Influence of Aridity and Weather on Breeding-Season Survival of Northern Bobwhites in South Texas, USA

Weather plays a major role in northern bobwhite Colinus virginianus population dynamics in arid environments (Guthery et al. 2002, 2005; Hernández et al. 2005; Tri et al. 2013; Parent et al. 2016). Northern bobwhites (hereafter, bobwhite) exhibit boom-and-bust population dynamics that correlate with weather (Stoddard 1931; Rosene 1969; Lehmann 1984; Hernández et al. 2005). During years of average or above-average precipitation, bobwhites are abundant in South Texas (Bridges et al. 2001; Lusk et al. 2002). However, weather does not affect all aspects of population dynamics equally. Vital rates such as survival may be affected by aridity and temperature (Hernández et al. 2005), but production is influenced by precipitation during spring and summer (Kiel 1976; Guthery et al. 2000, 2002; Perez et al. 2002; Tri et al. 2013), in an interactive manner.

Precipitation and temperature have strong impacts on bobwhite demographics in South Texas (Guthery et al. 1988; Hernández et al. 2005; Tri et al. 2013). Abundant rains can partially offset negative effects of hot summers on bobwhite production (Guthery et al. 2002). Survival, production, and number of nests initiated during breeding season are greater during relatively wet periods than they are during dry periods (Hernández et al. 2005). In addition, breeding season (April–August) cumulative precipitation explains 94% of variation in bobwhite age ratios in the harvest (a proxy for reproductive output) across South Texas (Kiel 1976; Tri et al. 2013).

Wildlife managers and biologists believe that weather drives bobwhite population dynamics. Although the link between weather and bobwhite production is well-established (Stoddard 1931; Kiel 1976; Lehmann 1984; Guthery et al. 1988; Tri et al. 2013), the empirical link between weather and adult survival of bobwhites is not clear (Guthery et al. 2000; Hernández et al. 2005). It is important to understand this potential link because survival during the breeding season (summer) is the second most important demographic parameter of bobwhite population change (Sandercock et al. 2008). Our objective was to determine the relative influence of abiotic factors (weather and aridity) on breeding season (April–August) survival at the landscape scale in South Texas (Figure 1). We conducted a longitudinal analysis of four bobwhite populations to assess these impacts temporally. We deduced that environmental factors (e.g., temperature, precipitation, drought severity, and growing season length) should have a strong impact on survival (Guthery et al. 1998; Hernández et al. 2005). We hypothesized that drought severity would be an important variable in predicting breeding season survival of bobwhites because this measure incorporates both temperature and precipitation.

Our study areas were four properties in the South Texas Plains ecoregion (Gould 1975): 1) a private ranch in Brooks County (9,100 ha; Ranch One); 2) a private ranch in Duval County (1,020 ha; Ranch Two); 3) a private hunting lease on the Encino division of King Ranch (4,585 ha; Ranch Three); and 4) Chaparral Wildlife Management Area in Dimmit and LaSalle Counties (6,154 ha; Ranch Four). We sampled bobwhites on Ranches One and Two from 2008 to 2009 (Buelow 2009; Tri 2010), Ranch Three from 2001 to 2005 (DeMaso 2008), and Ranch Four from 2005 to 2006 (Sands 2007; Table 1). Ranches were separated by ≥10 km (range = 10–136 km; Figure 1) along an aridity gradient (Guthery et al. 1988; 2001). Western areas of the gradient were more arid than eastern areas (Guthery et al. 1988). Plant communities at each site were predominantly mixed-brush communities characteristic of the South Texas Plains (Gould 1975; McLendon 1991). The primary wildlife management priority for the private ranches was producing wild bobwhites for recreational hunting. The Chaparral Wildlife Management Area was managed by the Texas Parks and Wildlife Department for nonconsumptive wildlife activities such as birding, limited public hunting, and wildlife research. Strip-disking, mechanical brush control, grazing, and prescribed burning were used to manage vegetation on all ranches. Ranches One and Two used broadcast supplemental feeding during the breeding season. Ranches Three and Four did not use supplemental feed. Detailed descriptions of study site vegetation and soils are found in Sands (2007), DeMaso (2008), Buelow (2009), and Tri (2010).

We trapped bobwhites using baited funnel traps from February to July during 2000–2006 and 2008–2009. We trapped no bobwhites during 2007. We marked bobwhites with a mass ≥150 g with radiotransmitters (6 g, necklace; Advanced Telemetry Systems, Isanti, MN, and American Wildlife Enterprises, Monticello, FL) and relocated each bird one or three times per week during the reproductive (nesting and brood-rearing) period (April–August). We removed relocations of radiomarked bobwhites that died <1 wk after capture so as to reduce effects of capture on mortality. We also removed data from any bobwhite with collars that did not function for the entire season. We captured, handled, and marked bobwhites within guidelines of Texas A&M University–Kingsville's Animal Care and Use Committee (Permit numbers 2003-3-3 and 2008-09-30A).

We used weather station data (mean maximum daily temperature averaged from April to August and total precipitation from April to August of each year) from the National Oceanic and Atmospheric Administration weather station nearest to each ranch (National Climatic Data Center 2015). We used data from National Oceanic and Atmospheric Administration weather station 413063 (Falfurrias, Brooks County, Texas; 27°13′37″N, 98°8′39″W) for Ranches One and Three, station 414058 (Hebbronville, Jim Hogg County, Texas; 27°18′24″N, 98°40′42″W for Ranch Two, and station 412048 (Cotulla, La Salle County, Texas; 28°29′10″N, 99°13′20″W for Ranch Four). We obtained monthly Keetch–Byram Drought Index (KBDI) values on each ranch (Texas Weather Connection 2015). The KBDI is an index used to calculate wildfire risk; it calculates a daily water balance that quantifies the extent to which drought is influenced by temperature, precipitation, and soil moisture (Keetch and Byram 1968). Fully saturated soils have a value of 0 and completely dry soils have a value of 800. We chose KBDI in favor of other drought indices (such as the Palmer Modified Drought Index) because KBDI was derived by remote sensing and can be sampled on a much finer scale (4 km × 4 km). We determined growing season length (number of days from the start to end of detectable photosynthetic activity) using the Moderate Resolution Imaging Spectroradiometer Normalized Difference Vegetation Index data set at 250-m resolution (United States Geological Survey 2015). This variable allowed us to account for differences in overall photosynthetic activity (e.g., bunchgrass productivity) across the landscape (Groten and Ocatre 2002). The resolution of the Normalized Difference Vegetation Index data was 250 m × 250 m; therefore, multiple measurements of growing season length were available for a given ranch within each breeding season. We used the map algebra tool in ArcGIS 10.2 (ESRI, Inc., Redlands, CA) to derive a mean value for each ranch from the constituent pixels in the growing-season-length layer.

We collected survival data on 580 bobwhites, but removed data from 66 because of radiotransmitter failure (Data S1, Supplemental Material). We assessed survival of 514 bobwhites during the breeding seasons (April–August) of 2001–2006 and 2008–2009 using a Generalized Linear Mixed Model in Program R (R Core Team 2015; http://www.R-project.org). If a bobwhite survived from April to August, we considered that bird a success (“1”); if a bobwhite did not survive the entire period, we considered that bird a failure (“0”). Sample sizes ranged widely among breeding seasons and ranches, but in all except two year–ranch combinations, sample size was ≥31 bobwhites per season (Table 1). Bobwhite survival often differs between males and females, as well as between juveniles and adults (Burger et al. 1998; Cox et al. 2004; Terhune et al. 2007).

We calculated descriptive statistics (mean, range, and CV) for each input variable. We developed a set of a priori candidate models using Generalized Linear Mixed Models using demographic variable(s) from the previous analysis and all possible linear combinations of weather variables. We conducted a Pearson's correlation analysis among all variables in the data set prior to fitting models. We removed any models that contained collinear predictors (r >0.50) but did fit all other possible linear combinations of our predictor variables. We fit all models with important demographic variable(s) as fixed effects and fit ranch as a random effect. We fit ranch as a random factor to account for site-specific variation in survival due to factors we did not measure (e.g., habitat structure, habitat quality, and predator density) among ranches as well as the lack of independence among years. We built nested subsets of the full logistic Generalized Linear Mixed Model (Survival = Sex + Age + Sex × Age + 1|Ranch) and ranked them with Akaike's Information Coefficient (AIC). If multiple models had similar support, we used likelihood ratio tests to determine if all model parameters were important or if some were spurious and uninformative (Arnold 2010). We also fit all possible two-way interactions. We used an information-theoretic approach to determine relative level of support for models in our candidate set (Burnham and Anderson 2002). Our cutoff value for all marginal P-values in the regression was α = 0.05. Ranch One served as the reference level in the regression output. We calculated two pseudo-R² statistics for each model to assess how much variation was explained (Nakagawa and Schielzeth 2013).The marginal R² is the proportion of variance explained by the fixed factor(s) alone; the conditional R² is the proportion of variance explained by both the fixed and random factors. We also conducted a deviance-based, goodness-of-fit test on our best model to assess how well our model fit the data (Agresti 1990).

Descriptive statistics indicated that the precipitation contained the greatest amount of variation among the predictor variables (Table 2). We found strong, positive correlation (r = 0.97) between length of the growing-season and mean maximum temperature, so we removed the mean maximum temperature from models in which the pair occurred. We found no evidence of simple collinearity among any predictor variables (all other pairwise r-values were <0.50). Based on likelihood ratio tests, bobwhite age at capture was not related to survival during the breeding season in our analysis, but sex was a useful predictor (χ² = 4.00, df = 1, P = 0.05). There was no support for an interaction between age and sex (χ² = 0.76, df = 1, P = 0.38) nor age in the model (χ² = 0.55, df = 1, P = 0.46).

Our best-supported model was (Survival = Sex + KBDI + Length of the growing-season + Ranch), which explained 35.3% of the variation in the data (Table 3). We found no evidence for a lack of fit (χ² = 545.41, df = 509, P = 0.73). In addition to sex, both weather variables (GROWSSN and KBDI) were important terms in the model (Table 4). Our next best model was (Survival = Sex + KBDI + GROWSSN + KBDI × GROWSSN + Ranch). It contained ∼25% of the AIC weight; however, additional analyses indicated that the interaction variable was uninformative (χ² = 0.80, df = 1, P = 0.37; Arnold 2010).

Bobwhite survival probability was greatest when conditions were wet and growing season was long. Odds of a bobwhite dying during the breeding season increased 10% for every 100-unit increase in KDBI, after accounting for all other variables (Table 4). Odds of a bobwhite surviving the breeding season increased 1% for every 10-d increase in GROWSSN; odds of surviving the entire breeding season were 83% greater for males than females (Table 4). Trends of KBDI on survival probability were similar among ranches but overall survival was greater on Ranch Three (Figure 2). On each ranch, probability of survival was greater for males than females in relation to KBDI. The downward trend on survival in relation to KBDI value was intuitive—as drought conditions increased, probability of survival decreased. The trend of growing season length on survival probability was also similar among ranches, but overall survival probability was greatest on Ranch Three (Figure 3). Effect of growing season length on probability of survival was statistically similar between sexes (Figure 3). We found no evidence that provision of supplemental feed influenced survival because survival rates were similar for Ranches One, Two, and Four. Ranches One and Two had supplemental feed; Ranches Three and Four did not. Survival was greater on Ranch Three than all others (Figures 2 and 3).

Most managers and biologists in South Texas understand that weather is related to bobwhite population dynamics (Forrester et al. 1998; Guthery et al. 2005). Many assume that survival during the breeding season has a strong link with weather, but that assumption is for the most part untested. We understand the relationship between breeding season weather and reproduction, but we know of the relationship between adult survival and weather at large scale (Guthery et al. 1998). Because temperatures in much of Texas frequently cause thermal stress and can reach lethal limits for bobwhites, we expected that there would be a strong relationship between breeding season survival and weather (Forrester et al. 1998; Guthery et al. 2005). We failed to find such a strong relationship—our data support only a relatively weak relationship between weather variables and adult bobwhite survival in South Texas.

The link between drought conditions and reduced production in bobwhites has long been established (Leopold and Ball 1931; Errington 1935); however, few empirical studies have assessed links between survival and drought. Drought can reduce game bird reproduction (Tri et al. 2013) and abundance (Hernández et al. 2005). This pattern is reflected among the demographics of many other North American galliforms such as wild turkey Meleagris gallopavo (Schwertner et al. 2007), scaled quail Callipepla squamata (Lusk et al. 2007), ring-necked pheasant Phasianus colchicus (Randell 2009), and sharp-tailed grouse Tympanuchus phasianellus (Flanders-Wanner et al. 2004).

Drought affects bobwhite survival indirectly through insect abundance and vegetation cover (Roseberry 1989). Bobwhites can survive solely on metabolic and preformed water during drought (Guthery 1999; however, drought can reduce available preformed water through reduced insect abundance (Roseberry and Klimstra 1984) and drier, dormant grassland plants in those conditions (Paulson and Reid 1970). Others suggest that drought causes a reduction in the quality and quantity of foods (vegetation and insect) as well as cover, and thus increases stress and mortality risk of bobwhites (Stoddard 1931; Haugen 1955; Reid and Goodrun 1960). Our results indicated that drought was related to bobwhite survival after accounting for growing season length, sex of the bobwhite, and site-specific ranch effects. Survival decreased as conditions became drier and hotter.

Composite variables (such as KBDI and GROWSSN) are often better predictors of demographic parameters than single precipitation and temperature variables (Bridges et al. 2001; Lusk et al. 2007). Drought severity was a useful predictor of bobwhite survival during the breeding season after accounting for growing season length, sex of bobwhites, and site-specific effects of each study site (ranch). This is because KBDI accounted for temperature, precipitation, and soil evapotranspiration (Keetch and Byram 1968). The GROWSSN variable was also useful because it represented the length of photosynthetic activity in a given year. It likely served as a proxy for grassland growing conditions. Our results were concordant with Bridges et al. (2001) and Lusk et al. (2007); composite variables explain the relationship between weather and bobwhite survival better than singular metrics of precipitation or temperature.

Drought was likely a proximate factor—rather than an ultimate factor—of bobwhite survival (Guthery et al. 2005; Hernández et al. 2005). Abundant woody cover can buffer bobwhite populations in South Texas from the worst effects of drought (Parent et al. 2015). We found some support for this hypothesis in our data. Site-specific effects explained ∼7% of the total variation (20% of variation explained by the model) in our bobwhite survival data set. We posit that site-specific effects of each ranch represented the ranch-wide variation in habitat components on the landscape (e.g., woody cover, nesting cover), but did not collect habitat metric data to assess this.

Survival is a complex demographic metric, influenced by dynamic conditions and a host of different factors. We found a relatively weak relationship between weather and breeding season survival of bobwhites, after accounting for differences between the sex of bobwhites and site-specific differences among ranches. We view our model with skepticism because of its generally mediocre (35% of variance explained) performance; however, it contains useful information because our results are consistent with other studies on quail demographics and weather (Forrester et al. 1998; Guthery et al. 2001, 2002; Hernández et al. 2005). Survival of males was higher than females after accounting for all other variables in the model. This is consistent with much of the literature (Roseberry 1974; Curtis et al. 1988; Pollock et al. 1989; Taylor et al. 1999).

The relationship between weather and bobwhite survival was weak because weather is often a correlative cause of mortality, rather than a causal one. Our model explained only 35% of the variation in the data set; and ∼28.5% of the total variation explained in the data set (81% of the data explained by our model) was from weather variables. For example, hot and dry conditions reduce usable space for bobwhites during much of the day (Guthery et al. 2005). Drought and extreme temperatures can induce lethal hyperthermia in bobwhites (Forrester et al. 1998; Guthery et al. 2000, 2001). Bobwhites under heat stress must loaf in cooler habitat coverts to reduce heat loads (Forrester et al. 1998). Bobwhites often move to areas with abundant woody cover to escape lethal temperature (Parent et al. 2016). Bobwhites suffer a higher rate of predation when drought-stricken habitats provide inadequate cover and food availability (Hernández and Peterson 2007; Rader et al. 2007). Although weather was not a proximate cause in this example, it could indirectly influence bobwhite mortality (DeMaso et al. 2014).

Another possible reason for the weak relationship between weather and survival could be that our own research efforts may have influenced survival. First, radiocollars can potentially handicap bobwhites (Osbourne et al. 1997; Guthery and Lusk 2004). If bobwhite mortality increased because of capture handling and radiomarking, our model would not be able to attribute variation to a latent parameter. Second, two ranches (Ranches One and Two) used supplemental sorghum during the breeding season. Though we found no trend in survival when supplemental feed was provided, supplemental feed can decrease (DeMaso et al. 1998; Guthery 2000; Guthery et al. 2004) or increase (Buckley et al. 2015) bobwhite survival rates during the breeding season. Influences of supplemental feed are often inconsistent (Hernández and Guthery 2012) and dependent on weather (Townsend et al. 1999; Sisson et al. 2000) or soil type (Doerr and Silvy 2002).

Our results indicate a relationship exists, albeit one with considerable uncertainty, among breeding season mortality and weather conditions, sex of the bobwhite, and site-specific effects of each ranch. Bobwhite mortality increased with drought severity, and decreased with longer growing seasons. Hot, dry conditions can suppress bobwhite reproduction entirely (Guthery et al. 2005). Those conditions can also result in increased bobwhite mortality during the breeding season—a critical time in their annual life cycle (Sandercock et al. 2008). Such conditions could create a two-fold blow to a bobwhite population—no reproduction and low breeding-season survival.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Data S1. Raw data from 515 radiomarked northern bobwhites Colinus virginianus on four ranches (Ranch) in South Texas during 2001–2009. Survival during the breeding season (Survived), Sex, Age (hatch-year = 0, after hatch-year = 1), total breeding-season precipitation (precip), length of the growing-season derived from the normalized difference vegetation index (GROWSSN), mean maximum breeding season temperature (MMax, degrees C), and Keetch-Byram Drought Severity Index (KBDI = 0–800 score) were all used to understand relationships between survival and demographic and weather variables.

Found at DOI: http://dx.doi.org/10.3996/122014-JFWM-092.S1 (27 KB).

Reference S1. Keetch JJ and GM Byram. 1968. A drought index for forest fire control. Research Paper SE-38. Asheville, North Carolina: U.S. Department of Agriculture, Forest Service, Southeastern Forest Experiment Station.

Found at http://dx.doi.org/10.3996/122014-JFWM-092.S2; also available at http://www.srs.fs.usda.gov/pubs/rp/rp_se038.pdf (3 MB PDF).

Reference S2. Paulson HA, Reid EH. 1970. Range and wildlife habitat evaluation—a research symposium. U.S. Department of Agriculture Miscellaneous Publication No. 1147.

Found at http://dx.doi.org/10.3996/122014-JFWM-092.S3; also available at https://archive.org/download/rangewildlifehab1147unit/rangewildlifehab1147unit.pdf (18 MB PDF).

The Richard M. Kleberg Jr. Center for Quail Research, Caesar Kleberg Wildlife Research Institute, South Texas Quail Associates Program, Texas Quail Coalition, Texas Parks and Wildlife Department, and Houston Safari Club provided support for this project. We thank B. Ballard and W. Kuvlesky, Jr. for reviewing various drafts of this manuscript and assistance with this publication. We also thank all of the anonymous reviewers for their comments and editorial assistance.

The C.C. Winn Endowed Chair supported LAB, and the Alfred C. Glassell Jr. Endowed Professorship supported FH. This manuscript is Caesar Kleberg Wildlife Research Institute Publication Number 13-121. Any use of trade, product or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.