Faye L. Jackson

Global warming is causing rapid and profound impacts on northern ecosystems. Shifting precipitation, permafrost thaw, and rapid shrub advance into tundra environments are combining to alter stream temperature in uncertain ways. The expansion and increased abundance of shrub vegetation, termed shrubification, may moderate warming stream temperatures through greater riparian shading. However, there has been no evaluation of how shrub expansion will affect the radiative balance and subsequent stream temperature of northern streams to date. This study (a) assesses the impact of shrubification in northern latitudes on the incident shortwave and longwave radiation received by streams across a range of stream widths, orientations, gradients, and latitudes, and (b) quantifies the potential magnitude of summer stream temperature response to the altered incident radiation. The potential temperature response was assessed regionally using a metric, , that considers changing incident radiation () between scenarios of three increasing riparian shrub structures and two streamflow conditions. Subsequently, a local case study was undertaken comparing an energy balance‐driven equilibrium stream temperature analysis () between the shrub scenarios. Simulation results of the hourly response to substantial shrubification ranged from −250.2 to +17.1 W m ⁻² , with a mean of −41.6 W m ⁻² , across all stream scenarios. The response was greatest in streams <5 m wide, although stream orientation influences response for wider streams. The analysis indicated a variable temperature response related to the progression of shrubification. These results suggest that stream temperature response to shrubification in headwater catchments may be complex and context dependent.

River temperature is an important control on ecosystem health. Concerns around climate change impacts and availability of lower cost sensors have led to an increase in temperature monitoring of aquatic environments. While equipment biases are typically small compared to daily, seasonal, and spatial temperature variability at catchment scales, they can be large in the context of long‐term temporal trends. Failure to address issues of quality control and calibration can thus affect the detection, direction, and magnitude of reported temperature trends. This study uses a 35‐year river temperature time series, collected in the Girnock Burn, Scotland, to illustrate the potential influence of equipment (datalogger/sensor) bias on reported trends. Non‐linear, seasonally varying trends were observed, with summer months showing the greatest positive trends. Calibration experiments comparing historical and contemporary equipment with a calibrated internal reference logger revealed that older dataloggers (pre‐2007) exhibited coarse reporting resolutions and measurement rounding resulting in positive and negative biases depending on the equipment make and model. Recent dataloggers collected data at a higher reporting resolution, where operational rounding issues were not relevant, and exhibit only relatively minor biases, even before calibration. Comparison of raw and bias‐corrected temperature trends showed that uncorrected time series exhibited moderated temporal trends due to positive biases prior to 1998 and negative biases between 1998 and 2007. This study illustrates the importance of metadata in interpreting long‐term time series and the need for appropriate calibration and quality control procedures for temperature monitoring networks, especially where long‐term trends are of interest.

High river water temperature (T_w) extremes have been widely reported during drought conditions as extreme low‐flows often coincide with high atmospheric energy inputs. This has significant implications for freshwater ecosystem health and sustainable river management practices globally. However, the extent to which different meteorological and hydrological processes interact during droughts to govern T_w dynamics, and how this varies between environmental contexts, remains poorly understood. Here, we review the mechanisms controlling T_w dynamics during droughts across temperate, maritime environments, using the United Kingdom as a detailed case study. We evidence that T_w spikes have widely occurred during extreme low‐flow events observed within droughts, but such trends have been inconsistent due to varying hydroclimatic conditions and river basin controls. To better understand this, we re‐conceptualize the mechanisms governing drought‐induced T_w dynamics operating across three ‘process sets’: (i) ‘energy flux dynamics’ as non‐advective controls on T_w; (ii) the role of ‘reach‐scale habitat conditions’ in mediating non‐advective controls on T_w, including hydraulic properties (e.g., residence time) and physical conditions (e.g., riparian vegetation coverages, wetted perimeters); (iii) ‘water source contributions’ (surface water and groundwater) as advective heat and water flow controls. We review natural and anthropogenic influences affecting T_w controls within each process set and discuss how such mechanisms are likely to change under drought conditions. More systematic research (spanning various river environments and drought severities) is required to test such concepts, with existing scientific knowledge on drought‐induced T_w dynamics being largely gleaned from studies examining non‐extreme low‐flow conditions or with broader focuses (e.g., annual thermal dynamics). We conclude by highlighting critical future research questions that need to be answered to better model T_w dynamics during future droughts and for unmonitored sites. Such scientific advances would more effectively inform how high T_w extremes could be better managed through evidence‐based mitigation and adaptation strategies.

Increasing river temperatures are a threat to cold water species including ecologically and economically important freshwater fish, such as Atlantic salmon. In 2018, ca. 70% of Scottish rivers experienced temperatures which cause thermal stress in juvenile salmon, a situation expected to become increasingly common under climate change. Management of riparian woodlands is proven to protect cold water habitats. However, creation of new riparian woodlands can be costly and logistically challenging. It is therefore important that planting can be prioritized to areas where it is most needed and can be most effective in reducing river temperatures. The effects of riparian woodland on channel shading depend on complex interactions between channel width, orientation, aspect, gradient, tree height and solar geometry. Subsequent effects on river temperature are influenced by water volume and residence time. This study developed a deterministic river temperature model, driven by energy gains from solar radiation that are modified by water volume and residence time. The resulting output is a planting prioritization metric that compares potential warming between scenarios with and without riparian woodland. The prioritization metric has a reach scale spatial resolution, but can be mapped at large spatial scales using information obtained from a digital river network. The results indicate that water volume and residence time, as represented by river order, are a dominant control on the effectiveness of riparian woodland in reducing river temperature. Ignoring these effects could result in a sub‐optimal prioritization process and inappropriate resource allocation. Within river order, effectiveness of riparian shading depends on interactions between channel and landscape characteristics. Given the complexity and interacting nature of controls, the use of simple universal planting criteria is not appropriate. Instead, managers should be provided with maps that translate complex models into readily useable tools to prioritize riparian tree planting to mitigate the impacts of high river temperatures.

There has been increasing use of spatial statistical models to understand and predict river temperature (T_w) from landscape covariates. However, it is not financially or logistically feasible to monitor all rivers and the transferability of such models has not been explored. This paper uses T_w data from four river catchments collected in August 2015 to assess how well spatial regression models predict the maximum 7-day rolling mean of daily maximum T_w (Tw_max) within and between catchments. Models were fitted for each catchment separately using (1) landscape covariates only (LS models) and (2) landscape covariates and an air temperature (T_a) metric (LS_T_a models). All the LS models included upstream catchment area and three included a river network smoother (RNS) that accounted for unexplained spatial structure. The LS models transferred reasonably to other catchments, at least when predicting relative levels of Tw_max. However, the predictions were biased when mean Tw_max differed between catchments. The RNS was needed to characterise and predict finer-scale spatially correlated variation. Because the RNS was unique to each catchment and thus non-transferable, predictions were better within catchments than between catchments. A single model fitted to all catchments found no interactions between the landscape covariates and catchment, suggesting that the landscape relationships were transferable. The LS_T_a models transferred less well, with particularly poor performance when the relationship with the T_a metric was physically implausible or required extrapolation outside the range of the data. A single model fitted to all catchments found catchment-specific relationships between Tw_max and the T_a metric, indicating that the T_a metric was not transferable. These findings improve our understanding of the transferability of spatial statistical river temperature models and provide a foundation for developing new approaches for predicting T_w at unmonitored locations across multiple catchments and larger spatial scales.

The use of spatial statistical models to understand and predict river temperature (Tw) from landscape covariates has increased rapidly. However, it is not logistically feasible to monitor all rivers and the transferability of such models has not been explored. This paper uses Tw data from four river catchments collected in August 2015 to assess how well spatial regression models predict the maximum 7 day rolling mean of daily maximum Tw (Tw_max) within and between catchments. Models were fitted for each catchment separately using (1) landscape covariates only (LS models) and (2) landscape covariates and an air temperature (Ta) metric (LS_Ta models). All the LS models included upstream catchment area and three included a river network smoother (RNS) that accounted for unexplained spatial structure. The LS models transferred reasonably to other catchments, at least when predicting relative levels of Tw_max. However, the predictions were biased when mean Tw_max differed between catchments. The RNS was needed to characterise and predict finer scale spatially correlated variation. Because the RNS was unique to each catchment and thus non-transferable, predictions were better within catchments than between catchments. A single model fitted to all catchments found no interactions between the landscape covariates and catchment, suggesting that the landscape relationships were transferable. The LS_Ta models transferred less well, with particularly poor performance when the relationship with the Ta metric was physically implausible or required extrapolation outside the range of the data. A single model fitted to all catchments found catchment-specific relationships between Tw_max and the Ta metric, indicating that the Ta metric was not transferable. These findings improve our understanding of the transferability of spatial statistical river temperature models and provide a foundation for developing new approaches for predicting Tw at unmonitored locations across multiple catchments and larger spatial scales.

Water temperature is an important control on processes in aquatic systems and particularly for freshwater fish, affecting growth, survival and demographic characteristics. In recognition of this importance, the Scottish Government has prioritised developing a robust national river temperature monitoring network. Advances in geographical information systems, spatial statistics and field data loggers make large-scale river temperature monitoring increasingly possible. However, duplication of environmental and thermal characteristics among monitoring sites means many networks have lower than expected statistical power. This paper describes a novel methodology for network design, illustrated by the development of the Scotland River Temperature Monitoring Network. A literature review identified processes controlling stream temperature and associated landscape controls. Metrics indicative of these landscape controls were calculated for points every 500 m along the river network. From these points, sites were chosen to cover the full range of observed environmental gradients and combinations of controlling variables. The resulting network contains sites with unique characteristics covering the range of relevant environmental characteristics observed in Scottish salmon rivers. The network will thus have minimal redundancy, often not seen in large networks, and high statistical power to separate the relative importance of predictor variables thereby allowing large-scale water temperature predictions.