1. Introduction
Water temperature has long been recognized as an important aquatic environmental variable [
1,
2,
3] that directly and indirectly affects numerous ecological processes [
4,
5,
6] and as such is regulated in the United States under the Clean Water Act, Section 303 (d) [
7,
8]. Increasing water temperature values and variability are known to induce thermal stress in aquatic species that can affect growth, reproductive success, and mortality [
9,
10,
11,
12]. A recent review of phenology research of aquatic species [
13] also identified water temperature as the most important environmental cue for life history behaviors, particularly spawning migration behavior [
14,
15,
16,
17,
18,
19].
In the last decade, numerous studies have focused on quantifying a stream’s thermal regime and drivers of water temperature variability [
20,
21,
22,
23,
24]. Thermal regime is a term that refers to the stream temperature characteristics and dynamics that we describe based on stream temperature data collected over time [
24]. At local spatial scales, important factors that affect stream temperature include riparian vegetation [
25], hydrology (e.g., discharge, groundwater source volume, and hyporheic exchange) [
1,
26,
27], and locations where tributaries enter the main channel [
1,
28]. Local scale variability related to groundwater and tributary connection are relative to baseflow hydrology according to stream size and volume [
29,
30]. Factors that may affect water temperature variability at the catchment and watershed scales include climate, elevation, and land cover, and geology [
1,
31,
32].
Thermal regimes are sensitive to anthropogenic watershed development that can result in warming and increased variability due to runoff over impervious surfaces and influence of stormwater pipes [
31,
33,
34,
35,
36]. Previous research has documented the relationship between impervious surface cover and greater incidence and magnitude of stormflow events [
37]. These stormflow events can elevate temperatures 3.5 to 7 °C with 3 to 7-h dissipation times, respectively [
38]. In addition, extensive subsurface pipe networks, including stormwater pipes, have been added to developed areas that can transfer stormwater with elevated temperatures directly to streams. These pipes can also indirectly interact with groundwater to affect water temperature and baseflow variability, depending on local site conditions. Direct connection of pipes to groundwater can add a constant, stabilizing baseflow from outflow and leaks [
33,
35].
Land cover within riparian areas and strategic placement of riparian trees can affect shading from solar radiation, heat fluxes in riparian areas, and water temperature variation [
2,
25,
39,
40,
41]. Aside from direct shading effects, riparian trees can also create humid microclimates over streams that can stabilize water temperature variability [
42,
43], with consensus in the literature that riparian tree effects on microclimate generally occur up to about one tree height (15 to 60 m) away from the edge of the stream [
39]. Daily maximum water temperature differences between forested and non-forested stream sites can be 4.2–4.9 °C cooler in forested stream reaches [
44,
45].
While extensive research has been conducted on thermal regulation by riparian tree shading in agricultural and mixed ag-forest watersheds [
39,
46], little is known about the thermal influence of riparian trees in urban watersheds (>15% impervious land cover) [
47,
48,
49,
50]. To our knowledge, there is only one research study currently available [
49] that has investigated the relationship between tree canopy cover in riparian buffers and water temperature within urban catchments, finding no significant relationship between canopy cover and water temperature. Therefore, the aim of this study was to quantify the effects of riparian and loggershed scale variables on water temperatures in stream sites with 4 to 62% impervious land cover within the “loggershed.” We introduce the term “loggershed”, which refers to the watershed of the natural and build network draining into each temperature logger point location. We calculated water temperature metric values for each logger location relevant to aquatic species life history, thermal stress, and critical thermal maximum water temperatures to explore relationships between land cover, human development, and potential water temperature changes within the loggershed.
We hypothesized that: (1) The greater the percentage tree canopy within riparian areas along stream networks at the loggershed scale, the lower the frequency and duration of exceedance of water temperature stress threshold values, magnitude of change in water temperature, and variability of water temperature; (2) The greater the percentage impervious surface within riparian areas along stream networks at the loggershed scale, the greater the frequency and duration of exceedance of water temperature stress threshold values, magnitude of change in water temperature, and variability of water temperature; (3) The greater the percentage tree canopy within the loggershed, the lower the frequency and duration of exceedance of water temperature stress threshold values, magnitude of change in water temperature, and variability of water temperature; (4) The greater the percentage impervious surface within the loggershed, the greater the frequency and duration of exceedance of water temperature stress threshold values, magnitude of change in water temperature, and variability of water temperature; (5) Land cover quantified within wider, 30 m riparian areas along both sides of the stream network at the loggershed scale will have a greater effect on water temperature than narrower, 5 m riparian areas; and (6) The greater the length of stormwater pipes (km) in the loggershed, the greater the water temperature variability.
4. Discussion
To our knowledge, this is the first time that the relationship between stormwater pipes and water temperature metrics has been explored to better understand thermal dynamics in urban watersheds. The results highlight important aspects of thermal habitat quality and water temperature variability for aquatic species living in urban streams based on thermal thresholds relevant to species metabolism, growth, and life history [
9,
10,
78]. Although this study was based on a limited number of study sites from one year of logger data, results highlighted significant negative correlations (19 metrics) between percent of tree canopy in the 5 m riparian area and water temperature.
Our results were consistent with the literature showing a negative effect of urbanization on aquatic organisms, especially sensitive species such as trout, due to thermal stress and unsuitable thermal habitat [
79,
80]. Spring to early summer is an important period for spawning behavior for many fish species, including blacknose dace [
81] and rainbow trout [
82] that occur at sites of this study. Increases in water temperature variability during these months associated with impervious surface cover may result in changes in timing of spawning or changes in emergence and length of adult life stages for aquatic macroinvertebrates [
83,
84]. Additionally, early life stages of fish that typically develop during the spring and early summer require available prey at the right time for survival and growth into the adult stage [
16,
17,
59,
60]. Change in the thermal triggers associated with spawning or emergence can result in seasonal asynchronicity (match mismatch hypothesis) (e.g., [
85]). Using water temperature metrics to monitor thermal habitat conditions and to identify times of the year when life history behaviors, such as timing of spawning migrations may be affected, can be a useful tool for managers to better identify the thermal habitat being impaired [
22,
62,
74,
86]. Our results showed that such as ADMax values for the months of April, May, and June could be used to monitor the effects of thermal stress and water temperature variation on early life stages of fish.
Urbanization and Thermal Degradation Mitigation
Although significant relationships between landcover and thermal metrics were observed, this consistency of significant correlations was not the case for relationships between water temperature and percent tree canopy in the 30 m riparian area (8 metrics), percent impervious cover in the 5 m (3 metrics), or percent impervious cover in the 30 m riparian area (4 metrics). In addition, tree canopy cover at the loggershed scale was the only significant variable retained in the mixed effects model to predict October variability. These results highlighted the importance of investigating other variables that can offset the benefits of riparian trees and influence thermal results in urban streams.
Moreover, this study only found significant correlations between land cover variables in the 5 m and 30 m riparian areas with the April, May, June and October variance and October maximum range water temperature metrics. The only significant candidate mixed effects models for monthly variability were for June, which included impervious surface in the 5 m riparian area and length of stormwater pipes; October maximum range included tree canopy cover at the loggershed scale.
This study showed the greatest values in monthly variance and maximum range in April, May, and June (
Figure 5). This variability in April and May was positively correlated to impervious surface in the 5 m and 30 m riparian area. In addition, the June variability was positively correlated to impervious surface in the 30 m riparian area and significant mixed-effects models included impervious surface in the 5 m riparian area and total length of stormwater pipes in the loggershed. These results are consistent with previous research that documented greater fluctuation in water temperatures in urban streams relative to percent impervious surface cover [
87,
88,
89,
90]. However, in addition to effects of impervious surface cover on water temperature variability, this study identified significant negative correlations between water temperature variability and tree canopy cover. Specifically, this study identified significant negative correlations between April and May water temperature variability and percent tree canopy cover in the 5 m riparian area and between variability in June and percent tree canopy cover in the 30 m riparian area. Results of this study also predicted a reducing effect of tree canopy cover in 5 m riparian areas on May and June water temperatures of ~5–7 °C.
The study found positive significant correlations between the total length of stormwater pipes at the loggershed scale and 8 magnitude water temperature metrics, in loggersheds with up to 5.34 km of pipe to 0.48 km of stream (
Figure 2). Although significant mixed effects models to predict water temperature variability included length of stormwater pipes, this result is based on geospatially extracted data and was not based on direct measurement of stormwater pipe leaks, stormwater inflows, green infrastructure effectiveness for treating thermal pollution from stormwater effects, or groundwater input at study sites. Variability of water temperature is known to be influenced by local, reach scale inputs such as groundwater inputs [
1,
27] and locations where tributaries enter the main channel [
1,
28].
Our results showed that the presence of stormwater pipes could potentially offset the benefit of riparian trees, which highlights the need for further investigation of other important variables which may affect riparian restoration outcomes. Similarly, other underground structures could leak water with different temperatures (e.g., cooler), and mask the effect of features such as impervious surfaces that usually cause thermal degradation. Our results highlighted the importance of understanding connections between specific urban features and thermal regimes. This is especially important for a project aiming to mitigate the impact of urban thermal degradation. Additionally, our studies provided insight on how the scale of the study could also influence the scale of the restoration action and outcomes, as some variables may have a more significant influence over larger scales (e.g., impervious surface). In contrast, others might have a more localized impact within the loggershed space (e.g.,) tree canopy. This study highlighted that urban streams are a complex mosaic of intertwining variables that ultimately influence the thermal regime. More research on the thermal sensitivity for each variable is needed in order to develop more meaningful management and mitigation recommendations for thermal degradation.
5. Conclusions
The most commonly applied thermal cooling best management practice (BMP) is riparian tree planting, a strategy that has been applied in the United States since the 1970s to mitigate the impacts from logging and agriculture [
46,
91,
92,
93]. Considering the results of this study, which found significant correlations and predicted effects of stormwater pipes, impervious surface, and riparian tree canopy cover on water temperature, further research is needed to identify additional urban variables of importance and if riparian tree canopy cover could still be used to mitigate increases in water temperature in urban catchments [
50,
94,
95]. Future studies using multiple years of water temperature data at the loggershed scale and on-the-ground surveys of tree species within riparian area widths 5 to 30 m and greater, in addition to geospatial data, would be helpful to further investigate characteristics of riparian areas that affect thermal regime along urban streams.
In urban areas with stormwater pipe networks such as the sites in this study with 10 out of 14 sites, ≥4 to 6 km of pipes per 1 km of stream, the local influence of stormwater pipe outlets on water temperature is likely [
96]. However, our study design was set up with a logger at every ~50 m and geospatial data were extracted at the loggershed scale. This study design was not set up to effectively quantify the effects of every tributary and stormwater pipe outlet, which would require a more extensive network of loggers upstream and downstream of each tributary confluence and stormwater pipe input longitudinally throughout the stream network. In addition, the baseflow index was estimated using USGS gage watershed locations at a scale too large to quantify local groundwater inputs effectively [
97]. Riparian vegetation is also known to influence local variability [
25], and local reach scale, transect based surveys of tree canopy cover, solar radiation, and tree species influence on local hydrology and microclimate may offer insights into local water temperature variability [
42,
98].
Our results confirmed that one of the greatest impacts of urbanization for aquatic species is the induced thermal stress. The extent of thermal stress for aquatic species depends on the availability of habitat with temperatures below thermal stress temperature thresholds and the ability of species to disperse to those habitats. This highlight the importance of conservation and creating coldwater refugia within connected stream networks that would offer refuge during times of thermal stress [
28,
99,
100,
101]. Previous studies have documented coldwater refugia in highly urbanized watersheds (17% impervious) where groundwater enters the stream [
102,
103]. Further direct measurement of groundwater availability within stream networks at the loggershed scale and catchment (HUC 12) scale may offer opportunities to identify other urban thermal refugia locations, keeping in mind that groundwater can be an important source of water temperature variability at the local reach scale. Understanding how managers can use cold water refuges to create a mosaic of thermal habitat for fish to thermoregulate may help prevent urban areas from becoming thermal barriers to dispersal throughout stream networks.