Identifying volatile organic compounds used for olfactory navigation by homing pigeons
Regional observations of volatile organic compounds
Measurements were conducted in summer 2017 and spring 2018 in the area of Pisa, Italy, as part of the HOMING project (Hunting Organic Molecules In NaviGation). Measurements consisted of: (i) a pilot study (summer 2017) to investigate the volatile organic compounds (VOCs) emitted by three representative local ecosystems surrounding the bird’s home aviary; (ii) a 2-month intensive field campaign (spring 2018) at the bird’s home aviary to monitor VOCs along with meteorological variables; (iii) three flights on board of a Cessna aircraft to sample at ca. 180 m, within the birds typical flight altitude (spring 2018). Specifically, (i) was conducted to identify the chemical composition of surrounding forest sites and test whether they can be smelled distinctly from the aviary. (ii) Was performed to derive the olfactory maps developed by birds based on the assumption that birds and our analytical equipment have comparable detection thresholds. (iii) Was needed to examine any regional scale gradient used by pigeons when flying. The home aviary is operated by the University of Pisa and is located at the rural site Arnino (43°39′25.7″N 10°18′14.7″E), outside the city of Pisa (11 km North–East), close to the Tyrrenian sea (1.8 km West) and the mouth of river Arno (3.2 km North) (Fig. 1).
Figure 1
Map of the sampling area. The point designated as “home” refers to the birds aviary of Arnino (43°39′25.7″N 10°18′14.7″E) located in Tuscany in the area of Pisa-Livorno. Points on the map show the three sites used for releasing birds during a navigational experiment and the main biogenic and anthropogenic sites in the area. Photos show an aerial view of the Arnino field site during the airborne sampling (left) and a homing pigeon equipped with a GPS-tag before being released for testing its homing performance (right). Figure drawn with Igor WaveMetrics.
Full size image
Arnino houses hundreds homing pigeons (Columba livia) used for navigational experiments. Volatile organic compound mixing ratios were determined with state-of-the-art on-line and off-line analytical techniques (PTR-MS and GC–MS) and speciated in their isomeric and enantiomeric forms.
Navigational experiments entailed releasing 174 pigeons equipped with GPS loggers (N. 80/2013-A), who had never left their home before, from unfamiliar sites 50–70 km from their home (Fig. 1). For logistical reasons (power availability and shelter for technological equipment at release points), no simultaneous measurement of air composition and bird release was possible. Therefore, we used results from the most recent available flight experiments (summer 2016 and summer 2017) to determine general homing performance indicators and examined the latter in conjunction with available meteorological information, as modelled air masses trajectories.
The daily variation of volatile organic compounds at the aviary site
Meteorological parameters including, temperature, relative humidity, wind direction and wind speed were measured in May–June 2018 during the ground-based campaign at the home aviary (Supplementary Fig. 1). Air temperature and relative humidity ranged between 12–25 °C and 60–85%, respectively (Fig. 2 and Supplementary Fig. 1), showing the expected mirrored diel cycle, whereby temperature is highest and RH lowest by day. Wind direction and speed exhibited a repeating daily pattern during the 2 months of measurements (Fig. 2 and Supplementary Fig. 1).
Figure 2
Diel cycle of meteorological parameters (a,b) and measured VOCs (c–f). The volumetric mixing ratio (VMR) of the volatile organic compounds are reported as their measured mass fragment by PTR-MS: m/z 63 (c), m/z 69 (d), m/z 81 (e) and m/z 107 (f); identified respectively as dimethyl sulphide (c), isoprene (d), sum of monoterpenes (e) and sum of xylenes (f). Meteorological parameters are plotted as mean campaign values (marker), median (line) and interquartile range (shaded area). VOC box plots report mean campaign values (marker), median (line), interquartile range (box) and 10th and 90th percentiles (whiskers).
Full size image
The proximity to the coast exposes the measurement site to influence by the local sea breeze system. Air masses reaching the aviary came from inland (mean wind angle 150°) every day until midday, and thereafter between 12:00 and 20:00–21:00 (local time) the air masses were advected from the sea (mean angle 270°). The wind speed, like the wind direction, also showed a diurnal profile. Figure 2 shows the diurnal profile of some compounds measured by proton transfer reaction mass spectrometry (PTR-MS). They have contrasting diel cycles and hereafter investigated further: dimethyl sulphide (DMS, m/z 63), isoprene (m/z 69), sum of monoterpenes (MT, m/z 81) and sum of xylenes (m/z 107). The full mass list and identities are reported in Supplementary Table 4 with other compound diel variations in Supplementary Fig. 2. We refer to the abbreviations included at the beginning of the manuscript for a reader not familiar with the terms used herein.
Dimethyl sulphide is a biogenic compound emitted by phytoplankton activity in seawater21,22,23; it is insoluble in water, and so readily transfers to the air at a rate dependent on the temperature and wind speed. Our measurements show that DMS mixing ratios at the site remained low and constant until 12:00, possibly reflecting weak soil emissions from the land24, to increase to a maximum at 20:00 due to continued marine emission, until the reversal of the wind back to offshore. The maximum hourly averaged mixing ratio of DMS is 0.17 ± 0.8 ppbv, comparable to 0.12 ppbv measured by Derstroff et al.25 at a coastal site on Cyprus.
Isoprene mixing ratios showed a bimodal distribution with two maxima at 11:00–12:00, and 18:00–20:00; similar to methyl ethyl ketone (MEK) and isoprene oxidation products methyl vinyl ketone, methacrolein and isoprene peroxides (MVK + MACR + ISOPOOH) and acetic acid (Supplementary Fig. 2). At both morning and afternoon maxima, the isoprene mixing ratio reached 0.7 ± 0.6 ppbv (maximum hourly average ± 1σ standard deviation). This is 10 times higher than the value observed during wintertime at a similar rural Mediterranean site in Spain26, but lower than the value measured in summertime by Steinbacher and coauthors27 at a rural site in the Po valley, Italy. Isoprene emission from terrestrial vegetation is mainly light dependent28, therefore near-source measurements typically exhibit a broad maximum around noon when irradiation is highest29,30. The two isoprene maxima observed here are the result of the sea breeze system, which brings isoprene poor marine influenced air to the site around midday.
Monoterpenes (C10H16) are emitted from vegetation, and are to a large part responsible for the scent of forests. Their sum at the site reached 1.6 ± 1 ppbv as a maximum hourly average at 01:00. Despite monoterpene emissions being also a function of temperature they do not show the same profile as isoprene. Monoterpene mixing ratios were higher at night, decreasing slowly from 03:00 and then sharply at 08:00, and remaining low through the day before rising back to early morning levels in the late evening. While isoprene is formed in the plant and released directly, mainly in response to light, many plant species emit monoterpenes day and night from resin duct storage pools contained in the leaves or needles31. The rates of monoterpene emission therefore depend strongly on temperature, and although lower in the night, continued emission into a shallower boundary layer can generate significant mixing ratios, as has been observed elsewhere (for example in the boreal forest32). As expected, air advected from the more vegetated inland had higher monoterpene mixing ratios than air from the sea. Although, marine monoterpene and isoprene emissions have been observed previously33, in this region terrestrial vegetation emissions clearly dominate. A similar profile (higher nighttime daytime ambient mixing ratio) was reported by Davison et al.34 from the Mediterranean forest of Castelporziano, however, the average daytime and nighttime concentrations were 3 times lower than reported here.
Xylenes (Fig. 2), benzene, toluene, and sum of trimethylbenzenes (Supplementary Fig. 2) are aromatic compounds predominantly emitted from anthropogenic sources, for example fossil fuel use. Mixing ratios increased in the morning at 8:00, probably due to traffic emissions, and decreased steadily between 9:00–12:00, as the wind direction changed. They remained low in the afternoon as the wind came from the sea, reflecting the expected absence of significant sources from this sector. The mixing ratios of all compounds increased again after 17:00, slowly for benzene and toluene and faster for xylenes and trimethylbenzene, suggesting different anthropogenic sources of such species (i.e. road traffic and industries). Although evidence for biogenic aromatic species has been documented35,36, here no evidence of natural sources of aromatic compounds was found. The maximum mixing ratios for benzene, toluene, sum of xylenes and trimethylbenzene, were 0.16 ± 0.08 ppbv, 0.32 ± 0.7 ppbv, 0.22 ± 0.09 ppbv and 0.11 ± 0.9 (hourly average maximum ± 1σ standard deviation), respectively. Similar results were found for benzene and toluene at a rural Mediterranean site in Spain (0.19 and 0.41 ppbv, respectively26).
Chiral monoterpenes
Many monoterpenes exist in two chiral forms, meaning they exist in nature as two non-superimposable mirror image forms (enantiomers), often with different biological activities. Insects can perceive each enantiomer differently and each may act as entirely distinct chemical signals37. We investigated the abundance of the prevalent chiral monoterpenes at the home aviary and at three ecosystems surrounding the aviary North (lake), East (mixed forest) and South (pine forest). This was to test whether unique mixtures could be ascertained at each site which is a condition of the mosaic hypothesis. At all sites, including the aviary, the (−) configuration dominated over the (+) for the monoterpene species α-pinene, β-pinene and limonene. This has been also observed in the tropical rainforest for α-pinene38. Relative ratios for the same monoterpene enantiomeric pair do not change significantly across the investigated sites, however, their absolute concentrations and relative abundance to the other terpene species do differ (see Fig. 3).
Figure 3
Volume mixing ratio (VMR) distribution of enantiomers from three ecosystems surrounding the birds’ aviary and at the birds’ aviary. Category plots show mean measured values with their standard deviation from three days of sampling during summer 2017. Box plots show mean campaign values (marker), median (line), interquartile range (box) and 10th and 90th percentiles (whiskers) from 2 months intensive field campaign at the birds’ aviary in spring 2018 (home) and from three flights conducted at 180 m altitude over the region during the airborne campaign. The x axes shows the retention time (min) of the chiral molecules in the gas chromatography mass spectrometer (GC–MS) and shows the good resolution achieved with the method. The left y axes indicate the volume mixing ratio of all molecules, except for (−)limonene measured at the aviary, whose VMR is indicated by the right y axes.
Full size image
At the aviary, (−)limonene was the most abundant species (0.89 ± 1.7 ppbv and 2.5 ± 2.7 ppbv), followed by (−)α-pinene (0.42 ± 0.57 ppbv and 1 ± 1 ppbv), (+)limonene (0.26 ± 0.29 ppbv and 0.37 ± 0.42 ppbv) and (−)β-pinene (0.17 ± 0.45 ppbv and 0.27 ± 0.49 ppbv, values indicate mean daytime and nighttime values ± 1σ standard deviation, respectively). The lake site emitted the least monoterpenes, followed by the pine forest and the mixed forest, the latter showing the highest concentration of measured terpenes. Since (−)limonene dominates the terpene blend measured at the aviary, we can surmise that the surrounding pine forest had a strong impact on the air chemical composition at the site (Fig. 3). A similar suite of biogenic molecules was found at all locations, albeit with differing ratios. In other words, there were no clear unique chemical markers for a particular area. Therefore, the source of biogenic molecules near to the aviary has the potential to interfere with or mask, odours from more distant sites. These findings are not consistent with the original “mosaic” hypothesis in which each location in the region has a unique chemical signature that can be determined from the aviary.
Olfactory maps: spatial and temporal distribution of VOCs
Soon after fledging, the young pigeons are housed in a loft with an associated aviary, and from the end of May are allowed to perform spontaneous flights around the loft typically until October, when the season of the experiments usually ends. During the first months after fledging39 young pigeons are exposed to the changing chemical conditions at the home site and according to the Papi olfactory hypothesis, they learn to associate wind-borne odours with wind directions thereby generating an olfactory map. It is thought that this olfactory map is continually updated throughout their life depending on conditions experienced40,41. A visual representation of this spatio-temporal information can be expressed in the form of a bivariate plot of the measured volatile organic compounds. Bivariate polar plots represent the mean campaign mixing ratio of a given compound as a function of the wind direction (angle) and wind speed (radius), similarly to a wind rose they can highlight spatial gradients in the surrounding of a measurement site. Figure 4 shows that methanol (a biogenic compound and biomass burning marker) mixing ratios were higher for higher wind speed and easterly winds (from inland). However, mixing ratios were elevated at both high and low wind speeds; indicating some methanol sources were also local. Dimethyl sulphide mixing ratios were larger for higher wind speeds originating from the North–West and South–West sectors, spanning from the mouth of the river Arno to the sea. The largest mixing ratios for isoprene came from inland, in particularly from the West–North–East for higher wind speeds and the South–East for lower ones. Smaller mixing ratios are found from the sea for large wind speed (North–West), suggestive of a weaker marine isoprene emission source. Small marine isoprene emissions have been measured previously, especially in chlorophyll rich waters12,33,42. Monoterpene mixing ratios are higher at lower wind speeds, mostly for winds coming from South–East but also East–North–East and South–West. The polar graph highlights the multiple sources of monoterpenes, and that the levels at the site are also locally influenced. The anthropogenic aromatic species xylenes and trimethylbenzene were larger when the air was transported from inland, in particular from the South-East, where heavy-traffic roads such as highway E80, the road FI-PI-LI and via AureliaSS1 are located. Additional sources of aromatics appear South of the aviary for high wind conditions, especially for trimethylbenzene. Trimethylbenzene (C6H3(CH3)3), is an aromatic hydrocarbon characterized by a strong odour, which is generally isolated from the C9 fraction of aromatics during petroleum distillation. South of the aviary, in the Stagno industrial area, there is a large petroleum refinery (Fig. 1).
Figure 4
Spatial distributions of measured VOCs reported as protonated masses and identified as: methanol (m/z 33), DMS (m/z 63), isoprene (m/z 69), sum of monoterpenes (m/z 81), sum of xylenes (m/z 107), trimethylbenzene (m/z 121). Mean campaign values of VOCs (colored scale) are showed as a function of wind direction (angle) and wind speed (radius). Figures drawn with R.
Full size image
This air composition measured at the aviary is the air breathed and smelled by the birds during their first months after fledging. The birds are therefore likely to be aware of, mixed biogenic and anthropogenic sources to the South-East, a more biogenic dominated source (with differing composition) to the West-North–East, and a marine source to the West. Thus, the ground based measurement campaign mapped several chemical compounds that vary distinctly at the site according to the wind direction and time of day. Which VOCs are useful for navigation will also depend on their respective atmospheric lifetimes. Following emission to the atmosphere VOCs are oxidized mainly by OH radicals, with small contributions by O3, NO3 and Cl radicals. Methanol, the most abundant species measured, has a lifetime of 12 days (based on OH reactivity only, assuming OH radical concentration of 2 × 106 molecules cm–343). The corresponding lifetime of isoprene is 1.4 h, therefore with a typical windspeed of 2.2 m/s the concentration of isoprene will be reduced to 1/e of the initial value in 11 km, while limonene in 2 km, xylenes in 33 km and DMS in 88 km. Therefore, over distances of ca. 50–100 km, VOCs with moderate lifetimes such as DMS and the aromatic species appear capable of creating regional gradients. The more reactive VOCs are usually photochemically transformed into less reactive products which themselves, or their combination, may create regional gradients. Significantly, the three regional gradients highlighted in Fig. 4, namely DMS, the aromatics and the monoterpenes are not aligned, rather they slope in different directions. Theoretically then, by comparing the level of these compounds to that experienced at the aviary the pigeon may, with reference to the wind-odour experience at the aviary, orient homeward. For example if the DMS level is lower at the release site than at the aviary, and the bird “knows” from the olfactory map learning phase that DMS comes from the West, then the birds homeward direction will have a westerly component. Having multiple gradients available would enable the pigeon to triangulate a homeward direction.
Regional spatial gradients
To verify that the spatial gradients observed from the olfactory maps exist at a larger regional scale we measured the air composition at the altitudes and over distances typically flown by homing pigeons (10–300 m, 100 km). Results from a flight campaign conducted on 26/05/2018 and 27/05/2018, in conjunction with the ground based campaign, are depicted in Figs. 3, 5 and Supplementary Fig. 3. Figure 3 shows the speciation of chiral molecules from ground measurements at four distinct sites (including the aviary) and above the sampled region considering all the airborne samples taken with the three flights. Interestingly, the regionally predominant compound at 180 m is α-pinene, rather than limonene observed at the ground. This is because α-pinene has a longer atmospheric lifetime (2.6 h), in comparison to β-pinene (1.8 h) and limonene (49 min)43. Figure 5 shows the mixing ratio of a biogenic precursor compound (−)α-pinene and the common terpene photochemical oxidation product nopinone44 sampled at different locations above the home aviary, and the release sites. The terpene airborne concentrations compare well with the ground-based ones (higher above the mixed Mediterranean forest, followed by the pine forest, followed by the lake site, as reported in Fig. 3). However, as expected, the terpene airborne concentrations are smaller compared to ground concentrations (40–50 pptv and 80–208 pptv, respectively) consistent with the ground based emissions being progressively oxidized by OH radicals and ozone as well as being mixed with the relatively clean air in the free troposphere above. This means that the birds will encounter strong vertical gradients of primary emitted species during flight with lower concentrations aloft. Furthermore oxidation chemistry will generate entirely different species, complicating the mixture, weakening the primary chemical signal and possibly obfuscating the originally emitted olfactory signals. In Fig. 5 we observe during the afternoon flights (with onshore winds) the nopinone mixing ratios generally increase to the East. This demonstrates that monoterpenes emitted at ground level during the day and in onshore winds are being oxidized to (among other species) nopinone on the time and space scales of the pigeon release experiments. Therefore, spatial gradients of reactive VOC do exist for the primary emitted species and their oxidation products, at a spatial scale compatible with the distances used for navigation experiments.
Figure 5
Spatial distributions of measured (−)α-pinene (a) and nopinone (b) from airborne sampling during three flights conducted on 26/05/2018 and 27/05/2018. Colored scales indicate respectively volume mixing ratio and counts of chromatograms peak area. Figures drawn with Igor WaveMetrics.
Full size image
Pigeon flight tracks and air masses trajectories analysis
To test whether the intensity of atmospheric odour signals could potentially aid in the homing of pigeons, we provide here a first preliminary analysis based on real pigeon tracks and simulated air masses trajectories. We hypothesize that overall stronger gradients of indicative aerial chemicals enable better navigational performance in pigeons18. To do this we examine the GPS logged tracks of homing birds in relation to the general gradients of atmospheric chemicals determined in the previous sections, and to the origin of air at the release sites. Bird release experiments were conducted over six days in summer 2016 and summer 2017 from three release sites simultaneously (Fig. 1). Those experiments were chosen for being the most recent results available for indices analysis, and for being conducted on sunny days, when air temperature, humidity and atmospheric pressure were comparable to those encountered during VOCs measurements (Supplementary Fig. 4). Each release experiment involved 9 or 10 individual birds, released singly every 5–10 min, over approximately 1 h and involved 27–30 individuals released on the same day from three sites (N = 174). The total number of tracks obtained was 143, due to either the loss of the device by some birds or the misfunctioning of the GSM GPS logger. Bird flight tracks were used to determine: the pigeon initial flight direction, the homing capabilities en route with the homing efficiency index (HEI45) and for the whole track with a mean aggregate azimuth penalty (MAAP); see Methods and supplementary information. Due to lack of power and shelter at the release sites, chemical information of atmospheric composition was not available during those experiments, therefore a meteorological approach, based on modelled air trajectories, was developed in order to examine how air mass transport was related to the birds’ homing performance. Specifically, we generated forward and backward trajectories of the air masses for each release day, for the site and time of release. Forward trajectories analysis showed that the prevalent wind direction during the days and time of pigeons’ flights was from west (supplementary information). We cannot yet test the bouquet of potential chemical information, but concentrate here on one traceable chemical gradient: DMS (Dimethyl sulphide) emanating from the Tyrrhenian sea. DMS, a chemical compound of marine origin (see Supplementary Fig. 5), is here identified to be a suitable candidate for homing by olfaction. Figures 2 and 4 show that it was among the most abundant compounds measured during daytime at the aviary, it follows a regular pattern of emission, is known to be detectable by birds, and it is atmospherically stable enough to survive transport over longer distances, likely decreasing along the West–East direction. Air masses reaching the aviary diurnally during the ground campaign were found to spend considerable time ( > 20 h in most cases) in the marine boundary layer in the past 24 h (see table 1 supplementary material). For these reasons, we tested if more marine-influenced air masses can influence the pigeon initial orientation (Fig. 6) and the homing path en route (homing efficiency index, supplementary material). We examined where each air mass was located (during 24 h prior to the release) reported in latitude and longitude coordinates and determined the time the air masses spent in the marine boundary layer, and the time the air masses spent over non-marine areas. As shown in Fig. 4, pigeons at the home loft are exposed to DMS associated with westerly winds; easterly winds are unlikely to carry DMS. According to the olfactory navigation hypothesis5 it is expected that a low atmospheric level of DMS at the release site is likely to produce a bias towards west in the orientation of the birds, regardless of their ultimate home direction. To test this hypothesis, the deviation from west of the individual mean vectors computed on the initial part of each track (within 10 km from the release site) is tested against the ratio of the time the air masses spent over sea versus over land (ratio sea/land) and the opposite (ratio land/sea). The Spearman ranking test highlighted that the deviation from west is positively correlated to the sea/land ratio of the air masses trajectories (n = 135, S = 0.163, p More
