (Carneiro et al., 2016). Simulated mean annual discharge at gauge Campo Largo (84 km 2 ) reached approx. 2 m 3 s À1 (2010 through 2018).

| Observational data
The investigation period of the present study covers 1 year from March 2018 through February 2019. During that period, water temperature was monitored in the Passaúna River and reservoir. At Passaúna River, water temperature was measured approximately 3 km upstream of the reservoir inflow near the gauging station Campo Largo using a temperature sensor (miniDOT, Precision Measurement Engineering Inc.) with a temporal resolution of 15 min, an accuracy of ±0.1 C and a resolution of 0.01 C (Figure 1). In the reservoir, a vertical thermistor chain with 11 temperature sensors (Minilog-II-T, Vemco) was deployed close to the intake station of the waterworks (Figure 1), at a mean water depth of 12 m. The chain was fixed at the bottom with the first logger being 1 m above the bed and all remaining sensors were arranged with a fixed vertical spacing of 1 m. The sampling interval was 1 min, precision and accuracy of the sensors was ±0.1 C and of 0.01 C, respectively.
Additional data for modeling discharge and river water temperature was provided by the Federal University of Paran a (UFPR). Measurements of mean daily discharge were available for four gauges within the catchment (Figure 1). These discharge data were used to calibrate and validate the water balance model over a period of several years (see below; Krumm, Haag, & Wolf, 2019). To simulate discharge and river water temperature during the period of investigation (March 2018 through February 2019) we used daily precipitation from four rain gauges within the catchment along with additional measurements of air temperature, global radiation, humidity and wind speed from two meteorological stations west of the catchment (Figure 1).
2.3 | Integrated water balance and stream water temperature modeling 2.3.1 | Model overview River discharge and stream water temperature were modeled using LARSIM-WT (Large Area Runoff SIMulation Model -Water Temperature - Haag & Luce, 2008). LARSIM is a process-oriented and spatially distributed water balance model, which simulates all major aspects of the terrestrial water cycle (LEG, 2019). LARSIM also includes an optional water temperature module (WT), which simulates water temperature throughout the complete river network on a physical basis (Haag & Luce, 2008).
Heat transport within the river network was modeled using the one-dimensional advection-dispersion equation. The local heat balance, that is, the source-sink term in the advection-dispersion equation, accounts for heat exchange with the atmosphere and at the river bed: with WT denoting water temperature ( C), t time (s), cp W the specific heat capacity of water (J kg À1 C À1 ), ρ W water density (kg m À3 ), h average water depth of the river reach (m), R S net shortwave radiation (W m À2 ), R L net longwave radiation (W m À2 ), H S turbulent flux of sensible heat (W m À2 ), H L turbulent flux of latent heat (W m À2 ) and H bed the conductive heat flux at the riverbed (W m À2 ).
In general, the parametrization of the heat fluxes follows the approach of Sinokrot and Stefan (1993) The turbulent fluxes of latent heat H L and sensible heat H s were simulated with an aerodynamic approach: ðWT À T air Þ ð 4Þ with L denoting the latent heat of vaporization (J kg À1 ), e sat,Wt the saturation vapor pressure (hPa) at the water surface with temperature WT, e air the actual measured vapor pressure in the air (hPa), γ the psychometric constant at normal pressure (0.655 hPa C À1 ), P the measured atmospheric pressure (hPa) and T air the measured air temperature ( C). The aerodynamic coefficient for turbulent exchange of water vapor K L (m s À1 hPa À1 ) was derived as a function of measured wind speed v wind (m s À1 ) by the approach of Rimsha and Donschemko (1958), which produces realistic results over a wide range of environmental conditions. Within this formula, we accounted for the effect of wind sheltering by riparian vegetation with a wind shield factor f wind , where f wind = 0 corresponds to no wind sheltering and f wind = 1 corresponds complete wind sheltering: Following commonly applied models in stream water temperature modeling (e.g., Bogan

| Application to the Passaúna catchment
In LARSIM-WT, the catchment of Passaúna reservoir was represented by sub-catchments and their corresponding river reaches ( Figure 1).
The sub-catchments and corresponding river reaches were delineated based on a Digital Elevation Model and a digital river dataset. Forests and riparian vegetation within Passaúna catchment is mainly made up by broad-leaved mostly evergreen trees, resulting in an almost constant mean leaf area index throughout the year. Therefore, with respect to river shading, we did not have to take into account seasonal changes of the leaf area index of riparian vegetation.
Moreover, almost all river reaches in the Passaúna catchment were less than approximately 5 m wide. With respect to river shading, we thus assumed that overhanging canopies of typical riparian trees on both banks may cover the river width completely. Consequently, for dense riparian vegetation we assumed a constant sky view factor that is uniformly distributed throughout the complete hemisphere above the river surface. Thus, we did not take into account additional influencing factors for estimating the shading factor f shade , such as height of trees, width of river reaches, solar angle, stream orientation or the proportion of direct and diffuse global radiation linearly (e.g., Caissie, 2006;Regenauer et al., 2019;Sinokrot & Stefan, 1993).
Exact values of the present state of stream shading along the river reaches were not available, since land use data were not accurate enough to account for small strips of riparian vegetation along the rivers and precise ground mapping of riparian vegetation was not existent. Therefore, we used the proportion of forest within each subcatchment as a proxy for the spatial distribution of the proportion of riparian vegetation. We then introduced a factor to multiply the relative proportion of forest within each sub-catchment to get the relative proportion of riparian vegetation of the river reaches. We optimized this factor to fit the measured stream water temperature with the model, but did not allow f shade to exceed a maximum value of 0.85 in any river reach. Within our approach, we thus assumed that the actual proportion of riparian vegetation was proportional to the proportion of forest within the sub-catchments. We obtained a mean factor of 1.6. Based on observations in the field, this factor seemed reasonable, since there were many riparian vegetation strips outside forested areas, which increased the first estimate of riparian shading. The optimized factor yielded degrees of shading ranging from 55% to 85% (i.e., f shade varied between 0.55 and 0.85), for the present state of stream shading. The wind shield factor was assumed to vary linearly with the shading factor, as described above. Thus, f wind varied between 0.26 and 0.40.
To analyze the potential effect of stream shading on river water temperatures and inflow dynamics into the reservoir, we defined two scenarios: "full shading" and "no shading." In the full shading scenario for all river reaches we used f shade = 0.85 and f wind = 0.40. In the no shading scenario, both parameters were fixed at zero for all river reaches.

| Classification of reservoir inflow dynamics
The inflow regime of the river into the reservoir depends on the difference in water density between river water and the seasonally stratified water in the reservoir. Water samples from the analyzed period had maximum total solid concentration of 0.16 g l À1 (Oliveira et al., 2019) and the density of the river and reservoir water was mainly controlled by the water temperature.
The intrusion depth of density currents formed at the reservoir entrance was assessed by the difference between the temperature of the inflowing river water and the temperatures observed at the reservoir surface and bottom. The inflow was considered as an overflow, where the river water floats on top of the reservoir, when the temperature of the Passaúna River was higher than the measured surface temperature in the reservoir. When the inflow temperature was lower than the temperature measured close to the reservoir bed, the inflow was classified as underflow, where the river flows along the reservoir bottom. For all other periods, the inflow was considered as interflow, similar to the analysis made by Ishikawa, Bleninger, and Lorke (2021) accepted at Inland Waters. This analysis was done based on daily mean temperatures.

| Inflow temperature
Observed river water temperature near the reservoir inflow (location Minidot in Figure 1) varied between 15.3 ± 1.7 C in winter and 21.3 ± 1.2 C in summer. The mean temperature was 18.3 ± 2.9 C, with the lowest temperature (10.9 C) in July and the highest temperature The scenario with full shading (f shade = 0.85 for all river reaches) resulted in a moderate reduction of the river water temperature at the reservoir inflow in comparison to the present state (Figures 2 and   3). The simulated temperature differences ranged between À0.1 C and À 2.0 C with an average of À0.8 C. Owing to the already relatively high degree of shading for the present state (f shade = 0.55 through 0.85), maximum shading only showed moderate effects on river water temperature at the reservoir inflow.
The absence of shading (f shade = 0 for all river reaches) resulted in significantly higher inflow temperature (Figures 2 and 3). Compared to the present state of shading, the increase of daily mean inflow temperature varied between +0.1 and +4.7 C with an average of +2.2 C. This increase is slightly non-linear with higher values in summer, when water temperature was high ( Figure 3). This non-linearity is due to the higher contribution of shortwave radiation to the overall energy balance during summer. The largest differences occurred in December 2018, when high shortwave radiation and low flow situations coincided (Figure 2). The difference in water temperature for the two contrasting scenarios without shading and with full shading ranged between +0.3 and +6.7 C with an average difference of +3.0 C (Figure 2c).

| Reservoir temperature stratification
Water temperature in the reservoir was 21.6 ± 3.5 C at the water surface, and 19.1 ± 2.0 C at the bottom. With persistent temperature differences between the surface and the bottom (1.4 to 7.2 C), the reservoir was stably stratified until the middle of April and from September 2018 (Figure 4). During the stratified period, water surface temperature showed synoptic variability, but with smaller amplitude than stream temperature (Figure 4) August. The seasonal mixing dynamics of the reservoir can therefore be classified as discontinuous warm polymictic (Lewis Jr, 1983). showed that forest clear-cutting leads to an increase of maximum summer water temperatures of headwater streams commonly in the range of +4 C to +9 C (see reviews in Beschta et al., 1987; D. R. Moore et al., 2005) F I G U R E 4 Time series of daily-mean water temperature in Passaúna reservoir and its main inflow: The filled gray area marks the range of temperature measured at the surface and the bottom of the reservoir. Colored lines show simulated temperature of the Passaúna River at the reservoir inflow, blue represents the present state of shading, green represents the full shading scenario and red the no-shading scenario of the river network upstream of the reservoir [Color figure can be viewed at wileyonlinelibrary.com] temperature reductions of less than 1 C (e.g., Broadmeadow, Jones, Langford, Shaw, & Nisbet, 2011;Brown, Cooper, Holden, & Ramchunder, 2010;Crisp, 1997;Dugdale et al., 2018;Stott & Marks, 2000;B. Webb & Crisp, 2006). However, mean summer temperatures may be reduced more effectively by approximately 1-3 C, and maximum reduction of water temperature may be as high as 5 C (Broadmeadow et al., 2011;Brown et al., 2010;Dugdale et al., 2018;B. Webb & Crisp, 2006).  et al., 2005). Also, the difference between the yearly average and the maximum effect in summer is in good accordance with the findings from the United Kingdom (Brown et al., 2010;Dugdale et al., 2018; B. W. Webb et al., 2008). However, our simulated effect of riparian shading on stream water temperature of Passaúna is somewhat higher than the empirical and modeling results from Europe and North America. This may partly be due to differences in residence time. We considered the effect of shading of the complete river network of a relatively large catchment, whereas the other studies mainly looked at small headwater streams with shorter residence times. In the case of river Pinka in Austria only the main river was assumed to be shaded, whereas inflowing tributaries were assumed to have the same temperature in all scenarios. Thus, shorter residence times in other studies may have contributed to the less pronounced effects of shading.

| Reservoir inflow regime
Moreover, we neglected the effect of riparian vegetation on longwave radiation in our study, which increases the energy input at the water surface slightly. This might lead to a small systematic overestimation of the effect of stream shading on water temperatures in our study.
Finally, and probably most importantly, the effect of riparian shading via blocking of incoming shortwave radiation is likely to be considerably higher at a low latitude of À25 (Passaúna) than at the much higher latitudes of the other study sites ($45 to 57 ). Therefore, in summary, our scenario results for the effect of shading on river water temperature are broadly in line with the literature and can be considered as realistic, even though we could not compare them to other findings from the tropics. Considering the fact that the focus of the present study is not on precise predictions, but results are rather used to demonstrate the potential impact of stream shading in a catchment on a downstream reservoir, the accuracy of our shading scenarios appears to be sufficient.

| Stream shading in the catchment affects reservoir mixing and stratification
Vertical mixing in reservoirs can be strongly suppressed by temperature stratification, which develops as a consequence of enhanced heating of the surface layer, or by lateral density currents formed by inflowing rivers. The latter mechanism depends on stream temperature of the inflow and has been well documented in many reservoirs and studied in laboratory experiments (Alavian, Jirka, Denton, Johnson, & Stefan, 1992;Imberger & Hamblin, 1982;Wells & Nadarajah, 2009)

| Implications for reservoir water quality
The observed change in inflow regime for the no shading scenario potentially affects water quality in the reservoir. The occurrence of overflows, which are not present for the present state of shading, facilitates nutrient transport to the photic zone and therewith promotes algae growth (Ayala et al., 2014;Rueda et al., 2007). The combination of excessive nutrient supply at the water surface and reduced vertical mixing during these conditions provide ideal conditions for harmful cyanobacterial blooms (Paerl & Otten, 2013), which are currently not present in Passaúna reservoir. In consequence of the reduced underflows, the transport of oxygen with the inflowing water to greater depths would be reduced, leading to a prolongation of periods of anoxia. Anoxic bottom water further increases internal loading with nutrients from the sediments (Søndergaard, Jensen, & Jeppesen, 2003), as well as the release of anoxic products such as methane, hydrogen sulfide and metals (Beutel & Horne, 1999). In consequence, the water quality in the reservoir can be expected to deteriorate for the scenario without stream shading in the catchment in comparison to the present and full shading conditions.
Given the high degree of forested area in the catchment of Passaúna reservoir (44%), no shading may represent a scenario of extreme land use change at first glance. However, increasing urbanization and agricultural land use in the growing metropolitan area of Curitiba exert a strong anthropogenic pressure on the Passaúna catchment, which may lead to significant deforestation. Furthermore, stream shading does not have an instantaneous effect on water temperature, but rather needs some residence time (i.e., flow distance) to exert its effect (Bartholow, 2000;Kalny et al., 2017;Regenauer et al., 2019). Thus, even if deforestation is restricted to the upstream part of the catchment, it may still lead to increased water temperatures at the downstream inflow of the reservoir.
The strong control of catchment properties on reservoir water quality has been extensively studied in terms of hydrological characteristics and in respect to the input of suspended and dissolved substances, including nutrients and pollutants (Beaver et al., 2014;Jones, Knowlton, & Obrecht, 2008;Knoll et al., 2015). The effect of riparian shading, which is closely linked to land use in the catchment, has not been considered. Our results demonstrate, that changes in stream shading should be included in management scenarios of the catchment area which aim at safeguarding reservoir water quality.

| Limitations of the present study
The good agreement between stream temperature simulations for the present state of shading with observations suggests that the applied integrated water balance and stream water temperature model LARSIM-WT is a robust tool for estimating the effect of shading on stream water temperature. However, by only considering the temperature difference between the inflowing water and reservoir stratification, we applied a rather crude approach for characterizing the inflow conditions. The plunging depth of density currents in reservoirs is known to depend on geometry of the inflow region and volumetric discharge, wind mixing and many other factors, while the plume formed by the density current is subject to dispersion Imberger & Hamblin, 1982). Moreover, the changing inflow will further change reservoir stratification, which was neglected in this study where we used observed reservoir temperature in all scenarios. In contrast, the most relevant change was the increase of overflows, where the inflowing water stays in the upper mixed layer.
Because surface water temperature is mainly driven by air temperature, only minor changes of reservoir stratification are expected in this case. More realistic descriptions of density currents and projected changes in water quality in response to changing inflow temperature requires more detailed hydrodynamic modeling (e.g., Long, Ji, Liu, Yang, & Lorke, 2019;Rueda et al., 2007) and further assumptions on boundary conditions, which would make the results more accurate, but also more case specific.
Our results were obtained for a small reservoir in the tropics, which was chosen for reasons of data and model availability. Although Passaúna reservoir can be considered as being representative for a large number of impoundments in terms of reservoir size, both the effect of riparian shading on stream temperature and the inflow dynamics are affected by many factors, including catchment and reservoir size, water depth and geographic location. For example, in larger rivers (more than approximately 15-30 m wide), shading certainly has limited effect on water temperature, as only a fraction of the water surface is subject to shading by riparian vegetation (e.g., DeWalle, 2008; R. Moore et al., 2014;Regenauer et al., 2019). Given the complex and site-specific conditions of the underlying processes, a more detailed assessment of the relevance of stream shading in the catchment on the inflow regime in reservoirs requires further analysis for a broader range of reservoirs and catchments in future studies.

| CONCLUSIONS
Stream shading is a relevant factor for river temperature, and its alteration can significantly affect reservoirs hydrodynamics and potentially water quality. Deforestation in the catchment and the removal of tall vegetation along riparian zones of streams may lead to increased river water temperatures, as can be predicted robustly by a combined water balance and water temperature model, such as LARSIM-WT. These changes of the water temperature of inflowing rivers can be expected to lead to a degradation of water quality in the downstream reservoir due to changes in reservoir hydrodynamics. Despite rather crude assumptions with respect to the hydrodynamics at the reservoir inflow and site-specific simulations, our findings revealed a so-far overlooked mechanism by which reservoir water quality can be affected and potentially also manipulated by catchment properties and land use management. Given the potential relevance of this process for reservoir water quality, the site-specific effects of riparian shading in the catchment should be considered with more realistic approaches to the hydrodynamics at the inflow and for a broader range of reservoirs.