Seismological Research Letters Volume 82, Number 5 September/October 2011 661 doi: 10.1785/gssrl .82.5.661 Computing Green’s Functions from Ambient Noise Recorded by Accelerometers and Analog, Broadband, and Narrow-Band Seismometers I. M. Tibuleac, D. H. von Seggern, J. G. Anderson, and J. N. Louie I. M. Tibuleac, D. H. von Seggern, J. G. Anderson, and J. N. Louie Nevada Seismological Laboratory, University of Nevada, Reno INTRODUCTION he objective of our study is to supplement the regional P/ S 3-D velocity model in the Reno basin with shear-wave velocity models derived from ambient noise interferometry (Aki 1957; Claerbout 1968; Shapiro et al. 2005; Sabra et al. 2005; Lin et al. 2008; Yang et al. 2008). We use a variety of seismic sensors in the Reno-Carson area (Figure 1). here is a gap for demonstrated extraction of Green’s func- tions (GFs) from ambient noise between short and long inter- station distances. he lateral resolution of existing tomographic models exceeds the dimensions of the Reno basin area (~60 km 2 ), and their depth resolution is larger than the Reno basin depth to basement (<3 km). GFs have been recently retrieved in the western United States from data recorded at broadband sensors, such as the EarthScope Transportable Array (TA) deployment (Figure 1C) with lateral resolution of 60–100 km (Yang et al. 2008; Lin et al. 2008) and for periods exceeding 8 s ( i.e., sampling more than 8 km deep). Using the Re-Mi meth- ods, noise-extracted Rayleigh waves for local shallow shear- wave velocity studies provided information on structure less than 0.2 km from the surface (Scott et al. 2004). In order to estimate P and S velocity models deeper than 3 km, earthquake tomography was until recently the only cost efective alterna- tive to active source experiments (Frary et al. 2009). he body- wave tomographic studies in the Reno basin (Preston and von Seggern 2008) do not allow precise control of the velocity or depth of shallow structures in the basin, due to low resolution in the upper 3 km. Another disadvantage of body-wave tomog- raphy is the high level of uncertainty in the S-arrival time picks. GF extraction from noise cross-correlations at scales less than 60 km using non-broadband instruments is possible and has been recently demonstrated by several research groups (Picozzi et al. 2009; Gouedard et al. 2008; Cho et al. 2007). To obtain higher resolution three-dimensional models at less than 60-km scale, using ambient noise-extracted GFs, the density of broadband (here named BH, from the three- component SEED channel names) stations must be increased considerably. Such deployments of broadband instruments are not cost efective. Alternatively, all the instruments available in the region must be used. he advantage of Reno basin as a pop- ulated region is that in addition to broadband sensors, dense networks of short-period instruments (analog or digital) and accelerometers are already deployed (Figure 1, let plot). his allows analysis of hundreds more inter-station paths, with the potential to greatly improve velocity-model resolution. We deine as “unconventional” a sensor pair contain- ing at least one non-broadband sensor or at least an analog short-period sensor. As shown below, extracting useful data with ambient noise interferometry from unconventional sen- sor pairs is not a trivial exercise. Few research groups have attempted this type of study, and to our knowledge, only one group used accelerometers (Cho et al. 2007). he irst challenge in using unconventional sensor pairs is data quality. Short-period recordings oten have poor data qual- ity for the desired frequencies, especially analog recordings. he short-period instruments are predominantly narrow-band (cor- ner frequency ≥1 Hz), with the response rapidly decreasing at periods longer than 1 s. he accelerometer response de-empha- sizes low frequencies (<1 Hz) because it falls of as the inverse of period squared. Also, accelerometers usually operate in trigger mode. In this study we show recovery of GFs from ambient noise for sensor pairs including: 1) digital narrow-band seismometer (EH) recordings (for instance, S-13 seismometers), 2) analog narrow-band seismometer (SH) recordings, digitized ater trans- mission, 3) digital accelerometer (HG) recordings, and 4) digital broadband instrument (BH) recordings, including USArray sta- tions. Each of these instrument classes presents its own problems for GF recovery and requires appropriate processing. he second signiicant challenge in using high-density unconventional sensor pairs is that measurements of Rayleigh- wave phase and group velocity are di icult for stations sepa- rated by less than 15 km. he rule of thumb is that the lon- gest Rayleigh wavelength that can be well resolved is one-half to one-third of the inter-station distance. For example, at 15 km distance, in order to extract 3 km/s group velocity Rayleigh