Measurement of microclimates in a warming world: problems and solutions Free

Global warming already has had profound biological effects (see IPCC, 2022), especially in regions like the poles where the rate of warming is well above average (Rantanen et al., 2022). We can expect increases in the frequency and intensity of heatwaves (Domeisen et al., 2023; Perkins-Kirkpatrick and Lewis, 2020; Thompson et al., 2023), with potentially catastrophic thermal consequences. But exceeding the arbitrary threshold of 1.5°C in global mean surface temperature (Masson-Delmotte et al., 2022), in itself, will have little impact on the heat balance (or energy balance; see Glossary) of a lizard on a rock or a cow in a pasture.

The heat balance of terrestrial animals facing global warming depends on their rates of convective heat exchange with the air, conductive heat exchange with surfaces (typically solid or static liquid), radiant heat exchange (typically, but not always, also with surfaces but without physical contact), metabolic heat production and evaporative cooling. When scientists discuss the biological responses of terrestrial animals to global warming, they usually consider two metrics of the thermal environment: air (dry-bulb) temperature (see Glossary) and relative humidity (RH; see Glossary), either individually or combined in an index (e.g. de Castro Junior et al., 2023; Rahimi et al., 2021). Other than the fact that these metrics are reported routinely by meteorological services, the rationale for choosing them is not clear. Air temperature is rarely the main factor that defines heat load on terrestrial animals in warm environments: ‘the impacts of climate change on species cannot be assumed to be proportional only to changing air temperature’ (Gardner et al., 2024). Furthermore, there is no physiological process in vertebrates that is governed by RH, including evaporative cooling.

Meteorological services measure variables in the free-stream environment; that is, the environment undisturbed by surface structures. For example, wind speed and direction typically are measured on a 3 m or 10 m mast (e.g. National Centers for Environmental Information, USA, www.ncei.noaa.gov/access/monitoring/wind/, accessed 13 December 2023). By contrast, terrestrial animals interact with surfaces and air immediately adjacent to their bodies, with which they are in direct contact, and, at a distance, with the radiation sources and sinks that prevail at their location, some of which, such as night sky, can be remote. Their thermoregulatory responses are not responses to the regional climate or weather, but to their specific microclimate (Coleman and Downs, 2010; Kearney et al., 2021; Pincebourde et al., 2016; Varner and Dearing, 2014). Thus, analyses of heat balance and predictions of an animal's thermal responses require knowledge of those properties of the microclimate that affect the transfer of heat and water vapour (see Box 1).

In this Commentary, we discuss those properties of microclimates that affect the heat balance and thermoregulatory responses of terrestrial animals. The Commentary concentrates on hot microclimates, which will be most problematic in a warming world. We pay specific attention to the practices and problems related to the measurement of relevant microclimate properties. We conclude by proposing a simple, low-cost approach to microclimate measurements.

The ‘sensible’ (Bakken, 1976) or dry (IUPS, 2001) heat transfer between an animal and its microclimate is the sum of radiant, convective and conductive heat flow (see Glossary). In each case, the net transfer of heat depends on the temperature difference between the animal's body surface and the environment. Depending on the direction of the difference, the microclimate directs heat towards, or accepts heat from, an animal's surface. In the case of an animal with fur, feathers or scales, that surface is not its skin, and radiation, convection and conduction may not occur at the same surface (Mitchell et al., 2018). Although conduction is important for a bear lying in a den, for example, active animals, with some exceptions (such as snakes), seldom have appreciable contact with solid surfaces, so we shall not pursue conduction. For convective heat transfer, the relevant temperature difference is that between microclimate air and the surface of the animal exposed to the air, modulated by wind speed (Mitchell, 1974). Historical accounts of animal convective heat transfer usually report microclimate air temperatures lower than the animal's surface temperature, but with global warming there are likely to be more hours of the day, in more places, when air temperature exceeds surface temperature. Convection then will load heat on the animal. In itself, an air temperature that exceeds body surface temperature may drive some terrestrial animals, especially small ones, beyond their limits of thermal tolerance. However, for exposed terrestrial animals, a greater threat to their thermal welfare will come from radiant heat load, and that threat may put animals at risk even if air temperature is well below body surface temperature (Mitchell et al., 2018).

We suggest that mean radiant temperature (MRT, see Glossary) is more important than air temperature in defining the thermal threat to an exposed animal. MRT is the mean temperature of the surrounding sources and sinks of radiation, assuming that they are perfect radiators. It is the microclimate variable that best defines the contribution of radiation to an animal's dry heat transfer. The importance of MRT over air temperature can be illustrated with an example. On 17 August 2020, a reliable air temperature of 54.4°C was recorded at Furnace Creek, CA, USA. A naked person standing outside on that day probably would have been sweating heavily, and so would have had a skin temperature of about 35–36°C (Notley et al., 2023). Assuming a wind speed of 1 m s⁻¹ would give a whole-body convective heat transfer of 8.3 W m⁻² °C⁻¹ (Mitchell and Whillier, 1971) and so a heat gain from the air of about 160 W m⁻² of skin. About half of the person's 1.8 m² of skin would have been exposed to direct solar radiation, some of which would be reflected from the skin, depending on skin tone (46% for pale skin and 25% for a dark skin). Thus, the direct solar heat load absorbed by the body would range between 486 W (1000 W m⁻²×0.9 m⁻²×0.54 for pale skin) and 675 W (1000 W m⁻²×0.9 m⁻²×0.75 for dark skin), more than double the convective heat load in either case. To dissipate this additional heat gained from solar radiation, the person would need to drink and evaporate additional water: three more standard glasses per hour for a pale-skinned person and 4.5 more glasses if the person were dark skinned. Those are underestimates; the total radiant heat load on the body would have been more than that imposed by direct solar radiation (see below). As Virginia Finch (1972) showed, radiation coming from the ground nearly doubles the radiant heat load on an antelope standing on the ground of high albedo (see Glossary), and the same applies to dairy cows (Da Silva et al., 2010). So, although the 54.4°C air temperature in the above example was extreme, it was not the main thermal threat to our naked person. Under global warming, MRT will remain far more important than air temperature in defining the heat load on animals. However, meteorological services do not measure MRT (and although they measure solar irradiance they usually do not report it). Nor is MRT measured by most biologists who explore the biological responses of animals to global warming.

The typical radiant environment for an animal outdoors extends far beyond direct solar radiation and radiation from the ground (Fig. 1). Therefore, the measurement of MRT is not straightforward (Kántor and Unger, 2011). The gold-standard method, at least for the built environment, is to position combinations of pyranometers and pyrgeometers (see Glossary) at the site of the occupant of the environment, pointing up, down and in the four horizontal directions. The measured radiant flows then are integrated with the ‘view’ that the occupant has of the radiant sources or sinks in every direction (e.g. Acero et al., 2023; Vanos et al., 2021). This method is impracticable for biologists working in the field, but a simpler alternative is available (Johansson et al., 2014), developed originally to measure MRT for humans indoors (Bedford and Warner, 1934; Vernon, 1932). The method involves placing a hollow sphere painted matte black (a globe or black-bulb thermometer) at the site of the occupant (or the site of an animal in the field), allowing it to reach thermal equilibrium (see Glossary), and measuring its temperature. At thermal equilibrium, convective and radiant heat transfer at the sphere balance each other out. Thus, if the rate of convective heat transfer can be calculated (from sphere size and temperature, wind speed, and air temperature), then the rate of radiant heat transfer at the sphere can be derived, and MRT calculated (e.g. Kuehn et al., 1970). The original sphere used in this context was ∼150 mm in diameter, a copper float from the cistern of a domestic toilet (Aitken, 1888). However, for many applications, especially in the field, that sphere is too big or responds too slowly, so other spheres of between 30 mm (Fuller et al., 1999; Hetem et al., 2007) and 100 mm (Vega et al., 2020) diameter have been used, with ‘miniglobes’ of 38 mm being popular (legacy table tennis balls; De Dear, 1988; Humphreys, 1977; Thorsson et al., 2007).

For nearly a century, there was confidence in the derivation of MRT from globe temperature (Bond and Kelly, 1955; Graves, 1974; Kuehn et al., 1970; Oliveira et al., 2019). That confidence has waned, mainly because of doubts about the convective heat transfer coefficient that is assigned to the globe thermometer, and especially to miniglobes (d'Ambrosio Alfano et al., 2021a,b; Teitelbaum et al., 2022). The original formula for the convective heat transfer coefficient does not apply when the globe is in still air and therefore under free or mixed convection (see Glossary entry for ‘convective heat flow’; Acero et al., 2023; d'Ambrosio Alfano et al., 2021a; Teitelbaum et al., 2020). The formula breaks down when the air flow is turbulent (Acero et al., 2023), in rain (Acero et al., 2023) or if wind speed at the globe is estimated rather than measured (Guo et al., 2018). Furthermore, sometimes when the globe is used outdoors, asymmetrical surface heating can result in errors in calculated MRT of up to 20°C (Guo et al., 2020; Kántor and Unger, 2011; Vanos et al., 2021). However, these sources of error do not render the globe obsolete. The calculation of MRT is essential to evaluate the heat balance of terrestrial animals, and the globe is the only practical way of measuring MRT in the field. Making users aware of the potential errors should help them avoid or account for these errors. Below, we suggest an alternative to the calculation of MRT when heat balance measurement is not required: using globe temperature as a metric in its own right.

Given the danger of overheating under a radiant heat load, it is not surprising that many small terrestrial animals that are diurnally active retreat periodically to refuges such as burrows, tree cavities and deep rock crevices where they can escape solar radiation (e.g. Chappell and Bartholomew, 1981; Fick et al., 2009). Even the largest of terrestrial animals, elephants, which are clearly excluded from burrows, seek shade under trees (Mole et al., 2016). The temperature of air under tree canopies tends to be a few degrees Celsius lower than that of the surrounding air, but the ground is much cooler than adjacent ground exposed to solar radiation (Lin and Lin, 2010). Shade-seeking under a tree, however, will not allow an animal to escape radiation as well as would entering a burrow, as solar radiation reflecting off the ground can contribute substantially to the radiant heat load on animals (Da Silva et al., 2010; Kelly et al., 1950; Finch, 1972), and there are several other routes by which radiation can penetrate into shade (Fig. 1).

Surprisingly few comparisons have been made of the radiant microclimate in shade and simultaneously in nearby open habitats. Table 1 gives some examples. It shows that making an incomplete measurement of the radiant heat load may lead to a false conclusion regarding the protection provided by shade. For example, pyranometers may lead to the conclusion that trees exclude almost all of the radiation. But pyranometers usually are set up to record only downcoming radiation and not radiation from the other directions illustrated in Fig. 1, much of which penetrates into and below the tree canopy (Kelly et al., 1950). When globe thermometers are used to measure radiation, they detect the complete radiant heat load in shaded microclimates, and reveal that the protection offered by trees is much less than is indicated by pyranometers (Table 1). Typically, animals under trees, or under artificial shelters (e.g. roofing or solar panels), receive 60% or more of the radiation that they would in the sun. During daylight hours, MRT in the shade of an overhead shelter can far exceed air temperature under that shelter (Fig. 2A). Thus, although MRT may be close to air temperature in deep shade, one never should assume that MRT and air temperature are equivalent. Radiant heat has to be measured even in the shade.

Although deep shade, such as dense forest canopy or a burrow, clearly must benefit the thermoregulation of animals that can reach it, Table 1 indicates that animals in sparse shade, such as those in the Box 1 figure, may not enjoy the full benefit. Indeed, in the tropics, the mean (Gaughan et al., 2010; Gebremedhin et al., 2011) and maximum (Faria et al., 2023; Tucker et al., 2008) daily deep-body temperature does not differ significantly between cattle in the shade and those in direct solar radiation. So why do large animals, including humans, seek shade so routinely? Perhaps shade helps the animal to reduce the evaporative water loss required to maintain body temperature (e.g. Maia et al., 2023; Wolf and Walsberg, 1996). It may also help the animal to reduce body surface temperature, to escape high ground temperature or to avoid the burning short-wave radiation from the sun, especially on exposed skin.

Air movement has a large impact on both convective and evaporative heat exchange. Just as MRT has to be measured in an animal's microclimate, so too does wind speed, as the example in Fig. 2B illustrates. The figure shows wind speed measured at different times of the day and at different locations on a Namib Desert sand dune. Notably, according to the authors, ‘wind speeds indicated by an anemometer in the free air stream bear no resemblance to the wind regimen experienced by lizards and other organisms living close to the slipface surface’ (Seely et al., 1990). This wind profile is not peculiar to sand dunes of the Namib Desert. In any environment, the profile of wind speed must decrease progressively from the free stream to zero at the ground surface. The profile shape depends on the physical structure of the environment (Garratt and Hicks, 1973), and the wind speed that affects an animal depends on what part (or parts) of the profile the animal occupies (Meyer et al., 2023). As Fig. 2B shows, it may not be possible to predict wind speed near the ground from wind speeds measured higher, for example, by anemometers on weather stations.

Another example from the Namib Desert illustrates how narrow the microclimate might be in which wind speed would have to be measured to obtain an accurate picture of conditions experienced by an animal. Ocymyrmex barbiger is a 4 mg Namib ant, active in full sunlight in the middle of the day (Marsh, 1985), even when sand temperature reaches 67°C, which is well above the ant's critical thermal maximum (see Glossary). The ants avoid thermal death by engaging in short bouts of activity (5–9 s), between which they climb onto any pedestal available, typically reaching 10–20 mm above the sand surface. There, they use a few limbs to grip the refuge object and ‘rapidly flail their other limbs about in the air’ (Marsh, 1985). Climbing to 20 mm puts the ant in an air layer that could be 20°C below the temperature of air at the ground surface. Flailing increases convection from their narrow limbs (Mitchell, 1974; Mitchell et al., 2018). Historically, the only anemometers small enough to make valid measurements of wind speed in a 20 mm thick microclimate have been hot-wire anemometers, which use a fine heated wire to determine wind speed. However, hot-wire anemometers are not presently amenable to continuous operation with output collected by data loggers. Thus, the future of wind speed measurement for studies in thermal biology will lie in ultrasonic anemometers, which also provide measurements of wind direction, air temperature and humidity, with no moving parts (Nakayoshi et al., 2015). Palm-sized ultrasonic anemometers that connect to data loggers already are available commercially (e.g. TriSonica™ 91×91×52 mm; ULSA™ 75×75×54 mm).

For most animals, however, the most important reason to measure wind speed in the microclimate will not be to determine convection, because, as we have already established, convection will probably play a minor role in comparison to radiation as the major thermal challenge. Wind speed affects evaporation in the same way that it affects convection (Mitchell et al., 2018). Effective evaporation is driven by the difference in water vapour pressure (WVP; see Glossary) between an animal's evaporating surfaces and its immediate environment (see below), but the rate of evaporation is amplified by increased wind speed. Effective evaporative cooling will be increasingly important if animals are to avoid hyperthermia under an increasing frequency and intensity of heatwaves.

It is commonly known that evaporative cooling is hindered in environments with high RH and high air temperature, but it is a misconception that air saturated with water vapour – that is, with 100% RH – prevents evaporation from the skin or upper respiratory tract (e.g. Luber and McGeehin, 2008; Mora et al., 2017). Evaporation can occur into saturated air. It can do so because water vapour in air, like any other uncharged motile substance, moves down its concentration gradient, which, according to the ideal gas equation, is proportional to the WVP gradient. If the water vapour concentration is expressed as absolute humidity (g m⁻³; see Glossary), the relationship between WVP and concentration depends on air temperature, but if it is expressed as specific humidity (g kg⁻¹; see Glossary), it does not.

So, during evaporative cooling, water vapour will move from the surface of the skin or upper respiratory tract into the microclimate, provided that the WVP in the microclimate is lower, even if its RH is 100% (Mitchell et al., 2018). If microclimate RH is 100%, for surface WVP to be higher than that of the microclimate, surface temperature has to be above microclimate air temperature. If sweat evaporates into the saturated air, the excess water vapour in the air will condense out; what happens to the water vapour after it leaves an animal has no thermal consequences for the animal.

RH, in its own right (i.e. without the context of prevailing air temperature) provides no information about an animal's ability to cool by evaporation. As an illustration of this concept, we can compare a hot, sweaty human in two different environments: Heathrow Airport and a hot dry desert (Fig. 3A). On New Year's Day 2023, the RH at Heathrow Airport reached 94%. Would that high RH have impeded evaporative cooling? The maximum air temperature at Heathrow on that day was 11°C. So, even at 94% RH, WVP at Heathrow would have been far below that on the body surface (Fig. 3A). Indeed, the WVP at Heathrow would have been similar to that of the hot dry desert at 40°C and 17% RH (Fig. 3A). So the WVP gradient (the driving force for evaporative cooling) from saturated skin at 36°C would have been the same at Heathrow as in the desert (ca. 6.0−1.2=4.8 kPa), despite the RH being 94% and 17%, respectively. In neither location would evaporative cooling be impeded. If there was a burrow in the desert with a temperature of 25°C and RH of 30% (WVP ca. 1.0 kPa), some water vapour would have flowed from the atmosphere into the burrow mouth, down the WVP gradient but against the RH gradient.

There is another way that reporting RH can be misleading. There are many diagrams in the literature similar to that shown in Fig. 3B, here depicting average air temperature and RH over 12 days in Brazil. The commentary on such diagrams may conclude that the deleterious thermoregulatory effects of the morning increase in air temperature are offset by the coincident decrease in RH, allegedly aiding evaporative cooling. But WVP hardly changes over the course of the day; RH decreases because air temperature increases, not because WVP changes (Fig. 3C). A decrease in RH, not accompanied by a decrease in WVP, does not aid evaporative cooling.

With global warming, at sites where sufficient liquid water is available (Simpson et al., 2024), the quantity of water in the Earth's atmosphere will increase, while RH will stay much the same (Trenberth, 2011; Trenberth et al., 2003). Hotter air will hold more water vapour: approximately 7% more water per degree Celsius rise in air temperature (Trenberth et al., 2003). A focus on RH may well miss the increase in atmospheric water vapour with a warming climate, leading to a failure to consider its consequences (Matthews et al., 2022). As we have emphasised, in the context of evaporative cooling, the relevant metric is WVP (or absolute or specific humidity), but not RH (Jenkins et al., 2023; Mitchell et al., 2018). As well as being important in studies on animal responses in a warming world, this notion is also relevant to work on human health, but is sometimes ignored. Of 260 publications between 2013 and 2016 related to human heat stress, 65% reported RH and only 5% reported absolute humidity (Davis et al., 2016). The manufacturers of instruments designed to measure water vapour should play a role in emphasising the importance of WVP; their instruments usually have RH as their primary output, but should instead give the calculated WVP.

Given the complexity of microclimates, is it possible to implement a measurement of the microclimate to which one can relate the biological responses of terrestrial animals in the field? Is there a method that measures variables that influence dry heat exchange and evaporative cooling in a microclimate? Is there a method that is straightforward, size-appropriate, widely applicable and affordable? We believe that these requirements can be met best by measuring globe temperature (the equilibrium temperature of which depends on MRT, air temperature and wind speed), calculating WVP (or absolute or specific humidity) and measuring wind speed, all within the relevant microclimate, or as close to it as possible.

The derivation of MRT from globe temperature has some issues (see above), which require attention by those calculating MRT, but the calculation of MRT will remain essential in studies aiming to analyse the full heat balance of an animal in the field. Some of the issues can be avoided by using only the original 150 mm diameter globe when MRT has to be calculated, because there is less uncertainty about how it is affected by air temperature, so the derived MRT is more accurate than that derived from smaller globes. However, in the many studies of animal responses to thermal challenge that do not require a detailed breakdown of heat balance, the errors associated with the calculation of MRT could be avoided by employing globe temperature itself – rather than MRT – as a thermal metric, a role it fulfils in the widely used wet-bulb globe temperature index (see Glossary) of human heat stress (Yaglou and Minard, 1957). More recently, globe temperature in its own right has been used occasionally, and successfully, as a metric of the microclimate in studies of the responses of free-living animals (e.g. Carroll et al., 2016; Cunningham et al., 2015; Fick et al., 2009; McFarland et al., 2020; Panagakis, 2011; Seely et al., 1990). We envisage globe temperature as a primary metric of the microclimate, not as a substitute for operative temperature (see Glossary; Bakken, 1976; Walsberg and Weathers, 1986). Operative temperature (which can be measured with taxidermic models or phantoms such as copper casts and agar models) certainly will be better than globe temperature in determining the environmental thermal stress on an animal of a specific size and shape, in a specific posture, with a specific orientation to the sun and wind, but that operative temperature will not apply to other animals in the same microclimate.

We propose that in experimental studies of the responses of animals to global warming, globe temperature should be measured routinely and biological responses should be correlated with globe temperature, not with air temperature. We propose further that globes of just two diameters should be used. For humans (e.g. Acero et al., 2023; Kuehn et al., 1970) and other large animals (e.g. Santos et al., 2017; Valtorta et al., 1997), the original 150 mm diameter globe has an established track record. For small animals and small microclimates, we propose a 40 mm diameter miniglobe; this is the diameter of current readily available and inexpensive table-tennis balls, which can be painted matte black and fitted with an internal thermometer (De Dear, 1988). These acrylic balls have low thermal inertia and so have a fast response time (Nikolopoulou et al., 1999). If wind speed is known, the equivalent temperature for the 150 mm globe can be calculated from miniglobe temperature (Hetem et al., 2007).

If animals of different sizes are under study, every weather station used in the field should have both globes installed (see Fig. 1). Miniglobes also can be attached to some study animals; for example, affixed to a stalk on a collar, so that globe temperature can be measured in the microclimate selected by the animal (Fuller et al., 1999; Hetem et al., 2007; Nakayoshi et al., 2015). The temperature of attached sensors of radiation inevitably will be influenced by the surface temperature of the animal itself, but so will the animal's microclimate; the attributes of an occupied and unoccupied microclimate, such as that of a burrow, are not the same. Where an attached miniglobe is not feasible, an array of miniglobes with a random distribution across the habitat would allow animal preferences to be assessed. Furthermore, measurements should not be confined to daylight hours, because radiation to the cold night sky, which the globe thermometer will detect, can provide substantial heat loss (Adelard et al., 1998; Swinbank, 1963).

Small data loggers (e.g. iButtons™) are widely used as measures of microclimate in studies of thermal biology (e.g. Levy et al., 2016; Sepulveda et al., 2014; Tillman et al., 2021; Van Jaarsveld et al., 2021; Vitt and Sartorius, 1999). When attached to the animal, these loggers allow detection of whether animals are in refuges (e.g. Murray and Smith, 2012). However, they will not measure air temperature reliably (their usual intended role) whenever they are not fully protected from radiation (Maclean et al., 2021). Furthermore, with their irregular shapes and age-dependent surface properties, they also will not properly integrate radiant heat load, air temperature and wind speed, as standard globe thermometers can do. The solution is to put the logger inside a table-tennis ball painted matte black.

As we have discussed above, when considering an animal's thermal environment, WVP also will be required. WVP usually will be calculated from RH and air temperature, and air temperature has to be measured with a thermometer protected from thermal radiation. Such protection is not achieved well by most naturally ventilated radiation shields (Nakamura and Mahrt, 2005), whereas fine-wire (<1 mm diameter) thermocouples painted white do surprisingly well in the measurement of air temperature even when unprotected (Christian and Tracy, 1985; Maclean et al., 2021). Furthermore, thermocouples are cheap and convenient for field work. The finer the wire, the less it is contaminated by radiation. Although wind speed often can be measured by using small cup anemometers (provided that wind speed is not too low), we expect that they will be replaced entirely by new technologies, perhaps miniaturised ultrasonic anemometers, in the suite of instruments for measurement of microclimate.

If we are to accurately assess the performance and predict the persistence or demise of terrestrial animals in a warming world, we must understand the biological responses of animals to their immediate microclimate, within which they operate and exchange thermal energy. To realise this goal will require the measurement of variables that allow us to properly understand the exchange of heat and water vapour between animals and their environment. This Commentary, we hope, will help to achieve those aims.