Climatic and catchment drivers of monthly water temperature of UK rivers

Water temperature is a key control on many river processes including ecology and biogeochemistry. Consequently, the effect of climate change on river and stream temperatures is a major scientific and practical concern. River thermal sensitivity to climate change/ variability is controlled by complex drivers that need to be unravelled to better understanding patterns of spatio-temporal variability and the relative importance of different controls to inform water and land management, specially climate change mitigation and adaptations strategies. To address these research gaps, we aim: (1) to quantify the relative importance of different climatic drivers of water temperature across a set of UK benchmark monitoring sites; and (2) to assess the effect of basin properties as modifiers of the climate-temperature relationships. For the UK, previous water temperature studies focussed either on a limited number of monitoring sites or climatic drivers. Water temperature data were collated across several long-term UK national capability projects, totalling 35 sites with a nationwide spread. Data were processed to create seasonal water temperature series (i.e. 3-month averages as follow: December - February for winter, March - May spring, June - August summer, and September - November autumn). Modelled climate data [daily, 1-km gridded forcing data for the Joint UL Land-Environment Simulator (JULES) for air temperature, short and long wave radiation, wind speed, specific humidity, and precipitation] were extracted for each of the water temperature site and seasonal averages were derived also. We modelled the response of water temperature (dependent variable) to the six climatic variables (predictors). One model per season was fitted to investigate season-specific controls; and one model was fitted for all seasons together to investigate the overall response (i.e. a total of five models). Methodologically, the study used a combination of two statistical techniques that are quite novel in the field: (1) a multi-level modelling approach was chosen to account for the hierarchical structure in the data sets (i.e. observations at a site, sites on a river); (2) model selection was based on an information theory criterion within a multi-model inference framework (i.e. sets of good models were selected rather a single best model). This approach showed that all six climatic variables exert some influence in some or all models. The most influential were air temperature for all seasonal models, short wave radiation for all seasons except summer, specific humidity for winter, wind speed and precipitation for summer. Spatial patterns were then investigated by mapping site-specific responses. These patterns were crossedchecked with selected basin properties in order to identify in what extent basins act as modifiers of the climate/ water temperature response.