Landform transitions in a fluviokarst landscape

Abstract

Flow diversions and landform transitions between channelized surface (fluvial) and concentrated subsurface (karst conduit) flows may be common in fluviokarst landscapes. Identifying landforms associated with fluvialto- karst or karst-To-fluvial transitions shows this to be the case at three study sites in the Inner Bluegrass karst region of Kentucky. Forms representing the capture or diversion of stream flow to subsurface conduits include sinking streams, dry karst valleys, paleovalleys resulting from karst stream piracy, and alluvial collapse dolines. Features indicating karst-To-fluvial transitions include stream incision into dry karst valleys, doline breaching by surface runoff or stream incision, and formation of karst pocket valleys and karst windows. Many smaller transitional landforms also exist (e.g., stream swallets). The three study sites have about three larger transitional landform features per km2, with karst-To-fluvial features slightly more common. Dissolution in bedrock-controlled stream channels leading to karst piracy is the most common cause of fluvial to karst transitions, while general stream incision driven by Kentucky River downcutting is the main driver of karst-To-fluvial shifts. The landform transitions are examined via a network model based on hydrological probability and flow partitioning. The model is dynamically unstable. Instability indicates that local changes and disturbances that modify moisture fluxes, local relief, conduit or surface channel conveyance capacity, or hydraulic slopes are likely to persist and grow, resulting in hydrogeomorphic transitions. Evolution of the Kentucky fluviokarst is best understood as mutual reinforcement, whereby fluvial dissection can be intensified and accelerated by the presence of karst features, and karstification is enhanced by stream incision. © 2017 Gebr. Borntraeger Verlagsbuchhandlung, Stuttgart, Germany.

Full text

Zeitschrift für Geomorphologie, Vol. 61/2 (2017), 109–122 Article
Published online June 2017; published in print August 2017
© 2017 Gebr. Borntraeger Verlagsbuchhandlung, Stuttgart, Germany www.borntraeger-cramer.de
DOI: 10.1127/zfg/2017/0452 0372-8854/zfg/2017/0452 $ 3.50
Landform transitions in a uviokarst landscape
Jonathan D. Phillips
Department of Geography, University of Kentucky, Lexington, KY 40508, USA. jdp@uky.edu
With 12 gures and 3 tables
Abstract: Flow diversions and landform transitions between channelized surface (uvial) and concentrated subsur-
face (karst conduit) ows may be common in uviokarst landscapes. Identifying landforms associated with uvial-
to-karst or karst-to-uvial transitions shows this to be the case at three study sites in the Inner Bluegrass karst region
of Kentucky. Forms representing the capture or diversion of stream ow to subsurface conduits include sinking
streams, dry karst valleys, paleovalleys resulting from karst stream piracy, and alluvial collapse dolines. Features
indicating karst-to-uvial transitions include stream incision into dry karst valleys, doline breaching by surface run-
off or stream incision, and formation of karst pocket valleys and karst windows. Many smaller transitional landforms
also exist (e.g., stream swallets). The three study sites have about three larger transitional landform features per km2,
with karst-to-uvial features slightly more common. Dissolution in bedrock-controlled stream channels leading to
karst piracy is the most common cause of uvial to karst transitions, while general stream incision driven by Kentucky
River downcutting is the main driver of karst-to-uvial shifts. The landform transitions are examined via a network
model based on hydrological probability and ow partitioning. The model is dynamically unstable. Instability indi-
cates that local changes and disturbances that modify moisture uxes, local relief, conduit or surface channel convey-
ance capacity, or hydraulic slopes are likely to persist and grow, resulting in hydrogeomorphic transitions. Evolution
of the Kentucky uviokarst is best understood as mutual reinforcement, whereby uvial dissection can be intensied
and accelerated by the presence of karst features, and karstication is enhanced by stream incision.
Keywords: Fluviokarst; geomorphic transitions; hydrological probability; ow partitioning, mutual reinforcement
1. Introduction
Fluviokarst systems are characterized by an intercon-
nected combination of surface and subsurface hydrologi-
cal processes and ux paths, and both uvial and karst
landforms. Mass ux patterns in uviokarst can be quite
complex, and landforms may be transitional or hybrid
forms. Both the uvial and karst components of the sys-
tem are linked to the same base level and regional drain-
age controls, and coevolve accordingly. Previous work in
a variety of uviokarst settings and over a range of time
scales has identied both karst-to-uvial and uvial-to-
karst transitions as landscapes evolve (see below). A
better understanding of these transitions is not only rel-
evant to uviokarst landscape processes and evolution
per se, but also to landscape evolution more broadly,
as several studies indicate that a combination of uvial
and karst processes produces denudation rates consider-
ably higher than either process alone (Kaufman & Braun
2001, Jaillet et al. 2004, Simms 2004, Worthington 2005,
Bahtijarevic & Faivre 2016). Further, geomorphic pro-
cesses and hydrologic uxes are not as well understood
in bedrock-controlled streams or in karst systems as
compared to alluvial streams and non-karst groundwater
systems. The uvial-karst interactions make these phe-
nomena even more complex, with important implica-
tions for water resource and land management (see, e.g.,
Thrailkill et al. 1991, Currens & Graham 1993, Ray &
Blair 2005, Reed et al. 2010, and Currens 2012 for exam-
ples relevant to the central Kentucky study area). The
purpose of this study is to document uvial:karst geo-
morphic transitions in a uviokarst landscape in central
Kentucky, USA; to gain insight into the factors driving
them; and to evaluate the signicance of such transitions
in landscape evolution. Specically, this research seeks
to determine: (1) the extent to which karst-to-uvial or
uvial-to-karst transitions are dominant in the region; (2)
whether such transitions appear to be unidirectional at
the landscape scale (i.e., progressive development of a
more karst- or uvial-dominated landscape; and (3) the
dynamical stability of uviokarst transitions.
Previous research has identied various uvial-to-
karst (FK) shifts, where uvially dominated forms and
surface ow paths transition to karst and subsurface domi-
nation (e.g., Jennings et al. 1976, Mylroie & Mylroie 1990,
eschweizerbart_xxx

110 J.D. Phillips
Fabel et al. 1996, Mills & Mills 2001, Jaillet et al. 2004,
Ortega Becerril et al. 2010, Woodside et al. 2015). Like-
wise, karst-to-uvial (KF) transformations have also been
documented (e.g., Jaillet et al. 2004, Ortega Becerril et al.
2010, Tiria & Vijulie 2013, Lipar & Ferk 2015). The Inner
Bluegrass karst region of Kentucky, the study area for this
project, is no exception (Thrailkill et al. 1991, Currens &
Graham 1993, Phillips et al. 2004, Phillips & Walls 2004,
Ray & Blair 2005, Phillips 2015, Jerin & Phillips 2017).
Previous studies of geomorphic transitions in uviokarst
have generally focused on specic features or phenomena
(e.g., karst capture or meander cutoffs in streams; Jennings
et al. 1976, Mylroie & Mylroie 1990, Fabel et al. 1996,
Jerin & Phillips 2017), or landscape scale transformations
(e.g., Sauro 2002, Bocic et al. 2015). This study focuses
on FK and KF transitions at the scale of reaches of surface
or cave streams, low-order valleys, and dolines.
An evolutionary progression from an initially u-
vial landscape to uviokarst and nally to holokarst was
postulated by a number of karst researchers (e.g., Cvijic
1918, Roglic 1964, Ford 2007). However, early on it was
recognized that this progression is both far from inevita-
ble, and potentially reversible (e.g., Sawicki 1909, White
2009). For instance, Sauro’s (2002) analysis of the Monte
Berici karst, Italy, shows uvial development as the main
morphogenetic process, driven by climate change and
tectonic uplift. Karst forms developed later on relatively
inactive or relict uvial forms as the river abandoned its
uplifted planation surface. By contrast, Ortega Becerril
et al. (2010) examined a switch from domination by karst
and dissolutional erosion to dominance by uvial forms
and mechanical erosion in Spain. In central Kentucky,
there exists a tendency for divergent evolution into
channel-rich, karst-poor and karst-rich, channel-poor
zones within the broader uviokarst landscape (Phillips
et al. 2004, Phillips & Walls 2004).
Some landscapes, such as the Una-Korana plateau
in the Dinaric karst, do approximate the uvial-to-karst
sequence (Bocic et al. 2015). However, even here the
palaeodrainage network is still evident, the uvial-to-
karst transition is still underway, and the uvial network
contains both active, intermittently active, and relic or
inactive segments (Bocic et al. 2015). In other areas there
exist strong contrasts between nearby karst- and uvi-
ally-dominated landscapes, or between karst areas with
or without uvial imprints (e.g., Benac et al. 2013, Bahti-
jarevic & Faivre 2016).
1.1. Diagnostic landforms
The landforms identied and mapped for this project
include four associated with conversion of uvially-
dominated to karst-dominated landforms or diversion of
stream ow to groundwater conduits: sinking streams,
dry karst valleys, palaeochannels, and alluvial collapse
dolines. Landforms representing KF transformations
inventoried here include karst pocket valleys, karst
windows, headcuts that convert karst valleys to incised
stream channels, and breached dolines. These are dened
and described below.
A sinking stream (also called a losing stream) is
a characterized by surface ow in the channel that is
diverted to groundwater via swallets or sinks in the val-
ley bottom, or by inltration into the bed via joints and
fractures (Fig. 1). Such streams include reaches that have
Fig. 1. Examples of dry karst valley/sinking stream and incision into dry
karst valley landforms in the Raven Run study area. The basin indicated as
a possible breached doline likely has this origin, but could also be formed by
other processes. Such features were not included in the inventory unless
they could be condently identied. Base topographic relief map derived
from 1.5 m LiDAR data.
eschweizerbart_xxx

Landform transitions in a uviokarst landscape 111
Fig. 2. Examples of karst window, paleovalley created by
karst capture, and karst pocket valley landforms in the
Shawnee Run study area. Base topographic relief map
derived from 1.5 m LiDAR data.
perennial or seasonal surface ow, and reaches that are
entirely underground (no surface channel) or where the
surface channel or valley is ephemeral and conveys ow
only during wet periods when karst cavities and conduits
are lled. A dry karst valley (Fig. 1) is predominantly
under-drained by conduits, but may convey surface dis-
charge during wet periods when conveyance capacity of
conduits is exceeded. From a surface hydrology perspec-
tive they are considered ephemeral streams or unchan-
nelled valleys. Dry karst valleys in the study region occur
only on unincised uplands, and have generally low slopes
(<0.004; all slopes are given as gradients in m m–1). A
swallet is a vertical or subvertical depression connected
to karst conduits. They may be small or incipient dolines,
drains within dolines or other depressions, or openings to
vertical shafts. Of interest here are valley bottom stream
swallets representing FK transitions. Individual swal-
lets are not mapped separately in this study, but play an
important role in sinking streams and dry karst valleys.
Piracy or cutoff of surface streams may occur due to
subsurface ow diversions, leaving uvial palaeovalleys
(Fig. 2) that indicate FK transitions. Most often these
occur in the form of meander cutoffs (see Jennings 1976,
Mylroie & Mylroie 1990, Fabel et al. 1996, Mills & Mills
2001), though other types of capture also occur (Ortega
Becerill et al. 2010, Jerin & Phillips 2017). In the incis-
ing Kentucky River gorge area, these cutoffs are elevated
above modern channels. That the process is active is indi-
cated by the fact that about 25 percent of the perennial
springs in the area studied by Ray & Blair (2005) are
associated with cutoff or bypassing by conduits of short
segments of surface stream. Alluvial collapse dolines
(Fig. 3) occur along Kentucky River oodplains and
alluvial terraces due to the collapse of insoluble alluvium
into karst cavities in the underlying carbonate rocks.
Karst pocket valleys (Tiria & Vijulie 2013, Lipar &
Ferk 2015) are formed by collapse or erosional exhu-
mation of cave passages or larger conduits (see Fig. 2).
They are characterized by limited surface drainage areas,
steep, often ampitheater-like valley heads, exposed bed-
rock within the valley, and groundwater seepage along
joints, bedding planes, and small conduits exposed within
the valley. In Kentucky, these have a perennial trickle of
downvalley ow, and sometimes larger discharges. In
the study region pocket valleys have mean slopes >0.04.
Karst windows are underground karst streams exposed by
erosional exhumation, and have perennial ow (Fig. 2).
Mean slopes are lower than those of pocket valleys. They
are exposed portions of subsurface streams, as opposed
to broader scale transitions from underground to surface
streams. A breached doline occurs when a karst depres-
sion is penetrated by a uvial channel, such that the
doline is incorporated into the surface watershed (Fig. 4).
This can occur externally, due to headward migration
of incising streams into sinkholes. Breaches can also be
generated internally, as plugging or clogging of doline
drains or swallets, or of underlying conduits, leads to sur-
face overow and resulting erosion.
As the incision signal from the Kentucky River
migrates up the tributaries over time (see below), the
headcuts of downcutting streams incise into dry karst
valleys, converting these to dominantly uvial forms and
ow. These are termed karst valley to incision transitions
(Fig. 1). Photographic examples of the indicator land-
forms are shown in Figures 5 and 6.
Fig. 3. Examples of alluvial collapse dolines on the Kentucky
River oodplain in the Polly’s Bend study area. Base topo-
graphic relief map derived from 1.5 m LiDAR data.
eschweizerbart_xxx

112 J.D. Phillips
2. Study area and methods
The Inner Bluegrass physiographic region of Kentucky
is characterized by a humid subtropical climate. Land
use is predominantly agriculture, particularly grazing,
with forests concentrated mainly along river and stream
corridors. Bedrock is dominated by horizontally bedded
Ordovician limestones, with small amounts of dolomite,
calcitic shale, and bentonite. Karst features, particularly
surface landforms, are concentrated (but by no means
limited) to the Lexington Limestone formation, which
comprises the unincised upland surfaces. Older literature
characterized this surface as the Lexington peneplain, but
it was apparently formed by unroong during uplift of
an epeirogenic feature (the Cincinnati Arch) and is not a
planation surface. Older faults are found throughout the
region, but the area has not been tectonically active dur-
ing the Quaternary.
The Inner Bluegrass is drained by the Kentucky River,
which provides the base level for both karst and uvial
systems (Fig. 7). The river began incising through the
Ordovician limestones 1.3 to 1.8 Ma in response to gla-
cial rearrangement of the ancestral Ohio River system to
which the Kentucky River drained (Teller & Goldthwait
1991, Andrews 2004). The river has incised 60 to 130 m
in that time, forming the Kentucky River gorge, and the
tributaries have downcut in response. Larger tributar-
ies include a strongly incised lower reach graded to the
Kentucky River (though often with structurally-controlled
knickpoints), and smaller tributaries may have hanging
valley morphology where the tributary downcutting could
not keep pace (Phillips & Lutz 2008). The incisional
response has typically not reached the upper portions of
the larger tributaries, which are unincised and include
sinking streams and dry karst valleys.
Three study areas were selected within the Inner
Bluegrass, based on access, minimal topographic dis-
turbance (by construction, agriculture, mining, etc.),
and inclusion of both uvially incised and unincised
areas (Fig. 7). Sites were selected to also include areas
on the inside and outside of Kentucky River meander
bends, as these have experienced fundamentally dif-
ferent landscape responses with respect to uviokarst
development (Phillips 2015).
Raven Run is on the outside of a meander bend,
and the core of the area is managed as a public park
and nature preserve. The site is mostly forested, but
includes some areas of grassy meadow. The 7.55 km2
study area includes a small portion on the east (left)
bank of the river. The Polly’s Bend site is at Dupree
Nature Preserve, on the interior of a bend at Polly’s
Bend, a compound Kentucky River meander loop. The
3.68 km2 study area is mainly forested, with small areas
of meadow. The 9.14 km2 Shaker Preserve study site is
within the Shaker Village Nature Preserve, within the
watershed of Shawnee Run. This site includes a combi-
nation of forest, recently ungrazed meadow, and grazed
pastures. Shawnee Run joins the river on the outside of
a meander bend.
2.1. Landform mapping
The FK and KF transition landforms were mapped using
a combination of geographical information system and
eld approaches. The study areas were rst extensively
reconnoitered in the eld on foot, and via kayak from the
Kentucky River. Based on this (and literature review),
the key FK and KF landforms were identied, and the
locations of some individual landforms mapped.
GIS data were used to delineate the indicative land-
forms throughout the study sites. This included digital
elevation model (DEM) data at a 1.5 m (5 ft) horizon-
tal resolution, derived from LiDAR data. Both true and
false-color imagery at horizontal resolutions as ne as
0.3 m (1 ft), was used, with imagery taken between
2006 and 2015. Supporting GIS layers included sur-
face geological maps at a 1:24,000 scale, and cover-
age of springs and water wells. A karst potential map
of Kentucky (1:24,000 scale), and coverage of sinkhole
outlines were also used. All data (except for some of
the true-color aerial photographs, accessed through
Google EarthTM) were obtained through the Kentucky
Geological Survey (KGS; http://kgs.uky.edu/kgsweb/
main.asp#tabs-2 ). The Harrodsburg sheet of the Karst
Fig. 4. Examples of breached dolines in the Polly’s Bend
study area. Base topographic relief map derived from 1.5 m
LiDAR data.
eschweizerbart_xxx

Landform transitions in a uviokarst landscape 113
Atlas of Kentucky (Currens et al. 2003) was also used.
The atlas shows groundwater ow paths determined by
dye tracing, mapped karst groundwater basins, springs,
and stream sinks.
The DEM data proved most useful. The imagery was
useful mainly in the nonforested portions of the study
sites, due to heavy canopy cover in other sections. The
karst potential maps are based on surface lithology, and
provided redundant information to the geological maps.
The sinkhole layer was useful but incomplete, as some
dolines identied in the eld and via the DEM do not
appear in this data.
The FK and KF landforms identied from the GIS
data were then visited individually in the eld to verify
or ground-truth the interpretations. Landforms which
turned out to have equivocal properties when examined
in the eld, or which could not be accessed, were not
included in the nal maps (see e.g., Fig. 1).
Fig. 5. Indicator landform examples in the study areas. (A) Dry karst valley. (B) Dry karst valley with incipient chan-
nel active during wet weather. (C) Incised channel just downstream of a dry valley. (D) Sediment transported by
surface runoff transported into recently opened swallet (arrow). (E) Icicle Bend on Shawnee Run. While not a clas-
sic karst window, this section is part of an exhumed underground stream (see Jerin & Phillips 2017). (F) Alluvial
collapse dolines on the Kentucky River oodplain. All photos by author.
eschweizerbart_xxx

114 J.D. Phillips
Fig. 6. Indicator landform examples in the study areas. (A) Karst pocket valley. (B) Breached doline. Photo A by
author; B is Google EarthTM image with relief exaggerated 3X.
Fig. 7. Study areas (boxes) within the Inner Bluegrass region of Kentucky.
eschweizerbart_xxx

Landform transitions in a uviokarst landscape 115
2.2. Fluviokarst transition model
To explore transitions and evolutionary trajectories in
uviokarst systems, Smart (1988) developed a model
based on hydrological probabilities. The essence of the
model is that karst conduit conveyance capacity is related
to relief, dened as height or thickness of the karst aqui-
fer above its base level. When imposed discharge is less
than this conveyance capacity, ow is dominantly sub-
surface and karst evolution is dominant. When imposed
discharge is higher than conduit conveyance capac-
ity, ow is displaced to the surface, triggering surcial
hydrological processes and uvially-dominated land-
scape evolution. The model accounts for ow rates as a
function of hydraulic head and conduit size, exceedence
probabilities of discharge, conduit enlargement via disso-
lution, and sedimentation. Details are provided by Smart
(1988), but key for the current study are the feedback
mechanisms identied by Smart involving surface and
groundwater ow, conduit size, and relief.
Fluvial-karst interactions in Kentucky uviokarst
were previously analyzed by the author using a ow
partitioning model considering the apportioning of
surplus moisture to surface concentrated (channel)
discharge, overland diffuse (sheet) ow, concentrated
groundwater (conduit) discharge, and subsurface dif-
fuse ow. Phillips & Walls (2004) developed the origi-
nal conceptual model based on competition between
the ow paths, and Phillips (2015) developed threshold
criteria for the likelihood of competing ow paths. For
sub-banktop or sub-conduit-lling ow, cross-sectional
area or hydraulic radius do not limit discharge. In these
cases the threshold criterion is dominated by hydraulic
slopes (though roughness or friction factors may also
play a role).
These conceptual frameworks lead to the model
shown in Figure 8, showing the interrelationships among
surface and groundwater ow, conduit size and slope,
ground or surface slope, and karst relief (as dened by
Smart 1988). Surface and groundwater ow are com-
petitive (negatively inuence each other) due to a nite
amount of surplus moisture. Groundwater ow and con-
duit size are mutually reinforcing, due to the positive
feedbacks between discharge and dissolutional enlarge-
ment. As in Smart’s (1988) model, karst relief inu-
ences groundwater conveyance capacity by providing
additional space for subsurface ow. Thus karst relief
positively inuences groundwater and negatively affects
surface ow. In karst, ground and conduit slopes must
be treated separately. Steeper conduit slopes promote
groundwater ows, and inhibit surface runoff by direct-
ing more ow to the subsurface. Ground surface slopes
have the opposite effect – steeper slopes promote over-
land runoff and negatively inuence groundwater ow.
Ground slopes are also shown as self-limiting, owing
to nite available elevation above base level, and to
geomech anical limits on slope stability.
In addition to the models of Smart (1988) and Phillips
(2015), a number of empirical studies support the feed-
back links shown in Fig. 8. The discussion of factors
determining cave patterns by Audra & Palmer (2015)
either explicitly includes or implies feedbacks indicated
in Fig. 2, as does Bailly-Comte et al.’s (2009) work on
groundwater-surface water interactions on an event
(rather than landscape evolution) time scale.
In south central Kentucky karst, Lavalle (1967)
showed that structurally aligned karst depression con-
centrations and elongation are both positively related to
karst relief and subsurface hydraulic gradients. Williams
(1985) laid out the critical role of topographic slope in
favoring either karst or uvial features, which was con-
rmed for central Kentucky by Phillips & Walls (2004).
Tüfeçki & Sener (2007) identied the importance of base
level (related to karst relief) and slope in evolution of a
Turkish karst landscape. The key role of base level low-
ering for both karstication and uvial incision was out-
lined by Worthington (2005), among others.
Dissolution and chemical denudation are not directly
gradient dependent, as long as the base of the epikarst is
above the water table. However, slope advantages play a
role in ow partitioning, and strongly impact ow veloci-
ties. When slows are too slow, water becomes Ca satu-
rated and non-aggressive. More importantly, to the extent
hydraulic gradients increase conduit discharge, positive
feedbacks tend to increase conduit size via dissolution,
at least until reaction kinetics rather than moisture sup-
ply become limiting. These feedbacks have long been
included in simulation models, and Worthington (2015)
showed their applicability to channel (conduit) develop-
ment in carbonate aquifers. He also found that hydraulic
gradients are dominant for guiding channel development
Fig. 8. Fluvial-karst transition feedback model. Positive
feedback indicates that an increase or decrease in one com-
ponent results in a change in the component at the end of the
arrow in the same direction. Negative feedback indicates a
change in the opposite direction.
eschweizerbart_xxx

116 J.D. Phillips
at the regional scale. Interestingly, Worthington (2015)
also showed that development of caves (vs. smaller con-
duits) is favored where sinking stream recharge rather
than matrix percolation is dominant, implying that FK
transitions enhance cave development. Filipponi et al.
(2009, 2010) demonstrated the importance of initial
permeability or conduits, as well as the ow/dissolution
feedbacks.
An alternative version of the feedback model is
shown in Figure 9, which incorporates effects of topo-
graphic change on relief and slope. Aggregate conduit
capacity (inuencing exceedence probability; Smart
1988) may be reduced by surface denudation associated
with surface ow and uvial erosion – that is, surface
lowering increases the probability of “spillover” from
karst conduits. Fluvial downcutting increases relief, and
as the tributaries are partly incised this can be considered
as an internal feedback as well as an external control due
to Kentucky River incision. One could also postulate – or
for a specic case, identify – various positive and nega-
tive links between relief and surface slope, as indicated
by the double arrow in Fig. 9.
The network of interactions shown in Figures 8 and
9 was analyzed for dynamical stability using the Routh-
Hurwitz criteria. The rst criterion is that the system is
dynamically stable if and only if the (real parts of the)
eigenvalues of the interaction matrix of the network
are all negative. The interaction matrix is constructed
based on positive, negative, or zero entries according to
whether the row component has a positive, negative, or
no effect on the column component (components are the
boxes in Figs. 2, 3). Methods are standard in qualitative
asymptotic stability analysis (Puccia & Levins 1985),
and are described in detail in a geomorphological context
by Phillips (1999). The analysis was performed for the
situation shown in Fig. 8, and for Fig. 9 with the surface
ow to relief link as both positive and negative, and vari-
ous positive and negative links between relief and ground
slope.
Dynamical stability in this context means that the
effects of disturbances and changes tend to be dimin-
ished or offset over time, returning the uviokarst sys-
tem toward its predisturbance state. Thus, for example,
a newly-opened or widened swallet, or storm-related
headcut migration, would deterioriate or become clogged
with sediment or vegetation. Dynamical instability indi-
cates that the effects of disturbances tend to persist or
grow disproportionately large – thus a relatively small or
localized change to a stream or karst feature could result
in a transition. This analysis can thus provide insight as to
whether large events or major disturbances are required
to drive KF and FK transformations.
3. Results
Figures 10 to 12 show the mapped indicator landforms,
and summaries are shown in Tables 1 and 2. At Raven
Run, a total of 17 features were mapped and conrmed
(seven FK and 10 KF), with a density of 2.25 features
km–2. Note that these are minima, as only unequivo-
cally classiable features that were directly observed in
the eld were included, and small individual features
such as stream swallets are not included. Polly’s Bend
had the highest density of features (5.43 km–2), which
were dominantly (70 percent) FK. At Shaker Preserve
two thirds of the 24 mapped features were KF, with a
density of 2.63 km–2. The sites are discussed individu-
ally below.
3.1. Raven Run
The Raven Run site on the west side of the river and in
the nature sanctuary has a prevalence of KF transitions
(Fig. 10). Fluvial dissection is common on the outside of
Kentucky River bends in the gorge area (Phillips 2015).
As Raven Run and its tributaries actively incise, the
headcuts encroach on sinking streams and dry karst val-
leys in the upper reaches of the watershed. Similar pro-
cesses have breached at least one doline. There are also
at least three cases of pocket valleys on the steepening
valley side slopes of the incising streams. However, in
the lower reaches of Raven Run, two paleovalleys asso-
ciated with karst piracy exist, indicating FK transitions.
The other FK transitions are on the other side of the river
in the form of alluvial collapse dolines on the Kentucky
River oodplain or lower alluvial terrace.
The karst window feature in lower Raven Run is not
a classic window, exposing a portion of an otherwise
underground stream. Rather, it represents an exposed
conduit or cave passage that was apparently part of a
karst piracy event that was later subaerially exposed.
This is indicated both by the channel/valley morphology,
Fig. 9. Modied version of Fig. 8, including potential surface
erosion-relief and ground slope-relief feedbacks.
eschweizerbart_xxx

Landform transitions in a uviokarst landscape 117
and the relationship to the abandoned channels in the
lower portion of the creek valley.
3.2. Polly’s Bend
This site is mainly on the interior of a river meander
bend. In addition to the FK and KF features, Fig. 11
shows a linear zone of dolines across the bend. A high
density of dolines on meander bend interiors is typical
in the Kentucky River gorge, and more-or-less linear
alignments are not uncommon (Phillips 2015). Some-
times these are associated with mapped faults or struc-
tural trends. The formation of these features is unclear,
though they are associated with generally lower surface
slopes on bend interiors, and may be formed by col-
lapse into cross-bend conduits that convey groundwater
bypass ow (Phillips 2015). Though they are not
included in the inventory of FK or KF features, they are
clearly associated with uvial-karst interactions due to
their concentration relative to river meander bends, and
their presence on slip-off slopes associated with river
bend migration.
The dominance of FK transition features here is
mainly due to a large number (10 conrmed) of alluvial
collapse dolines on the river oodplain. No large tributar-
ies join the Kentucky River on meander bend interiors,
apparently being diverted during downcutting and bend
migration (Phillips 2015). Thus there are no karst valley-
incised channel transitions or paleochannels here, though
incision of small, local features has resulted in a breached
doline and several pocket valleys.
Table 1. Indicator landforms of uvial-karst transitions (number mapped in Figures 10–12).
Landform Transition Raven Run Polly’s Bend Shaker Preserve
Karst valley→incised channel Karst to uvial 5 09
Karst pocket valley Karst to uvial 3 5 5
Karst window Karst to uvial 101
Breached doline Karst to uvial 111
Sinking stream/dry karst valley Fluvial to karst 2 4 6
Stream swallet Fluvial to karst NM NM NM
Paleovalley: karst piracy Fluvial to karst 2 0 2
Alluvial collapse doline Fluvial to karst 310 0
NM = present but not mapped
Table 2. FK and KF landforms by study site.
Study site Area (km2) Maximum
Elevation (masl)
Total relief
(m)
Fluvial to Karst
(N, N/km2)
Karst to Fluvial
(N, N/km2)
Total
(N, N/km2)
Raven Run 7.55 307 139 7
0.93
10
1.32
17
2.25
Polly’s Bend 3.68 288 129 14
3.80
6
1.63
20
5.43
Shaker Preserve 9.14 285 129 8
0.88
16
1.75
24
2.63
All sites 20.37 29
1.42
32
1.57
61
2.99
Table 3. Interaction matrix for the uviokarst transition feedback model shown in Fig. 8, with positive and negative links shown
as 1 or –1.
Con size Sconduit Sground Relief Qsurface Qgroundwater
Conduit size 0 0 0 0 –1 1
Slope, conduit 0 0 0 0 –1 1
Slope, ground 0 0 0 0 1 –1
Relief 0 0 0 0 –1 1
Surface discharge (Q) 0 0 1 –1 0 –1
Groundwater discharge 1 0 0 0 –1 0
eschweizerbart_xxx

118 J.D. Phillips
3.3. Shaker preserve
Though this study area does not include the Kentucky
River, the Shaker site otherwise contains the greatest
variety of transitional features. Undercounting is likely
more pronounced here, as the establishment of a number
of stock ponds in the upper areas of the Shawnee Run
watershed has obscured or modied potential transitional
features.
The incision of Shawnee Run and its tributaries
has apparently triggered a number of KF transitions
as upstream migration of knickpoints and headcuts
encroaches on karst valleys and depressions. Pocket val-
leys are probably more common than the ve mapped in
Fig. 12, because many candidate features lack the amphi-
theater-like upper valley and/or the conduit inputs could
not be observed during eldwork.
Like Raven Run, the karst window shown in Fig. 12
is not a true window, but an exposed portion of a karst
capture (described in detail by Jerin & Phillips 2017).
3.4. Stability analysis
The network of interactions shown in Fig. 8 and Table 3 are
dynamically unstable by the Routh-Hurwitz criteria. The
addition of various combinations of the additional links
shown in Fig. 9 does not change this outcome. The results
are qualitatively insensitive – that is, given the interaction
matrix values greater than, less than, or equal to zero, vary-
ing the quantitative values will not produce a negative larg-
est eigenvalue (Lyapunov exponent). Thus the interactions
among surface and groundwater ow, ground surface and
conduit slope, conduit size, and karst relief are dynamically
unstable and prone to persistence and growth of perturba-
tions. Implications are discussed further below.
4. Discussion
The study sites include abundant evidence of both FK
and KF transitions, and do not indicate a landscape
Fig. 10. Fluvial-karst transition indicator landforms at the Raven Run study site. Base map is shaded relief from
~1.5 m (5 ft) resolution DEM.
eschweizerbart_xxx

Landform transitions in a uviokarst landscape 119
proceeding along a monotonic pathway toward either
karst or uvial domination. This is primarily due to three
factors. First, both uvial and karst processes are strongly
inuenced by base level and hydraulic gradients. The
geologically recent incision of the Kentucky River is the
single most important factor driving landscape evolution
in the study area, and strongly inuences and promotes
both karst and uvial processes. Second, karst and uvial
processes and subsurface and surface ows are strongly
interconnected and cannot change independently of each
other. Finally, the network of interconnections is dynami-
cally unstable, indicating that relatively small local dis-
turbances or variations can trigger transformations, thus
mitigating against either karst or uvial dominance. Thus
neither the primary driving factor of landscape change,
the karst-uvial interactions, or the system dynamics
promote landscape-scale dominance of either karstic or
uvial development.
The dynamical instability explains, at least phe-
nomenologically, the variety of KF and FK transitional
landforms. Local changes or perturbations tend to per-
sist and grow, so at any moment both FK and KF forms
are present, rather than a systematic landscape-scale
trend toward either a uvial or holokarst system. Karst
to uvial transitions are driven primarily by incision of
the Kentucky River and its tributaries, as channel inci-
sion and headward erosion encroach on dry karst valleys,
occasionally breach dolines, and form pocket valleys.
Shifts from uvial to karst dominance occur in both
incising and unincised areas. This in turn reects the rela-
tive independence of subsurface processes from surface
topographic slope gradients.
While the changing baselevel of the downcut-
ting Kentucky River and the incision of tributaries are
major drivers of both uvial and karst activity, the per-
vasive soluble bedrock makes karst processes common
throughout the Inner Bluegrass. Thus uvial and allu-
vial landforms are always liable to FK transformations,
particularly alluvial collapse dolines and karst capture
of stream ows. Slip-off slopes on river meander bend
Fig. 11. Fluvial-karst transition indicator landforms at the Polly’s Bend study site. The polygon outlined in the
heavier black line is a line of dolines (see text). Base map is shaded relief from ~1.5 m (5 ft) resolution DEM.
eschweizerbart_xxx

120 J.D. Phillips
interiors are hotspots for karstication created by long-
term uvial system evolution (Phillips & Walls 2004,
Phillips 2015) and upland paleochannels are also favora-
ble locations for doline development (Andrews 2004,
Phillips & Walls 2004).
Overall, a picture emerges of mutually reinforc-
ing uvial and karst activity that transcends both types
of activity responding to the same baselevel controls.
This implication is at least broadly consistent with ear-
lier studies. Karst processes such as cave collapse and
stream piracy were found by Woodside et al. (2015)
to strongly inuence stream longitudinal proles in a
uviokarst setting, for example. Conversely, Jaillet
et al. (2004) showed that uvial erosion opens “hydro-
geological windows” into karst systems, and forms
groundwater springs that stimulate new karstication.
Karst-uvial interaction can accelerate either process,
as in the formation of pocket valleys along valley side
slopes of incising streams in the study area, and the
stimulation of cave development by karst capture of
surface channels (Worthington 2015).
The uviokarst interaction model in this study, and
previous work supporting it, are based on the partition-
ing of ow and a competition metaphor for available
runoff. Other work also invokes similar metaphors for
uviokarst, such as White (2009), who portrayed the
evolution of Appalachian uviokarst in terms of compe-
tition between stream erosion, cave development, sur-
face denudation, and tectonics. While local divergence
into karst-rich or uvial-rich zones occurs on inner and
outer Kentucky River meander bends (Phillips et al.
2004, Phillips 2015) as a whole the Inner Bluegrass u-
viokarst region is similar to the Appalachian uviokarst
described by White (2009) in that endpoint situations
(karst-rich, channel-poor zones or limestone uplands
planated to local based levels vs. channel-rich, karst
poor zones or uvial dissection by deep canyons) are
rare. Rather, active, variable, and reversible uvial-karst
!""#$
%&'#()&*+
,)--.'#+/#
01
234
0
*.
1
#*+&.)$
5/4(.+#,)--.'
6)&*+#7031/7
8&.)49.1
1/-03.
:03(03;#*+&.)$
%&'#()&*+#,)--.'
5)-./49)33.-
<&/$#()&*+#
4)=+>&.
4)=+>&.
Fig. 12. Fluvial-karst transition indicator landforms at the Shaker Preserve study site. Base map is shaded relief
from ~1.5 m (5 ft) resolution DEM.
eschweizerbart_xxx

Landform transitions in a uviokarst landscape 121
interactions rather than progressive dominance by one
or the other are the rule.
5. Conclusions
Four major landform types representing transitions from
karst to uvially-dominated landforms and processes
were identied in the Inner Bluegrass region of central
Kentucky. These are sinking streams, dry karst valleys,
paleovalleys resulting from karst stream piracy, and
alluvial collapse dolines. Four other common landforms
indicate shifts from uvial to karst: stream incision into
dry karst valleys, doline breaching by surface runoff or
stream incision, karst pocket valleys, and karst windows.
These features were mapped at three study sites. All three
showed both KF and FK transitional landforms, with a
mean density of about three such landforms per km2.
Because of constraints on denitive identication of tran-
sitional forms, conservative criteria for inclusion, and the
fact that smaller features (such as stream swallets) were
not mapped, the actual occurrence of landforms indicat-
ing shifts in process dominance is higher.
A uviokarst structural network interaction model
representing interactions among surface channel ow,
subsurface conduit ow, surface topographic slope, sub-
surface gradients, and relief was analyzed. These inter-
actions are dynamically unstable under any plausible
conguration, indicating that local disturbances altering
hydrologic ow paths are likely to persist and grow over
time, facilitating transformations between surface ow
and uvial domination vs. subsurface ow and karst
domination. Thus, while local divergence into karst or
uvial domination is common, the uviokarst landscape
as a whole is characterized by variable and reversible
karst-uvial interactions, with no general long-term trend
toward a holokarst or a uvially-dominated landscape.
A competition metaphor based on ow partitioning
at the process scale or river vs. karst at the landscape
evolution scale is useful in understanding uviokarst
development. However, the functioning of the Kentucky
uviokarst studied here is better understood as mutual
reinforcement. Fluvial dissection can be intensied and
accelerated by the presence of karst features, and karsti-
cation is enhanced by stream incision.
References
Andrews, W.M., Jr. (2004): Geologic Controls on Plio-
Pleistocene Drainage Evolution of the Kentucky River in
Central Kentucky. – PhD dissertation, University of
Kentucky, Lexington.
Audra, P. & Palmer, A.N. (2015): Research frontiers in speleo-
genesis. Dominant processes, hydrogeological conditions
and resulting cave patterns. – Acta Carsolog. 44: 315–348.
Bahtijarevic, A. & Faivre, S. (2016): Quantitative comparative
geomorphological analysis of uvial and karst relief of
Florida. – Environ. Earth Sci. 75: article no. 428, doi:
10.1007/s12665–016–5397–8.
Bailly-Comte, V., Jourde, H. & Pistre, S. (2009): Concep tu-
alization and classication of groundwater-surface water
hydrodynamic interactions in karst watersheds: case of the
karst watershed of the Coulazou River (southern France). –
J. Hydrol. 376: 456–462.
Benac, C., Juracic, M., Maticec, D., Ruzic, I. & Pikelj, K.
(2013): Fluviokarst and classical karst: examples from the
Dinarics (Krk Island, northern Adriatic, Croatia). –
Geomorph. 184: 64–73.
Bocic, N., Pahernik, M. & Mihevic, A. (2015): Geomorpholo-
gical signicance of the palaeodrainage network on a karst
plateau: The Una-Korana plateau, Dinaric karst, Croatia. –
Geomorph. 247: 55–65.
Currens, J.C. (2012): Model ordinance for development on
karst terrain: Kentucky, USA. – Carbonates & Evaporites
27: 133–136.
Currens, J.C. & Graham, C.D.R. (1993): Flooding of Sinking
Creek, Garretts Spring karst drainage basin, Jessamine and
Woodford Counties, Kentucky, USA. – Environ. Geol. 22:
337–344.
Currens, J.C., Paylor, R.L. & Ray, J.A. (2003): Mapped Karst
Groundwater Basins in the Harrodsburg 30 X 60 Minute
Quadrangle (1:100,000 scale map). – Kentucky Geological
Survey, https://www.uky.edu/KGS/water/research/kaatlas
.htm.
Cvijic, J. (1918): L’hydrographie souterraine et l’evolution
morphologique du karst. – Revue de Géographie Alpine 6:
375–426.
Fabel, D., Henricksen, D. & Finlayson, B.L. (1996): Nickpoint
recession in karst terrains: an example from the Buchan
karst, southeastern Australia. – Earth Surf. Proc. Land. 21:
453–466.
Filipponi, M., Jeannin, P.Y. & Tacher, L. (2009): Evidence of
inception horizons in karst conduit networks. – Geomorph.
106: 86–99.
Filipponi, M., Jeannin, P.Y. & Tacher, L. (2010): Understanding
cave genesis along favorable bedding planes. The role of
primary rock permeability. – Z. Geomorph. N.F. Suppl. 54:
91–114.
Ford, D. (2007): Jovan Cvijic and the founding of karst geo-
morphology. – Environ. Geol. 51: 675–684.
Jaillet, S., Pons-Branchu, E., Brulhet, J. & Hamelin, B. (2004):
Karstication as geomorphological evidence of river inci-
sion: the karst of Cousance and the Marne valley (eastern
Paris Basin). – Terra Nova 16: 167–172.
Jennings, J.N., Brush, J.B., Nicoll, R.S. & Spate, A.P. (1976):
Karst stream self-capture London bridge, Burra Creek,
N.S.W. – Austral. Geogr. 4: 238–249.
Jerin, T. & Phillips, J.D. (2017): Local efciency in uvial sys-
tems: Lessons from Icicle Bend. – Geomorph. 282:
119–130.
Kaufman, G. & Braun, J. (2001): Modelling karst denudation
on a synthetic landscape. – Terra Nova 13: 313–320.
eschweizerbart_xxx

122 J.D. Phillips
Lavalle, P. (1967): Some aspects of linear karst depression
development in south central Kentucky. – Ann. Assoc. Am.
Geogr. 57: 49–71.
Lipar, M. & Ferk, M. (2015): Karst pocket valleys and their
implications on Pliocene-Quaternary hydrology and cli-
mate: Examples from Nullarbor Plain, southern Australia.
– Earth-Sci. Rev. 150: 1–13.
Mills, H.H. & Mills, R.T. (2001): Evolution of undercut slopes
on abandoned incised meanders in the eastern Highland
Rim of Tennessee, USA. – Geomorph. 28: 317–336.
Mylroie, J.E. & Mylroie, J.R. (1990): Meander Cutoff Caves
and Self Piracy: The Consequences of Meander Incision
into Soluble Rocks. – Nat. Speleol. Soc. Bull. 52: 33–44.
Ortega Becerril, J.A., Heydt, G.G. & Duran, J.J. (2010): The
role of sculpted forms and endokarstic active conduits in
the development of uviokarstic canyons. The Rio Puron
cave conduit (Spain). – In: Andreo, B. (ed.), Advances in
Research in Karst Media. Berlin, Springer, 387–392.
Phillips, J.D. (1999): Earth Surface Systems. Complexity,
Order, and Scale. – Oxford, UK, Basil Blackwell.
Phillips, J.D. (2015): Badass geomorphology. – Earth Surf.
Proc. Landf. 40: 22–33.
Phillips, J.D. & Lutz, J.D. (2008): Prole convexities in bed-
rock and alluvial streams. – Geomorph. 102: 554–566.
Phillips, J.D., Martin, L.L., Nordberg, V.G. & Andrews, W.A.
(2004): Divergent evolution in uviokarst landscapes of
central Kentucky. – Earth Surf. Proc. Landf. 29: 799–819.
Phillips, J.D. & Walls, M.D. (2004): Flow partitioning and
unstable divergence in uviokarst evolution in central
Kentucky. – Nonlin. Proc. Geophys. 11: 371–381.
Puccia, C.J. & Levins, R. (1985): Qualitative Modeling of
Complex Systems. – Harvard University Press, Cambridge,
MA.
Ray, J.A. & Blair, R.J. (2005): Large perennial springs of
Kentucky: Their identication, base ow, catchment, and
classication. – In: Beck, B.F. (ed.), Sinkholes and the
Engineering and Environmental Impacts of Karst.
Geotechnical Special Publication 144, American Society of
Civil Engineers, 410–422.
Reed, T.M., Mcfarland, J.T., Fryar, A.E., Fogle, A.W. & Taraba,
J.L. (2010): Sediment discharges during storm ow from
proximal urban and rural karst springs, central Kentucky,
USA. – J. Hydrol. 383: 280–290.
Roglic, J. (1964): Karst valleys in the Dinaric karst. – Erdkunde
18: 113–116.
Sauro, U. (2002): The Monti Berici: A peculiar type of karst in
the southern Alps. – Acta Carsolog. 31: 99–114.
Sawicki, L.R. (1909): Ein Beitrag zum geographischen Zyklus
im Karst. – Geograph. Z. 15: 185–204, 259–281.
Simms, M.J. (2004): Tortoises and hares: dissolution, erosion,
and isostasy in landscape evolution. – Earth Surf. Proc.
Landf. 39: 477–494.
Smart, C.C. (1988): A deductive model of karst evolution based
on hydrological probability. – Earth Surf. Proc. Landf. 13:
271–288.
Teller, J.T. & Goldthwait, R.P. (1991): The old Kentucky River;
a major tributary to the Teays River. – Geol. Soc. Am. Spec.
Pap. 258: 29–42.
Thrailkill, J., Sullivan, S.B. & Gouzie, D.R. (1991): Flow
parameters in a shallow conduit-ow carbonate aquifer,
Inner Bluegrass karst region, Kentucky, USA. – J. Hydrol.
129: 87–108.
Tiria, L. & Vijulie, I. (2013): Structural-tectonic controls and
geomorphology of the karst corridors in alpine limestone
ridges, southern Carpathians, Romania. – Geomorph. 197:
123–136.
Tüfeçki, K. & Sener, M. (2007): Evaluating of karstication in
the Mentese Region of southwest Turkey with GIS and
remote sensing applications. – Z. Geomorph. N.F. 51:
45–61.
White, W.B. (2009): The evolution of Appalachian uviokarst:
competition between stream erosion, cave development,
surface denudation and tectonic uplift. – J. Cave Karst Stud.
71: 159–167.
Williams, P.W. (1985): Subcutaneous hydrology and the devel-
opment of doline and cockpit karst. – Z. Geomorph. N.F.
29: 463–482.
Woodside, J., Peterson, E.W. & Dogwiler, T. (2015):
Longitudinal prole and sediment mobility as geomorphic
tools to interpret the history of a uviokarst stream
system. – Int. J. Speleol. 44: 197–206.
Worthington, S.R.H. (2005): Evolution of caves in response to
base-level lowering. – Cave & Karst Sci. 32: 3–12.
Worthington, S.R.H. (2015): Characteristics of channel net-
works in unconned carbonate aquifers. – Geol. Soc. Am.
Bull. 127: 759–769.
Manuscript received: February 7, 2017
Revised version accepted: May 16, 2017
eschweizerbart_xxx