We documented the impact of Late Pleistocene–Holocene climate change on terrace deposits and preserved channels in the unglaciated drainage of the Colorado River in central Texas (south-central United States) using integrated channel morphology and provenance analysis. Detrital zircon (DZ) U-Pb ages (n = 1850) from fluvial terrace deposits and new quantitative analysis of fluvial channel morphology based on LiDAR data were used to reconstruct sediment provenance and shifts in paleohydraulic conditions during Late Pleistocene to Holocene aridification. These data reveal a reduction in fluvial channel size and discharge temporally coupled with a rapid shift in erosion locus and dominant sediment sourcing, from the Southern Rocky Mountains to the Llano area, during the glacial-interglacial transition. Geomorphic mapping and morphometric analysis show narrowing of river channels linked to diminishing Colorado River discharge. DZ data show an abrupt shift to erosion in the lower drainage basin and the remobilization of older terraces due to river incision and lateral channel migration. We attribute these systematic changes to upper-basin contraction caused by drainage reorganization and aridification during the Late Pleistocene, as well as the onset of enhanced convective precipitation sourced from the Gulf of Mexico, driving focused erosion along the topographic edge of the Llano uplift in central Texas since the early to mid-Holocene.

In response to climate perturbations, fluvial drainage basins record sedimentologic and hydraulic changes, including changes in river morphology, discharge, erosion loci, or sediment production. Several provenance studies have focused on basin-wide signals of deglaciation in response to Late Pleistocene–Holocene glacial-interglacial transitions (Sharman et al., 2021; Jonell et al., 2018, 2017; Alizai et al., 2011), but less is known about the response of mid-latitude unglaciated fluvial drainage systems like the Colorado River basin of Texas (south-central United States). Existing research has examined the impact of glacio-eustatic climate variations on upstream fluvial morphology and discharge (Hassenruck-Gudipati et al., 2022; Blum et al., 2013; Baker and Penteado-Orellana, 1977). However, questions remain about the impact of changes in climate and intensity of convective storms on sediment generation and transfer in the drainage basin since the last glacial period.

Recently, detrital zircon (DZ) U-Pb provenance data have been used as tracers of erosion and sediment transfer in Quaternary systems to track climate-driven changes in sediment sourcing of submarine fans fed by large river systems like the Mississippi and Amazon Rivers (Mason et al., 2019; Blum et al., 2018; Fildani et al., 2018; Sickmann et al., 2016). This study combines DZ U-Pb provenance and fluvial morphometric analysis to investigate erosion and sediment sourcing changes in the drainage of the Colorado River of Texas to reconstruct its erosional and discharge evolution since the Late Pleistocene. The Colorado River is a prime example of an unglaciated drainage in the Late Pleistocene and flows through several physiographic provinces with variable bedrock geology with distinct DZ U-Pb age signatures. This study focuses on the Quaternary fluvial terraces in the lower drainage basin that recorded the last glacial and post-glacial periods (Blum et al., 1995). We present ~1850 new DZ U-Pb ages integrated with airborne LiDAR data characterizing preserved terrace and paleochannel generations to document coupled changes in fluvial morphology (river width), paleodischarge, and locus of sediment sourcing and erosion within the Colorado River drainage since the onset of the last glacial period (115 ka).

The Colorado River drainage spans 109,500 km2, extending across Texas from the Gulf of Mexico to headwaters in New Mexico (Figs. 1A and 1B). The river drains a basin with heterogeneous geology and flows from a relatively steep mixed-bedrock valley to a low-relief coastal plain underlain by Cenozoic siliciclastic and Quaternary alluvial strata. The modern river delivers an estimated water discharge of 806 m3/s and sediment load of ~13.3 MT/yr to the Gulf of Mexico, reflecting reduction in catchment area and river length following the last glacial period (Blum and Hattier-Womack, 2009). We aim to further explore the role of upstream changing climate in these changes, building upon extensive work done to characterize climate in the basin. The ages of Quaternary terraces span the last glacial and post-glacial periods, being entrenched within the alluvial deposits of the lower basin and resting unconformably atop Cretaceous and Tertiary strata. They are classified as follows based on age and geomorphology: (1) Late Pleistocene terraces (115.4 ka) corresponding to the Montopolis terraces as defined by Weber (1968) and the 6R channel of Baker and Penteado-Orellana (1977) and mapped as the Eagle Lake Alloformation by Blum and Valastro (1994); (2) latest Pleistocene–mid-Holocene terraces (20–5 ka) equivalent in age to both the Eagle Lake Alloformation (20–14 ka) and Columbus Bend Allomember (CBA)-1 (12–5 ka) as defined by Blum and Valastro (1994) and to the 6, 6A, 6B, and 5 channels as defined by Baker and Penteado-Orellana (1977); (3) late Holocene terraces (4–1 ka) corresponding to CBA-2 (5–1 ka) as defined by Blum and Valastro (1994); and (4) latest Holocene terraces (<1 ka) corresponding to CBA-3 (ca. 0.6 ka) as defined by Blum and Valastro (1994).

Using diagnostic DZ U-Pb age signatures from the different geological and physiographic provinces within the catchment based on existing data, it is possible to leverage the DZ U-Pb provenance record to elucidate basin-wide temporal changes in erosion locus and sediment sources including the Southern High Plains; the North Central Plains and Edwards Plateau that combined are the Plains-Plateau province; the Llano region; and the Inner and Outer Coastal Plain (Fig. 1C). Overall observed DZ age modes fall into the following categories: Yavapai-Mazatzal (1800–1600 Ma), Mid-Continent (1500–1300 Ma), Grenville (1300–900 Ma), Peri-Gondwanan–Pan-African (760–300 Ma), Cordilleran arc (252–85 Ma), Laramide (85–50 Ma), and Cenozoic (<50 Ma). The bedrock geology of the Colorado River drainage basin is characterized by five distinct composite DZ signatures, namely (1) a Rocky Mountains DZ signature comprising age ranges of 1800–1600 Ma, 1500–1300 Ma, 1300–900 Ma, 252–85 Ma, 85–50 Ma, and <50 Ma, with a dominant Yavapai-Mazatzal (1800–1600 Ma) age mode; (2) a Southern High Plains DZ signature included as part of category (1), given their recycled Rocky Mountains nature; (3) a Plains-Plateau DZ signature that exhibits a dominant Peri-Gondwanan–Pan-African age mode (760–300 Ma) and subsidiary age modes at 1300–900 Ma, 1800–1600 Ma, 1500–1300 Ma, and 252–85 Ma (in decreasing order); (4) a Llano signature that comprises only Grenville (1300–900 Ma) and Mid-Continent (1500–1300 Ma) age modes; and finally (5) a Coastal Plain DZ signature that includes 1800–1600 Ma, 1500–1300 Ma, 1300–900 Ma, 252–85 Ma, 85–50 Ma, and <50 Ma age modes. While having a signature similar overall to the Rocky Mountains signature, the Coastal Plain can be subdivided on the basis of its DZ signature, with the Outer Coastal Plain showing Triassic zircons (ca. 230 Ma) in the dominant Cordilleran arc (252–85 Ma), Laramide (85–50 Ma), and Cenozoic (<50 Ma) age modes and the Inner Coastal Plain being dominated by Cenozoic (<50 Ma) zircons. These different DZ signatures allow us to attribute detritus to erosional source regions within the drainage to reconstruct the erosional evolution during Late Pleistocene–Holocene climatic changes.

We used a 50-cm-resolution LiDAR-derived digital elevation model (DEM) (detrended) to map fluvial terrace levels and measure preserved paleochannel widths (n ≈ 200) along a ~150-km-long segment between the Smithville and Columbus locations in the Colorado River of Texas (Fig. 1B). We identified and mapped three distinct terrace levels that include Late Pleistocene (115.4 ka optically stimulated luminescence [OSL] age), latest Pleistocene to mid-Holocene (20–5 ka), and latest Holocene (4–1 ka) terraces. The three mapped terrace levels exhibit detrended elevation ranges of 12–14 m, 8–12 m, and ~8 m, respectively. Within the two lowest levels, we identified paleochannel–point bar assemblages that show a dramatic decrease in paleochannel width from ~585 m during the Late Pleistocene to ~325 m during the Holocene (Figs. 2A and 2B). The progressive downcutting of younger and narrower channels leads to incision of older terraces (Fig. 2A) and remobilization of sediment from Pleistocene terraces as observed in the modern river channel.

From the mapped fluvial terraces at the Smithville (Fig. 2A) and Columbus (Fig. S3C in the Supplemental Material1) locations, we collected 14 medium-grained sandstone to gravelly conglomerate samples for DZ U-Pb provenance. These samples underwent standard mineral separation techniques and laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) zircon U-Pb analyses at the UTChron laboratory at the University of Texas at Austin. Multi-dimensional scaling (MDS) and dissimilarity analyses were used to compare DZ age distributions. Detailed methods and data sets are provided in the Supplemental Material. DZ signatures in chronological order include Late Pleistocene terraces (115.4 ka) characterized by recycled Grenville (34%) and Peri-Gondwanan–Pan-African (17.2%) age modes and lack of Cenozoic zircon (<1%). MDS and dissimilarity matrix (Euclidean distance, Ed) analyses suggest a provenance affinity linked to the Plains-Plateau (Ed = 0.24) and Rocky Mountains (Ed = 0.4) provinces (Fig. 3). In contrast, the late Holocene terraces (4–1 ka; OSL age: 2.6 ka) are marked by an increase in Grenville DZ input (42%) and a reduction in Peri-Gondwanan–Pan-African zircons (13%), suggestive of a spatial shift in sediment sourcing to the Llano uplift. MDS plots and a dissimilarity matrix value of Ed = 0.47 support this shift to a Llano DZ signature. However, younger late Holocene terraces (OSL age: 2.5 ka) appear to resemble Late Pleistocene deposits but with a lower proportion of Grenville zircons (32%). Statistical analysis supports a lower affinity of late Holocene terraces (2.5 ka) with the Plains-Plateau DZ signature (Ed = 0.26) and a significant increase in affinity with Coastal Plain DZ signatures (inner, Ed = 0.37; outer, Ed = 0.4) compared to the late Holocene terraces (4–1 ka; OSL age: 2.6 ka). Latest Holocene deposits (<1 ka) exhibit DZ signatures diagnostic of the Cenozoic Coastal Plain marked by an increase in the contribution of Cenozoic (14%) and Cordilleran (9%) zircon modes compared to older terrace deposits (Late Pleistocene and Holocene). MDS and dissimilarity analyses support a higher affinity with the Coastal Plains (Inner, Ed = 0.23; Outer, Ed = 0.32) and also show affinity with Southern High Plains (Ed = 0.23) DZ domains. The analysis is limited to paleochannel and fluvial terrace preservation, with negligible uncertainty in DZ U-Pb ages.

This study reconstructs erosional and hydraulic changes in the central Texas Colorado River drainage basin in response to early to mid-Holocene climate and precipitation changes. The Late Pleistocene terrace deposits (115.4 ka) show an average detrended height of 12–14 m and a dominance of Plains-Plateau and Rocky Mountains DZ signatures. The headwaters region included eastward-flowing streams that originated at the Rocky Mountains and flowed across the Llano Estacado and Simanola Valley, New Mexico (Hall, 2001; Holliday, 1995). A longer Colorado River compared to the modern river implies a larger drainage basin that extended farther northwest (Fig. 4A). MDS-DZ provenance data suggest the southeastern Rocky Mountains were the dominant sediment source during the Late Pleistocene. The latest Pleistocene to mid-Holocene terrace deposits (20–5 ka) show an average detrended height of 8–11 m. Paleochannels preserved within these terraces show an average channel width of ~585 m, consistent with elevated river discharge during the Late Pleistocene. Toomey et al. (1993) proposed high regional effective moisture during the last glacial maximum at 22 ka. High effective moisture conditions, coupled with a larger catchment area, strongly supports a larger river with higher discharge.

For the early to mid-Holocene (Fig. 4B), central Texas speleothem records suggest warmer and more arid conditions within the Colorado River drainage following the glacial-interglacial transition (Ellwood and Gose, 2006; Musgrove et al., 2001). Basin-wide aridification resulted in a reduction in river discharge as manifested by markedly smaller channels (~323 m) in late Holocene (2.6 ka) terrace deposits and a switch to DZ provenance reflective of dominant sourcing from Llano uplift. We propose that this provenance change is associated with a shift in the locus of erosion within the drainage to the Llano uplift flanked by the Balcones escarpment. This change is likely driven by (1) the beheading of the upper Colorado River drainage as headwater rivers diverted southward during post-glacial times and the onset of more arid interglacial conditions (Hall, 2001; Holliday, 1995), particularly in the Southern High Plains, dramatically reducing river discharge, and (2) upper basin aridification accompanied by focused convective precipitation (Sun et al., 2021; Toomey et al., 1993) and erosion in the Llano uplift, as seen today, driven by moisture from the Gulf of Mexico. Increasing DZ sediment provenance from the Llano uplift is also corroborated by high Holocene denudation rates and thinner soil coverage based on cosmogenic and strontium data (Hidy et al., 2014; Cooke et al., 2003). Blum et al. (1994) also argued that Holocene storms removed hillslope sediment in the Llano and Edwards Plateau region, contributing to the observed change in our DZ provenance data (Fig. 3B). The DZ age distributions of younger late Holocene terrace deposits (2.5 ka) resemble a mixture of diagnostic DZ age signatures from the Coastal Plain and recycled Late Pleistocene terrace deposits. This trend is likely linked to the cannibalization of Pleistocene terraces and mixing with Llano and Coastal Plain–derived detritus during late Holocene incision. The significant latest Holocene (<1 ka) increase in Cenozoic zircons (14%), typical of the Inner Coastal Plain, likely reflects increased river incision within the coastal plain that reached Cenozoic bedrock and sediment generation through remobilization and/or recycling of Late Pleistocene to Holocene terraces. Ongoing sediment remobilization is also readily apparent along the modern river and driven by storm-induced bank failure.

In conclusion, this integrated morphological and provenance study of Late Pleistocene to Holocene terrace deposits along the Colorado River in central Texas demonstrates the impact of climate change on river morphology, discharge, and erosion loci since the Late Pleistocene. Our new data reveal that since the onset of the last glacial period, the Colorado River drainage experienced (1) contraction of the upper portions of the catchment due to capture and southward diversion of originally eastward-flowing Rocky Mountain tributaries during the latest Pleistocene, (2) Holocene aridification of the upper drainage basin resulting in lower average discharge as manifested by a reduction in river channel width, (3) Holocene onset of increased denudation in the Llano uplift due to focused precipitation of Gulf of Mexico derived moisture, and (4) Holocene incision of the lower Colorado River and recycling of Pleistocene terraces and Cenozoic strata. These results demonstrate a relatively rapid response time (~104 k.y.) of the Colorado drainage to climate and precipitation changes. The study underscores the power of integrating geomorphology and DZ provenance as a tracer for constraining climate-modulated impacts on sediment sourcing, erosion evolution, and paleohydraulic conditions.

1Supplemental Material. Detailed provenance, LA-ICP-MS and geomorphic methodologies, U-Pb data tables, and supplemental figures. Please visit https://doi.org/10.1130/GEOL.S.23671341 to access the supplemental material, and contact editing@geosociety.org with any questions.

We thank Jacob Covault for insightful discussion. We thank Matt McKay for thoughtful reviews. Financial support for this work was provided by UTChron Laboratory at the University of Texas at Austin, and by the members of the Quantitative Clastics Laboratory research consortium at the Bureau of Economic Geology (BEG), The University of Texas at Austin. This project was supported by Geological Society of America and American Association of Petroleum Geologist Foundation grants to Gutiérrez.

Gold Open Access: This paper is published under the terms of the CC-BY license.