Comparison of Methods for Estimating Wet-Bulb Globe Temperature Index From Standard Meteorological Measurements

Transcription

1 MILITARY MEDICINE, 178, 8:926, 2013 Comparison of Methods for Estimating Wet-Bulb Globe Temperature Index From Standard Meteorological Measurements Tejash Patel, MS; Stephen P. Mullen, MS; William R. Santee, PhD ABSTRACT Environmental heat illness and injuries are a serious concern for the Army and Marines. Currently, the Wet-Bulb Globe Temperature (WBGT) index is used to evaluate heat injury risk. The index is a weighted average of dry-bulb temperature (T db ), black globe temperature (T bg ), and natural wet-bulb temperature (T nwb ). The WBGT index would be more widely used if it could be determined using standard weather instruments. This study compares models developed by Liljegren at Argonne National Laboratory and by Matthew at the U.S. Army Institute of Environmental Medicine that calculate WBGT using standard meteorological measurements. Both models use air temperature (T a ), relative humidity, wind speed, and global solar radiation (R G ) to calculate T nwb and T bg. The WBGT and meteorological data used for model validation were collected at Griffin, Georgia and Yuma Proving Ground (YPG), Arizona. Liljegren (YPG: R 2 = 0.709, p < 0.01; Griffin: R 2 = 0.854, p < 0.01) showed closer agreement between calculated and actual WBGT than Matthew (YPG: R 2 = 0.630, p < 0.01; Griffin: R 2 = 0.677, p < 0.01). Compared to actual WBGT heat categorization, the Matthew model tended to underpredict compared to Liljegren s classification. Results indicate Liljegren is an acceptable alternative to direct WBGT measurement, but verification under other environmental conditions is needed. INTRODUCTION Heat exhaustion and heat stroke are serious concerns for military organizations. In the U.S. military in 2011, a total of 2,652 cases of heat exhaustion, an incident rate of 1.82 per 1,000 person-years, were reported. 1 Preventing casualties is a command responsibility during mission planning and execution in warm and hot environments or during hard work in cooler weather. Heat indices that provide estimates of injury risk because of physical activity and heat exposure are commonly used to assess and reduce the risk of heat causalities. 2 5 At present, the U.S. military uses the Wet-Bulb Globe Temperature (WBGT) heat stress index to guide work/rest and hydration practices and reduce the risk of heat illness and injury. 2 However, accurate and timely WBGT information needed to make informed decisions is often not readily available because of the cost and overhead of WBGT sensor systems, including recalibration of electronic monitors and the need to regularly replenish the wet- bulb thermometer s H 2 O reservoir. Fortunately, mathematical models have been developed that estimate WBGT from standard weather measurements, i.e., air temperature (T a ), air velocity, radiant load (e.g., direct, diffuse, or reflected solar radiation), and humidity. Biophysics and Biomedical Modeling Division, U.S. Army Institute of Environmental Medicine, 42 Kansas Street, Natick, MA Citations of commercial organizations and trade names in the report do not constitute an official Department of the Army endorsement or approval of the products or services of these organizations. The opinions or assertions contained herein are the private views of the author(s) and are not to be construed as official or as reflecting the views of the Army or the Department of Defense. doi: /MILMED-D This approach enables users with access to standard meteorological data to accurately estimate WBGT values without needing to rely on specialized, high-maintenance WBGT monitors. In addition, WBGT values needed for training and mission planning can be reconstructed from historic data, as well as projected using forecasts. The WBGT index, originally proposed by Yaglou and Minard 3, is based on a weighted average dry-bulb temperature (T db ), black globe temperature (T bg ), and natural wet-bulb temperature (T nwb ). Unfortunately, 2 of the instruments used to measure those environmental parameters, i.e., T bg and T nwb, are nonstandard and the T nwb sensor typically requires daily maintenance. The WBGT index would be more accessible, and easier to use, if it could be calculated using standard meteorological measures without the need for data from specialized instruments. Budd, 3 in a comprehensive review of WBGT, reviewed alternatives to the standard WBGT index methods, including the American College of Sports Medicine WBGT equation that uses only T a and relative humidity (RH). Another example given is the Swedish method of calculating WBGT that substitutes psychrometric wet-bulb temperature for T nwb, but has a correction for air velocities below 0.5 m/s. Another alternative heat index that uses standard meteorological measures is the environmental stress index, which compares favorably to physiological variables that reflect heat strain such as heart rate, sweat rate, and rectal temperature. 4 Unlike the American College of Sports Medicine WBGT, the output of these alternatives will allow the use of the same WBGT reference tables and guidance as the standard WBGT index. 5 A second approach, not involving the introduction of a new index, is to calculate WBGT sensor inputs from standard 926 MILITARY MEDICINE, Vol. 178, August 2013

2 meteorological sensor data. Liljegren, 6 Parsons, 7 and Matthew et al. 8 have developed models for this task. However, Parsons model requires inputs of clothing biophysical characteristics in addition to the standard meteorological measures. This study focuses on the models by Liljegren 6 and Matthew et al., 8 which use only standard meteorological measures as inputs. The goal is to determine whether either of these models accurately estimates the directly measured WBGT index values. The WBGT index has a long history and is in wide use across the military 2 and industry 7 to quantify environmental heat exposure. The equations for calculating outdoor and indoor WBGT are: WBGT = 0.7 equation, in Cor F) WBGT = 0.7 +T nwb T bg (indoor equation, in C or F) Three inputs are required: T db ; T bg, i.e., the temperature inside a black-painted 15-cm-diameter hollow copper sphere; and T nwb provided by a thermometer fitted with a wetted wick. All inputs should be measured at a 1.2 m (4 ft) elevation. In both equations, T nwb, which primarily reflects humidity, is weighted the most heavily. In many indoor environments T bg» T db, but at industrial sites or aboard ships other asymmetric high-temperature radiant heat sources such as boilers or furnaces may be of greater impact than T a. Part of the acknowledged appeal of WBGT is its simplicity of calculation. After calculation, the user refers to a table (e.g., Table I) to identify the corresponding heat category designated by a colored flag that can be green, yellow, red, or black. Typically, WBGT is measured with commercially manufactured instruments. A source list of WBGT monitors may be found in Appendix B of TB MED Currently, the U.S. Army does not specify standards for WBGT monitors, and thus no test programs exist to evaluate WBGT systems. Many commercially available instruments incorporate smaller-than-standard black globes, compute a correction factor, and digitally display T bg and WBGT. The ISO 7726 standard 9 states that the WBGT black globe sensor may be of any diameter, but smaller sensors are disproportionately influenced by T a and air velocity because of their faster response T nwb T bg T a (outdoor TABLE I. Current WBGT Heat Category Guidelines for Maximum Continuous Work Time in Minutes During Warm Weather Training 2 Heat Category WBGT Index ( C) Easy (250 W a ) Work (min) Moderate (425 W) Work (min) Hard (600 W) Work (min) No Flag Green Yellow Red Black > a W = watts. FIGURE 1. QuesTEMP 34 (Quest Technologies, Oconomowoc, Wisconsin) WBGT system with a 5 cm black globe (left sensor), natural wet-bulb sensor with water reservoir (middle), and a shielded dry-bulb thermometer (right sensor). times to environmental changes as diameter decreases. Figure 1 shows the WBGT sensor (QuesTEMP34, Quest Technologies, Oconomowoc, Wisconsin) used in this study. Dry-bulb temperature is T a measured with a shaded thermometer at a standard 1.2 m elevation since T a varies with height. Direct sunlight will affect the thermometer reading, so the thermometer is shielded from sunlight although allowing air exchange with the environment. Radiant load is estimated from T a measured within a hollow copper globe painted flat black, which is called black globe temperature. The temperature of the sphere is influenced by convective cooling (primarily a function of wind speed [ws] and T a ) and environmental radiation. Incoming radiation consists of direct solar radiation, diffuse solar radiation, reflected solar radiation, and long-wave radiation from the sky and ground. The temperature measured in the MILITARY MEDICINE, Vol. 178, August

3 sphere or globe effectively integrates all of these into a mean radiant temperature regardless of the orientation of the sun or other radiant sources. T nb or naturally aspirated wet-bulb temperature is measured by fitting a wetted wick over a temperature sensor exposed to the natural air flow. It is used to measure humidity but, as is the case with the globe thermometer, the measurement is affected by both air velocity and radiant load. Unlike other humidity sensors that require little attention in the field, most natural wet-bulbs require daily attention to keep the wick clean and to refill the reservoir with distilled water. For dry, warm, or hot conditions, the water must be checked and refilled daily. Our objective was to compare Liljegren 6 and Matthew et al. 8,10 models estimates of WBGT using data from standard weather sensors as input, i.e., WBGT. Both models use T a and ws as inputs to calculate T nwb and T bg. The Liljegren model uses humidity to calculate both T nwb and T bg. The Matthew model uses both RH and T db to calculate T nwb but does not use humidity for the calculation of T bg. Both models require the input of a value for global solar radiation (R G ) in W/m 2. R G is measured with a level pyranometer exposed to the sky, combining direct and diffuse solar radiation effects that contribute to T bg. Both models recognize the different radiant sources (direct, diffuse and reflected solar radiation, and ground emissions of infrared energy) and convective gain or loss of energy based on ws and T a. The Liljegren model uses a fraction of global radiation to estimate direct solar radiation, whereas the Matthew model uses a value for mean radiant temperature derived from R G. Ground temperature provides a mean to estimate thermal/ infrared radiation emitted from the ground surface in accordance with the Stefan Boltzmann equation. However, given that ground temperature is not commonly reported, assumptions were made regarding the relationship between T a and ground temperature. METHODS WBGT data were collected in Griffin, Georgia, which is in west-central Georgia 110 km from Fort Benning, a major U.S. Army training area. Since the 1990s, the U.S. military and allied countries have been engaged in major deployments to desert regions, and frequently train in hot/dry climates, which differ significantly from the southeastern United States where WBGT was originally developed. An additional data set representing this hot/dry environment was collected at Yuma Proving Ground (YPG), Arizona and was also used to evaluate the WBGT models. Tables II and V show summaries of the Georgia and YPG data. All data collection instruments were calibrated before use. The QuesTEMP34 (Quest Technologies) with a 5-cm globe sensor (Fig. 1) was used to collect the WBGT data. The dry-bulb sensor had a miniaturized plastic housing for shade and ventilation. Quest monitors were also used to collect data for the development of the Liljegren model. TABLE II. Quest WBGT and Weather Data Summary, for Weekdays 8:00 a.m. to 4:00 p.m., July 1 to September 28, 2007 (N = 3013) QuesTEMP34 15 Minute Average WBGT T db T nwb T bg WBGT Mean SD Min Max Campbell Base Station 15 Minute average T a RH R G ws (m/s) Mean SD Min Max The mean, standard deviation, minimum, and maximum data points are shown for the Campbell and QuesTEMP34 instruments for 15 minute segments. Griffin, Georgia Site: Our 12-month Georgia data set was collected in in Griffin, Georgia (latitude , longitude , elevation 285 m). Meteorological data were collected using (1) Campbell Scientific instruments Remote Automated Weather System base station, (2) QuesTEMP34 WBGT system, and (3) additional solar from silicon diode pyranometer. Yuma Proving Center Site: A smaller 3-day data set was collected at YPG on August 28 30, 2007 (latitude , longitude , elevation 99 m). The instrumentation used included a Campbell Scientific instrument 10 logger but was otherwise identical to the instrumentation used for the Griffin, Georgia data collection. Subsets of the data collected from the Griffin, Georgia and YPG sites were used for model evaluation. Specifically, weekday daytime data from the July to September warmweather period were used in the analysis. The weekend and evening data sets were not included in the analysis because of artifacts introduced by the fact that the WBGT system s wetbulb sensor reservoir was not refilled before the evenings or over the weekends. If the wet-bulb wick is not kept wet, there will be no evaporative heat loss, and T nwb readings will be higher, thus introducing a bias that may exaggerate the risk. The final weekday daytime data sets represented Monday to Friday 8:00 a.m. to 4:00 p.m.. The T bg, T nwb, and WBGT were calculated using the Matthew and Liljegren models. The inputs of each model for WBGT calculation were standard meteorological measures (date, time, air temperate, RH, barometric pressure, ws, and R G ). + data RESULTS Griffin, Georgia For the July to September daytime data set a minimum Category 1 WBGT index value of 25.6 C was used to determine if WBGT values would be of concern (see Table I). 928 MILITARY MEDICINE, Vol. 178, August 2013

4 TABLE III. RMSD of Measured Values of WBGT, T nwb, and T bg from Griffin Data Set Compared to Predicted Values From Matthew and Liljegren Models RMSD Between Measured and Calculated Values ( C) RMSD Error WBGT T nwb T bg Matthew Liljegren Table II summarizes the data used for analysis from the Griffin site. The Quest system was used to collect T db, T nwb, T bg, and WBGT values. The Liljegren and Matthew models predict these same measures using standard meteorological inputs. Table III shows root mean square deviation (RMSD) of the measured and predicted values for the Griffin data set during weekday working hours. The prediction of T bg is significantly more accurate using the Liljegren model, although the Matthew model is more accurate in calculating T nwb. Figure 2 shows correlations and Bland Altman plots. The Liljegren-predicted WBGT performs best (Liljegren: R 2 = 0.854, p < 0.01; Matthew: R 2 = 0.677, p < 0.01). The residual plots show the variance in the actual data compared to the correlation. The Matthew model shows higher numbers of outlying residual points than Liljegren. Measured WBGT heat categories were matched more accurately by the Liljegren model. The Bland Altman chart shows less difference from actual with the Liljegren estimation. It also shows a wider spread in categorization with Liljegren, whereas the Matthew model underperforms for higher heat categories discussed later. Table IV shows the WBGT heat categories for the July to September Griffin data set. The Matthew model tends to produce WBGT heat categorizations lower than the actual values resulting in a classification shift to lower flag conditions. For example, Matthew s classification is close to actual classification in the no category condition, it exhibits a spike of about 20% in the no flag condition, resulting in significantly fewer classifications for the green and yellow conditions and no classification for red and black. This represents a lower shift in classification. The Liljegren model closely follows the actual WBGT for the green condition, remains close to actual WBGT in yellow and red conditions, and is 10% higher than actual in the black condition. This represents a more balanced heat category classification than the Matthew model (Matthew: R 2 = 534.3, p < 0.01; Liljegren R 2 = 221.6, p < 0.01). Yuma, Arizona Table V summarizes the daytime (8:00 a.m. to 4:00 p.m.) weekday subset of the YPG data used for analysis. As in the Griffin data collection, the Quest system was used to collect T nwb, T bg, T a, and WBGT data. Table VI shows RMSD of the measured (WBGT) and predicted (WBGT ) values for the YPG data set. Figure 3 shows correlations and Bland Altman plots. The Liljegren-predicted WBGT performs best (Liljegren: R 2 = 0.709, p < 0.01; Matthew: R 2 = 0.630, p < 0.01). The residual plots describe the unexplained variance between the correlation modeled and the actual data. The Bland Altman plots clearly show how Matthew only predicts in a very narrow band whereas Liljegren predicts across more of the WBGT spectrum. The effects of this on WBGT heat categorization for the YPG data set are shown in Table VII. The WBGT flag conditions predicted by the Matthew model varied significantly from the actual Quest WBGT classifications. The Liljegren classification tended to follow much closer to the actual gold standard (directly measured) WBGT Quest values (Matthew: R 2 = 20.7, p < 0.01; Liljegren R 2 = 11.9, p < 0.01). DISCUSSION The prevention of heat illness and injuries is a command responsibility during mission planning and execution. Estimates of heat injury risk because of heat exposure rely on the WBGT index. The ability to obtain WBGT from standard meteorological measurements can greatly increase the accessibility of this risk assessment tool. The Liljegren model estimates of WBGT, using standard meteorological measurements is sufficiently accurate, particularly during periods of high heat injury risk, to be useful for heat casualty prevention. However, additional verification in regions with varied weather and terrain would help confirm the Liljegren s accuracy in many more operational environments. Data were collected in two distinct weather environments to test the accuracy and precision of the models in varying local conditions. At the Griffin site, the weather was warm (25.5 ± 4.4 C) with high humidity (70.7 ± 19.5%). Both the Matthew and Liljegren models were run for the July to September working hours. Compared to the actual readings, Matthew WBGT had an RMSD of 3.28 C and Matthew T nwb had an RMSD of 1.42 C. However, most of the error derived from the Matthew T bg estimate showed an RMSD of C. The Liljegren model brought the calculated black globe error down to 1.03 C, but with a slightly higher calculated T nwb RMSD of 2.63 C. Overall, Liljegren provided a WBGT prediction with a 1.71 C RMSD error. Although this error is less than that of Matthew, it is larger than the heat category ranges in the higher flag conditions. Table IV shows the percentage breakdown of heat category classification for Quest WBGT data, and the Matthew and Liljegren WBGT estimate. Matthew categorizes no data into the red or black flag conditions although 11.6% of the actual data fall within this range. Matthew underestimates risk of heat stress so much that it predicts that 76.6% of the time points will have no heat category compared to 55.6% of the actual data. Liljegren provides a more accurately distributed heat category prediction. However, its calculated heat categories seem to be MILITARY MEDICINE, Vol. 178, August

5 FIGURE 2. Comparison of Quest WBGT with Matthew and Argonne predicted WBGT for Griffin, Georgia data. Left column shows correlations, residual scatter grams, and Bland Altman plot for Matthew model vs. Quest WBGT measurements and right column shows the same analyses for the Liljegren model vs. Quest WBGT measurements. slightly more conservative (showing slightly higher risk classification) than actual WBGT (Matthew: R 2 = 534.3, p < 0.01; Liljegren R 2 = 221.6, p < 0.01). Fewer points are classified as no risk (43.4% vs. 55.6%) and more points are classified as a higher black flag risk (14.7% vs. 4.8% for actual WBGT). At YPG, the weather was hotter (38.0 ± 4.3 C) and less humid (29.7 ± 12.4% RH) than Griffin, Georgia. When compared to actual measurements, the Liljegren model showed lower RMSD error than Matthew for all three measurements (T db, T nwb, and T bg ) by 1 2 C. The Quest WBGT data show that 89.2% of data are in the red and black conditions 930 MILITARY MEDICINE, Vol. 178, August 2013

6 TABLE IV. WBGT Heat Category Classification as a Percentage for Weekdays From 8:00a.m. to 4:00 p.m. During July to September at the Griffin, Georgia Data Collection Site WBGT Category None No Flag Green Yellow Red Black Quest (%) Matthew (%) Liljegren (%) None indicates values fall below the minimum value (25.6 C) in Table I. TABLE V. Quest WBGT and Weather Data Summary, for Weekdays 8:00 a.m. to 4:00 p.m., August 28 30, 2007 (46 hours) QuesTEMP34 T db T nwb T bg WBGT Mean SD Min Max Campbell Base Station T a RH R G ws (m/s) Mean SD Min Max The mean, standard deviation, minimum, and maximum data points are shown for the Campbell and QuesTEMP34 instruments. TABLE VI. RMSD of Measured Values of WBGT, T nwb, and T bg From YPG Data Set Compared to Predicted Values From Matthew and Liljegren Models RMSD Between Measured and Calculated Values ( C) WBGT T nwb T bg Matthew Liljegren with 78.3% being black. The Matthew model again provided a lower estimation of heat category with only 34.8% estimated to be in black and 47.8% in red, totaling 82.6% with the balance allocated to the yellow category. The Liljegren model follows actual WBGT more closely but also with a slightly lower estimation (Matthew: R 2 = 20.7, p < 0.01; Liljegren R 2 = 11.9, p < 0.01). Liljegren predicts 87% will be in the red and black conditions, with 67.4% being in black. Interestingly, the lowest condition seen in actual WBGT is green at 2.2% of total points; Matthew shows no points in this category although Liljegren shows 2.2%. Both models seem to follow the most general trends in the actual WBGT heat categorization. With the Griffin data set, Matthew places all weight on the heat categories with lower risk of injury but with none at the high risks. This is problematic because occasions of high heat injury risk would go completely unidentified. The same can be seen in the YPG data with a full 43.5% of data not categorized in the black category, which has the highest heat injury risk. The Liljegren model showed a more balanced estimation of heat category. Each heat category was represented with significantly less error than Matthew. This shows that moments when the risk of thermal injury is high the Liljegren model performs better at identifying those heat categories. Despite the overall level of agreement between the observed Quest WBGT values and the Liljegren model estimations, there is some divergence. WBGT-based heat categories are narrow and decrease as the risk of injury increases. A small error in sensor values may result in significant costs either of unnecessary interventions to prevent injury or failure to act to prevent injury. 3,6 Because of the accuracy needed in estimating WBGT, especially for the narrower high heat risk categories, there is a need for highly accurate WBGT data or estimations. The simplest source of error is that the WBGT monitors and some weather stations report temperature values rounded to the nearest whole number. To obtain the most accurate estimates of WBGT, it is better to use the inherent accuracy of the meteorological sensors to provide more accurate input. The real concern is to separate erroneous instrument readings from a response to a real but unanticipated event. Error events may be associated with instrument failure, or the drying out of the wet-bulb sensors because of high radiant load. For example, under most conditions, T db ³ T nwb, thus for the difference (T db T nwb ), a negative value is suspect. It is useful to allow a slightly lower threshold, perhaps < 2 F. Another relatively simple solution is to perform data smoothing or moving average techniques to weather observations to mitigate the effects of transitory anomalies. A benefit of using standard weather observations is that weather forecasts may be used to project WBGT index values. As R G values are not included in forecasts, they must be estimated using models or algorithms that estimate solar based on location, date and time of day, and cloud cover. However, if a solar model is substituted for a direct measurement of R G, that will require validation. Accuracy using forecasts will clearly be less than direct measurement, but allows mission planning. Development or integration of existing solar calculation models would enhance the forecasting capability. Because WBGT is for a worldwide application, one concern over improving the Liljegren WBGT -to-wbgt fit is preventing the model to overfit for a specific weather condition observed at a specific location with a specific set of circumstances. To further validate WBGT, additional data should be collected for a variety of environments that differ MILITARY MEDICINE, Vol. 178, August

7 FIGURE 3. Comparison of Quest WBGT with Matthew and Liljegren predicted WBGT for Yuma, Arizona data. Left column shows correlations, residual scatter grams, and Bland Altman plot for Matthew model vs. Quest WBGT measurements and right column shows the same analyses for the Liljegren model vs. Quest WBGT measurements. TABLE VII. WBGT Heat Category Classification as a Percentage for Weekdays From 8:00 a.m. to 4:00 p.m. During August 28 30, 2007 at the YPG Data Collection Site Daytime None No Flag Green Yellow Red Black Red + Black Quest (%) Matthew (%) Liljegren (%) None indicates values fall below the minimum value (25.6 C) in Table I. 932 MILITARY MEDICINE, Vol. 178, August 2013

8 from those in this study. Variations in location, season, terrain, periods of extreme weather, and proximity to radioactive sources can affect WBGT readings. Also addressing how these varying conditions may affect physiological status would be a useful addition to WBGT. WBGT is a scale to communicate the risk of heat injury, and as such does not wholly capture the complex interaction of environment and human physiology. WBGT could supplement meteorological data with models addressing relationships between weather and physiological heat strain to strengthen heat risk categorization. A pressing need exists for methods to directly calculate WBGT from standard weather measurements. This study suggests WBGT estimated from standard weather observations using the Liljegren model are accurate. However, additional verification of the model is required. Budd 3 characterizes WBGT as an initial tool and emphasized the failure of WBGT to address the issue of maximum sweating capacity. Clearly, including information on the capacity of the human thermoregulatory system into the WBGT would enhance and strengthen WBGT heat categorization. Are efforts to implement the use of a WBGT model merited? The simplicity of WBGT, the fact that it is ingrained in present practices, its contribution to reducing heat casualties, and its potential for improvement are major factors supporting the continued and expanded use of WBGT. Replacing WBGT sensors with a mathematical model that provides WBGT estimation from standard meteorological sensors, such as the Liljegren model, would help provide improved and more reliable guidance. Standard weather sensors are highly reliable and enable historical and forecasting WBGT capabilities. The ability to obtain valid WBGT estimates in the absence of direct WBGT measurements will allow more widespread use of this injury prevention tool. However, future improvements should also include the addition of biomedical models that predict physiological heat strain in response to environmental heat stress. CONCLUSIONS WBGT is in wide use, but has significant limitations including the need for specialized maintenance-intensive sensors. WBGT estimates provided by the Matthew and Liljegren models were evaluated. The Liljegren model is the more accurate method of estimating WBGT using standard meteorological sensors, and effectively provides reasonable WBGT estimates in the absence of direct WBGT data. Additional verification of the Liljegren model under a range of environmental conditions is warranted to establish its validity for widespread use. ACKNOWLEDGMENTS We thank Dr. James Liljegren and his coauthors at Argonne National Laboratory provided access to their model, the staff of the Georgia Automated Environmental Monitoring Network at the University of Georgia Griffin campus for collecting the Georgia weather data, Julio A. Gonzalez for providing logistic support for the Georgia project, and him and Laurie Blanchard for the weather data collection during the Yuma, Arizona study. REFERENCES 1. Armed Forces Health Surveillance Center. Update: Heat Injuries, Active Component U.S. Armed Forces, Medical Monthly Surveillance Report 2012; 19(3): Available at fulltext/u2/a pdf; accessed June 24, US Departments of the Army and Air Force: Occupational and Environmental Health, Prevention, Treatment, and Control of Heat Injury. TB MED 507/AFPAM (I). Washington, DC, Headquarters, March Budd GM: Wet-bulb globe temperature (WBGT): Its history and its limitations. J Sci Med Sports 2008; 11: Moran DS, Pandolf KB, Shapiro Y, Labor A, Heled Y, Gonzalez RR: Evaluation of the environmental stress index for physiological variables. J Therm Biol. 2003; 28(1): Santee W, Gonzalez J: WBGT index alternative? (Abstract) Aviat, Space Environ Med 2008; 79(3): Liljegren JC, Carhart RA, Lawday P, Tschopp S, Sharp R: Modeling the Wet Bulb Globe Temperature using standard meteorological measurements. J Occup Environ Hyg 2008; 5: Parsons K: Heat stress standard ISO 7243 and its global application. Ind Health 2006; 44: Matthew WT, Santee WR, Berglund LG: Solar Load Inputs for Thermal Strain Models and the Solar Radiation Sensitive Components of the WBGT Index. Technical Report T Natick, MA, US Army Research Institute of Environmental Medicine, Available at accessed June 24, ISO 7726: 2001: Ergonomics of the Thermal Environment Instruments for Measuring Physical Quantities. Available at home/store/catalogue_tc/catalogue_detail.htm?csnumber=14562; accessed June 24, Matthew WT, Berglund LG, Santee WR, Gonzalez RR: USARIEM Heat Strain Model: New Algorithms Incorporating Effect of High Terrestrial Altitude. Technical Report T03-9. Natick, MA, US Army Research Institute of Environmental Medicine, Available at accessed June 24, MILITARY MEDICINE, Vol. 178, August