The role of riparian vegetation density, channel orientation and water velocity in determining river temperature dynamics

Abstract A simulation experiment was used to understand the importance of riparian vegetation density, channel orientation and flow velocity for stream energy budgets and river temperature dynamics. Water temperature and meteorological observations were obtained in addition to hemispherical photographs along a ∼1 km reach of the Girnock Burn, a tributary of the Aberdeenshire Dee, Scotland. Data from nine hemispherical images (representing different uniform canopy density scenarios) were used to parameterise a deterministic net radiation model and simulate radiative fluxes. For each vegetation scenario, the effects of eight channel orientations were investigated by changing the position of north at 45° intervals in each hemispheric image. Simulated radiative fluxes and observed turbulent fluxes drove a high-resolution water temperature model of the reach. Simulations were performed under low and high water velocity scenarios. Both velocity scenarios yielded decreases in mean (≥1.6 °C) and maximum (≥3.0 °C) temperature as canopy density increased. Slow-flowing water resided longer within the reach, which enhanced heat accumulation and dissipation, and drove higher maximum and lower minimum temperatures. Intermediate levels of shade produced highly variable energy flux and water temperature dynamics depending on the channel orientation and thus the time of day when the channel was shaded. We demonstrate that in many reaches relatively sparse but strategically located vegetation could produce substantial reductions in maximum temperature and suggest that these criteria are used to inform future river management.


INTRODUCTION
It is anticipated that a changing climate will alter river temperature regimes. Elevated temperatures relative to historical baselines are expected for most watercourses [e.g. Beechie et al., 2013;van Vliet et al., 2013;MacDonald et al., 2014a;Hannah and Garner, 2015]. Such changes, particularly increased maxima, may diminish the spatial and temporal extent of suitable cool-water habitat for temperature sensitive organisms with potential impacts on the composition and productivity of aquatic ecosystems [Wilby et al., 2010;Leach et al., 2012]. Consequently, there is substantial interest in adaptation strategies that may ameliorate the effects of climate warming, including: riparian planting [e.g. Hannah et al., 2008;Brown et al., 2010;Imholt et al., 2013;Ryan et al., 2013;Garner et al., 2014], reconnecting rivers to their floodplains [e.g. Poole et al., 2008;Opperman et al., 2010], restoring or enhancing hyporheic exchange [Beechie et al., 2013;Kurylyk et al., 2014], reducing and retaining urban runoff [e.g. Booth and Leavitt, 1999] and reducing rates of water abstraction [Poole and Berman, 2001]. However in upland streams, where catchment hydrology and geomorphology have not been altered significantly by human activities, fewer of these strategies may be implemented to protect aquatic ecosystems from thermal extremes [Beschta, 1997;Poole and Berman, 2001]. Observational datasets, frequently in combination with deterministic modelling approaches, have demonstrated that the summer temperature of headwater streams is generally dominated by: (1) advected heat from upstream (2) heat exchange at the air-water column interface [e.g. Westhoff et al., 2011;Leach and Moore, 2014;MacDonald et al., 2014a;Garner et al., 2014], predominantly solar radiation gains [Hannah et al., 2008;Leach and Moore, 2010;MacDonald et al., 2014a], and at some locations (3) groundwater inflows [e.g. Westhoff et al., 2007]. Recognising the important role of energy exchange between the atmosphere and the water column and in response to the increasing scientific literature, river managers (e.g. The River Dee Trust; Upper Dee riparian scheme) are increasingly advocating the use of riparian vegetation to reduce total energy inputs to the water column, and thus thermal variability and extremes [e.g. Gomi et al., 2006;Johnson and Jones, 2000;Hannah et al., 2008;Imholt et al. 2011Imholt et al. , 2013Garner et al., 2015].
Although there is a clear requirement for understanding of the effects of riparian cover on stream temperature, there have been relatively few robust process based studies that provide realistic predictions of the likely effects of landuse change.
Moore et al. [2014] discussed various methods for representing the effects of vegetation on radiative energy fluxes above streams. However, to date river temperature models [e.g. Rutherford et al., 1997;Watanabe et al., 2005;DeWalle, 2008;Roth et al., 2010;Lee et al., 2012] have not considered the importance of vegetation structure (i.e. leaves, trunks and branches) and location relative to the position of the sun and the receiving waterbodies. Therefore, they were unable to adequately account for the temporally variable influence of discontinuous vegetation on the radiation budget. Furthermore, vegetation also has a significant effect on riparian microclimatic variables such as wind speed, relative humidity and air temperature, resulting in large reductions in latent heat losses (e.g. 60-87 % was observed by Garner et al., 2015) in comparison to open reaches [e.g. Hannah et al., 2008;Garner et al., 2015]. However, most modelling studies [e.g. Rutherford et al., 1997;Watanabe et al., 2005;DeWalle, 2008;Lee et al., 2012] have not considered the effects of changing microclimate as a result of riparian landuse change and so likely over-estimated the effect of forest canopies on reducing net energy fluxes and thus water temperature. Consequently, attempts to simulate the effects of riparian landuse change on water temperature have lacked the necessary physical realism to produce accurate estimates of effect sizes.
This study aims to generate systematic, process-based information on the effects of: (1) channel shading, (2) channel orientation and (3) water velocity on river temperature. Previous modelling and observational studies suggest that these three variables play an important role in determining river temperature dynamics. Firstly, because water temperatures are lower when vegetation is present [e.g. Hannah et al., 2008;Hrachowitz et al., 2010;Roth et al., 2010;Garner et al., 2015] and instantaneous differences in temperature between forested and open locations are greatest at sites under the densest canopies [e.g. Roth et al., 2010;Broadmeadow et al., 2011;Groom et al., 2011;Imholt et al., 2013]. Secondly, because the orientation of the channel [LeBlanc, 1997;DeWalle, 2008;Li et al., 2012] and therefore the location of vegetation relative to the path of the sun is important in controlling solar radiation inputs [Lee et al., 2012]. Finally, because longitudinal temperature gradients are reduced in steeper, faster flowing reaches compared with flatter, slower flowing ones [e.g. Danehy et al., 2005;Subehi et al., 2009;Groom et al., 2011]. Knowledge of these controls and their interactions is important to inform optimal tree planting strategies and to assess likely outcomes.
In this context, we simulate the effects of varying riparian vegetation density and channel orientation on the stream energy budget and quantify their influence on water temperature dynamics under scenarios of high and low water velocity. The effects of riparian vegetation on river temperature are modelled using hemispheric photographs of different riparian canopy densities under field observed conditions and local measurements of micro-climate, thereby providing improved realism to estimates of likely effect size while at the same time being sufficiently generalisable to provide useful information to inform riparian planting strategies.

STUDY AREA
We collected field data within a 1050 m study reach of Glen Girnock. This upland basin is located in north east Scotland and drains into the Aberdeenshire Dee ( Figure   1). The catchment upstream of the reach has an area of ~ 22 km 2 in which heather (Calluna) moorland dominated landuse. Riparian landuse along the reach transitioned from moorland to semi-natural forest composed of birch (Betula), Scots pine (Pinus), alder (Alnus) and willow (Salix) . Basin soils are composed predominantly of peaty podsols with some peaty gleys. Basin geology is dominated by granite at higher elevations and schists at lower elevations and is thus relatively impermeable [Tetzlaff et al., 2007]. Within the study reach the riverbed is composed primarily of cobble and boulder with gravel accumulation in localised patches. The reach is 280 m above sea level (asl) at the upstream reach boundary and 255 m asl at the downstream reach boundary. During field data collection the mean wetted width of the channel was 9.5 m. Previous work within the study reach demonstrated that there are no substantial groundwater inflows and consequently that groundwater does not significantly modify water temperature dynamics [Malcolm et al., 2005;Garner et al., 2014]. Thus, the influence of canopy density, channel orientation and water velocity on water temperature could be investigated in the absence of confounding groundwater influences [e.g. Story et al., 2003;Westhoff et al., 2011].
The UK Meteorological Office record daily averages of air temperature and totals of precipitation at Balmoral (< 10 km north west of the catchment). During the period 1950-2013 annual average air temperature was 6.6°C , maximum temperatures occurred in June and July (daily averages 13.0 and 12.6 °C respectively) and minima occurred December to February (daily averages 2.4, 2.2 and 1.6 °C respectively).
Between 1950 and 2013 annual average precipitation totalled 846 mm, October to January were the wettest months (daily average totals ranged from 85.7 mm in December to 92.5 mm in October) and February to September were the driest (daily average totals ranged from 55.1 mm in April to 70.8 mm in August). River discharge is monitored continuously by the Scottish Environmental Protection Agency (SEPA) in a rated natural section of the Girnock at Littlemill (Figure 1). Annual mean flow is 0.530 m 3 s -1 . Summer flows (i.e. June-August) are typically < 0.100 m 3 s -1 but the flow regime is highly responsive to precipitation and so high flow events (e.g. ≥ Q 10 , 1.126 m 3 s -1 ) occur year-round.

Experimental design
Spatially distributed field data were used to parameterise a simulation experiment that investigated the influence of: (1) riparian vegetation density, (2) channel orientation (and thus vegetation orientation relative to the sun's path), and (3) water velocity (a proxy for stream gradient) on heat exchange patterns and water temperature dynamics within a 1050 m reach of the Girnock Burn. A single time series of discharge was used for each velocity scenario thereby separating the the effects of velocity and residence time from those of varying water volume. Consequently, the effects of each vegetation and channel orientation scenario were simulated for a low (i.e. slow velocity: 0.023 ms -1 ) or high gradient (i.e. fast velocity: 0.155 ms -1 ) river. We did not investigate the effects of changing discharge because we were primarily interested in the effects of riparian woodland on river temperature under summer low flow conditions, when the most extreme high water temperatures are expected to occur.
Firstly, a process-based water temperature model (herein referred to as the 'base model') driven by spatially distributed energy flux data temperature [Garner et al., 2014after Bartholow, 2000Boyd and Kasper, 2003;Rutherford et al., 2004;Westhoff et al., 2007Westhoff et al., , 2010Leach and Moore, 2011;MacDonald et al., 2014a, b] was parameterised for observed conditions within the Girnock Burn. Previous work suggested that the base model adequately described spatio-temporal variability in river temperature [Garner et al., 2014], and thus is capable of providing realistic Stream temperature was predicted along the reach at a resolution of 50 m.  Figure 1) were high, while discharge was very low. Consequently, the effects of vegetation density, channel orientation and water velocity on water temperature were evaluated under a 'worst-case scenario' of high energy inputs and low flows [after Garner et al., 2014].

Micrometeorological measurements
Three AWSs (automatic weather stations) were installed within the reach (Figure 1) to characterise spatio-temporal variability in energy fluxes: the first was located in open moorland at the upstream reach boundary (AWS open ), the second was located in semi-natural forest 190 m downstream of the upstream boundary (named "AWS forest upstream" or AWS FUS ) and the third was located in semi-natural forest 685 m downstream of the upstream boundary (named "AWS forest downstream" or AWS FDS ). Hydrometeorological variables measured by each AWS were: air temperature (°C), relative humidity (%), wind speed (ms -1 ), incoming solar radiation, net radiation and bed heat flux (all Wm -2 ). The instruments deployed on the AWSs are detailed in Hannah et al. [2008]. AWSs measured meteorological variables ~2 m above the stream surface. Bed heat flux measurements were made using heat flux plates buried (to avoid radiative and convective errors) at 0.05 m depth within the riverbed below each AWS. Heat flux plates provided aggregated measurements of convective, conductive, advective and radiative heat exchanges between the atmosphere and the riverbed and the riverbed and the water column [after Evans et al., 1998;Hannah et al., 2008;Garner et al., 2014]. All AWS sensors were sampled at 10-second intervals and averages were logged every 15-minutes.

Stream temperature measurements
Stream temperature measurements were used to evaluate the performance of the base model under observed conditions [i.e. Garner et al., 2014] and provided initial conditions at the upstream reach boundary. Water temperature was measured at 15minute intervals using ten water temperature TinyTag Aquatic 2 dataloggers (manufacturer stated accuracy of +/-0.5 °C) and three Campbell 107 thermistors  Figure 1). Prior to installation the sensors were compared [following Hannah et al., 2009] over the range 0-30 °C and were in agreement by < +/-0.1 °C.
Sensors were deployed within white plastic PVC tubes to shield them from direct solar radiation.

Hydrology and stream geometry
Discharge (m 3 s -1 ) was obtained from a Scottish Environmental Protection Agency (SEPA) gauging station at Littlemill (Figure 1). Discharge was required as input to the water temperature model (see "3.4 Modelling approach"). The time series of discharge from 6 th July 2013 ( Figure 3e) was used as input to the base model run and for the simulation experiment runs; values were very low (average 0.089 m 3 s -1 , which is equal to Q 96 calculated for June-August during the period 1983-2013), stable (0.082-0.096 m 3 s -1 ) and exhibited no sudden changes. Water velocity (ms -1 ) for the base model was calculated from discharge using a discharge-mean velocity function for Littlemill derived by Tetzlaff et al. [2005] and was used to route discrete parcels of water through the reach in order to drive the flow-routing component of the water temperature model (see '3.4 Modelling approach'). For evaluation of the base model velocity was allowed to vary temporally (at hourly intervals) in response to changing discharge. For the simulation experiments constant values of high (0.155 ms -1 ) and low velocity (0.023 ms -1 ) were used at all locations and time steps. Wetted width was required as input to the water temperature model. Spatially varying values measured at 50 m intervals along the reach were used for the base model evaluation, but a fixed value of 9.5 m (the mean wetted width) was used for the simulation experiments.

Hemispherical images
Hemispherical images were taken at 5 m intervals along the stream centreline using a Canon EOS-10D 6.3 megapixel digital camera with Sigma 8 mm fisheye lens. Prior to taking each image the camera was orientated to north and levelled ~20 cm above the stream surface [after Leach and Moore, 2010]. All images were used to parameterise the radiation component of the base model and thus represent the baseline (current) riparian vegetation condition in the reach [i.e. Garner et al., 2014] for the model validation. Data derived from nine of these images (each representative of 10-90 % canopy density at 10 % increments; Figure 2) were used to parameterise the vegetation scenarios.

Net energy
Net energy (Qn, Wm -2 ) available to heat or cool the water column was calculated as: Where Q n is net energy, Q * is net radiation, Q e is latent heat, Q h is sensible heat and Q bhf is bed heat flux (all Wm -2 ). Heat from fluid friction was omitted because it makes a negligible contribution to the energy budget in this reach [after Garner et al., 2015].
Herein, positive energy fluxes represent gains to the water column while negative energy fluxes represent losses. Moore et al. [2005] and then extended and evaluated by Leach and Moore [2010] was used to compute net radiation (Q * ) at the location of each hemispherical image. At each location net radiation was calculated as:
Where α is the stream albedo, D(t) is the direct component of incident solar radiation at time t (Wm -2 ), g(t) is the canopy gap fraction at the position of the sun in the sky at time t, s(t) is the diffuse component of solar radiation (Wm -2 ), f v is the sky view factor, ɛ a, ɛ vt and ɛ w are the emissivity of the temperatures of the air, vegetation and water respectively (all °C), σ is the Stefan-Boltzmann constant (5.67 x 10 -8 Wm -2 K -4 ), and T a and T w are air and water temperature respectively (both °C).
Values for atmospheric emissivity were calculated for clear-sky day and night conditions using the equation presented in Prata [1996;used also by Leach andMoore, 2010, Garner et al., 2014] and were subsequently adjusted for cloud cover using equations in Leach and Moore [2010]. The emissivity and albedo were taken to be 0.95 and 0.05 for water, and 0.97 and 0.03 for vegetation respectively [after Moore et al., 2005 and used subsequently by Garner et al., 2014].
Hemispherical photographs were converted to binary images by setting a threshold that determines whether a pixel should be classified as sky (white) or another object (black) such as river banks, tree trunks, leaves or branches. An optimum threshold value of 130 was selected from candidate values of 120-190 at 10 unit increments.
This threshold value minimised RMSE between observed and modelled incoming solar radiation at AWS FUS during 1 and 7 July 2013 [see Garner et al., 2014]. The solar zenith and azimuth angles were computed as a function of time (t, minutes) using equations in Iqbal [1983] so that the canopy gap at the location of the sun could be derived from g*(θ,ψ) as a function of time, g(t). Sky view factor was computed as: were generated by linear interpolation between the two nearest AWSs to the point along the stream centreline at which the hemispherical photograph representative of the vegetation scenario was taken. Net longwave radiation is a function of water temperature; therefore initial values for this flux at the upstream reach boundary were calculated using observed water temperature at AWS Open .

Latent and sensible heat fluxes
To compute heat lost by evaporation or gained by condensation, latent heat was estimated after Webb and Zhang [1997] (Equation 6).
Where U is wind speed (ms -1 ) and e a and e w are vapour pressures of air and water (both kPa), respectively. Saturation vapour pressure (e sat ) was calculated as a function of air or water temperature, T (K), after Stull [2000] (Equation 7).
Vapour pressure of water (e w ) was assumed to be equal to e sat (T w ). Vapour pressure of air (e a ) was calculated using Equation 8.
For the simulation experiments, time series of meteorological variables (i.e. air temperature, wind speed and relative humidity) required to calculate turbulent fluxes were generated for each vegetation scenario by linear interpolation between the two nearest AWSs to the point along the stream centerline at which the hemispherical photograph representative of the scenario was taken. Turbulent fluxes are a function of water temperature; therefore initial values at the upstream boundary were calculated using observed water temperature at AWS Open .

Modelling approach
A Lagrangian modelling approach was used to simulate river water temperature [after Garner et al., 2014] in which the trajectory of discrete parcels of water is followed through the reach in order to determine the energy exchange conditions the parcels are exposed to and thus calculate changes in their temperature as they flow downstream and time elapses.
The reach was divided into a series of 1 m segments (s) bounded by nodes (x). At hourly intervals a discrete parcel of water (i) with an initial temperature was released from the upstream boundary at AWS Open and routed through the reach using the discharge-mean velocity function [Tetzlaff, 2005]. The distance travelled by each water parcel from its location (x) at time t to its next location (x+1) at time t+Δt was calculated as the product of the length of each 15-minute time step (Δt, i.e. 900 seconds) and either: (1) for evaluation of the base model, the average velocity of the parcel at times t and t+Δt or (2) for the simulation experiments, 0.023 or 0.155 ms -1 for the low and high velocity scenarios, respectively. As the water parcel travelled downstream from x towards x+1 the model determined the mean of each of the meteorological variables the parcel was exposed to along its trajectory through the segments at times t and t+1. This information was used to calculate the water temperature of each parcel at 50 m intervals by integration of Equation 11 in the deSolve package [Soetaert et al., 2010] for R (Version 3.0.2, R Group for Statistical Computing, 2013).
Where ‫ݓ‬ ௦̅ is the mean wetted width of the stream surface (m) within segments ‫̅ݏ‬ , Energy exchange due to bed heat flux, which accounted for < 1 % of the stream energy budget [Garner et al., 2014], was retained within the model structure for

Base water temperature model evaluation
The performance of the base water temperature model was evaluated previously by

Vegetation density and channel orientation effects on simulated water temperature dynamics
Water temperature metrics were derived from all values simulated throughout the reach (i.e. n= 483 temperatures). Typically, mean and maximum water temperatures

Effects of water velocity on simulated water temperature dynamics
The velocity under which simulations were performed determined the residence time of water parcels within the reach. The high velocity scenario resulted in shorter residence time (cf. low). For example the parcel of water released from AWS Open under the high velocity scenario at 23:00 on 23 July left the reach around 00:45 on 24 July (Figure 9a and b) whereas under the low velocity scenario the water parcel did not leave the reach until around 11:30 on 24 July (Figure 9d and e).
Shorter ( Figures 9g and 9h for a scenario of 30 % canopy density in which the channel was orientated SE-NW (i.e. exposed to the strongest solar radiation gains) and a scenario of 30 % canopy density in which the channel was orientated NW-SE (i.e. shaded from the strongest solar radiation gains) respectively. Most notably, when the channel was exposed under the low velocity condition the highest temperatures (> 25.0 °C) occurred throughout most of the reach and persisted for longer (Figure 9g).
When the channel was shaded under the low velocity condition the lowest daytime temperatures (< 20 °C) occurred throughout the reach and persisted for longer ( Figure   9h).

DISCUSSION
This study quantified the influence of riparian vegetation density on energy exchange and water temperature dynamics in channels of varying orientation and with varying water velocity. The latter is a control of hydraulic retention time within the reach, which increases for lower gradient streams if wetted width and discharge are unchanged. The following discussion considers the effects of: (1) interactions between vegetation density and channel orientation on stream heating and cooling processes and (2) water velocity, and we identify the limitations of our approach. The implications of the findings are discussed in the context of river management in a changing climate.

Vegetation density, channel orientation and effects on stream heating and cooling
Riparian vegetation reduces solar radiation inputs and consequently net energy available to heat the water column [Hannah et al., 2004[Hannah et al., , 2008Leach and Moore, 2010;Garner et al., 2014Garner et al., , 2015. During the study period (Northern Hemisphere summer) at this relatively high latitude site (57°02'N) riparian vegetation had the greatest effect on net solar radiation and net energy inputs when it overhung the stream centreline and therefore shaded the stream from the greatest solar radiation inputs. Consequently during summer, when river flows are lowest and water temperature highest, riparian planting is only likely to be effective in reaches where river width is sufficiently narrow and/ or trees are sufficiently tall.
Around half of riparian vegetation scenarios did not typically reduce solar radiation sufficiently to produce net energy losses and therefore drive cooling of water as it Previous studies have demonstrated that summary daily water temperature metrics (especially maxima) are reduced under the densest riparian canopies [Broadmeadow et al., 2011;Groom et al., 2011;Imholt et al., 2013] and that the orientation of vegetation relative to the path of the sun is important in determining the magnitude of this reduction [Lee et al., 2012]. Our study demonstrated that for intermediate canopy densities, the effect of riparian vegetation on maximum and mean temperatures is strongly dependent on channel orientation and thus the location of vegetation relative to the path of the sun. A canopy of 30 % density could be as effective at reducing maximum and mean temperatures as a canopy of 60 % density, provided that it shaded the water column when potential solar radiation gains were greatest (i.e. when the sun was between south-easterly and south-westerly sky positions in the Northern Hemisphere), while a canopy cover of up to 60 % could have little effect in reducing maximum and mean temperatures if it did not shade the channel while the sun was in these sky-positions.
River managers are increasingly searching for ways to reduce deleterious maximum temperatures. Re-introduction of riparian shading offers one of the most promising management approaches. Nevertheless, river managers must work within a broader social and economic context, where riparian planting (and associated fencing) comes with significant financial costs and has the potential to conflict with other landuses, which in the uplands of Scotland includes deer stalking and grouse shooting. Our study suggests that the channel must be shaded almost entirely to generate the greatest reductions in mean and maximum temperatures, so this is an 'expensive' and potentially unachievable way to create thermal refugia. Such dramatic reductions may be desirable at locations where water temperatures are near, or anticipated to exceed, lethal or sub-lethal thresholds for an organism of interest [Beechie et al., 2013].
Consequently, the introduction of minimal shade targeted to appropriate headwater reaches may be the most cost-effective and ecologically beneficial method to generate cool-water refugia. Based on our results for Northern Hemisphere streams, optimal planting would take place on the most southerly bank of channels flowing east-west, northeast-southwest, or northwest-southeast, and vice versa. These planting locations could achieve considerable reductions in mean and maximum temperatures at minimal cost while minimising potential negative ecological consequences associated with dense shading. Channels that are orientated north-south, and vice versa, and thus do not have abundant southerly banks would require denser vegetation on their west and east banks to shade the water column from the highest solar radiation gains and thus yield reductions in water temperature. As such, they are likely to be a lower priority for targeted riparian planting schemes when reductions in stream temperature are a stated objective.

Effects of water velocity on stream heating and cooling
Mean and maximum water temperatures were increased and (to a lesser extent) minimum temperatures were decreased when water travelled at a low velocity (cf. high velocity) due to a longer residence time within the reach and thus greater accumulation/ dissipation of heat [Subehi et al., 2009;Danehy et al., 2005;Groom et al., 2011]. Consequently, our results suggest that riparian planting should be targeted in slow-flowing reaches, where retention times are longer and heat accumulation, and thus water temperatures, can be minimised most efficiently.

Limitations
Models are always simplifications of reality; therefore they must incorporate assumptions [Westhoff et al., 2011]. Garner et al. [2014] discuss in full the assumptions and consequent limitations of the base model. Here we identify the assumptions made in conducting the simulation experiments and make suggestions for improvements in future model applications.
In the experiments presented herein we sought to represent spatial variability in micro-climate through linear interpolation between relatively closely spaced AWS.
The effects of spatially variable micro-climate have been often ignored in previous studies [e.g. Rutherford et al., 1997;Watanabe et al., 2005;DeWalle, 2008;Lee et al., 2012] but can modify turbulent fluxes and thus the energy budget significantly [e.g. Hannah et al., 2008;Garner et al., 2015]. We considered this approach to be reasonable for the base scenario and the good evaluation statistics suggest that this simple method was reasonable and appropriate.
Unfortunately, with only three AWS sites, it was not possible to separate the influence of riparian landuse from wider landscape effects on micro-climate. Consequently, we were unable to scale turbulent fluxes appropriately for the different landuse scenarios where this resulted in a spatial distribution of vegetation or channel orientation characteristics that differed from the base model. Garner et al. [2014]  Finally, we investigated the effects of changing velocity on water temperature but did not investigate the potential effects of spatially or temporally varying discharge and did not consider the effects of changing velocity (for a fixed discharge) on wetted width. A full investigation of the effects of velocity, wetted width and discharge on river temperature could be conducted in future using channel geometry data in combination with hydraulic and hydrological models.

CONCLUSIONS
This study used field data from an upland Scottish salmon stream to underpin simulation experiments and provide systematic, mechanistic understanding of the effects of riparian shading scenarios, channel orientation and velocity on water temperature dynamics. The information gained from the novel modelling approach allows scientists and river managers to make better-informed decisions on optimal riparian tree planting strategies, through improved understanding of the interrelationships between channel orientation and vegetation density that influence the effectiveness of riparian vegetation as a strategy for mitigating thermal extremes. The magnitude of reductions in water temperature under a given canopy density will depend on local conditions [Ryan et al., 2013] including the magnitude of net energy exchange (linked to meteorological conditions but also vegetation cover density and channel orientation), water velocity and hydrology. The experiments presented here demonstrate that where southerly banks (in the Northern Hemisphere) may be afforested then relatively sparse, overhanging vegetation is able to produce spatially and temporally extensive cool-water refugia when thermal extremes occur. Only in reaches where a southerly bank cannot be afforested is dense, overhanging vegetation required, and potentially deleterious effects should be considered in these circumstances. Additionally, planting should be targeted in slow-flowing (e.g. low gradient) reaches where flow retention times are longer and within which large heat loads can accumulate in the absence of shade.
Scientists and river managers can use models such as those presented here to quantify potential changes in river thermal conditions associated with riparian planting schemes under both present and future climates at relatively small spatial scales.
However these models require large observational datasets that are rarely available, and are logistically and financially unfeasible to collect in many circumstances.
Consequently, future research should also seek to upscale the information yielded in this study to identify readily defined proxies for sensitivity (e.g. channel orientation and gradient) that can be combined with rapid riparian canopy density assessments [e.g. Imholt et al., 2013] in statistical models capable of predicting water temperatures at large spatial scales [e.g. Hrachowitz et al., 2010].