Survival of eggs to third instar of late-summer and fall-breeding monarch butterflies (Danaus plexippus) and queen butterflies (Danaus gilippus) in north Texas

Eastern migratory monarch butterflies (Danaus plexippus) have declined over 80% in recent years, but little is known about fall reproduction in the southern U.S. where monarchs may compete with queen butterflies (Danaus gilippus). We provide data on the survival to third instar, associated arthropods, and phenology of fall breeding monarchs and queens in Texas. Monarch and queen survival was relatively high, but varied among years. Oleander aphids (Aphis nerii), spiders, and red imported fire ants (Solenopsis invicta) had minor negative effects on survival. The abundance of monarchs and queens on the study site peaked three to four weeks before the main passage of monarchs in the area. Queens had similar phenology and exhibited a migratory pattern similar to monarchs but on a smaller scale. Survival of fall monarchs is relatively high and potentially important for winter roost recruitment. Fall survival was not greatly affected by any particular arthropod taxon, but may be affected by precipitation. Fall reproduction is a response to available host plants and its timing enables pupae to eclose in time for migration to winter roosts. Management of Asclepias viridis and other native milkweeds to facilitate fall reproduction could facilitate recovery of monarchs if it buffers variable productivity further north. Management should use mowing and burn schedules that promote high quality host plants. Populations of queens should be monitored for their potential to compete with monarchs especially in response to the potential impacts of parasite resistance and climate change.


Introduction
The monarch butterfly (Danaus plexippus) is an iconic species distributed across much of North America (Gustafsson et al. 2015). Based on winter roost data, the eastern migratory population of this species has shown alarming declines in recent years Brower et al. 2018) with some models projecting extinction thresholds within the next few decades (Flockhart et al. 2015;Semmens et al. 2016;Oberhauser et al. 2017;Voorhies et al. 2019;Nail 2020). In response, the Monarch Joint Venture established the Monarch Conservation Implementation Plan (Monarch Joint Venture 2021) focused on enhancing breeding habitat, migratory habitat, and protecting overwintering sites. A major emphasis of the Monarch Conservation Implementation Plan is to restore and enhance pollinator habitat and to increase milkweed host-plant populations. The efficacy of these strategies depends on spatial and temporal assumptions about monarch survival. However, there are gaps in this data, particularly with regard to the reproductive output of fall-breeding individuals in the southern U.S. (Flockhart et al. 2013).
There are small populations of monarchs in Florida, on Caribbean islands, and along coastal areas of the Gulf of Mexico that are permanent residents (Howard et al. 2010;Dockx 2012;Zanden et al. 2018). In Florida and likely in other locations with resident populations, there is mixing between resident individuals and migratory individuals (Knight and Brower 2009;Zanden et al. 2018). Aside from these small populations, almost all of the eastern migratory population of monarchs spend their winter at roost sites in central Mexico (Vidal and Rendón-Salinas 2014). In the spring, these butterflies migrate north and re-colonize most of North America east of the Rocky Mountains (Flockhart et al. 2013). Traditionally, it was assumed that the last, most northern generation enters reproductive diapause, migrates south to Mexico, and migrates north again in spring to renew the cycle (Malcolm et al. 1993).
However, many monarchs migrating through southern latitudes in late summer or early fall either come out of reproductive diapause or have never entered diapause (Batalden and Oberhauser 2015;Majewska and Altizer 2019). These individuals lay eggs in the southern U.S., thereby resulting in an additional generation (Calvert 1999;Borland et al. 2004;Prysby and Oberhauser 2004;Flockhart et al. 2013;. There is evidence that fall reproduction supplements the overwintering population in Mexico (Flockhart et al. 2013). Stable isotope data demonstrate that about 11% of monarchs at overwintering sites in Mexico originate from south-central portions of the U.S. and, in some years, up to 30% of the wintering population may originate from areas of Texas, Oklahoma, Louisiana, and Arkansas (Flockhart et al. 2017). Importantly, fall breeding may augment the overwintering population in years when productivity in the U.S. Midwest is poor. These data emphasize the need to focus monarch conservation across the entire breeding distribution in appropriate places at appropriate times (Flockhart et al. 2017). Current monarch breeding habitat conservation focuses on spring and summer reproduction and largely neglects fall breeding in the south because fall breeding is poorly understood, especially in the critically important migratory corridor in Oklahoma and Texas.
There is little detailed information on the survival of fallbreeding monarchs along the main corridor in the southern U.S. or on the ecological context of this generation and its associated host-plant arthropods. Most information on fall reproduction emphasizes milkweed availability along with egg and larval abundance (Calvert 1999;Prysby and Oberhauser 2004;Baum and Sharber 2012;Baum and Mueller 2015;Nail et al. 2015) or emphasizes the potential negative impact of exotic milkweed on fall monarchs (Batalden and Oberhauser 2015;Satterfield et al. 2015;Majewska and Altizer 2019;Clement and Crawford 2020). For example, one study demonstrated that field collected late instars show high levels of parasitism (Mueller and Baum 2014), but did not quantify overall survivorship of eggs to any particular stage of development. Knowledge of the survivorship of fallbreeding monarchs is needed to determine how productive this generation is and to inform efforts that might facilitate the recovery of eastern monarchs. In this paper we provide the first detailed information on fall-breeding monarch egg survival to the third instar at a site in northeast Texas. In addition, we provide information on the phenology of adult and larval site occupancy in order to put fall reproduction into a temporal context.
The focus was fall reproduction in monarch butterflies. However, it quickly became apparent that we also needed to document queen butterflies (Danaus gilippus) which breed in the same area in fall, share the same host plants, and are potential competitors (Brower 1962;Brym et al. 2020). Furthermore, while the eastern subspecies of the queen is well studied in Florida (Brower 1961(Brower , 1962McLaughlin and Myers 1970;Ritland 1991a, b;Moranz and Brower 1998), and there is some literature on the western subspecies in the Mojave and Sonoran Deserts (Sakai 1993;Grodsky et al. 2020;Saul-Gershenz et al. 2020), there is almost no literature on the biology of the queen butterfly in Texas. What little is known indicates that queen butterflies in Texas migrate back and forth into Mexico (Einem 2003;Hobson et al. 2021). This migratory tendency coupled with the general potential for tropical and subtropical butterflies to expand northward in response to warming climates ( Kocsis and Hufnagel 2011) makes it likely that queen butterflies will expand their distribution northward (Moskowitz 2003). Monarchs and queens exhibit density-dependent competition (Brower 1961(Brower , 1962, but monarchs are more vulnerable to protozoan ectoparasites (Barriga et al. 2016). Monarchs might also be negatively affected by climate change (York and Oberhauser 2002;Zipkin et al. 2012;Zylstra et al. 2021). If so, then the competitive dynamic between monarchs and queens could be important under future climate scenarios. This makes it essential to monitor queen populations as they expand into the U.S. Because queens are an important aspect of the ecology of fall reproduction by monarchs, we felt it necessary to provide data on the reproductive success, abundance, and phenology of queens at our study site.
Lastly, we wanted to document the ecological context of fall reproduction for both monarchs and queens relative to the arthropod community associated with milkweed host plants. The aim was to see how these arthropods might affect egg and larval survival. We pay added attention to the possible effects of red imported fire ants (Solenopsis invicta) (hereinafter referred to as fire ants) as this species has been implicated as a major predator of monarchs in the southern U.S. (Calvert , 2004.

Study site
The study was conducted on a 28-hectare old-field on the outskirts of the city of Sulphur Springs, Hopkins County, Texas (33° 07′ 49.1″ N, 95° 39′ 03.3″ W). The site, owned by the Sulphur Springs Economic Development Corporation (SSEDC), was vacant land typically mowed in spring, summer, and fall, in order to suppress woody vegetation and to harvest hay. During the two years of our study, the field was mowed in early May, mid-July, and after we had finished collecting data, at the end of October in 2017 and the middle of November in 2018. Historically, the site was cropland, but more recently, and prior to acquisition by the SSEDC, the land was converted for hay production and planted with Bahia grass (Paspalum notatum) and Bermuda grass (Cynodon dactylon). The site had relatively low plant diversity and was largely dominated by grasses. However, at least ten years of minimal management led to the invasion of this field by other grasses and forbs. In particular, the field contained a high density of Green Milkweed (Asclepias viridis), the only species of milkweed on the study site. On 25 September 2017, an analysis of 35 randomly placed 50 m 2 circular quadrats found the density of the milkweed plants to be 2069 ± 373.8 plants/ha.

Data collection
Data were collected daily from 15 August 2017 through 26 October 2017 and from 15 August 2018 through 3 November 2018. Field seasons ended when eggs could no longer be found and when the last of our focal individuals had either disappeared or reached the third instar. An exception to daily monitoring occurred in early September of 2018 when a human body was found on the study site. As a result, from 2 September through 4 September 2018, we were unable to access the study site due to the resulting criminal investigation.
Eggs were found by watching females lay eggs or by searching potential host plants for eggs. When an egg was found, the leaf, stem, or flower on which the egg was attached was marked with a non-toxic ink marker. A white numbered flag was placed adjacent to the plant so that it could be subsequently located. Following the method used in other focal individual studies (Prysby 2004;De Anda and Oberhauser 2015;Stevenson et al. 2021), each egg was visited daily between 08:00 h and 14:00 h CDST to monitor its status. Monarch larvae sometimes temporarily leave the host plant (Rawlins and Lederhouse 1981;Borkin 1982) and early instars can be difficult to locate even if present on the host plant (Prysby and Oberhauser 2004). Therefore, when an egg or larva was missing from a host plant, the plant was revisited for three subsequent days to ensure that the larva had not simply been off the plant or overlooked. As in other studies on monarch survival (Zalucki and Kitching 1982;Zalucki et al. 1990), we considered eggs and larvae that were missing to be mortalities and the date of the mortality was recorded as the first day the egg or larva was missing from the host plant. Monarch larvae begin to emigrate from the host plant once they reach the third instar Fisher et al. 2020) and at that point we could not distinguish mortality from emigration. For this reason, our survival estimates are from egg to third instar.
It was not possible to distinguish the eggs of monarchs from the eggs of queens in the field (Calvert 1999;Batalden and Oberhauser 2015;; Scott 2019) so we initially had to combine all of the eggs for analyses of egg survival to the first instar. First instar queens can be distinguished from first instar monarchs by the presence of a pair of black dots corresponding to the third pair of tubercles present in queen larvae. This enabled us to document the survivorship from first to third instar separately for each species.
During daily visits to the host plants, all arthropods on the plant were documented. To do so, we approached each plant carefully and recorded any volant insects during that approach. More sedentary arthropods were noted upon closer inspection of the plant and while checking on the status of the egg or larva. In order to minimize disturbing the arthropods, we could not collect them. This technique clearly has limitations. First, it only provides a snapshot of the arthropods on a host plant. Second, many arthropods require close inspection of genitalia or other fine structures to make precise identification. Though some species were clearly identifiable, many arthropods could only be identified to family. Consequently, our interpretation of these data is highly generalized and, though the data provide information on arthropod community composition, our ability to correlate particular arthropods with survival is limited.
Fire ants have been implicated as having catastrophic effects on monarch eggs and larvae in Texas (Calvert , 2004 and this merited closer evaluation of the potential effects of fire ants. In addition to noting fire ants on the host plants, following the method of Hudman (2018) we quantified the density of fire ants adjacent to the host plant by counting all fire ant mounds within 4 m of the host plant on the day that an egg was first located. The purpose of this data was to determine if monarch or queen survival was correlated with fire ant mound density adjacent to the host plants.
While checking the status of eggs and larvae and while searching for new eggs on new host plants, we kept track of the number of non-focal larvae found on non-focal host plants and on the numbers of adult monarch and queen butterflies present on the study site. We also kept track of how many plants were examined to provide a crude measure of search effort. We used these data to generate a measure of the phenology of monarch and queen adults and larvae on our study area. For adults we compared our data on monarch phenology with data provided by Journey North (Sheehan and Weber-Grullon 2021). Data collected from the Journey North website included only fall monarch sightings recorded between 15 August and 15 November for both 2017 and 2018. We restricted the data to observations that occurred between 32° N and 34° N and between 94° W and 98° W, corresponding to an area 111 km north or south of the study site and 186 km east or west of the study site. We compared our data on the phenology of adult queens with that provided by Butterflies and Moths of North America (BAMONA) (Lotts and Naberhaus 2021). This data was much sparser than the Journey North Data. For that reason, and because the overall seasonal phenology of queens is so poorly documented in Texas, we evaluated BAMONA data that spanned the entire annual cycle for 20 years from 1 January 2000 through 31 December 2020. These data were stratified into three broad geographic regions in Texas, all east of 100 o W: South Texas (Mexican border and Gulf of Mexico to 28 o N), Central Texas (28 o N to 31 o N, 96° W to 100° W), North Texas (31 o N to Oklahoma border and 100 o W east to Louisiana border).
Weather conditions varied dramatically among the two years of the study. To quantify these differences, daily weather data on precipitation and temperature were obtained from a National Oceanic and Atmospheric Administration (NOAA) weather station located 2.3 km from the study site. Statistical analyses were conducted using SAS® Studio 3.8 software. For frequency data, we used Chi-square Contingency Tables. Where cells had expected values of less than five, Fisher's Exact Tests were used. Logistic Regression was used to analyze the effect of arthropods on egg and larval survival. This analysis used a step-wise procedure to isolate the best subset of models. AICc was then used to select the best model from that subset. For data on site occupancy by adults and non-focal larvae, two-sample t-tests were used to compare among species or years. Folded F-tests were used to determine if variances were equal. Where variances were unequal, Satterthwaite's method for unequal variances was used to estimate the test statistic, otherwise the pooled variance estimate was used to estimate the test statistic. For all statistical analyses we used p-values to assess the level of confidence associated with effect sizes (Murtaugh 2014).

Weather conditions
We present these data because markedly different weather conditions between the two years of the study may have had important impacts on the results. In particular, drought conditions during 2017 resulted in obvious water stress to the host plants that was not evident during the much wetter year of 2018.
Temperatures did not differ markedly from normal for either year, though August of 2018 was hotter than normal (Table 1). For our site, the average first day of frost is November 11. However, in 2017 the temperature fell below freezing on the mornings of 28 October and 29 October, two days after the last two eggs disappeared, four days after the last monarch larva reached the third instar, and 13 days after the last queen reached the third instar. The low temperature on each of these mornings was − 0.6 °C. In 2018, temperatures remained above freezing until 10 November when the morning temperature was − 1.7 °C. This was seven days after the last egg was missing, 13 days after the last monarch reached the third instar, and 14 days after the last queen reached the third instar.
Precipitation differed markedly among years (Table 1). The 30.3 cm of precipitation for August 2017 is misleading because that total reflects a single precipitation event on 13 August when 20.83 cm of rain fell. Otherwise, the fall of 2017 was extremely dry. Only 0.66 cm of rain fell between 19 August and 2 October and there were only nine days of precipitation over the course of the study. These dry conditions led to observable wilting and loss of condition of many plants, including some milkweed host plants.
In contrast, the fall of 2018 was extremely wet (Table 1). Precipitation in September was 167% above normal and precipitation in October was 43% above normal. In 2018, it rained on 28 of the 81 days of the study. Though there was ponding of water on many parts of the study area, there were no cases in which a mortality could be linked to a host plant immersed in water. The increased precipitation in 2018 resulted in a noticeably better condition of the host plants.

Survival of eggs and larvae
Over two years, 463 eggs were found on 414 plants; 231 eggs in 2017 and 232 eggs in 2018. The peak period of egglaying occurred in September. Despite marked differences in weather conditions, the density of eggs per plant checked varied little among the two years of the study. In September of 2017, an average of 36 potential host plants had to be examined to find one egg (0.028 eggs per plant). In September of 2018, an average of 31 potential host plants had to be searched to find one egg (0.032 plants per egg). Among host plants that held eggs, 49 harbored multiple eggs; 44 with two eggs, four with three eggs, and one with four eggs. The number of plants with multiple eggs was higher than expected by chance (Poisson Distribution, X 2 = 381.92, df = 2, p < 0.0001). Because we could not distinguish monarch eggs from queen eggs (Scott 2019), hatching success included both monarchs and queens. Across both years, 51% of eggs survived to hatch. However, hatching success differed among years (Fig. 1). In 2017, 44.6% of eggs hatched and the daily survival rate was 83%. In 2018, 67.2% of eggs hatched and the daily survival rate was 92%. In 2017, 11.7% of first instars survived to the third instar and in 2018, 26.3% of first instars survived to the third instar (Fig. 1). Overall, 5.2% of eggs survived to the third instar in 2017, whereas in 2018, 17.7% of eggs survived to third instar.
First instar monarchs and queens are extremely small and, in some cases, owing location on the plant, light conditions, and the need to avoid disturbing the larvae, it was not possible to immediately identify the larvae. Some of these unidentified individuals disappeared before they could be identified. Thirty-seven of the 103 eggs that hatched in 2017 and 12 of the 153 eggs that hatched in 2018 succumbed before they could be identified. Thus, we identified 56 first instar monarchs and 10 first instar queens in 2017 and 73 first instar monarchs and 68 first instar queens in 2018.
Among first instars that were identified, 19.6% of monarchs and 10% of queens survived to the third instar in 2017 whereas, in 2018, 20.5% of monarchs and 38.2% of queens survived to the third instar. These values are inflated because they exclude individuals that hatched but were not identified prior to their disappearance. However, the proportion of adults of each species observed in the field was highly concordant with the number of non-focal larvae of each species observed in the field (see phenology of monarchs and queens below). Based on the assumption that unidentified mortalities among first instars were also concordant with these proportions, we refined our estimates of survival for each species. In 2017, 80% of adults and non-focal larvae were monarchs, so we assigned 80% (n = 30) of the 37 Fig. 1 Percent of all eggs that hatched or survived from first to third instar. These data combine monarchs and queens. In 2017, 44.6% of eggs hatched. In 2018, 67.2% of eggs hatched, 51% more than had hatched in 2017 (Chi-square 2 × 2 Contingency Table; X 2 = 16.25, df = 1, p < 0.0001). Survival from hatching to third instar was 11.7% in 2017. Survival from hatching to third instar in 2018 was 26.3%, 125% higher than it was in 2017 (Chi-square 2 × 2 Contingency Table; X 2 = 16.13, df = 1, p < 0.0001). Numbers over bars indicate the total number of eggs or first instars observed in each year unidentified larvae as monarchs and the remainder (n = 7) as queens. In 2018, 50% of adults and non-focal larvae were monarchs, so half of the 12 unidentified larvae (n = 6) were assigned as monarchs and the other half as queens. These values were used to estimate the survival of each species from hatching to third instar (Fig. 2). In 2017, monarch survivorship from hatching to third instar was 12.8% and 5.9% for queens. Odds ratio estimates for these data indicate that a monarch first instar was 2.3 times more likely to survive than a queen egg (Table 2). In 2018, monarch survival from hatching to third instar was 19%, and 36.5% for queens (Fig. 2). In that year, a monarch first instar was only about 0.4 times as likely to survive as a queen first instar (Table 2). Survival tended to be higher for both species in 2018 though this was less evident for monarchs than it was for queens (Fig. 2, Table 2), Combining these estimates with the hatching success data, in 2017, we estimate survival from egg to third instar was 5.7% for monarchs and 2.6% for queens. In 2018, survival from egg to third instar was 12.8% for monarchs and 24.5% for queens.

Host plant arthropods
Other arthropods on the host plants were documented to understand the frequency and abundance of predatory and non-predatory arthropods associated with monarch and queen eggs and larvae. Small sample sizes required that we combine data for both monarchs and queens. We also had to correct inherent biases in the data. Individuals that survived to the third instar were monitored longer than individuals that did not survive and this difference inflates the observed number and abundance of arthropods on the host plants. To correct this, we removed from the analyses individuals monitored for fewer than eight days. This left 105 eggs on 100 host plants; of these, 38 survived to the third instar and 67 died. The mean and variance for the length of time that individuals were monitored was comparable for eggs that survived and those that died (Mean ± SE, survived = 10.92 ± 0.51, died = 10.64 ± 0.38, F-Test for Equal Variances, F = 102, df = 66, 37, p = 0.9802; t-test for means, t = 0.44, df = 103, p = 0.6621). This approach standardized the mean and variance and created more balanced samples. However, among eggs that died, the analyses are restricted to individuals who survived at least eight days and our conclusions are made with this limitation in mind.
Arthropods associated with the 105 eggs included 2192 individuals of 39 taxa (Online Appendix 1). Fourteen taxa were predatory whereas the remaining 25 were primarily non-predatory. The most abundant arthropods were oleander aphids, mainly because they formed large aggregations on some host plants. Because of this, the actual frequency of oleander aphids was very low (Online Appendix 1). Most other arthropod taxa had low abundance. With the exclusion of oleander aphids, the average host plant only harbored 2.9 ± 0.58 (SE) arthropods, of which 1.4 ± 0.36 (SE) were predators and 1.6 ± 0.36 (SE) were non-predators. The most frequent taxon, and the most frequent predator, was jumping spiders, associated with 14% of host plants (Online Appendix 1). Thirty-seven of the 39 taxa had frequencies of less than 10% and a third of the host plants had no arthropods documented on them at all.
Because frequencies and abundances were low, it was difficult to isolate specific taxa that had an impact on monarch or queen survival. Furthermore, sparse data can lead to inflated parameter estimates when using logistic regression (Greenland et al. 2016). To avoid this, we combined some taxa into broader, more inclusive categories. The categories included jumping spiders, grasshoppers, fire ants, and oleander aphids. The remaining taxa were combined into two groups; all other predatory taxa and all other non-predatory taxa. Logistic regression generated four predictive models, none of which were particularly satisfactory ( Table 3). The best model contained four arthropod groups: fire ants, jumping spiders, all other non-predatory arthropods, and oleander aphids. All of these groups, except all other non-predatory arthropods, were negatively associated with monarch survival. The largest effects were negative impacts of fire ants and jumping spiders on monarch and queen survival (Table 3).
Other studies have implicated fire ants as having strong negative impacts on monarch egg and larval survival (Calvert , 2004, so we evaluated survival relative to the number of fire ant mounds around each host plant. We used the data from 462 eggs for which fire ant mound density was measured. On average there were 0.74 ± 0.05 (SE) fire ant mounds within 4 m of the host plants, representing a density of 147 fire ant mounds per hectare. Slightly more mounds were observed in 2018 than in 2017 (Kruskal-Wallis Test, Chi-square Approximation, X 2 = 3.84, df = 1, p = 0.0502) ( Table 4). Since we could not differentiate monarch and queen eggs, we conducted separate analyses for all eggs and for those monarch and queen larvae that were identified at the first instar (Table 4). Though none of the comparisons had p-values less than 0.05, a couple trends are evident. In 2017, more than four times as many fire ant mounds were adjacent to host plants with mortalities as compared to host plants without mortalities. For monarch larvae in both years combined, almost twice as many fire ant mounds were adjacent to host plants with mortalities as compared to host plants without mortalities. These trends were not evident for eggs in 2018 or for queen larvae in both years combined.

Phenology of monarchs and queens
Data were collected on site occupancy by monarch and queen adults and larvae while searching potential host plants for eggs. For these analyses, days when fewer than ten plants Table 3 Stepwise logistic regression analysis of the effects of arthropod groups on the survival of fall monarch and queen eggs or larvae Stepwise selection was used to generate the best subset of models using a p-value of 0.50 for entry or removal of a parameter from the model. Models are sorted in order of ascending AICc. Best model is based on minimum corrected AIC score (AICc), w i is the Akaike weight of each model were searched were excluded to ensure sufficient effort was represented. The number of adults or non-focal larvae observed was divided by the number of plants searched to generate a crude value of individuals per unit effort. There were over four times as many adult monarchs observed in August of 2017 as were observed in August of 2018 (Fig. 3). The average date of site occupancy was 6 September in 2017 and 15 September in 2018. However, dates in September when large flights of monarchs occurred were similar for both years though some flights in 2018 were larger than any flights in 2017 (Fig. 3). In 2017 large flights of monarchs occurred on 10, 17 and 19 September. In 2018, there were large flights of monarchs on 11 and 20 September (Fig. 3). These dates are about three to four weeks earlier than the average dates for monarch observations recorded by Journey North (Fig. 4) The scales on the vertical axes are fixed to emphasize differences in abundance between years. In 2017, the average date of occupancy by monarchs was 12 days earlier than that of queens (t-test, Satterthwaite's method for unequal variances; t = 3.06, df = 35.8, p = 0.0042). In 2018, the average date of site occupancy by monarchs was three days earlier than that of queens (t-test, Pooled variance estimate; t = 1.12, df = 112, p = 0.2692). In 2017, 80% of the adults were monarchs (t-test, Satterthwaite's method for unequal variances; t = 3.04, df = 68.9, p = 0.0033). In 2018, across the entire season, there were approximately equal numbers of monarchs and queens (t-test, Pooled variance estimate; t = 0.25, df = 112, p = 0.8067) The pattern is consistent with a seasonal colonization of Texas starting in south Texas in winter and early spring and moving northward in two successive generations that peak in July and October in central and north Texas by the end of September of 2018, adult queens slightly outnumbered adult monarchs. In 2017, 80% of adults observed on the study site were monarchs. In 2018, monarchs represented only 50% of the adults observed on the site. Because of large numbers of adult monarchs in August of 2017, the average date of site occupancy was twelve days earlier for monarchs than it was for queens. In 2018 the average date of site occupancy by monarchs was only three days earlier than it was for queens.
Queens are poorly studied in Texas and we wanted to know where the queens on our study site may have originated. Data on the annual phenology of queens for Texas were obtained from the Butterflies and Moths of North America (BAMONA) website. There were clear geographic and temporal patterns to queen occupancy in the three regions of Texas (Fig. 5). In south Texas, queens were mostly observed from January through March and again in October and November. No queens were recorded in south Texas for the months of July through September. In central and north Texas, peaks occurred in July and October. In these two regions, there were no observations of queens from January through March (Fig. 5). These data are consistent with an annual colonization of Texas that starts in January in south Texas and along the gulf coast, continues northward in April, and peaks in July and October in central and north Texas. At our study site, adult queen abundance was higher in September than it was in October. However, the eggs laid by our queens in September would mature and eclose sometime in October and the large numbers observed in the BAMONA database in October at our latitude might represent newly emerged adults from this generation.
Data on the phenology of monarch and queen non-focal larvae were combined for all instars observed while searching for new eggs on new host plants. In 2017, the phenology of monarch and queen non-focal larvae was similar to that of the adults (Fig. 6) and monarch larvae were much more common than queen larvae. There was a peak in both monarch and queen larvae in mid-to late September. In 2017, the average date for monarch larvae was 12 September, six days earlier than the average date for queen larvae. In 2018, the numbers of both monarch and queen larvae increased as the season progressed. In that year, the average date for monarch larvae was 28 September, two days earlier than for queen larvae. In 2017, 82% of larvae observed while checking potential host plants were monarchs. In 2018, monarchs represented 44% of the larvae observed in the field. These percentages are similar to the percentages of adults observed in the field and, like the adults, the average dates of monarch larvae on the study site were well in advance of the average passage date of adults recorded by Journey North.

Discussion
Our estimates of the survival of fall-breeding monarchs and queens from egg to third instar varied among the two years of the study; for monarchs, success was over two times higher in 2018 than it was in 2017. For queens, success was almost ten times higher in 2018 than it was in 2017. We attribute these differences among years to the very dry conditions of 2017 and the much wetter conditions of 2018. In 2017, many host plants exhibited visible signs of water stress in the form of wilting, reduced growth, and die-back of some ramets. We suspect that these effects of precipitation on host plant quality had an important impact on monarch and queen survivorship. Clearly, as suggested by others (Prysby and Oberhauser 2004;Majewska and Altizer 2019), more data are needed on the effects of host plant quality on fall reproduction in monarchs and queens. Fig. 6 Study site occupancy by non-focal monarch and queen larvae in 2017 (upper graph) and 2018 (lower graph). The vertical axes are fixed to emphasize differences in abundance between years. In 2017, the average date for monarch larvae was six days earlier than it was for queen larvae (t-test, Satterthwaite's method for unequal variances; t = 1.79, df = 39.9, p = 0.0816). In 2018, the average date for monarch larvae was only two days earlier than it was for queen larvae (t-test, Pooled variance estimate; t = 0.65, df = 92, p = 0.5191). In 2017, 82% of the larvae were monarchs (t-test, Satterthwaite's method for unequal variances; t = 3.40, df = 86.13, p = 0.0010). In 2018, across the entire season, there were approximately equal numbers of monarch larvae and queen larvae (t-test, Satterthwaite's method for unequal variances; t = 0.88, df = 96.42, p = 0.3807) For monarchs, the 12.8% survivorship to third instar estimated for 2018 was comparable to estimates reported for spring generation monarchs in Texas (Stevenson et al. 2021) and in Florida (Cohen and Brower 1982;Zalucki and Brower 1992;Brower et al. 2018). Even in 2017, our estimated survivorship of 5.7% was higher than rates reported for studies reporting survivorship to third instar using similar protocols in summer in Minnesota (De Anda and Oberhauser 2015), Wisconsin (Borkin 1982;Prysby 2004), and Michigan (Haan and Landis 2019). On the other hand, using different protocols, Nail et al. (2015) found that monarch survival was similar for spring, summer, and fall generations. Collectively these comparisons suggest that fall reproduction by monarchs, based on survival estimates, has the potential to be as productive as earlier generations further north. However, our egg density was between 25% and 35% lower than egg densities recorded in the midwestern U.S. in summer (Nail et al. 2015). Likewise, the lower larval densities for fall monarchs documented on a broader regional scale (Tracy et al. 2022) are probably also a result of lower egg densities rather than low survivorship. Though we feel that our study site represents a typical old-field in north Texas, monarch survival can vary considerably among sites within a geographic region. For example, spring survival in north Texas and northwestern Louisiana from egg to third instar varies from 0 to 40% among sites (Lynch and Martin 1993). Also, our study is limited to survival from egg to third instar. Spring and summer survival of older instars is higher than that of younger instars (Zalucki and Kitching 1982;Lynch and Martin 1993;Zalucki et al. 2001;Oberhauser et al. 2001;Prysby and Oberhauser 2004), but we have no comparable data on the fall survival of older instars. High levels of parasitism may occur among fall-generation monarchs in the south (Mueller and Baum 2014), so the assumption of higher survival of older instars needs further investigation for fall-breeding monarchs. Data are needed covering a broader geographic area, multiple land use types, and older instars to consolidate the assessment of fall survival and productivity of monarch eggs and larvae.
Arthropod predators are important determinants of monarch egg and larval survival (Prysby 2004;Koch et al. 2005;De Anda and Oberhauser 2015;Oberhauser et al. 2015;McCoshum et al. 2016;Hermann et al. 2019;Myers et al. 2019Myers et al. , 2020Nestle et al. 2020;Baker and Potter 2020) and this applies to queens and other lepidoptera as well (Zalucki et al. 2002). The milkweed plants on our study site harbored 40 arthropod taxa of which 14 were predatory and 25 were non-predatory arthropods. However, the number of arthropods per host plant was small, making it difficult to isolate specific taxa of importance in affecting survival.
Three taxa seemed potentially important: oleander aphids, jumping spiders, and fire ants. Aphids are known to attract predators onto host plants (Prysby 2004), but in our study aphids caused extensive tissue damage to the host plants. This may directly affect monarch survival or force younger instars to leave the plant which, prior to the third instar, would likely be fatal (Zalucki and Kitching 1982;Zalucki et al. 2002;De Anda and Oberhauser 2015). Jumping spiders and other spiders were the most frequent predators and occurred on one in five host plants. Spiders are important predators on monarch eggs and larvae (Smithers 1973;Zalucki and Kitching 1982;Lynch and Martin 1993;De Anda and Oberhauser 2015;Hermann et al. 2019;Myers et al. 2020;Stevenson et al. 2021) and the stout growth form of A. viridis represents an attractive location for spiders to place webs and harborages (Vasconcellos-Neto et al. 2017). Fire ants may be the most important predator of monarch eggs and larvae in the southern U.S. (Calvert , 2004. On our study site the mound density was at the lower end of the average range typical of North American populations (Macom and Porter 1996;Porter et al. 1997) and fire ants seemed most important during the very dry conditions of 2017. This could be due to the effect of weather conditions on food resources and the resulting changes in foraging strategies by the ants (Stein et al. 1990;Cassill and Tschinkel 1999;Vogt et al. 2002). We suggest that future studies of fall reproduction pay particular attention to the importance of the three taxa cited above.
Queen butterflies in Texas are poorly studied. In our study, the occurrence of breeding queens in fall corresponded well with that of breeding monarchs, especially relative to peaks in September. Queens exhibit southward migratory movements in south Texas and Mexico in the fall (Einem 2003). However, northward movements of queens have not been studied. We found that for most of Texas east of the 100th meridian, there are two peaks in the abundance of queens. In south Texas, these peaks occur in January-February and again in October-November. In central and north Texas, these peaks occurred in July and October. Though we can't be certain, this pattern suggests an annual northward colonization of Texas followed by a southward migration in fall. Analyses of queens migrating southward through northern Mexico in October and November show that these migrants originate from north Texas and the southeastern U.S. (Hobson et al. 2021). Collectively, the phenology of queens across Texas suggests that queens engage in migratory movements that are similar, albeit on a much smaller scale, to that of monarchs. However, because queens are a mainly tropical and subtropical species, they may be poorly adapted to weather fluctuations at the northern periphery of their summer and fall distribution. This would explain why queens on our study site were far less abundant and had very low survivorship in the very dry fall of 2017 as compared to the much wetter fall of 2018. Variations in weather conditions and their effect on southern generations might also explain why queens frequently, but unpredictably, show up far to the north in the central plains (Lotts and Naberhaus 2021) and northeastern U.S. (Moskowitz 2003).
Monarchs and queens utilize the same host plants and can compete for access to these plants. In Florida, queens are displaced from ovipositing on preferred food plants by the presence of large numbers of monarchs (Brower 1961) and, reciprocally, monarchs are displaced from ovipositing on preferred food plants by the presence of large numbers of queens (Brower 1962). In our study it was not overtly evident that the two species were competing and, in both years, there were over 30 vacant host plants for every egg found. However, we found multiple eggs on host plants in spite of those vacant plants implying that females are selective either for microhabitat or host plant quality. Such a situation makes competition more likely. Because they are a broad-ranging, highly mobile species, queens have a potential to expand into North America in response to climate change as noted for tropical and subtropical lepidoptera in general (Kocsis and Hufnagel 2011). Furthermore, queens are less susceptible than monarchs to the protozoan ectoparasite Ophryocystis elektroscirrha (OE) (Barriga et al. 2016). If climate change and parasite resistance favor queens, then fall reproduction by monarchs could be negatively impacted in the future.
The phenology of fall reproduction is important as it affects the survival of the resulting generation. At our latitude the average first frost occurs on November 11th, though freezing temperatures occasionally occur earlier. The peak in reproduction of monarchs and queens from September to mid-October in Texas provides sufficient time for the majority of the resulting adults to eclose before the advent of freezing weather (Baum and Sharber 2012). An earlier study found that most monarchs that eclose in fall in Texas are in diapause and show migratory behaviors (Batalden and Oberhauser 2015) and stable isotope data show that 11% to 30% of overwintering monarchs in Mexico originate from the southern great plains of the U.S. (Flockhart et al. 2017). Collectively these data and the results of our study indicate that fall reproduction is an important component of the annual cycle of eastern monarch populations.
For monarchs, the temporal displacement between the peak in fall reproduction and the peak period of migrants observed in our study is consistent with other studies on fall breeding in Texas and Oklahoma (Prysby and Oberhauser 2004;Baum and Sharber 2012;Batalden and Oberhauser 2015) and supports the prevailing hypothesis that fall-breeding monarchs in the southern U.S. are early migrants that originated from adults that eclosed further north (Calvert 1999;Zalucki and Rochester 1999;Borland et al. 2004;Batalden and Oberhauser 2015;Satterfield et al. 2018; Majewska and Altizer 2019). However, in 2017, monarchs were observed in small numbers all summer in north Texas (JGK Pers. Obs.). On our study site in that year there were large numbers of reproductive monarchs in mid-August and the source of these individuals is uncertain. Winter temperatures at this latitude are too cold for monarchs to overwinter (Zalucki and Rochester 1999;Howard et al. 2010), so the summer monarchs might have hatched in the area in spring and failed to migrate or they could be individuals that diffused northward from the resident populations along the coast of the Gulf of Mexico, 400 km further south. Resident populations of monarch butterflies along the gulf coast of Texas appear to be a relatively recent phenomenon (Howard et al. 2010;Majewska and Altizer 2019) and the dynamics of these populations are largely unknown and need further study. It is possible that some of our fall-breeding individuals were not migrants from further north.
In general, fall reproduction in monarchs is believed to be caused either by early migrants which never entered diapause or when early migrants come out of diapause upon reaching the southern U.S. Among adults captured in fall in Minnesota and Wisconsin, only 3% of females had mated whereas among females captured in fall in Texas, over 17% had mated (Borland et al. 2004). The presence of viable populations of milkweed at these lower latitudes has been implicated as causing adults to oviposit on the available milkweed (Borland et al. 2004;Prysby and Oberhauser 2004;Baum and Sharber 2012). Fall breeding is also suggested to be exacerbated by the cultivation of tropical milkweed (A. curassavica) in gardens and landscaping plots. In Texas, fall egg densities were much higher in gardens planted with A. curassavica than in plots with native milkweeds (Prysby and Oberhauser 2004) and monarchs, when given a choice, prefer to oviposit on A. curassavica rather than on the native milkweeds A. asperula and A. oenotheroides (Batalden and Oberhauser 2015). In Georgia, females in fall were more likely to exhibit egg development when exposed to A. curassavica than when exposed to A. incarnata (Majewska and Altizer 2019). It has been proposed that the widespread cultivation of A. curassavica might be disrupting migratory patterns of monarchs (Malcolm 2018;Clement and Crawford 2020) and increasing their exposure to parasites (Satterfield et al. 2015;Malcolm 2018).
However, fall reproduction is also likely to be a normal part of the phenotypic plasticity of monarchs. Native milkweeds are widely available in late summer and fall in the south-central U.S. (Tracy et al. 2022) and these plants are extensively utilized for fall breeding by monarchs (Calvert 1999;Tracy et al. 2022). Mowing (Baum and Mueller 2015;Knight et al. 2019;Dee and Baum 2019) and summer burns (Baum and Sharber 2012) cause A. viridis to sprout new growth. At our study site, mowing in late July generated new growth through September and into October, resulting in a high milkweed density compared to studies conducted in other regions (Pleasants 2017;Kaul and Wilsey 2019). The studies cited in the preceding paragraph that were conducted in Texas and Georgia on milkweed preferences found that native milkweeds also stimulated fall reproduction. In the Texas study (Batalden and Oberhauser 2015), male monarchs were as likely to exhibit mating behavior when caged with the native milkweed A. oenotheroides as they were when caged with A. curassavica. In the Georgia study (Majewska and Altizer 2019), males exposed to A. incarnata were more likely to be reproductive than those exposed to A. curassavica and females exposed to A. incarnata were only slightly less likely to show reproductive behavior than were females exposed to A. curassavica. Majewska and Altizer (2019) suggested that a lower affinity of monarchs for A. incarnata was due to the low cardenolide concentrations found in that species as compared to A. curassavica. The Georgia study needs to be replicated using A. viridis because, unlike the other native species mentioned above, the cardenolide content and the cardenolide polarity profiles of A. viridis and A. curassavica are very similar (Rasmann and Agrawal 2011). Furthermore, A. viridis is abundant in Texas and Oklahoma in the fall (Baum and Sharber 2012;Dee and Baum 2019;Tracy et al. 2022), and at the landscape scale much more abundant than A. curassavica (Calvert 1999). Our study site contained only A. viridis and it seems unlikely that exposure to A. curassavica was an important factor in stimulating reproduction on our study site. It is more plausible that fall reproduction in the presence of healthy populations of A. viridis is a normal response for monarchs (Freedman et al. 2018;Green 2021).
While the cultivation of exotic milkweeds should be discouraged, the importance of fall monarch reproduction on native milkweeds should not be ignored. Native milkweeds, most importantly A. viridis, are widely available in the southern plains in late summer and fall (Calvert 1999;Tracy et al. 2022). As such, fall reproduction is a normal aspect of the species' biology and likely represents an adaptive trade-off between the benefits of current reproduction and those of continuing migration to Mexico. Most importantly, fall reproduction could represent an important buffer against reproductive failures in earlier generations (Flockhart et al. 2017). For these reasons, land management for successful fall breeding in the southern U.S. should be an important aspect of monarch recovery plans.

Implications for conservation
Fall reproduction by monarchs is an important yet neglected component for the management of the eastern monarch butterfly particularly given the potential for the offspring produced through fall reproduction to contribute to wintering populations in Mexico. Fall reproduction needs greater emphasis in monarch recovery plans. Land managers in the southern U.S. wanting to contribute to increased monarch recruitment need to coordinate prescribed burns and mowing schedules to maximize production of healthy A. viridis and other native milkweed plants in the fall. In Texas, mowing or burning should be conducted in July and should be avoided in September and October when monarch eggs and larvae occupy host plants.
Queen butterflies showed variable survival in north Texas. However, the characteristics of this species make it likely to expand its distribution northward in the U.S. in response to climate change. Because queens could displace monarchs from favorable host plants, monitoring of this species and its populations is essential to understand the impact of the queen on the future dynamics of fall reproduction in monarchs.

Consent for publication Not applicable.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.