Temperature-Humidity Dependent Wind Effects on Physiological Heat Strain of Moderately Exercising Individuals Reproduced by the Universal Thermal Climate Index UTCI

1. Introduction

1.1. Sustainable heat stress mitigation by wind

Climate change accelerating with fossil energy use will exacerbate heat extremes, which will provoke physiological heat strain in terms of increased heart rates, core and skin temperatures, and sweat rates imposing health risks [1,2] and potentially affecting human performance and productivity [3,4]. Elevating air movements by outdoor wind or ventilators is a cost-effective and sustainable measure for mitigating physiological heat strain [5,6,7]. However, policies and public health authorities have advised against electric fan use under heat wave conditions with air temperatures above the typical skin temperature of 35 °C, because convective cooling will then turn to convective heating of the body [8,9,10]. On the other hand, the well-known enhancement of sweat evaporation with increased wind speed [11,12,13] challenged these postulations and advocated for an evidence-based approach [14]. Recent studies, mostly with sedentary participants [15,16], demonstrated cooling effects of ventilation due to enhanced sweat evaporation with even higher temperatures at elevated humidity levels.

Comparative studies involving experimental heat exposures are usually restricted to a few well-defined climatic conditions [15]. Thus, for separating beneficial cooling from detrimental heating wind effects over a grid of temperature-humidity conditions in heat wave scenarios, several studies applied biophysical modelling based on human heat balance calculation [17,18,19,20]. Interestingly, qualitatively similar threshold curves based on modelling the naturally aspirated wet-bulb temperature had been developed one century ago [21]. These curves indicated wind cooling at elevated humidity levels, but additional heat load in hot-arid climates, as flagged by the term ‘poison wind’ (simoom) for potentially fatal outdoor conditions in Southwest Asian desert regions [21]. Notably, the simulations of heat wave scenarios only considered resting conditions concerning metabolic rates of about 1 MET (1 MET=58.2 W/m²), whereas physical load at workplaces and many home and outdoor activities are associated with metabolic rates higher than 2 MET [22,23].

Concerning physiological heat strain under moderate exercise levels, the recorded heart rates, core and skin temperatures, and sweat rates in Figure 1 from an extensive database of climate chamber experiments [24,25] exemplify the aforementioned findings for an acclimated young male walking 4 km/h on a treadmill corresponding to 2.3 MET metabolic rate. While increasing air velocity from 0.3 to 2 m/s lowered physiological heat strain in a hot-humid climate with air temperature exceeding skin temperature (Figure 1a), increased wind amplified heat strain under hot-dry conditions (Figure 1b). Figure A1 in the Appendix A includes a supporting example from another participant. In both cases, the transient overshooting of sweat rate with low wind (v_a = 0.3 m/s) under warm-humid conditions (Figure 1a) may be explained by hidromeiosis [26], which did not occur when increasing wind speed to v_a = 2.0 m/s enhanced sweat evaporation [27].

A recent study with exercising participants [28] examined how wind speed at various temperature-humidity combinations affected physical work capacity, defined by the workload related to metabolic rate, which the participants could tolerate with heart rate clamped at 130 beats per minute (bpm). Coupling their experimental data with heat balance calculations, they showed beneficial wind effects for air temperatures up to 44 °C with relative humidity exceeding 50%, but detrimental wind effects for higher temperatures. Similar findings had been reported concerning the shift of the critical, i.e. tolerable humidity level by wind in hot environments [29]. Another study [30] showed beneficial wind effects for sports activities at specific conditions with 37 °C air temperature and 50% relative humidity.

However, there is a lack of data from controlled laboratory experiments specifying wind effects on physiological heat strain in terms of heart rates, core and skin temperatures, and sweat rates while covering a wide grid of air temperature-humidity combinations.

In addition to data requirements, there seems to be a need for improving the thermo-physiological modelling beyond the heat balance approach, as recent findings suggest that the simple models will underestimate physiological heat strain under electrical fan use, e.g. for the vulnerable elderly population [31].

When looking for advanced, but easy-to-use modelling approaches, the Universal Thermal Climate Index UTCI [32] developed for moderately active persons (2.3 MET) constitutes a noteworthy alternative. The index was constructed from the dynamic physiological thermal strain simulated by the advanced UTCI-Fiala model of human thermoregulation [33] coupled with an adaptive clothing model [34]. UTCI allows for the assessment of the thermal environment covering the range from extreme cold to extreme heat stress conditions. The model was extensively validated [35], and showed good agreement with thermal environment standards and experimental data in occupational settings [36,37,38]. Though UTCI relies on sophisticated models, the operational procedure [39] provides algorithms making it easily applicable, e.g. for assessing the wind cooling potential of actual and future urban scenarios in relation to wind direction [40]. Concerning the effects on physical work capacity, UTCI outperformed any other considered thermal index in capturing not only the effects of wind over a wide grid of temperature-humidity conditions [28], but also in combination with heat radiation [41].

1.2. UTCI sensitivity to wind

Following meteorological conventions, UTCI calculations rely on air velocity 10 m above ground (v_a,10m). For conversion to any other measurement height, e.g. 1 m for person level, the operational procedure [39] provides a logarithmic formula shown in Eq. 1, indicating that air velocity at person level (v_a,1m) is computed as the 10 m value (v_a,10m) divided by 1.5, in accordance with international standards [42].

v_a,1m = v_a,10m × log(1/0.01) / log(10/0.01) = v_a,10m / 1.5

(1)

As the UTCI person is assumed to move on the level with 4 km/h, corresponding to a walking speed v_w = 1.1 m/s, this is taken into account by calculating the resulting or relative air velocity at person level (v_ar,1m) according to Eq. 2. Here, α denotes the angle between the directions of walking and wind, assigned to zero for indicating the same direction. As UTCI does not assume a specific angle, v_ar,1m is calculated by integrating Eq. 2 over all α between zero and 2π [34,43].

$$v_{ar,1m}=\sqrt{{\left(v_w{-v}_{a,1m}\times cos\alpha \right)}^2+{\left(v_{a,1m}\times sin\alpha \right)}^2}$$

(2)

Figure 2 presents the resulting v_ar,1m used by UTCI for calculating the convective and evaporative heat loss [33]. Notably, reducing wind speed below v_a,10m = 0.5 m/s (v_a,1m = 0.3 m/s) will hardly impact v_ar,1m, which is limited by v_w, while v_ar,1m will approach v_a,1m for v_a,10m above 3 m/s.

1.3. Study objectives

Based on a comprehensive physiological heat strain database [24], this study aimed at exploring and quantifying whether reports on beneficial wind effects with air temperature exceeding skin temperature under humid conditions for sedentary individuals are transferable to moderate exercise levels, and whether the Universal Thermal Climate Index UTCI reproduces these effects.

2. Materials and Methods

2.1. Experimental data

We compiled a database of climate chamber experiments conducted previously at IfADo [24,44] according to the ethical principles of the Declaration of Helsinki after approval by IfADo’s local Ethics Committee.

We searched our database for series of heat stress experiments with participants exercising on a treadmill under both reference (RefWind, with air velocity v_a = 0.3 m/s) and high wind (HiWind, v_a = 2 m/s) conditions. Inclusion criteria were a minimum number of 15 experiments per series with comparable workload and clothing. We retrieved 198 trials organized in 10 series, which originated from five acclimated semi-nude (basic clothing insulation I_cl = 0.1 clo, 1 clo = 0.155 K·m²·W⁻¹) young fit males under RefWind and HiWind conditions, respectively. The number of experiments in each series varied inter-individually and depending on wind condition between 16 and 25 experiments, with total numbers of 97 trials for RefWind and 101 for HiWind. The average personal characteristics (mean ± SD) of the five participants were 20.1 ± 0.9 years of age, 1.87 ± 0.02 m of body height, 70.5 ± 2.1 kg of body weight, 1.94 ± 0.02 m² of body surface area, and 47.9 ± 6.4 mL/min/kg of maximal oxygen consumption. Before exposure, the participants had undergone a heat acclimation protocol enduring at least three weeks [45].

We briefly summarize the procedures as detailed descriptions are available elsewhere [24]. Each trial consisted of treadmill work with constant workload of walking 4 km/h on the level, which corresponded to the activity level of 2.3 MET, as assumed for UTCI [39]. Each trial lasted for at least three hours and consisted of 30 min work periods interrupted by 3 min breaks for determining body weight loss, from which sweat rate (SR) was calculated correcting for ad libitum fluid ingestion. Under both HiWind and RefWind, the participants were exposed to varying levels of heat stress with conditions characterized by different combinations of air temperature (T_a; range 25 – 55 °C) and humidity, expressed as water vapour pressure (p_a; 0.5 – 5.3 kPa). Mean radiant temperature (T_mrt) was equal to T_a. Trials were stopped prematurely if rectal temperature exceeded 38.5 °C or on the participant’s demand.

Rectal temperatures (T_re) were recorded continuously using a thermistor probe (YSI 401, YSI Inc., Yellow Springs, Ohio) inserted 10 cm past the anal sphincter. Local skin temperatures were measured with thermistors (YSI 427, YSI Inc., Yellow Springs, Ohio) at the forehead (T_sk,head), chest (T_sk,chest), back (T_sk,back), upper arm (T_sk,arm), thigh (T_sk,thigh), and lower leg (T_sk,leg), and used to calculate mean skin temperature (T_sk) as weighted average of the local skin temperatures according to [24], shown in Eq. 3:

T_sk = 0.05×T_sk,head + 0.2×T_sk,chest + 0.15×T_sk,back + 0.2×T_sk,arm + 0.25×T_sk,thigh + 0.15×T_sk,leg

(3)

Heart rates (HR) were obtained using ECG electrodes, and stored in one-minute intervals, as were T_re and T_sk. As illustrated by the yellow shaded areas in Figure 1, the averages of T_re, HR, T_sk, and SR over the third hour of exposure representing steady-state [25] were submitted to the following data analysis.

2.2. Data analysis and statistics

Penalized regression splines fitted by generalized additive models (GAM) provide a flexible modelling framework for irregularly distributed data [46] allowing for non-linear [47] and random effects [48].

E[Y_ij] = µ + s(ID_i) + te(T_a,ij, p_a,ij) + te_Δv(T_a,ij, p_a,ij) + ε_ij

(4)

Using the notation from Eq. 4 [46], we predicted HR, T_re, T_sk, and SR, respectively, as responses from experiment j of participant i (Y_ij) by separate GAMs. These models included an overall intercept µ, and a bivariate penalized tensor regression spline te(T_a,ij, p_a,ij) for the effects of air temperature and humidity depending on the (T_a, p_a)-combinations, accounting for the repeated measurements by subject-specific intercepts as random coefficients s(ID_i) [48]. Adding another bivariate spline te_Δv(T_a,ij, p_a,ij) as so called factor smooth interaction [46] of the regression splines with the wind condition (RefWind vs HiWind), we obtained estimates for the wind effect Δ_v, i.e. the response difference under HiWind compared to RefWind over the (T_a, p_a) grid supplemented by p-values [49]. These were then visualized in a difference plot [50].

More specifically, s(ID_i) refers to a cubic regression spline with basis dimension (or rank) set to k=5 (i.e. with k–1=4 as upper limit of the associated degrees of freedom). Similarly, te(T_a,ij, p_a,ij) and te_Δv(T_a,ij, p_a,ij) represent tensor products of two cubic regression spline bases for the T_a and p_a dimension, respectively, each of rank k=9, and are hence of total rank k=81. Maximum likelihood parameter estimates with standard errors and p-values were obtained assuming Gaussian error ϵ_ij [49]. Calculations were performed with the R software version 4.2.1 [51] using the package mgcv [46] together with mgcViz [50] and tidymv [52].

2.3. UTCI calculations

We calculated UTCI values using the regression polynomial as described in the UTCI operational procedure [39] for given combinations of air temperature (T_a), ambient water vapour pressure (p_a), air velocity 10 m above ground (v_a,10m), and mean radiant temperature (T_mrt). Then, we computed the effect of actual wind speed v_a,10m on UTCI (Δ_vUTCI) compared to the reference wind speed (v_a,10m,ref = 0.5 m/s), while keeping the other parameters constant, as shown in Eq. 5:

∆_vUTCI = UTCI(T_a;p_a;v_a,10m;T_mrt ) - UTCI(T_a;p_a;v_a,10m,ref;T_mrt)

(5)

Matching the range of experimental climatic conditions (section 2.1), but respecting the upper limits of UTCI validity [39,53], we computed UTCI and Δ_vUTCI for combinations of T_a between 25 – 50 °C and of p_a from 0.1 – 5 kPa with relative humidity rH ≤ 100%, while setting T_mrt = T_a as in the experiments. We performed the calculations for the reference wind speed (v_a,10m,ref = 0.5 m/s) matching RefWind with v_a,1m = 0.3 m/s (Figure 2) and increased wind speeds with v_a,10m = 3 m/s matching HiWind with v_a,1m = 2 m/s.

Supplemental calculations were performed for assessing the modifying effect of radiant heat load by increasing ΔT_mrt=T_mrt-T_a from 0 to 30 K in steps of 10 K. In addition, we considered conditions with higher wind speeds v_a,10m = 4 and 6 m/s, respectively. Here, v_a,10m = 4 m/s corresponds to an increase of 1.7 m/s in relative air velocity at body level (v_ar,1m) according to Figure 2, matching the settings of the climate chamber experiments; whereas v_a,10m = 6 m/s (v_a,1m = 4 m/s) parallels the wind speed applied in several simulations and experimental studies [15,17,18,19,20,28].

3. Results

Section 3.1 presents the wind effects (HiWind vs RefWind) evaluated by UTCI, and section 3.2 contains the corresponding analyses for the experimental heat strain data, while section 3.3 relates the physiological wind effects to the UTCI assessment. Eventually, section 3.4 shows the results for the supplemental UTCI analyses considering higher wind speeds and the influence of thermal radiation.

3.1. Wind effects assessed by UTCI

The psychrometric charts in Figure 3 display UTCI contours for selected index values representing the upper limits of the UTCI heat stress categories in relation to air temperature and humidity for the different wind speeds (Figure 3a). The contours were bended upwards to the left indicating that heat stress increased with temperature and humidity, as expected [36,39]. The dashed lines (HiWind) were above the solid lines (RefWind) indicating wind cooling at temperatures below 34–35 °C and for hot-humid conditions. On the other hand, dashed lines below solid lines indicating heating by wind were found for hot-dry conditions. The dot-dashed line connecting the intersections of the RefWind and HiWind contours represents zero wind effect Δ_vUTCI = 0.

Figure 3b quantifies the magnitude of the wind effect Δ_vUTCI depending on temperature and humidity. Wind cooling (blueish) increased with decreasing temperature and increasing humidity while heating due to wind (reddish) increased with increasing temperature and decreasing humidity.

3.2. Wind effects on physiological heat strain

Table 1 summarizes the results of the GAMs fitted separately to the physiological heat strain variables. While data for T_re were complete, there were a few missing values for HR, T_sk and SR. The estimated mean values of 103 bpm (HR), 37.6 °C (T_re), 35.5 °C (T_sk), and 744 g/h (SR), respectively, deemed typical for light to moderate activities under heat stress [24,38]. Goodness-of-fit was good to excellent with more than three quarters of variance explained (R²>75%) and small residual standard errors for HR, T_re and T_sk. SR showed almost 90% explained variance, but a slightly increased standard error, which, however, was still below 125 g/h, considered as the relevant accuracy limit acceptable in occupational or military settings [54]. Moreover, graphical model checking as displayed in the Appendix A by Figure A2 did not reveal any problematic issues concerning the underlying model assumptions.

For all heat strain variables, the splines te(T_a, p_a) modelling the influence of temperature and humidity were statistically highly significant (p<.0001). Similarly, the moderating effects of wind speed as considered by the bivariate interaction splines te_Δv(T_a, p_a) were statistically significant for HR, T_re, T_sk and SR.

Inter-individual variability, as indicated by the random term s(ID) played a considerable role (p<.0001) for T_re, T_sk and SR, respectively, whereas individual influences on HR appeared to be less pronounced (p = .20).

Analogous to the UTCI results from Figure 3a, Figure 4 displays the effect of temperature and humidity separately for RefWind and HiWind as contours for the physiological heat strain responses of HR, T_re, T_sk and SR. Although the shape of the lines varied between the four variables, and additionally depended on the strain level, a general pattern emerged with contours bended upwards to the left indicating strain levels increasing with temperature and humidity, as it had also occurred for UTCI (Figure 3a). Again similar to UTCI, the dashed lines (HiWind) were above the solid lines (RefWind) at low temperatures and under hot-humid conditions demonstrating reduced strain due to wind cooling, whereas for hot-dry climates, the opposite was found with solid lines above dashed lines indicating higher strain levels with increased wind speed.

The dot-dashed lines representing zero wind effect Δ_v = 0, thus marking the threshold separating cooling from heating effects, was bended upwards right similar as for UTCI (Figure 3b), but with a greater curvature. They indicated reduced heat strain with HiWind at low temperatures and for hot-humid conditions, with limiting vapour pressure above 2 kPa for HR and T_re, and above 3 kPa for T_sk and SR, respectively.

The difference plots in Figure 5 show the intensity of the cooling (blueish) and heating (reddish) wind effects over the temperature-humidity grid estimated by GAM, where white areas indicate indifferent temperature-humidity combinations with statistically non-significant differences from Δ_v = 0.

Whereas no significant core temperature increase due to wind was observed, detrimental wind effects with increased HR, T_sk and SR were found under hot-arid conditions with temperature above 40 °C and vapour pressure below 1 kPa. Temperatures below 35 °C yielded no detrimental wind effects.

Though significant decreases of core temperature and heart rates by wind under high temperatures started at vapour pressure above 2 kPa, pronounced effects with reductions by more than 10 bpm HR or 0.15 °C T_re only appeared for very humid conditions with vapour pressure approaching 4 kPa.

3.3. UTCI assessment related to physiological wind effects

The scatterplots in Figure 6 comparing the wind effects (Δ_v) on UTCI with the physiological effects showed positive correlations with the closest agreement (r≈0.9) for skin temperature and sweat rate, also concerning the transition from cooling (Δ_v<0) to heating (Δ_v>0) effects. Slightly lower correlations were observed with the effects on heart rates and core temperature.

In addition to correlation analyses, the plots give an indication for the capacity of UTCI for correctly classifying conditions with physiological wind cooling or heating [56]. The dashed reference lines in Figure 6 divide the plot in quadrants with the lower left quadrant showing the physiological cooling events correctly assessed by UTCI and the upper right containing the correctly assessed heating events, which dominate concerning T_sk and SR.

Whereas hardly any data points were present in the upper left quadrant with wind cooling falsely predicted by UTCI, the conditions collected in the lower right quadrants (falsely predicted heating) concur with a more conservative UTCI assessment of the transition from cooling to heating wind concerning HR and T_re.

3.4. UTCI assessment with higher wind speeds and thermal radiation

Figure A3 in Appendix A visualizes the wind effects (Δ_v) on UTCI over the temperature-humidity grid for the different wind speeds and radiant heat loads, expressed as ΔT_mrt = T_mrt – T_a, with the upper left panel with v_a,10m = 3 m/s and ΔT_mrt = 0 K repeating the data from Figure 3b. Figure 7 summarizes the wind effects Δ_vUTCI depending on wind speed and radiant heat load as boxplots calculated over the temperature-humidity grid.

As expected, increasing wind speed amplified both the cooling (Δ_vUTCI<0) and heating effects (Δ_vUTCI>0) at the extremes, as indicated by the minima and maxima, but also the 1^st and 3^rd quartiles in Figure 7. In addition, the threshold line for Δ_vUTCI = 0 slightly shifted towards higher temperatures for higher wind speeds (Figure A3), thus contributing to the decreasing trend with wind speed observed for the median effect in Figure 7.

Concerning the influence of heat radiation on Δ_vUTCI, all percentiles showed a decreasing trend with increasing radiant heat load ΔT_mrt (Figure 7). The reduced reddish surface areas with increased radiant heat load in Figure A3 suggest that this was attributable to a massive shift of the threshold line towards higher temperatures. Thus, wind-cooling effects occurred for a higher portion of climatic conditions over the temperature-humidity grid, shifting the Δ_vUTCI distribution towards more negative values (Figure 7).

The wind effect Δ_vUTCI at v_a,10m = 4 m/s (with ΔT_mrt = 0 K) corresponded to the effect of increasing relative air velocity (v_ar,1m ) by 1.7 m/s (Figure 2), conforming to the change in (relative) air velocity realized in the experiments. As the relevant UTCI calculations of convective and evaporative heat losses depend on relative air velocity [33], we repeated the correlational analyses from Figure 6 for this condition, as shown in Figure A4. The outcome was very similar as before in Figure 6, though we observed slightly lowered correlations of the physiological wind effects with Δ_vUTCI.

4. Discussion

4.1. Temperature-humdity depending wind effect thresholds

With a view on sustainable heat stress mitigation for moderate activity levels occurring during domestic or industrial work, we assessed the wind effects on physiological heat strain parameters connected with the use of, e.g. electric fans or ventilators. Linking a comprehensive database originating from climate chamber experiments [24] with flexible spline regression models [46], we quantified the physiological heat strain responses of moderately exercising individuals to increasing wind speed over a wide grid of temperature-humidity combinations. Thus, we could derive the temperature-humidity dependent thresholds marking the transition from cooling (Δ_v<0) to heating (Δ_v>0) wind effects on physiological heat strain based on experimental data. This is a distinguishing feature and particular strength of our study, because corresponding experiments with sedentary participants were limited to a few selected climatic conditions [15,16], so that mapping wind effect thresholds in relation to air temperature and humidity for low-active persons had to rely on simulation studies [17,18,19,20]. Recently, a study with exercising participants [28] aimed at the interaction of wind with combinations of temperature and humidity over a likewise extensive grid, but focused on the physical work capacity with cardiac strain clamped at 130 bpm HR.

Our experimental findings suggest beneficial, i.e. strain reducing wind effects for temperatures below 35 °C, conforming to common public health recommendations [8,9,10]. However, beneficial wind effects were also observed at higher temperatures with humidity levels above 2 kPa water vapour pressure concerning heart rate and core temperature, and 3 kPa concerning skin temperature and sweat rate, respectively (Figure 5). This agreed qualitatively with corresponding charts produced in previous studies [19,28]; though there were discrepancies concerning the humidity level separating beneficial wind cooling from detrimental heating effects, which was specified at 50% relative humidity in the former studies, while our study suggested thresholds in terms of absolute humidity (vapour pressure). Some of these discrepancies may be explainable by the 3-h exposures in our study, whereas shorter exposures (1 h) had been used in previous studies [28], with longer exposure times usually resulting in higher strain levels [25,57]. The lower humidity limits for skin temperature and sweat rates compared to core temperature and heart rates in our study may reflect the direct influence of wind speed on the heat transfer from the skin surface by convection and evaporation, and on sweating efficiency [11,12,13,58].

For UTCI, no limiting humidity in terms of vapour pressure were identified, but wind-cooling effects were evident for any temperature condition with relative humidity above 50% (Figure 3b). UTCI assessment of wind effects showed high positive correlations with the observed changes in physiological responses (Figure 6), exhibiting the closest agreement (r=0.9) for skin temperature and sweat rate, where wind is known for elevating the relevant convective and evaporative heat transfer as discussed before. UTCI showed also a good agreement concerning the transition from cooling (Δ_v<0) to heating (Δ_v>0) wind effects on skin temperatures and sweat rates. The UTCI assessment was more conservative, i.e. overpredictive concerning the potentially hazards due to additional heating by wind with respect to core temperature and heart rates (Figure 6).

The expanded UTCI calculations suggested that further increasing wind speed would amplify the cooling and heating effects depending on temperature and humidity (Figure A3, Figure 7). In addition, these calculations pointed to a relevant interaction with radiant heat load, e.g. arising from outdoor exposure to solar radiation. Elevated radiant heat load will increase UTCI by approximately 3 °C per 10 K increase in ΔT_mrt [36]. On the other hand, the limit for beneficial wind-cooling was shifted towards higher temperatures, counteracting, at least partly the heat gained from increased radiant heat load. Similar interaction effects with wind decreasing the heat gain from radiant heat had been reported from thermal manikin measurements with work clothes [59,60]; however, further supporting experimental studies with human participants concerning the interaction of wind with radiant heat load are required.

4.2. Limitations and outlook

The selected study population with young healthy fit males certainly limits the generalizability of our findings, though recent research suggests only minimal influences of individual factors on the heat stress assessment under moderate activity levels [61], as applied in our study. As another limitation, the responses of semi-nude participants (I_cl = 0.1 clo) might differ from the effects with clothed humans, e.g. as assumed by UTCI with clothing insulation not falling below I_cl = 0.3 clo [34]. On the other hand, the prior studies with sedentary participants [15,16] applied similar minimal clothing settings. In addition, limited differences were found concerning the wind effects on physical work capacity with minimal clothing compared to light and covering clothing [28], and the corresponding temperature-humidity charts in that study were produced for a person wearing I_cl = 0.28 clo.

Given these limitations, future experimental and modelling studies should consider the moderating influence of individual factors like age, sex or body constitution, and of solar radiation and/or clothing with relevance, e.g. for desert conditions (simoom) [21], but also for (semi-)outdoor workplaces.

5. Conclusions

Widening previous findings with resting persons to moderately exercising individuals, our results confirm the wind cooling effects for air temperatures below 35 °C, and corroborate and quantify the potential wind cooling effects at air temperatures above typical skin temperature values in humid climates.

In addition, our findings suggest that UTCI has the capability for adequately assessing the heat stress mitigation by wind depending on temperature and humidity under such conditions. Thus, UTCI constitutes an easy-to-use tool that may be helpful for establishing evidence-based policies aiming at sustainable heat mitigation strategies under climate change scenarios.