RESUMEN
Bark beetles (Coleoptera: Curculionidae, Scolytinae) have specialized feeding habits, and commonly colonize only one or a few closely related host genera in their geographical ranges. The red turpentine beetle, Dendroctonus valens LeConte, has a broad geographic distribution in North America and exploits volatile cues from a wide variety of pines in selecting hosts. Semiochemicals have been investigated for D. valens in North America and in its introduced range in China, yielding apparent regional differences in response to various host volatiles. Testing volatiles as attractants for D. valens in its native and introduced ranges provides an opportunity to determine whether geographic separation promotes local adaptation to host compounds and to explore potential behavioral divergence in native and introduced regions. Furthermore, understanding the chemical ecology of host selection facilitates development of semiochemicals for monitoring and controlling bark beetles, especially during the process of expansion into new geographic ranges. We investigated the responses of D. valens to various monoterpenes across a wide range of sites across North America and one site in China, and used the resulting information to develop an optimal lure for monitoring populations of D. valens throughout its Holarctic range. Semiochemicals were selected based on previous work with D. valens: (R)-(+)-alpha-pinene, (S)-(-)-alpha-pinene, (S)-(-)-beta-pinene, (S)-(+)-3-carene, a commercially available lure [1:1:1 ratio of (R)-(+)-alpha-pinene:(S)-(-)-beta-pinene:(S)-(+)-3-carene], and a blank control. At the release rates used, (+)-3-carene was the most attractive monoterpene tested throughout the native range in North America and introduced range in China, confirming results from Chinese studies. In addition to reporting a more effective lure for D. valens, we present a straightforward statistical procedure for analysis of insect trap count data yielding cells with zero counts, an outcome that is common but makes the estimation of the variance with a Generalized Linear Model unreliable because of the variability/mean count dependency.
