Inadequacy of typical physiological experimental protocols for investigating consequences of stochastic weather events emerging from global warming

Abstract

Increasingly variable, extreme, and nonpredictable weather events are predicted to accompany climate change, and such weather events will especially affect temperate, terrestrial environments. Yet, typical protocols in comparative physiology that examine environmental change typically employ simple step-wise changes in the experimental stressor of interest (e.g., temperature, water availability, oxygen, nutrition). Such protocols fall short of mimicking actual natural environments and may be inadequate for fully exploring the physiological effects of stochastic, extreme weather events. Indeed, numerous studies from the field of thermal biology, especially, indicate nonlinear and sometimes counterintuitive findings associated with variable and fluctuating (but rarely truly stochastic) protocols for temperature change. This Perspective article suggests that alternative experimental protocols should be employed that go beyond step-wise protocols and even beyond variable protocols employing circadian rhythms, for example, to those that actually embrace nonpredictable elements. Such protocols, though admittedly more difficult to implement, are more likely to reveal the capabilities (and, importantly, the limitations) of animals experiencing weather, as distinct from climate. While some possible protocols involving stochasticity are described as examples to stimulate additional thought on experimental design, the overall goal of this Perspective article is to encourage comparative physiologists to entertain incorporation of nonpredictable experimental conditions as they design future experimental protocols.

WEATHER, CLIMATE, AND COMPARATIVE PHYSIOLOGY

As the American humorist Mark Twain quipped, “Climate is what we expect. Weather is what we get.” Long before “climate change” became a household phrase, Mark Twain recognized the important difference between the relatively stable nature of “climate” (measurable environmental characteristics that are consistently observed year after year) and the typically less predictable nature of much shorter-term environmental changes that we call “weather” (events measured in hours, days, and weeks). Predictions of ever more extreme weather are associated with climate change (10, 13). Long-standing theoretical predictions, some accompanied by experimental data, indicate that extreme weather events decrease biodiversity and shift geographic species distribution through population eradication, if not actual species extinction (12, 15–18, 20, 23, 27, 28, 30, 31, 33–38, 40). As such, weather (not climate) may be the most challenging characteristic of climate change facing both developing and mature terrestrial animals (3, 21, 36), or indeed animals in any environment that is not buffered from weather-induced changes (e.g., large lakes, oceans).

Many of the studies associated with the effects of weather “correlate” variable environments with species survival, diversity, and reproduction. Comparative physiologists are in a position to actually create experimental environmental stressors and so move beyond correlation to causation by contributing vital data and even testing hypotheses emerging from ecological and population biology. In particular, we can play more prominent roles in assessing how nonpredictable weather events (especially temperature change) can lead to decreased fitness, and even ascribe mechanisms (behavioral, physiological, molecular, biochemical) for fitness reduction. Moreover, stochastic environmental variation could lead to as-yet-unexplored longer-term individual and population effects, potentially acting through epigenetic inheritance (4).

TRADITIONAL PROTOCOLS FOR INVESTIGATING ENVIRONMENTAL EFFECTS

Loosely paraphrasing Mark Twain, “Responses to climate change are what comparative physiologists study. Weather is what animals get.” Indeed, most of us control our experimental temperatures to fractions of a degree Celsius, gas partial pressures within a millimeter of mercury, and salinities within a part per thousand. Moreover, when we do induce ambient changes, we are often slaves to powers of 10 or neat and tidy fractions thereof. Thus, there are hundreds of comparative physiological articles that have induced a step-wise, ramp-up, or (rarely) a sine wave form of environmental temperature change (as an example) of 3.0°, 5.0°, or 10.0°C in the laboratory (Fig. 1, A–F), followed by determination the animal’s physiological responses to the new temperature steady state. Mea culpa—e.g., (5, 29, 39). Yet, steady-state, mean temperatures typical of experiments rarely reflect natural conditions, where temperature variations occur on multiple complex timescales ranging from a few hours to years (8).

Fig. 1.

Experimental protocols typically used for inducing temperature or other changes in comparative physiological experiments exploring animals’ responses to climate change. A–D: the most frequently used step transition protocols. E and F: cyclic transitions (spike and especially sine wave protocols), which are much less frequently employed. G and H: semistochastic transitions that are more realistic, but less easy to induce. Modified from Colinet et al. (6) with permission. Further details of these protocols are provided in Ref. 6.

The carefully controlled, traditional experimental protocols described above have certainly been useful in revealing mechanisms. However, they may be inadequate for investigating the full repertoire of animals’ natural physiological regulatory responses associated with nonpredictable, extreme weather events attending climate change. How animals subjected to such stochasticity actually cope (or equally importantly, fail to cope) will be critical to research into future climate change. Yet, as numerous authors have observed, the typically carefully controlled laboratory conditions rarely represent those occurring in the field (e.g., 8, 9, 24). Of course, matching laboratory to field conditions is not a new concept by any means. Consequently, limitations of steady-state protocols in physiological studies have been pointed out many times, often by researchers in the field of thermal biology, which is increasingly emphasizing the importance of using non-steady-state experimental protocols to reveal the more realistic responses of animals to non-steady-state changes (e.g., 1, 2, 6–9, 19, 26).

WHY STOCHASTIC ELEMENTS IN EXPERIMENTAL PROTOCOLS?

Why are experimental protocols employing nonpredictable environmental conditions of importance to comparative physiologists? In fact, nonlinear and sometimes counterintuitive effects can emerge from non-steady-state protocols. For example, fluctuating temperatures can have generally positive effects at low temperatures, but often have negative effects at high temperatures (for an introduction to that literature, see Ref. 19). Some animals have been shown to develop more rapidly or show differences in physiological performance when exposed to stochastically varying environments compared with those in steady-state or cycling in predictable fashions (6, 11, 32). Yet, a close examination of that literature reveals that the terms “variable” and “fluctuating” are often used to describe step-wise changes of differing magnitudes, or to describe predictable, rhythmic cyclic changes, examples of which are schematically indicated in Fig. 1, E–G. Even the protocols shown in Fig. 1, G and H, while quite different from the other protocols and including stochastic elements, nonetheless employ an element of predictable cycling and so are only “quasi-stochastic”. Of course, even nonpredictable extreme environmental changes often show some predictable changes associated with diurnal or seasonal cycles or El Niño/La Niña events, as examples. Thus, quasi-stochastic protocols imposing nonpredictability upon basic cyclicity will certainly be informative. Yet, our knowledge of the effects of truly nonpredictable environmental changes on animal physiology is very limited.

Unfortunately, the message of the potential importance of employing variable (however that be defined) rather than steady-state conditions in our protocols appears not to be heavily penetrating into comparative physiological laboratories investigating implications of long-term climate change. Only infrequently are non-steady-state protocols employed, and these only rarely involve semistochastic or fully stochastic protocols. Indeed, such protocols may be likely to be questioned by contemporary referees of journal articles (just as some reviewers of this Perspective article questioned the approach).

CHALLENGES OF EXPERIMENTAL STOCHASTICITY

To be fair, there are reasonable concerns surrounding protocols with nonpredictable changes in variables. Let me raise some of these concerns and comment upon each in turn.

Defining a “control condition” comprises a significant concern in protocols employing nonpredictable experimental conditions. Establishing a constant acclimation period for a control is straightforward in constant condition protocols (Fig. 2A), or even those with an experimental phase with nonpredictable elements (Fig. 2B). Moving beyond these approaches, a “stochastic control” (is that very phrase oxymoronic?) could also be employed, which actually might be the most realistic for mimicking weather events. Thus, limited nonpredictable variation around a control set point could be combined with a subsequent elevated (or decreased) set point designated as the experimental level, around which the same degree of nonpredictable variation continues (Fig. 2C). A recovery period mirroring the initial control could then follow. In this scenario, stochasticity confined within upper and lower bounds equally applies to control, experimental, and recovery phases. Another example of a potential stochastic protocol would not artificially limit the nonpredictable variation of the environmental variable as in Fig. 2C, but would actually allow changes that incorporate extremes in the form of the highest (and/or lowest, not illustrated) previously measured environmental value (Fig. 2D).

Fig. 2.

Suggested experimental protocols that could be used individually or, better still, in combination, for studying physiological responses to changes in environmental variables associated with weather and climate. A: an initial control phase at a constant level is followed by an experimental phase of elevated constant level. B: an initial control phase at a constant level is followed by an experimental phase where levels fluctuate stochastically around the same constant level shown in A, with upper and lower values established to limit possible level fluctuations. C: an expansion of the protocol in B, in which the control phase, itself, is characterized by stochastic fluctuation similarly bound by upper and lower limits. Note that these protocols could be reversed in direction to examine low rather than the illustrated high values. D: modification of the protocol from C, allowing greater excursion limits coinciding with the highest known environmental values.

Standardization across emerging studies is one of the more vexing concerns of stochastic protocols, for by its very nature, a protocol employing unpredictable environmental variation should be … well … unpredictable. Yet, there must be some form of standardization, such that other researchers can attempt to replicate the results. The answer to standardization of experiments involving nonpredictable changes could involve ensuring that all changes in manipulated variables within the control, experimental, and recovery phases of the protocols are truly random. Thus, nonpredictable changes are induced in the rate, magnitude, and duration of the increase in the environmental variable, the rate of subsequent variable decrease to control levels and perhaps even the time interval between multiple tests. The “standardization,” thus, lies in random, bias-free temperature increases, which another researcher should be able to replicate by employing similar randomization.

Statistical analysis comprises another significant concern involving data emerging from stochastic protocols. Concurrently employing multiple combinations of protocols (e.g., Fig. 2, A–D), each with multiple replicates, would allow differentiation of the specific effects of nonpredictable changes. Within any given protocol, time series averaging by any one of a variety of different models could be employed to assess physiological changes within the nonpredictable phases (25). Another possible approach may involve decomposing stochastic environmental variable time series using Fourier transformation, which has recently been used to examine the biological effects of climate variability across multiple timescale effects (8). Yet another approach, which has been employed to test effects of nonpredictably varying environmental conditions, is an “ecomechanical” approach for analyzing data over time (see Ref. 9). Clearly, standardized statistical approaches will need to emerge in parallel with increased use of protocols, including stochasticity.

Induction of nonpredictable variability is another concern. Inducing true stochasticity is obviously more complex than just a one-time adjustment to a controller to create a new, constant steady-state condition. However, many regulation devices (especially for temperature) now accept computer-programmed input, and a random number generator could be employed to input truly random changes within fixed ranges (Fig. 2, B–D). Alternatively, experimental protocols could be built around so-called “stochastic weather generators” that have emerged from ecology, agriculture, and climate science to develop and test models for experimental environments mimicking various weather scenarios (14, 22). For the electronics-capable investigator, a small investment in electronic components can enable the programing of nonvariable changes in the variable of interest.

Comparability of physiological studies based on stochastic environmental changes with existing traditional studies is a final concern (additional concerns could be argued, of course). This is where inclusion of traditional step-wise protocols alongside stochastic ones (i.e., deploying all of the multiple protocols depicted in Fig. 2, each in a different experimental population) could be particularly illuminating. Indeed, of the relatively few comparative physiological studies employing quasi-stochastic elements, most also carry out more conventional step-wise protocols in parallel, creating what might be considered a “meta-control” (6, 11, 32). That data emerging from a single laboratory using experimental protocols with nonpredictable elements may differ from data from more conventional step-wise or ramp protocols used in that laboratory is highly informative per se, and it should stimulate additional thought and experimentation.

Perspectives and Significance

The comparative physiological community can contribute to the contemporary study of climate change by exploring the physiological effects of extreme weather events associated with global climate change. In doing so, we should be encouraged to increase the use of stochastic protocols that more realistically mimic short-term weather events to supplement protocols based on statistical calculations of long-term climate change.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

W.W.B. prepared figures; W.W.B. drafted manuscript; W.W.B. edited and revised manuscript; W.W.B. approved final version of manuscript.