Paired air-water annual temperature patterns reveal hydrogeological controls on stream thermal regimes at watershed to continental scales

Highlights

• Paired air-water annual signals are a promising tool for stream thermal regimes.

• Air temperature adds critical information for water thermal regime classification.

• Strong shallow to deep GW influence was indicated in small and large streams.

• Local hydrogeological controls are important for stream temperature heterogeneity.

• The methods described here can be efficiently applied to readily available data.

Abstract

Despite decades of research into air and stream temperature dynamics, paired air-water annual temperature signals have been underutilized to characterize watershed processes. Annual stream temperature dynamics are useful in classifying fundamental thermal regimes and can enhance process-based interpretation of stream temperature controls, including deep and shallow groundwater discharge, when paired with air signals. In this study, we investigated multi-scale variability in annual paired air-water temperature patterns using sine-wave linear regressions of multi-year daily temperature data from streams of various sizes. A total of 311 sites from two spatially intensive regional datasets (Shenandoah National Park and Olympic Experimental State Forest) and a spatially extensive national dataset spanning the contiguous United States (U.S. Geological Survey gages) were evaluated. We calculated three annual air-water thermal metrics (mean ratio, phase lag, and amplitude ratio) to deduce the influence of groundwater and other watershed processes on stream thermal regimes at multiple spatial scales. Site-specific values of the three annual air-water thermal metrics ranged from 0.69 to 5.29 (mean ratio), −9 to 40 days (phase lag), and 0.29 to 1.12 (amplitude ratio). Regional patterns in the annual thermal metrics revealed persistent yet spatially variable influences of shallow groundwater discharge and high levels of thermal variability within watersheds, indicating the importance of local hydrogeological controls on stream temperature. Furthermore, annual thermal metric patterns from the regional datasets were generally concordant with the national dataset suggesting the utility of these annual thermal metrics for analysis at multiple scales. Analysis of the national dataset showed that previously defined thermal regimes based on water temperature alone could be further refined using air-water metrics and these metrics were related to physiographic watershed characteristics such as contributing area, elevation, and slope. This research demonstrates the importance of spatial scale and heterogeneity for inferring hydrological process in streams and provides guidance for the interpretation of annual air-water temperature metrics that can be efficiently applied to the growing database of multi-year temperature records. Results from this research can aid in the prediction of future thermal habitat suitability for coldwater-adapted species at ecologically and management-relevant spatial scales with readily available data.

Introduction

As climate change and other anthropogenic alterations to watersheds transform the natural thermal regime of streams and rivers (Bassar et al., 2016, Isaak et al., 2012, Kaushal et al., 2010, Kędra and Wiejaczka, 2018), understanding the sensitivity and vulnerability of stream segments to these changes is increasingly imperative for long-term ecological management. Temperature is one of the most important properties controlling the water quality of streams (Caissie, 2006). Almost all physical, chemical, and biological processes of the stream corridor are influenced by temperature, including fish development and metabolism, dissolved oxygen concentration, biogeochemical cycling, and organic matter decomposition. Absent hydrologic alteration, channel water temperature is driven by meteorological and hydrogeological factors, such as incoming solar radiation, outgoing longwave radiation, air temperature, wind speed, humidity, stream channel dimensions (depth, width, and flow volume), and groundwater (GW) inputs (Caissie, 2006, Cluis, 1972, Westhoff et al., 2007). Because of its effect on ecosystem processes, stream temperature has received abundant attention in the scientific literature for more than a half-century and numerous modeling approaches have been used to predict different facets of its regime (key reviews include Anderson, 2005, Benyahya et al., 2007a, Caissie, 2006, Gallice et al., 2015, Kurylyk et al., 2019, Webb et al., 2008). Short-term air-water temperature relations have been used to map spatially variable stream water thermal sensitivity in summer (Kelleher et al., 2012). However, longer-term annual water temperature patterns, and how these relate to local air temperature patterns and watershed processes, have been underexplored.

Stream temperature models fall into deterministic or statistical groups (Benyahya et al., 2007a, Caissie, 2006). Deterministic models are based on the balance of energy (heat) and mass (flow) fluxes in a water body (Boyd and Kasper, 2003, Glose et al., 2017). These models are best for conducting impact studies that assess changes to one or more components of the stream heat budget or when exploring changes in temperature at multiple spatial scales (Benyahya et al., 2007a, Caissie, 2006, Westhoff et al., 2007). However, total heat budget models are complex, computationally expensive, and require numerous hydrological, physiographic, and meteorological inputs that may be excluded from typical measurement protocols, poorly defined at the spatial scale needed for ecological applications, or difficult to quantify. In contrast, statistical models are computationally simpler with minimal data requirements (Benyahya et al., 2007a) facilitating prediction at ecologically relevant spatial grains and extents. However, statistical models currently lack a clear understanding of the relationships between derived model coefficients and important watershed processes potentially limiting their utility. Further, unsampled spatial heterogeneity can lead to overly simplistic predictions at multiple scales. For instance, the development of national-scale models, such as The National Water Model (NOAA, 2019), requires understanding of intra- and inter-regional spatial variation for meaningful downscaled predictions.

Statistical models typically use stochastic methods to assess relationships between water and air temperatures for small time-steps (e.g., daily). These models result in the derivation of a long-term periodic component of water temperature data, which is commonly assumed to be temporally stable (Caissie et al., 1998). However, interannual and long-term trends in water temperature may be driven by temporally variable watershed processes, including solar radiation, GW discharge dynamics from adjacent aquifers and temporal runoff patterns. While tracing heat signatures has long been used to monitor the activity of these and other contributing factors (Anderson, 2005, Constantz, 2008, Halloran et al., 2016), few studies have explored their influence on patterns and properties of the long-term water temperature signal (i.e., mean, phase, and amplitude). Together, these properties can indicate the presence of stream thermal inertia (Letcher et al., 2016), influence of shallow (~upper 6 m) versus deeper GW discharge (Briggs et al., 2018b), riparian shading (Fabris et al., 2018, Johnson and Jones, 2000, Wawrzyniak et al., 2017), or dam operations (Rounds, 2007). However, it is unclear whether these observations are applicable across locations and climates and if annual stream temperature signals are relatively stable over time. Further, air temperature is often assumed to be a dominant control on stream temperature at seasonal timescales, but recent research indicates uncoupling of the annual signal amplitude and phase relations between air and water temperature may be used to infer other watershed controls on the stream heat budget (Briggs et al., 2018a).

Here, we assess the utility of simple statistical modeling approaches using multi-year paired air-water temperature data for developing ecologically relevant thermal metrics, and to evaluate how these diagnostic metrics vary across multiple spatial scales. The specific objectives are to: (1) compare annual air and water temperature signal parameters and combined air-water metrics across watershed, regional, and national spatial scales; and (2) provide guidance for the interpretation of three paired air-water annual thermal metrics (mean ratio, phase lag, and amplitude ratio) derived from sine-wave linear regressions with respect to physical watershed processes. These objectives are addressed using paired air and water temperature data collected at relatively high spatial resolution in mountain headwater streams from two different climatic regions of the United States (U.S.) (Pacific Northwest and Mid-Atlantic), along with data from streams of generally larger size distributed across the contiguous U.S. The use of the latter, national dataset, allows for a comparison between the paired air-water annual temperature signals method presented here and another multi-year thermal classification method that used the same sites (Maheu et al., 2016). This previously developed stream thermal regime classification system, using identified key environmental drivers, was based exclusively on water temperature data. Their results showed that sites could be clustered based on differences in the annual water temperature mean, amplitude, and phase parameters, and that these clusters could be predicted by air temperature and flow characteristics. We hypothesize that the inclusion of the comparison of annual water temperature patterns to local annual air temperature patterns in the present analysis will improve the ability to predict processes such as GW exchange and riparian vegetation characteristics that may vary across steep gradients or spatial discontinuities.