ABSTRACT We used geologic mapping, tephrochronology, and 40 Ar/ 39 Ar dating to describe evidence of a ca. 3.5 Ma pluvial lake in Eureka Valley, eastern California, that we informally name herein Lake Andrei. We identified six different tuffs in the Eureka Valley drainage basin, including two previously undescribed tuffs: the 3.509 ± 0.009 Ma tuff of Hanging Rock Canyon and the 3.506 ± 0.010 Ma tuff of Last Chance (informal names). We focused on four Pliocene stratigraphic sequences. Three sequences are composed of fluvial sandstone and conglomerate, with basalt flows in two of these sequences. The fourth sequence, located ~1.5 km south of the Death Valley/Big Pine Road along the western piedmont of the Last Chance Range, included green, fine-grained, gypsiferous lacustrine deposits interbedded with the 3.506 Ma tuff of Last Chance that we interpret as evidence of a pluvial lake. Pluvial Lake Andrei is similar in age to pluvial lakes in Searles Valley, Amargosa Valley, Fish Lake Valley, and Death Valley of the western Great Basin. We interpret these simultaneous lakes in the region as indirect evidence of a significant glacial climate in western North America during marine isotope stages Mammoth/Gilbert 5 to Mammoth 2 (MIS MG5/M2) and a persistent Pacific jet stream south of 37°N.

Bear Lake, on the Idaho-Utah border, lies in a fault-bounded valley through which the Bear River flows en route to the Great Salt Lake. Surficial deposits in the Bear Lake drainage basin provide a geologic context for interpretation of cores from Bear Lake deposits. In addition to groundwater discharge, Bear Lake received water and sediment from its own small drainage basin and sometimes from the Bear River and its glaciated headwaters. The lake basin interacts with the river in complex ways that are modulated by climatically induced lake-level changes, by the distribution of active Quaternary faults, and by the migration of the river across its fluvial fan north of the present lake. The upper Bear River flows northward for ~150 km from its headwaters in the northwestern Uinta Mountains, generally following the strike of regional Laramide and late Cenozoic structures. These structures likely also control the flow paths of groundwater that feeds Bear Lake, and groundwater-fed streams are the largest source of water when the lake is isolated from the Bear River. The present configuration of the Bear River with respect to Bear Lake Valley may not have been established until the late Pliocene. The absence of Uinta Range–derived quartzites in fluvial gravel on the crest of the Bear Lake Plateau east of Bear Lake suggests that the present headwaters were not part of the drainage basin in the late Tertiary. Newly mapped glacial deposits in the Bear River Range west of Bear Lake indicate several advances of valley glaciers that were probably coeval with glaciations in the Uinta Mountains. Much of the meltwater from these glaciers may have reached Bear Lake via ground-water pathways through infiltration in the karst terrain of the Bear River Range. At times during the Pleistocene, the Bear River flowed into Bear Lake and water level rose to the valley threshold at Nounan narrows. This threshold has been modified by aggradation, downcutting, and tectonics. Maximum lake levels have decreased from as high as 1830 m to 1806 m above sea level since the early Pleistocene due to episodic downcutting by the Bear River. The oldest exposed lacustrine sediments in Bear Lake Valley are probably of Pliocene age. Several high-lake phases during the early and middle Pleistocene were separated by episodes of fluvial incision. Threshold incision was not constant, however, because lake highstands of as much as 8 m above bedrock threshold level resulted from aggradation and possibly landsliding at least twice during the late-middle and late Pleistocene. Abandoned stream channels within the low-lying, fault-bounded region between Bear Lake and the modern Bear River show that Bear River progressively shifted northward during the Holocene. Several factors including faulting, location of the fluvial fan, and channel migration across the fluvial fan probably interacted to produce these changes in channel position. Late Quaternary slip rates on the east Bear Lake fault zone are estimated by using the water-level history of Bear Lake, assuming little or no displacement on dated deposits on the west side of the valley. Uplifted lacustrine deposits representing Pliocene to middle Pleistocene highstands of Bear Lake on the footwall block of the east Bear Lake fault zone provide dramatic evidence of long-term slip. Slip rates during the late Pleistocene increased from north to south along the east Bear Lake fault zone, consistent with the tectonic geomorphology. In addition, slip rates on the southern section of the fault zone have apparently decreased over the past 50 k.y.

Bear Lake is a long-lived lake filling a tectonic depression between the Bear River Range to the west and the Bear River Plateau to the east, and straddling the border between Utah and Idaho. Mineralogy, elemental geochemistry, and magnetic properties provide information about variations in provenance of allogenic lithic material in last-glacial-age, quartz-rich sediment in Bear Lake. Grain-size data from the silici-clastic fraction of late-glacial to Holocene carbonate-rich sediments provide information about variations in lake level. For the quartz-rich lower unit, which was deposited while the Bear River flowed into and out of the lake, four source areas are recognized on the basis of modern fluvial samples with contrasting properties that reflect differences in bedrock geology and in magnetite content from dust. One of these areas is underlain by hematite-rich Uinta Mountain Group rocks in the headwaters of the Bear River. Although Uinta Mountain Group rocks make up a small fraction of the catchment, hematite-rich material from this area is an important component of the lower unit. This material is interpreted to be glacial flour. Variations in the input of glacial flour are interpreted as having caused quasi-cyclical variations in mineralogical and elemental concentrations, and in magnetic properties within the lower unit. The carbonate-rich younger unit was deposited under conditions similar to those of the modern lake, with the Bear River largely bypassing the lake. For two cores taken in more than 30 m of water, median grain sizes in this unit range from ~6 μm to more than 30 μm, with the coarsest grain sizes associated with beach or shallow-water deposits. Similar grain-size variations are observed as a function of water depth in the modern lake and provide the basis for interpreting the core grain-size data in terms of lake level.

A continuous, 120-m-long core (BL00-1) from Bear Lake, Utah and Idaho, contains evidence of hydrologic and environmental change over the last two glacial-interglacial cycles. The core was taken at 41.95°N, 111.31°W, near the depocenter of the 60-m-deep, spring-fed, alkaline lake, where carbonate-bearing sediment has accumulated continuously. Chronological control is poor but indicates an average sedimentation rate of 0.54 mm yr ‒1 . Analyses have been completed at multi-centennial to millennial scales, including (in order of decreasing temporal resolution) sediment magnetic properties, oxygen and carbon isotopes on bulk-sediment carbonate, organic- and inorganic- carbon contents, palynology; mineralogy (X-ray diffraction), strontium isotopes on bulk carbonate, ostracode taxonomy, oxygen and carbon isotopes on ostracodes, and diatom assemblages. Massive silty clay and marl constitute most of the core, with variable carbonate content (average = 31 ± 19%) and oxygen-isotopic values (δ 18 O ranging from ‒18‰ to ‒5‰ in bulk carbonate). These variations, as well as fluctuations of biological indicators, reflect changes in the water and sediment discharged from the glaciated headwaters of the dominant tributary, Bear River, and the processes that influenced sediment delivery to the core site, including lake-level changes. Although its influence has varied, Bear River has remained a tributary to Bear Lake during most of the last quarter-million years. The lake disconnected from the river and, except for a few brief excursions, retracted into a topographically closed basin during global interglaciations (during parts of marine isotope stages 7, 5, and 1). These intervals contain up to 80% endogenic aragonite with high δ 18 O values (average = ‒5.8 ± 1.7‰), indicative of strongly evaporitic conditions. Interglacial intervals also are dominated by small, benthic/tychoplanktic fragilarioid species indicative of reduced habitat availability associated with low lake levels, and they contain increased high-desert shrub and Juniperus pollen and decreased forest and forest-woodland pollen. The 87 Sr/ 86 Sr values (>0.7100) also increase, and the ratio of quartz to dolomite decreases, as expected in the absence of Bear River inflow. The changing paleoenvironments inferred from BL00-1 generally are consistent with other regional and global records of glacial-interglacial fluctuations; the diversity of paleoenvironmental conditions inferred from BL00-1 also reflects the influence of catchment-scale processes.

During glacial (pluvial) climatic periods, Death Valley is hypothesized to have episodically been the terminus for the Amargosa, Owens, and Mojave Rivers. Geological and biological studies have tended to support this hypothesis and a hydrological link that included the Colorado River, allowing dispersal of pupfish throughout southeastern California and western Nevada. Recent mitochondrial deoxyribonucleic acid (mtDNA) studies show a common pupfish (Cyprinodontidae) ancestry in this region with divergence beginning 3–2 Ma. We present tephrochronologic and paleomagnetic data in the context of testing the paleohydrologic connections with respect to the common collection point of the Amargosa, Owens, and Mojave Rivers in Death Valley during successive time periods: (1) the late Pliocene to early Pleistocene (3–2 Ma), (2) early to middle Pleistocene (1.2–0.5 Ma), and (3) middle to late Pleistocene (<0.7–0.03 Ma; paleolakes Manly and Mojave). Using the 3.35 Ma Zabriskie Wash tuff and 3.28 Ma Nomlaki Tuff Member of the Tuscan and Tehama Formations, which are prominent marker beds in the region, we conclude that at 3–2 Ma, a narrow lake occupied the ancient Furnace Creek Basin and that Death Valley was not hydrologically connected with the Amargosa or Mojave Rivers. A paucity of data for Panamint Valley does not allow us to evaluate an Owens River connection to Death Valley ca. 3–2 Ma. Studies by others have shown that Death Valley was not hydrologically linked to the Amargosa, Owens, or Mojave Rivers from 1.2 to 0.5 Ma. We found no evidence that Lake Manly flooded back up the Mojave River to pluvial Lake Mojave between 0.18 and 0.12 Ma, although surface water flowed from the Amargosa and Owens Rivers to Death Valley at this time. There is also no evidence for a connection of the Owens, Amargosa, or Mojave Rivers to the Colorado River in the last 3–2 m.y. Therefore, the hypothesis that pupfish dispersed or were isolated in basins throughout southeastern California and western Nevada by such a connection is not supported. Beyond the biologically predicted time frame, however, sparse and disputed data suggest that a fluvial system connected Panamint (Owens River), Death, and Amargosa Valleys, which could account for the dispersal and isolation before 3 Ma.

The western pupfish clade (Cyprinodontidae: Cyprinodon ) consists of nine species that occur primarily in isolation from one another in the large area extending from the Guzmán Basin, northwestern Chihuahua, Mexico, to the Death Valley region of southeastern California and southwestern Nevada. This paper presents a reassessment of estimated divergence times based on a compilation of previously published mitochondrial DNA sequences (ND2, cyt b, and control region). The results agree with previous estimates, which state that the western pupfish clade originated in the late Miocene or early Pliocene and that Cyprinodon was present in the Death Valley region by at least the middle Pliocene and possibly earlier. Enigmatically, there is little geologic evidence of late Neogene surface-water connections among the Guzmán, Lower Colorado River, and Death Valley regions. This indicates that either our knowledge of such connections is incomplete or pupfish dispersal across basin divides via small, relatively transient, surface-water connections has been more common than expected based on the low-gradient, valley-floor habitats generally occupied by this group.

Stratigraphic and geomorphic analyses reveal that the regional drainage basin of the modern Amargosa River formed via multistage linkage of formerly isolated basins in a diachronous series of integration events between late Miocene and latest Pleistocene–Holocene time. The 275-km-long Amargosa River system drains generally southward across a large (15,540 km 2 ) watershed in southwestern Nevada and eastern California to its terminus in central Death Valley. This drainage basin is divided into four major subbasins along the main channel and several minor subbasins on tributaries; these subbasins contain features, including central valley lowlands surrounded by highlands that form external divides or internal paleodivides, which suggest relict individual physiographic-hydrologic basins. From north to south, the main subbasins along the main channel are: (1) an upper headwaters subbasin, which is deeply incised into mostly Tertiary sediments and volcanic rocks; (2) an unincised low-gradient section within the Amargosa Desert; (3) a mostly incised section centered on Tecopa Valley and tributary drainages; and (4) a west- to northwest-oriented mostly aggrading lower section along the axis of southern Death Valley. Adjoining subbasins are hydrologically linked by interconnecting narrows or canyon reaches that are variably incised into formerly continuous paleodivides. The most important linkages along the main channel include: (1) the Beatty narrows, which developed across a Tertiary bedrock paleodivide between the upper and Amargosa Desert subbasins during a latest Miocene–early Pliocene to middle Pleistocene interval (ca. 4–0.5 Ma); (2) the Eagle Mountain narrows, which cut into a mostly alluvial paleodivide between the Amargosa Desert and Tecopa subbasins in middle to late Pleistocene (ca. 150–100 ka) time; and (3) the Amargosa Canyon, which formed in late middle Pleistocene (ca. 200–140 ka) time through a breached, actively uplifting paleodivide between the Tecopa and southern Death Valley subbasins. Collectively, the interconnecting reaches represent discrete integration events that incrementally produced the modern drainage basin starting near Beatty sometime after 4 Ma and ending in the Salt Creek tributary in the latest Pleistocene to Holocene (post–30 ka). Potential mechanisms for drainage integration across paleodivides include basin overtopping from sedimentary infilling above paleodivide elevations, paleolake spillover, groundwater sapping, and (or) headward erosion of dissecting channels in lower-altitude subbasins. These processes are complexly influenced by fluvial responses to factors such as climatic change, local base-level differences across divides, and (or) tectonic activity (the latter only recognized in Amargosa Canyon).

The Death Valley system (southeastern California and southwestern Nevada) contains a locally endemic aquatic biota that has long been the subject of compelling biogeographic speculation, yet it remains little studied phylogenetically. Springsnails (Hydrobiidae: Pyrgulopsis ) are one of the most diverse elements of this fauna, and they are thought to have evolved in association with late Tertiary rearrangements of landscape and drainage. We assembled a molecular phylogeny for this fauna to investigate its evolutionary development in relation to regional geological history. Sequences for two mitochondrial genes were obtained from 80 populations representing 13 of the 14 Death Valley system springsnail species, and 31 extralimital congeners. Combined analyses of the 1188 base-pair data set consistently depicted the Death Valley system fauna as a polyphyletic assemblage of eight or nine lineages. Based on a molecular clock, the six lineages endemic to the Death Valley system were estimated to be minimally Pliocene in age, which is concordant with inception of regional topographic closure during this time period. The single endemic lineage with a well-resolved sister relationship was closest to a species from the upperGila Riverbasin, which also suggests an old divergence event. Three other lineages shared a pattern of shallow structuring (divergence events youngerthan 0.7 Ma) across multiple drainage basins, some of which have long been isolated. This suggests that, contrary to previous thought, regional springsnail biogeography has been shaped in part by geologically recent (Pleistocene) dispersal, and, in some places, it has occurred by means other than spread through continuous reaches of aquatic habitat.

The North American Great Basin is a useful venue for the study of dispersal, vicariance, and rates of molecular evolution among aquatic organisms because its Pleistocene hydrogeographic history is relatively well known. This study examines regional molecular variation in the amphipod Hyalella azteca using mitochondrial (mt) gene sequence (deoxyribonucleic acid [DNA]) data. Populations within several endorheic drainages in the southern Great Basin were analyzed to determine if they represent a monophyletic assemblage with respect to populations from the pluvial Lake Bonneville drainage in the northern Great Basin. We also tested whether the patterns of molecular diversification among populations in the southern Great Basin were consistent with a Pleistocene vicariance hypothesis, and if the magnitude of observed sequence divergence was concordant with standard molecular clock calibrations. Our results show that diversity and endemism among Hyalella populations in the southern Great Basin are high with respect to those in the Lake Bonneville Basin. We further demonstrate that hyalellid populations in the southern Great Basin are a polyphyletic assemblage with respect to their counterparts in the Bonneville Basin, suggesting that dispersal events have been partially responsible for the enigmatic relationships within this assemblage. The relationships among lineages within the southern Great Basin are largely enigmatic and are not concordant with Pleistocene hydrographic history. Our data also indicate that rates of molecular evolution have been heterogeneous; there is a 2.8-fold disparity in relative rates of mtDNA divergence among closely allied lineages. The magnitude of sequence divergence among lineages is inconsistent with standard molecular clock calibrations, and evidence indicates that accelerated rates of divergence may have contributed to the high diversity and endemism among Great Basin hyalellids, complicating reconstruction of the temporal sequence of biogeographic events.

From the late Miocene to the middle Pliocene, the current drainage basin of the Owens River probably consisted of a broad, moderate-elevation, low-relief plateau with radiating drainage toward the Pacific Ocean, the northwestern Great Basin (now Lahontan drainages), and the Mojave and Colorado drainages. This plateau probably contained shallow basins, created by an extensional pulse at 12–11 Ma, at the present locations of major valleys. Between 4 and 3 Ma, this plateau was disrupted by a rapid westward step of extensional Basin and Range Province tectonism, which reactivated the Miocene faults and resulted in a linear north-south valley (the Owens Valley) with high mountain ranges on each side. This tectonic event resulted in geographic isolation and fragmentation of aquatic habitats and may have been a critical driver for speciation of aquatic organisms. Subsequent to this remarkable transformation of the landscape, the predominant influence on aquatic habitats has been very large, climate-driven fluctuations in the regional water balance that have resulted in the repeated interconnection and disconnection of the various basins that make up the paleo–Owens system. The magnitude of these fluctuations appears to have increased markedly since the early Pleistocene. Searles Lake has generally been the terminus of the Owens River, but at least once, probably at ca. 150 and/or ca. 70 ka, the system overflowed into Death Valley. During the last interglacial (marine isotope stage 5) and the Holocene, Owens Lake has been the terminus, but apparently not frequently before. These very large fluctuations in the water balance undoubtedly produced large shifts in the nature and distribution of aquatic habitats over geologically short periods of time, as well as repeatedly creating and severing connections between various parts of the larger drainage basin. This dynamic hydrological system provided the setting, and no doubt much of the impetus, for speciation, extinction, and distribution of aquatic species within the paleo–Owens system, but any paleohydrological causes will have to be extracted from a complex history.

Pleistocene deposits in Panamint Valley, California, document the changes in pluvial lake level, source water, and elevation of the regional groundwater table associated with climate change. The oxygen isotope stage (OIS) 2 and 6 lacustrine record is well preserved in surficial deposits, whereas the OIS 3–5 lacustrine-paludal and lacustrine record is mainly derived from an archived core sample. Amino acid racemization ratios in ostracodes and gastropods suggest that the shoreline and groundwater-discharge features that lie between ∼600 and 550 m elevation formed during one highstand, probably during OIS 6. A fossiliferous part of the ∼100-m-deep core DH-1, which was drilled in the Ballarat Basin during the late 1950s, was resampled in this study. Comparison of DH-1 with core DH-3 from Panamint Valley and core OL-92 from Owens Lake suggests the 34–78-m-depth interval of DH-1 may span all or much of OIS 4. The microfauna from this depth interval indicate a saline marsh or shallow lacustrine environment, but not a large lake. The ostracode assemblage requires low ratios of alkalinity to calcium (alk/Ca) water likely indicative of solutes in deep regional groundwater sources rather than the high alk/Ca solutes common to the Owens River system. OIS 2–aged sediment from surficial deposits, a shallow auger hole, and core DH-1 contain faunas, including the ostracode Limnocythere sappaensis , which require the high alk/Ca evolved solutes common to the Owens River. The elevation of the lacustrine sediments further indicates a moderate-sized saline lake around 180–200 m depth. In the northern Lake Hill basin, a saline lake persisted until at least 16 ka, and it was succeeded by fresh, groundwater-supported wetlands, which were fully developed by ca. 12,575 14 C yr B.P. and which persisted until around 10,500 14 C yr B.P., when the basin became a dry playa.

The chronologic history of pluvial Owens Lake along the eastern Sierra Nevada in Owens Valley, California, has previously been reported for the interval of time from ca. 25 calibrated ka to the present. However, the age, distribution, and paleoclimatic context of higher-elevation shoreline features have not been formally documented. We describe the location and characteristics of wave-formed erosional and depositional features, as well as fluvial strath terraces that grade into an older shoreline of pluvial Owens Lake. These pluvial-lacustrine features are described between the Olancha area to the south and Poverty Hills area to the north, and they appear to be vertically deformed ∼20 ± 4 m across the active oblique-dextral Owens Valley fault zone. They occur at elevations from 1176 to 1182 m along the lower flanks of the Inyo Mountains and Coso Range east of the fault zone to as high as ∼1204 m west of the fault zone. This relict shoreline, referred to as the 1180 m shoreline, lies ∼20–40 m higher than the previously documented Last Glacial Maximum shoreline at ∼1160 m, which occupied the valley during marine isotope stage 2 (MIS 2). Crosscutting relations of wave-formed platforms, notches, and sandy beach deposits, as well as strath terraces on lava flows of the Big Pine volcanic field, bracket the age of the 1180 m shoreline to the time interval between ca. 340 ∼ 60 ka and ca. 130 ∼ 50 ka. This interval includes marine oxygen isotope stages 8–6 (MIS 8–6), corresponding to 260–240 ka and 185–130 ka, respectively. An additional age estimate for this shoreline is provided by a cosmogenic 36 Cl model age of ca. 160 ∼ 32 ka on reefal tufa at ∼1170 m elevation from the southeastern margin of the valley. This 36 Cl model age corroborates the constraining ages based on dated lava flows and refines the lake age to the MIS 6 interval. Documentation of this larger pluvial Owens Lake offers insight to the hydrologic balance along the east side of the southern Sierra Nevada and will assist with future regional paleoclimatic models within the western Basin and Range.

Owens Lake has existed for most of the past 800,000 yr, but the sequence of interconnected lakes and streams of which it was often part, the Owens River cascade, last flourished during late Pleistocene time. A fluctuating, increasingly saline, terminal lake survived into the late Holocene until upstream water diversions to the Los Angeles Aqueduct began in 1913. Shoreline fragments and beach stratigraphy indicate that the lake reached its highest late Pleistocene level around 23.5 ka, during the Last Glacial Maximum, when it was fed by meltwaters from Sierra Nevada glaciers and spilled southward to Searles Lake and beyond. The lake then fell to relatively low levels after 16.5 ka before experiencing terminal Pleistocene oscillations related to hydroclimatic forcing, which involved changing regional precipitation regimes rather than major inputs from Sierra Nevada glaciers. Two major transgressions occurred. The first culminated around 14.3 ka and was probably related to a cooler, wetter regional climate. The second culminated around 12.8 ka and was linked to the earlier wetter phase of the Younger Dryas cold event. However, the high late Pleistocene shoreline is deformed, and the highest beach ranges in elevation from 1140 m to 1167 m above sea level. If the terminal Pleistocene lake overflowed, as suggested here, then its outlet has also been raised since 12.8 ka. This deformation appears to have involved uplift of the Coso Range magmatic complex relative to subsidence and faulting within the Owens Lake graben between the Sierra Nevada and Inyo Mountains frontal faults. Such deformation confounds simple hydroclimatic explanations of lake behavior and must be incorporated into models that seek to interpret the changing form and geochemistry of Owens Lake and the frequency of its spillage southward to Searles Lake.

Lake Manix, in south-central California, was the terminal basin of the Mojave River until the late Pleistocene, when it drained east to the Lake Mojave Basin. Based on new field observations, radiocarbon ages, and soil development, we propose modifications to previously published hypotheses on the timing of the last 543 m above sea level (masl) highstand of Lake Manix, the timing of the first discharge eastward, and the time required to cut Afton Canyon between the two basins. Subtle beach barriers, wave-cut scarps, and lagged beach gravels indicate that Lake Manix reached highstands between 547 and 558 masl at least twice prior to its previously known 543 m highstands. Properties of soils formed on beach barriers at 547–549 masl compared to soils on dated deposits suggest an age of older than 35 cal ka for this highstand. Calibrated radiocarbon ages for three lacustrine highstands at or near 543 masl are ca. 40–35 ka, 33–30 ka, and 27–25 ka. Lake Manix periodically discharged down a drainage presently located on the north rim of Afton Canyon at 539 masl. Soil development estimated from multiple buried soils within fluvial deposits and overlying fan deposits suggests that discharge was coeval with or somewhat older than the 547–549 m highstand, and that fluvial aggradation in this drainageway was followed by a period of relative landscape stability and episodic burial by alluvial-fan deposits. Strath terraces below these highest fluvial deposits, but above the canyon rim, record initial incision of the Lake Manix threshold. Surface and soil properties indicate that they are latest Pleistocene to early Holocene in age, similar to the previously studied strath terraces that are inset well below the rim and below the basal lake sediments. We suggest that the higher straths above the rim formed no earlier than ca. 25 cal ka. We interpret the soils, stratigraphy, and fluvial landforms in the canyon to indicate relatively rapid incision of Afton Canyon to the depth of the bedrock floor of Lake Manix, followed by intermittent, gradual bedrock incision.

Lakes in one form or another have characterized the western Mojave Desert since at least Miocene time. The most recent of these, Lake Thompson, developed in the late Pleistocene, when it covered as much as 950 km 2 and rose to at least 710 m above sea level. During Holocene time, the lake desiccated, and is now represented by Rogers, Rosamond, and Buckhorn dry lakes, which may flood up to 200 km 2 during unusually wet phases. The spatial dimensions of the former lake are defined by modest geomorphic and lithostratigraphic units, mostly exposed lake beds and beach ridges interbedded with and later mantled by fluvial and eolian deposits. The lake's temporal devolution is revealed by four cores, and ages are constrained by accelerator mass spectrometry 14 C dating of organic sediment. These cores show a deep perennial lake from before 36 ka to at least 34 ka, a shallow but variable perennial lake from before 26 ka to 21 ka, followed by lowering and at least partial exposure of the lake floor to deflation and alluviation. A shallow perennial lake returned during the terminal Pleistocene, from around 16.2 ka to at least 12.6 ka, forming distinctive beach ridges beyond the margins of the present dry lakes, and it may have reappeared in the early Holocene. During subsequent Holocene desiccation, lake segmentation occurred as waves and currents generated lower sequences of beach ridges around contracting lakes. These ridges became mantled with eolian sand, but, as fluvial sediment inputs diminished with increasing aridity, these dunes were degraded, and their roots survive today as indurated yardangs.

An evaluation of the poorly understood Cenozoic hydrologic history of the American Southwest using combined geological and biological data yields new insights with implications for tectonic evolution. The Mesozoic Cordilleran orogen next to the continental margin of southwestern North America probably formed the continental divide. Mountain building migrated eastward to cause uplift of the Rocky Mountains during the Late Cretaceous to early Tertiary Laramide orogeny. Closed drainage basins that developed between the two mountain belts trapped lake waters containing fish of Atlantic affinity. Oligocene-Miocene tectonic extension fragmented the western mountain belt and created abundant closed basins that gradually filled with sediments and became conduits for dispersal of fishes of both Pacific and Atlantic affinity. Abrupt arrival of the modern Colorado River to the Mojave-Sonora Desert region at ca. 5 Ma provided a new conduit for fish dispersal. Great dissimilarities in modern fish fauna, including differences in their mitochondrial deoxyribonucleic acid (DNA), indicate that late Miocene runoff from the Colorado Plateau did not flow down the Platte or Rio Grande, or through the Lake Bonneville Basin. Fossil fishes from the upper Miocene part of the Bidahochi Formation on the Colorado Plateau have characteristics that reflect a habitat of large, swift-moving waters, and they are closely related to fossil fishes associated with the Snake and Sacramento Rivers. This evidence suggests that influx of fishes from the ancestral Snake River involved a major drainage, not merely small headwater transfers.

We estimated the timing of paleodrainage connections in the Colorado River Basin using mitochondrial deoxyribonucleic acid (DNA) sequence divergences among populations of the speckled dace, Rhinichthys osculus . Cytochrome b and ND4L sequences were analyzed by maximum likelihood methods to estimate phylogenetic branch lengths, which were calibrated to geological time with a fossil age estimate. We assume that heterogeneity in rate of evolution of mitochondrial DNA is caused in part by differences in body size, temperature, and correlated life-history traits; therefore, branch lengths are used directly to calculate rates of nucleotide substitution and ages of nodes on the phylogenetic tree. Rhinichthys osculus is estimated (by the corrected age of the oldest fossil) to have diverged from its sister species at 6.3 Ma. We estimate that speckled dace have been in the Colorado drainage for 3.6 m.y., and they have dispersed through the drainage and to former connectives, such as the Los Angeles Basin, in the past 1.9 m.y. Divergence among lineages of the upper and lower Colorado River drainages (above and below Grand Canyon) is estimated to have occurred ca. 1.9–1.3 Ma. Genetic divergence of allopatric lineages in the lower Colorado River drainage was accompanied by morphological adaptations to different stream gradients, but small genetic distances among these forms indicate recent gene flow and lack of reproductive isolation.

A model is proposed whereby a Miocene Colorado River precursor canyon, deeper than 600 m, formed on the western Hualapai Plateau by headward erosion along a strike-valley drainage. Basin and Range faulting of the margin of the Colorado Plateau initiated canyon formation. This canyon was occupied by a long narrow lake, and the surface of the lake was at or above the level of the Hualapai Limestone. Such a hypothesized lake would have trapped any coarse sediment derived from the surrounding basin at the head of the lake, well upstream from the Grand Wash Trough. The drainage area feeding into the lake would have included the Hualapai Plateau and the combined ancestral drainages of Kanab and Cataract Creeks. This >13,000 km 2 basin has been dominated by surface exposures of Paleozoic carbonates since at least late Eocene time and generates no more than 1%–2% of the runoff associated with the modern (predam) Colorado River discharge. Such a carbonate-dominated, sediment-deficient basin would supply carbonate-rich runoff to the structural depocenter in the Grand Wash Trough, possibly explaining the upward transition to the Hualapai Limestone facies in late Miocene time. The upstream canyon delta produced in this proposed model could have been removed by the Pliocene-Pleistocene integration and younger incision of the more powerful, modern Colorado River.