Methods for Estimating Wet Bulb Globe Temperature From Remote and Low-Cost Data: A Comparative Study in Central Alabama

Abstract

Heat stress is a significant health concern that can lead to illness, injury, and mortality. The wet bulb globe temperature (WBGT) index is one method for monitoring environmental heat risk. Generally, WBGT is estimated using a heat stress monitor that includes sensors capable of measuring ambient, wet bulb, and black globe temperature, and these measurements are combined to calculate WBGT. However, this method can be expensive, time consuming, and requires careful attention to ensure accurate and repeatable data. Therefore, researchers have attempted to use standard meteorological measurements, using single data sources as an input (e.g., weather stations) to calculate WBGT. Building on these efforts, we apply data from a variety of sources to calculate WBGT, understand the accuracy of our estimated equation, and compare the performance of different sources of input data. To do this, WBGT measurements were collected from Kestrel 5400 Heat Stress Trackers installed in three locations in Alabama. Data were also drawn from local weather stations, North American Land Data Assimilation System (NLDAS), and low cost iButton hygrometers. We applied previously published equations for estimating natural wet bulb temperature, globe temperature, and WBGT to these diverse data sources. Correlation results showed that WBGT estimates derived from all proxy data sources-weather station, weather station/iButton, NLDAS, NLDAS/iButton-were statistically indistinguishable from each other, or from the Kestrel measurements, at two of the three sites. However, at the same two sites, the addition of iButtons significantly reduced root mean square error and bias compared to other methods.

Keywords: WBGT; health; heat stress.

FIGURE 1

Map of Kestrel monitoring sites, weather stations, and iButtons used in Alabama. Distances between downtown Birmingham data sources are as follows: Kestrel monitor ➔ Birmingham Airport 8.8 km, Kestrel monitor ➔ iButton 0.6 km. Distances between suburban Birmingham data sources are as follows: Kestrel monitor ➔ Birmingham Airport 16.8 km, Kestrel monitor ➔ iButton 0.02 km. Distances between Wilcox County data sources are as follows: Kestrel monitor ➔ Crenshaw Airport 70.5 km, Kestrel monitor ➔ iButton 0.09 km.

FIGURE 2

Daily average time series of wind speed at each monitor site.

FIGURE 3

Daily average time series of temperature (T_a) at each monitor site.

FIGURE 4

Daily average time series of relative humidity at each monitor site.

FIGURE 5

Linear regressions of Kestrel reported T_g and T_nwb vs. calculated T_g and T_nwb based on daily averages. Black solid lines represent a 1/1 slope. Downtown Birmingham (n = 37), suburban Birmingham (n = 38), Wilcox County (n = 10).

FIGURE 6

Linear regressions of Kestrel reported WBGT vs. calculated WBGT based on daily averages. Black solid lines represent a 1/1 slope.

FIGURE 7

Daily average time series of instrument recorded and calculated values for T_g, T_nwb, T_a, and WBGT for downtown Birmingham, suburban Birmingham, and Wilcox County for the holdout period.

FIGURE 8

Diurnal cycles of T_a, T_g, T_nwb, and WBGT in downtown Birmingham from 19–22 September 2017.

FIGURE 9

Mean percentage of hours in recommended work‐rest schedule based on WBGT index across all iButtons (mean and 95% confidence interval) in moderate workload. Moderate workload has a metabolic rate of approximately 300 W, examples are normal walking and moderate lifting. Thresholds for % work recommended per hour: Continuous work <28°C, 75–100% work (45 to 60 min work/0 to 15 min rest per hour) >28°C and <29°C, 50–75% work (30 to 45 min work/15 to 30 min rest per hour) >29°C and <30°C, 25–50% work (15 to 30 min work/30 to 45 min rest per hour) >30°C and <31.5°C, 0–25% work (0 to 15 min work/45 to 60 min rest per hour) ≥31.5°C (NIOSH, 2017). n = 28 iButtons in Jefferson County and n = 14 iButtons in Wilcox County. Only one WS in each county.

FIGURE 10

Map of central Alabama showing NLDAS (grid) and NLDAS/iButton (points) calculated WBGT (°C) in August of 2017.