Global labor loss due to humid heat exposure underestimated for outdoor workers

There are a variety of methods for estimating humid heat exposure (Buzan and Huber 2020, Schwingshackl et al 2021). Here, we use an approximation of WBGT (Yaglou and Minard 1957), an internationally recognized heat stress metric that incorporates temperature and humidity measures and is used in occupational health studies (Lemke and Kjellstrom 2012, American Conference of Governmental Industrial Hygienists 2015, Jacklitsch et al 2016). The advantage of a metric such as WBGT is that it considers both temperature and humidity in relation to heat exposure and heat stress for working individuals. The term 'humid heat' here refers to conditions that are either hot (and dry) or hot and humid enough to cause decreases in labor productivity. Although WBGT has its limitations (Bates and Schneider 2008, Holmér 2010, Dehghan et al 2012, d'Ambrosio Alfano et al 2014), we use sWBGT to estimate humid heat exposure and labor impacts, due to its common use in previous studies (e.g. Watts et al 2018, 2021, Li et al 2020) and the availability of hourly reanalysis data to calculate sWBGT (near-surface air temperature, humidity, and pressure). Following the method of Li et al (2020), we use hourly European Centre for Medium-Range Weather Forecast 5th Generation Reanalysis (ERA5) output (1 January 1981–31 December 2020), including t2m (2 m air temperature), surface pressure (sp), and 2 m dew point temperature (d2m) to calculate sWBGT. ERA5 is provided at ∼31 km spatial resolution; when ERA5 sWBGT and hours lost data are compared to population data, the hours lost data are regridded to the population resolution (∼0.5° × 0.5° spatial resolution) via bilinear interpolation. The sWBGT metric used here assumes no solar radiation and is intended to estimate humid heat exposure in the shade or indoors with no air conditioning (Li et al 2020). WBGT may underestimate the impacts of heat in high humidity environments (d'Ambrosio Alfano et al 2014), and sWBGT in the sun can be at least 2 °C–3 °C higher than shade values (Kjellstrom et al 2013), so our estimates of productivity losses from heat exposure are conservative. Estimating global impacts of humid heat on labor using ERFs with global gridded data represents one method for investigating, at a large scale, the potential impacts of warming on workers.

2.2.1. Calculating per capita labor losses

2.2.2. Calculating population-weighted labor losses

2.2.3. Calculating productivity losses

We compare labor productivity loss (also called 'labor loss' or 'physical work capacity' loss) estimates using two methods: the LCHCC method (Watts et al 2021, 2018, Romanello et al 2021) and Foster et al (2021) (laboratory ERF). Figure S1 shows the fraction of work capacity lost as a function of WBGT for each ERF. The LCHCC exposure response relationship is based on epidemiological data to derive the fraction of work lost for a given hourly value of WBGT (Watts et al 2018, 2021), whereas the Foster et al (2021) ERF is derived from measurements of reduced human physical work capacity in a controlled laboratory environment. Although both ERFs provide estimates of work capacity or productivity losses due to humid heat exposure, neither study is based on a globally representative sample of agriculture and construction workers (details below). Nonetheless, use of a single ERF (i.e. the LCHCC ERF), to estimate global labor losses is a common practice in several prominent publications (Watts et al 2018, 2021, Kjellstrom et al 2018, Orlov et al 2020).

The LCHCC method employs estimated exposure response relationships for reduced hourly work capacity (labor productivity) for heavy manual labor conducted at 400 watt (W) intensity using a cumulative normal (ERF) function. This ERF is based on two studies of heat exposure and productivity in outdoor workers: one of Indian rice harvesters (Sahu et al 2013), and the other of African miners (Wyndham 1969). The Kjellstrom et al (2018) ERF employed in 2018 LCHCC begins to show work loss at WBGT greater than ∼23 °C (figure S1). Previous work (Watts et al 2021, Kjellstrom et al 2018) has assumed labor loss cutoffs of 10% and 90% of the hour (i.e. WBGT causing labor productivity losses <6 min h⁻¹ were counted as 'zero productivity loss', and losses >54 min h⁻¹ were 'capped' at 54 min). However, we use the approach employed in the 2020 LCHCC (Watts et al 2021) and 2021 LCHCC (Romanello et al 2021) that assumes no work loss cutoffs at 10% and 90% of work capacity. Although humid heat exposure can impact workers conducting light (e.g. services) and medium (e.g. manufacturing) labor, here we only consider heavy labor impacts because heavy labor losses account for the largest fraction of labor loss from humid heat exposure (Watts et al 2018, 2021, Romanello et al 2021).

We compare estimates of labor losses from humid heat using the 2020 and 2021 LCHCC (Watts et al 2021, Romanello et al 2021) ERF to the Foster et al (2021) ERF, which is derived from measurements of reduced human physical work capacity in a range of controlled heat and humidity relative to a cool environment. Whereas the LCHCC ERF is not based on data above WBGT of ∼34 °C, the laboratory ERF does include measurements of work capacity up to WBGT of 40 °C. The Foster et al (2021) method yields estimates of reduced work capacity at much lower humid heat than the LCHCC method, with ∼10% reduced work capacity at WBGT of 18 °C (figure S1). This lower threshold for reduced work capacity has been reported in other studies, such as in Deryugina and Hsiang (2014), who found that county-level earnings in the United States began to decline at a temperature threshold of ∼15 °C. Limitations of the Foster et al WBGT-Physical Work Capacity (PWC) relationship include that the study used young, healthy, fit British participants who were not acclimatized to heat. However, as discussed in Foster et al (2021), acclimatization to heat is less critical in fit individuals. Further discussion of the specific methods and limitations can be found in Foster et al (2021).

Between 2001 and 2020, exposure to high humid heat is associated with approximately 677 billion hours of lost labor per year (±30 billion hours yr⁻¹; figure 1), with 72% of the global working-age population (∼4 billion people) living in locations with background climate conditions that would cause heavy labor losses of at least 100 h person⁻¹ yr⁻¹ (figure S2). Previous estimates (using the LCHCC ERF) indicate that labor loss due to humid heat exposure was approximately 248 billion hours yr⁻¹ (±24 billion hours yr⁻¹), with 40% of the global working-age population (∼2.2 billion people) living in locations with heavy labor losses over 100 h person⁻¹ yr⁻¹. Using the laboratory ERF, estimates of labor loss for workers in heavy labor sectors are approximately 2.7 times as high as estimates from the LCHCC ERF (figure 1(a) and (b); table S1).

Zoom In Zoom Out Reset image size

Figure 2(a)–(c) shows the spatial distribution of per capita labor losses (h person⁻¹ yr⁻¹) due to humid heat exposure in the recent past using the LCHCC and the laboratory ERFs, respectively. In both global maps, large portions of the tropics have experienced major labor losses (>100 h person⁻¹ yr⁻¹, or >8 days person⁻¹ yr⁻¹), which is consistent with the hot and humid climate in this region. However, the laboratory ERF-based estimates show maximum labor losses upwards of 600 h person⁻¹ yr⁻¹ (>50 days person⁻¹ yr⁻¹) compared to <250 h person⁻¹ yr⁻¹ (<21 days person⁻¹ yr⁻¹) reported in the LCHCC (figure 1(c), (d) and 2). Additionally, laboratory ERF-based estimates show that labor losses extend to the mid-latitudes, with many locations in this region (e.g. southern and eastern continental United States) experiencing labor losses >100 h person⁻¹ yr⁻¹ (>8 days person⁻¹ yr⁻¹). Using the LCHCC ERF, losses of this magnitude are only seen in the worst-impacted areas in the tropics and South and Southwest Asia (figure 1(c)).

Zoom In Zoom Out Reset image size

Estimates of humid heat impacts also show significant country-level changes when using the LCHCC as opposed to the laboratory ERF. India, China, and Indonesia saw the largest differences, with approximately 149 billion, 49 billion, and 25 billion hours yr⁻¹ difference, respectively (table S1). Over 40 countries saw a difference of at least 1 billion hours yr⁻¹. However, when examining differences as a proportion of LCHCC, differences may be several orders of magnitude higher, with multiple high-latitude countries showing labor losses using the laboratory ERF that experienced almost no labor losses using the LCHCC ERF (table S1). Nearly all countries showed at least 100% higher estimates using the laboratory method. Just four countries (United Arab Emirates, Pakistan, Qatar, and Bahrain) saw <100% changes in labor loss between the estimates, and no countries showed declines in labor loss. In the locations with smaller labor loss changes, humid heat already reaches unsafe thresholds for continuous labor regardless of the employed ERF.

Labor losses weighted by the working-age population engaged in outdoor labor (section 2) are highest in South, East, and Southeast Asia (figure 2(d)–(f)), where there are large numbers of working-age people (figure S3) working in agriculture (table S1). Specifically, labor losses are most pronounced in India, which accounts for almost half of the global total losses and experienced over four times the labor losses of the second most impacted country, China (figure 1(a) and (b)).

Some coastal and island nations in Southeast Asia (e.g. Cambodia) and the western Pacific (e.g. Tuvalu) saw up to seven to eight weeks person⁻¹ yr⁻¹ of labor loss (figure S4), and humid heat contributed to at least four weeks person⁻¹ yr⁻¹ of labor loss in 87 countries (table S1). Many low-latitude, coastal or island nations (e.g. Tuvalu) with the largest annual per capita labor losses according to the laboratory ERF experience minimally seasonal climates. In these tropical locations, the minimum sWBGT remains above 17 °C (figure S5), the sWBGT at which the laboratory ERF shows noticeable losses (figure S1). Therefore, the laboratory ERF indicates the mean climate is already warm enough to create decreases in labor productivity throughout much of the year across a large swath of the tropics. By contrast, the coolest months of the year are still below the minimum sWBGT (23 °C) that causes noticeable labor losses using the LCHCC ERF. According to the LCHCC ERF, the largest per capita labor losses are experienced in locations with seasonally hot (and potentially humid) climates (figures 1(c) and S4). In these locations (e.g. Qatar, Bahrain, Bangladesh), the LCHCC ERF indicates the warm season sWBGT would cause relatively large labor productivity losses, but the cool season does not experience sWBGT high enough to cause noticeable labor losses (figure S5).

Per capita working-age population labor loss translates to a job-equivalent loss (section 2) of 155 million jobs globally per year (98 million more than LCHCC), with India accounting for nearly half of this loss (equivalent to approximately 62 million jobs lost). These annual losses are comparable to temporary work loss during the global COVID-19 lockdowns, which are estimated to have caused workhour losses equivalent to approximately 130 million full-time jobs in the first quarter of the pandemic (ILO 2020a). The impacts from high humid heat we report here exceed or are comparable to those caused by other pressing environmental health challenges, such as air pollution, which in 2016 caused 1.2 billion lost working days (OECD 2016), or lack of safe drinking water and sanitation, which was projected to cause an estimated 22 billion lost work days in 2015 (Hutton and Haller 2004).

These labor losses translate to significant economic costs (section 2). Annually, the humid heat-related global labor loss results in lost productivity (figure 1(e) and (f)) of approximately 2.1 trillion in 2017 PPP$ (±98 billion 2017 PPP$) using the laboratory ERF, nearly three times as high as the impacts that result when using the LCHCC ERF (733 billion 2017 PPP$, ±72 billion 2017 PPP$). Impacts also vary significantly by country, with China and India again experiencing the largest losses, and Indonesia and the United States showing over 90 billion PPP$ losses per year. To put this in perspective, India experiences annual productivity losses from high humid heat equivalent to almost 7% of its 2017 GDP. Humid heat impacts on labor cause productivity losses equivalent to >10% of GDP in several countries in tropical Africa, such as Sudan and Sierra Leone (figure S6 and table S2).

We also compare recent labor losses (2001–2020 mean; figure S7(d)–(f)) to losses at the end of the 20th century (1981–2000 mean; figure S7(a)–(c)). Here we use 2017 population and worker statistics for both time periods to focus on the impacts of climate. Over this time period, global-mean, near-surface air temperatures have increased by ∼0.4 °C (Lenssen et al 2019), resulting in increases in per capita labor losses of up to 150 h person⁻¹ yr⁻¹ (12.5 days person⁻¹ yr⁻¹) in some low-latitude locations for both ERFs (figure 3(a) and (b)). However, median increases in labor losses across all latitudes are typically higher for the laboratory ERF, peaking near the equator and 10°–20° N (figure 3(c)). These changes in labor losses over the last several decades have also resulted in large increases in global and country-level population-weighted labor losses, per capita losses, and associated decreases in production that impact the economy (figure S8). Specifically, both ERFs show increases in population-weighted global labor productivity losses of at least 9% between the two time periods (LCHCC: increase of 39 billion lost hours, laboratory: increase of 58 billion lost hours). Both ERFs also show increases in productivity losses of at least 100 billion (2017 PPP$), with the laboratory ERF showing an increase in global losses of 192 billion (2017 PPP$).

Zoom In Zoom Out Reset image size