Introduction

Rivers are dynamic ecosystems that are resilient to change when ecologically intact. Central to this resilience is the interaction between surface water and groundwater. Groundwater temperature remains stable as air temperature changes, making it an important buffer against climate change. During the summer, areas of groundwater upwelling can be significantly cooler than the ambient river temperature. This buffering effect plays a critical role in the growth and survival of cold-water fishes, which use these areas to avoid stressful river temperatures (Torgensen et al. 2012).

Our project focused on groundwater-induced thermal refuges. These are broadly defined as areas within a stream that fish preferentially use to avoid thermal stress (Sullivan et al. 2021). Thermal refuges have allowed cold water fish like salmon to persist and thrive in historically hot and dry systems, and are becoming increasingly important with the intensification of climate change. It is therefore crucial for fisheries managers to identify these sites, observe how fish use them, and understand the hydrological processes that create and maintain them. 

Our work took place in the Coldwater River, a tributary of the Nicola River and part of the Thompson and Fraser watersheds. Across the southern interior of B.C., summer air temperatures are rapidly warming, which results in warming of streams. Salmon require water temperatures below 18 ºC to thrive, with temperatures exceeding 22 ºC resulting in stress and potentially death. Recent observations in the Coldwater have recorded stream temperatures exceeding 25 ºC (Zdasiuk et al. 2024), and rising stream temperatures have been cited by scientists and traditional knowledge holders as a limiting factor to salmon recovery in the system (Ulaski et al. 2023). Salmon populations that use the Coldwater are especially vulnerable to heat stress due to their life history and migration timing. For example, Lower Thompson Chinook, Interior Fraser Coho, and interior Fraser Steelhead rear in the river year-round, and Chinook spawn in late summer, when flows are lowest and water temperatures are highest (Kosakoski & Hamilton, 1982). A legacy of intensive land use–including forestry and agriculture–coupled with recent climate-driven natural disasters (e.g., floods, fires, pine beetle infestation) have resulted in degradation of the riparian environment throughout much of the lower river. This, in turn, is exacerbating climate impacts and compromising the ability of the river to adapt to future change (Ulaski et al. 2025). 

The goal of this project was to map potential thermal refuges in the lower Coldwater River using a thermal infrared drone. In addition, we performed ground-truthing and snorkel surveys at select sites to characterize the habitats and their use by fish. This work is a foundational step for improving our understanding of thermal refuges in the Coldwater River and will be valuable for future restoration, enhancement, and management projects that support the resilience of the river and its salmon in the face of climate change. 

Acronyms list

RGB: Red, Green, Blue (visible light) imagery 
rKm: River kilometers
TIR: Thermal Infrared
VLOS: Visual Line of Sight

Methods

Study site

The Coldwater River is situated in nłeʔképmx and Syilx territory and home to five band governments: sp’ax̌ʔmi and nɬq’aɬməlʔx (Upper Nicola), sulús (Lower Nicola), nc̓ə́ɬetkʷu (Coldwater), sxéxn̓x (Shackan), and nwéyc (Nooaitch). Salmon are profoundly linked to the culture, identity, and the well-being of Indigenous peoples in the watershed and the decline of salmon returns has had significant cultural, food security, and economic impacts on these communities. 

The hydrograph of the river is snow dominated, characterized by peak flows during the freshet in May or June with flows diminishing steadily throughout the summer. Baseflows and maximum water temperatures are usually observed around mid-August. 

Our surveying focused on the lower reach of the river from the Brookemere interchange (49° 52’13″N , 120° 54’40″W) to the Coldwater-Nicola confluence in Merritt. This reach is often characterized as the most thermally stressful reach of the watershed due to its lower elevation, lack of cold tributary inputs, highly modified riparian zone, high anthropogenic water demands, low summer flows, and low gradient. In addition, the lower reach is heavily used by Chinook salmon for spawning in years where streamflows are favourable. 

Thermal infrared surveys

Video overflights were conducted on July 29th and August 1st, 2025. We used a DJI Matrice 300RTK drone paired with a ZenMuse H20N sensor to conduct all surveys. The drone was programmed to follow a pre-determined path that was between 5-7 km depending on access and suitable takeoff and landing sites. The Coldwater River is bounded along most of the lower river by Coldwater Road, which allowed us to position the pilot as a passenger in a follow vehicle. Surveys were conducted in a point-to-point fashion, maintaining visual line of sight (VLOS) and controller signal connectivity throughout the route. Additional visual observers were positioned in a separate follow vehicle. We set flight speed at 10 m/s, capturing thermal and RGB video simultaneously. We set a maximum survey altitude of 120m above the ground and manually operated the gimbal to ensure image coverage of the entire river. Flights were conducted at mid-day when temperatures were near daily maxima and the sun was at its highest. The daily high temperatures for July 29th were 30 ºC and 31 ºC respectively, with minimal to no cloud cover.

Thermal imagery analysis

Unlike previous thermal imaging efforts in the watershed, we opted to collect thermal video instead of still imagery. Our goal was to quickly and efficiently identify discrete areas of relative difference in water temperature along a large study reach (~45 river kms). As such, we did not record absolute temperature values collected by the sensor nor did we collect calibration data from the river using temperature loggers.

We classified temperature anomalies as “cold-water patches.” Cold-water patches are distinct from thermal refuges as the latter implies preferential use by fish during heat stress while the former describes a difference in water temperature at a discrete site visible to the infrared sensor.

During video analysis, cold-water patches were displayed as dark blue signatures relative to the surrounding water. Each cold-water patch was noted and the location georeferenced from the drone’s in-flight GPS. Cold water patches were categorized as one of five categories: side channel, alcove, springbrook, tributary confluence, or lateral seep, based on their morphology and hydrologic connectivity to the river (Table 1) (Wilms, 2024).

Table 1: Thermal refuge classifications 

 Thermal refuge classifications with example imagery collected during our overflights.

Site visits and snorkel surveys

We visited six identified cold-water patches to conduct habitat and fish abundance surveys. Sites were selected based on access, size, and distribution throughout the river. Sites were visited between August 12 and August 15. Each site was visited mid-day with daytime highs reaching 30 ºC with a cloudless sky. As infrared imaging is unable to detect temperature below the waters’ surface, we visited an additional six deep pool habitats (> 3 m maximum depth) previously identified by Scw’exmx Tribal Council habitat surveys to determine if they provided thermal refuge. These six sites were the only deep pool habitats within our study reach.

At each site, a qualitative score (high, moderate, none) of large woody debris volume and riparian vegetation percent cover was recorded. A minimum water temperature was recorded and compared to the mainstem water temperature. We also measured the wetted length, wetted width, and maximum depth of the cold water patch.

Each cold water patch was snorkeled once during the summer to document aquatic habitat metrics (substrate size, large wood abundance) and salmonid presence and abundance. Surveys were conducted by one or two snorkelers, depending on the size of the cold-water patch. We followed standard operating procedures for snorkel surveys in rivers from O’Neal (2007), outlined in the Salmonid Field Protocols Handbook (AFS, 2007). 

Results

Thermal imaging

We identified 38 cold water patches between the Brookemere interchange and the Coldwater-Nicola confluence. The majority of identified cold water patches were side channels and alcoves. 

Table 2: Identified cold-water patches by type

Refuge Type	Count
Side Channel	11
Alcove	15
Springbrook	4
Tributary Confluence	1
Lateral Seep	2

Cold water patches were relatively evenly distributed throughout the study reach, with the exception of three notable gaps– large distances between cold water patches. These gaps were roughly 3 river kilometres (rkm) long and may pose a thermal barrier to fish during the summer months. 

Thermal barrier 1

Thermal Barrier 2

Thermal Barrier 3

Site characteristics

Riparian vegetation abundance (% cover) was low to moderate at five of the six cold-water patches. Large woody debris abundance was high at two sites and low to none at four sites. Minimum recorded water temperatures ranged from 16 – 19 ºC. Sites were generally shallow, with maximum depths between 30 cm and 1.2 m, and varied widely in area, from 45 m² to 2,500 m². All sites were classified as either alcoves or springbrooks.

None of the sites exhibited cold water plume expression into the mainstem river. In addition, most sites warmed dramatically from their upwelling source to their downstream terminus; at some sites, temperature differences were as great as 5 ºC. Given the low velocities, shallow depth, and lack of riparian shade, this warming is likely due to solar radiation. As a result, the effectiveness of these areas as thermal refuges for fish – particularly during hot, cloudless conditions – may be limited.

Figure 1. Two cold alcove sites

Figure 2. Springbrook site

Salmonid abundance

Fish assemblages at all sites were dominated by juvenile Coho and Chinook salmon, with Coho generally outnumbering Chinook ten to one. Average fork length for both species was estimated between 50 – 80 mm. Juvenile Rainbow Trout and juvenile Mountain Whitefish were also common at all sites. Likely due to the nature of the sites as shallow, off-channel habitat with low velocities, we did not observe any adult salmonids at any sites. Fish at all sites were observed exhibiting normal holding and feeding behaviour and body condition was generally normal.

Discussion

Drone-based thermal mapping

Our project contributes to the growing field of drone-based thermal imaging of rivers (Willms 2024, Dugdale et al 2019, Casas-Mulet et al. 2020, Kuhn et al 2021). Drone-based mapping has a variety of advantages over conventional crewed aircraft surveys. Manned aircraft are expensive, carry an increased risk burden, contribute to air and noise pollution, and pose a greater disturbance to people and wildlife. Point-to-point video missions proved an effective method to quickly identify cold water patches at a coarse scale with minimal data processing effort. For areas of interest or where greater resolution is needed, traditional mapping missions should be flown to produce fine-scale thermo-orthomosaics.

A common challenge associated with aerial thermal imaging is the inability of infrared sensors to see through vegetation, potentially leading to missed cold water signatures. In our experience on the lower Coldwater, the wide channel width and lack of riparian vegetation allowed for adequate imaging of the entire channel. This could become a greater issue when surveying the upper reaches of the river where channel width narrows and riparian forests are intact.

Like all drone-based surveys, we encountered two operational challenges: battery capacity and controller connectivity. With two fully charged batteries, our expected flight time was roughly 20-25 minutes, corresponding to roughly 7 river kilometers of surveyable terrain. This posed a challenge where take off and landing sites were further than our maximum battery range, which required us to make shorter out-and-back flights to ensure that we captured the entire section. We suggest future teams scout potential take-off and landing sites prior to flying, and have a complete set of six (capacity of the DJI BS60 Intelligent Battery Station) fully charged batteries for each field day. Lastly, it is imperative to become familiar with the expected flight times for your drone model under various wind and weather conditions.

The Coldwater River is bounded by a small side road (Coldwater Road) along much of the study reach, which allowed the pilot and visual observers to follow the drone in a vehicle. This contributed greatly to mission efficiency as we were able to conduct point-to-point flights, rather than out-and-back flights. Using this method we were able to survey the entire study section (~45 rkm) over two days. This method may not be safe or effective on busier roads or highways.

We found it useful to determine if there were locations within the flight path and vehicle route where the distance from the controller to the drone would exceed the expected range of ~1.5 km. At these locations we would attempt to move closer to the drone on foot and/or enter manual flight mode to ensure connectivity. We also found it useful to position the pilot in the rear of the vehicle on the side closest to the drone (i.e. driver’s side if the drone is to the left of the vehicle) with the window open to maximize signal strength. 

Deep pools as thermal refuge

One assumption of this project was that deep pool habitats would serve as reliable thermal refuges for salmon. After snorkeling all six deep pools in the lower river, we were surprised to learn that none of the sites provided cold water habitat, either through hyporheic exchange or thermal stratification, both of which are common mechanisms that deep pool habitats provide thermal refuge in other systems (Sivakumaran 2014, Frechette 2018, Linnansaari 2023). 

We suggest that this could be the result of the combination of shallow depths, thermal mixing, streambed permeability, and lowering of the water table from environmental and anthropogenic factors. This is a significant finding as it shows that the availability of thermal refuge habitat in the lower river is limited to use by only juvenile fish. Deep pool habitats are critical for adult salmon as they hold in the river prior to spawning. The lack of thermal refuge sites suitable for adult salmon could be potentially limiting to the amount of spawners the river can support, especially considering that the life history of spring run Chinook accessing the Coldwater River relies on them holding for several months in the river before spawning (STC, 2025). This is reflected by the fact that the majority of these spawners are restricted to a few exceptionally deep pools in the higher elevation reaches of the Coldwater River (STC, 2025).

Cold water patches vs. thermal refuges

Sullivan et al. (2021) proposes a standardized nomenclature for thermal refuges which we incorporated into our analysis. They define thermal refuges as areas within a stream that fish preferentially use to escape thermal stress, while a cold water patch is an area that is cooler than the surrounding water. Not all cold water patches can be defined as thermal refuges, and without direct observations of fish use it is impossible to determine if a cold-water patch is acting as a thermal refuge. Habitat characteristics besides temperature such as overhead cover, substrate size, food availability, and underwater structure can all contribute to a cold water patch being used as a thermal refuge. An important next step for our work is to thoroughly survey the cold-water patches identified via drone surveys to determine if they are used as thermal refuges. 

Conclusion

Aerial thermal infrared surveys proved effective at identifying cold-water patches along the lower Coldwater River. Site visits and snorkel surveys allowed us to characterize six cold-water patches in detail to determine their use by fish during periods of heat stress. A key finding of our work was that thermal refuge habitat in the lower Coldwater River is limited to use by juvenile fish as they were mostly off-channel habitats with low velocities and shallow depths. In addition, many of these areas lack riparian shade and aquatic structure, limiting their value as quality habitats for fish. Our work demonstrates that there is opportunity in enhancing and restoring existing cold-water patches to both maximize their habitat quality for fish and ensure that they are resilient to future impacts of climate change. 

Lastly, we developed a novel model for thermal refuge surveying in rivers with low riparian vegetation and adequate road access. Video overflights, as opposed to the generation of still thermo-orthomosaics, allowed us to rapidly survey 45 km of the river and identify almost one thermal anomaly per kilometer. Our work demonstrates that thermal infrared drone surveys can be a reliable, lower-cost solution for mapping thermal refuge habitat in rivers.

References

TA Willms, M Shrimpton, T Pypker, E Martins, D Hill. 2024. Identification and characterization of stream thermal refuges and associated habitat use by juvenile pacific salmon (Oncorhynchus sp.) and steelhead (O. mykiss). University of Northern British Columbia. doi:https://doi.org/10.24124/2024/59567

CE Torgensen, JL Ebersole, DM Keenan. 2012. Primer for identifying cold-water refuges to protect and restore thermal diversity in riverine landscapes. US Environmental Protection Agency, https://faculty.washington.edu/cet6/pub/Torgersen_etal_2012_cold_water_refuges.pdf 

ME Ulaski, L Warkentin, SM Naman, JW Moore. 2023. Spatially variable effects of streamflow on water temperature and thermal sensitivity within a salmon-bearing watershed in interior British Columbia, Canada. River Research and Applications 39(10): 2036-2047, https://doi.org/10.1002/rra.4200

B Zdasiuk, J Moore. 2024. Nicola stream temperature monitoring, 2023-2024 season report. Salmon Watersheds Lab, Simon Fraser University. 

DH Johnson, BM Shrier, JS O’Neal, JA Knutzen, X Augerot, TA O’Neil, TN Pearsons. 2007. Salmonid field protocols handbook: techniques for assessing status and trends in salmon and trout populations. American Fisheries Society.

K Sivakumaran, S Villamizar, H Pai, A Aguilar, TC Harmon. 2014. Assessing water temperature zones in idealized holding pools for Chinook salmon: a hypothetical study based on the regulated Lower San Joaquin River. American Geophysical Union Fall Meeting.

DM Frechette, SJ Dugdale, JJ Dodson, NE Bergeron. 2018. Understanding summertime thermal refuge use by adult Atlantic salmon using remote sensing, river temperature monitoring, and acoustic telemetry. Canadian Journal of Fisheries and Aquatic Sciences https://doi.org/10.1139/cjfas-2017-0422

T Linnansaari, AM O’Sullivan, C Breau, EM Corey, EN Collet, RA Curry, RA Cunjak. 2023. The role of cold water thermal refuges for stream salmonids in a changing climate – experiences from Atlantic Canada. Fishes https://doi.org/10.3390/fishes8090471

CJ Sullivan, JC Vokoun, AM Helton, MA Briggs, BL Kurylyk. 2021 An ecohydrological typology for thermal refuges in streams and rivers. Ecohydrology https://doi.org/10.1002/eco.2295