Illustration of the sensitivity-accuracy trade-off.

<p>(a) Top row: Localization entropy plots for <i>M. microtis</i> as a function of echo strength () and azimuth-elevation position. The entropy in the centre, i.e. the most sensitive region, reaches a minimum around echo strength of . For echo strengths above entropy in the periphery (i.c. low elevation positions) is lower than around 0 degrees azimuth and elevation. Middle row: Plots of the localization entropy of <i>R. rouxii</i> as a function of echo strength () and azimuth-elevation position (noise level ). The plots are for a flutter rate of 200 Hz. These plots show that the localization entropy in <i>R. rouxii</i> is always lowest in a central region and that the region of low entropy expands as a function of echo strength. Bottom row: Similar as the middle row but for a noise level of . (b) Visual analogue to illustrate the sensitivity-accuracy trade-off. The top left and the bottom right quadrant of the classic Lenna image <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1002268#pcbi.1002268-Hutchinson1" target="_blank">[46]</a> were given an increased intensity. The quadrants serve as an analogue to the highly sensitive templates found in the central region of the echolocation systems of bats. The other two quadrants were given an increased contrast, i.e. as an analogue to the peripheral templates with a higher dynamic range. In the bottom figure the top figure has been corrupted by additive noise. The correlation between each quadrant in the top and the bottom figure is indicated by the numbers in the bottom figure. It can be seen that the quadrants with the highest contrast (i.e. the lowest average intensity or sensitivity) have the highest correlation. This indicates that high contrast images, while having a lower average intensity, are less affected by noise. Similarly, localization templates with a high dynamic range are less disturbed by noise but render the bat less sensitive.