Faster Seafloor Spreading and Lithosphere Production During The

Faster seaﬂ oor spreading and lithosphere production during the mid-Cenozoic

Clinton P. Conrad Department of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, Maryland 21218, USA Carolina Lithgow-Bertelloni Department of Geological Sciences, University of Michigan, Ann Arbor, Michigan 48109, USA

ABSTRACT Concurrent changes in seawater chemistry, sea level, and climate since the mid-Cretaceous are thought to result from an ongoing decrease in the global rate of lithosphere production at ridges. The present-day area distribution of seafl oor ages, however, is most easily explained if lithosphere production rates were nearly constant during the past 180 m.y. We examined spatial gradients of present-day seafl oor ages and inferred ages for the subducted Farallon plate to construct a history of spreading rates in each major ocean basin since ca. 140 Ma, revealing dramatic Cenozoic events. Globally, seafl oor spreading rates increased by ~20% during the early Cenozoic due to an increase in plate speeds in the Pacifi c basin. Since then, subduction of the fast-spreading Pacifi c-Farallon ridge system has led to a 12% decrease in average global spreading rate and an 18% or more decrease in the total rate of lithosphere production by the most conservative estimates. These rapid changes during the Cenozoic defy models of steady-state seafl oor formation, and demonstrate the time-dependent and evolving nature of plate tectonics on Earth.

Keywords: ridge spreading, seaﬂ oor ages, crustal production, plate tectonics, Farallon plate.

INTRODUCTION The youngest seafl oor, recently produced at seafl oor spreading allows us to estimate the age The motions of Earth’s tectonic plates rep- spreading ridges, occupies proportionally more distribution of subducted seafl oor, the spreading resent the surface manifestation of convection area than does older seafl oor because it is less counterpart of which is still preserved. We use in the Earth’s mantle, and, over time scales likely to have been subducted since the time of its these additional pieces of information to detect of 50–100 m.y., govern a variety of geologi- formation. By examining the area distribution of changes in spreading rate, ridge length, and sea- cal processes, including mantle heat loss, sea- seafl oor ages, Parsons (1982) and Rowley (2002) fl oor production rates as functions of time and level change (Kominz, 1984; Gaffi n, 1987), observed an approximately triangular shape for between ocean basins. Although these estimates the carbon cycle (Berner and Kothavala, the distribution of area per unit age versus age. become increasingly uncertain with age because 2001), and seawater chemistry (Hardie, 1996). Rowley (2002) noted that this distribution is the subduction is constantly destroying informa- Although the basic history of plate motions expected result of a constant rate of seafl oor pro- tion about past spreading environments, we can has been deduced for the Cenozoic and some duction at ridges combined with consumption observe trends, particularly during the Cenozoic, of the Mesozoic, controversy has emerged of seafl oor with a uniform age distribution. In that deviate from the constant spreading model regarding changes in the global rates of ridge the absence of any other information about past suggested by the seafl oor age distribution alone. spreading and lithospheric production during seafl oor production rates, the present-day sea- As in Rowley’s (2002) study, our estimates are this time period. Tectonic reconstructions sug- fl oor age distribution gives no reason to suspect a based only on seafl oor ages measured from the gest that plate velocities have increased dur- signifi cant change in production rates during the present-day seafl oor and are thus not subject to ing the Cenozoic, particularly in the Pacifi c past 180 m.y. (Rowley, 2002). Although it ignores potential uncertainty associated with reconstruc- (Cogné and Humler, 2004), yet global studies information about past rates that can be gleaned tions of past plate motions. of rates of ridge spreading (Kominz, 1984) or from tectonic plate reconstructions, Rowley’s subduction (Engebretson et al., 1992) suggest (2002) study is appealing because it relies only PAST SEAFLOOR SPREADING that lithospheric production rates at ridges on well-defi ned (Müller et al., 1997) age data RATES AND RIDGE LENGTHS have slowed during the Cenozoic. Rowley from the present-day seafl oor. Other studies that The age of the ocean fl oor at any point (θ, φ) = (2002) used the area-age distribution of the infer faster production rates for past times (e.g., (colatitude, longitude) can be determined by present-day ocean fl oor to suggest that sea- Kominz, 1984; Engebretson et al., 1992) rely on examining the pattern of parallel magnetic fl oor production rates have remained nearly tectonic reconstructions that inevitably carry a anomalies that record Earth’s magnetic fi eld at constant since 180 Ma. This controversy has degree of uncertainty. the time of their formation at a spreading ridge. introduced questions over whether Cenozoic Maps of present-day seafl oor ages (Müller By interpolating between the locations of dated trends such as sea-level lowering (Haq et al., et al., 1997), however, contain additional infor- magnetic anomalies and using additional infor- 1987), increasing Mg/Ca ratio in seawater mation about the spreading systems that origi- mation from fracture-zone traces, Müller et al. (Hardie, 1996), and decreasing atmospheric nally formed this seafl oor. For example, Cogné (1997) determined the seafl oor age on a 0.1° × carbon and associated cooling (Berner and and Humler (2004) recognized that the spacing 0.1° grid (Fig. 1A). We use it to estimate past Kothavala, 2001) are the results of decreasing of isochrons provides a time history of past spreading rates and ridge lengths through analy- lithospheric production rates. spreading rates. In addition, the symmetry of sis of seafl oor age gradients.

© 2007 Geological Society of America. For permission to copy, contact Copyright Permissions, GSA, or [email protected]. GEOLOGY,Geology, January January 2007; 2007 v. 35; no. 1; p. 29–32; doi: 10.1130/G22759A.1; 4 fi gures. 29 Figure 1. Half-spread- North ing rate at time oceanic Atlantic lithosphere was originally produced, as determined by applying Equation 1 to sea fl oor age map of West Müller et al. (1997). Light Small Pacific Basins gray indicates regions where ages or rates cannot be determined, middle gray indicates continental West East South Indian areas, and dark gray indi- East Pacific Atlantic cates plate boundaries. Indian Dashed black lines are 25 m.y. age contours. We South have defi ned eight dif- Pacific ferent ocean basins, the boundaries of which, shown in black, are drawn through continents and along transform faults.

0 1 2 3 4 6 8 10 15 20 Half spreading rate (cm/yr)

(θφ)= +Δ The gradient of seafl oor age, T(θ, φ), is ( ) = T , ttT2 increasingly uncertain for past times because Ltsp∫ dL sp , TtT(θφ, )=−Δ 2 inversely proportional to a rate of spreading, (2) they are based on a progressively smaller frac- θ φ vsp( , ): dA(θφ, ) tion of the original spreading length. where dL = , R sp ΔTv (θφ, ) v (θφ, ) = E = sp CENOZOIC AND CRETACEOUS sp ∇T (θφ, ) and dA(θ, φ) = dθ dφ sin(θ) is the local area. TECTONIC CHANGE ⎡⎛ ∂ (θφ)⎞ 2 Equation 2 determines spreading length by Using half-spreading rates determined from ⎢ 1 T , RE ⎜ ⎟ dividing the local area dA by a distance of Müller et al.’s (1997) data set (Fig. 1), we used ⎢⎝ sin(θ) ∂φ ⎠ (1) ⎣ spreading perpendicular to the ridge axis, given Equation 3 to measure average half-spreading (−12) by ΔTv (θ, φ). Because each spreading ridge rates as a function of time both globally and 2 ⎤ sp ⎛ ∂T (θφ, )⎞ + ⎥ , produces isochrons on either side of its axis, for each of the eight ocean basins defi ned in ⎝⎜ ∂θ ⎠⎟ ⎦⎥ in the absence of subduction the integration of Figure 1. To remove nonphysical slow spread- Equation 2 over all seafl oor of age t yields dou- ing associated with transform faults, we do not where RE is the radius of the Earth. When applied ble the ridge length at that time. The destruction include half-spreading rates slower than 15% to seafl oor age data, inverse age gradients calcu- of isochrons by subduction, however, causes of the average present-day spreading rate for a lated using Equation 1 yield half-spreading rates measured values of Lsp(t) to generally decrease given basin (Fig. 2), as measured by taking the that applied at the time of seafl oor formation with increasing seafl oor age. length-weighted average of full spreading rates θ φ (Müller et al., 1998). We determined vsp( , ) Average rates of spreading, Vsp, can be deter- given by DeMets et al. (1990). We fi nd that this for all points with defi ned ages in Müller mined for times in the past by taking the length- 15% cutoff removes nonphysical slow spread- et al.’s (1997) data set (Fig. 1). Some regions, weighted average of the local half-spreading ing rates associated with transform faults with- particularly within the Pacifi c’s poorly defi ned rate along an isochron: out eliminating spreading associated with past mid-Cretaceous quiet zone (118–83 Ma), show (θφ= +Δ ) lithosphere production. For the present day, ( ) = 1 T , ttT2 (θφ) anomalously fast spreading rates in excess of Vtsp ∫ vdLsp, sp , (3) the average spreading rates measured from the Lt( ) TtT(θφ, =−Δ 2) 20 cm/yr, which we omit from further analysis. sp Müller et al. (1997) age grid (Fig. 2, lines) gen-

We also ignore rates determined for seafl oor where dLsp is the local spreading length given erally are within the DeMets et al. (1990) error younger than 1 Ma because gradients measured by Equation 2 and vsp is given by Equation 1. estimates for present-day seafl oor spreading. from such points often span the ridge, and thus This length-weighted determination of average Given that similar uncertainty is likely associ- do not correspond to spreading rates. Because half-spreading rate is preferable to an area- ated with rates measured from the Müller et al. age gradients across transform faults are large, weighted average because it avoids the over- (1997) age grid (which also rely on picks of sea- our determination of spreading rate (Equation 1) weighting of fast-spreading seafl oor inherent fl oor magnetic anomalies), we take the favor- assigns slow apparent rates near transform faults to area weighting. Thus, Vsp from Equation 3 is able comparison of these two measurements of in all basins (Fig. 1). Because these slow rates an estimate of the average half-spreading rate present-day seafl oor spreading as a verifi cation are nonphysical artifacts, we remove them from along the length of a former ridge at time t. of the robustness of our method. For past times further analysis, as described in the following. Note that this estimate of spreading rate is the range of uncertainty is at least as large as that

The total length of preserved spreading, Lsp determined only along the preserved length of shown for the present day. along an isochron of age t can be determined the former spreading ridge, which, because of Average half-spreading rates (Fig. 2) describe by integrating local estimates of the spreading subduction, represents an increasingly smaller the tectonic evolution of each basin’s spread- length, dLsp, over points whose ages are within fraction of the original spreading length for ing systems as a function of time. The global Δ an age range T that is centered about t: older seaﬂ oor. Thus, estimates of Vsp become average of half-spreading rates (Fig. 2A) is

30 GEOLOGY, January 2007 4.5 120 that the resulting increase in the slab pull force Reconstructed Farallon 4.0 A) Global Average + Farallon Global Average 110 E. Pacific 3.5 drove faster Pacifi c plate motion, and thus faster W. Pacific 3.0 Pacifi c spreading rates. 100 S. Pacific 2.5 Small Bas. 90 E. Indian 2.0 Spreading rates in basins with passive mar- W. Indian 1.5 gins (Fig. 2C) remained fairly stable during the 80 S. Atlantic 10 70 N. Atlantic B) 9 East Pacific Cenozoic, exhibiting variations of only a few mil- West Pacific 60 8 South Pacific limeters per year (Fig. 2C). The rate changes that 50 7 Small Basins occur generally have a tectonic explanation. For East Indian 40 6 example, the West Indian basin shows an ~50% 30 5 increase in spreading rate between ca. 65 Ma 4 20 and ca. 50 Ma (Fig. 2C). This period of rapid 3 10 2 spreading may be related to a strong slab pull Preserved spreading length (103 km) 0 1 force on the Indian plate, which was eliminated 0 20 40 60 80 100 120 140 0 Seafloor age (Ma) 2.5 when India collided with Eurasia ca. 50 Ma or C) earlier. South Atlantic spreading was slower dur- 2.0 Figure 3. Total spreading length (isochron ing this period (Fig. 2C), and can be explained by length), calculated using Equation 2 and ΔT 1.5 changes in African plate motion. = 5 m.y., preserved on the seafl oor of each Average preserved half-spreading rate (cm/yr) 1.0 basin, as a function of seafl oor age. Lengths West Indian for subducted Farallon plate are estimated 0.5 South Atlantic RIDGE LENGTHS AND LITHOSPHERE North Atlantic as described in text. 0.0 PRODUCTION RATES 0 20 40 60 80 100 120 140 Estimates of total lithosphere production Seafloor age (Ma) rate require knowledge of ridge length as well vative because in many instances (e.g., seafl oor Figure 2. Average half-spreading rates (col- as spreading rate. We calculate the length of older than ca. 50 Ma beneath the Aleutians or ored lines) calculated as function of seafl oor preserved isochrons in each ocean basin using young lithosphere beneath Central America), age by applying Equation 3 to half-spreading Equation 2 and fi nd that preserved spreading both sides of the ridge have been destroyed. map of Figure 1, using a 5 m.y. running mean lengths generally increase toward the present day For these situations our reconstruction does not (ΔT = 5 m.y.). A: Global averages, both with and without subducted Farallon plate (see (Fig. 3). This trend is expected because older sea- recover spreading from either side of the ridge. text). B: Averages for fi ve basins (Fig. 1) fl oor is more likely to have been subducted since Thus, our estimate of Farallon spreading lengths whose margins are dominated by subduc- the time of its formation. We cannot directly (Fig. 3) represents only a fraction of the total tion zones. C: Averages for three basins measure isochron lengths for seafl oor that has spreading length at the time. Nevertheless, add- (Fig. 1) surrounded by passive margins. Computed rates for present day compare been subducted, but we can infer these lengths if ing reconstructed Farallon spreading lengths to favorably to independent estimates of pres- (assumed symmetric) spreading counterparts are the global summation (Fig. 3) shows that spread- ent-day spreading determined by taking preserved. This is the case for the Pacifi c, where ing lengths have decreased by at least ~5000 km length-weighted averages of spreading rate the Kula and Farallon plates, which dominated (~4.5%) during the past 20 m.y. estimates, shown as dots with error bars, the northern and eastern Pacifi c for most of the We can obtain a minimum estimate of past taken from DeMets et al. (1990). Because spreading rates depend directly on the time Cenozoic, have been largely subducted beneath rates of lithosphere production by multiply- scale used to assign ages to isochrons, we North America and South America. These plates ing preserved ridge lengths (Fig. 3) by rates have applied the Cande and Kent (1995) time were separated from the Pacifi c plate by a ridge of spreading (Fig. 2) for each basin. The result scale, which was used by Müller et al. (1997), system that produced most of the Cenozoic litho- (Fig. 4) is analogous to the plot of seafl oor area to the DeMets et al. (1990) data, which slows rates by a factor of 0.957. Because DeMets sphere of the Pacifi c plate (West Pacifi c basin). produced per unit age as a function of age. et al. (1990) gave full spreading rates across If we assume symmetrical spreading, we can Except for our use of a 5 m.y. running mean, ridges, orange and purple dots and error estimate spreading lengths for the subducted Far- Figure 4 reproduces the area-age distribution bars in part B are determined from spread- allon plate by examining isochrons that are pre- ing rates for combined East and West served on the Pacifi c side of the ridge. Pacifi c basins. Thus, this curve should be 3.5 To obtain a minimum estimate of subducted Reconstructed Farallon compared to the average of East and West E. Pacific Pacifi c curves slightly weighted toward the Farallon spreading length, we subtract the East 3.0 W. Pacific S. Pacific East Pacifi c curve because of its longer Pacifi c spreading length from its counterpart in Small Bas. present-day ridge length (Fig. 1). E. Indian the West Pacifi c basin (Fig. 3). For ages younger 2.5 W. Indian S. Atlantic than ca. 10 Ma, Cocos-Nazca spreading (Fig. 1) N. Atlantic 2.0 makes the East Pacifi c spreading lengths lon- ~2.1 cm/yr during the Late Cretaceous and ger than those in the West Pacifi c, so we obtain 1.5 early Cenozoic. Between ca. 40 Ma and ca. zero contribution for the missing length (cer- 30 Ma, the global average increased by ~20%, tainly an underestimate). For ages between 1.0 to ~2.5 cm/yr. This global increase was caused ca. 10 and ca. 40 Ma, East Pacifi c lengths get 0.5

by accelerating spreading rates in the West smaller because they are lost to subduction, so Seafloor production rate (km2/yr) Pacifi c basin, rates that almost doubled during our reconstruction of lost Farallon spreading 0.0 0 20 40 60 80 100 120 140 this time (Fig. 2B). The West Pacifi c increase grows to include a progressively larger fraction Seafloor age (Ma) is likely the result of eastward motion of the of the West Pacifi c spreading lengths. For ages Pacifi c-Farallon ridge system, which caused a older than ca. 40 Ma, Farallon subduction has Figure 4. Net seafl oor production rate for Cenozoic increase in the age, and thus weight, eliminated all of the seafl oor in the East Pacifi c, each basin as a function of age, calculated by multiplying the average half-spreading of slabs subducting along the northern and west- so our reconstructed Farallon spreading lengths rates (Fig. 2) by the total spreading lengths ern margins of the Pacifi c (Parsons, 1982). We equal the West Pacifi c spreading lengths. This (Fig. 3) for each basin, and reconstructed (Conrad and Lithgow- Bertelloni, 2004) showed reconstruction of subducted seafl oor is conser- Farallon plate.

GEOLOGY, January 2007 31 of Rowley (2002), who used its approximately lengths have decreased by at least 4.5% since ACKNOWLEDGMENTS triangular nature to infer nearly constant rates 20 Ma, primarily due to the subduction of the We thank R. Demicco for a helpful review, and of lithosphere production by ridges. When past Farallon-Pacifi c ridge system beneath west- an anonymous reviewer for extensive comments that improved the analysis and discussion; fi nancial sup- production of the Farallon lithosphere is added ern North America. Because the lost spreading port was provided by the David and Lucile Packard (Fig. 4), however, it becomes evident that litho- ridges included some of the world’s fastest, the Foundation and National Science Foundation grants spheric production rates were higher in the combination of slow decreases in both spreading EAR-0609590 and EAR-0609553. past. To determine Farallon production rates, rate and ridge length resulted in a net decrease we multiplied spreading rates from the West in the global rate of lithosphere production by at REFERENCES CITED Berner, R.A., and Kothavala, Z., 2001, GEOCARB Pacifi c basin (Fig. 2B) with the reconstructed least ~18% in the past 20 m.y. Using less conser- III: A revised model of atmospheric CO2 over Farallon spreading lengths (Fig. 3). In doing so, vative assumptions, production rates that were Phanerozoic time: American Journal of Sci- we assumed symmetrical spreading despite the more than 25% faster at 20 Ma are possible. ence, v. 301, p. 182–204. fact that asymmetrical spreading is common in Additional subduction of seafl oor prior to the Cande, S.C., and Kent, D.V., 1995, Revised calibra- spreading environments (Müller et al., 1998) mid-Cenozoic makes estimating earlier trends tion of the geomagnetic polarity timescale for the Late Cretaceous and Cenozoic: Journal of and is evident on the East Pacifi c Rise. However, in seafl oor production rate diffi cult. However, a Geophysical Research, v. 100, p. 6093–6095. spreading along the East Pacifi c rise is faster on faster seafl oor production rate in the past is con- Cogné, J.P., and Humler, E., 2004, Temporal varia- the Nazca plate side of the ridge, and has been sistent with observations of increasing Mg/Ca tions of oceanic spreading and crustal produc- for at least 40 m.y. (Fig. 2), so our reconstruction ratio (Hardie, 1996), decreasing sea level (Haq tion rates during the last 180 My: Earth and Planetary Science Letters, v. 227, p. 427–439. is likely conservative. If the half-spreading rates et al., 1987), and decreasing atmospheric carbon Conrad, C.P., and Lithgow-Bertelloni, C., 2004, The for the current East Pacifi c seafl oor, which are (Berner and Kothavala, 2001) during the Ceno- temporal evolution of plate driving forces: as much as ~50% faster than those for the West zoic, all of which can be attributed to decreasing Importance of “slab suction” versus “slab Pacifi c during 20–40 Ma, also applied to the lithosphere production rates. We show that this pull” during the Cenozoic: Journal of Geo- portion of East Pacifi c (Farallon) spreading that decrease occurred because the rapidly spread- physical Research, v. 109, p. B10407, doi: 10.1029/2004JB002991. has been lost to subduction, then the lithospheric ing Pacifi c-Farallon ridge system decreased in DeMets, C., Gordon, R.G., Argus, D.F., and Stein, production rates inferred for the subducted Far- length during the Cenozoic. S., 1990, Current plate motions: Geophysical allon plate would be as much as ~50% larger Our results contrast strongly with Rowley’s Journal International, v. 101, p. 425–478. than the estimate shown in Figure 4. (2002) suggestion that lithosphere production Demicco, R.V., 2004, Modeling seafl oor-spreading rates through time: Geology, v. 32, p. 485–488. By introducing seafl oor production inferred rates have not signifi cantly changed from their Engebretson, D.C., Kelley, K.P., Cashman, H.J., and for the Farallon plate (Fig. 4), we see that sea- current rate during the past 180 m.y. Instead, a Richards, M.R., 1992, 180 Million years of fl oor production rates have decreased during conservative reconstruction of isochrons for the subduction: GSA Today, v. 2, p. 93–95. the past 20 m.y. The average rate of decrease subducted Farallon plate shows that production Gaffi n, S., 1987, Ridge volume dependence on sea- (~0.025 km2/yr/m.y.) is a minimum estimate rates were at least ~18% higher (3.3 km2/yr; fl oor generation rate and inversion using long- term sea-level change: American Journal of because our assumptions certainly underesti- Fig. 4) and possibly as much as 25% larger Science, v. 287, p. 596–611. mate past spreading lengths and likely under- (3.5 km2/yr if we assume East Pacifi c spread- Haq, B.U., Hardenbol, J., and Vail, P.R., 1987, Chro- estimate the half-spreading rates that applied to ing rates for the subducted Farallon plate), as nology of fl uctuating sea levels since the Trias- them. This ~18% decrease in the global rate of recently as 20 Ma (cf. 2.8 km2/yr for the pres- sic: Science, v. 235, p. 1156–1167. Hardie, L.A., 1996, Secular variation in seawater lithosphere production occurred in only 20 m.y., ent day [Fig. 4]; see also Rowley [2002]). In chemistry: An explanation for the coupled and at a rate comparable to estimates of the rate addition, given that a simple reconstruction secular variation in the mineralogies of marine of decrease determined from rates of sea-level of the Farallon plate led to the discovery of limestones and potash evaporites over the past lowering (~0.024 km2/yr/m.y. over 80 m.y.) faster lithosphere production in the recent past, 600 m.y.: Geology, v. 24, p. 279–283. (Gaffi n, 1987) or changes in subduction rate additional reconstruction of subducted seafl oor Kominz, M.A., 1984, Oceanic ridge volume and sea- 2 level change—An error analysis, in Schlee, (~0.02 km /yr/m.y. over 50 m.y.) (Engebretson may show even faster lithosphere production J.S., ed., Interregional unconformities and et al., 1992). Our estimates indicate that, prior rates prior to ca. 30 Ma. Rapid spreading asso- hydrocarbon accumulation: American Asso- to ca. 30 Ma, the rate of seafl oor production was ciated with the now-subducted Pacifi c-Kula ciation of Petroleum Geologists Memoir 36, lower than it is today (Fig. 4). Although this and Kula-Farallon spreading systems may have p. 109–127. Müller, R.D., Roest, W.R., Royer, J.-Y., Gahagan, decreased production rate results partly from driven additional crustal production in the ear- L.M., and Sclater, J.G., 1997, Digital isochrons slower spreading rates in the early Cenozoic lier Cenozoic and Cretaceous. of the world’s ocean fl oor: Journal of Geophys- (Fig. 2A), some of this measured decrease is Demicco (2004) showed that the approxi- ical Research, v. 102, p. 3211–3214. due to the loss of seafl oor to subduction since mately triangular nature of the present-day Müller, R.D., Roest, W.R., and Royer, J.-Y., 1998, that time. area/age versus age curve does not necessarily Asymmetric sea-fl oor spreading caused by ridge-plume interactions: Nature, v. 396, require constant lithosphere production rates, p. 455–459. DISCUSSION AND CONCLUSIONS as suggested by Parsons (1982) and advocated Parsons, B., 1982, Causes and consequences of the We have examined gradients in present-day by Rowley (2002). We have shown that vary- relation between area and age of the ocean seafl oor ages to estimate both past ridge lengths ing rates of seafl oor production, combined with fl oor: Journal of Geophysical Research, v. 87, p. 289–302. and spreading rates during the Cretaceous and changing ridge geometries, seem to have pro- Rowley, D.B., 2002, Rate of plate creation and Cenozoic. Globally, the average spreading rate duced the approximately triangular distribution destruction: 180 Ma to present: Geological increased by ~20% during the fi rst half of the observed for the present day. Given the Ceno- Society of America Bulletin, v. 114, p. 927– Cenozoic (Fig. 2A) due primarily to a near zoic and Mesozoic history of changing plate 933. doubling of Pacifi c plate speed (Fig. 2B). The geometries, spreading rates, and ridge lengths, Manuscript received 24 February 2006 global average spreading rate has decreased by and the fact that plate motions represent the sur- Revised manuscript received 10 August 2006 ~12% during the second half of the Cenozoic face expression of vigorous and time-dependent Manuscript accepted 28 August 2006 (Fig. 2A). When we include an inference for mantle fl ow, it is not surprising that lithosphere subducted Farallon seafl oor, we fi nd that ridge production rates vary signifi cantly with time. Printed in USA

32 GEOLOGY, January 2007