INTRODUCTION

Understanding the effects of climate change on cold-blooded species, such as fish, is crucial for informing conservation. Lotic ecosystems are important to study in a changing climate because they encompass and connect aquatic ecosystems with heterogeneous temperatures, allowing fish to move in search of thermal refuges. For instance, Brook Trout Salvelinus fontinalis and Atlantic Salmon Salmo salar move to riverine coldwater refuges (e.g., groundwater springs, hyporheic discharge points, and tributary confluences) to reduce metabolic demands amid warm temperatures (Linnansaari et al. 2023). Thermal refuges protect coldwater and coolwater fish from suboptimal or lethal temperatures and promote population resilience, particularly in summer, making it important to study these habitats in a changing climate (Wilbur et al. 2020; Sullivan et al. 2021).

The dynamic nature and diverse habitats of rivers demand life history adaptations that can make organisms vulnerable to rapid environmental changes (Beamesderfer and Farr 1997). For example, Gulf Sturgeon Acipenser desotoi have life history and behavioral characteristics (e.g., high early life mortality, delayed maturation, anadromy, and seasonally restricted feeding) that complicate population estimates and render them susceptible to anthropogenic stressors (Beamesderfer and Farr 1997; Pine and Martell 2009). Despite an abundance of Gulf Sturgeon across Florida in the late 19th century, commercial fisheries were depleted after only three or four fishing seasons (U.S. Commission on Fish and Fisheries 1902), with population biomass decreasing by 90% (Pine and Martell 2009). By September 30, 1991, the species was listed as threatened under the U.S. Endangered Species Act (U.S. Fish and Wildlife Service [USFWS] and Gulf States Marine Fisheries Commission [GSMFC] 1995). Compounding their life history and behavioral vulnerabilities, Gulf Sturgeon may be particularly sensitive to the effects of climate change, inhabiting a latitudinally compressed range with limited opportunity to move north in search of cooler water (USFWS and GSMFC 1995). Such geographic susceptibility, combined with the species' anadromy (Huff 1975), makes it important to understand how Gulf Sturgeon use riverine thermal habitats to inform conservation efforts in a changing climate.

The relatively warm temperatures of subtropical rivers would be expected to increase the metabolic rate, energy demands, and food consumption of Gulf Sturgeon in summer. However, studies of their food habits demonstrate that Gulf Sturgeon feed minimally, if at all, during freshwater occupancy (Mason and Clugston 1993; USFWS and National Marine Fisheries Service [NMFS] 1995). Combined with taxing migrations to and within rivers in the northern Gulf of Mexico (GOM), warm water temperatures and minimal feeding generate a need for energetic refuge in summer. In northern GOM rivers, such as the Choctawhatchee and Suwannee rivers, Gulf Sturgeon form summer aggregations in habitats that typically encompass or are directly downstream from springs (Chapman and Carr 1995; Carr et al. 1996; Foster and Clugston 1997; Hightower et al. 2002) that provide cool, constant-temperature (~21°C) water (Rosenau et al. 1977). The scale at which springs may provide thermal refuge is uncertain, as Gulf Sturgeon are typically not observed on top of springs (Clugston et al. 1995; Foster and Clugston 1997). Indeed, there is little information on mechanisms explaining why large groups of Gulf Sturgeon use small areas near springs. Hypothesized reasons include cooling of water temperatures in the vicinity of springs, more localized influences of cool water, other habitat-related spring effects (e.g., deep scour holes serving as energetic refuges; Sulak and Clugston 1999), or some combination of these or related factors. Such mechanisms have not been substantiated in the field despite decades of population monitoring. Summer is an important fieldwork season to target, capture, mark, and monitor Gulf Sturgeon because the fish are located and aggregated in rivers at this time (Hightower et al. 2002). Current summer fieldwork involves tools such as gill netting, mark–recapture, and acoustic telemetry to study population dynamics, status, trends, and juvenile production as well as development of electronic logbooks, Bluetooth-enabled tag readers, and a modern tagging database (USFWS and NMFS 2022). However, the thermal suitability of summer fieldwork and the thermal quality of summer aggregation areas may shift with climate change, necessitating water temperature research to inform Gulf Sturgeon management.

Despite the potential role of springs in influencing river temperature or morphology in summer aggregation areas, other factors affect how Gulf Sturgeon use habitats throughout their life history. For instance, Gulf Sturgeon movements have been linked to increased precipitation and river discharge (Chapman and Carr 1995; Parauka et al. 2001). Because precipitation and groundwater affect river discharge and water temperature (Kurylyk et al. 2013; Merriam et al. 2017), these variables may independently or interactively influence Gulf Sturgeon thermal habitat suitability for normal physiological functioning and fieldwork (e.g., netting and tagging; USFWS and GSMFC 1995; Kahn and Mohead 2010). However, riverine water temperature models used by fisheries managers typically account for air–water temperature relationships as opposed to precipitation dynamics (e.g., magnitude and intensity), groundwater attributes (e.g., amount, temperature, and thermal sensitivity [TS]), or their thermal effects in a changing climate. Hence, there is a need for water temperature models that integrate the effects of air temperature, precipitation, and groundwater in ways that facilitate and improve management of riverine thermal habitats and fish populations, including Gulf Sturgeon and other species elsewhere. For instance, models that (1) quantify the thermal influence (TI) and TS (i.e., vulnerability to warming) of precipitation and groundwater and (2) predict how they may change under various climate change scenarios would enable fisheries managers to locate and protect important thermal habitats that cannot be identified through the use of air–water temperature models (Carlson et al. 2019). Such advancements are important for conservation of Gulf Sturgeon, which are summer occupants of coastal rivers, where precipitation and groundwater dynamics have not been thoroughly studied in the context of climate change or threatened-species management.

The Choctawhatchee River is an important system for Gulf Sturgeon summer residency (USFWS and GSMFC 1995; Fox et al. 2000; Hightower et al. 2002), although the species is understudied there compared to in rivers such as the Suwannee River (Foster and Clugston 1997; Sulak and Clugston 1999). Heavily influenced by springs (Northwest Florida Water Management District [NWFWMD] 2017) that may affect summer habitat use in northern GOM rivers (Chapman and Carr 1995; Carr et al. 1996), the Choctawhatchee River nonetheless has thermal habitats and associated water temperature drivers (e.g., precipitation and groundwater) that have not been directly investigated relative to Gulf Sturgeon habitat suitability. For these reasons, along with the river's known anthropogenic stressors (e.g., groundwater withdrawal and agricultural runoff; Hightower et al. 2002), the Choctawhatchee River is an ideal system for addressing our goal of developing an accurate, cost-effective approach to predicting how climate change may affect water temperatures and fish thermal habitat suitability that can be applied to diverse rivers, species, and fisheries management contexts. To generate such an approach, we focused on thermal habitat suitability for normal physiological functioning of and fieldwork involving Gulf Sturgeon. Our first objective was to identify the best performing water temperature models by comparing the accuracy of models that account for air temperature only, air temperature and precipitation, air temperature and groundwater, or all three variables. Our second objective was to use the highest performing models to quantify and forecast water temperatures and thermal habitat suitability—within and beyond Gulf Sturgeon summer aggregation areas—in wide-ranging climate change scenarios and thereby develop an approach that can be applied to other species and systems.

METHODS

Measuring water temperature, air temperature, and precipitation

To provide the information needed to address objectives 1 and 2, hourly water temperatures were measured every day during May–August (2021 and 2022) in five reaches of the Choctawhatchee River (two holding areas and three nonholding areas; Figure 1). Encompassing a warm, wet part of the year in the study area (Powell 2022, 2023), these months overlap a time period when predicted changes in climate (e.g., warming air and water temperatures, changing precipitation, and changing groundwater dynamics) would exert an important influence on the availability and quality of Gulf Sturgeon thermal habitats. Water temperature was measured using high-accuracy (±0.2°C), low-drift (0.1°C annually) HOBO Pro v2 temperature loggers (Onset Computer Corporation 2018; one logger per reach) that were attached to stationary, durable acoustic telemetry receivers (used for a separate study), which were anchored to the river bottom and affixed to sturdy shoreline trees with steel cable. The river bottom was selected because it was continuously inundated and structurally reliable and because Gulf Sturgeon generally use these depths rather than areas near the river surface. Loggers were protected from debris by using polyvinyl chloride (PVC) tubes with aeration holes.

Given that air temperature can be an important driver of water temperature, hourly air temperatures were measured every day in May–August (2021 and 2022) by attaching HOBO Pro v2 temperature loggers to shoreline structures (e.g., trees and fence posts) as close as possible to water temperature loggers. Housed in PVC tubes with aeration holes, the air temperature loggers were affixed to structures above the water level to minimize the potential for inundation during high-water events. Financial limitations prevented deployment of air temperature loggers at every water temperature monitoring location. In 2021, air temperature loggers were placed in the three middle reaches of the river; air temperatures for the other two water temperature monitoring locations were derived from the reach nearest to each (i.e., 6.6 and 7.2 km away, respectively). After the first summer, some air temperature loggers were affected by human interference (e.g., theft and other indications of potential tampering) or deployment error (e.g., insecure fasteners leading to logger movement, increased shading, and inundation). Hence, subsequent air temperatures were obtained from one reach with reliable data in summer 2022, when the mean distance between water and air temperature loggers was 6.8 km, compared to 2.8 km in summer 2021. These distances were less than 15.9 km, a distance that did not reduce model accuracy compared to site-specific models in previous research (Carlson et al. 2020).

Water and air temperature data were downloaded periodically throughout the study to avoid potential data loss due to extreme weather or other unexpected events. Daily precipitation measurements were obtained from Choctawhatchee River basin rain gauges within the Community Collaborative Rain, Hail, and Snow (CoCoRaHS) Network database (https://www.cocorahs.org/). Precipitation was summarized as cumulative daily precipitation (PR) because it is a comparatively unexplored precipitation metric that incorporates the effects of other metrics (e.g., precipitation frequency and intensity) and promotes straightforward data collection (Carlson et al. 2020).

RESULTS

Objective 1

Modeling water temperature

Traditional air–water temperature (i.e., MDAT-only) models were generally low performing and inaccurate. Across the 40 reach–month–year groups, MDAT-only models were least parsimonious (i.e., highest AIC_c value) in 48% of groups (N = 19), least or second-least parsimonious in 90% of groups (N = 36), and never the most parsimonious. Models including PR or ADD outperformed MDAT-only models by better predicting MDWT (Figure 2). In the majority of reach–month–year groups, MDWT was best modeled using MDAT + PR + ADD formulations (80% of groups; N = 32), followed by MDAT + ADD (12.5%; N = 5) and MDAT + PR (7.5%; N = 3) formulations (Table 1). These formulations improved the adjusted R² values by between +0.09 and +0.75 (mean improvement = +0.45) compared to MDAT-only models in the same reach–month–year group (Table 1). The MDAT-only models had an average adjusted R² of 0.28, whereas MDAT + PR + ADD models averaged 0.76, MDAT + PR models averaged 0.50, and MDAT + ADD models averaged 0.47. Differences in accuracy among model formulations were statistically significant (KW test: χ² = 61.52, df = 3, p < 0.001); MDAT + PR + ADD models were more accurate (Z = 4.11–7.77, p < 0.001) and MDAT-only models were less accurate (Z = 2.88–7.77, p ≤ 0.024) than the other three model formulations (Figure 3). The MDAT + PR and MDAT + ADD models had intermediate adjusted R² values that were statistically similar to each other (Z = 1.00, p = 0.99).

TABLE 1. Parameters of water temperature models developed for the Choctawhatchee River, Florida. Year represents the baseline weather conditions used to construct the model. Abbreviations include model intercept (Int); coefficients for mean daily air temperature (MDAT), accumulated degree-days above mean summer air temperature (ADD), and cumulative daily precipitation since May 1 (PR); p-value; adjusted R² value (Adj R²); improvement in adjusted R² compared to the MDAT-only model (ΔAdj R²); and normalized root mean square error (NRMSE).

Reach	Month	Year	Int	MDAT	ADD	PR	p	Adj R2	ΔAdj R2	NRMSE
1	May	Cool, wet	21.874	1.314 × 10−2	0.265	−1.361	<0.01	0.95	+0.66	0.021
		Warm, wet	19.063	0.269	0.212	−0.772	<0.01	0.68	+0.47	0.023
	Jun	Cool, wet	24.992	5.994 × 10−2	3.254 × 10−2	−0.255	<0.01	0.88	+0.58	0.008
		Warm, wet	15.004	0.441	5.631 × 10−2	–	<0.01	0.76	+0.37	0.031
	Jul	Cool, wet	23.578	0.122	7.466 × 10−2	−0.110	<0.01	0.92	+0.32	0.010
		Warm, wet	17.939	0.466	0.103	−0.218	<0.01	0.48	+0.09	0.027
	Aug	Cool, wet	39.832	−3.570 × 10−2	3.597 × 10−2	−0.509	<0.01	0.58	+0.57	0.021
		Warm, wet	28.286	0.170	0.108	−0.210	<0.01	0.45	+0.42	0.018
2	May	Cool, wet	21.799	1.677 × 10−2	0.269	−1.365	<0.01	0.94	+0.66	0.023
		Warm, wet	20.524	0.213	0.132	−0.530	<0.01	0.75	+0.57	0.021
	Jun	Cool, wet	26.337	4.768 × 10−3	3.540 × 10−2	−0.268	<0.01	0.86	+0.63	0.009
		Warm, wet	15.364	0.429	5.488 × 10−2	–	<0.01	0.73	+0.36	0.032
	Jul	Cool, wet	24.077	0.107	8.004 × 10−2	−0.122	<0.01	0.90	+0.33	0.011
		Warm, wet	22.514	0.506	0.201	−0.498	<0.01	0.68	+0.30	0.023
	Aug	Cool, wet	42.534	−6.867 × 10−2	4.914 × 10−2	−0.593	<0.01	0.54	+0.53	0.024
		Warm, wet	33.088	0.125	0.240	−0.317	<0.01	0.56	+0.55	0.016
3	May	Cool, wet	18.707	0.140	0.228	−0.450	<0.01	0.93	+0.51	0.024
		Warm, wet	19.721	0.239	0.117	−0.496	<0.01	0.76	+0.55	0.021
	Jun	Cool, wet	25.017	5.918 × 10−2	2.622 × 10−2	−0.187	<0.01	0.87	+0.53	0.008
		Warm, wet	14.354	0.465	4.795 × 10−2	–	<0.01	0.72	+0.29	0.031
	Jul	Cool, wet	26.097	7.218 × 10−2	8.161 × 10−2	−0.164	<0.01	0.91	+0.33	0.011
		Warm, wet	21.080	0.507	0.173	−0.426	<0.01	0.67	+0.23	0.022
	Aug	Cool, wet	31.430	−8.596 × 10−3	–	−0.158	<0.01	0.43	+0.42	0.020
		Warm, wet	31.650	0.170	0.254	−0.318	<0.01	0.61	+0.54	0.015
4	May	Cool, wet	22.459	−2.273 × 10−2	0.236	−0.592	<0.01	0.96	+0.75	0.018
		Warm, wet	19.263	0.256	0.117	−0.502	<0.01	0.72	+0.51	0.023
	Jun	Cool, wet	25.128	4.884 × 10−2	3.068 × 10−2	−0.210	<0.01	0.88	+0.58	0.008
		Warm, wet	14.137	0.471	4.958 × 10−2	–	<0.01	0.73	+0.30	0.031
	Jul	Cool, wet	26.331	4.086 × 10−2	8.490 × 10−2	−0.147	<0.01	0.94	+0.38	0.009
		Warm, wet	20.295	0.554	0.187	−0.460	<0.01	0.67	+0.23	0.024
	Aug	Cool, wet	35.249	−0.103	–	−0.216	<0.01	0.48	+0.47	0.025
		Warm, wet	33.173	0.155	0.290	−0.362	<0.01	0.66	+0.61	0.015
5	May	Cool, wet	19.461	0.106	0.240	−0.552	<0.01	0.95	+0.61	0.023
		Warm, wet	17.694	0.322	0.144	−0.560	<0.01	0.60	+0.38	0.033
	Jun	Cool, wet	25.426	4.200 × 10−2	2.851 × 10−2	−0.195	<0.01	0.81	+0.53	0.010
		Warm, wet	14.096	0.482	6.092 × 10−2	–	<0.01	0.77	+0.37	0.032
	Jul	Cool, wet	24.677	0.152	8.066 × 10−2	−0.193	<0.01	0.90	+0.29	0.012
		Warm, wet	19.150	0.615	0.194	−0.465	<0.01	0.70	+0.21	0.024
	Aug	Cool, wet	30.385	2.751 × 10−2	–	−0.147	<0.01	0.38	+0.37	0.021
		Warm, wet	36.324	0.108	0.308	−0.410	<0.01	0.65	+0.64	0.016

The air–water temperature correlation was significantly lower in holding areas than in non-holding areas across all months studied (t = 1.69, p = 0.05). Hence, summer water temperatures in areas where Gulf Sturgeon aggregated were driven by factors other than air temperature (e.g., precipitation and groundwater) to a greater degree than water temperatures in non-holding areas, which were more influenced by air temperature. Holding areas were cooler than non-holding areas in 2021 (t = 2.91, p < 0.01) and 2022 (t = 3.49, p < 0.001). Holding areas were consistently cooler than non-holding areas during the warmest summer months (June, July, and August) in both years (t = 2.24–7.20, p = 7.43 × 10⁻¹³ to 0.03), but May temperatures were equivalent in these habitat types (t = 0.20–0.45, p = 0.65–0.84).

Modeling thermal sensitivity and thermal influence of precipitation and groundwater

As TS_pr and TS_gw increased from 0.00 to 0.33, 0.66, and 1.00, the temperature of the Choctawhatchee River also increased. When precipitation and groundwater were thermally insensitive (TS_pr and TS_gw = 0.0), the mean change in water temperature across reaches was +0.20°C for each 1°C increase in air temperature. As precipitation and groundwater became more thermally responsive (TS_pr and TS_gw = 0.33, 0.66, and 1.00), water temperature increased by an average of +0.40, +0.60, and +0.81°C, respectively, per 1°C increase in air temperature. All water temperature increases were statistically significant (KW test: χ² = 93.48, df = 3, p < 0.001; pairwise comparisons: Z = 2.84–9.06, p = 7.83 × 10⁻¹⁹ to 0.03).

In holding areas, the same trend was observed: across the four levels of precipitation and groundwater warming, mean TS increased significantly from 0.17 to 0.38, 0.60, and 0.81 (ANOVA: F_{3, 60} = 25.4, p < 0.001; pairwise comparisons: p = 9.20 × 10⁻¹¹ to 0.05). However, in non-holding areas, changes in TS were not significant when moving from the first (0.21) to the second (0.40; KW test: χ² = 54.77, df = 3, p = 0.18) or from the third (0.60) to the fourth (0.80; p = 0.25) levels of precipitation and groundwater warming. The TS of holding areas reached a higher peak and exhibited a greater overall increase (+0.64) compared to non-holding areas (+0.59) as precipitation and groundwater warmed. The TI of precipitation and groundwater was measurably greater in holding areas than in non-holding areas during May (holding: +0.61°C; non-holding: +0.54°C), June (holding: +0.71°C; non-holding: +0.66°C), July (holding: +0.59°C; non-holding: +0.52°C), and August (holding: +0.67°C; non-holding: +0.65°C), but the difference was statistically significant only in July (t = 2.33, p = 0.05).

Objective 2

Predicting water temperature

In cool, wet conditions, projected water temperatures increased from May to June and from June to July, remaining relatively stable from July to August (Figures 4 and 5). In all months, water temperatures warmed as precipitation and groundwater became more thermally sensitive within MATW increments (Table A.1). However, the magnitude of warming across TS values increased as MATW intensified from +1°C (mean warming = 0.79°C across months) to +4°C (mean warming = 3.17°C; Figures 4 and 5).

In warm, wet conditions, predicted water temperatures in May–August were warmer and temporal trends in warming were different than those in cool, wet conditions (Tables A.1 and A.2; Figures 6 and 7). However, the overall pattern—increased water temperature warming with increased intensity of simulated climate change—was similar regardless of weather conditions. In warm, wet conditions, projected water temperatures warmed as the TS_pr and TS_gw increased within MATW increments (Figures 6 and 7). Similarly, the magnitude of warming across TS values increased as MATW intensified, ranging from a mean of 0.43°C (MATW = +1°C) to 1.72°C (MATW = +4°C; Table A.2; Figures 6 and 7).

Predicting thermal habitat suitability

Observed water temperatures in cool, wet conditions were below the temperature thresholds for Gulf Sturgeon surgery, sampling, and elevated physiological stress during all months (Table A.3). Future May water temperatures were predicted to be suitable for surgery, sampling, and normal physiology under all 16 climate change scenarios (Table A.3; Figure 4). Predicted June water temperatures were too warm for surgery in 20–100% of reaches under seven of the warmest scenarios, too warm for sampling in all reaches under the warmest scenario (Table A.3), and below temperature thresholds for elevated physiological stress under all scenarios (Figure 4). Projected July water temperatures, which were warmer than those in May and June, were too warm for surgery in 40–100% of reaches under 10 climate change scenarios and were too warm for sampling in 20–100% of reaches under six scenarios (Table A.3). July water temperatures were predicted to elevate Gulf Sturgeon physiological stress in all reaches under the warmest scenario (Table A.3; Figure 5). Similarly, predicted August water temperatures were too warm for surgery in 60–100% of reaches under 12 climate change scenarios and were too warm for sampling in 20–100% of reaches under six scenarios, elevating physiological stress in 60% of reaches under the warmest scenario (Table A.3; Figure 5).

Compared to 2021, the warm, wet weather of 2022 produced less suitable conditions for Gulf Sturgeon surgery, sampling, and normal physiology. Observed 2022 water temperatures exceeded the temperature threshold for surgery in 80–100% of reaches during June, July, and August and were too warm for sampling in 20% of reaches during June and July (Table A.4). Future May water temperatures were projected to be too warm for surgery in 20–80% of reaches under two of the warmest climate change scenarios (Table A.4) but were suitable for Gulf Sturgeon sampling and normal physiology under all 16 scenarios (Figure 6). Notably warm weather in June 2022 generated models in which future June water temperatures were predicted to be too warm for surgery and sampling in all reaches under all climate change scenarios, causing elevated physiological stress in 20–100% of reaches under nine of the warmest scenarios (Table A.4; Figure 6). Projected July water temperatures were too warm for surgery and sampling in all reaches under all climate change scenarios, although they were invariably less than 31°C and thus predicted to cause less physiological stress than June temperatures in corresponding scenarios (Table A.4; Figure 7). Predicted August water temperatures were too warm for surgery in all reaches under all climate change scenarios, and they were too warm for sampling in 20–100% of reaches under 10 scenarios, although elevated physiological stress was only expected in 20% of reaches under the warmest scenario (Table A.4; Figure 7).

DISCUSSION

Projected changes in air temperature, precipitation, and groundwater dynamics are likely to increase water temperatures in a river occupied by threatened Gulf Sturgeon. Predicted increases in water temperature varied subtly to considerably among reaches, months, years, weather conditions, and simulated climate change scenarios. However, the predominant warming trend indicates that Gulf Sturgeon in the Choctawhatchee River are likely to face new warmer; and, in some cases, less physiologically favorable thermal habitat conditions in a changing climate.

We found that the accuracy of water temperature models improved with the inclusion of precipitation and groundwater. The improvement in adjusted R² (mean improvement = +0.45; maximum = +0.75) for precipitation- and groundwater-corrected models indicates that precipitation and groundwater are key drivers of water temperature that shape the spatiotemporal structure of thermal habitats for Gulf Sturgeon in the Choctawhatchee River. In an important application for fisheries management, our research illustrates the utility of straightforward adjustments to air–water temperature models using inexpensive, readily available, or easily collectable rainfall and air temperature data while expanding the geographic area and management contexts in which precipitation- and groundwater-corrected models have proven useful. Previous studies have used this approach to account for spatiotemporal heterogeneity in groundwater (Snyder et al. 2015) or precipitation and groundwater (Carlson et al. 2019, 2020) as thermal habitat drivers for coldwater fishes (e.g., Brook Trout and Brown Trout Salmo trutta) in Virginia and Michigan. However, the approach has not been evaluated in the Southeast or in the context of threatened-species management until this study.

Florida exhibits considerable hydrological and climatic diversity; river hydrology is driven by groundwater discharge in some areas, and precipitation is relatively abundant compared to that in other U.S. states (PRISM Climate Group 2023) but is highly patchy throughout Florida (Powell 2022, 2023). Given these conditions, it is reasonable that precipitation- and groundwater-corrected air–water temperature models are well suited for fisheries management in Florida rivers. In contrast, these models would likely exhibit lower performance in western U.S. states, where factors such as elevation change and snowmelt have a greater influence on water temperature dynamics and for which mathematical approaches that incorporate these variables (e.g., spatial network models) have performed well (Peterson and Ver Hoef 2010). Our study expands the scope of river temperature research in Florida and the Southeast to include an approach that integrates the management-relevant efficiency and cost effectiveness of air–water temperature models with the reliability and exactitude of hydrometeorological models while minimizing their disadvantages (e.g., low accuracy and low management applicability, respectively; Carlson et al. 2019). The prevalence and patchiness of precipitation in this region render our models particularly informative for advancing the integration of water temperature dynamics into fisheries management programs. The cooling effect of precipitation was universal across reaches, months, and years but was spatially variable in intensity, which is important information for managing Gulf Sturgeon and other species within and beyond the Choctawhatchee River. Our approach for incorporating precipitation and groundwater into air–water temperature models is transferable to fisheries management scenarios in rivers and streams where air temperature, groundwater, and precipitation such as rain are major drivers of water temperature.

Previous researchers have observed summer aggregations of Gulf Sturgeon in small sections of northern GOM rivers (Chapman and Carr 1995; Foster and Clugston 1997), describing possible reasons for this pattern (e.g., access to spring water and hydrodynamic protection; Chapman and Carr 1995; Carr et al. 1996; Sulak and Clugston 1999). We found that non-holding areas were warmer and more thermally influenced by air temperature than holding areas. In contrast, holding areas had a greater TI of precipitation and groundwater and were more sensitive to precipitation and groundwater warming. A heavily used holding area located downstream from the Holmes Creek confluence was considerably cooler in June–August than reaches without Gulf Sturgeon, suggesting that the fish use thermal refuges in the warmest months. These findings are energetically and ecologically reasonable given that precipitation and groundwater inputs are often cooler than ambient temperatures and thus help to cool or offset increases in the water temperature (Rosenau et al. 1977; Merriam et al. 2017), thereby allowing coldwater and coolwater fish to conserve energy (Wilbur et al. 2020; Linnansaari et al. 2023). However, use of habitats that are more thermally buffered by precipitation and groundwater does not imply that Gulf Sturgeon, their summer holding areas, or their management programs will be unaffected by climate change. In fact, our models projected that holding areas would be as vulnerable as or more vulnerable than non-holding areas to warming water temperatures if climate change involves TS_pr or TS_gw greater than 0.0, an outcome that seems more likely than a future involving thermally insensitive precipitation and groundwater.

Spatiotemporal attributes of Gulf Sturgeon movement and summer residency in the Choctawhatchee River proffer important perspectives for interpreting our findings. In early May, fish move downstream from spawning grounds to establish summer residency in holding areas (Fox et al. 2000; Hightower et al. 2002), where, as we found, water temperatures are equivalent to those of non-holding areas for only a few weeks before they become significantly cooler in the warmest period of the summer. Movement to these about-to-be-cooler sites is noteworthy, representing an energetically efficient preparatory step before the upcoming warm conditions of June, July, and August. The TI of precipitation and groundwater was statistically greater in holding areas than in non-holding areas during July only. Biologically, however, holding areas evidently offer cooler, more favorable habitat conditions throughout the summer, reflecting stronger drivers of cooling—although the TI of precipitation and groundwater was difficult to assess at a small sample size—and perhaps the presence of hydrodynamic refuge or related factors. An early summer synchrony between postspawn movement, warming water temperatures, and holding area thermal buffering may help to explain why Gulf Sturgeon establish and maintain summer residency in these areas and not others. Occupying cool habitats that are influenced by hydrological factors promoting cool water temperatures is energetically and ecologically advantageous. Likewise, locating and establishing residency in these habitats before the heat of summer would enable Gulf Sturgeon to efficiently achieve what previous researchers have suggested is a principal reason for observed patterns in summer habitat use: minimizing energy expenditure during a taxing season with no or very limited feeding (Mason and Clugston 1993; Chapman and Carr 1995; Carr et al. 1996; Sulak and Clugston 1999). Our results are consistent with the notion that holding areas offer thermal refuge by providing favorable temperatures for conserving energy in summer.

Although knowledge gaps about holding areas remain, our results illustrate how drivers of cooling affect these habitats at a critical time of year, when Gulf Sturgeon finish spawning and move to summer residency locations in the Choctawhatchee River. Adding hydrological, thermal, and temporal context to established knowledge about holding areas in this river—including their vulnerability to water quality threats, such as agricultural runoff and wastewater treatment effluents (Hightower et al. 2002)—our findings underscore the importance of studying and protecting these habitats as components of Gulf Sturgeon management programs. Water quality monitoring, riparian and watershed habitat preservation, groundwater conservation (including efforts to reduce groundwater withdrawal in the Choctawhatchee River watershed; Hightower et al. 2002), and related activities (e.g., careful inspection of the watershed's 33,000 septic systems; NWFWMD 2017) could be useful tools for mitigating increases in water temperature and developing partnerships to protect holding areas and the Gulf Sturgeon that rely on them. Importantly, these approaches may also be germane to Gulf Sturgeon management in other northern GOM rivers; the transferability of our models to these systems highlights an opportunity to use the approach to identify when and where particular management strategies may be helpful.

Our results indicate that the intensity of climate change will shape outcomes for Gulf Sturgeon management activities. The present results also indicate that if future changes in climate involve warm, wet conditions, then Choctawhatchee River water temperatures in June and July would be too warm for surgery and sampling in all reaches across the range of projected air temperature increases for the study area (+1–4°C; U.S. Federal Government 2023). However, regardless of prevailing weather conditions, May was predicted to be thermally suitable for surgery and sampling in all but the most extreme climate change scenarios (i.e., MATW = +4°C; TS_pr or TS_gw = 0.66 or 1.00); even then, at least one sampled reach was projected to remain cool enough for surgery and sampling. August was a more nuanced month; it was predicted to be too warm for surgery in warm, wet conditions (all climate change scenarios) and in 50% of cool, wet conditions (MATW = +3°C or +4°C) but was generally suitable for sampling in at least one reach under both cool, wet conditions and warm, wet conditions. Collectively, these findings indicate that identifying thermally appropriate times for Gulf Sturgeon fieldwork activities will be increasingly important and time-restricted amid climate change, with June, July, and some parts of August projected to be thermally unsuitable for these efforts, especially with MATW of +3°C or +4°C (or +2°C with thermally sensitive precipitation or groundwater). Although previous researchers have noted that Gulf Sturgeon aggregation behavior makes summer a logistically strategic season during which to conduct population assessments, despite netting stress (Hightower et al. 2002), our results indicate that climate change will likely amplify stress from hands-on management activities and reduce the amount of time in which Gulf Sturgeon can be safely handled during summer within and beyond the Choctawhatchee River (Kahn and Mohead 2010). However, early summer may provide windows of opportunity for surgery and sampling in a changing climate.

In most months and climate change scenarios, predicted water temperatures were generally suitable for normal physiological functioning of Gulf Sturgeon, particularly in cool, wet weather conditions. However, in one or more reaches, midsummer water temperatures in warm, wet conditions commonly exceeded normal physiology thresholds under moderate to intense climate change scenarios. In these circumstances, the ability of Gulf Sturgeon to maintain equilibrium could conceivably be affected, as could survival. However, Gulf Sturgeon may have thermal resilience strategies that are not addressed herein, and the species has persisted amid changing weather and climatic conditions for over 200 million years (Florida Museum 2024). Moreover, Gulf Sturgeon do not have a formally established upper lethal temperature (Kahn and Mohead 2010), explaining why we used temperature thresholds for related species. Future research is needed to identify the upper lethal temperature of Gulf Sturgeon, both juveniles and adults, due to the importance of each life stage for population sustainability and the possibility of ontogenetic differences in holding area location and general thermal habitat use. Although Hightower et al. (2002) showed that immature and mature Gulf Sturgeon often used the same or nearby holding areas in summer, 14.6% of immature fish used adult non-holding areas in the lower portion of the Choctawhatchee River. Our use of juvenile temperature thresholds to evaluate holding areas with abundant adults was necessary given information gaps, but it is a limitation of the study; life stage-specific temperature thresholds would allow the outcomes of our research to be interpreted with additional clarity. Importantly, our research provides managers with a portfolio of water temperature models for the Choctawhatchee River and, more broadly, a reliable mathematical approach for predicting thermal habitat conditions for numerous rivers throughout Florida and the Southeast. Managers can use our modeling framework to decide when and where to safely target, capture, tag, and monitor Gulf Sturgeon and other species amid climate change.

Our study lays a foundation for research to explore questions not addressed herein. We investigated water temperature dynamics in years with cool, wet conditions and warm, wet conditions; expanding our approach to include other weather conditions (e.g., cool, dry; warm, dry) and extreme events (e.g., record warm temperatures, heavy rains, and hurricanes) will be important for fully understanding how climate change may affect Gulf Sturgeon thermal habitat suitability in the Choctawhatchee River. It is also important to identify potential habitats that Gulf Sturgeon could use if current summer holding areas—given their relatively high vulnerability to climate change (see above)—become less thermally suitable or unsuitable in a changing climate. It is conceivable that Gulf Sturgeon could avoid adverse conditions in current holding areas by moving elsewhere, but there is a shortage of information on the location, connectivity, thermal quality, or potential for repeated, long-term use of such habitats that adds to their importance for present and future Gulf Sturgeon management efforts. The Choctawhatchee River is an important system for Gulf Sturgeon summer residency, but our single-river study encompassed only part of the Gulf Sturgeon range (and only five reaches therein). Evaluating how changes in air temperature, precipitation, and groundwater may affect Gulf Sturgeon thermal habitat suitability in other rivers (e.g., the Apalachicola, Escambia, Pascagoula, and Pearl rivers) will be important for managing the species amid climate change; our modeling approach is readily transferable to those systems. Finally, using our modeling approach to compare the potential effects of climate change on different river populations and size-/age-classes (e.g., juveniles, subadults, and adults) may yield important information for managing Gulf Sturgeon and other species of interest (e.g., invasive species and sport fish) throughout their life span and across their range.

In conclusion, we developed an accurate, cost-effective, management-relevant approach for integrating and modeling water temperature drivers to forecast fish thermal habitat suitability in a changing climate. While incorporating the known importance and variable effects of air temperature on water temperature, our modeling approach accounts for interannual weather variability and the thermal effects of precipitation and groundwater, highlighting their role as cooling sources in a warm, taxing season. We demonstrated that holding areas are cooler than non-holding areas in summer, underscoring the management importance of holding areas while revealing their vulnerability to warming amid climate change. Managers of Gulf Sturgeon and other species can use our precipitation- and groundwater-corrected modeling approach and portfolio of water temperature models to locate key thermal habitats, predict future temperatures, and plan the spatiotemporal logistics of management activities, such as population sampling, mark–recapture, and acoustic telemetry, in a changing climate. By advancing the hydrological and thermal understanding of Gulf Sturgeon habitat use in a subtropical river, our research provides a modeling framework that can be applied to other species and systems to understand, predict, and prepare for the effects of climate change.

CONFLICT OF INTEREST STATEMENT

The authors confirm that this article involves no conflicts of interest.

ETHICS STATEMENT

This research was conducted in accordance with all relevant protocols. This research did not require the handling of fish, so Institutional Animal Care and Use Committee guidelines were not applicable.