We surveyed over 60 papers providing data or speculating on the relationship between rainfall and nest predation rates in freshwater turtles. The papers were found by searching the references sections of published work and by on-line searches using Google, Google Scholar, and the academic research databases of the University of Wisconsin Library System (>1100 e-collection content selections from Primo Central Index [PCI] from Ex Libris [ProQuest] including Web of Science and Scopus) using the keywords: “turtle:nest:predation:rain;precipitation.” The papers were mostly peer-reviewed publications but some unpublished M.Sc. and Ph.D. theses were also included. Papers that only cited previous publications without providing new data were not included or are distinguished as such in this review.

To supplement literature reviews, we also analyzed data on rainfall, nesting activity, and nest survivorship from two populations of Ouachita Map Turtles (Graptemys ouachitensis) located on the lower Wisconsin River within 10 km of Spring Green, WI, United States (43°10′38′′N and 90°04′02′′W). Both nesting sites are on sand terraces approximately 15 and 52 m from the main river channel, respectively, and are comprised of various xerophytic herbaceous vegetation covering approximately 20% of the surface, with the remainder being open sand (for a more complete description of these sites, see Geller, 2012a).

The study was carried out over 12 years from 2008 to 2021 excluding the years 2012 and 2018. Newly constructed turtle nests were located beginning in late May of each study year by review of images from pole-mounted trail cameras (RECONYX^®, Inc.; Holmen, WI, United States) monitoring each nesting area. Cameras were programmed to take continuous time-lapse images at 1-min intervals to document nesting events along with motion-triggered series of more closely timed photographs to provide detailed documentation and timing of nest predator (all Northern raccoons, Procyon lotor) visits. Surveillance ended each year with either the documented predation or hatchling emergence of the last monitored nest in each year.

We obtained total daily rainfall amounts (in mm) from two sources: (1) data from the Lone Rock Tri-County Airport, Sauk County, Wisconsin, approximately 9.7 km northwest of the study sites (downloaded from NOAA National Climate Data Center¹, station id: GHCND:USW00014921); and (2) from an on-site rain gauge. Where the two sources differed, we used the rainfall amounts from the on-site gauge, although these were not always available due to logistic constraints. In some years, rainfall timing on the site was determined by use of a funnel-driven waterwheel device enclosed within a Lucite^®-fronted housing placed within the peripheral camera field-of-view (Geller, 2012a). Date-stamped time-lapse images indicating waterwheel movement delineated rain event timelines. These units were accurate to within 0.2 h of rainfall duration (sprinkles to heavier amounts), as determined by field tests and camera-visible nocturnal rainfall. In all years, within-day rainfall event amounts and timing were estimated using hourly rainfall tracking charts from Lone Rock, Wisconsin via Weather Underground historical weather charts². We also used this database to derive metrics on daily high, low, and mean temperatures; and daily high, low, and mean air pressures.

Our unit of observation was a day on which turtle activity was monitored for which we had rainfall data. For modeling purposes, we defined the nesting season as starting one day prior to the first recorded turtle nest and ending one day after the last recorded turtle nest for each year (12 years from 2008-2021, excluding 2012 and 2018). We excluded 26 days when the sites were flooded so nesting could not take place and, to be conservative, also excluded 28 nests constructed within 5 days after flooding because the timing of these nests may have been affected by females retaining eggs during flooded periods. In some years, some nests were protected from predation using nest cages. Protected nests were included in the analysis of nesting behavior in relation to rainfall but not in examining how rainfall affects nest predation risk. Reported sample sizes reflect varying numbers of camera records available for analysis, as influenced by camera position, intervening vegetation, and other variables.

We modeled the number of nests constructed each day (counts from 0 to 6) as a function of four variables that capture how turtle nesting might respond to daily rainfall variation: (1) rain on the day of nesting or not; (2) rain on two consecutive days (high rain frequency periods); (3) no rain for two consecutive days (low rain frequency periods); and (4) no rain on the day and rain the next day (nesting in anticipation of oncoming rain). Although we had data on the amount of rainfall for each day, using rainfall presence/absence provided a better fit to the data as indicated by lower values of Akaike’s Information Criteria (AIC).

We fitted four models to the data, with the number of nests constructed each day as the response variable and one of the four rainfall variables above as a univariate predictor in each model. We specified a Poisson error distribution (appropriate for count data) and included an observation level factor and year as random effects in each model. The observation level factor accounted for any overdispersion in the data, while the inclusion of year as a random effect accounted for the possibility that the probability of nesting per day was higher, on average, in some years than others. We compared the fit of the four models to the data using AIC, with better fitting models having lower AIC values. The models were fitted using the lme4 package (Bates et al., 2015) in the statistical analysis software R (R Core Team, 2017). We also tested for relationships the between number of nests constructed and the measures of daily temperature and air pressure by including these as univariate predictors as described above.

We reviewed 42 studies that quantified or addressed the potential association of nesting with rainfall in 23 species of freshwater turtles. Of these, 29 (69%) studies demonstrated or suggested that the propensity to nest was associated with rainfall, compared to 13 (31%) studies that found no association (Tables 1, 2). The research to-date is, thus, somewhat equivocal with regards to the association of rainfall and turtle nesting activity.

However, when associations have been found, our review revealed a general consensus for turtle nesting to occur during or after rainfall, rather than before it (Table 1). In addition, the few studies that systematically examined the effect of rainfall amount on nesting activity found larger rainfall amounts more likely to stimulate turtle nesting during or after the rain event than lesser amounts (≥10 mm, Jackson and Walker, 1997; >2.5 cm, Tucker, 1997; see also Walde et al., 2007; Bernstein et al., 2015; Table 1). However, Buckardt et al. (2020) found no relationship between rainfall amount and the propensity to nest, while others noted that heavy rainfall suppresses nesting activity (Legler, 1954; Burger and Montevecchi, 1975), although in some cases this is likely due to a concurrent decrease in air temperatures to non-optimal levels rather than rainfall amount effects per se (e.g., Harding and Bloomer, 1979; Vogt, 1980).

Some early reports (e.g., Pallas, 1960) are primarily anecdotal observations about particular rainfall events and do not provide the ecological context in which to evaluate the uniqueness of the association, such as the numbers of turtles nesting in other conditions. Limitations in study designs or data acquisition have restricted some other studies to basic reports on the association of nesting with rainfall on a simple binary, rain present/absent basis within a given calendar day, and do not allow assessments of the impact of rainfall within different temporal periods (e.g., previous 24 h) on nesting propensity or, importantly, how rainfall amounts were quantitatively distributed in relation to the timing of nesting events. For example, 77% of the reviewed studies that failed to find an association of nesting activity and rainfall did not record and include rainfall amounts in their analyses (Table 2). These methodological differences may have contributed to some of the disparity in findings regarding the association of rainfall and freshwater turtle nesting activity.

However, even better-quantified studies sometimes report different responses of turtles to rainfall patterns. For example, Bowen et al. (2005) noted that Chrysemys picta nested on virtually every day during the nesting season and suggested that timing of nesting in that species was primarily a function of physiological readiness and sufficiently warm temperatures, rather than of rainfall or other environmental variables. However, Bowen et al. (2005) were unable to provide explanations for why this behavior contrasted with those for Emydura macquarii and Chelodina expansa in that same study, whose nesting was correlated with rainfall and associated changes in air temperature. Disparity in study findings sometimes appear even for the same species at the same nesting sites (e.g., for Chrysemys picta at the Thompson Causeway, Illinois [Bowen and Janzen, 2005; Muell et al., 2021]; see Tables 1, 2).

Nonetheless, when associations have been found, many studies have concluded that rainfall on a given day is an important determinant of nesting propensity in freshwater turtles, either when considered as a single factor (Congello, 1978; Georges, 1984; Burke et al., 1994, 1998; Roosenburg, 1994; Wilson et al., 1999; Bowen et al., 2005, in part; Walde et al., 2007; Bernstein et al., 2015; Espinoza et al., 2018) or in conjunction with other meteorological variables such as rising or relatively high air temperatures and falling barometric pressures above certain minimums (Hammer, 1969; Burger and Montevecchi, 1975; Clay, 1981; Congdon et al., 1987; Stott, 1988; Kuchling, 1993; Tucker, 1997; Najbar and Szuszkiewicz, 2005; Muell et al., 2021).

Still other studies have found air and/or water temperatures alone to be principal cues to initiate nesting, with a preference for nesting on days with the relatively high daily temperatures that optimize locomotor performance and reduce the time spent nesting (e.g., for Chrysemys picta, Bowen et al., 2005; Frye et al., 2017) or that allow turtles that nest during the evening to maintain their body temperatures at functioning levels (e.g., for Emydoidea blandingii, Buckardt et al., 2020). Pig-nosed turtles (Carettochelys insculpta) ceased nesting for up to several days during cool periods in the tropical (winter) dry season (Doody et al., 2003). Rather than rainfall, additional studies have noted associations of nesting activity with bright moon phases, possibly a function of social facilitation and/or as a mechanism to reduce individual or nest predation (Escalona et al., 2019; Buckardt et al., 2020). These complexities show that single-factor analyses of rainfall frequencies alone (e.g., Aresco, 2004; Geller, 2012a) are not likely to be as informative for a given species and context as studies incorporating rainfall amounts and a broader array of meteorological parameters in their analyses, such as the concurrent effects of air and water temperatures, and time since last rainfall (noted by Jackson and Walker, 1997; Muell et al., 2021). Overall, the literature thus shows significant variation in the assessment of whether or not freshwater turtles appear to nest in greater numbers after rainfall, suggesting that this association is not a generalizable aspect of chelonian reproductive biology.

A total of 245 G. ouachitensis nests were constructed on 147 of the 320 days on which activity was monitored, giving an average of 1.7 nests constructed on days when nesting occurred.

Rain fell on 142 of the 320 days (44.3%) when activity was monitored across all study years. Most nests (57.0%, n = 139) were constructed on calendar days without rainfall. Similarly, most nests were constructed more than 24 h after previous rainfall (54.9%, n = 134), more than 24 h before the next rainfall (63.1%, n = 154), and many (38.1%, n = 93) were without rain during both pre- and post-nest construction periods. A small number of nests (2.5%, n = 6) were constructed during rainfall itself. The all-year mean number of nests constructed on calendar days without rain (0.84 nests/d) did not differ from the mean number constructed on days with rain (Model 1 in Table 3 and Figure 1). Similarly, there was no association between the number of nests constructed on a given day and (any) rainfall amount (p = 0.81) or on calendar days with ≥ 20 mm of rain (p = 0.84).

On a calendar day basis, Model 2 (rain on the day of nesting and rain on the previous day) provided the best fit to the data on the number of nests constructed per day (Table 3). The negative parameter estimate for this model indicated fewer nests were constructed when it had rained two days in a row. There was strong evidence that the parameter estimate for Model 2 was negative: the 95% confidence interval, given by ±2 times the standard error, did not include zero and the outcome was associated with a low P-value. Figure 2 summarizes the pattern of turtle nesting in relation to whether it rained for two days or not. While there was a much lower probability of nesting given rain two days in a row (nesting probability = 0.31 relative to 0.51 if this did not occur), there was less of a difference in the mean number of nests constructed if nesting occurred on either occasion (1.7 versus 1.48, Figure 2). There were no significant effects of temperature or barometric pressure on the propensity to nest.

We reviewed 21 studies that quantified or addressed the potential for rainfall to increase nest survival in 7 turtle species. In our review, 11 (52%) of the studies demonstrated or suggested that rainfall increased nest survival, compared to 10 (48%) studies that found no effect (Tables 4, 5).

The primary cues used by predators to locate newly constructed freshwater turtle nests are believed to be largely olfactory for mammalian predators, with varying degrees of visual use based on predator species (reviewed in Geller and Parker, 2022); largely olfactory for certain nest-depredating lizards (e.g., Soanes et al., 2015); and largely visual by bird predators (e.g., Jackson and Walker, 1997). Whether visual or olfactory, many researchers have suggested that rainfall may dilute these nest location cues (Table 4). One proposed mechanism by which rainfall diminishes olfactory cues suggests that water percolating through the soil column flushes out the odors produced by soil microbes (Lindbo et al., 2012), thereby reducing what were formerly point-source olfactory signals of disturbed soil at nest locations (Geller, 2015; see also, Buzuleciu et al., 2016). Greater rainfall amounts, percolating to greater soil depths (reviewed in Hess et al., 2018), are likely more thorough in aerosolizing the microbe-produced compounds within newly constructed nest cavities as well as from surrounding substrates, reducing the olfactory gradient. Both the amount of rain and its timing relative to nest construction are, thus, potentially important influences on the degree to which rain reduces nest predation. As a corollary, rainfall during and after nest construction is likely to be more effective in reducing both olfactory and visual nest location cues than when rainfall precedes nesting (Bowen and Janzen, 2005).

These inferences are supported our review, in that all of those which noted an effect of rain in reducing nest predation, either qualitatively or quantitatively, found larger rainfall amounts on the day of or soon after nest construction to be more effective in reducing nest predation than smaller amounts or none (Table 4); No studies suggested that rainfall in advance of nesting would decrease nest predation rates. In contrast, it is difficult to fully evaluate studies that did not show an enhancing effect of post-construction rainfall on nest survival (Table 5) because (like some that did show an effect) none reported the relevant rainfall amounts except for the 7.3 mm reported by Wilhoft et al. (1979) and the experimental work of Buzuleciu et al. (2016), wherein 2 cm of dechlorinated tap water was applied to individual artificial nests. However, as pointed out by Czaja et al. (2018), experimentally applied water at nests themselves does not scale in effect to that of larger areas affected during natural rainfall and possibly explains why these negative results differ from other research where the effects of natural rainfall were assessed.

Importantly, none of the studies finding no effect of rainfall on predation rates on newly constructed natural turtle nests with raccoons as predominant predators (Schwanz et al., 2010; Wirsing et al., 2012; Bougie et al., 2020) assessed the temporal association between rainfall and individual nesting events, which is likely central to the dynamics of how rainfall affects the strength of nest location signals (see above). Some other studies that discounted the role of rainfall in reducing turtle nest predation rates recorded red foxes (Vulpes vulpes) as nest predators (Congdon et al., 1983, 1987; Spencer, 2002; Dawson et al., 2014), which may utilize a wider array of nest location cues than some other predators (e.g., raccoons), including, potentially, the scent of eggs themselves (see Congdon et al., 1987; Geller and Parker, 2022). However, whether these differences in sensory abilities explain the apparent lack of rainfall effect on nest survival is unclear. Similarly unexplained are reported increases in nest predation during or soon after rainfall for late-season turtle nests (Congdon et al., 1983; Brooks et al., 1992). However, late-season nests may have a different suite of nest location cues than those at recently constructed nests (see reviews in Riley and Litzgus, 2014; Geller and Parker, 2022), including, potentially, hatchling vocalizations (e.g., Ferrara et al., 2012; Geller and Casper, 2019). As hatchling emergence onto the surface is often associated with rainfall and warm or rising ambient temperatures (e.g., Tucker, 1997; Nagle et al., 2004; Geller et al., 2020), it is possible, although speculative, that rain during late, pre-emergence stages may increase overall hatchling activity, leading to increases in odors from disturbed soils or from the hatchlings themselves, hatchling-produced sounds, or other surface-detectable cues to nest locations.

Nest predation rates on our two nesting sites were high (overall 93.0%, yearly range 75–100%, n = 128) and only 9 unprotected (uncaged) nests escaped predation across all study years. All predation was from raccoons, which were present nearly every night during nesting periods (Geller, 2012b). Ninety-five percent of the nests depredated by raccoons were destroyed within 24 h of construction (mean nest survival = 11.8 h, range 0.0–7.6 days, n = 111), typically by the first raccoon within approximately 1 meter of a given nest (ca. 90% of the time; Geller, 2012b). Most depredated nests had no post-construction rain before predation (84.5%, n = 116 total depredated nests), and all of those without intervening rain were found by raccoons before or during the first night. Raccoons depredated 94.3% of the nests constructed without rain in the previous 24 h (n = 70) and 90.7% of the nests made when rain (median = 4.3 mm, mean = 9.5 mm, range 0.3–78.5 mm) did occur within 24 h before (n = 50) or during nest construction (n = 4) (Yate’s χ² = 0.164, p = 0.685, n = 124), suggesting that rain falling before nest construction had little effect in reducing nest predation rates.

Given the high rate of predation and short nest survival timelines, we did not have sufficient data to reliably identify the characteristics of nests that escaped predation. However, surviving nests tended to have larger amounts of rainfall during both the first night after nest construction (mean = 7.8 mm, SD = 10.27, range 0.0–25.0, n = 9) and through the first four nights after nest construction (mean = 19.4 mm, SD = 14.63, range 0.8–35.1, n = 9) than did depredated nests (first night mean = 0.3 mm, SD = 1.23, range 0.0–10.4, n = 109) (Figure 3), and all surviving nests had either single or multiple rainfall bouts during the 4 days following nest construction. Moreover, raccoons depredated fewer nests having post-nest construction rain within 24 h (85.7%, n = 42 initial nests) than those without rainfall shortly after nest construction (96.4%, n = 83 initial nests), a difference unlikely to arise by chance alone (Fisher’s exact test; p = 0.060, n = 125). Further review of the camera data showed that 58.3% (21 of 36) of the depredated nests constructed within 24 h of the next rain event were detected by raccoons before any of this rainfall occurred. When these nests are removed from the depredated nests within 24 h of post-construction rain counts, the depredation rate of nests actually exposed to post-construction rain decreases to 71.4% (n = 21), strengthening evidence for a non-random difference (Fisher’s exact test; p = 0.002, n = 104). This result supports the idea that raccoons depredate a smaller proportion of nests when rainfall follows nest construction and highlights the importance of determining the exact timing of rainfall and nest predation in the field rather than using metrics based on the simple presence or absence of rain on a given day (see also Bowen and Janzen, 2005).

To date, most research on the role of changing barometric pressure—declines in which are a key component of oncoming rainfall that could be a nest initiation cue to ovipositing turtles—has involved marine species. Drops in air pressure, in conjunction with other meteorological variables, have been suggested to be a cue to deteriorating habitat conditions for loggerhead sea turtles (Caretta caretta), resulting in movements away from nesting areas (Schofield et al., 2010). In other studies, higher barometric pressures were positively associated with higher rates of successful nesting (i.e., non-aborted attempts) by C. caretta and leatherback sea turtles (Dermochelys coriacea) (Pike, 2008; Palomino-González et al., 2020; respectively) and, although correlative in nature, were considered cues promoting nesting behavior. Within freshwater turtles, movements to nesting areas were believed triggered by falling barometric pressures in advance of rainfall by Chelodina colliei (then as C. oblonga) (Clay, 1981). In contrast, research on the allopatric congener, C. longicollis, indicated that, while there was a correlation of increased nesting-related movements out of ponds with falling barometric pressure, the association was not statistically significant and was weaker than that of rainfall itself, and was thus considered indirect and corollary (Stott, 1988, M.Sc. Thesis).

Our review of the literature reveals that air pressure data have rarely been collected in chelonian studies and we, thus, lack basic understanding of its potential role as a cue to nest initiation in freshwater turtles or how this environmental variable may affect turtles at different taxonomic and geographic scales. However, results from our case study indicated no relationship between decreasing barometric pressure or temperature and nesting activity, suggesting that turtles on our sites were not responding to these potential indicators of future rainfall.