Study area:
The Hengill geothermal catchment of SW Iceland in May 2018 (transplant assays) and August 2022 (in situ assays; see Fig. 1) was our focal field system. Headwater streams in the system differ in mean annual temperature from 3–20 °C due to geothermally warmed groundwater, but are otherwise alike in their physical and chemical characteristics. Thus, this model systems allows us to embed short-term manipulative assays within a long-term temperature gradient.

Two types of assays were conducted within the Hengill geothermal streams: (1) transplant respirometry (SW Iceland_AC): fish were collected from cold (IS12 with a mean ± standard deviation annual temperature of 7.8 ± 4.2 °C) and warm streams (IS1 = 11.3 ± 4.0 °C and IS5 = 13.8 ± 1.6 °C), and their metabolism was measured in five different streams (i.e. transplant of fish between rivers) to examine the effects of acute thermal exposure; and (2) in situ respirometry (SW Iceland_CH): metabolism was measured in the same nine streams where the fish were caught (i.e. no transplant of fish between streams) to examine the effects of chronic temperature exposure.

To test the generality of the chronic temperature exposure effects, we conducted in situ assays in three additional locations across the natural latitudinal range of brown trout and its widespread close congener, the Atlantic salmon: (1) UK_CH, consisting of two carriers of the River Frome (East Bourton Boundary Stream and Woodsford North Stream), sampled in August 2021; (2) Spain_CH, consisting of 10 rivers spread evenly across the Deva and Pas catchments, sampled in July 2021; and (3) NE Iceland_CH, consisting of four rivers (Hafralónsá, Hofsá, Selá, and Vesturdalsá) sampled in June 2018.

Assay procedure:
Fish were captured by electrofishing in all study sites and the methodology for field respirometry was consistent. Assays chambers consisted of 7.2 L round (32 cm diameter), airtight, and transparent plastic chamber (LocknLock brand), which were submerged in a 50 L container of river water, filtered through a 250 µm sieve. One individual fish was placed in each chamber and the lid was sealed underwater, ensuring there were no air bubbles in the chamber. Chambers were secured in shallow water for 1.5-3 hours approximately. Each time assays were run, one chamber contained only filtered river water (i.e. no fish) to act as a control for background photosynthesis and/or respiration of micro-organisms. A miniDOT logger (Precision Measurements Engineering “PME”) was inserted into each chamber to measure dissolved oxygen concentrations and water temperature every minute. At the end of each assay, fish were weighed and measured (fork length) and released into the same river from which they were captured. Weights of all fish individuals from Hengill 2018 and 20 individuals from NE_Iceland were estimated according to length-weight relationships obtained from empirical data collected at those locations.

Quantifying metabolic rates:
Fish exhibited some activity during the assays since movement is necessary to maintain their position in the water, thus we consider the decline in dissolved oxygen consumption rate as a proxy for routine metabolic rate here. To reduce potential effects of stress associated with collection and handling of fish, the first 30 minutes of recorded oxygen data in each assay were excluded from further analysis. The maximum duration of the recorded data was also standardised to 120 minutes, resulting in a 90-minute assay measurement period (after excluding the first 30 minutes) or maximum duration of the recorded data. The ‘auto_rate’ function in the ‘respR’ package v2.0.259 of R v4.1.360 was used to calculate oxygen depletion rates through a combination of rolling regression and Kernel density estimation (KDE) algorithms. This procedure identifies the most linear portion of the data (representative of routine metabolic rates) through rolling linear regressions of 30-minute time frames. Metabolic rates were only retained for further analysis if the r2 value for the regression was greater than 0.861, which was the case for 87% of the data. Background respiration rates were calculated as the slope of the linear regression through the entire 90-minute period of the control chamber and subtracted from the corresponding fish metabolic rates for that assay block. Positive oxygen depletion rates after background correction were excluded from further analysis (which was the case in just one assay). Per volume rates (mg O2 m 1 L-1) were converted into whole organism rates (mg O2 h−1) using the effective volume of the chamber, estimated as total volume minus the volumes of the miniDOT and the fish (assuming a density of 1,000 kg m-3).