Our meta-analysis showed that water loss not only rarely remains constant throughout time but also that temporal profile strongly varies across species often deviating from the summed/average profile. We could distinguish five different temporal patterns in our dataset. Majority exhibited initial acclimation and alternating drops and rises in water loss across 12 hours of the experiment. Few cases exhibited a steady decline or increase or a middle peak of EWLi values. We emphasize that all temporal patterns are likely to cause statistical problems in comparative studies where summed or average values of water loss rates are used. Moreover, the length of experiment in the tested dataset show that 12 hours is sufficient to detect temporal patterns but shorter experiments might hamper recovering fine variation trends. A high occurrence of such complex temporal patterns also supports the possibility that they hold biological relevance that needs future research endeavours.
The high share of temporal patterns found in the diverse dataset clearly suggests time must be correctly considered in water loss studies. However, traditionally, the average or the sum of hourly water loss are used for comparative purposes. We showed that this might cause statistical problems (Fig. 2) and consequently results’ interpretation. Comparing the averages across different experimental intervals (2–12 hours) showed averages calculated at different time-points will be significantly affected by temporal patterns. Specifically, the “initial acclimatisation” with the high initial value gives a non-normal distribution, which makes the average value meaningless and artificially inflates the total water loss value. The “steady decline” pattern could pull down the sum and the average value of the water loss and oppositely, the “steady increase” pattern would pull them up. For the “fluctuation” pattern, the sum is potentially going up or down, while for the “middle peak” it may go up. The normality of the data set could be compromised in all datasets with temporal patterns. With non-normal distribution it is more difficult to interpret the average value. Therefore, we suggest that in the future average and sum value of EWLi should systematically be avoided in comparative studies, unless no temporal pattern is statistically proven and normality of data is achieved.
Moreover, when comparing temporal patterns, it is of high importance that the measurement durations are equal between datasets. It is for example recommended that if initial acclimatisation period is omitted, this should be done for all datasets in the comparison, even if not showing initial acclimatisation. Furthermore, with this analysis we show that there is a high chance that temporal variation will be missed (83% chance detected in our study data set) if the duration of the experiment is too short. In the light of refinement principle (3R principles; Russel and Burch 1959) it would be beneficial to shorten the duration of the experiment, however, this cannot be recommend based on the present evidence. We have shown that even if initial stabilisation in the data is observed after the first 6-hours of the experiment, temporal patterns may emerge later in the second part of the 12-hour experiment. Some studies of water loss follow experimental procedure where the weight is recorded only at the beginning and the end of the experiment (e.g. after 12 hours). We highly recommend to avoid such procedure because information obtained will be very limited and inferences about temporal variation will not be possible, while the animal will be used in the experiment.
On the other hand, the absence of patterns detected in approximately 17 % of analysed datasets should also not be neglected. Especially high share of datasets exhibited the “initial acclimatisation” pattern One potential explanation for found absence of “initial acclimatisation” pattern that is likely caused by higher rates of activity at the beginning of the experiment, followed by lower rates of activity due to individual variability in susceptibility to handling stress (e.g. Rodríguez-Prieto et al. 2011). Thus these patterns may not be relevant to the population/species, but are more an artefact of taking hourly measurements over the course of a controlled experiment. Tested individuals may also have wet surface of the body due to housing conditions that dries up in the first hour of the experiment. Therefore, tested individual should always be allowed to calm down and dry the surface of the skin of any excessive humidity.
Found temporal patterns, “fluctuation”, “steady decline”, “steady increase” and “middle peak”, may be connected to several mechanisms that modulate skin/ocular resistance to water loss and respiration rates. Known behavioural mechanisms in reptiles for coping with dehydration are decreasing activity and hiding to burrows, shutting down eyes to prevent excessive evaporation through ocular surface, and decreasing respiratory rate (Mautz 1980, 1982; Araya-Donoso et al. 2021). Eyes have been shown to contribute significantly to water loss, since their surface is very permeable and minimization of time spent with the eyes open may be a form of hydroregulatory behaviour (Lanham and Bull 2004; Mathews et al. 2000). The permeability of the eyes may be important especially for species with relatively big eye surfaces compared to body. Furthermore, it is known that the variability of lipid content in the skin of reptiles is supposed to be the most responsible for the rate of cutaneous water loss (Roberts and Lillywhite 1980). Also, different fluid repartition among body compartments may occur as a response to dehydration (Nose et al. 1983; Arad et al. 1989). Overall, these existing evidence of behavioural and physiological hydroregulation not only provide potential explanations for findings of temporal patterns of water loss but also supports their significance.
In conclusion, our analysis of water loss data sets using diverse dataset shed a light on problematics of global comparative studies using average and sum values of water loss. All organisms today are suffering from impacts of global climate change that involve experiencing higher ambient temperatures and drying conditions, therefore, for improving our understanding of water loss in connection with complex impacts of global climate change we need to first fully understand properties of this physiological and behavioural functional trait. Simple and straightforward statistical approach and reporting should be used and the length of experiments should maintain long enough and comparable to be able to detect temporal patterns. Overall, our main recommendation for future is performing 12-hour experiments, collecting hourly measurements and providing raw data discriminated by time and individual. A standardized framework for analysing and reporting temporal patterns of water loss may follow the same procedure as used in this study: 1) comparing hourly values of water loss (EWLi) with GAMM, using “hour” as factor, 2) plotting fitted hourly EWLi values, 3) if “initial acclimatisation” pattern is identified, the first value should be removed and the remaining dataset should again be tested following steps 1) and 2). Reporting should include results of GAMM – non-linear explanatory variable “hour”, size and weight as covariates, individual as random effect. Plots should use fitted values for each hour and made per species/population. We believe that in the future, especial care should be taken not to ignore temporal patterns when using water loss data in order to improve our understanding of functional responses of organisms to dehydrating conditions.