Geomorphology 130 (2011) 17–28
Contents lists available at ScienceDirect
Geomorphology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / g e o m o r p h
Universal and local controls of avulsions in southeast Texas Rivers Jonathan D. Phillips Tobacco Road Research Team, Department of Geography, University of Kentucky, Lexington, KY 40506-0027, USA
a r t i c l e
i n f o
Article history: Accepted 4 October 2010 Available online 7 October 2010 Keywords: Conﬁgurational complexity Connectance entropy Avulsion Southeast Texas Universal Local
a b s t r a c t This study addresses the relative importance of universal and local factors in river avulsions, from the perspective of the conﬁgurational complexity of the ﬂuvial system as it relates to the potential for avulsions. Because local factors cannot be addressed without a speciﬁc context, several rivers of the southeast Texas coastal plain (Brazos, Navasota, Trinity, Neches, and Sabine Rivers) are studied based on ﬁeld observations and surveys, digital elevation models, and geographical information system (GIS) analysis. Avulsions are inﬂuenced by a combination of universal factors relevant to any alluvial river, and local factors at least partly contingent on the environmental setting and history of the study rivers. The universal controls are factors that create conditions under which avulsions can occur: channel aggradation, banktop discharge, and cross-valley slope advantages. A rate of channel aggradation greater than that of ﬂoodplain accretion is an important setup factor, but superelevation (channel bed elevation greater than or equal to ﬂoodplain elevation behind the natural levee) is not required. Slope advantages are necessary, but not sufﬁcient, for avulsions to occur. Local factors include abandoned channels on the ﬂoodplain, and basins or depressions associated with higherdischarge Pleistocene conditions. The local controls also include deﬂection factors that divert ﬂow from the main channel during high discharges, and levee weaknesses. Relationships among the universal and local factors were represented as directed graphs, and the conﬁgurational complexity determined using connectance entropy. Results show that the relative importance of universal and local factors varies with the scale or scope of analysis. In the broadest context, universal factors contribute more than 75% of the connectance entropy, signifying the importance of identifying the key setup factors. Within a given aggradation or valley-ﬁlling context, however, local factors account for more than 80% of the connectance entropy, indicating that knowledge of the trigger factors and local ﬂoodplain morphology is the key to understanding and predicting avulsions. This study suggests a three-stage process for predicting avulsions. First, the universal factors can be used to identify reaches with the potential to avulse. Second, superelevation (a sufﬁcient but not necessary condition) can be used to identify speciﬁc locations where avulsions are imminent, and (as a necessary but not sufﬁcient condition) the absence of cross-valley slope advantages can be used to identify locations where avulsions cannot occur. The ﬁnal stage—predicting avulsions at other locations within systems or reaches where the setup factors are present—requires speciﬁc consideration of local factors. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Abrupt channel changes in rivers—avulsions— are generally understood via a setup-and-trigger framework. Setup factors represent necessary conditions for avulsions to occur—channel aggradation and cross-valley slope advantages. These allow for the possibility of avulsion during ﬂows sufﬁcient to breach natural levees, while triggers are speciﬁc events that divert ﬂow from the main channel or focus levee breaching at local points of weakness (Nanson and Knighton, 1996; Jones and Schumm, 1999; Makaske, 2001; Slingerland and Smith, 2004). The latter may include channel changes or obstructions that displace water from the channel or deﬂect ﬂow toward banks, and local gaps or weaknesses in natural levees.
E-mail address: [email protected]
0169-555X/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.geomorph.2010.10.001
The setup factors are readily accommodated in models (e.g. Mackey and Bridge, 1995; Slingerland and Smith, 1998, 2004; Tornqvist and Bridge, 2002; Jerolmack and Paola, 2007; Stouthamer and Berendsen, 2007). However, those models require empirical parameterization or stochastic forcings to account (implicitly or explicitly) for the trigger factors, that by their nature are highly spatially and historically contingent. Likewise, aggradation and crossvalley slopes are reﬂected as generally applicable principles in conceptual models of avulsion and the formation of river anabranches (e.g. Nanson and Knighton, 1996; Aslan and Blum, 1999; Jones and Schumm, 1999; Makaske, 2001; Slingerland and Smith, 2004). However, work on a number of river systems shows that while the setup factors are necessary conditions for avulsions to occur, they are not sufﬁcient, and do not allow for prediction of where and when avulsions occur (e.g. Jones and Schumm, 1999; Makaske et al., 2002; Aslan et al., 2005; Taha and Anderson, 2008; Phillips, 2009).
J.D. Phillips / Geomorphology 130 (2011) 17–28
The problem of river avulsions is a good example of the general principle that geomorphic phenomena tend to be governed or inﬂuenced by two sets of controls. Universal or global controls are laws, principles, and relationships that are widely and generally applicable. Local controls are speciﬁc to time and place (see Phillips, 2007). The controls of aggradation, slope advantages, and overbank ﬂows apply to any river and any avulsion. The local factors, by contrast, are geographically and historically contingent, and are not necessarily widely applicable. Ice jams, for example, are potential ﬂow diversions that may be important in some rivers and irrelevant in others. The hippopotamus pathways linked to avulsions in the Okavango River, Botswana by McCarthy et al. (1992) are not only restricted to the geographic range of the hippo, but locally contingent on hippopotamus habits and habitats. In rivers of southeast Texas, Phillips (2009) found that valley aggradation state was a ﬁrst-order control of differences in avulsion in regimes, but that antecedent, inherited morphology was an important second-order control, and local, contingent trigger factors a thirdorder control. Thus ﬁve rivers with the same general set of environmental controls showed ﬁve different avulsion regimes. The universal controls, such as channel and valley aggradation rates and frequency of overbank ﬂow, can be used to identify rivers or reaches where avulsions are more or less likely. The availability of potential cross-valley slope advantages or occupiable paleochannels can further reﬁne these probabilities, as well as deﬁne speciﬁc locations where avulsions are possible or not. Given that slope advantages are far more common than avulsions (e.g., Aslan et al., 2005), however, attempts to predict or interpret speciﬁc avulsion locations may require addressing the local factors. Because models based on universal factors are (by deﬁnition) more readily generalizable, and local factors (again, by deﬁnition) require more situationspeciﬁc knowledge, data, or expertise, the relative importance of universal and local factors is critical to the understanding, prediction, and management of channel shifts. This study addresses the relative importance of universal and local factors in river avulsions, from the perspective of the conﬁgurational complexity of the ﬂuvial system as it relates to the potential for avulsions. Because local factors cannot be addressed without a speciﬁc context, the southeast Texas coastal plain rivers studied in earlier work by the author are used. 1.1. Local and global Three general principles appear to apply to essentially all avulsions and ﬂuvial systems prone to avulsions (c.f. Jones and Schumm, 1999; Slingerland and Smith, 2004). First, avulsions are overwhelmingly associated with aggrading valleys, as evidenced by (among other things) the strong association of avulsions with deltas and with anastamosing channel patterns. Second, avulsions begin with levee breaches or crevasses, and thus are triggered during overbank ﬂows. Third, a successful avulsion (a crevasse resulting in a channel shift or the formation of a persistent distributary or anabranch) requires a slope advantage. That is, the new channel must have a steeper gradient than the original channel. These three factors—aggradation, likelihood of overbank ﬂow, and valley slope—can thus be considered universal in the sense that they are applicable to the avulsion regime of any ﬂuvial system, anywhere, anytime. These generally applicable universal factors may be associated with external, allogenic forcings and controls such as climate change, regional tectonics, and land-use change. They may also be associated with autogenic factors or events such as sediment inputs from surrounding hillslopes, inherited ﬂoodplain and valley topography, or morphological changes within channels and ﬂoodplains. Similarly, local trigger factors may be associated with either allogenic (e.g., climate-driven changes in riparian vegetation) or autogenic factors. Thus the universal or global vs. local distinction should be understood
in terms of their relevance in understanding or evaluating avulsions in every ﬂuvial system vs. only some ﬂuvial systems, rather than as exogenous or allogenic vs. endogenous or autogenic. Universal as used here is comparable to the term global as the latter is often used in systems theory. However, at least four different conceptions of local:global are potentially relevant to geomorphology. These are related and overlapping, but not equivalent. First, global and local may be deﬁned sensu stricto: global as planetary and local as some signiﬁcantly smaller sub-planetary level. Global and local may also be used more generally to refer to broad spatial (and temporal) resolutions or scales vs. narrower, more restricted scales, implying that the distinction is related to position in a scale hierarchy. The environmental modeling and geomorphology literature widely recognizes the simultaneous and interrelated effects of processes and controls at different scales or levels. This can sometimes be addressed via formal nested hierarchies and hierarchy theory (e.g. Haigh, 1987; DeBoer, 1992; Hoosbeek and Bryant, 1992). The contrast between global laws and local, contingent factors does not equate with formal scale hierarchies, however, because global or contingent factors may operate at a range of scales and manifest themselves in either top-down or bottom-up fashion. A third global:local conceptualization is that of exogenous or extrinsic vs. endogenous or intrinsic. This may also be expressed in terms of allogenic and autogenic processes (see e.g. Stouthamer and Berendsen, 2007). This can be a most useful dichotomy, for example in the distinction between intrinsic and extrinsic thresholds. However, both general laws and contingent controls may be either intrinsic or extrinsic to geomorphic systems, as discussed above. Finally, there is universal or global vs. local as discussed in this paper, where universal (global) factors are those that are independent of time or place. Global laws, relationships, or generalizations operate everywhere and always. Local factors are not independent of time and space. This is not to say that local factors are somehow immune to biophysical laws and principles, but that they have to be applied on a case-by-base basis. For example, riparian vegetation is subject to edaphic and physiological controls that may be described with global principles, and the hydraulic effects of vegetation on ﬂow are likewise reducible to physical laws. However, the manifestation of vegetation/ geomorphic interactions is highly local, and there is no generalization by which a riparian vegetation factor or equation or relation can be uniformly applied to any stream corridor. Similarly, the history and sequence of events such as coastal storms or ﬂoods is important in conditioning the geomorphic responses of coasts and streams, but these factors are unavoidably local and idiosyncratic. This concept of local and global is consistent with use of the terms in mathematics and symbolic logic, where global properties or relationships apply to an entire system or space, and local to a restricted portion thereof. The importance of local factors in this sense is increasingly being recognized in geomorphology (Lane and Richards, 1997; Harrison, 1999, 2001; Phillips, 2007; Sauchyn, 2001; Vandenberghe, 2002). However, “universal” is used in this paper to avoid confusion with the other notions of global mentioned above.
1.2. Conﬁgurational complexity This study is concerned with the complexity (as measured by entropy) of the structure of the interrelationships between the factors or components controlling or inﬂuencing avulsions. This reﬂects the complexity of a representation of the components and relationships as a network graph, box-and-arrow model, or interaction matrix. This form of complexity is sometimes referred to as structural complexity, but in geomorphology and ecology the latter term is often used to describe the physical or spatial heterogeneity of landforms, landscapes, or habitats. Thus the term conﬁgurational complexity is employed, broadly analagous to the conﬁguration complexity concept
J.D. Phillips / Geomorphology 130 (2011) 17–28
used in computing and engineering to describe the complexity of interlinked technological components. Conﬁgurational complexity of geomorphic systems is inﬂuenced by the number of components, the number of links between them, and the (un)evenness of linkages among components. As previous applications in geomorphology and related ﬁelds have shown, entropy is an appropriate measure of the structural or conﬁgurational complexity of systems of this type (e.g. Phillips, 2002; Phillips and Walls, 2004; Saldana and Ibanez, 2007). Considerable debate exists in environmental modeling—climate, hydrology, and ecosytem models as well as geomorphological—about the relative merits of different general approaches to modeling complex environmental systems. One approach is to develop ever larger and more complex models in an attempt to capture ever more of the real-world complexity. An alternative approach is toward simpler, leaner models that seek to capture essential phenomena. In a hydrological modeling context, Sivakumar (2008) contrasts these two approaches as attempts to “model everything” versus efforts to “capture essential behaviors.” For discussions of these issues in geomorphology, see, e.g. Phillips (1992), Werner (1995, 1999), Brasington and Richards (2007), and Murray (2007). In hydrological modeling the “dominant processes concept” has emerged, the essential argument of which is that there are too many potentially relevant hydrological processes to feasibly or efﬁciently include them all in a single model. However, in any given watershed a handful of processes dominate the hydrological response, and an effective model may be developed based on those (Sivakumar, 2004, 2008). The dominant processes concept can be generalized to a “dominant controls concept” (DCC) in geomorphology. The DCC implies that, while there may exist a very large number of factors and processes that can inﬂuence avulsions, in any given ﬂuvial system some will be irrelevant and others of comparatively negligible inﬂuence, leaving a few dominant controls to deal with. With respect to avulsions, three paths are evident. One is to focus on universal controls, either accepting the variation associated with local factors or folding it into empirical parameters. This can be effective when the concern is primarily or solely with the universal factors. A second strategy is a “model everything” effort. The problem is such a model must either be restricted in its applicability, or must include a wide range of local factors—ice jams, woody debris, animal trails, uprooted trees, ﬂood basins, inherited morphology, landslide dams, etc.—which are locally but not widely relevant. The third approach is based on the DCC, in which models have universal components and links that may be widely or generally applied, and local components and relationships that are problem-speciﬁc. The conﬁgurational complexity of the ﬁrst and third approaches is inherently lower than that of a model everything strategy. In choosing between a global-only or DCC approach, a connectance entropy analysis such as that used here is helpful because it allows the relative importance of local and universal factors in conﬁgurational complexity to be determined. The entropy analysis also provides a means for determining the scale of analysis or scope of problem at which local or universal factors may assume predominance. 1.3. Local and universal entropy Several entropy concepts are used in geomorphology, deriving from thermodynamics, information theory, and nonlinear dynamical systems theory. The latter two are related in that the change in Shannon entropy (from information theory) over time in an evolving dynamical system is equal to Kolomogorov entropy, which measures the divergence or convergence in state space. Here the concern is with entropy as an indication of the complexity of a geomorphic system rather than thermodynamic entropy. Geomorphic models such as those discussed below may be generalized by considering the fundamental system components
and their qualitative relationships to each other. For example, a rapid rate of channel aggradation relative to ﬂoodplain accretion tends to promote avulsions. While the speciﬁc quantitative nature of these relationships varies from one ﬂuvial system to the next, the qualitative relationships are more generally applicable. The system or phenomenon of interest—such as avulsions—can be represented as an interaction matrix or box-and-arrow diagram. Such a system has entropy associated with its connectance or web of interaction. This in turn may be amenable to decomposition into its local and universal components. The connectance entropy is given by C = −∑½ðxi = xÞ logðxi = xÞ
where x is the total number of links (arrows or nonzero matrix entries) in the system, and xi is the number of links originating from the ith component. If the system includes both universal and local factors, by the principle of entropy decomposition the total entropy is equal to the sum of the entropies associated with the universal and local components: C ðTÞ = C ðUÞ + C ðLÞ:
The information in the directed graph is given by I = lnðxÞ–C
This formalism is discussed in more detail, and examples given, by Phillips (2002). 2. Avulsions The geomorphology and sedimentology of avulsions has been reviewed elsewhere, both in general (Makaske, 2001; Slingerland and Smith, 2004), and with respect to the Texas coastal plain (Aslan and Blum, 1999; Blum and Aslan, 2006; Phillips, 2009). General agreement exists that avulsions are associated with aggrading systems, that a successful avulsion (i.e., a crevasse channel that persists and captures all or a signiﬁcant portion of the river ﬂow) requires a slope advantage, and that avulsions occur during ﬂoods or near-ﬂood ﬂows. However, because of the importance of local inﬂuences on those controls, and the inherently local trigger factors, considerable uncertainty exists with respect to the causes of speciﬁc avulsions (Jones and Schumm, 1999; Makaske, 2001; Slingerland and Smith, 2004). In Mackey and Bridge's (1995) model of ﬂoodplain development, the probability of an avulsion is modeled as eQ
pðavulsionÞ = ðQ =Q a Þ ðkScv =Sdv Þ
where Q is discharge and Q a a critical threshold discharge for crevasses or avulsions to occur; Scv, Sdv are the cross- and downvalley slopes, and k, eQ, eS are empirical or “tuning” parameters. The latter are necessary in part to keep the model from predicting an unrealistically high number of avulsions, a testament to the necessary but not sufﬁcient nature of the discharge and slope controls. The Mackey and Bridge (1995) framework was incorporated into Stouthamer and Berendsen's (2007) model of avulsions for the Rhine–Meuse delta. Jerolmack and Paola (2007) developed a cellular automata model of avulsions in aggrading channels. In each model grid cell two elevations are tracked—the levee top and channel bottom. The channel aggrades in place until a superelevation threshold is met, triggering an avulsion. This threshold (not to be confused with the superelevation of the water surface sometimes associated with helical ﬂow at the outside of bends) occurs when the levee top is one channel-depth above the ﬂoodplain (and thus the elevation of the
J.D. Phillips / Geomorphology 130 (2011) 17–28
channel bed is ≥ﬂoodplain elevation). The new channel then follows the path of steepest descent. This model is based on the analysis of conditions for branching in depositional rivers by Jerolmack and Mohrig (2007), who emphasize the importance of vertical channel accretion rates vs. lateral migration rates in favoring avulsions. While gradient advantages appear to be necessary for avulsions to occur, they are not sufﬁcient. Aslan et al. (2005) found that though potential gradient advantages on the Mississippi River ﬂoodplain are common, avulsions are relatively rare. Based on this, Aslan et al. (2005) suggested that substrate composition and ﬂoodplain channel distributions are key controls of avulsions, consistent with Hudson and Kesel (2000).
2.1. Southeast Texas avulsions The study area (Fig. 1) includes the Brazos River from state highway (SH) 21 near Bryan to the Gulf of Mexico; the Navasota River from the Lake Limestone Dam to the Brazos River conﬂuence; the Trinity River from (Lake) Livingston Dam to Trinity Bay; the Neches River from the SH 21 crossing to Beaumont; and the Sabine River from Toledo Bend Dam to the Sabine Lake estuary. The river distances are 469, 185, 175, 340, and 214 km, respectively. The climate is humid subtropical, and the topography ranges from virtually ﬂat in the coastal marshes to gently rolling. The Quaternary geologic framework and ﬂuvial and sea level history are described in some detail elsewhere (Blum et al., 1995; Morton et al., 1996; Blum and Aslan, 2006; Phillips, 2008, 2009; Phillips and Slattery, 2008). Generally, the valleys are cut into early to mid Pleistocene marine, coastal, and alluvial sediments. Within the valleys there occur up to three levels of late Pleistocene terraces referred to as the “Deweyville” terraces or formations (Blum et al., 1995).
Discharge regimes for the lowermost gaging stations with longterm discharge records are summarized in Table 1. The ﬂood discharge was determined from minimum ﬂood stages identiﬁed by the National Weather Service for the respective stations. These are generally associated with the threshold between bankfull and overbank ﬂow. The high probabilities for mean daily ﬂows exceeding banktop levels in Table 1 reﬂects the tendency at some stations for ﬂows to remain at or above ﬂood stage for several days or weeks. The lowermost Brazos River is more incised and has a larger channel capacity relative to discharge than the other study rivers, accounting for the less frequent overbank ﬂow—though note that even there, mean daily ﬂows at or above ﬂood stage occur, on average, more than three days per year. The high probability at the Navasota station reﬂects rapid aggradation and ﬂow displacement in the channel (Phillips, 2009). At the Trinity (Liberty) and Sabine River stations, the morphology of the river valley is such that overbank ﬂows are common and channel–ﬂoodplain connectivity is high, with conditions at these stations more or less representative of the valley reaches in which they occur (Phillips and Slattery, 2007, 2008; Phillips, 2008). The Neches station underestimates the longer-term ﬂood frequency due to the presence of a large ﬂood control reservoir upstream. While dams and reservoirs also exist on the lower Navasota, Trinity, and Sabine Rivers, these are water supply or hydropower impoundments and do not greatly inﬂuence ﬂood regimes downstream. According to Aslan and Blum (1999), Texas Gulf of Mexico coastal plain rivers undergo two distinct styles of avulsion—reoccupation of former channels, and diversion into ﬂood basins. The Nueces and Trinity Rivers are examples that represent early stages of sedimentary inﬁlling in response to Holocene sea level rise, and avulse by reoccupying late Pleistocene channels cut during falling and lowerstand sea levels. The Colorado River represents a later stage of inﬁlling where most of the accommodation space is ﬁlled. Avulsions here
Fig. 1. Study area.
J.D. Phillips / Geomorphology 130 (2011) 17–28
Table 1 Flow characteristics of lowermost gaging stations on the study rivers. Station name
Harmonic mean ﬂow cms or m3 sec-1a
Flood discharge cms or m3 sec-1b
Probability of ﬂood Qc
Brazos River @ Rosharon Navasota River @ Easterly Trinity River @ Romayord Trinity River @ Libertyd Neches River @ Evadale Sabine River @ Ruliff
08116650 08110500 08066500 08067000 08041000 08030550
45.930 0.710 41.739
1812 85 2407 909 1189 377
39.45 119.72 57.67
0.011 0.033 0.001 0.153 0.010 0.222
From Asquith and Heitmuller, 2008. Based on National Weather Service ﬂood stages. c Probability that meas daily ﬂow equals or exceeds ﬂood discharge. d Liberty station is representative of lowermost Trinity River, but tidal backwater effects result in zero or negative discharges which make mean ﬂow calculations problematic. Romayor is the next upstream gaging station. b
occur as repeated diversions into ﬂoodplain depressions (Aslan and Blum, 1999). The Sabine/Neches and Trinity incised valleys are unﬁlled, in the early stages of ﬁlling as described above, and avulsion has so far taken place by reoccupation of Pleistocene falling stage to lowstand channel belts. By contrast, the Brazos and Colorado valleys are ﬁlled, and have progressed through the entire sequence described above (Blum and Aslan, 2006). Phillips (2009) examined avulsions in ﬁve southeast Texas Rivers, and afﬁrmed the state of valley ﬁlling as a ﬁrst-order control. However, three rivers (Trinity, Neches, and Sabine) with the same state of valley ﬁlling had different avulsion regimes, attributed to local, contingent variations in antecedent morphology associated with Quaternary ﬂuvial and sea level changes. The lower Brazos River is characterized by an inﬁlled incised valley similar to that of the Colorado River, and as predicted by Aslan and Blum (1999), avulsions are characterized by repeated diversions into ﬂoodplain depressions (Sylvia and Galloway, 2006; Phillips, 2009). Recent movement along listric faults was linked to a Brazos avulsion cluster by Taha and Anderson (2008). The Brazos avulsion regime is also characterized by relocation avulsions (as opposed to the formation of anabranches, or local avulsions sensu Jerolmack and Paola, 2007). The Brazos has experienced no avulsions in the past ca. 300 years (Waters and Nordt, 1995), and the abandoned channels are fully or partially inﬁlled except where occupied by tributaries (Phillips, 2009). The lower Navasota River is an anabranching system over most of its length (except where it occupies a Brazos paleochannel), and is characterized by an active avulsion regime, including historic and recent channel shifts. While some relocation avulsions have occurred, most are local-type avulsions, resulting in the formation of anabranches. The abandoned or sub-dominant channels in the lower Navasota valley exist in all stages of development, from active anabranches, to semi-active high-ﬂow channels, to billabongs, and inﬁlled channels. This avulsion regime is attributed to active valley ﬁlling in a narrow valley with limited accommodation space (Phillips, 2009). The Trinity, Neches, and Sabine Rivers are part of the same ﬂuvial system, and their ancestral channels converge on the present continental shelf (Anderson et al., 2008; Milliken et al., 2008). All have experienced historic avulsions, though the Neches has the most active avulsion regime, possibly attributable to reoccupation of an anastamosed channel belt in the lower river basin (Phillips, 2009). Like the Navasota, the Neches is dominated by anastamosis-type avulsions, and has abandoned and sub-channels in all stages of inﬁlling. Avulsions in the lower Trinity River are controlled primarily by large paleomeanders, paleochannels, and depressions associated with a higher-discharge ancestral Pleistocene river system. Similar features exist in the Sabine, Neches, and (to a lesser extent) Brazos Rivers, but are not as inﬂuential with respect to avulsions. Sabine River avulsions are inﬂuenced by a variety of neotectonic, deltaic, and anthropic processes (Phillips, 2008, 2009), and both the Sabine and Neches include high-ﬂow and paleochannels indicating an older anabranching pattern (Phillips, 2009; cf. Leigh, et al. 2004).
3. Methods The general approach of this study consists of ﬁve general steps. First, the applicability of the universal factors (aggradation, overbank ﬂows, and cross-valley slope) and their manifestations in the study region was assessed. Second, the local controls relevant to the study region were identiﬁed. Third, the qualitative direct links between these factors, controls or components were identiﬁed—that is, whether an increase (decrease) in any component directly results in an increase (decrease) or no effect on the other components. This forms the basis for representation of the avulsion system as a signed digraph or interaction matrix. The ﬁrst three steps were based on a combination of GIS-based analyses and ﬁeld investigations, as described below, and a review of the literature on the geomorphology, hydrology, and Quaternary history of ﬂuvial systems in southeast Texas. Step four is an analysis of the connectance entropy of the system to determine the relative importance of the universal and local factors. The ﬁfth step is a rudimentary test of a model based on general, universal factors with respect to applicability in the study area. This is the channel bed superelevation model (SM) of Jerolmack and Mohrig (2007; see also Jerolmack and Paola, 2007), described earlier. 3.1. Avulsion data Digital orthographic aerial photographs (digital ortho quarter quads or DOQQs), satellite imagery, digital elevation models and 1:24,000 topographic maps were used to identify abandoned channels or anabranches representing potential avulsions on the lower Brazos, Navasota, Trinity, Neches, and Sabine Rivers. These data were obtained from the Texas Natural Resources Information Service and examined using geographic information system (GIS) tools (ARCGIS™, and RiverTools™). Field examinations were conducted to conﬁrm the identiﬁed features as former river channels, and to assess their state of inﬁlling. The avulsions were classed as relocations or anastamosis-type avulsions depending on whether a downstream connection with the modern river could be discerned. Some relocations were further classiﬁed as distributary-type avulsions. The channel shifts were further identiﬁed as local (occurring entirely within Holocene meander belts) or non-local (involving reoccupation of Pleistocene paleochannels or ﬂow into or through Pleistocene ﬂoodplain depressions). Finally, abandoned channels were categorized as active (conveying ﬂow at all or nearly all times), semi-active (regularly conveying ﬂow during high-discharge events), billabongs (ﬂooded segments with no connection to the active channel except during ﬂoods), tributary-occupied, and inﬁlled. The methods and resulting inventory are described in detail in Phillips (2009). The extent to which the superelevation model of Jerolmack and Mohrig (2007) is applicable can be addressed by comparing the elevations of the active channel, abandoned channel, and ﬂoodplain surface (Hc, Hp, Hf, respectively). If an avulsion occurs due or
J.D. Phillips / Geomorphology 130 (2011) 17–28
subsequent to superelevation of the original channel relative to the ﬂoodplain, then immediately after completion of the avulsion Hc bHf bHp : An avulsion not associated with channel superelevation would result in Hc bHp bHf or Hc ≈Hp bHf : However, these outcomes could also be produced after a superelevation-driven avulsion due to subsequent incision of the original channel if it is occupied by an active or high-ﬂow anabranch or a tributary. Assuming that Hf ≥ (Hc, Hp) and Hc ≤ (Hp, Hf), one other possibility exists:
3.2. Model construction
Hc bHp ≈Hf : In this case an inﬁlled paleochannel would rule out the SM, as the original channel bed would have been lower than the ﬂoodplain elevation. An SM-type avulsion resulting in this relationship would have some former levees or relict levees in the form or ridges above the general level of the ﬂoodplain surface. Therefore the likelihood of the applicability of the SM can be assessed based on the relative elevations of the new and old channels and the ﬂoodplain surface below levees, and the state or condition of the abandoned channel, as summarized in Table 2. Thirty past avulsion sites were selected where ﬁeld access could be obtained to either directly survey the relative elevations of channels, ﬂoodplains, and abandoned channels, or where the relative elevations indicated from DEM-based measurements could be ﬁeld-veriﬁed. In the former cases comparisons of Hc, Hp, Hf, were based on ﬁeld surveys using a laser level and prism in concert with a weighted line or sonar depth ﬁnder deployed from a boat or canoe. The laser level was used to survey topography using break-of-slope methods, and water surface elevations at the water's edge. Maximum depths below the water surface elevation were determined using a weighted line where possible. If ﬂow velocities precluded this, a sonar depth ﬁnder was used. The entire cross-section was probed or sampled to determine the thalweg or point of maximum depth. Cross-sections obviously inﬂuenced by bars or pools were avoided. However, particularly in sub-channels and at high ﬂows, this was not always evident. Thus water depths may not be representative of the reaches they occur in. However, as the key assessment is the relative elevation of the ﬂoodplain and paleochannels to channel bottom elevation, this uncertainty is of limited concern. In the latter cases relative elevations indicated from DEM-based measurements were assessed by identifying ﬂoodplain and paleochannel surfaces in the ﬁeld, and comparing their elevations using the laser level. In addition, topographic transects from DEMs were used to
Table 2 Interpretation of applicability of the SM to avulsions (see text for details). Relative elevations
Hc b Hf b Hp Hc b Hp b Hf
SM Non-SM if old channel is inﬁlled or slough Unlikely SM if old channel is semi-active or occupied by aggrading tributary Possible SM if old channel is active or tributary-occupied Non-SM if old channel is inﬁlled or slough Unlikely SM if old channel is semi-active or occupied by aggrading tributary Possible SM if old channel is active or tributary-occupied Non-SM if old channel inﬁlled or signiﬁcantly aggraded Probable SM otherwise.
Hc ≈ Hp b Hf
Hc b Hp ≈ Hf.
conﬁrm that the qualitative relationships (relative values of Hc, Hp, Hf) were consistent in the vicinity of the ﬁeld measurements. This was done because all the study rivers are predominantly sand-bed, with numerous active bars and bedforms. The elevations, particularly in the main channels, are therefore likely to be locally variable. Thus the ﬁeld-based measurements presented in the results section are conﬁrmed to be typical of nearby valley cross-sections with respect to general relative elevations by DEM analysis, and the DEM-based measurements are conﬁrmed with respect to relative, qualitative relationships by ground-truthing. The abandoned or subordinate channels were classiﬁed in the ﬁeld as active, semi-active, inﬁlled, tributary-occupied, or billabongs (sloughs), as described above based on ﬁeld observations of ﬂow, channel morphology and connectivity, channel vegetation, and presence or absence of bedforms or other ﬂow indicators.
Based on the data and information described above and previous work on avulsions and other channel changes within the study area (Waters and Nordt, 1995; Aslan and Blum, 1999; Wellmeyer et al., 2005; Sylvia and Galloway, 2006; Phillips and Slattery, 2007, 2008; Phillips, 2008, 2009; Taha and Anderson, 2008), the primary factors controlling avulsions were identiﬁed. These factors were then categorized as universal (potentially applying to any alluvial river), and local (directly related to the environmental setting of the study area, and not applicable to all alluvial rivers). The relationships among these factors were represented as directed graphs indicating the causal relationships of the components or factors on each other. Three separate graphs were constructed, one representing the interactions of all the local and universal factors, and a second representing interactions of only the universal factors. The third graph represents a contextualized analysis within a speciﬁc aggradation situation (e.g., where the channel and ﬂoodplain aggradation relationship is held constant and thus removed from the model). The connectance entropies and information of the three models were determined and compared using Eqs. (1)–(3). 4. Results 4.1. Avulsion controls Analysis of avulsions within the study area conﬁrms the general relationship between aggradation and a tendency to avulse. All ﬁve of the study reaches are in some stage of valley ﬁlling, and avulsion styles and regimes are directly related to these stages (Aslan and Blum, 1999; Blum and Aslan, 2006; Phillips, 2009). As suggested by the SM, general aggradation alone is insufﬁcient to set up avulsions; within-channel aggradation exceeding ﬂoodplain accretion, and a domination of vertical as opposed to lateral adjustments are required (Jerolmack and Mohrig, 2007). In several locations within the study area, vertical accretion on ﬂoodplains has occurred or lateral channel bars have been built in recent decades even as channels have incised. Examples from the Sabine and Trinity Rivers, respectively, are given by Phillips (2003) and Phillips et al. (2005). Fieldwork for this project found evidence of an unsuccessful avulsion (a crevasse channel which was formed but does not capture any ﬂow except during ﬂoods) at the Sabine site (Fig. 2). The most active avulsion regime in the study area is in the Navasota River, where active channel aggradation is common, as evidenced by the very high frequency of overbank ﬂow at gaging stations in the lower Navasota (Table 1). With respect to overbank ﬂows, the most active avulsion regimes are associated with the portions of the study area that experience the most frequent overbank ﬂows. This is well illustrated in the lower Navasota River, where all but the lowermost 25 km, where the Navasota occupies a Brazos River paleochannel with rare overbank
J.D. Phillips / Geomorphology 130 (2011) 17–28
rivers (Fig. 3). Likewise, focusing of thalweg ﬂow roughly perpendicular to banks by bar formation is also evident (Fig. 4). Levee weaknesses may be associated with local topographic variability, paths and trails across levee tops created by fauna (including humans), and sites of former crevasses. Another possibility is collapsed animal (e.g., beaver) burrows, though this was not observed in the ﬁeld. A frequently observed cause of local depressions and denudation in levee sides and tops in the study area is tree uprooting (Fig. 5). Two additional local factors were identiﬁed in the study area— abandoned channels, and ﬂoodplain depressions associated with “Deweyville” paleomeanders. Avulsion by reoccupation of older Pleistocene channel belts was identiﬁed by Aslan and Blum (1999; Blum and Aslan, 2006) as the dominant avulsion style in the Trinity and Nueces Rivers, and in earlier stages of valley ﬁlling in the Brazos and Colorado Rivers, Texas. Reoccupation and channel-switching is also observed in the Navasota, Neches, and Sabine Rivers. Floodplain depressions—in ﬁeldwork for this study the largest depressions were always associated with Deweyville paleomeanders —create potentially steeper cross-valley gradients from banktops and levees toward the depressions. These have been shown to play a role in avulsions in the lower Brazos (Taha and Anderson, 2008), Trinity (Phillips and Slattery, 2008), and Sabine (Phillips, 2008). These depressions do not invariably promote avulsions, however. While river-to-depression gradients may be steep relative to the river channel, because downvalley slope gradients out of the depressions are often not advantageous, avulsions may occur away from the
Fig. 2. Crevasse channel on the Sabine River. Photograph is taken from the levee breach. While this channel conveys ﬂow to a ﬂoodplain depression during ﬂoods, it was not a successful avulsion because there is no cross-valley slope advantage at this point.
ﬂows, is characterized by multiple channels. The lowermost portions of the Trinity and Sabine Rivers, where overbank ﬂows occur with greater than annual frequency (Table 1), also contain the majority of the avulsions cataloged by Phillips (2009). As is apparently the case elsewhere (c.f. Slingerland and Smith, 2004; Aslan et al., 2005), cross-valley slope advantages relative to the river channel are a necessary but not a sufﬁcient condition for avulsions to occur. This is well illustrated by the Pickett's Bayou area of the lower Trinity River. Five separate crevasse channels were observed connecting the Sabine to Pickett's bayou. All slope away from the river channel, and at high ﬂows Pickett's Bayou functions as a distributary, with Trinity River water ﬂowing through the high-ﬂow channels into the bayou, and thence to Old River, an anabranch in the Trinity River delta. Maintenance of these high-ﬂow channels is made possible by a strong local cross-valley slope advantage associated with a Pleistocene paleodepression. However, the Pickett's Bayou–Old River ﬂow path is less steep than the Trinity River channel (Phillips and Slattery, 2007). By contrast, 50 to 75% of the Sabine River ﬂow is diverted through Cutoff Bayou to Old River, Louisiana, which is both deeper and steeper than the Sabine channel (Phillips, 2008). While either some ﬂow deﬂection or a localized levee weakness is necessary to initiate a crevasse, both need not be present. Thus these are considered local factors. Other than human agency, the most likely ﬂow deﬂectors in the study rivers are large woody debris jams and channel bars. Successful avulsions generally obliterate evidence of the original trigger. However, LWD is an important component of rivers in the Gulf Coastal Plain, and the historical record includes much evidence of logjams and ﬂow displacements and diversions (Phillips and Park, 2009). Examples of ﬂow diversions due to woody debris are common in the region on tributaries and distributaries of the major
Fig. 3. Flow blockages or deﬂections by large woody debris in a distributary on the Sabine River delta (top) and in a tributary of the lower Navasota River (bottom).
J.D. Phillips / Geomorphology 130 (2011) 17–28
Tectonic movements are another local factor which can trigger avulsions, and have been implicated in a Quaternary avulsion cluster on the lower Brazos River (Taha and Anderson, 2008). However, because such movements are independent of the other components they were not included in the analysis. 4.2. Linkages
Fig. 4. Flow deﬂections by channel bars in the lower Sabine River.
paleomeander depressions. This is the case in Grama Grass Bottom on the Trinity River, for example, where the Trinity River paleochannel leads into the depression, while the modern channel from the avulsion site follows a gradient that is, overall, steeper.
The components and their interconnections are shown in Fig. 6. The links reﬂecting the inﬂuences of overbank discharges, channel and ﬂoodplain aggradation, cross-valley slope, and the ﬂow deﬂection and levee weakness triggers follow from the principles discussed earlier. The positive and negative inﬂuences, respectively, of channel and ﬂoodplain aggradation on overbank ﬂow are based on basic geometric reasoning. Independently of other factors, channel sedimentation reduces capacity and will displace ﬂow from the channel more often; while ﬂoodplain accretion increases bank heights and channel capacity. The positive link from overbank ﬂow to ﬂoodplain aggradation is based on the necessity of such ﬂows for vertical accretion. The negative relationship between ﬂoodplain aggradation and cross-valley slope (i.e., ﬂoodplain sedimentation tends to reduce gradients) is based on the inﬁlling of ﬂoodplain depressions and paleochannels. This can be inferred from the stages of valley ﬁlling identiﬁed by Blum and Aslan (2006), and from the burial of such features in the lowermost valleys affected by Holocene sea level rise (Anderson et al., 2008; Milliken et al., 2008). As indicated in the previous section, paleomeander depressions may locally increase slopes, while decreasing downvalley slopes. Fig. 6 shows a positive link, reﬂecting the former, but in some cases the link could be negative. Note what while the sign of the links inﬂuences the stability properties of the system, the signs do not affect the connectance entropy. Paleochannels may locally inhibit avulsions if inﬁlled with resistant clay plugs, but this was not observed in the study area, where the presence of potentially reoccupiable paleochannels is an important inﬂuence on avulsions. While some large older paleomeanders are clay-ﬁlled, in only one case (on the Trinity River, see Phillips and Slattery, 2008) was one observed to intersect the modern channel. The reverse link, from avulsions to abandoned channels, is usually positive, as shown in Fig. 6. In the case of relocation avulsions and multiple anastamoses or distributaries where not all channels are active, avulsions create paleochannels. However, the link could conceivably be negative if an avulsion results in the reoccupation of a paleochannel without abandonment of the original channel. Avulsions are self-limiting in the immediate vicinity and in the short run, in that once a channel shift occurs then the setup factors for avulsion are eliminated (Stouthamer and Berendson, 2007). Further, where avulsions create anabranches, or abandoned channels persist, crevasses drain to an existing channel rather than cutting a new one. However, at broader scales, avulsions create potentially reoccupiable paleochannels that may facilitate avulsions from upstream sites or at longer time scales, providing the setup factors redevelop. Avulsions also feed back to ﬂoodplain aggradation, as they are an important mechanism of ﬂoodplain construction. 4.3. Connectance entropy
Fig. 5. Levee weakness created by undermining and uprooting of trees, lower Sabine River.
Fig. 6 shows the graph representing the entire set of relationships among the key components or factors. This includes the local factors of ﬂow deﬂection and levee weakness triggers and the presence of abandoned channels that directly promote avulsion, and paleomeander depressions, that inﬂuence avulsions through cross-valley slopes. The latter, as well as banktop discharges, are universal factors promoting avulsions, as is a rate of channel aggradation greater than the rate of ﬂoodplain aggradation. Fig. 7 shows only the universal factors.
J.D. Phillips / Geomorphology 130 (2011) 17–28
Fig. 6. Relationships among the major universal and local factors inﬂuencing avulsions in southeast Texas rivers.
Treating Fig. 6 as a directed graph, its connectance entropy is 2.00. The connectance entropy of the universal or global version (Fig. 7) is 1.56, indicating (from Eq. 2) the local contribution to entropy as 0.44. This indicates that universal factors account for 78% and local factors 22% of the total conﬁgurational entropy. The results from the information perspective are qualitatively similar—i.e., local factors account for a signiﬁcant, but very much smaller, proportion of the information (Table 3). Fig. 8 is a constrained graph which can be thought of as a withinriver formulation for a ﬂuvial system with a given aggradation regime. That is, it shows the controlling factors and their relationships in the context of a constraining aggradation regime that allows the possibility of crevasses and avulsions. The entropy of this conﬁguration is 1.89, with entropy accounting for only the local factors 1.56, or 83% of the total. The universal factors in this case account for only 17% of the entropy and 41% of the information (Table 3). 4.4. Channel, abandoned channel, and ﬂoodplain elevations The elevations of a number of paleochannels and ﬂoodplains relative to the main channel, and the state of the abandoned channels are shown in Table 4. These are interpreted with respect to the SM as described in Table 2. No cases were found where Hc b Hf b Hp. In two cases Hc ≈ Hp b Hf. One of these is a distributary in the lower portion of the Trinity River delta; the other a tributary-occupied channel in the lower Brazos River valley. In two cases Hc b Hp ≈ Hf. In both cases the paleochannels are inﬁlled, indicating that the original channel bed at the time of abandonment was lower than the ﬂoodplain surface. In one case (Grama Grass Bottom on the Trinity River), another portion of the
same paleochannel is occupied by a tributary, and the bed of the tributary-occupied section is nearly 3.5 m below the ﬂoodplain surface. These abandoned channels thus do not support the superelevation model. Table 4 shows 24 cases where Hc b Hp b Hf. If the abandoned channel is fully or partially inﬁlled this indicates that the SM does not apply. This occurs in 15 cases. In the eight other cases, active anabranches, distributaries, or tributaries occupy the paleochannels. Thus channel bed superelevation-driven avulsions could have occurred, with subsequent incision of the abandoned channel. The two entries in Table 4 where Hp b Hc are associated with the Sabine River and Cutoff Bayou and are a special case (see Phillips, 2008 for more details). More than half the Sabine's ﬂow is diverted through Cutoff Bayou to Old River, the lower portion of a former Sabine channel. This reoccupied paleochannel has both steeper slope and a lower-elevation bed than the Sabine River in the same vicinity. The bayou channel elevation is slightly higher (b1 m) than that of the Sabine immediately upstream of the conﬂuence, and slightly lower than that downstream. The Old River channel is also 0.7 m below that of the Sabine. This accounts for the negative Hp entries in Table 4. Overall, in the 30 cases examined here, in 18 (60%) the SM does not apply. The 12 other cases represent possible SM-type avulsions. 5. Discussion 5.1. Local and universal entropies The local and universal entropies of the qualitative interrelationships among factors inﬂuencing avulsions in southeast Texas suggest that local factors are signiﬁcant. However, the relative importance
Fig. 7. Subgraph of Fig. 6, showing only the universal factors.
J.D. Phillips / Geomorphology 130 (2011) 17–28
Table 3 Connectance entropy (C) and information (I) of the avulsion models (Figs. 6–8). Graph/model
Total (Fig. 6) Universal (Fig. 7) Local Constrained (Fig. 8) Constrained universal Constrained local
2.00 1.56 0.44 1.89 0.33 1.56
1.00 0.78 0.22 1.00 0.17 0.83
0.89 0.75 0.14 0.39 0.16 0.23
1.00 0.84 0.16 1.00 0.41 0.59
Table 4 Relative elevations of active and paleochannels and ﬂoodplains, state of original channel, and interpretation relative to the superelevation model (SM)a. Sites based on or conﬁrmed by ﬁeld measurements are shaded.
Location Sabine, Sand Lake Sabine R./Cutoff Bayou Upstream of confluence
5.2. Superelevation Results from the southeast Texas coastal plain, where avulsions are common, suggest that superelevation—where channels aggrade to the point that their elevation exceeds that of the ﬂoodplain beyond the natural levee—is not a necessary condition for avulsions. Both the geomorphological reasoning and empirical evidence supporting the superelevation model (Jerolmack and Mohrig, 2007) are sound, suggesting that superelevation may well be a sufﬁcient, but not necessary, condition for avulsion. Where channel aggradation exceeds ﬂoodplain accretion to create superelevated conditions, an avulsion is
Billabong; non-SM Active; possible SM
Active; possible SM
-0.7 1.0 2.0 1.1 3.1 2.1 1.7 2.9 1.0
4.8 3.0 3.0 4.2 4.3 4.3 4.3 4.2 2.5
Trinity R./Mussel Shoals Cr.
Trinity, Grama Grass Bottom Trinity R./Green’s Bayou
Trinity R./Pickett’s Bayou Lower Trinity delta Navasota, Sulphur Springs Navasota, Sulphur Springs Navasota, Bundic Rd. Navasota, Bundic Rd. Navasota, Democrat Crossing Navasota, Democrat Crossing Navasota, Democrat Crossing Navasota, Democrat Crossing Brazos near Millican
3.9 0.0 3.1 2.4 4.1 4.6 2.7 2.7 1.0 1.4 2.4
5.7 1.9 3.7 3.7 5.4 5.4 5.5 5.5 5.5 5.5 6.2
Brazos near Bryan Brazos near San Felipe
Brazos near Harris Reservoir Brazos near Harris Reservoir
Active; possible SM Distributary; possible SM Distributary; possible SM Active; possible SM Semi-active; non-SM Active; possible SM Active; possible SM Semi-active; non-SM 12 subchannels, tributary occupied or infilled; non-SM Tributary occupied & infilled; non-SM Infilled; non-SM Occupied by aggrading tributary; non-SM Semi-active; non-SM Active; possible SM Infilled; non-SM Billabong; non-SM Semi-active; non-SM Infilled; non-SM Billabong; non-SM Billabong; non-SM Active; possible SM Active; possible SM Tributary occupied; possible SM Infilled; non-SM Occupied by aggrading tributary; non-SM Semi-active; non-SM Occupied by aggrading tributary; non-SMs
Sabine R./Cutoff Bayou Downstream of confluence Sabine R./Old R./Cutoff Bayou Sabine Delta Sabine Delta Neches, John’s Lake Neches, John’s Lake area Neches, John’s Lake area Neches, John’s Lake area Neches, John’s Lake area Neches, Timber Slough Rd.
of universal and local factors varies considerably according to the context of analysis. In a between-river or between-reach context, universal factors contribute nearly 80% of the connectance entropy, signifying the importance of identifying the key setup factors—channel aggradation, cross-valley slope advantages, and likelihood of banktop ﬂow. The interrelationships involving local factors contribute a signiﬁcant but clearly subordinate amount to the connectance entropy. This suggests that a DCC-based approach to assessing avulsion potential at a broad scale could be based on the universal factors. In the constrained model context, corresponding roughly with a within-reach or within-river analysis where the (qualitative) relationship between channel and ﬂoodplain aggradation is constant, the situation is reversed. Local factors account for more than 80% of the connectance entropy, indicating that knowledge of the trigger factors and local ﬂoodplain morphology is the key to understanding and predicting avulsions. In this case a DCC-based approach should be based on the local factors most relevant—in the study area, ﬂoodplain topography as inﬂuenced by former channels and paleomeander depressions, large woody debris and bar patterns, and levee disturbances.
Channel elevation Hc is set to zero in each case. Elevation (m) of abandoned river channel (Hp). Elevation (m) of ﬂoodplain surface beyond the natural levee.
Fig. 8. Constrained case: relationships among the major universal and local factors inﬂuencing avulsions in southeast Texas rivers, assuming a valley ﬁlling or aggradation regime potentially conducive to avulsions.
J.D. Phillips / Geomorphology 130 (2011) 17–28
likely inevitable and imminent. However, avulsions do not necessarily require such conditions. The SM directly or indirectly incorporates the universal, setup factors involved in avulsion. Aggradation is explicitly included, and because the discharge regime is considered constant, changes in bankfull ﬂow frequency are a function of the relative aggradation of channel beds and ﬂoodplains, if channel width does not change. Local cross-valley slope gradients are a function of the relative channel and ﬂoodplain elevations. Where the SM explains avulsions this indicates that universal factors constitute adequate general explanations, though local trigger factors may still be required to explain speciﬁc locations and timings of channel shifts. Avulsions inconsistent with the SM suggest that local factors such as ﬂow deﬂection or levee weakness triggers, and/or local details of ﬂoodplain topography are required to explain the avulsions. Results here suggest that in the majority of cases the SM does not apply. A signiﬁcant minority of cases (40%) are possible or likely SM-type avulsions, but no unequivocal evidence of superelevation was found.
divert ﬂow from the main channel during high discharges, and levee weakness factors. Relationships among the universal and local factors were represented as box-and-arrow diagrams, and the conﬁgurational complexity determined using connectance entropy. Results indicate that local factors are signiﬁcant, but that the relative importance of universal and local factors varies with the scale or scope of analysis. In the broadest context, universal factors contribute nearly percent of the connectance entropy, signifying the importance of identifying the key setup factors. In a constrained situation where the aggradation regime is suitable, local factors account for more than 80% of the connectance entropy, indicating that knowledge of the trigger factors and local ﬂoodplain morphology is the key to understanding and predicting avulsions. Consideration of conﬁgurational complexity, connectance entropy, and universal vs. local factors facilitates a dominant control concept approach to predicting or modeling avulsions.
5.3. Predicting river avulsions
This work has been supported by several research contracts in connection with the Texas Instream Flow Program, administered by the Texas Water Development Board (TWDB). Greg Malstaff and Mark Wentzel of the TWDB have been particularly helpful in facilitating this work. The Sabine River Authority (SRA) of Texas, and the Big Thicket Research Association also assisted with ﬁeld work and logistics. Able ﬁeld and GIS assistance has been provided by Sarah McCormack, Calvin Harmin, and Lauren Bolender of the University of Kentucky; Jerry Wiegreffe, James East, and Ken Wilson of the SRA; and Jordan Furnans and Mark Wentzel of the TWDB. Two anonymous reviewers provided helpful and insightful comments on an earlier draft.
Results of this study suggest a three-stage process for predicting avulsions. First, the universal factors of channel aggradation, banktop discharge, and cross-valley slope relative to channel slope can be used to identify reaches with the potential to avulse. Second, as a sufﬁcient but not necessary condition, superelevation can be used to identify speciﬁc locations where avulsions are imminent. As a necessary but not sufﬁcient condition, the absence of cross-valley slope advantages can be used to identify locations where crevasses may occur, but successful avulsions cannot. The ﬁnal stage—predicting avulsions at other locations within systems or reaches where the setup factors are present, requires speciﬁc consideration of local factors. In southeast Texas, this points to the need for more research on the effects of antecedent and inherited morphology, on mechanisms of ﬂow deﬂection, and the prevalence and causes of levee weaknesses or depressions. Woody debris is clearly a possible cause of deﬂections. Transverse bars are common in the Trinity, Neches, and Sabine Rivers, and are likely present in the other rivers as well. The potential growth of these bars as a deﬂection mechanism also deserves study. Field observations suggest that previous crevasses, tree uprooting, local erosion, animal trails, and off-road-vehicle tracks are all commonly associated with low points in natural levees. Based on conﬁgurational complexity, the universal factors can be considered the dominant controls with respect to between-river or between-reach variations in avulsion regime. However, within a given aggradation setting, or where the aggradation regime allows avulsions, the dominant controls are local factors controlling local ﬂoodplain topography, availability of occupiable paleochannels, and ﬂow deﬂection and levee weakness triggers. 6. Conclusions Avulsions in alluvial rivers of southeast Texas are inﬂuenced by a combination of universal factors relevant to any alluvial river, and local factors at least partly contingent on the geographic/environmental setting and history of the study rivers. The universal factors are setup factors that create conditions under which avulsions can occur: channel aggradation, banktop discharge, and cross-valley slope advantages. A rate of channel aggradation greater than that of ﬂoodplain accretion is an important setup factor, but superelevation (channel bed elevation greater than or equal to ﬂoodplain elevation behind the natural levee) is not required. Slope advantages are necessary, but not sufﬁcient, for avulsions to occur. Local factors include abandoned channels on the ﬂoodplain, and basins or depressions associated with higher-discharge Pleistocene conditions. The local controls also include deﬂection factors that
References Anderson, J.B., Rodriguez, A.B., Milliken, K.T., Taviani, M., 2008. The Holocene evolution of the Galveston estuary complex, Texas: evidence for rapid change in estuarine environments. In: Anderson, J.B., Rodriguez, A.B. (Eds.), Response of Upper Gulf of Mexico Estuaries to Climate Change and Sea-Level Rise: Geological Society of America Special Paper, 443, pp. 89–104. Aslan, A., Blum, M.D., 1999. Contrasting styles of Holocene avulsion, Texas Gulf Coastal Plain, USA. In: Smith, N.D., Rogers, J. (Eds.), Fluvial Sedimentology VI. Special Publication, International Association of Sedimentology. Blackwell, Oxford, pp. 193–209. Aslan, A., Autin, W.J., Blum, M.D., 2005. Causes of river avulsion: insights from the late Holocene avulsion history of the Mississippi River, USA. Journal of Sedimentary Research 75, 650–664. Asquith, W.H., Heitmuller, F.T., 2008. Study of annual mean and annual harmonic mean statistics of daily mean streamﬂow for 620 U.S. Geological Survey Streamﬂow Gaging Stations in Texas Through 2007: Austin, TX: U.S. Geological Survey Data Series, 372. 259 pp. Blum, M.D., Aslan, A., 2006. Signatures of climate vs. sea-level change within incised valley-ﬁll successions: Quaternary examples from the Texas Coastal Plain. Sedimentary Geology 190, 177–211. Blum, M.D., Morton, R.A., Durbin, J.M., 1995. “Deweyville” terraces and deposits of the Texas Gulf coastal plain. Gulf Coast Association of Geological Societies Transactions 45, 53–60. Brasington, J., Richards, K. (Eds.), 2007. Reduced-complexity geomorphological modelling for river and catchment management: Special Issue, Geomorphology, 90, pp. 171–366. DeBoer, D.H., 1992. Hierarchies and spatial scale in process geomorphology: a review. Geomorphology 4, 303–318. Haigh, M.J., 1987. The holon: hierarchy theory and landscape research. Catena suppl. 10, 181–192. Harrison, S., 1999. The problem with landscape. Geography 84, 355–363. Harrison, S., 2001. On reductionism and emergence in geomorphology. Transactions, Institute of British Geographers 26, 327–339. Hoosbeek, M.R., Bryant, R.B., 1992. Towards the quantitative modeling of pedogenesis— a review. Geoderma 55, 183–210. Hudson, P.F., Kesel, R.H., 2000. Channel migration and meander bend curvature in the lower Mississippi River prior to major human modiﬁcation. Geology 28, 531–534. Jerolmack, D.J., Mohrig, D., 2007. Conditions for branching in depositional rivers. Geology 25, 463–466. Jerolmack, D.J., Paola, C., 2007. Complexity in a cellular model of river avulsion. Geomorphology 91, 259–270. Jones, L.S., Schumm, S.A., 1999. Causes of avulsion: an overview. In: Smith, N.D., Rogers, J. (Eds.), Fluvial Sedimentology VI. Special Publication, International Association of Sedimentology. Blackwell, Oxford, pp. 171–178.
J.D. Phillips / Geomorphology 130 (2011) 17–28
Lane, S.N., Richards, K.S., 1997. Linking river channel form and process: time, space, and causality revisited. Earth Surface Processes and Landforms 22, 249–260. Leigh, D.S., Srivastava, P., Brook, G.A., 2004. Late Pleistocene braided rivers of the Atlantic Coastal Plain, USA. Quaternary Science Reviews 23, 65–84. Mackey, S.D., Bridge, J.S., 1995. Three-dimensional model of alluvial stratigraphy: theory and application. Journal of Sedimentary Research B65, 7–31. Makaske, B., 2001. Anastamosing rivers: a review of their classiﬁcation, origin, and sedimentary products. Earth-Science Reviews 53, 149–196. Makaske, B., Smith, D.G., Berendsen, H.J.A., 2002. Avulsions, channel evolution and ﬂoodplain sedimentation rates on the anastamosing upper Columbia River, British Columbia. Sedimentology 49, 1049–1071. McCarthy, T.S., Ellery, W.N., Stanistreet, G., 1992. Avulsion mechanisms on the Okavango fan, Botswana: the control of a ﬂuvial system by vegetation. Sedimentology 39, 779–795. Milliken, K.T., Anderson, J.B., Rodriguez, A.B., 2008. Tracking the Holocene evolution of Sabine Lake through the interplay of eustasy, antecedent topography, and sediment supply variations, Texas and Louisiana, USA. In: Anderson, J.B., Rodriguez, A.B. (Eds.), Response of Upper Gulf of Mexico Estuaries to climate change and sea-level rise: Geological Society of America Special Paper, 443, pp. 65–88. Morton, R.A., Blum, M.D., White, W.A., 1996. Valley ﬁlls of incised coastal plain rivers, southeastern Texas. Gulf Coast Association of Geological Societies Transactions 46, 321–331. Murray, A.B., 2007. Reducing model complexity for explanation and prediction. Geomorphology 90, 178–191. Nanson, G.C., Knighton, A.D., 1996. Anabranching rivers: their cause, character, and classiﬁcation. Earth Surface Processes and Landforms 21, 217–239. Phillips, J.D., 1992. Qualitative chaos in geomorphic systems, with an example from wetland response to sea level rise. Journal of Geology 100, 365–374. Phillips, J.D., 2002. Global and local factors in earth surface systems. Ecological Modelling 149, 257–272. Phillips, J.D., 2003. Toledo Bend Reservoir and geomorphic response in the lower Sabine River. River Research and Applications 19, 137–159. Phillips, J.D., 2007. The perfect landscape. Geomorphology 84, 159–169. Phillips, J.D., 2008. Geomorphic controls and transition zones in the lower Sabine River. Hydrological Processes 22, 2424–2437. Phillips, J.D., 2009. Avulsion regimes in southeast Texas rivers. Earth Surface Processes and Landforms 34, 75–87. Phillips, J.D., Park, L., 2009. Forest blowdown impacts of Hurricane Rita on ﬂuvial systems. Earth Surface Processes and Landforms 34, 1069–1081. Phillips, J.D., Slattery, M.C., 2007. Downstream trends in discharge, slope, and stream power in a coastal plain river. Journal of Hydrology 334, 290–303.
Phillips, J.D., Slattery, M.C., 2008. Antecedent alluvial morphology and sea level controls on form-process transition zones in the lower Trinity River, Texas. River Research and Applications 24, 293–309. Phillips, J.D., Walls, M.D., 2004. Flow partitioning and unstable divergence in ﬂuviokarst evolution in central Kentucky. Nonlinear Processes in Geophysics 11, 371–381. Phillips, J.D., Slattery, M.C., Musselman, Z.A., 2005. Channel adjustments of the lower Trinity River, Texas, downstream of Livingston Dam. Earth Surface Processes and Landforms 30, 1419–1439. Saldana, A., Ibanez, J.J., 2007. Pedodiversity, connectance, and spatial variability of soil properties, what is the relationship? Ecological Modelling 208, 342–352. Sauchyn, D.J., 2001. Modeling the hydroclimatic disturbance of soil landscapes in the southern Canadian plains: the problems of scale and place. Environmental Monitoring and Assessment 67, 277–291. Sivakumar, B., 2004. Dominant processes concept in hydrology: moving forward. Hydrological Processes 18, 234–235. Sivakumar, B., 2008. Dominant processes concept, model simpliﬁcation and classiﬁcation framework in catchment hydrology. Stochastic Environmental Research and Risk Assessment 22, 737–748. Slingerland, R., Smith, N.D., 1998. Necessary conditions for a meandering-river avulsion. Geology 26, 435–438. Slingerland, R., Smith, N.D., 2004. River avulsions and their deposits. Annual Review of Earth and Planetary Sciences 32, 257–285. Stouthamer, E., Berendsen, H.J.A., 2007. Avulsion: the relative roles of autogenic and allogenic processes. Sedimentary Geology 198, 309–325. Sylvia, D.A., Galloway, W.E., 2006. Morphology and stratigraphy of the late Quaternary lower Brazos valley: implications for paleo-climate, discharge, and sediment delivery. Sedimentary Geology 190, 159–175. Taha, Z.P., Anderson, J.B., 2008. The inﬂuence of valley aggradation and listric normal fauling on styles of river avulsion: a case study of the Brazos River, Texas, USA. Geomorphology 95, 429–448. Tornqvist, T.E., Bridge, J.S., 2002. Spatial variation of overbank aggradation rate and its inﬂuence on avulsion frequency. Sedimentology 49, 891–905. Vandenberghe, J., 2002. The relation between climate and river processes, landforms, and deposits during the Quaternary. Quaternary International 91, 17–23. Waters, M.R., Nordt, L.C., 1995. Late Quaternary ﬂoodplain history of the Brazos River in east-central Texas. Quaternary Research 43, 311–319. Wellmeyer, J.L., Slattery, M.C., Phillips, J.D., 2005. Quantifying downstream impacts of impoundment on ﬂow regime and channel planform, lower Trinity River, Texas. Geomorphology 69, 1–13. Werner, B.T., 1995. Eolian dunes: computer simulation and attractor interpretation. Geology 23, 1107–1110. Werner, B.T., 1999. Complexity in natural landform patterns. Science 284, 102–104.