Journal of Applied Physiology
Vol. 83, No. 3, pp. 1017-1032, September 1997
ENVIRONMENT

Heat strain models applicable for protective clothing systems: comparison of core temperature response

R. R. Gonzalez¹, T. M. McLellan², W. R. Withey³, S. K. Chang¹, and K. B. Pandolf¹

¹ US Army Research Institute of Environmental Medicine, Natick, Massachusetts 01760-5007; ² Defence and Civil Institute of Environmental Medicine, North York, Ontario, Canada M3M 3B9; and ³ Centre for Human Sciences, Defence Research Agency, Farnborough, Hampshire GU146TD, United Kingdom

ABSTRACT

INTRODUCTION

MODEL CHARACTERISTICS

METHODS

RESULTS

DISCUSSION

APPENDIX

ACKNOWLEDGEMENTS

FOOTNOTES

REFERENCES

ABSTRACT

Gonzalez, R. R., T. M. McLellan, W. R. Withey, S. K. Chang, and K. B. Pandolf. Heat strain models applicable for protective clothing systems: comparison of core temperature response. J. Appl. Physiol. 83(3): 1017-1032, 1997.Core temperature (T_c) output comparisons were analyzed from thermal models applicable to persons wearing protective clothing. The two models evaluated were the United States (US) Army Research Institute of Environmental Medicine (USARIEM) heat strain experimental model and the United Kingdom (UK) Loughborough (LUT25) model. Data were derived from collaborative heat-acclimation studies conducted by three organizations and included an intermittent-work protocol (Canada) and a continuous-exercise/heat stress protocol (UK and US). Volunteers from the US and the UK were exposed to a standard exercise/heat stress protocol (ambient temperature 35°C/50% relative humidity, wind speed 1 m/s, level treadmill speed 1.34 m/s). Canadian Forces volunteers did an intermittent-work protocol (15 min moderate work/15 min rest at ambient temperature of 40°C/30% relative humidity, wind speed 0.4 m/s). Each model reliably predicted T_c responses (within the margin of error determined by 1 root mean square deviation) during work in the heat with protective clothing. Models that are analytically similar to the classic Stolwijk-Hardy model serve as robust operational tools for prediction of physiological heat strain when modified to incorporate clothing heat-exchange factors.

heat acclimation; exercise; clothing heat exchange; core temperature; thermal models

INTRODUCTION

MATHEMATICAL MODELING of thermal responses allows testing of wide performance limitations in individuals exposed to environmental extremes. Use of models is, therefore, especially important when experimental settings with human subjects are restricted to finite thermal limits necessary to protect the individual. In essence, the ideal mathematical model of heat strain should incorporate all essential variables active in thermoregulation. Although it seems to be an almost insolvable task, a great many worthwhile models do a reliable job describing the heat-balance equation. Models describing steady-state responses apply best when quasi-heat balance exists (2, 29, 30). They are quite useful in a first approach prediction of physiological effector response (e.g., sweating rate, skin blood flow), particularly when a given metabolic activity stays constant over the time of the given heat exposure.

A regulating system is usually described in two distinct ways: in terms of a passive or controlled system and an informational or controlling system. In physiological terms, the controlled system in human thermoregulation is considered as the body with its inclusive anatomical features, heat capacities, and energy fluxes from various tissues: core, muscle, adipose, and skin sites. The controlling system includes the complete central nervous system transmitting information in a network manner (23, 30). Early forerunners of classic rational thermal models (which employ elements of heat exchange that predict physiological response) incorporated extensive descriptions of the passive system in terms of a steady-state bio-heat equation (18) or open-loop systems without a full description of control or regulatory action. Such models were established on a scanty experimental database (8, 29). As data became available, closed-loop characterizations of the thermoregulatory system appeared, which included a rudimentary feedback-control formulation of internal body temperature (8, 29).

A significant database has been collected by using human experimental studies and wide clothing systems from which predictive modeling equations can be developed for individuals working in temperate and hot environments (1, 4, 5, 8, 11, 25). However, few comparisons of the results from various model outputs have ever been carried out. The present study describes approaches conventionally used in both rational and operational models and presents a comparison between measured and predicted core temperature (T_c) responses during exercise and heat exposure. Data from collaborative efforts conducted by three separate laboratories were evaluated against two models with separate model formats that employ distinct mathematical approaches but have analogous utility in prediction of heat strain, since they derive their form from fundamentals of heat exchange. This report covers one aspect of the collaboration.

MODEL CHARACTERISTICS

Operational Model

Recent enhancements to the original equations (17) include three major components in the present United States (US) Army Research Institute of Environmental Medicine (USARIEM) operational model, incorporating thermal, sweating, and heart rate (HR) responses. Only the thermal and sweating aspects will be covered briefly here; an extensive documentation of the clothing coefficient analyses and equations used is found in the APPENDIX and other reviews (2, 17, 19).

Predictive equations for implementing work/rest disciplines with various clothing systems, environments, or workload sequences in the original Givoni-Goldman model are based on a series of functions (2, 6, 7). The main function is the equation that establishes the difference in T_c expected at equilibrium. The basic Givoni-Goldman (7) model has the mathematical form

(1)

where H₁ is the net metabolic heat load, H₂ is the environmental heat load, and H₃ is the difference between total heat load (H₁ + H₂) and maximum evaporative cooling.

The coefficients _i, where i = 0, 1, 2, 3, and 4, are determined empirically from actual experimentation or database values. Any one of the covariates is measurable during an experiment or precisely determined by theoretically applied heat-transfer equations (e.g., partitional calorimetry) or by the other covariates.

H₁ is expressed as a function of body weight, weight of clothing and load, walking velocity, terrain factors, and grade of walking.

H₂ is expressed as a function of body surface area (DuBois), dry bulb temperature, average skin temperature (_sk), and total dry thermal insulation of clothing worn.

H₃ is expressed as a function of H₁ and H₂ covariates and water vapor permeability index (Woodcock's i_m factor; Ref. 27) of clothing worn, which, in turn, is a function of effective wind speed and relative humidity of the air and skin (based on saturated vapor pressure of the skin and ambient) (2).

Days of heat acclimation, solar heat load, physical fitness [maximal oxygen uptake (O_{2 max})], gender, and state of hydration have been also incorporated as modifying factors in the present model (17, 21).

Parameter adjustments that describe delay characteristics after a change in environment or clothing and that predict the rate of change of T_c as a function of time during work have been substantially improved in the present USARIEM heat strain model described in the APPENDIX. Transient T_re equations as a function of time of exposure to heat (t) have been developed for the resting state and exposure to heat stress by the following expression

(2)

where X_f and X₀ are the final and initial temperatures computed, respectively; t is duration of heat stress; and 0 < _i < 1, where i = 5, 6, and ₇ are parameters determined empirically. The model is a modified form of the Gompertz curve forming a sigmoid shape (28). Duplications of the initial rise [inflection of T_re f (t)] and family of curves compared with experimental data for various work rates, heat-stress exposures, and clothing systems are critical to the model's utility.

For the case where increased heat stress includes the initiation of work and t = 0 

(3)

A delay function in T_c caused by exercise at a given metabolic work and the rate of change function is currently implemented by four subroutines. The delay time is a period where changes in T_c are driven by the rate of change of T_c at the start of the delay period (e.g., inflection at initiation of exercise) and the magnitude of the change in T_c necessary to reach equilibrium that results from a change in metabolic rate described by an exponential rate coefficient (K_work factor) (APPENDIX, Eqs. A17-A19). The T_c is calculated every minute during the delay time. T_c at every minute is dependent on the initial T_c, the equilibrium T_c, the time in the period from the end of a proceeding delay time during work or rest, and the exponential coefficient, K_work. The original delay time in the Givoni-Goldman model (7) was based on a minimal delay time of ~6 min observed when 580 W of work are initiated. The time lag (min) for work-induced T_re is now based on a fit from multiple studies in our laboratory covering work rates [metabolic free energy (M)] of 250-600 W and is estimated by

(4)

Thermal component evaluated from partitional calorimetry. The basic analytical form discussed may be easily calculated for coding in computers by analysis of heat-exchange properties (2, 3, 5) in which

(5)

where T_re,f (°C) refers to the final steady-state level at a given time interval (from an initial T_re,0), predicted as a function of the following variables: skin heat transfer (H_sk; in W) composed of (M  W_ex) (in W), with M being a function of weight, walking velocity, grade, and terrain factors; W_ex being rate of work done on an organism by a external system as a function of grade, terrain, body weight, and clothing plus equipment weight, the latter generally assessed at three wind speeds (2, 5, 6, 17). Dry (in W) incorporates the sensible environmental heat load (R + C) on the person, where Dry = 6.45/I_T · A_D(_sk  T_a), in which total clothing thermal insulation (I_T) (assessed over a minimum of three wind speeds), body surface (A_D), and average skin-to-ambient temperature (_sk  T_a) are generally evaluated on a copper manikin in a specific garment (Table 1) (2, 3).

Table  1.   Clothing insulation, thermal resistance, water vapor permeability, and evaporative potential of NBC clothing systems Clothing System V, m/s IT, clo RT, m2 · K · W1 icl, clo Rcl, m2 · K · W1 im/IT UK Army NBC 1.12 1.78 0.276 1.36 0.21  0.26 UK Royal Navy NBC 1.12 2.08 0.322 1.66 0.257 0.17 US Army 1.12 2.11 0.327 1.69 0.262 0.15 US Air Force 1.12 1.24 0.192 0.82 0.127 0.28 Canadian NBC 1.12 1.88 0.291 1.46 0.226 0.18 Intrinsic clothing value (icl) from direct manikin evaluation for each system [clothing surface area (Acl = 2.23 m2)]. clo, Unit of thermal resistance based on heat transfer of 0.155 m2 · K · W1; V, air speed; IT, total clothing thermal insulation; RT, thermal resistance; im/IT, evaporative potential; NBC, nuclear biological-chemical protective suit.

Evap is accounted for by an exponential function {0.8 exp[0.0047 · (E_req  E_max)]} of the difference in required evaporative cooling (E_req) and maximum evaporative heat exchange (E_max; in W), where

(6)

in which LR is the Lewis relation = 2.2 °C/Torr (16.5 K/kPa) (2, 5) that is used to evaluate, at sea level, evaporative heat transfer to radiative and convective heat transfer, respectively (LR = h_e/h_c); i_m/I_T describes the evaporative potential of a specific clothing system based on the Woodcock's factor i_m to thermal insulation (I_T), and (P_s,sk  P_a) is the body skin saturation vapor pressure (P_s,sk) to ambient water vapor pressure (P_a) gradient depending on an effective body surface area (A_eff) (2, 3, 5, 6).

Sweating rate and net water requirements. The change in T_re from rest to a given time point during transients depending on metabolic activity can be determined by T_re/t = (S · A_D)/ · _b, where S (W · m²) is rate of heat storage, evaluated from partitional calorimetry and accounting for all energy exchanges in the heat balance equation, M is the latent heat constant (680 W · h · kg¹), and _b is the nude body weight loss (kg) (3, 5, 7).

The necessary water to supplement that lost during work and environmental heat is an added parameter adjustment to the original Givoni-Goldman operational model in the present USARIEM heat strain model. This equation is derived from sweating rate _sw (g · m² · h¹), which is a function of the maximum evaporative power of the environment E_max (3, 5) and required evaporative heat loss from the heat balance equation (21) Eqs. A15-A16 in APPENDIX. Water requirements (Wtr; ml/min) for a wide range of heat-stress conditions can be analytically determined by the following relationship (21)

(7)

Thermoregulatory Models

The United Kingdom (UK) Loughborough (LUT25) model, as used here, was transcribed by Haslam and Parsons (9, 19) from the the Stolwijk-Hardy 25-node model of thermoregulation (23), which represents the human body as 25 compartments. In brief, the model depicts the head as a sphere and the trunk, arms, hands, and legs and feet as cylinders. A symmetrical construct of the body is assumed to limit the number of iterations required. Each segment is further divided into four layers: core, muscle, fat, and skin compartments to equal 24 compartments. The major arterial and venous vessels are represented as the 25th compartment. Each compartment is assigned a mass, volume, and specific heat. The values were obtained partly via direct experimentation and partly from the literature (23) and relate to an average-sized male with a body weight of 74.4 kg and 1.89 m² DuBois surface area A_D (3).

In the model's passive structure, heat flows radially by conduction from compartment to adjacent compartment. From segment to segment, heat flow is by convective transfer to and from the blood. Metabolic heat production is divided proportionately between the various segments and layers. External body compartments exchange heat with the environment by means of convection, radiation, and evaporation of thermoregulatory sweating.

The controlling system is based on a set-point theory of human thermoregulatory control. Signals controling vasodilation, vasoconstriction, sweating, and shivering are calculated as a function of the difference of the actual temperatures of the compartments, from reference temperatures for each respective compartment. Local thermal signals are modified based on density of thermoreceptors present in each respective compartment. The thermal signals are integrated to produce core, core and skin, and skin signals. The effector regulator interprets the integrated signals and produces effector commands. Each effector command is implemented by an appropriate effector outcome: shivering, vasodilation, vasoconstriction, and sweating, after first being modified according to local compartmental thermal state.

The LUT25 computer program used in this comparison is a recoded version of the 25-node model adapted from the original FORTRAN program listing given by Stolwijk and Hardy (S-H) (23). In the version provided by Centre for Human Sciences to USARIEM, there were modifications to the program listing and corrections to program logic to allow execution on a personal computer. The program code was changed to prevent the S-H model from shivering in the heat, which the original consistently demonstrates. The original S-H model predicts responses for the unclothed condition only. The published computer code of the model was adapted to account for clothing and thermal radiant loads by implementing the coefficients from the Gagge et al. (3) J. B. Pierce model, extended to include intrinsic clothing (i_cl) and water vapor permeability factors (Woodcock's i_m; Refs. 3, 6, 27). The LUT25 version used in this study implements clothing coefficients equally to each body compartment of the S-H model. Trunk temperature was used to predict the responses of T_re from the experimental trials. All other parameters and control coefficients were exact facsimilies of the original code.

METHODS

The experimental study was conducted in three separate laboratory sites (US, UK, and Canada). Comparisons of thermoregulatory responses were done on volunteer subjects dressed in completely encapsulating chemical protective (CP) ensembles. The emphasis of the study was to compare T_re responses vs. exercise time obtained from individuals exercising with the various protective ensembles. A standard level of environmental stress agreed on by country members was used: T_a  T_g = 35°C/50% relative humidity (RH), where T_g is globe temperature [water vapor pressure (P_w) = 2.81 kPa; 21 Torr], at constant wind speed of 1 m/s. Subjects attempted a 100-min walk on a level treadmill at a pace of 1.34 m/s (metabolic heat production ~300-400 W) until self withdrawal or until T_re reached 39°C. UK and US experiments followed the above common protocol design. Data garnered from all laboratory experiments were then compared with model simulations obtained by using the USARIEM heat strain model (our present experimental version) and the LUT25 model (9). Model predictions were also conducted on experimental data obtained from a previous intermittent-work protocol (1) carried out by Canada, utilizing the Canadian CP ensemble.

USARIEM Procedures

1

1

Experimental design. The study was conducted from February to early June in the US Army Doriot tropical environmental chamber during all phases of the study. Following a familiarization session to the test environment (35°C/50% RH, wind speed of 1 m/s), all 10 subjects underwent a 10-day heat-acclimation period. A 48-h period intervened between pre- and postacclimation experiments. The subjects were in good health and had not taken any prescribed or unprescribed medication or alcohol during the course of the experiments. A medical monitor was on site throughout the testing. Some exercise bouts were terminated before the 100-min time schedule when a subject voluntarily withdrew, when a subject's T_re reached a terminal point of 39°C, or when HR exceeded 180 beats/min for 5 min. Termination was also allowed at any time based on the medical monitor's decision. The relative percent O_{2 max} ranged from 27 to 30%.

Heat acclimation. During heat acclimation, subjects wore only gym shorts and gym shoes and walked on a level treadmill at the same speed as in the preacclimation phase (1.34 m/s) at a constant environment of T_a = 49°C/20% RH and wind speed of 1 m/s. Heat acclimation was confirmed when T_re and/or HR had leveled off by the 10th day of exposures. Water intake was allowed ad libitum.

Exercise-heat tests. Before and after the 10-day heat-acclimation program, subjects were exposed to the standard heat-stress test environment agreed on by the laboratories. Other than an initial hydration of 500 ml occurring 20 min before exercise, water was not given during pre- and postacclimation continuous work phases. In both the pre- and postacclimation experimental runs, subjects donned either the US Army battle dress overgarment (BDO), worn over the battle dress uniform (BDU), or the US Air Force CP ensemble worn over underclothing. In brief, the US Army temperate zone BDO consists of a two-layer, two-piece garment with coat and trousers. The outer garment shell is a 50:50 nylon-cotton twill, which is durable and water repellent against liquid agents. The outer shell is laminated to an inner layer of polyurethane foam liner impregnated with activated carbon. The outer-layer pattern is either olive green or four-color woodland camouflage. For maximum chemical protection, the BDO is worn over a regular issue BDU. The fully encapsulated configuration also entails donning a M17A1 CP mask, butyl rubber hood, butyl rubber gloves with cotton liners, and vinyl rubber overboots over the regular issue leather combat boots.

The Air Force chemical defense flight suit (CWU/77P) consists of a one-piece coverall with similar two-layer construction as the US Army BDO. The exterior color is tan. The coverall is intended primarily for ground-crew operations. During testing, the coverall was worn over undershirt and underwear, as dictated by US Air Force requirements. The same US Army M17A1 gas masks, protective gloves, and rubber overboots were used for both the Army nuclear biological-chemical (NBC) and the Air Force flight suit experiments. Thermal and water vapor resistance values shown in Table 1 were evaluated at USARIEM by using a copper manikin (6).

T_re, _sk, and HR were continuously monitored throughout the exercise. T_re was measured with a vinyl-covered calibrated thermistor probe (Yellow Springs Instruments 44033) inserted 10 cm past the anal sphincter. _sk was determined by using a calibrated six-point surface area weighting (16) (temperatures from the forehead, chest, back, upper arm, thigh, and calf) recorded from a skin temperature/skin heat flux harness (Concept Engineering, Old Saybrook, CT; FR-025-TH44018). All body temperatures were recorded every 10 s by using a personal computer data-acquisition system.

Metabolic heat production was calculated by open-circuit spirometry (5). Total body sweating rates were determined from body weight changes before and after exercise utilizing a Sauter balance (±0.005 kg). Oxygen uptake (O₂) was measured by indirect calorimetry with the Douglas bag method (5) at the 10th minute and 30th minute time periods of the treadmill walk. The chemical mask was detached for a 2-min collection of expired air while the subjects continued to walk on the treadmill. After each collection period, subjects reconnected the chemical mask. These O₂ data were used to evaluate the transient and steady-state metabolic heat production. The electrocardiogram was monitored continuously with a dedicated telemetry system (Hewlett-Packard 78100A, 78101A). HR data were recorded every 10 min.

UK Procedures

Subjects. Thermal responses were observed in 13 male subjects over the full 60 min, although subject attrition was noticeable before the 100-min milestone of the experiment. For this paper, data were truncated past the 65th minute for the comparison with model output. Subjects were dressed in either the UK Army NBC clothing or UK Royal Navy NBC clothing system. Average (±SD) physical characteristics were age (25.1 ± 3.1 yr), weight (75.6 ± 7.8 kg), height (1.73 ± 0.13 cm), and body fat (4 skin sites; 13.6 ± 4.5%). Clothing and water vapor transfer characteristics were determined at USARIEM for all the UK ensembles (6), shown in Table 1.

Experimental design. All experiments were conducted at the Centre for Human Sciences, Defence Research Agency, Farnborough, UK. The experimental protocol was approved by a local Human Ethics Committee.

Heat acclimation. Experiments were conducted in the standard heat-stress environment before and after 10 successive days of heat acclimation. The heat-acclimation phases were conducted with subjects wearing light clothing (shorts, T-shirts, boots) on a level treadmill (1.33 m/s) in a hot environment (calculated WBGT of 38-40°C) until T_re reached 38.8°C, after which subjects rested in the chamber. T_c was maintained at this level for an additional hour by further rest or intermittent exercise, as needed. T_re, a weighted four-site skin temperature (20), HR, and total weight loss were recorded at 1-min intervals. Endurance times were determined by time to self-withdraw or, whenever withdrawal limits were reached, based on the Defence Research Agency UK Ethics Committee criteria.

Canadian Procedures

1

1

Experimental design. All trials were conducted during the months of November through March when outside T_a varied from 20 to 10°C. Following a 60-min familiarization session to the heat-stress test environment (T_a = 40°C/30% RH, air movement ~0.4 m/s), subjects were assigned to one of two groups: 1) one group underwent heat acclimation for 6 days (n = 8); and 2) the other group underwent a 12-day heat-acclimation procedure (n = 8). The two groups were matched as closely as possible on the basis of their initial physical characteristics and O_{2 max}. Because there was no difference between groups in the reduction in heat strain while wearing the CP clothing after 6 or 12 days of heat acclimation (1), the data were analyzed as one group (n = 16) for this report pre- and postacclimation. Physiological responses to a standard heat-exercise stress test were collected twice before and twice after no more than 4 days following the heat-acclimation period. A 48-h period intervened between these pre- and postacclimation trials. Before and after heat acclimation, subjects were evaluated while wearing lightweight combat clothing and the CP clothing ensemble used in the Canadian Forces. All experiments were performed at the same time of day between 0800 and 1200.

Heat acclimation. Subjects wore jogging shorts and a T-shirt. Heat acclimation was carried out by a daily 1-h bout of treadmill exercise (1.34 m/s, 3-12% grade) in a hot environment (T_a = 40°C, 30% RH). Water was given ad libitum. The exposure was repeated over 6 consecutive or 12 days (two 6-day periods, separated by 1 day).

Exercise-heat tests. Before and after the heat-acclimation program, subjects were exposed to the heat-stress environment while wearing the Canadian Forces NBC protective clothing ensemble. This ensemble consists of a one-piece protective overgarment with a similar two-layer construction as the US BDO system. The overgarment was worn over lightweight cotton combat clothing and underwear, together with impermeable rubber gloves and overboots (worn over jogging shoes or combat boots to the subject's preference) and a C4 respirator and cannister. Clothing and vapor characteristics of this clothing configuration were determined at USARIEM (Table 1). Subjects alternated 15 min of level treadmill walking at 1.34 m/s with 15 min of seated rest. These trials continued for a maximum of 150 min or until T_re (measured 15 cm beyond the anal spinchter) reached 39.3°C, HR reached or exceeded 95% of the individual's maximum for 3 min, or when nausea or dizziness precluded further exercise. Dependent parameters besides T_re included _sk (12-point area weighted average), HR (Sport tester), and O₂ (from open-circuit spirometry determined as a 2-min average every 15 min), and whole body sweating rates were determined from pre- and posttrial nude and clothed body weight measurements corrected for respiratory and metabolic weight losses.

Statistical Evaluation

RMSD. Comparison of each T_c time series data from model predictions and experimental data was accomplished by using RMSD (9, 19). The statistic was used as formulated in Haslam and Parsons (9) for goodness of fit comparison of model output predictions. The RMSD (°C) of model prediction output to observed T_c is defined as where d_i is the difference between observed (from experimental trials) and predicted T_c response at each time point (°C) and n is the number of time points examined.

Model input techniques. The USARIEM heat strain model has the option of querying the user for initial T_c levels (T_re,start) to match observed levels. In using the model, we employed a recursive programing operation (24, 28) to establish K_work (Eqs. A17a-A19, APPENDIX), which matched the rise in T_c based on the actual average metabolic heat production (Table 2) observed in experimental runs.

Table  2.   Metabolic heat production before and after heat acclimation in subjects wearing various countries' protective clothing ensembles Ensemble Unacclimated 30th min 10-day Heat Acclimation 30th min %Change UK Royal Navy NBC 362 ± 91  323 ± 59.6   10.8 (n = 6)  (n = 12) P < 0.22, NS UK Army NBC 362 ± 67  316 ± 63.2   12.7 (n = 9)  (n = 10) P < 0.13, NS US Army BDO + BDU 449 ± 51.3  436 ± 63.9   2.9, NS (n = 10) (n = 10) US Air Force CWU/77P 419 ± 57.0  413 ± 55.0   1.5, NS (n = 10) (n = 10) Values are means ± SD in W; n = no. of subjects. BDO, battle dress overgarment; BDU, battle dress uniform.

In the LUT25 model, the trunk T_c option was utilized to simulate observed T_re response (9). Additionally, intrinsic thermal resistance values were determined by a copper manikin analysis over three wind speeds at USARIEM and utilized as input into the LUT25 model (Table 1).

For each specific standard heat-stress experiment, a T_re ceiling level of 39°C (102.2°F) was implemented as a heat-casualty limit, instead of the conventional 39.5°C (103.1°F) (USARIEM Human Use Review Committee limits). This T_c limit helped curb excessive fatigue in the subjects, brought on by repeated daily exposures.

Evaporative efficiency. Differences in nude and dressed weights before and after each experiment were corrected for urine, respiratory, and metabolic weight losses (1, 3, 5, 14). The amount of sweat secreted (_sw, kg/h) was calculated as follows: pretrial minus posttrial nude weight (corrected) plus the weight of water drunk during the trial. Evaporative sweat loss (Ev, kg/h) from the clothing was calculated as the ratio of the difference in corrected dressed weight to the amount of sweating (Ev/_sw) and expressed as a percentage (Ev/_sw × 100). This percentage determined the evaporative potential observed through a clothing system with which to compare the predicted evaporative potential from manikin values of i_m/I_T (Table 3).

Table  3.   Comparison of final , sweating rates, and evaporative potential observed and predicted for each clothing system   Final, °C  sw, kg/h Ev, kg/h Ev/sw, %  Predicted Evaporative Potential From im/clo, %  UK Army NBC   Preacclimation (n = 13) 37.2 ± 0.5  1.40 ± 0.5  0.40 ± 0.3  28   Postacclimation (n = 13) 36.8 ± 0.6  2.00 ± 0.6* 0.40 ± 0.1  20 26 UK Royal Navy NBC   Preacclimation (n = 13) 37.6 ± 0.5  1.80 ± 0.5  0.40 ± 0.5  22   Postacclimation (n = 13) 37.1 ± 0.2* 2.20 ± 0.9  0.30 ± 0.4  14 17 US Army   Preacclimation (n = 10) 37.05 ± 0.67  1.25 ± 0.25  0.25 ± 0.08  20   Postacclimation (n = 10) 36.73 ± 0.21  1.48 ± 0.24* 0.24 ± 0.10  16 15 US Air Force   Preacclimation (n = 10) 35.95 ± 0.49  1.32 ± 0.16  0.51 ± 0.23  38   Postacclimation (n = 10) 35.23 ± 0.48* 1.51 ± 0.16* 0.54 ± 0.27  35 28 Canadian NBC   Preacclimation (n = 16) 37.9 ± 0.2  1.00 ± 0.1  0.28 ± 0.07  28   Postacclimation (n = 16) 37.2 ± 0.3  1.13 ± 0.10* 0.30 ± 0.07  27 18 Values are means ± SD; n = no. of subjects. sw, amount of sweat secreted; Ev, sweat evaporation. * Significant decrease in final mean skin temperature and increases in sweating rate postacclimation (P < 0.05). Predicted evaporative potential evaluated by im/clo on completely wetted skin (parallels fully heat-acclimated person).

RESULTS

Table 2 shows the steady-state mean ± SD metabolic heat production (30th minute) before and after heat acclimation from the UK and US experimental runs.

Time Series

In the experiments with US Air Force CP clothing, there was an initial rapid rise in _sk, followed by leveling off in the final values of _sk to 35.95 ± 0.49°C in the preacclimation runs, but it declined to 35.23 ± 0.48°C (P < 0.05) after the heat acclimation (Table 3).

Because sensible heat loss (R + C) was reduced by having _sk  T_a = 35°C, the only avenue of heat exchange was brought about by the enhanced sweating rate cooling the skin under the clothing (some 0.8°C lower). Skin cooling was facilitated adequately while wearing the US Air Force CP ensemble, as indicated by the high evaporative potential (i_m/I_T) evident from this ensemble (Table 3). No apparent differences in the T_c responses were evident before and after heat acclimation, nor was there any evidence of convergence of _sk with T_c. T_c rose steadily (1.2°C/h) and leveled off to ~38.5°C.

During exposure to the standard heat stress, in both the pre- and postacclimation experiments with the US BDO + BDU, _sk values rose rapidly but exhibited a leveling off at ~37.05 ± 0.67°C pre- and at 36.73 ± 0.23°C (not significant), indicating that both sensible heat exchange (by R + C) and some evaporative cooling potential were possible. T_c values rapidly rose (2.14°C/h) to the limiting threshold of 39°C in ~70 min of exercise without any obvious steady state or leveling off. There was no apparent convergence of _sk with T_re in any of the experiments at this metabolic heat production. There was a significant increase in sweating rate after heat acclimation (Table 3) in both the US Army and US Air Force experiments.

UK experiments. Figure 2 presents results of the UK experiments to the heat-stress exposure before and after the 10-day heat acclimation, showing changes found in T_re response.

The runs with the UK Army clothing system show three clear-cut observations: 1) after the heat acclimation there was a lower final T_re (P < 0.05); 2) there was an offset in the slope of the T_re vs. time plot, suggesting that significant evaporative cooling (Table 3, P < 0.05) by sweating aided in increasing the endurance times by delaying the rise in T_c at this metabolic activity; and 3) the rate of rise of T_c up to a 65-min period in all subjects averaged 1.63°C/h before acclimation and 1.57°C/h after the heat acclimation. Final skin temperatures pre- and postacclimation were not significantly different (Table 3).

In the experiments in which subjects wore the Royal Navy clothing system, curves of T_re plotted for the first 60 min of exercise displayed a similar lowered T_c offset after heat acclimation as in the UK Army runs. Rate of rise of the T_c vs. minutes of exercise averaged 1.6°C/h before acclimation and some 1.35°C/h after heat acclimation (P < 0.05). Final skin temperatures were lower after heat acclimation (P < 0.05) (Table 3).

Canadian experiments with intermittent work. No differences were evident in heat production attributable to heat acclimation during intermittent work or rest phases, although a gradual elevation was apparent at each 15-min rest-cycle phase heat production over time (from 200 W to as much as 300 W). The gradual increase in resting M appeared affected by the antecedent elevated rate of heat storage ensuing during each exercise bout.

Figure 3 shows the T_re response to the heat-stress exposures (T_a = 40°C/30% RH, air speed = 0.4 m/s) plotted as a function of time before and after the heat-acclimation phases. There was an overall offset toward a lower T_re after heat acclimation. A significant increase in sweating rate (Table 3) was observed postacclimation.

USARIEM Heat Strain Model vs. Experimental Data Comparisons

Relative fit of the model output compared with average T_re ± SD of the observed data was generally overpredictive for the first 40-50 min of exercise with the US Army CP ensemble. The final predicted T_c values were within ±1 SD of the observed values during both the pre- and postacclimation runs. Relative fit of model output to mean data observed with the US Air Force clothing system was within ±1 SD during the unacclimated T_re time course and overpredictive up until the final time point in the postacclimation experiments.

Sensitivity analysis of the US data prediction vs. observed T_c response was carried out by calculation of RMSD over all 1-min time points of the exercise until a final exercise period (9, 31) (Table 4).

Table  4.   Sensitivity analysis of USARIEM model and LUT25 model predictions using RMSD on time series rectal temperature response data 1) USARIEM Model vs. Experimental Data (Unacclimated) 2) USARIEM Model vs. Experimental Data (Heat-Acclimated) 3) LUT25 Model vs. Experimental Data (Unacclimated) US BDO + BDU 0.364 0.203 1.34 US Air Force CP 0.300 0.353 0.85 UK Army NBC 0.189 0.147 0.346 UK Royal Navy NBC 0.249 0.107 0.292 Canadian NBC 0.228 0.240 0.870 USARIEM, US Army Research Institute of Environmental Medicine; LUT25, United Kingdom Loughborough model; RMSD, root mean square deviation (°C); CP, chemical protective garment.

For all US data shown in Table 4, performance of model predictions of T_re over the whole time course of runs was in reasonable agreement to observed T_re with values occurring within a RMSD of 0.4°C for both the unacclimated and acclimated state.

UK experiments. Figure 5 compares the USARIEM model predictions of T_c with the experimental observations obtained from the UK studies comprising a 65-min period achieved by all subjects.

To fulfill the 100 min at the required average metabolic rate while wearing the UK Army NBC ensemble (Table 1 clothing insulation), the USARIEM model calculated that final T_re values would reach 39.5°C when subjects are unacclimated to heat but ~38.6°C after becoming heat acclimated. Figure 5 shows that the model results matched the observed T_c time records accurately, except for a slight initial overrise shown in the first 15-20 min in the UK Army experiments in the preacclimation state and a slight underprediction evident in the postacclimation phase around minutes 60-65. The actual cessation of exercise in 13 individuals wearing the UK Army chemical clothing system occurred at around 62 min (based on a final T_re of 39°C) and ~82 min after the heat-acclimation phase (32% higher, P < 0.05).

Similar reliable predictions using the USARIEM model were obtained while evaluating the UK Royal Navy CP clothing suit data before and after heat acclimation. The RMSD analysis is shown in Table 4, encompassing both UK experimental data vs. the model output. Model predictions of T_re over the whole time course were in reasonable agreement with actual T_re values falling within a RMSD of 0.25°C for both the unacclimated and acclimated state.

Canadian experiments. Figure 6 shows the USARIEM model simulations carried out on the Canadian heat-stress intermittent-exercise experiments before and after heat acclimation.

The USARIEM model output reliably tracked the T_re recorded during the work/rest cycles; both curves from the USARIEM model tracked the changes observed before and after heat acclimation. The superimposed curves from the model output on the actual mean T_c data during both the preacclimation and heat-acclimation model curves overlap within ±1 SD of the actual observed data. Table 4 shows the RMSD values obtained in the simulation vs. observed data comparison. Model performance of predictions of T_re for the intermittent exercise over the whole time course were in reasonable agreement to actual T_re falling within a RMSD of 0.24°C for both the unacclimated and acclimated state.

LUT25 Model Evaluations

1

Figure 7 displays the results of the LUT25 trunk node temperature simulations in comparison with the mean ±1 SD T_re data observed from the US and UK experiments.

The LUT25 model consistently showed an initial decrease in the T_c response for the first 5-10 min of exercise observed in all UK experiments. The model output was overpredictive in the T_c response when comparing both US data sets, with forecasting temperatures approaching a hyperthermic state within 40-50 min. However, the LUT25 model matched closely (within ±1 SD) the observed T_c response obtained from both UK experimental runs.

Canadian experiments. The clothing thermal and evaporative resistances were first assessed in terms of intrinsic clothing resistance before input into the LUT25 model (Table 1). Figure 8 shows the T_re results from the LUT25 model modified to simulate the Canadian 15 min work/15 min rest procedure for the unacclimated state.

The LUT25 model suitably matched observed T_c values up to the first 105 min of intermittent exercise/rest experiments. The saw-tooth pattern shown by intermittent work cycles and rest periods (as well as the slightly elevated T_c values occurring during the rest periods) were tracked within ±1 SD. Model output was slightly overpredictive of T_c response toward the end period (105-120 min).

Table 5 presents an analysis of the mean difference for RMSD values (from Table 4) covering pooled model output results vs. the T_c responses. Characteristically, the lower the RMSD, the closer is the fit of a particular model simulation of this parameter to the experimental observations.

Table  5.   Paired sample t-test comparison of RMSD output from different models Data/Model Compared With Table 4 Columns Mean Difference of RMSD Value (±SD) t-Value (Two-Tail) P <  All country's unacclimated data/USARIEM model vs. LUT25 model (1 vs. 3) 0.474 ± 0.169* 2.803 0.049 All country's acclimated data/USARIEM model (1 vs. 2) 0.056 ± 0.042  1.335 0.253, NS * Data assuming equal variances.

There was a significant difference between output of the USARIEM and LUT25 models in the RMSD value for the unacclimated data (P < 0.05). No statistical differences in the RMSD value were apparent in the pre- and postacclimation analyses when using the USARIEM heat strain model.

DISCUSSION

This study investigated whether experimental results can be suitably matched with physiological outputs determined from various model predictions in individuals dressed in protective clothing. Simulations from two disparate thermal model constructs were tested as predictors of T_c. It is useful to discuss several positive features and limitations of each of the two models and point out the physiological implications that may explain some of the model results.

USARIEM Heat Strain Model

In the original model construct, abrupt rates of rise in T_c values as a function of time are often observed when applying the Givoni-Goldman (7) T_re equation during exercise with different protective clothing systems. In the Givoni-Goldman equations, a time lag equation for rise in T_c is based on a best estimate of 6 min for 580 W of work. The curve prediction of T_re,f is an exponential rise to maximum. Consequently, the rate constant becomes too elevated because of the assumed high-intensity exercise. This rate constant overexaggerates T_re,f at each time point, resulting in overprediction of model output vs. experimental values. The original Givoni-Goldman T_c equation, used during uncompensable heat stress, frequently generates an RMSD >1.0 in the respective time-series comparisons between model and observed T_c and skin temperatures. RMSDs >1.0 observed from model comparisons of a given physiological variable typically indicate that a model does not reasonably predict observed values. When this occurs, corrections should be made to the computer code or algorithm control coefficient structure before use of the model. We implemented a time-delay feature in the model that matches each average metabolic heat production (described by Eq. A19 in APPENDIX) (Figs. 4, 5, 6). This function accurately buffered the rise of T_re paralleling observed mean T_c values, as observed in the experimental trials.

LUT25 Model

One could assume that the values of evaporative potential were too small when determined from measurement of evaporative and thermal resistance by using a copper manikin. The consequences would be that latent heat loss simulated by the LUT25 model through the semipermeable membrane laminates of a given protective clothing system would be underpredictive and T_c would be estimated too high. This hypothesis is not likely, since Table 3 shows that at 1 m/s wind speed, quite similar values of i_m/I_T of 0.26 and 0.28 were obtained from the UK Army NBC and the US Air Force suit, respectively. Yet, T_c prediction was greater than observed with the US Air Force garment when using this model. Evaporative heat loss predictions through these clothing systems were also comparable at ~88 and 95 W · m².

Overprediction of T_c by use of the trunk node in the LUT25 model is possibly affected by the estimations of distribution of blood flow to the core (trunk node) and shell. In the unclothed state, effective shell thickness is a variable that increases with vasoconstriction but narrows with vasodilation. The fraction of body mass in the shell at any time (t) during exercise may be described by  (3, 12) in which

(9)

where the proportion of body mass in the core is given by (1  _t) and SkBf is skin blood flow. In the S-H clones, the algorithm for controlling sweating rate (_sw) weights these proportions according to the relative influence of core and skin contribution. When subjects are unclothed or when wearing thin porous clothing, prediction of various responses is quite accurate from thermoregulatory models employing the above algorithm. However, a more complex model algorithm incorporating axial and radial heat conductance that deals with latent heat release or storage in multilayered clothing may be a future goal necessary to predict finite T_c changes with protective clothing (2, 3, 30, 33).

We noted a curious dip in T_c (trunk node temperature) occurring from the LUT25 output simulations at the start of exercise when this variable was plotted as a function of time (Fig. 7). One reason for the unique dip observed when applying the LUT25 simulation may be because the model follows passive state predictions at each node held over from the original S-H model (8, 23). Both models reflect axial heat conductance through each layer established for a horizontal cylinder. Additionally, in the original passive equations of the S-H model's algorithms, the redistribution of cooler skin blood flow occurs from the extremities toward the trunk core on the initiation of exercise. This often happens at the initiation of treadmill or cycle ergometer exercise. A recent report (15) suggests that the latter transient sequestration of cool blood from the lower body extremities toward the pelvic area is a reasonable interpretation for the initial drop in T_c evident in the LUT25-node model.

One feature simulated equally well by both models was the tracking of T_c response during intermittent exercise (Canadian experiments). The model prediction confirms the results of other studies (1, 2, 4, 17). These studies show that employment of work/rest cycles in which the work cycle is 1 l/min (or, roughly, ~25-27% O_{2 max}) reduces final T_c, improves tolerance to an environmental challenge, diminishes rate of heat storage, and maximizes evaporative potential through CP garments.

Along with the LUT25 model used in this evaluation, several other physiological models have appeared that have embraced the original thermoregulatory control algorithms formulated in the S-H 25-node prototype (12, 26). One model developed by Kraning (12) is worth mentioning. This model has merged the clothing heat and mass transfer characteristics derived by copper manikin evaluations with the heat and mass transfer equations appearing in the S-H model. Kraning's model also integrates many positive features of both the S-H model and the prediction capabilities found in the USARIEM heat strain model. Kraning's SCENARIO simulation routines have coupled 1) the combined effects of posture, metabolic activity level, clothing coefficients, T_c, and skin temperature influences on cardiac stroke volume; 2) the effect of "cardiovascular overload" during work-in-the-heat routines on increasing muscle oxygen extraction, thereby relieving the overload; 3) effects of both T_c and skin temperature as modulators of the central temperature "set-point" for controlling skin blood flow; and 4) effects of age as a factor in decreasing maximal HR on thermoregulatory responses.

Modeling of Heat-Acclimation Response

Heat acclimation during sustained operations with light work clothing or combat fatigues is certainly obvious, since evaporative potential and performance are maximized. However, its operational importance while subjects are wearing protective clothing is not certain and possibly becomes a limiting variable because of excessive sweating, which is dependent on work intensity and magnitude of evaporative potential through a clothing system. The enhanced sweating at a lower T_c produces excessive skin wettedness (E_sk/E_max) with limited cooling benefits (3, 25) when exercise continues for an extended period of time underneath protective clothing. If the water lost is not replaced adequately, individuals can become hypohydrated (22). Further investigations of sweating and skin blood flow responses might reveal additional interesting observations useful to modeling of the UK and Canadian heat acclimation phases not within the scope of this report.

Simulations using either the USARIEM heat strain or UK LUT25 models offer reliable predictions of T_c responses (within 1 RMSD) during work in the heat while subjects are wearing protective clothing. However, only the USARIEM heat strain model simulated responses adequately for both the unacclimated and heat-acclimated phases during continuous work (UK and US trials) and intermittent work (Canadian trial). From the data vs. model comparisons, it is clear that direct thermal and vapor resistance evaluations are especially important when protective clothing is used in which new semipermeable laminates are structured into the garment (Table 1). If manikin data are not available, a method to transform clothing and vapor transfer resistance values for civilian clothing may be applicable for protective clothing as well (19).

In summary, our study demonstrated that analytically similar thermoregulatory models modified to predict clothing heat transfer can easily serve as discerning tools suitable for various occupational heat-stress scenarios. These models are appropriate tools for use as guidelines in guarding against overheating, ascertaining work/rest cycles, predicting appropriate stay times, and cooling power requirements based on radiative and evaporative heat loss in civilian workers wearing protective clothing. Prediction analysis by use of such models is a practical way to obviate direct T_c monitoring in hazardous-materials workers wearing CP clothing systems.

ACKNOWLEDGEMENTS

We thank the soldiers who volunteered to participate as subjects in this investigation. We acknowledge many contributors who assisted in the experimental study, including Dr. Stefan Constable, Armstrong Laboratory, Brooks Air Force Base, TX, for providing the US Air Force CP ensemble; Dr. Claire E. Millard and Mike Neale, Research Assistant, Centre for Human Sciences, Defence Research Agency, Farnborough, UK; and Robert Wallace, Statistician, USARIEM, Natick, MA.

FOOTNOTES

Address for reprint requests: R. R. Gonzalez, Biophysics and Biomedical Modeling Division, USARIEM, Natick, MA 01760-500 (E-mail: RGonzalez{at}natick-ccmail.army.mil).

Received 22 October 1996; accepted in final form 12 May 1997.

APPENDIX

An iterative software model employing the Givoni-Goldman equations was developed for implementation on a personal computer. A full explanation of the modifications to the original Goldman-Givoni (7) program with current algorithms may be found in Pandolf et al. (17). The modifications present in the version used in this study are given in the following sections.

Improvements in parameter equations and the model's implementation on PC-DOS and Macintosh platforms include the ability to use multiple data sets for individual input variables; a user-friendly, pull-down menu user interface; the ability to use a variety of units when entering data; graphic analysis of data; and automatic file saving for future use and/or retrieval. These improvements greatly enhance the program's ability not only to generate final results for work and recovery categories but also to allow the user to graphically review the generated results.

USARIEM Experimental Heat Strain Model

(A1)

where T_re,0 is initial value for T_re based on an experiment, defaults to 36.8°C; T_re,f is equilibrium T_c (°C); M is energy expenditure or metabolic rate (W); W_ex is external work (W); Dry is radiative and convective heat exchange (W); E_req is evaporative heat exchange required (W); and E_max is maximum possible evaporative heat exchange (W)

(A2)

where  is incline of slope (%grade); Wt is weight of person (kg); L is weight of clothing, equipment, and load (kg); and V_w is velocity of walking (m/s).

Heat Exchange Routines

(A3)

1

(A4)

(A5)

Clothing-Related Routines

(A6)

(A7)

(A8)

(A9)

DuBois Body Surface Area

(A10)

(A11)

(A12)

2

1

1

T_re and Change Calculations

(A13)

where T_re,0 is the initial value for T_re (°C), default is 36.8°C; T_re,f is final T_re (°C); E_max is maximum evaporation heat transfer of the environment (W); and E_req is required evaporative heat transfer (W) to achieve thermal balance

(A14)

where _Tre,f is change in final T_re for heat acclimation and dehydration (°C); and DIH is days in heat (heat acclimation level base).

Water Requirements

(A15)

(A15a)

(A15b)

(A15c)

(A16)

Maximum Work Time Calculation

(A17)

Else, if T_re,fwork + T_re,fwork > T_re,fstart and if maximum work time < 0.0, then maximum work time = 0.0. Else

(A17a)

where t_delay is time lag (min) for work to induce T_c changes at given M. Comparison of T_re f(t) for various M (W) indicates that this value is shorter as M is higher (evaluated from a database of multiple studies by the equation t_delay = 27.84  0.072 · M + 5.93 × 10⁵ · M²). K_work is exponential rate coefficient for work output (°C/min).

The value of the change of T_c, T_re,fwork (°C) is used to calculate K_work (°C/min) as well as the T_c predicted at time t (T_re,t) is calculated from

(A18)

where T_re,0 is the T_c at the end of the delay time period.

(A19)

REFERENCES