Volume 125, Issue 6 e2019JF005443
Research Article
Free Access

Quantifying Coastal Fluvial Morphodynamics Over the Last 100 Years on the Lower Rio Grande, USA and Mexico

John M. Swartz

Corresponding Author

John M. Swartz

Department of Geological Sciences, Jackson School of Geosciences, The University of Texas at Austin, Austin, TX, USA

Institute for Geophysics, Jackson School of Geosciences, The University of Texas at Austin, Austin, TX, USA

Correspondence to: J. M. Swartz,

[email protected]

Search for more papers by this author
Timothy A. Goudge

Timothy A. Goudge

Department of Geological Sciences, Jackson School of Geosciences, The University of Texas at Austin, Austin, TX, USA

Search for more papers by this author
David C. Mohrig

David C. Mohrig

Department of Geological Sciences, Jackson School of Geosciences, The University of Texas at Austin, Austin, TX, USA

Search for more papers by this author
First published: 07 June 2020
Citations: 3


Downstream changes in fluvial channel morphology are commonly observed in association with the backwater zone, where rivers transition from quasi-uniform flow with normal-flow depth to gradually varying flow. This transition is linked to changes in channel morphology and mobility and resulting fluvial stratigraphy. However, the majority of systems studied to date are perennial rivers with relatively consistent flow conditions. Here we investigate the evolution of a large river with significant flood-to-baseflow variability as it transverses and builds a large delta. We provide the first comprehensive study of the morphology and morphodynamics of the lower Rio Grande, a major continental drainage system that enters the western Gulf of Mexico. We quantify the morphology of the current Rio Grande channel and document spatial trends in channel geometry and kinematics using lidar, historical surveys, and hydrographic analysis. The modern Rio Grande channel morphology does not significantly vary toward the coast. Rather, the channel width, levee, and bed slopes remain nearly constant for ~200 river km. We find historical migration rates between 10 and 100 m/yr with no significant reduction toward the coast in contrast to previously studied systems. We propose that this invariant channel geometry and sustained high migration rates are signatures of the channel not requiring adjustment within the lower coastal reach to accommodate baseflow conditions, and the channel remains continuously adjusted solely to peak flow conditions.

Key Points

  • The lower Rio Grande maintains a consistent channel geometry upstream and downstream of the estimated backwater transition
  • Flow intermittency drives sustained lateral migration rates leading to channel bend cutoffs near the coast and little variation in channel belt width
  • Alluvial rivers with a high range of discharge variability have different morphodynamics than perennial systems and their resulting geomorphology and stratigraphy may not show the same trends

1 Introduction

The geomorphology of alluvial rivers and how they evolve in response to changing boundary conditions, whether natural or anthropogenic, has long been an area of intense study (e.g., Lane, 1957; Leopold & Maddock, 1953; Schumm, 1963). Particularly of interest are the dynamics of rivers within the coastal zone: the critical transitional region where fluvial processes act to build deltas and alluvial plains as well as transfer material to the marine environment. While advances have been made in capturing the modern sediment transport processes of rivers, as well as experimental work exploring the role of varying discharge and sediment flux in fluvial morphodynamics, the number of studies that explicitly examine the morphology and kinematics of rivers as they approach the coastline remain relatively small and as a result gaps remain in fully understanding the morphodynamics of alluvial systems. We examine the Rio Grande, a continental-scale river that drains western North America to the Gulf of Mexico, as it provides an example of a coastal alluvial system with a large range of discharge variability and which exhibits the unusual kinematic behavior of meander bends growing to the point of cutoff all the way to the coast (e.g., Campbell, 1927; Fernandes et al., 2016). Rivers approaching a receiving basin are known to undergo a hydraulic transition from quasi-uniform to gradually varying flow, known as backwater flow, an effect long studied by engineers (e.g., Chow, 1959). The backwater effect has a characteristic length that can be approximated by Lb = H/S, where H is the bankfull channel depth and S is the channel gradient of the normal reach (Paola & Mohrig, 1996). In coastal rivers, Lb is well approximated by the position where the mean channel bed elevation falls below sea level and can occur tens to hundreds of kilometers upstream from the coastline in gently sloping alluvial rivers. More recent work has begun to characterize how this effect plays a role in myriad fluvial processes and helps shape resulting landforms and stratigraphy by modulating sediment transport on varied spatial and temporal scales. Processes and products thought to be influenced include where avulsion occurs on river deltas (e.g., Chatanantavet et al., 2012; Edmonds et al., 2009; Jerolmack, 2009), how channel geometry changes downstream (e.g., Lamb et al., 2012; Nelson & Smith, 1989; Nittrouer et al., 2012; Smith et al., 2020), rates and spatial patterns of lateral channel migration (Fernandes et al., 2016; Hudson & Kesel, 2000; Smith & Mohrig, 2017), and depths of scour expected to occur autogenically (Ganti et al., 2019; Trower et al., 2018), among others. River bend migration has been recognized to be the sum of bank pull and bar push (Eke et al., 2014). These two components are largely independent from each other, but over relatively short time scales compensate so to preserve a statistically constant width to depth ratio and therefore are a primary control on alluvial channel geometry (Mason & Mohrig, 2018). Within backwater mediated alluvial rivers systematic downstream changes in width, depth, and cross-sectional area are observed in conjunction with the transition from normal to backwater flow (e.g., Nittrouer et al., 2012). In addition, alluvial systems commonly exhibit a marked decrease in lateral migration rates as they approach the coast, and this decrease is associated with these changes in channel cross-sectional shape (Fernandes et al., 2016).

The specific observed trends in channel processes and geometry in coastal alluvial rivers such as increasing channel depths and decreasing channel widths and lateral migration rates have all been linked to the backwater transition due to the assumption of persistent transport of water and sediment within the uniform-flow reach (e.g., Chatanantavet et al., 2012). Continuous bedload transport at all discharge conditions within the uniform flow zone necessitates adjustment at the backwater transition and associated spatial declaration, which leads to minimal sediment transport within the backwater during low and mean flow conditions in particular channel deepening and a reduction in point bar volumes (e.g., Mason & Mohrig, 2018; Nittrouer et al., 2012). In order for these spatial trends to exist persistent sediment transport must exist at all discharge conditions above the backwater transition, an assumption that is wholly reasonable for perennial systems but which may not be applicable to intermittent or highly variable rivers. Much of our understanding of coastal fluvial processes stems from physical and numerical experiments as well as modern observations. However, many of the best characterized natural systems that have helped develop our understanding of these morphodynamics are perennial rivers where baseflow conditions are relatively high year-round. This contrasts with rivers where the baseflow is minimal or close to 0 for much of the year. While intermittent river systems are becoming an area of increasing research from the lens of hydrology, water resources, and ecosystem dynamics (e.g., Allen et al., 2019; Datry et al., 2014; Nadeau & Rains, 2007), they have received less attention in terms of geomorphology and sedimentary processes. This study adds to the recent work on intermittent systems and fluvial morphodynamics by providing a detailed geomorphic characterization of a large coastal alluvial river using time-lapse observations to define its geometry and kinematics on longer time scales. While the location of the lower Rio Grande may have discouraged the level of interest and study that other river systems have seen, it also provides opportunity in the form of detailed surveys and measurements spanning over 100 years. We deploy a suite of measurements and methods to accomplish the following goals: (1) develop a quantitative geomorphic characterization of the lower Rio Grande, using aerial lidar data, (2) quantify the rate of channel bend migration for a reach of over 200 km using time-lapse analysis of digitized historical topographic surveys, and (3) perform hydrographic analysis of flow intermittency and the anthropogenic effects of dam construction. From the results of this work, we illustrate intriguing trends in the Rio Grande's channel geometry throughout the coastal reach, and similarly document the properties of the channel that lead to the unusual migration of river bends to the point of cutoff all the way to the coast prior to the damming of the river.

2 Study Area

This study provides the first quantitative assessment of the lower Rio Grande, a dynamic and underappreciated sedimentary system. The Rio Grande, or Rio Bravo del Norte, is a 3,050 km long continental river that drains an area of nearly 550,000 km2 including the southern Rocky Mountains in Colorado, USA, New Mexico, and much of northeastern Mexico and southwestern Texas (Patiño-Gomez et al., 2007; Figure 1). Despite its size, the majority of the basin is within a semiarid climate and discharge is highly variable and dominated by extended periods of minimal baseflow punctuated by large flooding events (Blythe & Schmidt, 2018). The upper Rio Grande discharge primarily corresponds with snowmelt from the southern Rocky Mountains, while the lower Rio Grande flow regime is dominated by rainfall in the Rio Conchos tributary originating in the Sierra Madre Occidental of Mexico (Blythe & Schmidt, 2018; Gonzalez-Escorcia, 2017). For almost 2,000 km it forms the border between the United States and Mexico and is subject to significant anthropogenic modification and water use in both countries. As it approaches the Gulf of Mexico it forms what is known as the lower Rio Grande Valley and delta, a 13,000 km2 area surrounding the modern channel comprised of floodplain deposits and abandoned meander belts formed by avulsions of the Rio Grande during the Holocene. This delta forms one of the largest depositional centers in the Gulf of Mexico. The western Gulf of Mexico coast is microtidal with less than a 0.5 m range (Lohse, 1958). The area is home to over 3 million people and forms a critical agricultural region for both the United States and Mexico. Additionally, the landscape of the valley and delta surrounding the modern Rio Grande River, dominated by avulsions and oxbow lakes, forms critical riparian habitat for numerous indigenous and migratory species. Over the last 150 years its dynamic nature has led to numerous political disputes, as the common floods and avulsions led to rapid cutoff and transfer of lands from one countries side of the river to the other (Liss, 1965). Despite its importance economically, ecologically, politically, and societally the lower Rio Grande has been poorly studied. To our knowledge there have been only a handful of works focused on understanding the fluvial dynamics and sedimentary processes that have led to the creation of this unique landscape (Beck, 1928; Fulton, 1976; Lohse, 1958).

Details are in the caption following the image
Map of the study area. (a) The Rio Grande Delta of United States and Mexico. The black line indicates the modern course of the Rio Grande/Rio Bravo del Norte. The yellow star shows the position of the Falcon Dam. The inset panel shows the position of the delta in the western Gulf of Mexico and the outline of the Rio Grande and Rio Conchos basins (labeled RG and RC, respectively). (b–d) Lidar DTM showing the sinuous character of the Rio Grande at the uppermost, middle, and coastal portion of the studied reach. The lower Rio Grande has a meandering, aggradational character for the entirety of the coastal reach with numerous preserved cutoffs present along the modern channel.

3 Data and Methods

The study area (Figure 1) extends from the mouth of the Rio Grande at the Gulf of Mexico (0 river km), to near Hidalgo, Texas (250 river km). This region is the bulk of coastal Rio Grande located on the delta below the Falcon-Amistad Dam and reservoir and was chosen to maximize coverage by available data sets.

The Rio Grande River's location defines the boundary between the United States and Mexico and as such has been repeatedly surveyed since the 1850s. Historical maps have the potential to help constrain the geomorphic behavior of systems on time scales beyond that typically covered with modern methods (James et al., 2012). Three historical topographic surveys were located and used in this work (Figures 2a2c). The first is a trans-National boundary survey conducted by the International Boundary and Water Commission (IBWC) between 1910 and 1911 (Figure 2a). Covering the entire study area, this map was created using advanced surveying techniques for the time and contains detailed subaerial elevation data in both the United States and Mexico and also bathymetric measurements of the channel bed. Additionally, the map includes the channel course of the Rio Grande in 1897, providing an additional channel centerline for use in migration rate analysis. The second is a U.S. Geological Survey (USGS) topographic survey collected in 1930 from the Gulf of Mexico to Barreda, Texas (Figure 2b). The third is also a USGS topographic survey collected in 1955–1956 to survey channel position following construction of the Falcon Dam from 1951–1954 (Figure 2c). Together, these three maps constrain the position of the Rio Grande in 1897, 1912, 1930, and 1956. All maps were converted to TIFF format and georeferenced based on map grid coordinates as well as surveyed monument locations using ESRI ArcGIS 10.3 georeferencing tools. The USGS topographic maps were surveyed using the NAD1927 datum, while the 1912 survey used a system that formed the precursor to this standard.

Details are in the caption following the image
Example maps of the historical surveys and lidar data set used in the time-lapse analysis. Approximately ~100 river km upstream from the coast line, all panels have the same extent. North of the Rio Grande is the United States, while south is Mexico. Flow is to the right. (a) The 1912 IBWC survey that includes detailed topographic and bathymetric points as well as the course of the 1897 channel. (b) The 1930 USGS survey digitized from 1:10000 scale panels. (c) The 1956 USGS survey digitized from 1:10000 scale panels. (d) The 2011 lidar DTM at 3 m resolution. Notice the change in position of the channel centerline between (a) and (c) but the similarity of (c) and (d). IBWC: International Boundary and Water Commission. USGS: U.S. Geological Survey.

Past topographic information was used only from the 1912 survey owing to its unusual detail in both topographic and bathymetric data. To assess the vertical accuracy of these measurements, we located permanent IBWC and USGS monuments within the 1912 maps that served as base reference locations during the survey and often still exist today. Located several hundred meters away from the Rio Grande channel, they should not have undergone significant elevation changes in the intervening 100 years. The surveyed 1912 elevation of these monuments (converted to modern mean sea level, MSL) was compared to their elevation as measured in a 2011 lidar survey also conducted by the IBWC (Table 1).

Table 1. Comparison of Vertical Accuracy of 1912 Survey and 2011 Lidar Topography
Monument no. Northing Easting 1912 elevation (m) 2011 elevation (m) Difference (m)
53 49,659.5 176,200 1.1 1.6 −0.5
52 51,655.1 171,076.5 8.66 8.31 0.35
51 52,836 166,713.6 3.82 3.95 −0.13
50 54,253.5 165,070.8 3.93 eroded N/A
49 56,999.9 163,633.1 5.13 5.33 −0.2
48 61,255.1 163,765.2 7.9 7.58 0.32
47 61,405.2 159,897.4 8.57 8.68 −0.11
46 58,671 156,979.3 9.37 eroded N/A
45 56,221.6 155,176.1 9.8 9.55 0.25
44 56,257.6 152,438.4 10.39 10.51 −0.12
43 55,643.2 151,653.5 10.69 10.9 −0.21
42 52,731.6 149,058.6 11.47 10.93 0.54
41 50,999.8 146,948.6 11.11 11.07 0.04
40 49,588.9 143,491.3 12.88 13.27 −0.39
39 46,930.7 141,784.8 14.05 13.78 0.27
38 42,942.5 140,468.1 13.72 13.03 0.69
37 41,868.6 137,002.7 15.55 eroded N/A
36 40,012 134,646.5 15.13 15.45 −0.32
35 39,998.4 130,287.4 16.28 16.59 −0.31
34 39,146.4 125,737.7 17.86 17.71 0.15
33 38,144.9 122,235.6 18.56 18.15 0.41
32 36,776.3 117,761 19.57 19.66 −0.09
31 35,510.1 114,995.6 20.9 20.59 0.31
30 35,775.1 113,356.7 20.3 20.2 0.1
29 35,772 110,191.4 21.45 21.61 −0.16
28 37,220.3 107,041.6 22.99 23.13 −0.14
27 35,657.4 101,854.2 22.66 23.31 −0.65
26 36,783.3 97,524.1 24.91 25.11 −0.2
25 36,411.1 93,927.1 25.33 25.16 0.17
24 37,400 91,090.7 27.4 26.49 0.91
23 35,820.8 84,696.9 29.1 28.93 0.17
  • Note. Survey monument elevations from the 1912 survey and the 2011 lidar elevation measured at same location. Three monuments (50, 46, and 37) had been eroded by lateral river migration. The average difference between the surveys was 0.04 m. There is no systematic trend or bias observed between the two data sets.

Modern elevation data for the Rio Grande were provided by a bare-earth aerial lidar digital terrain model (DTM). The lidar data set was collected by the IBWC from December 2010 to March 2011 using a Leica ALS50 dual sensor lidar system with a reported vertical accuracy of 5 cm and a nominal point spacing of 0.7 m. DTM tiles were imported and mosaicked using ArcGIS 10.3 and analyses carried out using Spatial and 3-D analyst toolboxes (Figures 1a1d and 2d).

The 2011 lidar survey provides excellent detail of the subaerial geomorphology of the studied reach but did not penetrate the water surface and therefore provides no details of the subaqueous channel. However, the IBWC and the Texas Commission on Environmental Quality (TCEQ) conducted a field theodolite survey for use in flood modeling in 2013. This 2013 survey consists of elevation cross sections approximately 1 km in length spaced in 100–500 m intervals from just below Brownsville to near Rio Grande City (Gonzalez, 2017; Figures 3a and 3b). The cross sections were imported into ArcGIS and the channel bed and levee crest elevations extracted.

Details are in the caption following the image
Example of TCEQ topographic cross sections collected in 2013. (a) Location of example Cross Section TCEQ 11LVRG371. These profiles were collected for flood modeling and so cross over the Rio Grande, with increasing point density within the channel. (b) Topographic cross section of the Rio Grande. Note the similarity in levee crest elevations on either side of the river, and the channel base. Levee crest elevations and channel bed elevations were extracted from each cross section to create river longitudinal profiles of each.

Nearly 250 river km were included in this study from the Gulf of Mexico to near Hidalgo, Texas. This reach is captured in all of our historical, topographic, and bathymetric data sets in 1912 and 2011–2013. To assess potential trends in morphology along the lower Rio Grande, we construct longitudinal profiles of levee crest elevations, channel bed elevation, and bankfull channel width along continuous profiles in the 2011–2013 data set and at 1,000 m intervals in the 1912 data. All values are presented relative to MSL. Levee crest elevations were measured for both the north and south levees, and the average reported. The levee crest elevation relative to the bed elevation defines the bankfull channel depth (Pizzuto, 1987). Bankfull channel width is defined as the distance between the respective levee crests of a river. The natural levees of the Rio Grande appear to still be intact and measurable in both 1912 and 2011 as the artificial levees built for flood control are emplaced several hundred meters away from the active channel, and the region in between left to flood naturally.

The Rio Grande is a highly sinuous river and was historically subject to rapid migration, loop cutoffs, and avulsion (Stanley & Randazzo, 2001) The channel centerline was digitized for the four historical intervals (1897, 1912, 1930, and 1956) as well as 2011. Bends in each centerline were identified on the basis of inflection points at each bend apex (Figure 4a). Values of lateral migration were calculated for each bend for each interval by measuring the distance between each bend apex at each time step (Figure 5). Due to the significant bend deformation that occurs between each time step, we conservatively chose to only measure lateral migration at bend apex, or inflection point, for bends that could be reliably observed across time intervals. We acknowledge that this method potentially underestimates the actual maximum migration rate of each bend compared to other methods (e.g., Finotello et al., 2018; Sylvester et al., 2019), but we have chosen it to allow for direct comparisons to previously published migration rates of the Trinity and Mississippi Rivers (Hudson & Kesel, 2000; Smith, 2012). The migration distances are divided by the time in between each observation to produce an average annual migration rate in meters/year. We additionally standardize the resulting migration rates by dividing the rate by mean channel width (averaged over the study reach) to allow for comparison of the Rio Grande system to other rivers where migration rate analysis has been conducted (Hudson & Kesel, 2000; Smith, 2012). The significance of changes in migration rates both between time intervals was tested using Welch's analysis of variance (ANOVA) applied to each groups migration rates followed by Tukey multiple comparison procedures (MCP) (Tukey, 1977; Welch, 1951). Additionally, potential changes in migration rates within each time interval as the river enters the estimated backwater reach were assessed using unpaired t tests between the backwater zone (0–70 km) and the normal reach (70–200 km). This allows for a more detailed understanding of how overall migration rates may have changed over the last century as well as testing for any systematic changes associated with position relative to the shoreline.

Details are in the caption following the image
Conceptual schematic detailing elements of an alluvial, meandering river. (a) Sketch detailing the different fluvial elements and parameters measured in this work. The Cross Section X-X′ is the location of (b). (b) Idealized cross section of an alluvial river. Bankfull width is defined from levee crest to levee crest, while bankfull depth is the elevation of the levees minus the channel bed elevation.
Details are in the caption following the image
Example of migration rate analysis. The 2011 lidar DTM of a river bend along the Rio Grande with channel centerlines digitized from historical maps and the 2011 lidar. White dots indicate the bend apex, or inflection point, and black arrows show the distance between each point that is used to measure the migration rate. Note the significantly shorter distance moved between 1956 and 2011 compared to previous time steps.

The channel belt of the Rio Grande, defined as the corridor of maximum extent of lateral migration and cutoff, was also measured. A polygon was created using ArcGIS 10.3 that bounds the outer edges of the furthermost channel bends and cutoffs identifiable within the 2011 lidar data (Figure 4a). The width of this polygon along the river corridor was measured at 5 km intervals and is used to represent the channel belt width.

A single IBWC/USGS gaging station is located 75 river km upstream from the Rio Grande mouth at Brownsville, TX (USGS 08475000), and provided discharge data. This station provides mean daily discharge data from 1934 to 2012, spanning the majority of our interval of interest. Importantly, the record covers a significant time period prior to the closure of the Falcon Dam in 1954 that led to drastically reduced flow in the lower coastal reach of the Rio Grande (Figure 6a). We use the resulting discharge record to understand flow conditions and variability of the lower Rio Grande in the predam (1934–1951) and postdam (1954–2011) intervals and build empirical cumulative density functions for each interval to assess differences in median flow as well as peak floods (Figure 6b).

Details are in the caption following the image
Mean daily discharge record of the Rio Grande at Brownsville. Hydrograph recorded mean daily discharge from January 1934 to January 2014. The total number of records is 28,488. The Falcon Dam was constructed from 1951–1954 and began impounding flow in 1954. Black dots represent predam discharge, while blue are postdam. (a) Daily hydrograph from 1934–2011. (b) Empirical cumulative density function of mean daily discharge and summary statistics. Note the significant shift of the hydrograph distribution postdam, where flows of below 100 m3/s make up over 90% of the record compared to 50% predam.

4 Results

The Rio Grande is an alluvial, meandering river system for the entirety of the studied reach. It exhibits cutoffs along the length of the river including numerous bend cutoffs present at the coast (Figure 1d). Morphology of the modern (2011–2013), postdam Rio Grande shows linear longitudinal trends across all measured morphometric characteristics (Figure 7a). The 2011 levee crest elevation profile reaches a maximum of 28 m at ~225 river km and linearly grades to 0 m at the river mouth, with a slope of 1 × 10−4 (Figure 7a). There exists some variability along this extracted profile that is due to small-scale incision and slope failure of the levee crest into the river channel, but the overall trend exhibits no slope breaks (concavity/convexity) or changes along the entirety of the studied reach. Similarly, the 2013 channel base is at 17 m at 225 river km and descends linearly to −2 m at 50 river km, where the cross-section data set ends (Figure 7a). Bankfull channel depth, defined as the height of the levee crest above the channel base, is therefore also linear as the channel base and levee crest slopes are the same. Bankfull channel depth maintains values of 5–10 m but does not exhibit any slope breaks or change as the lower Rio Grande approaches the coast. However, the 2013 cross-sectional channel base data do not extend all the way to the mouth. For the covered reach the average bankfull channel depth is 7 m. Measurements of channel belt width made from the 2011 lidar data show an average width of 3,000 m and minimal narrowing toward the coast (Figure 8).

Details are in the caption following the image
Results of geomorphic characterization from the 2011 lidar and 1912 survey. Longitudinal profiles of various geomorphic parameters. All plots are shown in relation to distance upstream from the coast line in river kilometers, with flow to the right. (a) Elevation profiles of the 2011 levee crests from lidar and 2013 channel base from the TCEQ cross sections. Linear regressions for both levee crests and channel base are shown by the dashed lines, and regression parameters provided in inset. (b) Elevation profiles of the 1912 levee crests and 1912 channel base digitized from the 1912 IBWC survey. The backwater length scale, Lb, is shown by a black arrow and is estimated to be 70 river km. This coincides with the distance that the channel bed initially falls below MSL. Linear regressions for both levee crests and channel base are shown by the dashed lines, and regression parameters provided in inset. (c) The 1912 and 2011 bankfull depths as calculated by levee elevation minus channel bed elevations. (d) The 1912 and 2011 bankfull channel widths measured as the distance between both levee crests. (e) The 1912 bankfull width:depth ratio calculated from their respective measurements. The linear regression model and 95% confidence intervals are shown by the black line and shaded area, respectively. Regression parameters are provided as well as the test statistic and p values of the regression significance testing, which found no linear trend as a function of distance to the coast.
Details are in the caption following the image
Channel belt width along the modern Rio Grande. Channel belt width measurements as a function of distance upstream from the river mouth as defined in Figure 4a.

The 1912 Rio Grande exhibited a morphology nearly identical to the modern channel. The maximum levee elevation was 30 m at 250 river km and graded linearly to 0 m at the river mouth, with a slope of 1 × 10−4 (Figure 7b). The 1912 channel base had a maximum elevation of 20 m at 250 river km, and linearly descended to −8 m at the coast, also with a slope of 1 × 10−4 (Figure 7b). The 1912 bankfull channel depth was constant with respect to river km, with an average value of 7 m and a range between 5 and 10 m (Figure 7c). Bankfull channel width was likewise approximately constant in the downstream direction, with an average value of 125 m and a range between 80 and 200 m (Figure 7d). As a result, the ratio of bankfull channel width-to-depth was also similarly constant across the study area (Figure 7e). By fitting a linear regression to the channel width:depth ratio, we observe no trend as a function of distance to the coast, and statistical testing of the regression parameters finds no significant difference between the regression slope (10−6) and a slope of 0 (p = 0.06).

Lateral migration rates for channel bends of the Rio Grande were assessed for the time intervals of 1897–1912, 1912–1930, 1930–1956, and 1956–2011. These rates are plotted in Figure 9, along with trend lines representing a 5 km moving average for the values in each time interval. Descriptive statistics for each time interval as well as for the regions above and below the estimated backwater length Lb are provided in Table 2. Inspection of the historical maps and the 2011 lidar survey indicates that throughout the study area, migration occurred in the form of bend expansion, translation, and channel straightening via cutoffs with no apparent dominant mode of deformation. For all studied intervals of time, there were no identifiable spatial trends in bend migration rates across the 150 river km nearest the coast (Figures 9a9d). The postdam interval (Figure 9d) is remarkable for how small the rates of bend migration are compared to the predam era, with rates below 1 m/yr for most of the studied coastal reach. On average, the predam migration rates are 15–20 times greater than postdam rates (Table 2) Migration rates of the four time intervals were compared using ANOVA F testing and a significant difference found (p < 0.0001). Tukey MCP determined that migration rates in 1897–1912 and 1912–1930 are not significantly different (p = 0.32) [17.5 and 15.3 m/yr, respectively]. Significant differences between all other groups are found (p < 0.01), confirming the decrease in migration rates through time to the present (Figure 9 and Table 2). Additionally, we assess the difference in rates above and below the backwater transition within each time interval by performing unpaired t testing (Figure 10). We find that there are no significant differences above and below the characteristic backwater length scale for the first three intervals (p = 0.71, 0.54, 0.09). A significant difference is observed (p = 0.0006) for the final interval 1956–2011, with average rates of 1.2 m/yr within the backwater reach and 0.7 m/yr above it (Figure 10).

Details are in the caption following the image
Lateral migration rates of the Rio Grande for the last 100 years. Results of the lateral migration rate analysis across four time steps as a function of distance upstream. All rates are in meters per year. Individual measurements are shown as open circles, while black lines represent a 5 km moving average function. Black arrows show the position of the estimated backwater length, Lb. (a) 1897–1912 rates extend slightly further upstream than other times due to better data coverage in the 1912 map. (b) 1912–1930 rates. (c) 1930–1956 rates. (d) 1956–2011 rates.
Table 2. Migration Rate Summary Statistics
Avg. migration rate (m/yr) Max. migration rate (m/yr) Min. migration rate (m/yr)
1897–1912 19.1 ± 14.4 104.7 0.8
0–70 km 14.8 ± 11.8 71.5 0.8
70–200 km 15.7 ± 13.2 104.7 5.7
1912–1930 15.3 ± 8.9 41.9 2.7
0–70 km 14.7 ± 8.7 37.1 2.7
70–200 km 15.9 ± 9.3 41.9 3.3
1930–1956 11.6 ± 6.6 33.8 2.8
0–70 km 10.3 ± 6.7 33.8 2.8
70–200 km 12.5 ± 6.3 32.0 3.3
1956–2011 0.9 ± 0.8 4.7 0.03
0–70 km 1.2 ± 1.0 4.7 0.05
70–200 km 0.7 ± 0.6 4.1 0.03
  • Note. Average, minimum, and maximum migration rates for each of the four time intervals measured, as well as for below and above the characteristic backwater length scale, Lb = 70 km.
Details are in the caption following the image
Migration rate descriptive statistics. Box plot summaries of migration rates above and below Lb = 70 km for each of the four time intervals: 1897–1912, 1912–1930, 1930–1956, and 1956–2011. P values are provided for the statistical comparison between each group, with only 1956–2011 exhibiting a statistically significant difference in migration rates.

The Brownsville IBWC gage data provided a detailed history of flow in the Rio Grande (Figure 6a). While numerous dams were constructed on the upper Rio Grande in New Mexico since Elephant Butte in 1916, it was the emplacement of the Falcon Dam in 1951–1954 that acted to divide the lower Rio Grande record into its predam and postdam intervals (Small et al., 2009). Predam, the median discharge was 48.7 m3/s, while the maximum recorded discharge was 872 m3/s. P90 and P95 exceedance values were 245 and 439 m3/s, respectively. This contrasts with the postdam period where the median discharge was 5.6 m3/s, the maximum recorded discharge reached only 459 m3/s, and the P90 and P95 values were 63 and 152 m3/s (Figure 6b). The postdam record is one dominated by minimal flow, and few large events of the scale are seen in the predam era.

5 Discussion

A characteristic backwater length, Lb, was estimated using data from the average channel depths from the 2011 and 1912 surveys that set H at 7 m and S, the bed slope at 1 × 10−4. These values provided an estimated Lb of 70 km (Paola & Mohrig, 1996). This length is consistent with the position at which the average bed elevation for the two surveys goes below median sea level, which is an operational definition of the backwater position (e.g., Ganti et al., 2016). This length scale identifies the onset of the backwater zone, and we investigate changes in river morphology and kinematics upstream and downstream of this point (e.g., Chow, 1959; Lane, 1957; Maselli et al., 2018; Mason & Mohrig, 2018). A significant finding in this work are the remarkably linear profiles for levee crest elevation and channel bed elevation that extend all the way to the coastline (Figures 7a and 7b). This profile structure is recorded in both the pre- and post-Falcon Dam surveys of 1912 and 2011–2013. The lack of change in profile slope is matched by approximately constant mean values for bankfull channel depth and width between 150 and 0 river km (Figures 7c and 7d). In contrast to recent studies that have found systematic changes in channel bed slope, widths, and depths downstream of the backwater transition (Lamb et al., 2012; Maselli et al., 2018; Nittrouer et al., 2011; Smith, 2012), we could not identify a change in river form moving across and into the estimated backwater zone as indicated by the lack of any significant trend in width:depth ratio across the reach (Figure 7e). This indicates that the Rio Grande is maintaining a consistent channel geometry and cross-sectional area throughout the coastal reach irrespective of position relative to the estimated backwater.

The predam Rio Grande saw no coherent reduction in the lateral migration rates of channel bends across the transition into and through the backwater zone (Figure 9). Within each of these three time intervals, migration rates show a high degree of variability both above and below 70 river km but no reduction in the average rate (Figure 10). Averaged across the three intervals, migration rates were 15.6 m/yr upstream of 70 river km and 14.4 m/yr downstream of this point. Postdam (1956–2011) a significant difference was observed between 0.7 m/yr above 70 km and 1.2 m/yr below, but these absolute values are small and likely relate to ongoing incision of the Rio Grande into its banks and channel bed (Gonzalez, 2017). This is reinforced by the observations of relative stability in levee position and bankfull geometry from 1956–2011, so the measured centerline migration rates reflect both incision as well as exposure and vegetation of previously submerged point bars along bends as the number of bankfull floods drastically decreased rather than a change in transport behavior associated with the estimated backwater length. The Rio Grande from 1897–1956 migrated on average between 10% and 20% of its channel width per year, and these rates persisted to the coastline (Figure 9). This lack of a spatial trend stands in marked contrast to observations of systems such as the Rhine, Po, Trinity, Teshio, and Mississippi Rivers, which see a drastic reduction in the lateral migration of channel bends as they approach the coast (Fernandes et al., 2016; Hudson & Kesel, 2000; Ikeda et al., 1981; Maselli et al., 2018; Nittrouer et al., 2011; Smith, 2012). This reduction is linked to the changes in channel and point bar properties that occur within the backwater reach, specifically the narrowing and deepening of these channels and a loss of point bar definition. This narrowing and deepening is the result of channel adjustment to low- and moderate-flow conditions. Comparing the Rio Grande with the Trinity and Mississippi Rivers specifically and standardizing by each river's respective channel width, as well as each system's estimated backwater length, the difference between these rivers is quite apparent (Figure 11). Both the Trinity and Mississippi see a systematic reduction in migration rate that begins at approximately the backwater length, while the Rio Grande rate remains high to the coast. It is therefore not surprising that loop-cut offs can be observed along the entirety of the Rio Grande coastal reach, including three within 5 km of the river mouth (Figure 12). Overall, the mobility of Rio Grande channel bends is not correlated to the backwater length and thus is inconsistent with the proposed primary control of the backwater length on the channel kinematics of coastal, alluvial rivers (Fernandes et al., 2016).

Details are in the caption following the image
Migration rates of the Rio Grande, Mississippi, and Trinity rivers. Migration rates of the Rio Grande compared to other Gulf of Mexico rivers, standardized by channel width and estimated backwater length. Rates are plotted on a log scale to allow better interpretation of downstream trends. Rates of the Rio Grande in open circles, Trinity in orange squares, and Mississippi in blue triangles. The Rio Grande maintains high migration rates through the backwater zone, while both the Trinity and Mississippi see a reduction to effectively 0 (Fernandes et al., 2016).
Details are in the caption following the image
Close in of 2011 lidar of the mouth of the Rio Grande with historical channel centerlines. The 2011 lidar DTM showcasing the morphologic signature of historical migration. Within 5 km of the coast there was significant bend translation, expansion, and three loop cutoffs occurred within our time-lapse data set. From 1956–2011 the only significant changes observed have been associated with coastal erosion at the river mouth.

An interesting byproduct of this sustained bend migration to the coast is found in the resulting channel belt width (Figure 8). Numerous studies have found a downstream narrowing of fluvial channel belts both in modern systems as well as subsurface stratigraphy (e.g., Fernandes et al., 2016; Gouw & Berendsen, 2007; Törnqvist, 1993). Fernandes et al. (2016) proposed that this sharp reduction is due to the decrease in lateral migration rates as alluvial rivers transition to the backwater zone. As the Rio Grande does not appear to have any meaningful backwater related control on its lateral migration, we do not expect to find a similar relationship. While the width of the most recent Rio Grande channel belt does decrease downstream from 5,000 to 2,000 m, this is mostly due to very wide points found far upstream and is not in association with the backwater length scale of 70 km (Figure 8). The measured channel belt widths are also quite large compared to other systems when formational channel width is considered: The Rio Grande (average channel width = 125 m) maintains a dimensionless width (channel belt width/channel width) of 15–20 for the entirety of the study area compared to the Mississippi (average channel width = 1,200 m) which has a maximum channel belt width/channel width of 20 and transitions to near one at the coast (Fernandes et al., 2016; Martin et al., 2018).

The impact of dams on rivers has long been studied to understand potential reductions in sediment delivery to the coast, as well as changes to river dynamics upstream (e.g., Mackin, 1948; Syvitski et al., 2005). Analysis of sand-bedded rivers suggests that even when sediment is impounded by a dam, the downstream flow can effectively mine the river bed and substratum to reestablish its bed material load (Galay, 1983; Nittrouer & Viparelli, 2014; Schmidt & Wilcock, 2008; Smith & Mohrig, 2017). The lateral mobility of rivers is the product of the coevolution of point bars and cut banks along channel bends (Hickin & Nanson, 1975; Hooke & Thorne, 1997; Howard & Knutson, 1984). In systems where bed material load is reestablished by mining the channel bottom, channel migration is reestablished when this load is sufficient to maintain and grow point bars (Grams et al., 2007; Smith & Mohrig, 2017). This is not observed on the Rio Grande. For this system the emplacement of Falcon Dam has resulted in a near complete cessation of lateral migration. A key difference between systems is found in how damming affected the distributions of water discharge. On the Trinity River, Wellmeyer et al. (2005) found that damming did not meaningfully alter the mean discharge nor did it significantly change the shape of the hydrograph. This stands in contrast to the Rio Grande, where not only was median flow reduced but more importantly the number of floods significantly decreased (Figure 4, Small et al., 2009). Prior to dam emplacement, the Rio Grande was subject to periodic and catastrophic floods. The available hydrograph record from 1934–1954 shows mean daily discharge at Brownsville varied between 0 and 872 m3/s (Figure 6). However, only the mean was recorded and peak discharged could reach as high as 5,000 m3/s for significant flood events (IBWC, 1932). Since the construction of Falcon Dam, median discharge dropped from ~49 to ~5 m3/s, while maximum recorded discharge fell from 872 to 459 m3/s (Figure 6). Predam the river was in flood for over ~400 days from 1934–1954, or on average 21 days a year, while postdam the number of flood days was ~38 from 1954–2011 or on average less than 1 day a year (Small et al., 2009; Figure 6). This massive reduction has led to extreme ecological and hydrologic effects in the lower river and across the delta (Small et al., 2009; TCEQ, 2012), and as observed in this study a similarly drastic change to river-channel kinematics in the coastal reach (Figure 9d).

The intermittent hydrograph of the predam Rio Grande provides a possible explanation for the lack of channel adjustment as the Rio Grande approaches the coast (Figures 7 and 9). Many of the proposed linkages between the hydrodynamic backwater and fluvial geomorphology are connected to specific spatiotemporal variations in sediment transport within the coastal zone (e.g., Lamb et al., 2012). In particular, it is assumed that there exists a persistent transport of water and sediment within the uniform-flow reach (e.g., Chatanantavet et al., 2012; Lamb et al., 2012; Nittrouer et al., 2011). This leads to channel adjustments in the backwater reach that are necessary to accommodate the flow of water and sediment during both low and moderate flows. Specifically, increases in cross-sectional channel area that are associated with a change in channel geometry with relatively large increases in channel depth and relatively small decreases in channel width to ensure that water continuously flows to the coast (Chatanantavet et al., 2012; Nittrouer et al., 2012). This change in channel geometry (i.e., the channel deepening and narrowing) is observed to occur within the backwater zone and is correlated with a reduction in point bar volume, slope, and fraction of the channel occupied to the point where scroll bars defining lateral migration of point bars are no longer observed on the flood plain (Mason & Mohrig, 2018). Transport in the lower Rio Grande is quite different than the above scenario. Baseflows are associated with transport that appears not to require an overdeepening of the Rio Grande channel to handle the transport of water and sediment in its lower 70 river km. In fact, historical predam records document springtime conditions where the bed of the coastal river was completely dry in the normal flow reach (IBWC, 1932). We hypothesize that based on its intermittent hydrograph properties, the Rio Grande channel is adjusted solely to peak flows, and during these times the backwater transition is essentially pushed to the coastline. As a result, no systematic changes to river geomorphology are observed throughout the coastal zone and channel bends migrated to the point of cutoff within only a few kilometers of the shore (Figure 12). Historically, during baseflow conditions flows from upstream have been sufficiently small as to leave the coastal reach of the Rio Grande essentially a standing body of water (TCEQ, 2012).

6 Conclusions

This study captures the long-term kinematics and geomorphology of the coastal reach of a significant continental-scale river, the Rio Grande. Using a novel data set of historical surveys and modern techniques, we show that the morphodynamics of intermittent rivers, even large ones, may differ significantly from more commonly studied systems. We also show that dam construction in the 1950s and associated decreases in flooding led to a near complete cessation of lateral migration of the river, transforming what was previously a highly active channel to a static feature. The lower Rio Grande shows no systematic changes in channel geometry and lateral migration behavior as a function of distance to the coastline in sharp contrast to many other previously studied coastal river systems. No overdeepening, narrowing, or reduction in bend migration is observed within the estimated backwater zone, a length scale that has become increasingly associated with numerous kinematic and morphologic effects. We hypothesize this is due to the large degree of discharge intermittency of the Rio Grande, where baseflow is often minimal or nonexistent and flooding severe. With little water moving through the system during low-flow conditions there is no channel adjustment required to transport this water to the coastline. Rather, the channel remains adjusted to high-flow conditions at all times. The Rio Grande offers an important counterpoint to the better studied perennial systems that show a systematic change in channel geometry toward the coast and a corresponding decrease in migration rate.


The authors would like to thank Paola Passalacqua and Daniel Stockli, as well as all the members of the Mohrig research team. We thank reviewer Alvise Finotello, two anonymous reviewers, Editor Amy East, and Associate Editor Evan Goldstein for the helpful comments, criticisms, and insights that greatly improved the manuscript. We would like to thank Jose Gonzalez of the Hidalgo County Drainage District for providing the TCEQ topographic cross sections. Support for author JMS was provided by a Bureau of Ocean Energy Management cooperative agreement (M16AC00020). TAG gratefully acknowledges financial support for this work through the National Center for Earth-Surface Dynamics 2 (NSF Grant EAR-1246761).

    Data Availability Statement

    The 2011 lidar data are freely available from Texas TNRIS (data.tnris.org). Hydrograph data for Brownsville are available from the International Boundary and Water Commission (www.ibwc.gov). The 1912 historical map is located in the University of Texas at Austin Perry-Castenada library as part of the Texas historical maps collection (http://legacy.lib.utexas.edu/maps/historical/history_texas.html). The 1930 and 1956 historical topographic maps are available from the U.S. Geological Survey Earth Explorer and the topoView downloader (https://ngmdb.usgs.gov/topoview/).