HR: 0800h
AN: B51D-0247    [Abstracts]
TI: Seismic Census Technique for African Elephants
AU: * Wood, J D
EM: jasonwood@centurytel.net
AF: Department of Geophysics Stanford University, Mitchell Bldg., Stanford, CA 94305-2215 United States
AU: O'Connell-Rodwell, C E
B51D-0247 AF: Department of Geophysics Stanford University, Mitchell Bldg., Stanford, CA 94305-2215 United States
AU: Klemperer, S L
B51D-0247 AF: Department of Geophysics Stanford University, Mitchell Bldg., Stanford, CA 94305-2215 United States
AB: Large mammal populations are difficult to census and monitor in remote areas. In particular, elephant populations in Central Africa are difficult to census due to dense forest making aerial surveys impractical. Conservation management would be improved by a census technique that was accurate and precise, did not require large efforts in the field, and could record numbers of animals over a period of time. We report a new detection technique that relies on sensing the footfalls of large mammals. Geophones were used to record the footfalls of elephants and other large mammal species at a water hole in Etosha National Park, Namibia. We were able to discriminate between species using the spectral content of their footfalls with an 85% accuracy rate while only using a single geophone. This was done using correlation coefficients comparing the shape of the spectra for various species. An ANOVA found significant differences between these correlation coefficients (F_4,1785 = 147.78, P = 0.000). An estimate of the energy created by passing elephants (the area under the amplitude envelope) can be used to estimate the number of elephants passing the geophone. Our best regression line plotting number of elephants versus energy recorded in the geophone explained 55% of the variance in the data. Much of the unexplained variance is due to the variation in distance from the geophone to the passing elephants. By subjecting the recordings to a narrow band-pass filter and using beamforming techniques on array data, we believe that we can control for the variation in distance between animal and geophones, and thus achieve better estimates of the number of animals passing the array. Using 7 pairs of geophones in a linear array, we offset the pairs of time series to correspond with the time delay associated with the signal intersecting the pair of geophones at increments of 2 degrees. The offset time series were then summed and the RMS value calculated. The largest RMS value was then taken to indicate the most likely angle from which the signal intersected the pair of geophones. Lines were then plotted from the pairs of geophones at the angle with the highest RMS. The centroid of the line intersections was taken to be the location of the source signal. Thus far we have managed to localize the signal source within 15 meters of its true location.
DE: 0416 Biogeophysics
DE: 0452 Instruments and techniques
SC: Biogeosciences [B]
MN: Fall Meeting 2005