Observed and predicted North American teleseismic delay times

•New measurements of absolute teleseismic delay times in North America using AIMBAT.•Measured mantle delays increase sharply across the Rocky Mountains and Appalachians.•A 4.2 s E–W mean delay difference consistent with a continental lithosphere thinner than 250 km.•Tomographic models underestimate observed delays to varying degrees.•Non-teleseismic tomographic models overestimate smoothness delay patterns.

Using a newly developed cross-correlation-based computer tool we measured absolute delay times of teleseismic P and S waves recorded by about 2000 broadband seismic stations from EarthScope's USArray, previous PASSCAL arrays, and additional permanent networks in North America. We estimate contributions to the delays from crustal structure, using various prior crustal models, and from event-side heterogeneity, using the delay time distribution itself. We then subtract these contributions from our measurements and map the average delay at each station location. We analyze these average delay times to investigate the structure of the North American mantle and formation of the North American continent. Mantle S delay times from stations west of the Rocky Mountains are 4.2 s larger than delay times from stations within the US portion of stable North America. Locally, S delays at Yellowstone are another 4 s larger than those west of Rocky Mountains. Decreasingly sharp delay time contrasts occur across the Rocky Mountains, the Appalachian Mountains, and on the southern edge of the North American Craton, respectively. These delay time gradients of various steepness broadly coincide with surface geological boundaries, but are offset to either side of these boundaries. Predictions of teleseismic S delays from twelve three-dimensional tomographic mantle S velocity models agree generally with observed delay time patterns, but underestimate the delays and advances to varying degrees. Tomographic models derived from different data than teleseismic arrival times predict similarly valid delays as models derived from teleseismic arrivals, while overestimating the smoothness of delay time patterns to varying degrees. We further utilize these tomographic models to predict and study the size and distribution of delay contributions from modeled heterogeneity in different depth ranges. This study shows that the 80–240 km depth range is the dominant contributor to delay time contrasts and variance. This depth range generally corresponds to asthenosphere in the western US and lithosphere in the central and eastern US. The average depth to which the observed delays require the central US lithosphere to extend is likely shallower than 240 km and consistent with the seismic bottom of lithosphere imaged in seismic-velocity models not derived from teleseismic delay times.