Patterns of nocturnal bird migration in southern Mexico

Introduction

Almost half of the bird species that inhabit the US travel south to winter in the Neotropics every year ( Rappole, Morton, Lovejoy, & Ruos, 1983 ). These include a wide range of taxonomic groups, from small passerines that depend on powered flight to raptors that use thermal wind updrafts for soaring flights ( Gill, 2007 ). Differences in flight style between these 2 groups of birds are related to the time of day used for migration: while soaring birds (e.g., raptors) almost always migrate during daytime, powered fliers (e.g., passerines) are predominantly nocturnal migrants ( Kerlinger, 1995).

Some migratory flyways used by Nearctic-Neotropical nocturnal migrants converge at the Isthmus of Tehuantepec (UNEP, 2009), the narrowest landmass between the Gulf of Mexico (GOM) and the Pacific Ocean (∼220 km at its narrowest stretch). Located in southern Mexico, the Isthmus of Tehuantepec has been considered “an avenue for migrants moving between North and South America”, and both the GOM and Pacific lowlands are thought to be “important wintering areas for North American migrants” ( Binford, 1989).

The Isthmus of Tehuantepec is part of the transitional area between the Nearctic and Neotropical realms ( Ríos-Muñoz, 2013 ) with its local topography playing an important role in creating major differences between slopes. The Sierra Madre Oriental and Sierra Madre del Sur run parallel to the GOM and Pacific coastlines, respectively; these important ridgelines meet at the western side of the isthmus, continuing to the east under the name of Sierra Madre de Chiapas ( Barber & Klicka, 2010 ) after being interrupted by a low elevation gap known as Chivela Pass. One of the characteristics that define the identity of each slope is a gradient in barometric pressure, which is consistently higher in the GOM throughout the year ( Steenburgh, Schultz, & Colle, 1998 ). This promotes a constant flow of wind toward the Pacific through the Chivela Pass ( Chelton, Freilich, & Esbensen, 2000 ), creating a rain shadow on the northward side of the mountain range, and promoting the establishment of rain forests in the GOM side of the isthmus ( Rzedowski, 1990 ). In contrast, the Pacific side is characterized by a dryer climate and tropical deciduous forests ( Conabio, 1998).

In the last decade, the Pacific side of the isthmus has gained much relevance due to its high potential for wind-energy production ( Alemán-Nava et al., 2014; Elliott et al., 2003; Jaramillo & Borja, 2004 ), promoting avian studies on different topics ( Villegas-Patraca, Cabrera-Cruz, & Herrera-Alsina, 2014; Villegas-Patraca, MacGregor-Fors, et al., 2012; Villegas-Patraca, Macías-Sánchez, Macgregor-Fors, & Muñoz-Robles, 2012 ), but few have been published regarding the flow of nocturnal migrants. Cabrera-Cruz, Mabee, and Villegas-Patraca (2013a) published the first reports on nocturnal bird migration for the area, derived from radar observations made from a single site during one fall season, but with no information on the species migrating through the area.
The objective of this study is to describe the patterns of nocturnal migratory traffic rates (MTR), defined as the number of targets passing across 1 km of migratory front per hour (targets/km/h), flight directions, flight altitudes, and species of nocturnal migrants observed from the Pacific side of the Isthmus of Tehuantepec, within one of the most important Wind Resource Areas of Mexico ( Alemán-Nava et al., 2014 ). For this, we used data collected with a single marine radar unit and point counts during 5 consecutive fall seasons (2009–2013), from different operational or proposed wind farms, under request by different wind-energy companies that were required to carry out environmental studies by the environmental authorities of Mexico. This work also aims to make public information that otherwise would remain in gray literature, informing society, conservation organizations, decision makers, and others interested in bird migration in the region.

Materials and methods

We studied bird migration from different sites within the Wind Resource Area of the Isthmus of Tehuantepec, an area that extends from the Pacific Ocean coast northward approximately 60 km, and approximately 60 km to 80 km from east to west (Elliott et al., 2003 ), centered approximately around the town of La Ventosa. The main vegetation type is a matrix of pastureland, sorghum crops, and fragments of tropical deciduous forest, which was the original vegetation type in the area ( Muñoz-Jiménez, 2014).

We monitored nocturnal bird migration with the aid of a marine radar adapted as a mobile unit, operating during a different number of nights throughout 5 fall migratory seasons (2009–2013), and identified bird species with morning point counts. Our study sites were distributed around ∼17 km route of coastal plain, delimited to the north by the Tolistoque ridgeline and to the south by the Laguna Superior ( Fig. 1 ). Radar monitoring stations were located in flat terrain either within operational or proposed wind farms and in sites where surrounding vegetation acted as a partial radar fence to reduce ground clutter contamination.

Figure 1
Monitoring sites where we collected radar and point count data (gray polygons). Sites with black dots (wind turbines) were operational wind farms at the time of our observations, sites without dots were being proposed for development. Black circles represent the 1.5 km radius of radar range used during our observations, and black bars radiating from the center of the circles are the flight directions recorded from each site; bar length indicates the amount of data for each site. Data was pooled for sites that were monitored multiple years. The inset shows the state of Oaxaca in southern Mexico in light gray, and the Pacific side of the Isthmus of Tehuantepec enclosed by a black rectangle.

Given that our radar monitoring was made under request by different wind-energy companies, each year we worked from different sites but always within the Wind Resource Area of the Pacific coastal plains of the isthmus. In 2009 and 2010, we monitored nocturnal bird migration from a single site throughout the season (La Venta II and Unión Fenosa, respectively), visiting each site daily during the whole field season. In 2011, 2012 and 2013 we had 3, 2 and 3 monitoring sites, respectively, and we alternated our daily visits among sites to conduct the observations because we only had 1 radar unit. For example, our first night of observations in 2011 was at La Venta II, the next night we visited Oaxaca 1, and the following night we visited Alesco; we repeated this sequence throughout the duration of our field season.

At the time of our observations, La Venta II was an operational wind farm whose area was composed by a matrix of pastureland, sorghum crops, and patches of deciduous dry forest (16°35′29.45″ N, 94°48′41.91″ W; 36 m asl). In 2010, we worked at Unión Fenosa, at the time it was an area proposed for the development of a wind-energy project; most of its area was covered by deciduous dry forest but also had portions of pastureland (16°26′0.94″ N, 94°48′36.61″ W; 13 m asl). In 2011 we worked in 3 sites: La Venta II, Alesco and Oaxaca 1; Alesco was an area proposed for wind-energy development covered mainly by deciduous dry forest and pastureland (16°27′28.24″ N, 94°54′36.54″ W; 5 m asl); Oaxaca 1 was a wind farm on the final stage of construction (not operating yet), the area not occupied by service roads or service areas for wind turbines was covered by deciduous dry forest (16°33′42.60″ N, 94°42′42.93″ W; 39 m asl). In 2012, we worked at La Venta II and Peñoles; the latter was an operational wind farm, with most of its area covered by deciduous dry forest (16°35′12.18″ N, 94°59′53.62″ W; 66 m asl). In 2013, we worked in La Venta II, Oaxaca 1 and C.E. Santo Domingo; C.E. Santo Domingo was an operational wind farm with wind turbines intermixed with sorghum crops (16°30′31.82″ N, 94°46′27.76″ W; 22 m asl); Figure 1 shows the location of our study sites, and the arrangement of wind turbines for those operational wind farms. Monitoring periods, as well as starting and ending dates of observations, differed each year ( Fig. 2).

Figure 2
Monitoring periods. Extent of each radar monitoring season on the Pacific side of the Isthmus of Tehuantepec, Oaxaca, Mexico. The total number of nights when observations were made is depicted within the bars.

Every night, we moved the radar unit to the corresponding sampling location. We used a horizontal and vertical mode of radar operation to observe nocturnal bird migration, starting ∼45 min after sunset during 6 1-h sampling sessions/night to cover the peak of nocturnal migratory activity ( Gauthreaux, 1971; Mabee & Cooper, 2004; Mabee, Cooper, Plissner, & Young, 2006 ). Each 1-h session was subdivided into: (1) a 10-min period to adjust radar to the horizontal mode, (2) 10-min in horizontal mode to count targets with a manual counter, (3) 15-min in horizontal mode to record flight speeds and directions, (4) 10-min to switch radar to vertical mode, and (5) 15-min in vertical mode to record flight altitudes. This subdivision was necessary because data collection was done by hand in real time, and the operator could not perform all the actions at the same time. Following Mabee et al. (2006), we used a 1.5 km radius of observation for both horizontal and vertical modes of operation. When the radar was operating in the vertical mode, the antenna was rotated vertically 90° with respect to the leveled radar. Each night, there were 1–2 radar observers operating the equipment.

We studied nocturnal bird migration with an X band marine radar (Model FR-1525 Mark 3, Furuno, Nishinomiya, Japan) mounted on a truck adapted as a mobile unit; a description of a similar radar laboratory can be found in Cooper, Day, Ritchie, and Cranor (1991) and Harmata, Podruzny, Zelenak, and Morrison (2000) . The radar transmitted at a frequency of 9140 MHz through a 2 m-long slotted waveguide (antenna) with a maximum output of 25 kW and was operated with a pulse length of 0.07 μs. The display unit had a range resolution of 35 m, and the antenna emitted a beam with a width of 1.23° (horizontal) × 20° (vertical) with side lobes ± 10° (Furuno, 2002 ). The unit was powered with a low-noise electric generator. We estimated visually that ground clutter was <30% of the radar screen at every site.

We estimated the migratory traffic rate (MTR) from the count data; we also measured ground speeds (km/h) and flight directions (degrees) using a hand-held scale and inner compass from the radar monitor, and measured flight altitudes (m agl) with an index line from the monitor. We use the term ‘target’ to designate objects detected because radar does not allow unequivocal identification of vertebrates ( Larkin & Diehl, 2012 ), and concurrent direct observations were not made to confirm the identity of targets. To minimize counting of non-bird targets (such as insects), we omitted small targets appearing at the center of the screen (within a 600-m radius of the radar station) flying at slow speeds (<24 km/h), as slower speeds could not be accurately measured with our hand-held scale.

Flight speeds of bats partially overlap those of birds, mainly at the lower speeds ( Hedenström, Johansson, & Spedding, 2009 ), so it was not possible to distinguish bats based solely on flight speed. We did not record targets with erratic flight patterns because they may have been foraging bats; migrating birds tend to have linear paths ( Bruderer & Popa-Lisseanu, 2005 ). We acknowledge, however, that there could be some bats displaying a linear flight and similar speed to birds, considering that our study area harbors one of the highest bat diversities in Mexico ( Santos-Moreno, 2014 ). Although passerines seem to migrate together ( Moore, 1990 ), the distances between them ( Larkin & Szafoni, 2008; Millikin, 2001 ) and the fixed 35 m range resolution of our radar unit ( Furuno, 2002 ) allowed us to assume that most of the targets were individual birds. All data were recorded manually into a laptop computer.

Point counts were conducted during some of the same seasons when our radar observations took place, and were also required by different wind-energy companies. Point counts were coordinated by our working team, but had a temporal design independent of our radar observations. The most consistent effort was made at the La Venta II wind farm, where 48 established point counts were covered every month from August through November 2009–2013. These point counts were covered during a different number of days each month, depending on factors such as weather and availability of technicians. Observations usually started around 6 am and ended around 11 am. Additional point counts were made between 2011 and 2013 at other wind farms or areas proposed for wind farm construction ( Table 1 ). At each point, an experienced birder spent 5 min identifying bird species within and beyond a 25-m radius, either visually or acoustically.

Table 1

Point count sites, sampling effort, and dates where observations were made in the vicinity of our radar sites on the Pacific side of the Isthmus of Tehuantepec, Oaxaca, Mexico. Footnotes describe the status of the sites when point counts where conducted, and location reference.

a
Site proposed for wind energy development, located between the town named La Ventosa and the Peñoles wind farm.
b
Wind farm under construction (2012) and operational (2013), located immediately north of the Oaxaca 1 wind farm.
c
Wind farm under construction, located south of the La Ventosa town.

Diurnal point counts were not intended as proof of identity for our radar targets, but rather as simple evidence of what species of Nearctic-Neotropical nocturnal migrants are present in the Pacific side of the Isthmus of Tehuantepec during fall migration.

Radar data were not corrected for differences in detectability with distance from the radar unit, hence all estimates are indices. To reduce data contamination by insects, we corrected ground speeds for airspeeds of every target recorded, deleting from analysis targets with corrected airspeeds slower than 7 m/s because they were likely insects ( Cabrera-Cruz, Mabee, & Villegas-Patraca, 2013b; Larkin, 1991 ). Wind data used to correct ground speeds was collected at ground level because higher-elevation measurements were not available; hence for vertical radar we estimated minimum airspeeds.

We analyzed flight directions using ‘circular’ ( Agostinelli & Lund, 2013 ), a package of circular statistics in R ( R Development Core Team, 2016 ). We report mean flight direction ( μ ) and the length of the mean vector or concentration parameter ( r ) which ranges from 0 to 1, where a large value indicated that observations were clustered more closely around the mean. We performed Watson–Williams F -tests to compare mean flight directions among sites and among seasons. Although concentration parameters ( r ) were not equal among years and sites, and Watson–Williams F -test assumes equality, this test is robust against deviation from these assumptions ( Berens, 2009 ). We compared MTRs among years and hours of the night with repeated measures Anova tests, performed with the ‘nlme’ package ( Pinheiro, Bates, DebRoy, & Sarkar, 2011 ) in R, specifying MTR as the dependent variable with hourly working sessions and years as categorical predictors. We report MTR as the mean (±1 SE) number of targets passing through a 1-km migratory front in 1 h (targets/km/h). We fitted a generalized linear model to our flight altitudes data and used Tukey post hoc contrasts to compare flight altitudes among years. Flight altitudes are reported in meters above ground level (m agl). We used the function ‘gam’ from the package ‘mgcv’ ( Wood, 2011 ) in R to fit generalized additive models (GAMs) to our MTR and flight altitude data as a function of session. We limited the use of GAMs to create graphical representations of the trend of MTR and flight altitudes across the night for each year. We applied a Z -score transformation to our data before fitting the models because GAM requires the data to be normally distributed. We compared flight altitudes among years by fitting a generalized linear model (GLM) indicating altitudes a response and year a predictor, and then performing a multiple comparison of means (Tukey) post hoc analysis. We did not compare MTR or flight altitudes among nights in 2011–2013 because during those years we worked in >1 site and data pooling was not possible given the significant differences in MTR among nights (see Results). For all analyses, we used data from nights with ≥5 sessions/night.

The timing and duration of our field seasons were different each year, but the area covered within the isthmus remained constant, and our observations overlapped the peak of passerine migration for the GOM side of the isthmus: 22 September–01 November ( Winker, 1995a ). For comparisons among years, we used data collected only during October, a period from which all sites had data. From the point counts we only present a list of nocturnal migrants.

Results

During the 5 fall seasons and at all sites, most recorded targets were flying to the southeast ( Fig. 1, Table 2 ), with a high concentration around the mean direction (mean flight direction with all sites and years pooled = 131°, r = 0.75, N = 27,842 radar targets). Seasonal mean flight direction was different among the 4 years monitored from La Venta II (2009 and 2011–2013; F 3, 7200 = 47.47, p < 0.001) and the 2 years monitored from Oaxaca 1 (2011 and 2013; F 1, 2933 = 50.25, p < 0.001, Table 2 ). It was also different among the 3 monitoring sites of 2011 ( F 2, 3485 = 193.57, p < 0.001), the 2 sites of 2012 ( F 1, 11651 = 362.47, p < 0.001) and the 3 sites of 2013 ( F 2, 6697 = 202.39, p < 0.001; Table 2).

Table 2

Mean seasonal values of flight direction, migratory traffic rate, and flight altitude for all years and sites (±1 SE). Under flight direction, r = length of the mean vector or concentration parameter, and N = number of targets sampled. Min and Max are seasonal means for each site.