Elephant low-frequency calls and atmospheric conditions that influence the transmission and detection of these calls were recorded at a fixed location over a period of about 3 weeks at the end of the dry season in the Etosha National Park, Namibia. A bimodal distribution in elephant call detections is observed, with the primary maximum (42% of all calls) occurring in a 3-h period following sunset and a secondary maximum (29% of all calls) in a 2-h period following sunrise.
This distribution in calls detected is shown to depend upon marked and regular changes over 24 h in near-surface atmospheric stability and velocity, which determine propagation ranges. The observed bimodal distribution of calls detected depends upon these changes in atmospheric conditions as well as the location of the caller and the rate of calling. The findings are supported by results from an atmospheric acoustic model but are at variance with observations of the number of calls made from collared elephants in markedly different habitat.
Detection of calls heard at a location remote from the caller represents a valuable and noninvasive research and applied tool that, however, must take note of atmospheric conditions that govern the propagation and reception of such signals.
A number of terrestrial animals produce loud calls at frequencies below 100 Hz. These calls can carry long distances because they lose very little energy to atmospheric absorption. Elephants are especially noteworthy because they produce some of the loudest animal sounds, and much of the energy in their calls lies between 10 and 35 Hz (Payne et al. 1986). The lower portion of this band is inaudible to human ears; sounds with energy between 1 and 20 Hz are called near-infrasound. Elephants are able to hear near-infrasonic signals; playback experiments using infrasonic calls at a source level of 112 dB SPL (a reference pressure of 20 μPa) elicited responses from elephants at distances of up to 2 km (Langbauer et al. 1991). Poole et al. (Poole et al. 1988) documented source levels of up to 117 dB SPL from African elephants, which implies a potential range of detection of 4 km if spherical spreading were the dominant factor in call attenuation with distance.
Spherical spreading losses can provide poor predictions of acoustic attenuation when there are strong vertical gradients of temperature and wind in the first few hundred meters of the air above the ground (Wilson et al. 2003). Day-to-night changes in surface temperatures and winds and the vertical gradients of these quantities occur on a regular basis over land. These changes are present in virtually all habitats but are particularly pronounced in the dry savanna occupied by Loxodonta africana. Garstang et al. (Garstang et al. 1995) and Larom et al. (Larom et al. 1997) used atmospheric observations of temperature and wind velocities in the lower atmosphere (<400 m) together with an acoustic model to calculate the maximum distances over which elephant calls could be heard by other elephants. They assumed a hearing threshold of 50 dB SPL (Heffner and Heffner 1982), a value that is supported by the Langbauer et al. (Langbauer et al. 1991) playback experiment assuming spherical spreading losses: 112 dB – 20 log10 (2000) = 46 dB.
The model predicts that an elephant call with a source level of 117 dB SPL would be heard at ranges from as little as 1 km in the middle of the day to more than 10 km an hour after sunset. Nighttime conditions for propagation and hearing of low-frequency sounds are always better than daytime conditions. During the day, a negative temperature gradient with altitude causes sound to refract away from the ground, and turbulent dissipation and elevated wind noise present additional obstacles to long-range acoustical communication. At night the temperature gradient can become positive (a temperature inversion). A strong temperature inversion effectively traps sound energy in a thin layer of the atmosphere, so spreading losses become cylindrical. In addition, ambient noise is dramatically reduced because near-calm conditions prevail at the surface. Later at night, propagation conditions are degraded when strong winds above the inversion layer—the nocturnal jet—cause episodic incursions of higher winds that penetrate the inversion layer and reach the surface. Conditions improve in the early morning, as the nocturnal jet decays and surface winds once again approach calm conditions.
In this paper we show that calls detected by a fixed microphone array follow a bimodal distribution as predicted by the atmospheric acoustical model. This suggests that calls detected are strongly influenced by conditions in the near-surface atmosphere. Calls made as determined from microphones attached to collars on elephants have a diurnal cycle with less evidence that the calling rate of the animal is dependent upon atmospheric conditions. Well-recognized patterns of dawn and dusk choruses displayed by many species provide intriguing evidence of a possible coupling between calling, hearing, and atmospheric conditions. Clear evidence of behavioral responses to atmospheric conditions, especially in relation to the “active space” of vocalizations and the degradation of signals with distance, has broad implications for the study of animal behavior and ecology (Payne et al. 2003; McComb et al. 2003; Langbauer 2000). Decisive demonstration that animals adapt calling activity to propagation conditions requires a cross-disciplinary field experiment that is costly and complicated. Animals must be tracked accurately, a comprehensive record of acoustical activity must be obtained, and atmospheric measurements must be collected that can support accurate acoustical models. A pilot study was therefore carried out in 1999 to obtain simultaneous observations of elephant calling behavior from a fixed array of microphones and near-surface atmospheric conditions. This pilot program was conducted at the end of the dry season in September–October 1999 in the eastern part of the Etosha National Park in Namibia. The results are reported here.
2.1. Field site
The field program was based at the Namutoni Horse Camp (18°47′S latitude, 16°57′E longitude) and the Mushara water hole (18°32′S latitude, 16°33′E longitude) in the far eastern end of the Etosha National Park, Namibia, in September 1999 (Garstang et al. 2000; Garstang and Fitzjarrald 2002). An observation tower and network of microphones were installed at the water hole. The platform, housing observers, and recording equipment were located 50 to 100 m, depending upon water level, from the Mushara water hole (Figure 1). A second artesian water hole, Kameeldoring, lies 12 km ESE of Mushara. These two water holes have been the preferred drinking locations for decades and probably serve a region of about 3000 km2 during the dry season (April–September). With elephant population densities of between 0.4 and 1.5 animals per 10 km2, this yields a population of between 120 and 315 animals (Lindeque 1988). In the dry season, the elephant population visiting Mushara and Kameeldoring is stable; there is no evidence of significant movements of elephants into or out of the area. Dieudonné (Dieudonné 1998) used isotopes of carbon, nitrogen, and oxygen in ivory from tusks of known place and time of origin to determine the possible role of climate, vegetation, and water quality in the elephant’s diet in the Etosha National Park (ENP). Oxygen isotopes indicating differences in water quality in the dry season reflect the spatial distributions of the elephants. The area served by the two primary water holes, Mushara and Kameeldoring, lies within one of four regions in the ENP identified by the isotopic signals obtained by Dieudonné.
2.2. Meteorological measurements
The meteorological measurements were made at the Namutoni Horse Camp, 30 km away from Mushara. The terrain and habitat of eastern Etosha is flat, featureless sandveld with sparse bushveld vegetation with a mean canopy height of 3.0 m (Figure 1). The meteorological measurements made at Namutoni provide an accurate picture of conditions at Mushara, because this terrain and surface cover create relatively uniform near-surface atmospheric conditions.
Continuous measurements were taken for 18 consecutive days in September 1999 to provide a record of the diurnal cycle of atmospheric conditions that govern the transmission and reception of low-frequency sound. Three systems were operated at the Namutoni Horse Camp. A 7.5-m meteorological tower provided a detailed measurement profile near the ground. Table 1 provides details of the tower sensor suite. A 30-m3 tethered balloon lofted temperature and humidity sensors to 200 m. The tethered-balloon temperature soundings focused on the transitions at sunset and sunrise to depict the rapid changes in atmospheric stratification. Seventy-six pairs of up and down soundings were made over 18 days. An acoustic sounder (sodar) was used to provide an echo intensity profile, and the wind vector averaged over half-hour periods over 10-m vertical intervals to a height of 200 m. The height of the inversion is determined from the sodar echo intensity profile and is located at the base of the enhanced echo region.
Temperature profiles from the tethered balloon clearly defined the base of the inversion. Using the tethered-balloon temperature profiles, we found that the base of the temperature inversion was adequately described by multiplying the sodar-estimated height by 0.62. In this way, the record from the acoustic sounder could be used to produce a continuous record of the times of formation and decay, duration, and height of the nocturnal inversion.
The temporal and vertical spatial resolution of these measurements were dictated by the computational requirements of the following quantities: sensible heat flux across the ground–air interface and through the first few tens of meters above the surface; stability of that air column based on these fluxes; the times of transition from stably stratified nocturnal to unstable, turbulent daytime conditions; the vertical gradients of temperature and wind; and the delineation of the lapse rates or change of temperature with height such that the presence, magnitude, thickness, and height of the base and persistence in time of the nocturnal inversion could be determined. Measurements of humidity and the components of solar and terrestrial radiation were made but not used in any direct fashion to describe conditions for the propagation and reception of low-frequency signals. Computational procedures for the determination of these quantities can be found in Garstang et al. (Garstang et al. 1997), Larom et al. (Larom et al. 1997), and Garstang and Fitzjarrald (Garstang and Fitzjarrald 1999).
2.3. Acoustical measurements
Elephant low-frequency calls in the range of 0 to 700 Hz were recorded from eight microphones deployed radially around the outside of an observation platform at the Mushara water hole (Figure 1). The array of eight microphones permitted arrival times and relative amplitudes of signals to be detected. This was used to determine whether the elephant was in the immediate vicinity of one of the sensors and to ensure that the calls used in the analysis were sufficiently loud to propagate over long distances.
The acoustic sensors consisted of Panasonic WM61A electret capsules. These were recess mounted on a thick wooden plank with a fine screen over the microphone. The whole system was buried such that the top of the plank was flush with the surface of the sand. A thin layer of sand was brushed over the screen and board. This arrangement was relatively cryptic and diminished the risk of damage from animals. For low-frequency signals, a few millimeters of dry sand impart negligible attenuation. Measurements made at ground level are representative of what the elephants can hear (ear height of approximately 2–3 m), because of the very long wavelengths of these low-frequency signals. Microphones deployed at or within 1 m of the surface have been successfully used to record elephant calls by a number of researchers (Langbauer et al. 1991; Payne et al. 2003).
The signals from these units were amplified by 40 dB and low-pass filtered at 700 Hz. The amplified signals were fed through buried cables to a PCMCIA data acquisition card (Computer Boards, Inc.) hosted by a Windows 98 notebook computer. A custom data acquisition program digitized the signals at 2 kHz. The multichannel sound recordings were time and date stamped using the notebook system clock, which was synchronized with a GPS clock. The intrinsic noise floor of the WM61A is nominally 30 dB SPL, and the preamplifier gain was set such that this corresponded to the least significant bit of the A/D converter. Even allowing for attenuation of the signals due to the thin layer of sand over the microphone, this system should have been at least as sensitive as the assumed 50 dB SPL hearing threshold of the elephants. The 12-bit A/D system clipped the signals at a received level of 102 dB SPL.
An experienced technician used specialized software developed by the Cornell Bioacoustics Research Program to analyze the acoustic data. Figure 2 shows an example of the spectrographic displays generated from these eight channels. A technician manually searched for and marked (boxes in Figure 2) the characteristic time-frequency signatures of the elephant sounds. Each call was marked once, on the channel presenting the strongest signal. The maximum time delay across the array, in terms of the separation of the microphones and the speed of sound, was less than 0.5 s. Recordings were made for 18 days, from 11 to 28 September.
Strong, repeatable diurnal patterns in atmospheric structure are evident in the meteorological measurements. Figure 3 shows a sample of tower measurements that reflect the day/night changes in temperature and wind speed as a function of time and height. Figure 4 shows a typical temperature and humidity profile as obtained from the tethered-balloon sensors. Figure 5 shows a combined time–height section of acoustic sounder (sodar) and tethered-balloon measurements.
Twice daily the surface and air temperature are equal (vertical lines in Figure 3 and subsequent figures). These neutral conditions bound the stable and the convective conditions of night and day. Temperature fluctuations, T′ (Figure 3), reach minimum values, and vertical and horizontal winds are low in the evening transition (about 1800 LST). Figure 3 shows that with the onset of nighttime stability, there is a marked decay in temperature and vertical and horizontal velocity fluctuations, all of which approach zero. Though some wavelike behavior is observed in the temperature fluctuation at this time, near-calm conditions typically prevail until after midnight. The diurnal cycle of temperature and wind shown in Figure 3 is characteristic of most dry-season days and nights.
The morning and evening neutral periods divide the 24 h into strong daytime heating and nocturnal cooling of the atmosphere at and immediately above the surface. Three heat flux distributions over 24 h are shown in Figure 6.
Average daily values over an entire dry season for Tsumeb (∼100 km from Namutoni) from Walk and Wieringa (Walk and Wieringa 1988).
Eddy flux half-hourly median values from the 3-week ensemble of 1-min grab samples. A “grab sample” for every minute partitioned into every one-half hour of the 24 h for 21 days of w′T′ was used to arrive at the 24-h distribution of heat fluxes. This assumes that the daily heat flux pattern is stationary, a plausible assumption for our location (see, e.g., Kaimal and Finnigan 1994, 254–257).
- A bulk aerodynamic estimate of heat flux over a single 24-h period (14–15 September 1999) was made using the expression U7.5 and T7.5 are the half-hourly mean wind speed and temperature at 7.5 m, T0.01 is the half-hourly mean soil temperature at 0.01-m depth, and ρ is air density. A constant exchange coefficient of 0.002 was assumed.
The agreement in magnitude and timing over 24 h of the heat fluxes shown in Figure 6 between the long-term (dry season) field experiment (21 days) and single day (14–15 September 1999) shows the day-to-day reproducibility of these stable and unstable conditions.
Based upon the heat flux median values (red line in Figure 6), we conclude that stable conditions are initiated near 1800 LST in the evening and persist to between 0800 and 0900 LST in the morning. Once surface heat flux goes to zero and then negative (i.e., heat is being lost from the atmosphere to the surface), buoyant mixing ceases and the atmosphere is stratified into a layer of stable cold air at the surface and warm air overlying this cold layer. The vertical temperature gradient changes from negative during the day to positive (inverted) at night.
Figure 7 shows the average behavior of the nocturnal inversion over the nighttime hours based upon sodar and tethered-balloon measurements. Throughout our analysis we took advantage of the regularity in the diurnal cycles at Etosha to form composites. In Figure 7, the dots represent all of the data reported by the REMTECH sodar as the “INVI” parameter. This value is calculated based on the CT2 echo strength parameter, where CT2 is the sodar’s range-corrected estimate of the squared “structure parameter” (see Stull 1988, p. 301). The sodar used did not have a radio acoustic sounding system (RASS) unit that would provide direct estimates of the temperature profile. Determination of the height of the inversion depends, therefore, upon fluctuations in the refractive index and not on the temperature itself. The resolution of the sodar is nominally 10 m. Details of the program that calculates the INVI are proprietary, and there is no obvious way theoretically to relate inversion base to the INVI parameter. To find the regular central tendency in the data, we chose the median value by hour of day (Figure 7). The key value of the sodar is its regular recording of data; it serves as a complement to the tethered-balloon data, which could only be operated for selected periods. The 0.62 factor was selected to follow closely the lower envelope of the sodar data, and this also fit fairly well with the direct estimates of inversion base obtained in this study using the tethered balloon (X’s) and also with inversion base values reported by an earlier study (Z’s) (Figure 7). Once the inversion has deepened, the sodar adjusted height is valid until after the inversion decays (>0900). When the inversion strength reaches about 4°C (100 m)−1 (Figure 7), inversion heights stabilize between about 75 and 100 m above the ground. These conditions cause strong downward refraction of sound that last through the night until 0900 LST, some 2 h after sunrise. The persistence and duration of the nocturnal inversion is in agreement with the delineation of the surface fluxes (Figure 6).
Wind speed influences the transmission and reception of sound traveling in the atmosphere. Figure 3 illustrates a typical diurnal cycle of wind speed at two heights, 2.5 and 7.5 m. Highest winds are seen during daylight hours some 4 h after sunrise to about 1 h before sunset. Near-calm conditions, especially at the 2.5-m level, occur just before sunset, persisting until after midnight. Intermittent gustiness is, however, seen soon after 2000 LST at the 7.5-m level, growing during the remainder of the night. A secondary minimum in wind speed occurs near sunrise but unlike the evening primary minimum, seldom goes to calm.
Explanation of the observed regular changes in surface wind speed over the 24 h rests upon the vertical distribution of winds in the atmosphere and the transfer of horizontal momentum. Friction at the ground surface slows winds, and usually a gradient exists from low wind speeds at the surface to higher winds at a level above the effects of surface friction. This boundary layer of the atmosphere is not of constant thickness. As daytime heating is initiated, buoyant air rises and less buoyant, colder air sinks. Such mixing transfers the low momentum at the surface upward and the higher momentum downward. Winds at the surface increase quickly with this heat-driven mixing, producing a daytime maximum in surface winds.
As radiative cooling sets in before sunset the cold air forming at the surface decouples the air above from the surface layer. Contact with the rough surface is removed, and the air immediately above this cold (dense) layer is freed of surface friction and accelerates. The cold air at the surface follows the terrain and any topography present represents sloping surfaces down which the air riding on the cold surface layer accelerates. In the eastern Etosha National Park the land slopes upward from the pan 100 km to the mountains near Tsumeb. This slope is about 1:1000 but is sufficient enough to accelerate air toward the park in a northwesterly direction. The combined effect of removal of friction and slope-induced accelerations produces a nocturnal jet that can reach speeds in excess of 10 m s−1 in the middle of the night and during the early hours of the morning.
As the nocturnal jet becomes established on the inversion surface, the vertical gradient across this boundary of the horizontal wind increases to a point where wind shear overrides static stability. Downward mixing starts to occur and the higher momentum above the inversion is carried in bursts to the surface. The frequency distribution of winds over 4 m s−1 is shown in Figure 8. Of the 18 nights of observations only 3 were free of gusts over 5 m s−1 while on a few nights, gusts of more than 10 m s−1 occurred.
In the quiet period in wind speeds occurring after sunset (1900 to 2200 LST) gusts above 4 m s−1 occur less than 4% of the time. The frequency of gusts above 4 m s−1 increases dramatically between 2300 and 0500 LST, reaching a nighttime peak of more than 35% of the time at 0500 LST. A decrease in the frequency of gusts occurs about sunrise (0700 LST) as shearing stresses and slope effects decrease, with the percentage of gusts over 4 m s−1 increasing rapidly to the 24-h maximum at 1000 LST. These episodic bursts of wind energy have at least two impacts. First, they increase the ambient noise everywhere. Second, they cause increased flow noise when passing over an elephant.
Most elephant sounds were visible on all eight channels, and none of the elephant signals were clipped. Thus, most elephant calls were not produced in the immediate vicinity of any element of the array. The time delays, as seen in the spectrograms, formed sensible patterns given the spatial extent of the array. There was no ambiguity in recognizing the same call on all channels.
The hourly distribution of 1328 low-frequency elephant calls recorded over 7 days is shown in Figure 9. It is important for subsequent discussion to emphasize that the calls shown in Figure 9 were calls detected (and probably heard by elephants at that location). The calls detected are a function of the range over which elephants can hear calls of conspecifics. Since the range, and thus the area, over which one elephant can hear another elephant changes as a function of atmospheric acoustic conditions, the number of calls detected at a fixed location varies with atmospheric conditions even though the calling rate and number of calling elephants may remain constant.
Calls detected (Figure 9) cluster in the hours following sunrise and sunset. While the highest number of detected calls in any given hour of the day falls in the 2 h following sunrise (184 and 196, or 14% and 15%, respectively), the largest number of detected calls occur in the 3 h following sunset (558, or 42%). Just over 70% of all calls detected fall in the 3 h following sunset and 2 h following sunrise. Of the remaining 30% of all detected calls, 24% occurred at night. Including the calls from the two maxima, two-thirds of the calls detected were made at night.
Figure 10 compares the 8-day record of call detection with the surface wind speed data. With few exceptions, the elephant calls detected are conspicuously clustered in intervals when the wind speed reached a local minimum. The most intensive bouts of detected calls occurred during intervals when the wind speeds reached their minimum values.
3.3. Movement patterns
Records were kept on the Mushara observation platform of sightings of elephants at the water hole for seven consecutive 24-h periods during the experiment. Table 2 shows the sightings of lone bulls or small groups of bulls (B), breeding herds (H), and lions (L). During daylight hours an accurate count of the number of animals in a breeding herd could be determined. At night the presence of a breeding herd at the water hole was easily determined, but the number of individuals making up the herd was in doubt. The largest number of sightings both by category and number were at and following sunset. No breeding herds were seen at the water hole during the day. Except for two bulls, no elephants were seen in the 3 h preceding and following sunrise.
One of the authors (C. B.) actively tracked the elephants in the northeastern sector of the ENP using a light aircraft and radio signals from collared elephants. The tracking program was carried out for a total of 10 months over two dry seasons bracketing the period of the field experiment. During this entire period the morning movements of elephants were always away from the water hole, while the afternoon movements were always toward the water hole. Over the period of 10 months a particular herd (led by the matriarch Knobnose) was frequently within 1 km of the water in the 2 h before sunset, arriving at the water near sunset. When tracking in the morning, takeoff was at sunrise. Elephants found within 1 h of sunrise were usually no more than 3 km from a water hole, moving away from the water. In the 10-month period elephant herds were never seen at the water in the 2 h after sunrise.
A pronounced bimodal distribution over 24 h is evident in low-frequency elephant calls detected by a fixed array of microphones at a water hole. The primary maximum in detected calls (1900–2100 LST), shown in Figure 9, matches the model-predicted time of optimum atmospheric acoustic conditions (Figure 11). The secondary maximum in detected calls (0700, 0800 LST) occurs later than the predicted secondary maximum in Figure 11. However, the more complete atmospheric dataset collected during this project documented the persistence of inversion heights and strengths until nearly 0900 LST (Figure 7), which means that the model-predicted second period of optimum conditions will occur close to 0900 LST. The low daytime (1000–1700 LST) and higher nighttime (2200–0600 LST) frequency distributions in detected calls in Figure 9 is consistent with the modeled predictions.
Model-predicted 24-h distributions of the area over which calls can be heard (Figure 11) depends upon lower-atmospheric conditions and their cycles over 24 h. Agreement between the distribution of detected calls (Figure 9) and model-predicted areas of detection (Figure 11) strongly suggests that the detection of calls depends upon repeatable and pervasive cycles in the atmospheric acoustic state. Habitat (terrain and vegetation) alters cooling and heating rates at the surface and hence atmospheric structure and acoustic conditions. The habitat selected in this study was chosen to maximize atmospheric effects on the transmission and detection of low-frequency sounds, but atmospheric models will be relevant to acoustic propagation at all locations.
The diurnal distribution of wind speed and wind gusts at the surface is closely linked to atmospheric stability. The daily cycle in wind thus reflects stability and amplifies the effects of stability upon the transmission and reception of low-frequency sound. This important role of wind was not adequately treated in previous models of long-range elephant communication (Garstang et al. 1995; Larom et al. 1997). Calm or low-wind conditions in the early evening at the surface reduce wind noise everywhere, as well as the flow noise at the elephant’s ear. Assuming that elephant hearing thresholds are not the limiting factor for sound detection (ears have evolved to take advantage of the quietest possible conditions), this reduction in noise levels corresponds to increased range of communication. Wind-driven refraction also shortens upwind and lengthens downwind communication ranges, both in the presence and in the absence of an inversion layer. The data presented here (Figure 10) suggest that calls recorded are directly influenced by atmospheric conditions; the number of calls detected is more tightly related to wind speed than to time of day. In general, higher winds and gusts were observed in the middle of the day and between midnight and dawn, at the times when fewest calls were detected. In this study, the best daytime wind conditions were less favorable for long-range communication than the worst nighttime wind conditions.
The number of calls detected at the Mushara waterhole reflects three dynamic factors: the proximity of elephants to the waterhole, the range of detection due to meteorological effects, and the rate of elephant calling. Elephants were most often observed at Mushara near and immediately following sunset. Except for one interval involving two bulls, elephants were never seen at the waterhole during the day. These local observations are congruent with radio-tracking data that found movements away from waterholes in the morning, and movements toward waterholes in the afternoon. Thus, proximity likely contributes to the maximum in calls detected in the early evening. The decline in detected calls late at night and the maximum in recorded calls shortly after dawn are not readily explained by proximity. Elephants were not seen at or near the water hole at the time of the early morning maximum in detected calls (Figure 9). This maximum in detected calls must depend upon changes in detection range, changes in calling rates, or both.
Since elephants as well as instruments detect calls traveling through the atmosphere, reception will be subject to changes in atmospheric state. Because the range, area, and hence the potential number of calling elephants heard, change as a function of a regular cycle in atmospheric conditions it follows that the detection of calls will be subject to the same cyclic changes. This cycle in detection affects both elephant behavior and our interpretation of a record of detected calls. Calls from a female elephant in estrous could be heard over an area of 300 km2 at the time of day when the range at which the call can be heard approaches 10 km. At the times when the range of detection of the same call decreases to 1 km the area covered by the call is reduced by two orders of magnitude to 3 km2 (Garstang et al. 1995; Larom et al. 1997). A 100-fold increase in the active space of an acoustical advertisement dramatically expands the set of potential mates. In terms of conservation, any census of elephants based upon calls detected must either know the location of every caller or be able to compensate for the daily variation in range of detection and calling activity.
Calling activity is probably contagious. Elephants are almost certain to call more often when they hear more calls. It is thus likely that the observed distribution of calls heard at this Namibian site is a product of the proximity of elephants to the recording site, changing atmospheric acoustic conditions, as well as changes in elephant calling behavior.
The Namibian record of calls detected or heard is in contrast with the only other known record of calling by elephants in their natural habitat over 24 h. Figure 12 shows the number of loud, low-frequency calls made by 14 collared female elephants in the Sengwa Reserve in Zimbabwe (W. Langbauer and K. Payne 2000, personal communication). This record is of calls made (Figure 12) as opposed to calls detected (Figure 9) or heard.
An early evening maximum is seen in both records. The distribution of calls heard in Namibia reflects the dependence of range upon atmospheric acoustic conditions. The Zimbabwe record of calls made after the early evening peak shows lots of calling during daylight hours when acoustic propagation conditions in Namibia are worst and few calls about dawn when the Namibian conditions for propagation are quite good. The Zimbabwe record of calls made, however, was obtained in terrain and habitat substantially different from Namibia. Concurrent meteorological observations would be needed to draw any conclusions about the dependency of these calls made upon the atmospheric state.
More comprehensive (acoustic, meteorological, and behavioral) observations are needed to determine the relationship between calling rates and atmospheric conditions. It can, however, be concluded with certainty that atmospheric acoustic conditions play a major role in determining the number of calls detected at any fixed location. In the absence of concurrent knowledge of the location of the caller, simple conclusions as to the number of animals calling cannot be drawn from a fixed detector (human or instrumental) record alone. Any census of animals based upon the number of calls heard by fixed detector should be adjusted for atmospheric bias. Methods need to be developed or refined that quantify this bias particularly when introduced by regular and pervasive atmospheric daily cycles.
Both the fieldwork and analytical work was supported by grants from the National Geographic Society and the Knickerbocker Foundation. The Department of Environment and Tourism of Namibia granted permission to carry out the work in Etosha National Park and made facilities available at Namutoni and the Mushara water hole. We wish to express our gratitude and thanks to these organizations for their support.
The field work could not have been carried out without the extensive support given the project and ourselves by Ginger Mauney. Elsabé Garstang, Helen Mohrmann, and Gwen Spicer were full partners in carrying out the field measurements. Johannes Kappies and other National Park personnel assisted with many of the field operations. Logistical and administrative support as well as preparation of the manuscript was provided by Mary Morris. Acoustic data analysis was provided by Dimitri Ponirakis and Melissa Fowler. Katharine Garstang provided graphics assistance. We wish to express our heartfelt thanks to all of these individuals for their help.
* Corresponding author address: Prof. Michael Garstang, Department of Environmental Sciences, Clark Hall, P.O. Box 400123, Charlottesville, VA 22904. firstname.lastname@example.org