Due to small sample sizes, I pooled cohort data to investigate the shapes of the reaction norms responding to each temperature treatment. I also pooled temperature

treatment data to investigate the shapes of the reaction norms responding to each cohort treatment. I presented mean±1se throughout the text and figures.

Results

Some tadpoles did not make it to Gosner 42. Twenty-two tadpoles died. Ten tadpoles in the LT-LF treatment had extended larval periods, and did not enter G42 after 200 days left (week 28). I stopped the experiment on week 28 due to logistic reasons. The 32 tadpoles were excluded from analyses. Overall, the percentage of individuals reached G42 were 75%, 75%, 94%, and 87% in LT-LF, LT-HF, HT-LF, and HT-HF treatments, respectively, in the spring, and 20%, 60%, 87%, and 87% in the summer.

Metamorphic weight

Food and cohort had significant effects on metamorphic weight (three-way ANOVA in GLIMMIX, food effect, F1, 83 = 104.7, P < 0.001; cohort effect, F1, 83 = 24.3, P < 0.001;

Fig. 3.1a and Table 3.3a). Tadpoles metamorphosed with heavier weights under high (318.33±17.95 mg, Table 3.1) than low (194.00±11.23 mg) food treatment. Spring tadpoles (348.05±18.25 mg) generally metamorphosed with heavier weights than summer ones (288.61±17.66 mg). There was no temperature effect (temperature effect, F1, 83 = 1.52, P = 0.22), nor interaction among factors (P > 0.07 in all interactions).

Metamorphic age

All main effects and interactions were significant. Generally, tadpoles reached

metamorphosis (G42) significant faster in high than low food treatment (three-way ANOVA with permutation, food effect, F1, 83 = 250.1, P < 0.001; Table 3.3 and Fig.

3.1b); faster in high than low temperature treatment (temperature effect, F1, 83 = 370.2, P

< 0.001); and faster in spring than summer cohort (food effect, F1, 83 = 79.4, P < 0.001).

Tadpoles reached G42 the fastest under high temperature and high food treatment, and the slowest under low temperature and low food treatment (temperature × food

interaction, F1, 83 = 39.4, P < 0.001). This is especially true for summer cohort under low food condition (cohort effect, F1, 83 = 79.4, P < 0.001; temperature × cohort

interaction, F1, 83 = 60.3, P < 0.001; food × cohort interaction, F 1, 83= 113.6, P < 0.001;

temperature × food × cohort interaction, F1, 83 = 59.5, P < 0.001). In the spring cohort, the effect of food was stronger in high than low temperature. In contrary, in the summer cohort, the effect of food was much stronger in low than high temperature (Fig. 3.1b).

Summer tadpoles had 2.5 times longer larval period than spring ones in the LT-LF treatment (182 and 74 days, respectively, Table 3.1; Scheffé’s test, P < 0.001).

Growth rate

Nearly all main effects and interactions, except cohort main effect, were significant (three-way ANOVA in GLIMMIX, cohort effect, F1, 83 = 2.5, P = 0.117; Table 3.3c).

Generally, tadpoles had significant higher growth rates in high than low food treatment

(food effect, F1, 83 = 233.7, P < 0.001; Fig. 3.1c), and grew significantly faster in high than low temperature (temperature effect, F1, 83 = 158.7, P < 0.001). Temperature and food interacted (temperature × food interaction, F1, 83 = 69.4, P < 0.001). The effect of temperature was stronger in high than low food treatment. Tadpoles had the highest growth rates under high temperature and high food treatment, and the lowest growth rates under low temperature and low food treatment. Such an interaction was

cohort-dependent (temperature × food × cohort interaction, F1, 83 = 5.4, P = 0.022).

Spring tadpoles had faster growth rate (4.04±0.24 mg/day) than summer ones

(1.81±0.20 mg/day) under lower temperature, but the two cohorts had similar growth rates under high temperature (10.15±0.77 and 10.81±0.75 mg/day, spring and summer cohort, respectively).

Percentage of weight loss

Temperature and food had significant effects on percentage of weight loss (three-way ANOVA in GLIMMIX, temperature effect, F1, 81 = 13.5, P < 0.0004; food effect, F1, 81 = 6.7, P = 0.01; Fig. 3.1d and Table 3.3d). Tadpoles lost less weight before metamorphosis under high than low temperature treatment, and under high than low food treatment.

Although the cohort main effect was not significant, there was a significant temperature

× food × cohort interaction (F1, 81 = 8.26, P = 0.005; Fig. 3.1d). Summer tadpoles lost a

lot more weight under high food, low temperature condition (Scheffé’s tests, P = 0.01).

There was no significant difference among four treatments in spring (Scheffé’s tests, P

> 0.05 in all pair-wise comparisons).

Growth trajectories

Spring and summer cohorts showed different patterns of growth trajectories (Fig. 3.2).

Spring tadpoles raised under the HT-HF condition metamorphosed in 3-5 weeks, followed by the HT-LF (in 6-8 weeks) and LT-HF (in 9-14 weeks) treatments. The LT-LF tadpoles metamorphosed in 10.5-17 weeks (Fig. 3.2a). Summer tadpoles raised under the HT-HF condition metamorphosed in 2-3 weeks, followed by the HT-LF and LT-HF treatments (both in 8-13 weeks). The LT-LF tadpoles metamorphosed in 26 weeks (Fig. 3.2b). The majority of tadpoles in the LT-LF treatment did not

metamorphosed in 15 weeks when all larvae in the other three treatments did. I moved 5 tadpoles to high temperature (LT-LF-HT, Fig. 3.2b). They soon metamorphosed in 3-4 weeks after the transfer. Tadpoles remained in LT-LF treatment (LT-LF-LT, Fig. 3.2b) metamorphosed at the end of week 26.

Fitting the developmental threshold model

I found a significant L-shaped reaction norm of age and weight at metamorphosis for

summer, but not spring cohort (both temperature treatments combined, Fig. 3.3; Table 3.2). The exponential decay function for the summer cohort was metamorphic weight

=0.16 + 0.28 × e-0.03 × metamorphic age. Tadpoles raised under HF treatment generally metamorphosed with larger size. The variation in size was big. They delayed metamorphosis only when temperature was low. However, tadpoles raised with restricted food generally metamorphosed with small size. The variation in size was small. And it revealed the physical size threshold (154.0 mg, Table 3.2) for

metamorphosis in F. limnocharis, which was occurred in summer cohort under restricted food.

I also found a significant L-shaped reaction norm of age and weight at

metamorphosis at 29℃, but not 22℃ treatment (both cohort combined, Fig. 3.4; Table

3.2). The exponential decay function was metamorphic weight =0.13 + 0.37 × e^{-0.03 ×}

metamorphic age. It revealed the physical size threshold (186.0 mg, Table 3.2) for

metamorphosis in F. limnocharis at 29℃. However, an L-shaped reaction norm function could not be fitted to the observations at 22℃.

Discussion

The development threshold model postulates that there is a minimum size or condition that must be attained before a life-history transition such as metamorphosis can occur.

Given a developmental threshold, tadpoles inhabiting poor conditions acquire little resource, and require more time to reach the minimum size. They have little to gain in body size by further postponing the transition, and should metamorphose as soon as they reach the threshold. Thus, one should observe the tadpoles metamorphose at similar sizes, likely at various ages (due to individual variations). In contrast, tadpoles inhabiting superior conditions can reach the minimum size quickly. They should have a relatively small metamorphic age. For them, it also pays to postpone entering

metamorphosis a little because they can gain body size easily, and because an individual’s fitness often increases with size. Thus, one should observe the tadpoles

metamorphose at various sizes (due to individual variations). This leads to a negative relationship between superior and poor condition on a metamorphic age (x-axis) by size (y-axis) space, and an L-shaped reaction norm, where the asymptotic lowest body size depicts the developmental threshold size for metamorphosis.

This study shows that the life-history strategy of the Indian rice frog in response to different food and temperature regimes partially conform to the general theoretical

framework of the developmental threshold model. The L-shaped reaction norm between metamorphic age and size was found in summer, but not spring cohort; under high, but not low temperature. Under the high food with high temperature condition, both spring and summer cohorts have high growth rates (Fig. 3.1c). Particularly, there is an apparent short metamorphic age with little variation (coefficient variations of spring and summer cohorts are 11.2 and 7.2%, respectively, Table 3.1). On the other hands, the variations in metamorphic size are relatively large under the same high food, high temperature treatment likely due to individuals' plasticity of food assimilation (Fig. 3.3a).

Nevertheless, a strong vertical boundary occurs in both cohorts (Fig. 3.3). In contrast, both spring and summer cohorts have low growth rate under restricted food, regardless of temperature. Tadpoles of both cohorts reached metamorphosis at small sizes with relatively little variation (coefficient variations of spring and summer cohorts are 18.0 and 13.5%, respectively, Table 3.1), although the variations in metamorphic age are large (Fig. 3.3).

There is a strong cohort × temperature × food interaction in metamorphic age.

Particularly, under restricted food, low temperature made greatly delaying of metamorphosis for the summer but not spring cohort, which resulted in a very clear horizontal boundary in summer cohort (Fig. 3.3b). It seems that high rearing

temperature could compensate for the effect of low food, and fast-track the tadpoles to the threshold size. Yet, it happened only in the spring cohort. Accordingly, the

differential responses of spring vs. summer cohort to the low-temperature, high-food treatment play a key role in determining if one would find L-shaped reaction norm. If one removed data pertaining to the low-temperature, high-food treatment, then both spring and summer cohorts would show a significant L-shaped reaction norm. It is clear that one has to consider the interaction between food and temperature when applying the developmental threshold model (Day & Rowe, 2002), as suggested by Kuparinen et al. (2010) for a temperate species, R. temporaria. This study further suggests that one

need to consider the effect of cohort. It seems that the L-shaped reaction norm would be more evident for a cohort (or species) constrained by time: cohort breeds toward the end of breeding season, high latitude population, or temperate species.

Kuan and Lin (2011) found a strong cohort effect in the Indian rice frog. Spring tadpoles responded to low temperature with large metamorphic weight, whereas summer ones did not. In addition, both cohorts responded to high temperature with shorten larval period, the summer cohort responded with a much greater magnitude than the spring one. They concluded that the spring and summer cohorts used ''size'' strategy and ''rate'' strategy in response to background temperatures, respectively. Supporting

their conclusions, the present study found both spring and summer tadpoles had greater body growth rates (Fig. 3.1c) and less weight loss before metamorphosis (Fig. 3.1d) when raised under their respective field temperature with sufficient food. Spring cohort raised at low temperature has significant larger metamorphic weight than summer one, while summer cohort has concentrated and more shorten larval period than spring one at high temperature (Table 3.1). Although the results of metamorphic age between spring and summer cohorts do not have significant differences, the pattern is similar to that in Kuan and Lin (2011). Furthermore, upon checking the variability of metamorphic age within cohorts, I found that the coefficient variation of summer cohort is much smaller than spring one (7.2% & 11.2% spring & summer, respectively, Table 3.1). Thus, this study also supports that spring cohort responds to low temperature with metamorphic size advantage, while summer cohort grows faster with time advantage at high temperature under abundant resources.

This study provides further insights than Kuan and Lin (2011) by putting the tadpoles through the low food treatment. Both spring and summer cohorts extended their larval periods when facing the most extreme environment: restricted food with low

temperature. A pattern also found by Gomez-Mestre et al. (2010). However, the two

cohorts responded differently. Under restricted food, low temperature greatly delayed

metamorphosis for the summer but not spring cohort. Spring tadpoles have not only larger metamorphic weight, but also faster growth rate than summer ones. On the other hands, even at its optimal temperature (29℃), summer tadpoles have much longer larval periods than spring ones. This reveals that food limitation may be more critical to summer larvae. In nature, summer is usually high in food supply. Summer tadpoles may not have the plasticity to deal with food shortage. In the nature, summer cohort has a shorter growing season (3 months) than spring one (> 4 months). The spring crop is grown from March to June for more than 4 months; while the summer crop is grown from August to October, then harvested. There is no more irrigation water after October.

I observed that summer larvae have smaller metamorphic threshold sizes. It may be an adaptive strategy that allows them to leave water quickly. This observation is consistent with empirical findings that larvae had the lowest thresholds from the ephemeral

environments due to time-constrained stress when faced with low food conditions (Morey & Reznick, 2000; Morey & Reznick, 2004; Kuparinen et al., 2010).

Moreover, low food and low temperature may be signaling the coming of fall and winter. Summer cohort has not only lowest metamorphic percentage, but also

dramatically extended larval period under the LT-LF condition. Summer cohort took 2.5 times of the days more than spring one to reach metamorphosis. In subtropical lowland

of Taiwan, Rana adenopleura and Polypedates braueri larvae raised at lower

temperature would extend larval period with retarded development rate (Chuang, 2006;

Hsu, Kam & Fellers, 2012). Larvae growth rates were independent of development rate, thus they kept growing in size, and finally metamorphosed at larger size than early metamorphs after winter. In this study, the growth rate of LT-LF treatment was the lowest among all treatments (0.76 mg/day, Fig. 3.1c), and it seems almost retarded.

Interestingly, when I transferred summer tadpoles from LT to HT treatment with restricted food, they can quickly reach metamorphosis with smaller size and as early as 8 weeks with LT-LF larvae (LT-LF-HT in Figure 3.2b). This again reveals that high temperature can compensate the effect of low food in F. limoncharis tadpoles. The species have efficient phenotypic plasticity to cope with changing environments.

Tadpoles collected from the neighboring rice field are able to metamorphose at the average size of 159.0 to 359.8 mg, with the average metamorphic age of 18-182 days under various temperature and food level regimes. The incredible ranges in

metamorphic size and age suggest high level of phenotypic plasticity. This high plasticity may be the reason why F. limoncharis is the most distributed anuran in lowland Taiwan.

As mentioned earlier, the metamorphic size and age of the spring cohort are more

variable than the summer cohort. The reason might be behind the reproductive biology of F. limoncharis. A study in central Taiwan revealed that mature females might be able to lay eggs more than once during their breeding season (Alexander, Chang & Yang, 1963). Thus, females joining the spring breeding cohort may invest more energy on reproduction. In spite of lower and variable temperature in spring, tadpoles have longer growing seasons with lasting irrigation water to metamorphosis at larger size. Larger metamorphs have better fitness on land because of the advantages in survival rate, foraging ability, locomoter performance (Cabrera-Guzman et al., 2013). In contrast, females participating in the summer cohort are close to the end of breeding season. If a female F. limoncharis can lay eggs twice a year, I suppose that the second reproduction of females would investigate less reproductive energy in summer than in spring. Besides, summer tadpoles not only have shorter growing seasons, but also smaller metamorphic sizes. In the field, I observe a large amount of froglets in May, but only a few in September (unpublished data). This implies that spring cohort may be the major

breeding cohort of F. limoncharis. Therefore, females with good quality, such as bigger size individuals, would be more competitive while joining the spring cohort than small ones. Finally, it makes the results what I observed in bigger metamorphic size in spring larvae.

In conclusion, the results were similar to previous study (Kuan & Lin, 2011) when larvae were raised under ad libitum food. However, under restricted food, both spring and summer cohort reduced their metamorphic size. Spring tadpoles still had size advantage than summer ones at low temperature, whereas summer ones lost their ''rate advantage'' at high temperature and had slower growth rates than spring ones. Besides, this is a pioneer study to demonstrate that the subtropical species, F. limnocharis, has fitted the L-shaped reaction norm between metamorphic age and size. A more complete picture of how tadpoles cope with the diverse environments must consider the effect of cohort effect.

Table 3.1 The average metamorphic age and weight of spring and summer F.

limnocharis tadpoles among eight treatments

Cohort Spring Summer Spring Summer

Temperature 22℃ 22℃ 29℃ 29℃

Food availability^a LF HF LF HF LF HF LF HF

Metamorphic age

Mean (days) 74.4 63.0 182.0 57.8 43.6 23.4 56.2 17.8 SD 20.5 18.0 4.6 14.4 5.0 2.6 17.1 1.3 Coefficient variation

(%) 27.5 28.6 2.5 25.0 11.5 11.2 30.5 7.2 Metamorphic weight

Mean (mg) 217.7 359.8 159.0 258.2 230.6 336.3 168.7 319.0 SD 48.9 74.6 17.0 65.7 31.3 50.5 27.6 56.6 Coefficient variation

(%) 22.5 20.7 10.7 25.4 13.6 15.0 16.4 17.8

a HF and LF are refer to ad libitum and restricted food, respectively.

Table 3.2 Regression lines: Metamorphic weight = a + b × e-c × metamorphic age that fitted on the metamorphic weight by age space to describe L-shaped reaction norms. The 2.5% lower confidence limits of metamorphic weight are also given.

High temperature Low temperature Spring cohort Summer cohort

a 0.13±0.06* 0.19±0.06** -0.05±1.63 0.16±0.03***

b 0.37±0.05*** 1.28±1.65 0.40±1.59 0.28±0.07***

c 0.03±0.02 0.04±0.03 0.004±0.02 0.03±0.02

Limits of metamorphic weight (mg)

185.98 180.63 209.87 154.26

* p<0.05 ; ** p<0.01 ; *** p<0.001

(a)

Fig. 3.1 The (a) metamorphic weight, (b) metamorphic age, (c) growth rate and (d) percentage of weight loss of spring (circles) and summer (triangles) F. limnocharis tadpoles raised under four treatments (22℃-Low-Food, 22℃-High-Food, 29℃

-Low-Food, and 29℃-High-Food). Sample sizes are 12, 12, 15, 14, and 3, 9, 13, 13 for the spring and summer cohort, respectively. Values are given in MEAN±1SE. Significant differences based on post hoc of Scheffé’s test are indicated by vertical bars.

(a)

Fig. 3.2 Growth trajectories of (a) spring and (b) summer F. limnocharis larvae under four treatments. Body weights were recorded weekly from the start of the experiment to metamorphosis (Gosner 42). Values are given in MEAN±1SE. Arrows pointing different symbols on the x-axis indicate the average metamorphic age in each treatment.

(a)

Fig. 3.3 Reaction norms for metamorphic age and weight observed in (a) spring and (b) summer cohorts of F. limnocharis at 29°C (triangles) and 22°C (circles), under ad libitum (closed symbols) and restricted food (open symbols) regimes. Only panel (b) showed a significant L-shape reaction norm which was fitted with a non-linear regression. Dashed lines are given as 2.5% lower confidence limits of the ages and weights.

(a)

Fig. 3.4 Reaction norms for metamorphic age and weight observed in temperatures of (a) 29°C and (b) 22°C, the spring (circles) and summer (triangles) F. limnocharis larvae were raised under ad libitum (filled symbols) and restricted food (open symbols). Only panel (a) showed a significant L-shape reaction norm which was fitted with a non-linear regression.

Dashed lines are given as 2.5% lower confidence limits of the ages and weights.