Reanalysis of Historical Mourning Dove Nest Data by Using a Bayesian Approach

Mourning doves Zenaida macroura are an important migratory game bird, providing nearly 2.6 million days of hunting opportunity for 857,000 hunters harvesting 14.5 million birds (Raftovich et al. 2014; Reference S2, Supplemental Material); for example, mourning dove hunting in Texas alone generates ≥US$316.4 million in economic activity (Southwick 2007; Reference S3, Supplemental Material). Mourning doves are also an important backyard bird species, with approximately 46.7 million U.S. residents watching wild birds, accounting for 38% of all expenditures related to wildlife-related recreation (U.S. Fish and Wildlife Service [USFWS] 2011). Given the widespread interest in mourning doves, it is important that management actions maintain populations at levels that balance the desires of all stakeholders. Recruitment is a key component of managing game bird populations; however, it is often difficult to directly estimate for altricial game birds such as mourning doves. Therefore, estimates of nest survival are often used as an index to recruitment and as a measure of habitat quality and quantity for mourning doves (Best and Stauffer 1980; Drobney et al. 1998; Jones and Geuple 2007).

Although dove hunting is permitted under the Migratory Bird Treaty Act of 1918 (Bean and Rowland 1997), some organizations during the late 1970s and early 1980s suggested September hunting was interfering with late-nesting mourning doves (Baskett and Sayre 1993; Minnis 1998). Specifically, it was asserted early September hunting increased mortality of eggs and nestlings due to hunter harvest of nesting adults, and hunting activity led to increased nest disturbance, resulting in reduced nesting (Sayre and Silvy 1993). Because of these concerns, a national cooperative nesting study was conducted during 1978–1980 (Geissler et al. 1987; Reference S1, Supplemental Material). The monumental project involved 27 state conservation agencies, the USFWS, five land-grant universities, and the International Association of Fish and Wildlife Agencies (currently the Association of Fish and Wildlife Agencies). The original study had two objectives: 1) estimate the proportion of annual mourning dove nesting activity occurring during the hunting season along with seasonal changes in nest survival, and 2) determine nest survival in hunted and nonhunted areas. Results from this large-scale cooperative project demonstrated only a small proportion of total nesting activity occurred after the start of the hunting season, and there was no difference in nest survival between hunted and nonhunted study areas (Geissler et al. 1987).

The original field data sheets from the national mourning dove nesting study (Geissler et al. 1987) were rediscovered in 2010. Given the advancement and availability of modern nest success estimators (Jones and Geuple 2007), we chose Bayesian nest survival estimators (He 2003; Cao and He 2005; Cao et al. 2009) to revisit the unparalleled nesting dataset to confirm earlier results and explore factors effecting nesting survival not previously considered (Cao et al. 2008). These Bayesian models estimate daily nest survival and daily nest encounter rates, and they allow irregular nest visitation under the assumption the observed nests do not have homogeneous nest survival. Because nest survival studies are often conducted in different study sites and time periods, it is unrealistic to assume homogeneous nest survival. These models also allow for incorporation of covariates thought to effect nest survival. For purposes of this paper, we deal specifically with data from the first part of the original study investigating the proportion of nesting activity throughout the year, overall nest survival, and seasonal changes in nest survival (i.e., objective 1 in Geissler et al. 1987). Therefore, our objectives for this paper were to 1) reconstruct the dataset for the first objective from the original field data sheets to ensure the same or similar data were used, 2) estimate nest survival by using modern Bayesian hierarchical models and compare results with estimates from the Mayfield method (Mayfield 1961, 1975) used in the original monograph, and 3) investigate the effect of nest covariates on nest survival and nest encounter rates.

During 1979–1980 of the original project, 106 study areas were established in 27 states (Figure 1). Area size and shape of nest searching plots were dependent upon the amount of habitat an observer could thoroughly search one 6-h day each week; plots remained unchanged during the remainder of the project. Areas were selected in a nonrandom manner to accommodate logistical needs of cooperators and reflected what was locally considered “good dove nesting habitat” (Geissler et al. 1987), or convenience sampling.

During 1979–1980, nest searching occurred in the Eastern Management Unit (EMU), Central Management Unit (CMU), and the Western Management Unit (WMU); EMU and CMU regions were divided into North (-N) and South (-S) (Figure 1). Weekly nest searches were conducted from February to October because nesting activity during November–January was assumed to be negligible across the breeding range (Sayre and Silvy 1993; Otis et al. 2008). Active mourning dove nests were defined as containing one or more eggs or nestlings. Although nestling mourning doves fledge 12–15 d posthatch (Sayre and Silvy 1993; Otis et al. 2008), nestlings were considered fledged on day 10 because they could survive on their own by that age (Geissler et al. 1987). Active nests were visited weekly until day 10 posthatch or until the nest failed. Investigators attempted to visit nests on day 10 to confirm a successful nest (Geissler et al. 1987). Nestlings were aged to the nearest day posthatch according to photographic field guides (Hanson and Kossack 1957; Mirarchi 1993).

The 11,000 data sheets discovered in a storage room at the USFWS regional office in Denver, Colorado, were shipped to the Missouri Department of Conservation, Resource Science Center, Columbia, for sorting and data entry. We used a relational database management system to reconstruct the dataset. To ensure we did not lose any data, we entered all data sheets, although duplicate photocopies were present. Most duplicate records were removed after data entry by identifying identical records, but some duplicates remained. Further errors and inconsistencies were manually checked. Some states in the reconstructed dataset (Reference S3, Supplemental Material) had more nests than originally reported (Geissler et al. 1987; Table 1); however, no details were provided in the original report about exclusion criteria for unacceptable nests. Our reconstructed dataset contained 7,139 nests.

We compared the original figures and tables by using the same methods used to originally create them, and we verified the recreated dataset by comparing original results to our reconstructed estimates (Tables S1–S10, Supplemental Material). We created a bootstrapped dataset by resampling our reconstructed data with replacement 100,000 times. We conducted the same analysis of variance as in the original report (Geissler et al. 1987). Specifically, for spring, summer, and fall, we used weighted analysis of variance (ANOVA), where weights were the number of nests in each state, to compare different management units in terms of when the nests were found, when the eggs were laid, the number of eggs and nestlings present, the percentiles of when eggs first hatched, and when the young fledged (i.e., response variables in ANOVA; Geissler et al. 1987). The weighting adjusted for sampling that may not have been proportional to the dove population in some divisions (Geissler et al. 1987). Within the management unit effect, the following were tested: the four specific contrasts North vs. South, EMU vs. CMU, WMU vs. all others, and an interaction between North vs. South and EMU vs. CMU. All regional response variables were in the form of seasonal counts that were first converted to proportions by dividing these seasonal counts by the regional total across all seasons and then transformed with angular transformation defined as θ = sin⁻¹ to stabilize variance and meet the ANOVA test assumptions. Bootstrap estimates and P values for the contrasts were calculated. For a quantitative measure of successful replication of the original dataset, we examined results to see whether the 95% confidence intervals of the P values captured the original results.

The original nest survival project (Geissler et al. 1987) estimated overall daily nest survival rate by using the Mayfield method (Mayfield 1961, 1975) that assumes constant daily nest survival and requires knowing the number of days survived and the total nesting days lost. Using the reconstructed dataset, we compared nest survival estimated from the Mayfield method to the Bayesian estimation of age-specific nest survival probabilities (BEANSP) method (He 2003; Cao et al. 2009) by using the BEANSP package (Cao et al. 2009) in the R statistical programming environment (R Core Team 2016; Data S1, Supplemental Material). Furthermore, we conducted a covariate analysis of factors affecting nest survival by using BEANSP and estimated daily nest encounter rates. To compare nest survival estimates between Mayfield and BEANSP methods, we used the Mayfield estimator and BEANSP without nest covariates. For the BEANSP analysis with covariates, we used the variables year, height of the nest, day the nest was discovered, stage (egg, nestling, egg and nestling, or fledgling) when nests were found, and management unit (EMU-N or northeast, EMU-S or southeast, CMU-N or north-central, CMU-S or south-central, and WMU or west; Figure 1).

The original study (Geissler et al. 1987) collected habitat information based on 18 categories where many categories overlapped in form or function (e.g., shelterbelt, fencerow, hedgerow). We reclassified the original 18 habitat types into 5 general categories: human dominated, linear search areas, agricultural areas, forests, and other. Further analysis using the regrouped habitat variables remained uninformative, so they were not used in further analysis. The variables nest stage, region, and nest habitat category were categorical variables; other variables were continuous (Table 2). The reference category was egg stage when examining differences between nest stages. When examining the categorical variable for region, EMU-N was the reference level. Model contrasts among mourning dove management subunits, units, and larger North vs. South vs. West scales (Figure 1) were conducted to explore regional differences in nest survival for the entire nesting season as possible explanations for regional differences in population indices.

We evaluated 10 different models with variables thought to affect mourning dove nest survival (Table 3). Deviance information criterion (Spiegelhalter et al. 2002) values (Table 3) indicated models 9 and 10 were both reasonable candidate models relative to other models, but model 9 had the lowest deviance information criterion value and was the most parsimonious. The final model was in the following form,

where J = 26 is the number of days for a nest to be considered successful (i.e., 2 d for nest initiation and egg laying, 14 d for incubation, and 10 d to fledging), is an index for the nest in the dataset, δ_j is the encounter probability on day j, q_jk is the failure probability that nest fails on day j, q_(J+1)k is the success probability that nest succeeds on day J, A_j is the age effect on the failure probabilities, x_k is the vector of covariates, and β is the vector of regression parameters. Coefficient interpretation was similar to the typical logit model, with being the ratio of the odds of nest failure at a certain age over the odds of a nest failing at that age with one less unit of the covariate x_i. In other words, β_i values were interpreted as changes in log odds, with positive values indicating an increased chance of nest failure (Cao et al. 2009).

Our reconstructed nesting dataset contained 7,139 nests compared to 6,950 nests in the original study (Table 3). Comparison of original Mayfield nest survival estimates and estimates from the reconstructed dataset showed little difference (Table 4). Our replicated Mayfield nest survival estimates were ≤0.001 different from original estimates; EMU-N, CMU-N, and CMU-S regions showed slightly lower estimates compared to original estimates (Table 4). Visual inspection of the distributions of nests found, birds hatched, and birds fledged (Figures S6–S10, Supplemental Material) also had similar results. Adding further evidence that we had accurately reconstructed the original dataset, a bootstrapping procedure captured 65 of the 80 P values from the original analysis (Tables S1–S5, Supplemental Material compared to Tables 1–5 in Geissler et al. 1987); about 50% of the inconsistencies resulted from a lack of details of how tables were constructed in the original report (Geissler et al. 1987). Because the results of our Mayfield analysis with reconstructed data were similar to original estimates, we assumed we had reconstructed the nesting dataset as accurately as possible, and we proceeded with estimating nest survival and daily nest encounter rates by using the Bayesian models. We assumed any differences between original nest survival estimates and the Bayesian model estimates could be attributed to differences in analysis and not to differences resulting from data reconstruction.

Mayfield daily nest survival estimates were constant compared to the more biologically relevant Bayesian estimates of daily nest survival probability showing periods of mortality early during the nesting cycle and again around day 20 (Figure 2). Specifically, Bayesian nest survival estimates showed lower nest survival the first 5 d of the nesting cycle during nest initiation and egg laying and a slight decline around day 19 or 20 when fledglings would be 4 or 5 d old (Figure 2). When the covariates were added to the model, however, the daily survival probabilities did not change substantially. Nest survival was significantly greater in 1980 than in 1979, and increasing nest height substantially improved nest survival (Table 5). Nests found later in the nesting cycle had greater survival. In other words, nests located containing both eggs and nestlings showed greater survival compared to nests in the egg stage only. Nests in the nestling stage showed greater survival compared to nests only in the egg stage alone, and nests in the fledgling stage showed greater survival compared to nests in the egg stage (Table 5). Nest encounter probability showed an increase in nest detection at day 4 and again at day 11 of the nesting cycle (Figure 3).

Nest survival was greater in EMU compared to CMU, and nest survival in WMU was greater than that in CMU and EMU, but not significantly different (Table 5; Figure 1). Northern states showed greater nest survival compared to southern and western states (Table 5; Figure 1). EMU-N had greater nest survival among all subregions with EMU-N > CMU-N > WMU > EMU-S > CMU-S (Table 5; Figure 1). Our comparisons of nest survival showed EMU-S and CMU-S were not significantly different (Table 5; Figure 1).

Given the gravity of many conservation policy decisions, it is critical to occasionally review and reevaluate input data and analyses used to inform those decisions. Asking fundamental questions of assumed knowledge helps strengthen existing management hypotheses and allows managers and scientists to ask better management and research questions (Firestein 2012). In this case, our reanalysis of the reconstructed data from Geissler et al. (1987) demonstrated a similar distribution of annual nesting activity, and similar nest survival by using modern analysis techniques (He 2003; Cao et al. 2009). Although a myriad of nest survival estimation techniques have been developed since the original Mayfield analysis (Geissler et al. 1987; Johnson 2007), Bayesian nest survival estimates benefit from allowing for an irregular visitation schedule and inclusion of covariates that may provide clues into factors influencing mourning dove nest survival dynamics.

Comparison of the original Mayfield nest survival estimates (Geissler et al. 1987) and our Bayesian model estimates for the reconstructed data show the value of modern nest survival estimators (Figure 2). Compared to the constant daily nest survival rate from the Mayfield estimator, our most parsimonious model with the lowest deviance information criterion value (Table 3) showed a more biologically meaningful representation of periods of greater risk during the nesting cycle (Figure 3). For example, nest survival increased steadily from nest initiation and egg laying until incubation during the first 4 to 5 d of the nesting cycle when the male and female take turns sitting on the nest (Figure 3). This period of low nest survival early in the nesting cycle is consistent with previous information on mourning dove behavior where nest-building activities may render the nest visible and conspicuous during nest construction (Coon et al. 1981). Construction of structurally flimsy nests along with a fixed clutch size of two eggs, small egg size and mass, multiple annual nesting attempts, reuse of other birds' nests, and production of crop milk by both sexes demonstrates relatively low input of parental resources in the nesting process (Westmoreland et al. 1986; Blockstein and Westmoreland 1993). If nests are damaged or lost early in the nesting cycle, minimal resources are lost and another nest can be initiated relatively quickly (Blockstein 1986; Blockstein and Westmoreland 1993). Around day 19 or day 20 when nestlings are 4 or 5 d old, our survival estimates declined slightly (Figure 2). As nestlings get older and require more crop milk, adults make increased nest visits to feed the nestlings (Blockstein and Westmoreland 1993; Otis et al. 2008), and the increased activity may draw increased predators' attention to the nest (Thompson and Burhans 2003).

Our final nest survival model did not include a habitat variable (model 9; Table 5; Figure 2), confirming previous findings that mourning doves are widely considered habitat generalists (Stauffer and Best 1980; Sayre and Silvy 1993; Best et al. 1997; Otis et al. 2008). Although nest searching activity in the original study focused on “good dove nesting habitat” (Geissler et al. 1987), important habitat containing nests were not searched in the original study because of low nest density and extra search effort needed to locate nests (Drobney et al. 1998; Hanson and Kossack 1963; Knopf 1986; Hughes et al. 2000). Given the objectives of the original study to find as many nests as possible to understand the annual distribution of nesting effort and survival, our encounter probability results (Figure 3) show nests were more easily located during early in the incubation phase, approximately midway through incubation, and again early in the nestling stage. If all potential nesting habitats were searched, substantially fewer nests would have likely been found (Schulz and Sheriff 1995; Drobney et al. 1998). Given the nonsystematic and opportunistic approach to nest searching, future use of these reconstructed data should recognize this characteristic.

Best and Stauffer (1980) suggest habitat generalists often demonstrate lower nest survival because the wider variety of habitats contain a greater variety of predators, and a likelihood of not having effective antipredator behavior in the wide range of habitats encountered. Although this may apply to some species in rural and agricultural settings, mourning doves routinely are listed as one of most abundant species monitored by the North American Breeding Bird Survey (Sauer et al. 2013). Due to their relative abundance, mourning doves seem to have overcome this obstacle by evolving an adaptive reproduction strategy allowing them to invest minimal time and resources into multiple nests in a wide range of habitats (Blockstein and Westmoreland 1993).

Nest height was an important factor in our preferred nest survival model (Tables 3 and 4), although the positive effect of nest height on survival is likely curvilinear. Previous nest survival estimates for other songbirds showed similar effects of greater survival with increased nest height (Burhans et al. 2002; Thompson and Burhans 2003), whereas other studies indicate little positive effect of nest height (Germaine et al. 1998; Smith et al. 2012). Alternatively, ground-nesting birds have demonstrated nest survival rates comparable to birds with tree nests (Downing 1959; Best and Stauffer 1980; Drobney et al. 1998), possibly indicating the influence of nest height on nest survival may be an effect of annual variations in local habitat or weather. A more practical explanation for nest height improving survival in our models is the convenience sampling used to locate nests in “good dove nesting habitat.” Nest searchers may have developed a search image for nests in linear habitats 2–3 m above ground and missed locating other nests in different habitats and heights.

Original nest survival estimates used pooled nesting data for 1979–1980 (Geissler et al. 1987), but we found our Bayesian nest survival analysis showed significant differences between 1979 and 1980. Annual variation in nest survival among altricial species has been directly attributed to weather (Best and Stauffer 1980; Shaffer and Burger 2004) or indirectly to annual habitat or food availability differences (Martin 1995; Burhans et al. 2002; Thompson and Burhans 2003). Pooling nesting data among multiple years improved the precision of nest survival estimates in the original study (Geissler et al. 1987), but in this case it masked real differences occurring in nest survival between years.

Nest survival in the EMU was significantly greater compared to that in the CMU or WMU (Table 5; Figure 1), corroborating historical population trends where EMU states demonstrated stable to slightly increasing National Mourning Dove Call-Count Survey trends (Tomlinson et al. 1994; i.e., increased nest survival may have been benefiting the EMU populations). Population trends in the CMU and WMU, however, showed National Mourning Dove Call-Count Survey declining trends since 1966; during the same period, CMU trends were among the highest in the three management units (Tomlinson et al. 1994). Declines in National Mourning Dove Call-Count Survey trends for the CMU and WMU during this period could be attributed to decreased nest survival with CMU states having larger populations or higher annual survival. Regional differences in mourning dove nest survival from our reconstructed data also confirm existing hypotheses that northern states demonstrate greater nest success compared to southern states where regional differences may reflect physiological trade-offs associated with northern latitude breeding (Wilson et al. 2004), weather differences between years (Miller et al. 2001), or food availability (Martin 1995) between North and South study areas during years of the study.

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

Data S1. Reconstructed mourning dove Zenaida macroura nesting dataset containing 7,139 nests derived from the original report “Mourning Dove Nesting: Seasonal Patterns and Effects of September Hunting” (Geissler et al. 1987) conducted on 106 study areas in 27 states during 1979–1980. Data used for analysis is contained in a file titled Nests.csv.

Found at DOI: Nests.csv (787 KB). Also found at DOI: 10.3996/102015-JFWM-100.s1; (786 KB PDF).

Data S2. Zip file containing the main code script and other source code necessary to compute nest survival probabilities and nest detection probabilities.

Found at DOI: 10.3996/102015-JFWM-100.s2; (2352 KB Zip File).

Table S1. Comparison of P values from Table 1 in the original report “Mourning Dove Nesting: Seasonal Patterns and Effects of September Hunting” (Geissler et al. 1987) and P values from the reconstructed mourning dove Zenaida macroura nesting dataset for analysis of variance (ANOVA) of seasonal percentage of annual total nests first found and regional subdivisions. The P values from the ANOVA tables were bootstrapped using 100,000 replications of the 7,139 nests in the reconstructed dataset to examine whether the 95% confidence intervals (CI) of the P values captured the original results. Asterisk (*) denotes bootstrapped 95% CI that did not contain original P values.

Found at DOI: 10.3996/102015-JFWM-100.s3; (15 KB DOCX).

Table S2. Comparison of P values from Table 2 in the original report “Mourning Dove Nesting: Seasonal Patterns and Effects of September Hunting” (Geissler et al. 1987) and P values from the reconstructed mourning dove Zenaida macroura nesting dataset (1979–1980) for analysis of variance (ANOVA) of seasonal percentage of annual total eggs laid based on backdating from those eggs surviving to hatch. The P values from the ANOVA tables were bootstrapped using 100,000 replications of the 7,139 nests in the reconstructed dataset to examine whether the 95% confidence intervals (CI) of the P values captured the original results. Asterisk (*) denotes bootstrapped 95% CI that did not contain original P values.

Found at DOI: 10.3996/102015-JFWM-100.s4; (16 KB DOCX).

Table S3. Comparison of P values from Table 3 in the original report “Mourning Dove Nesting: Seasonal Patterns and Effects of September Hunting” (Geissler et al. 1987) and P values from the reconstructed mourning dove Zenaida macroura nesting dataset (1979–1980) for analysis of variance (ANOVA) of seasonal percent of annual total of weekly counts of individual eggs or nestlings present. The P values from the ANOVA tables were bootstrapped using 100,000 replications of the 7,139 nests in the reconstructed dataset to examine whether the 95% confidence intervals (CI) of the P values captured the original results. Asterisk (*) denotes bootstrapped 95% CI that did not contain original P values.

Found at DOI: 10.3996/102015-JFWM-100.s5 (15 KB DOCX).

Table S4. Comparison of P values from Table 4 in the original report “Mourning Dove Nesting: Seasonal Patterns and Effects of September Hunting” (Geissler et al. 1987) and P values from the reconstructed mourning dove nesting Zenaida macroura dataset (1979–1980) for analysis of variance (ANOVA) of percentiles of hatch dates for first egg in nests and length of nesting season; nesting season defined as the number of days between the 10th and 90th percentile. The P values from the ANOVA tables were bootstrapped using 100,000 replications of the 7,139 nests in the reconstructed dataset to examine whether the 95% confidence intervals (CI) of the P values captured the original results. Asterisk (*) denotes bootstrapped 95% CI that did not contain the original P values.

Found at DOI: 10.3996/102015-JFWM-100.s6; (17 KB DOCX).

Table S5. Comparison of P values from Table 5 in the original report “Mourning Dove Nesting: Seasonal Patterns and Effects of September Hunting” (Geissler et al. 1987) and P values from the reconstructed mourning dove Zenaida macroura nesting dataset for ANOVA of seasonal percentage of annual total of mourning doves fledged. The P values from the analysis of variance (ANOVA) tables were bootstrapped using 100,000 replications of the 7,139 nests in the reconstructed dataset to examine whether the 95% confidence intervals (CI) of the P values captured the original results. Asterisk (*) denotes bootstrapped 95% CI that did not contain original P values.

Found at DOI: 10.3996/102015-JFWM-100.s7; (17 KB DOCX).

Figure S1. Each of the four plots shows a weekly representation of the percentage of a particular variable; the plots on the left are from original Figure 7 in Geissler et al. (1987), and the plots on the right from the reconstructed dataset estimating seasonal distribution of nesting activity and nest survival conducted on 106 study areas in 27 states during 1979–1980. Nesting variables were as follows: percentage of mourning dove Zenaida macroura nests first found each week (A), number of individual eggs and nestlings present each week (B), counts of nest with first egg hatched each week (C), and weekly percentage of individual mourning doves fledged (D). For each plot, the larger boxplot indicates the 10th and 90th percentiles of the data and the inner boxplot indicates the 25th and 75th percentile.

Found at DOI: 10.3996/102015-JFWM-100.s8; (119 KB DOCX).

Figure S2. Plots on the left show percentage of mourning dove Zenaida macroura nests found each week by management unit (i.e., Eastern Management Unit-North and -South [EMU-N and EMU-S]; Central Management Unit-North and -South [CMU-N and CMU-S]; Western Management Unit [WMU]) from Figure 2 in Geissler et al. (1987) estimating seasonal distribution of nesting activity and nest survival conducted on 106 study areas in 27 states during 1979–1980. Graphs on the right are from the reconstructed dataset. For each plot, the larger boxplot indicates the 10th and 90th percentiles of the data and the inner boxplot indicates the 25th and 75th percentile.

Found at DOI: 10.3996/102015-JFWM-100.s9; (72 KB DOCX).

Figure S3. Replication of Figure 3 in Geissler et al. (1987) estimating seasonal distribution of mourning dove Zenaida macroura nesting activity and nest survival conducted on 106 study areas in 27 states during 1979–1980. Left side shows original plots and the right side shows the reconstructed dataset. Each plot shows weekly percentage counts of eggs or nestlings by management unit (i.e., Eastern Management Unit-North and -South [EMU-N and EMU-S]; Central Management Unit-North and -South [CMU-N and CMU-S]; Western Management Unit [WMU]). The larger boxplot indicates the 10th and 90th percentiles of the data and the inner boxplot indicates the 25th and 75th percentile.

Found at DOI: 10.3996/102015-JFWM-100.s10; (71 KB DOCX).

Figure S4. Replication of original Figure 4 in Geissler et al. (1987) estimating seasonal distribution of mourning dove Zenaida macroura nesting activity and nest survival conducted on 106 study areas in 27 states during 1979–1980. Left side shows original plots and the right side shows plots from the reconstructed dataset. Each plot shows percentage of nests with the first egg hatched each week by management unit (i.e., Eastern Management Unit-North and -South [EMU-N and EMU-S]; Central Management Unit-North and -South [CMU-N and CMU S]; Western Management Unit [WMU]). The larger boxplot indicates the 10th and 90th percentiles of the data and the inner boxplot indicates the 25th and 75th percentile.

Found at DOI: 10.3996/102015-JFWM-100.s11; (74 KB DOCX).

Figure S5. Replication of Figure 5 in Geissler et al. (1987) estimating seasonal distribution of mourning dove Zenaida macroura nesting activity and nest survival conducted on 106 study areas in 27 states during 1979–1980. Left side shows original plots and the right side shows the reconstructed dataset. Each plot shows the percentage of the number of young fledged each week for that management unit (Eastern Management Unit-North and -South [EMU-N and EMU-S]; Central Management Unit-North and -South [CMU-N and CMU S]; Western Management Unit [WMU]). The larger boxplot indicates the 10th and 90th percentiles of the data and the inner boxplot indicates the 25th and 75th percentile.

Found at DOI: 10.3996/102015-JFWM-100.s12; (73 KB DOCX).

Reference S1. Geissler PH, Dolton DD, Field R, Coon RA, Percival HF, Hayne DW, Soileau LD, George RR, Dunks JH, Bunnell SD. 1987. Mourning dove nesting: seasonal patterns and effects of September hunting. Washington, D.C.: U.S. Fish and Wildlife Service.

Found at DOI: http://pubs.usgs.gov/rp/0168/report.pdf (2.71 MB PDF).

Also found at DOI: 10.3996/102015-JFWM-100.s13; (2782 KB PDF).

Reference S2. Raftovich RV, Chandler S, Wilkins KA. 2014. Migratory bird hunting activity and harvest during the 2012–2013 and 2013–2014 hunting seasons. Laurel, Maryland: U.S. Fish and Wildlife Service, Division of Migratory Bird Management.

Found at DOI: https://www.fws.gov/migratorybirds/pdf/surveys-and-data/HarvestSurveys/MBHActivityHarvest2012-13and2013-14.pdf (2.68 MB PDF).

Also found at DOI: 10.3996/102015-JFWM-100.s14; (2749 KB PDF).

Reference S3. Southwick R. 2007. The 2006 economic benefits of hunting, fishing and wildlife watching in Texas. Fernandina Beach, Florida: Southwick Associates, Inc.

Found at DOI: https://tpwd.texas.gov/publications/nonpwdpubs/media/tx_fish_hunt_wl_view_economics.pdf (0.98 MB PDF).

Also found at DOI: 10.3996/102015-JFWM-100.s15; (1007 KB PDF).

We are grateful for reviews and suggestions on earlier drafts from anonymous reviewers and the Subject Editor. This study was possible with the assistance of D.D. Dolton (USFWS, retired) who saved the original field nesting data sheets from being destroyed and shared them with us. We are grateful for the database management and data entry supervision provided by J. Fleming, and data entry by J. Collantes and E. Buckner. We acknowledge Y. Yang for providing valuable statistical support and M. Mitchell and I. Vining for graphics support. Funding and support for this work were provided by the Missouri Department of Conservation, Resource Science Division and the University of Missouri, Department of Fisheries and Wildlife Sciences.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.