4. Gypsum and anhydrite dissolution proceed even in contact with static water, hut dissolution rates increase rapidly with increasing flow velocities. As a turbulent flow regime sets in, dissolution rates are boosted, probably by an order of magnitude.

5. The presence of other salts, such as sodium chloride, in solution considerably increases gypsum dissolution rates, but influences anhydrite dissolution rates even more drastically. This effect contributes particularly to speleogenesis in deep-seated environments.

4. Structural and hydrostratigraphical pre-requisites of speleogenesis

Initial permeabilities of common aquifers (e.g. some clastic rocks) in deepseated settings is normally greater then that of karstifiable units, particularly sulphates, before the onset of speleogenesis. Gypsum beds commonly act initially as separating beds. Groundwater comes into contact with gypsum either from adjacent aquifer formations, or via minor beds of other lithologies, such as marls or dolomites, that are intercalated with the gypsum beds. This view is somewhat similar to that suggested by Lowe (1992) in his "Inception Horizon Hypothesis", although the intercalations in gypsum sequences do not commonly generate specific dissolution chemistry, as is suggested for carbonate sequences. However, such horizons in gypsum formations may locally determine chemical processes that maintain the gypsum dissolution potential, due to removal of sulphates from solution (sulphate reduction, dedolomitization).

Late diagenesis (catagenesis) and/or tectonism impose fissure permeability upon gypsum formations, which then begin to play the major role in determining initial flow paths through the gypsiferous sequence. Gradient fields in confined conditions generally promote vertical, crossformational circulation, although the presence of water-conducting intercalations and a specific fissure distribution may support a significant lateral component of speleogenetic development. While major tectonic fissures normally cut through the entire thickness of a bed, lithogenetic fissures tend to form largely independent networks that are confined within certain textural intervals. Such networks are characterised by a good lateral connectivity, but are connected vertically only at a relatively small number of discrete points. This provides the structural prerequisites for the development of multi-storey maze caves (Klimchouk, 1992, 1994; Klimchouk et al., 1995; see also Chapter 1.1).

A theoretical approach to the understanding of the propagation of early dissolutional ope­nings in fissures has been developed by Palmer (1984, 1991), based on the combined considera­tion of mass-balance relationships, hydraulic equations for laminar flow, and chemical mass-transfer. Dissolutional enlargement of partings in soluble materials, gypsum in particular, has been investigated theoretically and experimentally by James & Lupton (1978) and James (1992).

It is generally believed that most proto-caves propagate through fissures where the connected apertures are small: limits between <10µm to 1mm have been suggested (Ford & Williams, 1989). Recent study by Groves & Howard (1994) suggests that a minimum aperture of 100µm is required for conduit development. Seepage through them is very slow and laminar. The rate and configuration of dissolutional widening depends primarily on discharge through the fissure and the change in solute concentration along its length (Palmer, 1984, 1991; James, 1992). Due to gypsum's fast dissolution kinetics, solute concentration increases rapidly to about the 90% saturation level, so that the penetration distance, L90, is quite short (for details of the 1,90 concept see Weyl, 1958; White, 1977; Ford & Williams, 1989). Fissures enlarge at their inlets, remaining almost unchanged downstream resulting in a tapered geometry. The mode of dissolutional enlargement of the fissure, or through a sequence of interconnected fissures, will change only when a breakthrough of the penetration distance to the output boundary occurs, so that a flow path enlarges enough to permit an increase in flow velocity and penetration of significantly undersaturated water beyond the exit.

It is not yet clear whether the main breakthrough mechanism is the propagation of a taper from the input end, or slow dissolutional enlargement along the entire fissure length by water close to saturation. Shapes of enlarged fissures depend upon rock solubility, the dissolution rate constant, and flow velocity. Other variables being equal, fissure enlargement at the input end will be more tapered in gypsum than in limestone due to faster dissolution kinetics. Lower initial velocities produce sharp tapers, whereas higher velocities promote a more gradual enlargement (Fig.1). The sensitivity of the enlargement configuration to flow velocities, and hence to differen­ces in the initial fissure width, should he more pronounced in fissures in gypsum than those in limestone.

	Fig. 1. Effect of flow velocity on enlarge-ment configuration (After James & Lupton, 1978).

Under sluggish flow conditions in gypsum it seems possible that the tapering is so localised that the first part of a fissure may reach human-penetrable sizes, while its downstream parts remain almost unchanged and very narrow. The common occurrence of "blind" ends of dissolutional passages in the labyrinthine caves of the Western Ukraine, with narrow guiding fissures recognizable along their apex, supports such an assumption. This introduces some potential confusion of the morphometric criteria that allow distinction between the inception and develop­ment stages in gypsum speleogenesis.

The duration of the initiation phase (until breakthrough is achieved) depends mainly on the length of fissure paths and on initial flow rates; the latter in turn depends strongly upon the width of the initial aperture (it is proportional to the cube of the width). This dependence is the primary factor responsible for the selective enlargement of openings (Fig. 2). The effect is more pronounced in gypsum than in limestone due to diffusion control of dissolution kinetics, which causes stronger dependence of the dissolution rate constant upon flow velocity. It is quite possible that duration of the initiation phase of conduits propagating through narrow fissures in gypsum is comparable with, if not longer, than that in carbonates, given that the boundary hydraulic condi­tions are equal. This apparent paradox is caused by dissolution kinetics being faster in gypsum than in carbonates.

	Fig. 2. Penetration distances or progress of the dissolution front for - L90 in massive gypsum, calculated for initial fissure widths ranging from 0.21 - 1.0mm. Time elapsed since initiation is in years. The hydraulic gra­dient is 0.2 and water temperature is 10°C Inset: the form of the dissolution taper into the fissure that is obtained from theoretical calculations such as these (After James & Lupton, 1978, as adapted by Ford & Williams, 1989).

At this point the distinction between phreatic and confined conditions becomes crucial to speleogenetic development.

Phreatic conditions. The controls of conduit development under phreatic conditions are best described by Palmer (1984, 1991). Given a substantial hydraulic gradient, the amount of flow through a fissure path is determined by its width. During the initiation stage water emerging at the output boundary is almost saturated, so that the rate at which any route enlarges depends upon the amount of flow rather than upon dissolution kinetics (discharge-controlled stage). For a given conduit, enlargement rates increase slowly, as the discharge through the path is severely restricted by the narrower downstream parts. When breakthrough of L90 has occurred, enlargement accelerates dramatically and promotes a further large increase of discharge, so that a "runaway" condition develops. Enlargement rates in gypsum are further accelerated as turbulent conditions set in due to the mass transport control of dissolution kinetics. Those conduits that achieve early breakthrough are able to increase their discharge either by capturing water from neigh­bouring conduits or by extending their primary catchment. This situation emphasizes the importance of initial differences in hydraulic efficiency between fissures, and also explains the competitive early development of conduits.

Major discrete tectonic fissures intersecting the whole gypsum bed represent a different development scheme (Fig.3-B). Their initial resistivity to flow is potentially low, if their width is large, so conduits grow fairly rapidly as breakthrough conditions might he present from the very beginning. However, enlargement rates do not accelerate significantly, as further increase of flow velocity is inhibited by the reduction in the head gradient between the upper and lower aquifers. As soon as an appreciable through space is created, much of the flow can reach the top of the gypsum without touching the gypsum walls. Under generally slow flow conditions, downward convection circulation cells develop, giving rise to an "inverted tapering" of the conduit shape (see more about the natural convection effect in the sub-chapter below). With further growth of through voids the head gradient between the adjacent aquifers tends towards zero, and circula­tion is driven largely by natural convection. This mechanism appears to he responsible for the continuing lateral growth of large voids in gypsum, and for triggering major breakdowns and the formation of vertical through structures (VTS) in coverheds (see Chapter 1.10 for details).

Dissolution also occurs along the contacts with the adjacent aquifers. In general, conduits initiate as tapers at inputs (Fig.3-C), however the actual mechanisms and shapes of the initiation and development depend also on the nature of the initial flow paths. These may follow bedding plane paths between solid rocks, be along the interface between solid gypsum and granular or porous aquifer material, or along water-conducting fissures in contact with the solid gypsum, and so on. Consequently, dissolved gypsum can be removed down-gradient along a beddingrelated flow path, or in a direction normal to the lithological interface by diffusion or convection, then ouflowing with the regional flow. The latter situation is most favourable for conduit initiation and development, as concentrations decrease at the interface, hence increasing the overall gypsum dissolution rate. Flow paths are guided by the arrangement of connected initial apertures along bedding planes, or by channels determined by intersections of fissures in adjacent insoluble rock with the bedding contact, or by more transmissive zones in adjacent granular material. When an aquifer underlies the gypsum, and some tapered space is created, further enlargement can he pro­moted by natural convection circulation. It is likely that some large voids can develop in this way.

5.3. Development at the water table and in the vadose zone

With the onset of the entrenched karst stage, vadose conditions become increasingly predominant, with the continued possible existence of water table and phreatic zones in the lower parts of massifs. Rapid enlargement of artesian and/or phreatic conduits occurs at the water table, parti­cularly if annual fluctuations of major surface river levels cause periodic backflooding into a cave. In more stable conditions, such as in the interiors of watershed massifs, extensive horizontal notching may develop, promoted by water density stratification (see the sub-chapter below). In the vadose zone, cave development is concentrated along vertical percolation paths and free stream courses, hut is very active locally. Hydrochemical data from different regions suggest that groundwaters in the vadose zone never attain saturation with respect to gypsum. More dispersed dissolutional enlargement may occur due to the action of condensation waters, hut this effect tends to he localized in certain zones (see sub-chapter 5.4).

Because dissolution always leads to solvent density increase, gravitational separation of water, and natural convection due to this effect, are inherently involved in, and affect, the cave development process. The effect may he significant in limestones (Curl, 1966) although is much more pronounced in gypsum (Kempe, 1972, 1975) and in salts (Frumkin, 1994) due to the higher solubility of these rocks. A recent overview and further elaboration of the issue has been provided by Klimchouk (1997h).

Water density difference effects generally become notable at the development stage, when substantial spaces are created by forced convection dissolution. In artesian settings the effects may also contribute to conduit initiation. When continuous or periodic recharge of fresh water occurs, dissolution of the rock sets up density gradients, which cause gravitational separation (stratification) of water and drive natural convection circulation. The phenomenon may operate at the local scale (e.g. in a cave lake) or at the scale of an aquifer (e.g. in artesian aquifers or across the water table zone).

Maze caves constitute about 21% of the 197 largest recorded gypsum caves in the world, hut their proportion increases to 41% among caves over 1000m long, and to 54% of gypsum caves over 2000m long. Considering the length of surveyed passages rather than the number of caves, just the five greatest mazes of the Western Ukraine comprise far more then half of all known gypsum passages in the world.

Palmer (1975) distinguished two major situations favourable for maze cave development: (1) where dispersed, aggressive recharge takes place uniformly into all available fissures in a soluble rock unit, entering from an overlying insoluble hut permeable formation, and (2) where floodwater recharge causes temporal variations in discharge and head in an evolving system to he so great that no fixed passage configuration is allowed to stabilize with respect to flow (stream caves). Klimchouk (1992, 1994) suggested a third type of a genetic environment that favours maze cave development: upward recharge into a karst unit from the basal aquifer in a multi-storey artesian system. No unambiguous example of a cave in gypsum that could fit the first of Palmer's types is known to the present author, although this setting probably could produce mazes in gypsum. The second environment does not produce maze caves in gypsum as, due to the high gypsum solubility and fast solution kinetics, stream caves rapidly adjust their sizes to the highest possible discharge, so that no constrained floods occur. It appears that the great majority of maze caves in gypsum belong to the type of artesian mazes. The relevant hydrogeological settings are discussed in the Klimchouk' works cited above and in Chapter 1.6; the theoretical substantiation of the ori­gin and development mechanisms is outlined in section 5.2.

For typical rectilinear maze caves to develop, not only the artesian flow architecture favouring cross-formational upward hydraulic communication is required, but also a fairly uniform, though not necessarily high, initial fissure permeability within a gypsum bed. Otherwise discrete voids will tend to form rather than maze caves.

Artesian maze caves commonly display a multi-storey structure. Storeys are not related to cycles of stability/uplift (base level control) hut are controlled stratigraphically: they develop at the lower and upper contacts of the gypsum, and inside the gypsum, within some intervals (beds) where initial fissures form networks that are well connected laterally but not well connected vertically. It is particularly common for lithogenetic fissures to form largely independent networks confined within certain horizons, with only local connections to adjacent fissured horizons (Klimchouk et al., 1995; see also Chapter 1.1).

The most outstanding examples of mazes in gypsum are the great caves of the Western Ukraine (see Chapter 11.9 for characteristics and maps). They are also known in the Belomorsko­Kulojsky, Povolzhje, Pre-Urals and some Siberian regions of Russia, in Germany and in Spain. Caves of this type are undoubtedly very common in many other regions of intrastratal karsts, but favourable conditions are required for them to survive uplift and denudation to become accessible and he explored.

CALAFORRA, J.M. 1996. Contribution al conocimiento de la karstologia de yesos. PhD Thesis, Universidad de Granada. 350 pp.

Curl, R.L. 1966. Cave conduit enlargement by natural convection. Cave Notes, 8 (1). 4-8. DUBLJANSKY, V.N. 1970. Condensation of moisture in the fissurekarst collectors of the Mountainous Crimea, Carpathians and Pridnestrovskaya Podolia. Doklady AN Uk.SSR, ser.B, 1, Kiev. 14-17. (in Russian).

DUBLJANSKY, V.N. & SOTZKOVA, L.M. 1982. Microclimate of karst caves. Zemlevedenie, v.XIY, Moscow. 79-91. (in Russian).

JAMES, A.N. & LUPTON, A.R.R. 1978. Gypsum and anhydrite in foundations of hydraulic structu­res. Geotechnique, 28. 249-272.

JENNINGS, J.N. 1971. Karst. The M.I.T.Press. 252 pp.

KEMPE, S. 1972. Cave genesis in gypsum with particular reference to underwater conditions. Cave Science 49. 1-6.

KLIMCHOUK, A. 1997b. Speleogenetic effects of water density differences. 12th Internat. Congress of Speleology, La Chaux-de-Fonds, Switzerland.

KLIMCHOUK, A.B., AKSEM, S.D., SHESTOPALOV, V.N. & RUD'KO, G.I. 1988. The regime study of gypsum karst activity in the Western Ukraine. Kiev: Inst. Geol. Nauk. 55 pp. (in Russian).

  KLIMCHOUK, A.B., ANDREJCHOUK, V.N. & TURCHINOV, 1.1. 1995. Structural pre-requisites of speleogenesis in gypsum in the Western Ukraine. Kiev. Ukrainian Spelcol. Assoc. 104 p. LOWE, D.J. 1992. The origin of limestone caverns: an inception horizon hypothesis. Ph.D. Thesis, Manchester Metropolitan University. 511 pp.

PALMER, A.N. 1975. The origin of maze caves. NSS Bulletin 37 (3). 56-76.

WEYL, P.K. 1958. Solution kinetics of calcite. J. Geol., 66. 163-176.