Rivers follow the path of least resistance from their headwaters upstream to their terminal reaches downstream. The shape of river channels and the regions they drain are governed by the existing topography of a landscape and the geology that underlies it. Over their entire course, rivers widen, contract, straighten out, meander, and divide in response to changes in the landscape and as a result of the physical and chemical effects of flowing water on landscape features.
Distinctive patterns are acquired by stream networks in consequence of adjustment to geologic structure. In the early history of a network, and also when erosion is reactivated by earth movement or a fall in sea level, downcutting by trunk streams and extension of tributaries are most rapid on weak rocks, especially if these are impermeable, and along master joints and faults. Tributaries from those streams that cut and grow the fastest encroach on adjacent basins, eventually capturing parts of the competing networks therein. In this way, the principal valleys with their main drainage lines come to reflect the structural pattern.
Flat-lying sedimentary rocks devoid of faults and strong joints and the flat glacial deposits of the Pleistocene Epoch (from approximately 2,600,000 to 11,700 years ago) exert no structural control at all: this is reflected in branching networks. A variant pattern, in which trunk streams run subparallel, can occur on tilted strata. Rectangular patterns form where drainage lines are adjusted to sets of faults and marked joints that intersect at about right angles, as in some parts of ancient crustal blocks. The pattern is varied where the regional angle of structural intersection changes. Radial drainage is typical of volcanic cones, so long as they remain more or less intact. Erosion to the skeletal state often leaves the plug standing in high relief, ringed by concentric valleys developed in thick layers of ash.
Similarly, on structural domes where the rocks of the core vary in strength, valleys and master streams locate on weak outcrops in annular patterns. Centripetal patterns are produced where drainage converges on a single outlet or sink, as in some craters, eroded structural domes with weak cores, parts of some limestone country, and enclosed desert depressions. Trellis (or espalier) drainage patterns result from adjustment to tight regional folding in which the folds plunge. Denudation produces a zigzag pattern of outcrops, and adjustment to this pattern produces a stream net in which the trunks are aligned on weak rocks exposed along fold axes and small feeder streams run down the sides of ridges cut on the stronger formations. Deranged patterns, in which channels are interrupted by lakes and swamps, characterize areas of modest relief from which continental ice has recently disappeared. These patterns may be developed either on the irregular surface of a till sheet (heterogeneous glacial deposit) or on the ice-scoured expanse of a planated crystalline block. Where a till sheet has been molded into drumlins (inverted-spoon-shaped forms that have been molded by moving ice), the postglacial drainage can approach a rectangular pattern. In glaciated highland, postglacial streams can pass anomalously through gaps if the divides have been breached by ice, and sheet glaciation of lowland country necessarily involves major derangement of river networks near the ice front. At the other climatic extreme, organized networks in dry climates can be deranged by desiccation, which breaks down the existing continuity of a net. The largely linear systems of ephemeral lakes in inland Western Australia have been referred to this process.
Adjustment to bedrock structure can be lost if earth movement raises folds or moves faults across drainage lines without actually diverting them; streams that maintain their courses across the new structures are called antecedent. Adjustment is lost on a regional scale when the drainage cuts down through an unconformity into an under-mass with structures differing greatly from those of the cover: the drainage then becomes superimposed. Where the cover is simple in structure and provides a regional slope for trunk drainage, remnants of the original pattern may persist long after superimposition and the total destruction of the cover, providing the means to reconstruct the earlier network.
Great advances in the analysis of drainage nets were made by Robert E. Horton, an American hydraulic engineer who developed the fundamental concept of stream order: An unbranched headstream is designated as a first-order stream. Two unbranched headstreams unite to form a second-order stream; two second-order streams unite to form a third-order stream, and so on. Regardless of the entry of first-and second-order tributaries, a third-order stream will not pass into the fourth order until it is joined by another third-order confluent. Stream number is the total number of streams of a given order for a given drainage basin. The bifurcation ratio is the ratio of the number of streams in a given order to the number in the next higher order. By definition, the value of this ratio cannot fall below 2.0, but it can rise higher, because streams greater than first order can receive low-order tributaries without being promoted up the hierarchy. Some estimates for large continental extents give bifurcation ratios of 4.0 or more.
Although the number system given here, and nowadays in common use, differs from Horton’s original in the treatment of trunk streams, Horton’s laws of drainage composition still hold, namely:
1. Law of stream numbers: the numbers of streams of different orders in a given drainage basin tend closely to approximate an inverse geometric series in which the first term is unity and the ratio is the bifurcation ratio.
2. Law of stream lengths: the average lengths of streams of each of the different orders in a drainage basin tend closely to approximate a direct geometric series in which the first term is the average length of streams of the first order.
These laws are readily illustrated by plots of number and average length (on logarithmic scales) against order (on an arithmetic scale). The plotted points lie on, or close to, straight lines. The orderly relationships thus indicated are independent of network pattern. They demonstrate exponential relationships. Horton also concluded that stream slopes, expressed as tangents, decrease exponentially with increase in stream order. The systematic relationships identified by Horton are independent of network pattern: they greatly facilitate comparative studies, such as those of the influences of lithology and climate. Horton’s successors have extended analysis through a wide range of basin geometry, showing that stream width, mean discharge, and length of main stem can also be expressed as exponential functions of order, and drainage area and channel slope as power functions. Slope and discharge can in turn be expressed as power functions of width and drainage area, respectively. The exponential relationships expressed by network morphometry are particular examples of the working of fundamental growth laws. In this respect, they relate drainage-net analysis to network analysis and topology in general.
The functional relationships among various network characteristics, including the relationships between discharge on the one hand and drainage area, channel width, and length of main stem on the other, encourage the continued exploration of streamflow in relation to basin geometry. Attention has concentrated especially on peak flows, the forecasting of which is of practical importance. And because many basins are gaged either poorly or not at all, it would be advantageous to devise means of prediction that, while independent of gaging records, are yet accurate enough to be useful.
A general equation for discharge maxima states that peak discharges are (or tend to be) power functions of drainage area. Such a relationship holds good for maximum discharges of record, but conflicting results have been obtained by empirical studies of stream order, stream length, drainage density, basin size, basin shape, stream and basin slope, aspect, and relative and absolute height in relation to individual peak discharges in the shorter term. One reason is that not all these parameters have always been dealt with. In any event, peak discharge is also affected by channel characteristics, vegetation, land use, and lags induced by interception, detention, evaporation, infiltration, and storage. Although frequency–intensity–duration characteristics (and, in consequence, magnitude characteristics) of single storms have been determined for considerable land areas, the distribution of a given storm is unlikely to fit the location of a given drainage basin. In addition, the peak flow produced by a particular storm is much affected by antecedent conditions, seasonal and shorter term wetting and drying of the soil considerably influencing infiltration and overland flow. Nevertheless, one large study attained considerable success by considering rainfall intensity for a given duration and frequency, plus basin area, and main-channel slope expressed as the height–distance relationship of points 85 and 10 percent of stem length above the station for which predictions were made. For practical purposes, the telemetering of rainfall in a catchment, combined with the empirical determination of its response characteristics, appears effective in forecasting individual peak flows.
To empirical analysis of the morphometry of drainage networks has been added theoretical inquiry. Network plan geometry is specifically a form of topological mathematics. Horton’s two fundamental laws of drainage composition are instances of growth laws. They are witnessed in operation, especially when a new drainage network is developing; and, at the same time, probability statistics can be used to describe the array of events and forms produced.
Random-walk plotting, which involves the use of random numbers to lay out paths from a starting point, can produce networks that respond to analysis as do natural stream networks (i.e., length and number increase and decrease respectively, in exponential relationship to order, and length can be expressed as a power function of area). The exponential relationship between number and order signifies a constant bifurcation ratio throughout the network. A greater constancy in this respect would be expected from a randomly predicted network than from a natural network containing adventitious streams that join trunks of higher than one additional order. The exponential relationship between length and order in a random network follows from the assumption that the total area considered is drained to, and by, channels; the power relationship of length to area then also follows. The implication of the random-walk prediction of networks that obey the empirically derived laws of drainage composition is that natural networks correspond to, or closely approximate, the most probable states.
The geometry of a river changes over its length. Such changes can be noted by observing differences in cross section between one point in the river or another, or they can be noted by viewing selected lengths of the river from above. In addition, some rivers contain waterfalls, which serve as major disruptions in river flow.
Hydraulic geometry deals with variation in channel characteristics in relation to variations in discharge. Two sets of variations take place: variations at a particular cross section (at-a-station) and variations along the length of the stream (downstream variations). Characteristics responsive to analysis by hydraulic geometry include width (water-surface width), depth (mean water depth), velocity (mean velocity through the cross section), sediment (usually concentration or transport, or both, of suspended sediment), downstream slope, and channel friction.
Graphs of the values of channel characteristics against values of discharge usually display some scatter or departure from lines of best fit. One main cause is that values on a rising flood often differ from those on a falling flood, partly because of the reduction of flow resistance, and hence the increase in velocity, as sediment-concentration increases on the rising flood. Bed scour and bed fill are also related. Nevertheless, the variations for a given cross section can be expressed as functions of discharge, Q. For instance, width, depth, and velocity are related to discharge by the expressions: w ∝ = Qb, d ∝ = Qf, and v∝ = Qm, where w, d, v and b, f, m are numerical constants. The sum of the exponents b + f + m = 1, because of the basic relation—namely that Q = wdv.
Similar functions can be derived for downstream variations, but, for downstream comparisons to be possible, the observed values of discharge and of channel characteristics must be referred to selected frequencies of discharge. When data are plotted on graphs with logarithmic scales for each of two discharge frequencies at an upstream and a downstream station, the four points for each channel characteristic define a parallelogram, whereby the hydraulic geometry of the stream is defined in respect of that characteristic. The values of exponents in the power equations differ considerably from one river to another: those shown here are theoretical optimum values. One common cause of difference is that many gaging stations are located where some channel characteristics are controlled, whether naturally as by rock outcrops or artificially as by bridge abutments. Constraints on variation in width, for instance, are mainly offset by increased variation in depth.
Analyses of downstream variation in channel slope with discharge commonly reveal contrasts between field results and the theoretical optima. The discrepancy is probably due in considerable part to the fact that the channel slope can vary in concert with channel efficiency, including channel habit, channel size, and channel form. Many past discussions of stream slope are invalidated by their restriction to the two dimensions of height and distance. In any event, the slopes of many natural channels are influenced by some combination of earth movement, change in baselevel, glacial erosion, glacial deposition, and change of discharge and load characteristics that result from change of climate. Consequently, although natural profiles from stream source to stream mouth suggest a tendency toward a smooth concave-upward form, many actually are irregular. Even without a change of baselevel, degradational tendency, or discharge, a change in channel sinuosity can produce a significant change of channel slope.
A marked downstream lessening of slope does not imply a decrease in velocity at a given frequency of discharge. Reduction of slope is accompanied, and offset, by an increase in channel efficiency mainly because of an increase in size. The lower Amazon, with a slope of less than 7.6 cm (3 inches) per 1.6 km (1 mile), flows faster at the bankfull stage than many mountain streams, at 2.4 metres (about 8 feet) per second. According to the assumptions made, an optimal velocity equation in hydraulic geometry can predict a slight increase, constancy, or a slight decrease in velocity downstream, for a given frequency of discharge. On the Mississippi, velocity at mean discharge (not a set frequency) increases downstream. Velocity at the overbank stages of the five-year and 50-year floods is constant downstream. Constant downstream velocity may well be first attained at the bank-full stage. The fact that relationships are highly disturbed at and near waterfalls and other major breaks of slope (the Paraná just below the site of the former Guaíra Falls, for instance, ran at 9 to 14 metres [30 to 46 feet] per second) has no bearing on the principles of hydraulic geometry, which apply essentially to streams in adjustable channels.
The interrelationships and adjustments among width, depth, width–depth ratio, suspended-sediment concentration, sediment transport, deposition, eddy viscosity, bed roughness, bank roughness, channel roughness, and channel slope in their relation to discharge, both at-a-station and in the downstream direction, plus the tendency at many sections on many streams for variation to occur about some modal value, all encourage the conception of rivers as equilibrium systems. The designation quasi-equilibrium systems is usually used, because not all variances can be simultaneously minimized, and minimization of some variances (e.g., of water-surface slope) can only be secured at the expense of maximizing others (e.g., channel depth).
Distinctive patterns in the plan geometry of streams correspond to distinctive combinations of cross-sectional form, calibre of bed load, downstream slope, and in some cases cross-valley slope, tendency to cut or fill, or position within the system. The full range of pattern has not been identified: it includes straight, meandering, braided, reticulate, anabranching, distributary, and irregular patterns. Although individual patterns are given separate names, the total range constitutes a continuum.
Straight channels, mainly unstable, develop along the lines of faults and master joints, on steep slopes where rills closely follow the surface gradient, and in some delta outlets. Flume experiments show that straight channels of uniform cross section rapidly develop pool-and-riffle sequences. Pools are spaced at about five bed widths. Lateral shift of alternate pools toward alternate sides produces sinuous channels, and spacing of pools on each side of the channel is thus five to seven bed widths. This relation holds in natural meandering streams.
Meandering channels are single channels that are sinuous in plan, but there is no criterion, except an arbitrary one, of the degree of sinuousity required before a channel is called meandering. The spacing of bends is controlled by flow resistance, which reaches a minimum when the radius of the bend is between two and three times the width of the bed. Accordingly, meander wavelength, the distance between two successive bends on the same side—or four-bend radii—tends to concentrate between eight and 12 bed widths, although variation both within and beyond this range seems to be related to variations in the cross-sectional form of the channel. Because bed width is related to discharge, meander wavelength also is related to discharge.
Meandering channels are equilibrium features that represent the most probable channel plan geometry, where single channels deviate from straightness. This deviation, and channel division in general, is related in part to the cohesiveness of channel banks and the abundance and bulk of midstream bars. When single channels are maintained, however, the meandering form is most efficient because it minimizes variance in water-surface slope, in angle of deflection of the current, and in the work done by the river in turning. This least-work property of meander bends is readily illustrated by the trace, identical with that of stream meanders, adopted by a bent band of spring steel. Meander plan geometry is simply describable by a sine function of the relative distance along the channel bend. The least-work and minimum-variance properties of the plan geometry, however, are secured only at the expense of maximizing the variance in depth. The longitudinal profile of the bed of a meandering stream includes pools at (or slightly downstream of) the extremities of bends and riffles at the inflections between bends. Increased tightness of bend, expressed by reduction in radius and increase in total angle of deflection, is accompanied by increased depth of pool. Where riffles are built of fragments larger than sand size, they behave as kinematic waves. In other words, the speed of transport of material through a given riffle decreases as the spacing of surface fragments decreases, and the total rate of transport attains a maximum where the spacing is about two particle lengths. Numerous sand-bed streams in dry regions, however, fail to develop pool-and-riffle sequences, maintaining approximately uniform cross sections even at channel bends.
Irregularities in meanders developed in alluvium relate primarily to uneven resistance, which is often a function of varying grain size. Variations in total sinuosity are probably in the main caused by adjustments of channel slope. The process of cutoff (short-circuiting of individual meanders) is favoured not only by the erosion of outer channel banks and by the tendency of meander trains to sweep down the valley but also by the stacking of meanders upstream of obstacles and by increases of sinuosity that accompany slope reduction.
Meandering streams that cut deeply into bedrock form entrenched meanders, the terminology of which is highly confused. It seems probable that, in actuality, the sole existing type of entrenched meanders is the ingrown type, where undercut slopes (river cliffs) on the outsides of bends oppose slip-off slopes (meander lobes) on the insides. For reasons not yet understood, lateral enlargement of ingrown meanders seems habitually to outpace downstream sweep, although the trimming of the upstream sides of lobes, and occasional cutoff, are well known. Many existing trains of ingrown meanders belong to valleys rather than to streams, relating to the traces of former rivers of greater discharge. Reconstruction of the original traces indicates approximate straightness at plateau level, as opposed to the inheritance of the ingrown loops from some former high-level floodplain.
In a broader context, meander phenomena cannot be understood as requiring cohesive banks of the kind usual in rivers. Meanders, with geometry comparable to that of rivers, have been recognized in the oceanic Gulf Stream and in the jet streams of the upper atmosphere. In this way, stream meanders are classed with wave phenomena in general.
Braided channels are subdivided at low-water stages by multiple midstream bars of sand or gravel. At high water, many or all bars are submerged, although continuous downcutting or fixation by plants, or both, plus the trapping of sediment may enable some bars to remain above water. A single meandering channel may convert to braiding where one or more bars are constructed, as downstream of a tight bend where coarse material is brought up from the pool bottom. Each of the subdivided channels is less efficient, being smaller than the original single channel. If its inefficiency is compensated by an increase in slope (i.e., by downcutting), the bar dries out and becomes vegetated and stabilized. However, many rivers that are largely or wholly braided along their length owe their condition to something more than local accidents. The braided condition involves weak banks, a very high width–depth ratio, powerful shear on the streambed (implied by the width-depth ratio), and mobile bed material. Thus, braided streams are typically encountered near the edges of land ice, where valleys are being filled with incoherent coarse sediment, and also on outwash plains, as the Canterbury Plains of South Island, New Zealand. Width–depth ratios can exceed 1,000:1. Studies on terraced outwash plains demonstrate that braided streams can readily excavate their valley floors—in other words, they are by no means solely an inevitable response to valley filling.
New Zealand’s Tasman Glacier terminus is a braided river. These form when there are weak banks, high width-to-depth ratio, strong shear on the streambed, and mobile bed material. Steve Casimiro/The Image Bank/Getty Images
Distributary patterns, whether on alluvial fans or deltas, pose few problems. A delta pass that lengthens is liable to lateral breaching, whereas continued deposition, on deltas and on fans, raises the channel bed and promotes sideways spill down the least gradient. The branching rivers of inland eastern Australia, flowing across basin fills that range from thin sedimentary plains to thick fluvial accumulations, have affinities with deltaic distributaries even though their patterns are only radial in part. A branch may run for tens of kilometres before joining a trunk stream, whether its own or another.
Waterfalls, sometimes called cataracts, arise from an abrupt steepening of a river channel that causes the flow of water to drop vertically, or nearly so. Waterfalls of small height and lesser steepness are called cascades, a term often applied to a series of small falls along a river. Still gentler reaches of rivers that nonetheless exhibit turbulent flow and white water in response to a local increase in channel gradient are rapids.
Waterfalls are characterized by great erosive power. The rapidity of erosion depends on the height of a given waterfall, its volume of flow, the type and structure of the rocks involved, and other factors. In some cases the site of the waterfall migrates upstream by headward erosion of the cliff or scarp, whereas in others erosion tends to act downward to bevel the entire reach of river containing the falls. With the passage of time, by either or both of these means, the inescapable tendency of streams is to eliminate so gross a discordance of longitudinal profile as a waterfall. The energy of all rivers is directed toward the achievement of a relatively smooth, concave-upward, longitudinal profile; this is a common equilibrium, or adjusted condition, in nature.
Even in the absence of entrained rock debris that serves as an erosive tool of rivers, it is intuitively obvious that the energy available for erosion at the base of a waterfall is great. Indeed, one of the characteristic features associated with waterfalls of any great magnitude—with respect to volume of flow as well as to height—is the presence of a plunge pool, a basin that is scoured out of the river channel directly beneath the falling water. In some instances the depth of a plunge pool may nearly equal the height of the cliff causing the falls. Its depth depends not only on the erosive power of the falls, however, but also on the amount of time during which the falls remain at a particular place. The channel of the Niagara River below Horseshoe Falls, for example, contains a series of plunge pools, each of which represents a stillstand, or period of temporary stability, during the general upriver migration of the waterfall. The significance of this profile will be discussed in the following text, but in general it may be said that the fate of most waterfalls is their eventual transformation to rapids as a result of their own erosive energy.
The lack of permanence as a landscape feature is, in fact, the hallmark of all waterfalls. Many well-known occurrences such as the Niagara Falls came into existence as recently as 11,700 years ago, when the last of the great ice sheets retreated from middle latitudes. The oldest falls originated during the Neogene Period (23 million to 2.6 million years ago), when episodes of uplift raised the great plateaus and escarpments of Africa and South America. Examples of waterfalls attributable to such pre-Pleistocene uplift (which occurred more than 2.6 million years ago) include Kalambo Falls, near Lake Tanganyika; Tugela Falls, in South Africa; Tisisat Falls, at the headwaters of the Blue Nile on the Ethiopian Plateau; and Angel Falls, in Venezuela.
Paulo Afonso Falls on the São Francisco River, Alagoas, Brazil. Antonio Gusmao—TYBA/Agencia Fotografica Ltda.
Available data suggest that the falls of greatest height are seldom those of greatest water discharge. Many falls in excess of 300 metres (about 980 feet) exhibit but modest flow, and, in some cases, only a perpetual mist occurs near their bases. By way of contrast, the Khone Falls of the Mekong River in southern Laos, drop only 22 metres (72 feet), but the average discharge of this cataract is about 11,330 cubic metres (400,000 cubic feet) per second. In general, considering height and volume of flow jointly, it is understandable that Victoria, Niagara, and Paulo Afonso, among others, have each been proclaimed “the world’s greatest falls” by various explorers and authorities.
The distribution of waterfalls is not uniform, and large parts of the world are free of any notable occurrence. This is not surprising in view of the relatively large proportion of the Earth’s land area that consists of deserts and semiarid areas, which are understandably devoid of modern falls on climatic grounds. Ice-covered polar regions and relatively unbroken, low-lying plains and plateaus also are unfavourable sites of development.
Considered on a global basis, waterfalls tend to occur in three principal kinds of areas: (1) along the margins of high plateaus or the great fractures that dissect them; (2) along fall lines, which mark a zone between resistant crystalline rocks of continental interiors and weaker sedimentary formations of coastal regions; and (3) in high mountain areas, particularly those that were subjected to glaciation in the recent past.
Notable falls along high plateaus include the world’s highest, Angel Falls of the Churún River, Venezuela, with a drop of 979 metres (3,200 feet) and overall relief of more than 1,100 metres (3,600 feet); Tugela Falls, issuing from the Great Escarpment, South Africa, which is 948 metres (3,100 feet) in height; Victoria Falls (108 metres [350 feet]) on the Zimbabwe–Zambia border; and Kalambo Falls (427 metres [1,400 feet]) on the Tanzania–Zambia border. The volume of flow at Victoria Falls is relatively large, approximately 1,080 cubic metres (38,000 cubic feet) per second, but Guaíra Falls, a series of falls that until their submergence by the waters of Itaipú Dam in 1982 totaled 114 metres (375 feet) along the Paraná River, Brazil–Paraguay, had the largest known average discharge—13,300 cubic metres (about 470,000 cubic feet) per second. During flood stages, however, even this figure is exceeded at some falls along the Orange River and elsewhere. Angel Falls, Iguaçu Falls (82 metres [270 feet]), in Brazil, and several others occur along the margins of high plateaus, east of the Andes, between Venezuela and Argentina.
Iguaçu Falls extend for 1.7 miles (2.7 km) and, during the rainy season, may rise as much as of 450,000 cubic feet (12,750 cubic metres) per second. Shutterstock.com
Waterfalls that occur along fall lines are in some cases relatively indistinguishable from plateau examples—the Aughrabies Falls (146 metres [480 feet]), for instance, which occur where the Orange River leaves resistant crystalline rocks of the plateau in southern Africa. The typical fall-line example, however, occurs at the junction of the crystalline rocks of the Appalachian Mountains and the sedimentary coastal plain along the eastern United States. A number of major cities, including Philadelphia, Baltimore, and Washington, D.C., are a geographic consequence of the existence of falls along this line or zone because they present barriers to further inland navigation. In England there is an analogous example with respect to the line of towns including Cambridge that borders the Fens. The most spectacular fall-line waterfalls, however, include Churchill (formerly Grand) Falls, Labrador, Canada (75 metres [about 250 feet]); Jog Falls (Gersoppa Falls), Karnātaka, India (253 metres [830 feet]); and Paulo Afonso Falls, Brazil (84 metres [275 feet]).
The last category, mountainous and formerly glaciated regions, include such well-known waterfalls as Yosemite Falls, California (739 metres [about 2,400 feet]), with a three-section drop; Yellowstone Falls, Wyoming (94 metres [about 300 feet]), with a two-section drop; Sutherland Falls, South Island, New Zealand (580 metres [1,900 feet]); and Krimmler Waterfall, Austria (380 metres [about 1,250 feet]). Other falls of considerable height or volume of flow occur elsewhere in mountainous and formerly glaciated regions—namely, in the Alps, the Sierra Nevada and northern Rocky Mountains of North America, and South Island, New Zealand. The ice-free parts of Iceland and the fjord (drowned-valley) region of Norway also should be cited. Both areas contain numerous falls by reason of suitable topography and climate. Australia also has a few falls, notably the Wollomombi, in the Great Dividing Range, New South Wales (482 metres [about 1,600 feet]).
The several types of waterfalls that occur in nature may be classified according to a variety of schemes. One of the simplest of these is based on principal region of occurrence—high plateaus, fall lines, and formerly glaciated mountains, as discussed earlier. More meaningful, however, is an alternate, threefold classification system that places more emphasis on the specific ways in which geologic and physiographic conditions produce and affect waterfalls. Thus, falls can be categorized as: (1) those attributable to natural discordance of river profiles, whether caused by faulting (vertical movements of the Earth’s crust), glaciation, or other processes; (2) those attributable to differential erosion, which occurs whenever weak and resistant rocks are juxtaposed in some way; and (3) those attributable to constructional processes that create barriers and dams, over which water must fall. These three basic types will be discussed in turn.
In one sense, all falls must be attributable to a discordance of river profile by their very definition. This category is here arbitrarily restricted, however, to exclude profile breaks that are caused by differential erosion and constructional processes. Remaining are waterfalls along fault scarps, uplifted plateaus and cliffs, glacial features of several kinds, karst topography—the caves and cave systems produced by solution of carbonate rocks—and falls that result from the issuance of springs from canyon walls high above valley floors.
The enormous rigid plates that make up the outer shell of the Earth continually move relative to one another, resulting in seafloor spreading, continental drift, and mountain building. These large-scale motions cause a buildup of strain within the rocks of the crust at some depth below the surface. Ultimately, the rocks must yield or shift in order to release this strain, and, when they suddenly do so, an earthquake results. Commonly, there will be some visible evidence of this sudden release at the Earth’s surface, perhaps manifested by the creation of a cliff or series of cliffs along a line or zone. The sloping surfaces that form the cliff fronts are called fault scarps. The vertical movements that produce fault scarps seldom amount to more than about 3 metres (about 10 feet) during an individual earthquake. Repeated faulting along the same line or zone, however, can produce scarps that are thousands of metres in height in relatively brief periods of geologic time. Waterfalls occur where the faults cross established drainage systems. The ultimate height of such falls depends not only on the total height of uplift but also on the rate of downcutting by the affected rivers. Rates of uplift tend to exceed rates of downcutting considerably in those parts of the world where uplift is ongoing today. Hence, it is normal for high waterfalls to exist due to uplift in many areas. In addition, some plateaus are produced by broader, regional uplifts that are relatively continuous and are not associated with earthquakes. The heights attained are nevertheless comparable after suitable time intervals. Major rift (fracture) systems of continental or subcontinental scale, some sea cliffs, and other features of this nature also are attributable to some form of faulting. All of them provide suitable sites for waterfall development.
The processes of glaciation have served this same end. Mountain ranges that formerly were glaciated contain falls at the outlets of cirques, bowl-shaped depressions in the headwaters of drainage areas that were formed by the accumulation of ice and its erosive action on the underlying bedrock. In addition, waterfalls are most common where hanging valleys occur. Such valleys generally form when glacier ice deeply erodes a main or trunk valley, leaving tributary valleys literally hanging far above the main valley floor. After the glaciers have melted and withdrawn, streams from such tributary valleys must fall in order to join the main valley drainage system below. Hanging valleys also can occur in response to faulting and in some other non-glacial situations: the chalk cliffs of England, for example, where small streams cannot cut downward with sufficient rapidity to keep pace with backwearing of the cliffs by marine erosion.
Other features that may result from glaciation include glacial potholes and glacial steps. The former are thought to originate principally as a result of the plastic flow of ice at the base of a glacier, which permits the gouging of semicylindrical holes in the bedrock beneath the path of flow. The holes or depressions are subsequently enlarged and deepened by meltwater runoff that is heavily laden with gravels, and they have become the sites of modern cascades in many instances.
Waterfalls may occur at the outlets of cirques, which were formed in once-glaciated ranges such as Montana’s Pioneer Mountains. Dr. Marli Miller/Visuals Unlimited/Getty Images
The steps (or glacial stairway, as this feature is sometimes called) consist of treads and risers on a relatively giant scale that have been produced by the passage of ice over bedrock, particularly when alternating rock properties or joints offer differential resistance to the flow of ice. Again, the establishment of runoff after wastage of the ice has occurred will lead to a series of waterfalls or cascades at the site of each riser in the stairway.
Most spectacular among glacial features, however, are the overdeepened valleys along formerly glaciated coasts, as in Norway. These fjords are intimately associated with falls because the valley walls typically are both high and steep and because hanging valleys are ubiquitous.
Like the potholes previously mentioned, the solution of limestones and other carbonate rocks leads to the formation of pits, sinks, caves, and interconnected systems of caverns, which together are termed karst topography. Terrain of this kind commonly contains water in many of the included passages in the form of standing pools, streams, and, where discontinuities of cavern levels occur, waterfalls. There are a few parts of the world where karst topography and its associated drainage are prominent features of the landscape, but, on the whole, falls attributable to cave-forming processes are not numerous. Springs that issue from canyon walls high above main valley floors are in the same category. Most of these artesian (free-flowing) systems result from the same type of solution phenomenon along joints and fractures that produce caves in carbonate rocks.
Rocks differ markedly with regard to their resistance to erosion by running water. Although no quantitative scales to express this difference have been developed, widespread agreement exists on certain generalities. Metamorphic rocks (those that are formed from preexisting rocks under the action of high temperatures and pressures), for example, are commonly more durable than are sedimentary rocks, and great differences can exist even among the latter because of a significant amount of variation in the degree of cementation and kinds of rock structure present in them. Thus, a quartz-rich sandstone whose constituent grains are cemented by silica tends to be much more resistant than a fissile shale, the clay-rich layers of which tend to split and separate. And the blocky character of some carbonate rocks (limestones and dolomites) and extrusive igneous rocks (formed by the cooling of lava flows) tends to enhance their resistance to fluvial erosion, notwithstanding their relatively low resistance to solution.
Regardless of the intrinsic toughness of any rock type, however, lengthy periods of weathering or the presence of intricate fracture patterns will render it easily erodible. There are, in fact, a veritable legion of factors that influence rock resistance to erosion, and it is for this reason that generalities must be invoked. Suffice it to say that some rocks are weak whereas others are strong and that waterfalls are promoted where these occur in certain geologic arrangements.
There are three such arrangements that are common in nature: (1) horizontal or nearly horizontal strata in which rocks of greater resistance overlie weaker rocks, forming a protective cap rock; (2) inclined strata involving beds or layers of alternating resistance; and (3) various kinds of non-sedimentary rock arrangements in which dikes or veins of hard crystalline rocks are juxtaposed with weaker rocks. In each of these cases the weaker rocks are eroded more readily and more rapidly by running water, and the harder, resistant rocks, as a consequence, stand higher and are “falls makers.” In the special case of the cap-rock arrangement, waterfalls migrate upriver because the protective upper layers break off as the weaker supporting strata are eroded from beneath. Niagara Falls is the most notable example involving sedimentary rocks (a blocky dolomite cap overlies a series of less-resistant shales and sandstones). More commonly, a lava flow caps erodible strata.
There are four principal constructional processes that can lead to the creation of dams or barriers and, hence, to the formation of waterfalls. These processes are (1) precipitation of calcium carbonate from solution; (2) disruption of drainage by lava flows or the deposition of volcanic ash and other pyroclastic sediments; (3) ice damming and the construction of moraines, or ridgelike sedimentary deposits left at the sites of former glaciers; and (4) the deposition of landslide and avalanche debris.
The first of these, carbonate precipitation, can accumulate to considerable dimensions as spring deposits of travertine or calcareous tufa, often in a series of terraces. Where these ultimately block avenues of normal runoff, waterfalls result. The water in limestone caves also is rich in calcium carbonate, and where ponds occur in the path of small subterranean streams there is preferential precipitation at the spillage rims. The barriers that are raised are self-perpetuating, can attain heights of about 15 metres (about 50 feet) under certain circumstances, and have been called rimstone dams and falls.
Volcanic activity, principally in the form of basaltic lava flows, is related to waterfall development in many parts of the world. The flows compose the bulk of such great plateau areas as the Columbia River region of the United States and the Deccan Plateau in India and often serve as cap rock. The association of falls with plateaus in general and with cap-rock arrangements was noted previously, but, in addition, some falls result from drainage diversion and the ponding of streams and rivers by lava dams. This has occurred in some parts of New Zealand, Iceland, and Hawaii and, in general, in regions where volcanic activity is a prominent aspect of the landscape.
Ice dams can produce similar effects. One of the most interesting examples is Dry Falls, a “fossil waterfall” in the Columbia River Plateau, Washington, which formed in late Pleistocene time. A large ice sheet blocked and diverted the then-westward-flowing Columbia River and formed a vast glacial lake. The lake drained to the south when permitted to do so by periodically occurring ice dams, and torrents of water were released during these breakouts. The water flowed through the Grand Coulee channel and eroded a canyon nearly 300 metres (about 1,000 feet) deep. Dry Falls, about 120 metres (about 400 feet) high and 5 km wide (about 3 miles), occurs along this flow path. The Columbia River has reestablished its path to the sea since the disappearance of the ice sheet, so the falls are dry today.
The magnitudes of flow that must have occurred during the Pleistocene, however, can be appreciated from data on some of the great glacier outburst floods (jøkulhlaups) of modern history. The breaching of an ice dam at Grímsvötn, Ice., in 1922, for example, released about 7.1 cubic km (1.7 cubic miles) of water, and the discharge attained a value of 57,000 cubic metres (about 2 million cubic feet) per second.
There are other depositional features that may pond and dam streams, notably glacial moraines—which attain heights as great as 250 metres (about 800 feet) in the formerly glaciated valleys of the Alps—and landslides, avalanches, and other downslope movements of earth materials into valleys. The associated falls tend to be rather ephemeral, however, because all such unconsolidated material is cut through relatively swiftly, and smooth stream gradients are reestablished. The damming action of lava flows and glacier ice is far more important in nature. The lava flows consist of more durable material, and ice damming leads to outburst floods and great attendant erosion.
With the passage of time a particular waterfall must either migrate upstream, as in the case of a cap-rock falls, or serve as the locus for general downcutting along the reach of river containing the falls. In either case, the process depends on the height of the falls, the volume of flow, and the nature and arrangement of the rocks involved. Any discussion of waterfall development requires knowledge of these three factors and, more importantly, knowledge of the former locations and configurations of any particular waterfall under consideration. If the changes of location through time are lacking, then rates of waterfall recession are basically indeterminate.
The available data on the recession of the Horseshoe Falls of the Niagara River are little short of astonishing in comparison to the general paucity of such information elsewhere. Instrumental surveys of the configuration and position of Horseshoe Falls were made in 1842, 1875, 1886, 1890, 1905, 1927, and 1950. Still earlier delineations of position were provided by visual observations as long ago as 1678. For this reason, general waterfall development must be considered in terms of the Horseshoe Falls example. It should be noted, however, that the recession rates pertaining to this cap-rock-type falls are not necessarily average rates for all falls of this kind. They certainly do not apply to non-cap-rock falls in crystalline rocks, for example, where much slower rates generally prevail.
The average rate of recession of any falls can be determined from knowledge of the total upstream distance of migration and the time period during which the migration occurred. In the case of Horseshoe Falls, the total distance involved is about 12 km (7.5 miles), and retreat of the falls has been accomplished in approximately 12,500 years, since the disappearance of the most recent ice sheet from the area. The average rate of recession is therefore about 1 metre (3 feet) per year. The several instrumental surveys, however, suggest that a rate of 1.2 metres (4 feet) per year occurred during the 1842–75 period and 2 metres (about 7 feet) per year during the 1875-1905 period.
By way of comparison, the average recession rate for the American Falls, which occur downstream and to one side of Horseshoe Falls because of branching by the Niagara River, is only 8 cm (3.1 inches) per year. And, in a comparable vein, upstream migration of the Gullfoss in Iceland during the last 10,000 years is estimated to have occurred at an average recession rate of 25 cm (10 inches) per year. This is, again, a far slower rate of falls recession than has occurred at Horseshoe Falls.
To some extent the various recession rates are related to differential resistance of the rocks to erosion. Indeed, the discrepancy between the 1842–75 and 1875–1905 rates for Horseshoe Falls have been attributed in the past not only to possible surveying errors but also to the relative abundance of joints (fractures) in different parts of the dolomite cap rock. One study of Horseshoe Falls suggests, however, that another factor is of still greater importance—namely, the configuration of the crest of the falls and the relative stability of differing kinds of configurations.
