The Heat Strain of Various Athletic Surfaces: A Comparison Between Observed and Modeled Wet-Bulb Globe Temperatures

We conducted an observational study to evaluate the concordance of meteorologic data among athletic surfaces with varying physical properties and how these athletic surface data agreed with meteorologic data from the closest regional NWS ASOS. The athletic surfaces in this study represent modern athletic surfaces for the most common collegiate and high school sports. Athletic surfaces consisted of red and black all-weather tracks (track), blue and black hard tennis courts (tennis), nylon-knit artificial green turf (AstroTurf; SportGroup, Burgheim, Germany), green synthetic turfgrass with black rubber pellets (synthetic turfgrass), volleyball sand, softball clay, natural grass (grass), and a natural lake (water). The blue tennis court was a colored concrete mix, and the black tennis court was an asphalt construction. The starting location for data collection was counterbalanced among data-collection days across surfaces to control for the influence of time of day and duration of solar radiation on athletic surfaces.

Meteorologic measurements were taken over 18 days from 2012 to 2014 in the northeastern United States from May through September when solar zenith angles were equal to or less than 45°. Data were collected between 1 pm and 4:30 pm to approximate outdoor practice and competition times and in direct sunlight on days when the NWS-reported air temperature was 22.5°C to 35.6°C. Each day, data were obtained at the same location on each athletic playing surface. Measurements were not taken if the surface was being repaired at the time of measurement (n = 2), if the WBGT device malfunctioned (n = 4), or in inclement weather (n = 10), resulting in a total of 143 data pairs (local and NWS data points).

Initial measurements were collected 0.61 m above the ground to determine if playing surface proximity affected meteorologic measures due to radiative heat gain from the surface and 1.2 m above the ground to approximate the environmental stress experienced by athletes at chest level.²³  Preliminary analysis of WBGT values between heights (0.61 versus 1.2 m) revealed no differences (P > .05) and a strong intraclass correlation coefficient (ICC; 2,1) among surfaces (ICC [2,1] ≥ 0.92). Therefore, only data from 1.2 m were subsequently analyzed because this height approximates the environmental stress experienced at the torso,²³  with the exception of water, which was measured 0.61 m above the water surface to mimic the environmental stress of crew athletes.

We collected onsite meteorologic data using 2 WBGT devices (Kestrel 4600 Heat Stress Tracker; Nielsen-Kellerman, Boothwyn, PA) and reported the averages. Kestrel WBGT devices measure DB directly using an externally mounted, hermetically sealed thermistor. The GT is estimated using a hermetically sealed 25.4-mm black-globe temperature corrected to a standard 150-mm black globe, air velocity, DB, and emissivity.²⁴  We estimated natural WB from shaded DB and calculated pressure by the piezoresistive effect, relative humidity by electronic capacitance, and air velocity by vane anemometer according to Bernard and Pourmoghani.²⁵  Manufacturer-reported accuracy is as follows: wind speed ±3%, air temperature ±0.5°C, GT ±1.4°C, relative humidity ±2%, WB ±0.8°C, and WBGT ±1.8°C.

The WBGT devices were calibrated according to the manufacturer's instructions, and data were recorded after meteorologic values stabilized (at least 7 minutes). High reliability among devices was observed for air temperature (ICC [2,1] = 0.97), relative humidity (ICC [2,1] = 0.96), WB (ICC [2,1] = 0.96), GT (ICC [2,1] = 0.91), DB (ICC [2,1] = 0.99), and WBGT (ICC [2,1] = 0.99). Accounting for travel between athletic surfaces and WBGT-device recalibration on each athletic surface, the average interval between measures was 17.8 ± 3.4 minutes.

The WBGT was modeled using the algorithm of Liljegren et al,¹⁶  which has repeatedly been shown to provide the most valid assessment of outdoor WBGT.^17,19  The model requires a variety of meteorologic input variables. Air temperature, relative humidity, and wind speed were obtained from the NWS ASOS located at the Willimantic-Windham Airport in Windham, Connecticut (41.74°N, 72.18°W). This station ranged from 8.6 to 12.7 km away from the athletic surfaces (Figure 1). The observing station used a standardized setup and equipment to record meteorologic measurements, with thermometer and hygrometer readings taken 1.5 m above a grass-covered surface. Wind speeds were measured at a height of 10 m and were mathematically corrected to 2 m to approximate onsite measurement height. Finally, hourly solar radiation values were obtained from the National Solar Radiation Database.²⁶  The WBGT was modeled at an hourly resolution to match that of the solar-radiation data. Observed WBGT was matched to the nearest modeled hourly WBGT, with an average difference of 15 ± 8 minutes between observed and modeled WBGTs. Throughout this article, the modeled WBGT will be referred to as NWS WBGT.

The modified WBGT heat-safety category cutoffs created by Grundstein et al¹³  were used to classify heat-safety thresholds. These cutoffs, based on the Georgia High School Association guidelines, are thought to conservatively adjust for regional climate differences and subsequent heat acclimatization. These adjusted heat-safety category cutoffs, therefore, represent a more robust reflection of traditional heat-safety guidelines.

Mean differences among the various athletic playing surfaces were initially calculated. We used a priori dependent-samples t tests to evaluate WBGT differences among the various athletic playing surfaces when the mean difference was equal to or greater than 0.5°C WBGT. Bland-Altman plots were used to assess WBGT agreement between local and NWS WBGT measurements among the various athletic surfaces. A χ² analysis was used to determine agreement of activity guidelines among local and NWS WBGT measurements. To characterize the magnitude of difference between WBGT measures, we calculated mean differences (athletic surface – NWS). Absolute mean difference (all athletic surfaces – NWS) was calculated to determine differences regardless of direction (overestimation or underestimation). Data are expressed as means ± standard deviations, percentages, mean differences, or ranges. We analyzed the data using SPSS (version 21; IBM Corp, Armonk, NY). The α level was set at .05.

The onsite WBGT recordings were greater than the NWS readings for black tennis (28.48°C ± 3.11°C versus 26.80°C ± 2.60°C, respectively; P = .001), black track (27.74°C ± 3.23°C versus 26.63°C ± 2.68°C, respectively; P = .041), red track (27.81°C ± 3.12°C versus 26.74°C ± 3.11°C, respectively; P = .02), and volleyball sand (28.13°C ± 2.77°C versus 27.14°C ± 3.01°C, respectively; P = .05) surfaces. In contrast, water WBGT measurements (25.89°C ± 3.03°C) were lower than the NWS measurements (26.91°C ± 2.81°C; P = .004). No other difference between athletic surface and NWS WBGT was observed. The distance between each athletic surface and the weather station did not appear to affect WBGT agreement, as differences were observed at 8.4 and 11.4 km but not at 12.7 km. Across all athletic playing surfaces, NWS underestimated athletic surface WBGT measurements (26.85°C ± 2.93°C versus 27.52°C ± 3.13°C, respectively; mean difference = 0.67°C ± 2.02°C; P < .001). As expected, the absolute mean difference between the NWS and onsite WBGT was high (1.71°C ± 1.32°C). The range of difference scores was wide (−4.42°C to 6.14°C).

We observed no WBGT differences between similar colors or types of athletic surfaces (black versus blue tennis, red versus black track, and grass versus synthetic turfgrass or AstroTurf; Table 1). The WBGT measurements for water were consistently lower than those for all other athletic surfaces by 1.40°C to 2.70°C WBGT (P = .04). This difference in WBGT can be explained by lower GT for water (37.16°C ± 3.92°C) than for the aggregate of all other surfaces (42.06°C ± 3.98°C; P < .001). The large WBGT measurement variability among athletic playing surfaces is shown in Table 1.

Across all athletic playing surfaces, the WBGT differences between the NWS and onsite measurements were explained by higher onsite WB (22.28°C ± 2.79°C versus 23.84°C ± 3.16°C, respectively; P < .01) and DB (28.83°C ± 3.43°C versus 30.29°C ± 3.21°C, respectively; P < .01). Wind speed (1.03 ± 1.22 m/s versus 0.95 ± 1.78 m/s, respectively; P = .11) and GT (41.83°C ± 4.79°C versus 41.59°C ± 4.19°C, respectively; P = .59) were not different between the NWS and onsite measurements. The NWS wind speeds were less variable than onsite wind speeds (ranges = 0.5–2.3 m/s versus 0.0–4.2 m/s, respectively). The relative humidity measurement was greater at the NWS stations than on the local surfaces (44.55% ± 6.99% versus 42.64% ± 8.55%, respectively; P < .001).

Bland-Altman analysis showed that the mean bias of NWS WBGT was less than that of the onsite WBGT measures for all athletic surfaces except for the red track and water (Figure 2). Almost all (136/144) NWS WBGT measurements were within the limits of agreement for onsite WBGT measures throughout our measurement range (19.60°C–33.65°C). In contrast, the difference between onsite and NWS WBGT measurements resulted in misclassification of the heat-safety activity category 45% (65/143) of the time (⁠ = 3.857, P = .05). The NWS underestimated heat-safety activity by 1 category 31% (44/143), by 2 categories 5% (7/143), and by 3 categories less than 1% (1/143) of the time. The NWS WBGT measurements rarely overestimated athletic-surface WBGT heat-safety activity categories, missing by 1 category 9% (13/143) of the time. We qualitatively evaluated whether category misclassification depended on athletic surface type. The NWS readings misclassified heat-safety category 33% to 60% of the time across all surfaces, with errors occurring most often on AstroTurf (60%; 9/15), red track (53%; 8/15), grass (53%; 8/15), and black tennis (50%; 7/14). As WBGT increased, agreement regarding heat-safety activity guidelines decreased between the NWS and onsite WBGT measurements (Table 2). This indicated that practitioners should assess local WBGT measurements instead of relying on data from the NWS to ensure appropriate activity-guideline categories at their specific locations. Misclassification of the heat-safety activity-guideline category between local and NWS WBGTs is illustrated in Figure 3 by data pairs that do not lie inside the shaded gray areas on the grid.

The first purpose of our study was to compare the NWS WBGT estimates with local athletic-surface WBGT measurements within the NWS reporting range and to assess how location (NWS versus onsite) affected WBGT heat-safety–category classification. We used 2 automated WBGT devices to record onsite measurements across 10 athletic surfaces and created a model NWS WBGT using an optimized algorithm for comparison.^16,19 

Our main finding was that NWS WBGT underestimated the onsite WBGT measurements by 0.67°C. Coyle¹⁸  (0.5°C ± 2.0°C) and Cheuvront et al¹⁵  (1.9°C ± 2.4°C) demonstrated underestimation of WBGT values when comparing regional NWS data and local athletic-surface recordings. The large difference observed by Cheuvront et al¹⁵  was likely due to the type and color of the surface where temperatures were recorded. They compared black asphalt with grass (NWS), whereas we compared grass (NWS) with 8 surface types and 5 surface-type colors. Our data supported these findings, as WBGT measurements on black athletic surfaces ranged from 1.11°C to 1.68°C greater than the NWS estimates. Given that the modeled WBGT data were collected over grass, it is logical that the model would perform better against natural surfaces. Researchers^{15,16,19,27,28}  have also observed greater differences in WBGT measurements between natural and synthetic surfaces than between 2 natural surfaces. Whereas WBGT measurements between the NWS and grass were similar in our study, misclassification of the heat-safety category using the NWS value remained high (53%), suggesting that onsite WBGT measurement remains a prudent choice. In summary, it appears that the WBGT derived from NWS data underestimates local athletic-surface heat stress, particularly if the surface color is black or red or the surface is made of synthetic material.

The practical implication of these WBGT differences is the large heat-safety category misclassification by the NWS. We are the first to evaluate the agreement of WBGT heat-safety category classification among NWS and onsite measurements across several athletic surface types and colors. We observed that, when using modeled NWS WBGT data, category misclassification occurred frequently (≥50%) on AstroTurf, red track, grass, and black tennis surfaces. In fact, category misclassification occurred across all athletic surfaces, albeit less often (33% to 44%). When all athletic surfaces were combined, the modeled NWS WBGT incorrectly identified the heat-safety category classification for 45% of the paired measurements. The modeled NWS WBGT underestimated the onsite WBGT by 1 category recurrently (31%), but we also observed mismatches by 2 (5%) and 3 (1%) categories (Figure 3). This finding mirrored the work of Cheuvront et al,¹⁵  who also showed dissonance of heat-safety category classification when using NWS data to model onsite WBGT.

In addition, we evaluated the congruence between NWS and onsite WBGT heat-safety categories as WBGT increased within our dataset. The heat-safety–category agreement of the NWS WBGT with the onsite WBGT decreased as ambient conditions became progressively oppressive (Table 1). This alarming trend could create an extremely unsafe situation. If NWS data are used, athletic trainers may not prudently implement proper physical activity modifications when environmental conditions warrant, resulting in a dangerous increase in the risk of EHI. Heat-safety activity guidelines use evidence-based strategies designed to mitigate internal and external heat production by providing frequent rest and hydration breaks, reducing activity duration and intensity, minimizing or removing equipment, and moving practice earlier or later in the day when weather conditions are typically milder.^10,13,29  In this way, the EHI risk is reduced.

The second purpose of our study was to compare WBGT measurements among the various athletic surfaces to determine whether WBGT recordings on athletic surfaces in close proximity are needed. We observed that WBGT measurements among athletic surfaces were not different except for water, which was consistently lower than for all other surfaces (Table 2). Albeit not statistically significant, the mean differences among the various athletic surfaces were large, ranging from −1.29°C to 0.87°C. This large range of mean differences suggests that WBGT measurements should be taken on each athletic playing surface at regular intervals to adequately capture the true environmental conditions that affect physical performance and the risk of EHI. Such variable contrasts among surfaces were likely caused by the shelter effect, whereby surrounding natural and artificial structures affect meteorologic measures that factor into the WBGT calculation (humidity, wind speed, and solar radiation), creating unique microenvironments. Kopec²⁷  evaluated athletic-surface WBGT congruence by comparing green AstroTurf, tennis courts (hard and soft), and an asphalt parking lot with grass. The deviation between grass and the athletic surfaces ranged from 0.16°C to 0.72°C, somewhat comparable with our data in which similar surfaces were compared (−0.57°C to 0.61°C). In all, our study provides empirical evidence supporting the WBGT measurement guidelines proposed in the National Athletic Trainers' Association¹⁰  and ACSM¹¹  position statements on EHI prevention and highlights the importance of onsite meteorologic assessment before and at frequent intervals when determining the need for activity modification or cancellation.

A limitation of our study was the use of only 1 NWS station to estimate WBGT. Whereas the equation of Liljegren et al¹⁶  for outdoor WBGT modeling is most accurate,^17,19  triangulating WBGT estimates with more than 1 surrounding weather station might have enhanced the algorithm's accuracy and should be considered for future investigations. Model performance would likely also improve when weather stations (and solar-radiation data) are in close proximity to athletic sites. Many states, such as Delaware, Georgia, and Oklahoma, have high-density weather station networks (mesonets), enabling enhanced model performance. We used the modified WBGT heat-safety category cutoffs created by Grundstein et al,¹³  which are thought to conservatively adjust for regional climate differences. In doing so, our category delineations were not uniform and in some cases were small (1.0°C–2.6°C), particularly as WBGT increased, which may help explain our finding that more category misclassifications were observed at higher WBGTs. Using larger WBGT heat-safety–category spans as employed by the ACSM for distance running (≥6°C)³⁰  and general physical activity (2.1°C)¹¹  or the military (1.1°C–1.7°C)⁹  would have resulted in fewer misclassifications. Practically, cautiously preparing (eg, hydration, cooling stations) for extreme weather conditions and implementing additional heat-safety protocols (eg, less equipment, rest breaks) when possible is encouraged regardless of heat-safety guidelines. Finally, data were collected in southern New England, and, whereas the data are convincing, the results may be not applicable to geographic regions in the United States with extreme dry heat, such as the Southwest, or humid heat, such as the Southeast.

The NWS WBGT underestimated athletic-surface WBGT measurements, resulting in a large misclassification of heat-safety category, especially for red and black artificial surfaces. The agreement for heat-safety category between the local WBGT and NWS measurements decreased as WBGT increased. These findings highlight the need for practitioners to assess environmental conditions before and at regular intervals during outdoor physical activity using an onsite WBGT device. In doing so, heat-safety activity guidelines can be implemented judiciously to increase athlete safety. The wide range of WBGT measurements among athletic surfaces indicated that WBGT should be measured on each surface to determine the need for physical activity modification or cancellation. Whereas products are commercially available to calculate WBGT from NWS data, our data and those of Cheuvront et al¹⁵  strongly suggest that onsite WBGT measurements are needed to accurately assess environmental conditions and determine the need for activity modifications.