Modern and ancient fluvial megafans in the foreland basin system of
Transcription
Modern and ancient fluvial megafans in the foreland basin system of
Basin Research (2001) 13, 43±63 Modern and ancient ¯uvial megafans in the foreland basin system of the central Andes, southern Bolivia: implications for drainage network evolution in foldthrust belts B. K. Horton*.and P. G. DeCelles². *Department of Geology and Geophysics, Louisiana State University, Baton Rouge, LA 70803, USA ²Department of Geosciences, University of Arizona, Tucson, AZ 85721, USA ABSTRACT Fluvial megafans chronicle the evolution of large mountainous drainage networks, providing a record of erosional denudation in adjacent mountain belts. An actualistic investigation of the development of ¯uvial megafans is presented here by comparing active ¯uvial megafans in the proximal foreland basin of the central Andes to Tertiary foreland-basin deposits exposed in the interior of the mountain belt. Modern ¯uvial megafans of the Chaco Plain of southern Bolivia are large (5800±22 600 km2), fan-shaped masses of dominantly sand and mud deposited by major transverse rivers (Rio Grande, Rio Parapeti, and Rio Pilcomayo) emanating from the central Andes. The rivers exit the mountain belt and debouch onto the low-relief Chaco Plain at ®xed points along the mountain front. On each ¯uvial megafan, the presently active channel is straight in plan view and dominated by deposition of mid-channel and bank-attached sand bars. Overbank areas are characterized by crevasse-splay and paludal deposition with minor soil development. However, overbank areas also contain numerous relicts of recently abandoned divergent channels, suggesting a long-term distributary drainage pattern and frequent channel avulsions. The position of the primary channel on each megafan is highly unstable over short time scales. Fluvial megafans of the Chaco Plain provide a modern analogue for a coarsening-upward, >2-km-thick succession of Tertiary strata exposed along the Camargo syncline in the Eastern Cordillera of the central Andean fold-thrust belt, about 200 km west of the modern megafans. Lithofacies of the mid-Tertiary Camargo Formation include: (1) large channel and small channel deposits interpreted, respectively, as the main river stem on the proximal megafan and distributary channels on the distal megafan; and (2) crevasse-splay, paludal and palaeosol deposits attributed to sedimentation in overbank areas. A reversal in palaeocurrents in the lowermost Camargo succession and an overall upward coarsening and thickening trend are best explained by progradation of a ¯uvial megafan during eastward advance of the fold-thrust belt. In addition, the present-day drainage network in this area of the Eastern Cordillera is focused into a single outlet point that coincides with the location of the coarsest and thickest strata of the Camargo succession. Thus, the modern drainage network may be inherited from an ancestral mid-Tertiary drainage network. Persistence and expansion of Andean drainage networks provides the basis for a geometric model of the evolution of drainage networks in advancing fold-thrust belts and the origin and development of ¯uvial megafans. The model suggests that ¯uvial megafans may only develop once a drainage network has reached a particular size, roughly 104 km2 ± a value based on a review of active ¯uvial megafans that would be affected by the tectonic, climatic and geomorphologic processes operating in a given mountain belt. Furthermore, once a drainage network has achieved this critical size, the river may have suf®cient stream power to prove relatively insensitive to possible geometric changes imparted by growing frontal structures in the fold-thrust belt. Correspondence: B. K. Horton, Department of Geology and Geophysics, Louisiana State University, Baton Rouge, LA 70803, USA. # 2001 Blackwell Science Ltd 43 B. K. Horton and P. G. DeCelles INTRODUCTION Fluvial megafans constitute volumetrically signi®cant depositional elements of sedimentary basins adjacent to mountain belts. A ¯uvial megafan is a large (103x105 km2), fan-shaped (in plan view) mass of clastic sediment deposited by a laterally mobile river system that emanates from the outlet point of a large mountainous drainage network (Gohain & Parkash, 1990.; DeCelles & Cavazza, 1999.). Modern ¯uvial megafans have been recognized in nonmarine foreland basin systems at the outlets of major rivers that drain fold-thrust belts, particularly in the Himalayas (Geddes, 1960.; Sinha & Friend, 1994.; Gupta, 1997.; DeCelles & Cavazza, 1999.) and northern Andes (RaÈsaÈnen et al., 1992.). Although ¯uvial megafans are similar to sediment gravity ¯owdominated and stream-dominated alluvial fans in terms of their piedmont setting, planform geometry and sedimentation related to expansion of ¯ow downslope of a drainage outlet, ¯uvial megafans are distinguished by their greater size (alluvial fans rarely exceed 250 km2), lower slope, presence of ¯oodplain areas and absence of sediment gravity ¯ows (see also Blair & McPherson, 1994., for a discussion of alluvial-fan classi®cation). The term `terminal fan' is commonly used for a large, distributary ¯uvial system in which surface water in®ltrates and evaporates before it can ¯ow out of the system (Friend, 1978.; Nichols, 1987.; Kelly & Olsen, 1993.; Newell et al., 1999.). A ¯uvial megafan therefore may be considered `terminal' in cases where ¯uvial channels run dry before reaching bodies of water downstream. Although ¯uvial megafans are clearly related to the emergence of a large mountain river onto a low-relief alluvial plain, their stratigraphic evolution in nonmarine foreland basins may also be critically dependent on variables such as sediment ¯ux, water discharge, drainage catchment size, catchment lithology and subsidence rate, factors that are ultimately controlled by tectonic, climatic and geomorphic processes (e.g. Burbank & Raynolds, 1988.; Heller et al., 1988.; Flemings & Jordan, 1989.; Fraser & DeCelles, 1992.; Tucker & Slingerland, 1996.; Gupta, 1997.; Schlunegger et al., 1997., 1998.; Robinson & Slingerland, 1998.). In an effort to explore both the modern and the ancient dynamics of these systems, this investigation utilizes an actualistic approach to sedimentological analysis of ¯uvial megafans. We ®rst document the geomorphic features and depositional processes of active ¯uvial megafans in the central Andean foreland basin system of southern Bolivia (. Fig. 1.). This provides a modern analogue for Tertiary foreland basin deposits exposed within the central Andean fold-thrust belt to the west. Analysis of these deposits in the Eastern Cordillera demonstrates that ¯uvial megafans are identi®able in the stratigraphic record and may be dominant elements in the stratigraphic Fig. 1. Map of the central Andes and modern foreland basin system showing ¯uvial megafans (Grande, Parapeti, Pilcomayo) and their source drainage basins, main rivers, major geological structures and national borders. Geological zones include the Altiplano (AP), Eastern Cordillera (EC), Subandean Zone (SZ), Chaco Plain and Beni Plain. AP±EC boundary is de®ned by the western (hinterland) drainage divide. EC±SZ boundary is shown by long dashed line. Barbed lines represent east-directed thrust faults present at the mountain front and at depth in the westernmost Chaco Plain. Approximate axis of forebulge is drawn after Horton & DeCelles (1997.). The polygon in the western Chaco Plain shows the location of a composite Landsat image of ¯uvial megafans (Fig. 2.). Several rivers that deposit ¯uvial megafans bifurcate and/or terminate in swampland regions (RV: Rio Viejo area; IZ: Izozog swamp; PA: PatinÄo swamp). Rectangle in Eastern Cordillera locates Camargo syncline study region (Fig. 4.). 44 # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia evolution of nonmarine foreland basin systems. We anticipate that recognition of ¯uvial megafans in the geological record will improve attempts to reconstruct the large-scale geomorphology of ancient mountain belts, possibly providing estimates on the size and geometry of palaeo-drainage networks. G EO LO G I C A L B A C K G R O U N D Mountain building in the Andes is associated with subduction of the oceanic Nazca plate beneath western South America. During the Cenozoic, the central Andes have grown progressively eastward and continental crust has thickened by regional horizontal shortening (Isacks, 1988.; Gubbels et al., 1993.; Jordan et al., 1997.). Crustal thickening and eastward propagation of the central Andes has developed a topographic load on South American lithosphere, producing a retroarc foreland basin system (Jordan & Alonso, 1987.; Jordan, 1995.; Horton & DeCelles, 1997.). Synorogenic erosion and transport of clastic sediment to this evolving basin system has produced a substantial record of Cenozoic nonmarine sedimentation, including young deposits ®lling the modern foreland region and ancient deposits exposed within the central Andes (e.g. Jordan & Alonso, 1987.; Beer et al., 1990.; Jordan, 1995.; Horton & DeCelles, 1997.; Sempere et al., 1997.; Horton, 1998.). The eastern slope of the Bolivian Andes consists of a belt of generally east-vergent folds and thrust faults in the Eastern Cordillera and Subandean Zone (Fig. 1.). This fold-thrust belt is ¯anked on the east by the Chaco Plain (Fig. 1.), a low-relief, low-elevation region interpreted as the aggradational surface of the proximal foreland basin system (the wedge-top and foredeep depozones; Horton & DeCelles, 1997.). During most of Tertiary time, however, the foreland basin system was located to the west within the area now occupied by the fold-thrust belt (Sempere et al., 1990., 1997.; Horton, 1998.). Eastward advance of the fold-thrust belt and foreland basin system throughout Tertiary time has uplifted and exposed older basin ®ll within the interior of the mountain belt (Horton, 1998.; DeCelles & Horton, 1999.; Horton et al., in press).. In this paper, we discuss modern ¯uvial megafans of the Chaco Plain in order to provide a depositional analogue for Tertiary strata exposed in the Camargo syncline of the Eastern Cordillera. FLUVIAL MEGAFANS IN THE MODERN FORELAND BASIN Geomorphology and depositional processes Fluvial megafans of the Chaco Plain constitute the most extensive depositional systems of the proximal sector (wedge-top and foredeep depozones) of the modern central Andean foreland basin system. The apex of each megafan is located along the Subandean topographic front at the outlet point of an extensive mountainous catchment draining the # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 fold-thrust belt. Three ¯uvial megafans exist along a 400-km-long segment of the north-striking topographic front of the central Andes between 18uS and 22uS (Fig. 1.). They are, from north to south, the Rio Grande (or Guapay), Rio Parapeti and Rio Pilcomayo megafans (.Fig. 2.). Additional ¯uvial megafans exist further south in the Andean foreland of Argentina, including the Rio Bermejo (23.3uS), Rio Salado (25.3uS), Rio Huaco (30.2uS) and Rio Jachal (30.7uS) megafans (Iriondo, 1984., 1993.; Damanti, 1993.). The Andean megafans (. Table 1.) are comparable to modern Himalayan megafans (Geddes, 1960.; Sinha & Friend, 1994.; DeCelles & Cavazza, 1999.) in terms of channel gradient (0.03±0.12u), drainage catchment area (< 104x105 km2), and depositional area (< 103x105 km2) (Table 1.). In the intermegafan areas between adjacent ¯uvial megafans, the westernmost Chaco Plain is composed of stream-dominated alluvial fans fed by small catchments in the Subandean Zone (Fig. 2.). These alluvial fans (Table 1.) are distinguishable from the ¯uvial megafans on the basis of their steeper slopes (1±4u), smaller depositional areas (< 5 km2) and smaller catchment areas (< 200 km2). Most alluvial fans are fed by small catchments con®ned to the easternmost anticlinal ridge (Fig. 2.; Table 1.), indicating that these fans are consequent features developed only after initial structural growth of the present-day Subandean topographic front. In contrast, the rivers depositing the ¯uvial megafans exhibit relatively straight courses across the Subandean Zone (Fig. 2.), suggesting they are antecedent features that pre-date motion on the frontal structures of the fold-thrust belt, structures that have been active since about 5±10 Ma (Gubbels et al., 1993.). Upon entry onto the Chaco Plain, rivers that deposit ¯uvial megafans are no longer con®ned by lateral topographic barriers and are free to migrate over a large area. Although each ¯uvial megafan exhibits a single, relatively straight, sand-dominated, active channel in which the majority of surface water and sediment is presently transported (Fig. 2.), evidence outlined below indicates that the location of this channel is highly unstable over relatively short time frames. Drainage patterns on the Chaco Plain can be attributed to large-scale abandonment of distributary streams in the lower segments of the ¯uvial megafans. The Rio Grande and Rio Parapeti each make broad < 90u turns to the north over a downstream distance of 90 km, starting from the apex of their respective megafans (Figs 1 and 2.). The Rio Pilcomayo makes a sharp 40u turn to the south-southeast 60 km downstream from its megafan apex (Figs 1 and 2.). In all three cases, these river in¯ections are ¯anked to the east by a number of separate small catchments that have headwater reaches limited to the present ¯oodplain (i.e. they are not connected to the active primary channel) (Fig. 2.). Individual streams within these small catchments ¯ow approximately eastward and are roughly parallel or radiating in plan view. On the Rio Parapeti megafan (. Fig. 3.), there are numerous large channels similar in size to, but south of, the presently active channel. These straight channels commonly bifurcate downstream into 45 B. K. Horton and P. G. DeCelles Fig. 2. Composite Landsat Multispectral Scanner (MSS) image and interpretive line drawing of ¯uvial megafans of the western Chaco Plain of Bolivia. Fluvial megafans include active channels (black), abandoned or ephemeral megafan channels (densely stippled pattern) and overbank areas (lightly stippled pattern). Intermegafan areas contain alluvial fans (generally too small to depict at this scale) and aeolian dunes. Table 1.contains information on alluvial fans directly north of the Rio Parapeti drainage outlet. Downslope of the ¯uvial megafans, the alluvial plain consists of swampland regions, aeolian dune ®elds and small channels with headwaters con®ned to the present-day alluvial plain. See Fig. 1 for location. 46 # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia Table 1. Geomorphic characteristics of selected modern subaerial fans. River±Fan Location Andes ¯uvial megafans Grande (Guapay) Chaco Plain, Bolivia Parapeti Chaco Plain, Bolivia Pilcomayo Chaco Plain, Bolivia-Argentina Bermejo Chaco Plain, Argentina Catchment area (km2) Fan area (km2) Slope < 70 000 < 8000 81 300 < 12 600 < 5800 < 22 600 16 000 < 10 000 (?) Salado Huaco Chaco Plain, Argentina < 30 000 Bermejo basin, 7100 Argentina Jachal Bermejo basin, 27 700 Argentina Pastaza-Maranon Amazon basin, Peru ? Central Andes alluvial fans in intermegafan area between Rio Grande Saipuru Chaco Plain < 118 Sanja Honda*. Chaco Plain < 2.8 Taputa Chaco Plain < 30.2 Tapulami Chaco Plain < 22.2 Acae*. Chaco Plain < 8.1 Tacuarembo*. Chaco Plain < 9.9 Huirapucuti*. Chaco Plain < 5.2 Sereque*. Chaco Plain < 3.6 Piriti*. Chaco Plain < 3.8 Capihuazuti*. Chaco Plain < 4.6 Ovai*. Chaco Plain < 8.1 Charagua Chaco Plain < 168 Himalayan ¯uvial megafans (Nepal and India) Kosi Gangetic Plain 59 000 Gandak Gangetic Plain 45 000 Tista Gangetic Plain < 12 000 Other Okavango south-central Africa (Botswana) 180 000 Reference < 10 000 (?) 700 IGM (1972, 1. 998.); this study IGM (1980a,.b, 1. 998.); this study Guyot et al. (1990.); IGM (1981, 1. 998.); this study < 0.06u (?) Iriondo (1984).; Guyot et al. (1990).; OAS (1971). < 0.06u (?) Iriondo (1984).; OAS (1971). ? Damanti (1993.) 1400 ? 60,000±70 000 and Rio Parapeti < 5.0 < 0.5 < 3.0 < 1.2 < 0.9 < 0.8 < 0.5 < 0.5 < 0.8 < 0.6 < 0.8 < 2.8 0.03±0.10u 0.10±0.12u 0.04±0.06u Damanti (1993.) < 0.06u (?) RaÈsaÈnen et al. (1992). megafans (Bolivia) < 1±2u IGM (1974.); this study < 4u IGM (1974.); this study < 2u IGM (1974.); this study < 2u IGM (1974.); this study < 3u IGM (1974.); this study < 3u IGM (1974.); this study < 3u IGM (1974.); this study < 4u IGM (1974.); this study < 4u IGM (1974.); this study < 3u IGM (1974.); this study < 3u IGM (1974.); this study < 1±2u IGM (1974.); this study < 16 500 < 17 500 < 16 000 0.06u 0.03u 0.04u Sinha & Friend (1994.) Sinha & Friend (1994.) Geddes (1960).; DeCelles & Cavazza (1999). 18 000 0.02u Stanistreet & McCarthy (1993.) *Alluvial-fan catchment area is limited to frontal (easternmost) anticlinal ridge of Subandean Zone smaller channels. Channels to the north, closer to the active Rio Parapeti, are better de®ned than the southern channels due to a greater amount of aeolian dune activity to the south (Figs 2 and 3.). For the Rio Pilcomayo system, separate channels east of the main river de¯ection exhibit a distributary ¯ow pattern that continues well beyond the downstream limits of the ¯uvial megafan (Fig. 2.). The separate drainage networks feeding these channels expand dramatically downstream such that the mainstem rivers (the Rio Melo, Rio Apa, Riacho Yacare Norte and Rio Verde) are widely dispersed by the time they reach the south-¯owing Paraguay River about 500 km downstream (Fig. 1.). Based on the Chaco Plain channel geometries described above, we suggest that the ¯anking stream catchments were once the site of distributary ¯ow on the megafan surface. We therefore attribute the overall planform morphologies of the megafan rivers and these smaller catchments to recent abandonment of distributary channel systems on the lower, downslope ¯anks of ¯uvial # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 megafans. For the two northern systems (Rio Grande and Rio Parapeti) in which the active channels ¯ow abruptly northward (Figs 1 and 2.), these abandonment episodes could represent recent, large-scale stream capture by the Amazon drainage system. Historical records for the past few hundred years also attest to high channel mobility on the central Andean megafans. For example, rapid lateral shifts in river course from several to tens of kilometres have occurred in the Rio Pilcomayo and Rio Bermejo systems since the late 1800s, mostly associated with ¯ooding during rainy seasons (Baker, 1978.; Iriondo, 1984., 1993.). Changes in river course may also be related to documented stream-capture events that affect reaches several hundred kilometres long and may rapidly change the magnitude of water discharge and sediment ¯ux (Baker, 1978.; Iriondo, 1984.). Collectively, these geomorphologic and historical observations suggest that megafan construction is dominated by shortterm deposition in one location followed by avulsion to an 47 B. K. Horton and P. G. DeCelles Fig. 3. Near-vertical, space-shuttle photograph of Rio Parapeti ¯uvial megafan. North is to the right; north± south distance across ®eld of view is < 70 km. Abandoned, divergent channels are better de®ned near the main active channel than they are to the south where linear (north-trending) aeolian dunes obscure the inactive channels. adjacent, lower-elevation, overbank area. Therefore, the locus of active sedimentation is constantly shifting about the megafan through channel avulsion, preventing simple lateral migration of channels by point-bar accretion (i.e. meandering rivers with scroll topography) and minimizing the development of relief between the channel and adjacent overbank areas. Downstream transport Fluvial megafans generally display downslope decreases in grain size and channel size compatible with a proximal-to-distal change from ¯ow con®ned to a proximal feeder channel to more distributed ¯ow, either as smaller distributary channels or crevasse splays, on distal parts of the megafan (e.g. Friend, 1978.; Nichols, 1987.; Kelly & Olsen, 1993.; Newell et al., 1999.). For the Chaco Plain ¯uvial megafans, the sizes of both active and abandoned channels generally decrease downstream. The mainstem rivers, approximately several hundred to 2000 m wide on the ¯uvial megafans, become abruptly narrower (generally less than a few hundred metres) at the downstream terminations of each megafan (IGM, 1998.). The downstream distance from the megafan apex to the megafan margin is 150 km, 130 km and 250 km for the Rio Grande, Rio Parapeti and Rio Pilcomayo, respectively. Whereas proximal megafan channels are presumably 10 m deep or greater, the distal channels are commonly less than a few metres deep (Iriondo, 1993.). Grain size also decreases downstream, from sand- and limited gravelsized detritus of the upper reaches of the megafans to the 48 sand- and mud-sized detritus in lower parts of the megafans. Historical data and stream gauge data indicate that the rivers depositing the ¯uvial megafans carry substantial suspended silt and clay, whereas coarser sediment is largely transported as bedload in mid-channel and bank-attached bars developed during individual ¯oods (Iriondo, 1984., 1997.). For the Pilcomayo, the annual sediment discharge is approximately 8 r 107 tons; 90% of this material is transported in suspension (Guyot et al., 1990.). Whereas most river systems of Himalayan ¯uvial megafans merge downstream into a major trunk river (the Ganges River), the processes operating downstream of the central Andean ¯uvial megafans are drastically different. Little surface water derived from the central Andes ¯ows out of the Chaco Plain. The Rio Pilcomayo enters the broad (10 000 km2) PatinÄo swamp directly downstream of its megafan (Fig. 1.; Iriondo, 1984.). Most water in®ltrates here or is lost by evapotranspiration, effectively isolating the proximal segment of the river from a downslope continuation of the Pilcomayo that ultimately ¯ows into the Parana River, the second largest river in South America (and world's ®fth largest drainage area). Similarly, the Rio Parapeti enters a swampland region at the terminus of its megafan, ¯owing into the 6800-km2 Izozog swamp (Figs 1 and 2.) (Iriondo, 1993.). In the Rio Grande, ¯ow may be moderately stalled in the Rio Viejo area 260±400 km downstream of the megafan apex (Fig. 1.). In this area, the Rio Grande bifurcates into a number of smaller channels and much of the ¯oodplain is subject to periodic inundation before ¯owing into the Rio # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia Mamore and ultimately the Amazon River (the world's largest drainage area). These observations suggest that a limited solid load is currently transported beyond the megafans of the Chaco Plain, a condition analogous to the Okavango terminal fan (Stanistreet & McCarthy, 1993.). Therefore, the vast majority of clastic sediment derived from the central Andean fold-thrust belt between about 18u and 22uS (an area of roughly 160 000 km2) may be deposited on the Chaco Plain, primarily on ¯uvial megafans, rather than transported to the major river deltas at the Atlantic margin. Such proximal storage of orogenic sediment within the wedge-top and foredeep depozones has implications for estimates of sediment budgets for the Amazon and Parana Rivers. STRATIGRAPHIC FRAMEWORK OF THE CAMARGO SYNCLINE About 200 km west of the modern Chaco Plain ¯uvial megafans, a nearly 3-km-thick section of Upper Cretaceous to Tertiary strata is spectacularly exposed in the Camargo syncline (.Fig. 4.). A persistent, north-trending ridge about 65 km long with up to 1500 m of relief exposes clastic deposits of the Tertiary Camargo Formation. These strata form the bulk of the stratigraphic section in the syncline and are the focus of this investigation. Although the Cretaceous±Tertiary stratigraphic interval generally represents foreland-basin conditions (Sempere et al., 1997.), it can be further divided into separate genetic intervals attributable to sedimentation in different depozones within the evolving foreland basin system (. Fig. 5.; DeCelles & Horton, 1999.). The succession begins with the Coniacian±Campanian Aroi®lla and Chaunaca Formations, a 100-m-thick interval of red ¯uvial mudstones and minor sandstones that unconformably overlies Ordovician phyllite. This interval is overlain by light grey carbonate strata of the Maastrichtian-early Palaeocene El Molino Formation, a 150-m-thick interval of lacustrine and marginal marine deposits. Upsection, the mid Palaeocene Santa Lucia Formation contains 80 m of cross-strati®ed, red ¯uvial sandstones and ¯oodplain siltstones that are overlain by the Impora Formation, a distinctive, 60-m-thick interval of paludal and ¯oodplain deposits dominated by nodular, grey, calcareous palaeosols. Cross-strati®ed, quartzose, white ¯uvial sandstones and interbedded red ¯oodplain siltstones comprise the overlying, 190-m-thick Cayara Formation. Above the Cayara, the remaining section is composed of the Camargo Formation, a well-de®ned coarsening- and thickening-upward succession composed of a lower interval (< 1100 m) of red ¯uvial sandstones and ¯oodplain siltstones transitionally overlain by an upper interval (< 1200 m) of sandy, buff-red ¯uvial conglomerates (Fig. 5.). Fig. 4. Map of the Camargo syncline, Eastern Cordillera, depicting local geology and measured section location. Note the persistent, north-trending ridge with up to 1500 m of relief along the syncline axis. The highest peaks, the Cerro Tonka peaks, occur directly north and south of the east-¯owing Rio Tumusla. The measured section consists of a lower interval (A) of Santa Lucia through Cayara Formations and an upper interval (B and C) of Camargo Formation (Fig. 6.). See Fig. 1.for location. # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 49 B. K. Horton and P. G. DeCelles Fig. 5. Schematic diagram of the Upper Cretaceous±Tertiary stratigraphic column in the Camargo syncline. Well-de®ned age control for Aroi®lla through Santa Lucia section is based on previous studies (see text). The poorly constrained age of the Impora±Cayara±Camargo succession is based on age data for possible Camargo-equivalent strata to the west (Potoco Formation) and north (Mondragon and Bolivar Formations). Column at left depicts interpretation of depositional setting (or depozone) within the foreland basin system (DeCelles & Horton, 1999.). DeCelles & Horton (1999.) ascribe the Cretaceous± Tertiary succession in the Camargo syncline to an eastward-migrating foreland basin system (Fig. 5.). In this interpretation, the Santa Lucia and possibly El Molino Formations represent deposition of ®ne-grained sediment in the distal part of the foreland basin system beyond the forebulge, a region called the back-bulge depozone (DeCelles & Giles, 1996.). The thick interval of stacked palaeosols in the Impora Formation, representing a long time interval of limited or no deposition (possibly up to 10±20 Myr), is consistent with stratigraphic condensation in the forebulge depozone. The Cayara strata and overlying thick Camargo interval are considered indicators of deposition in progressively more proximal locations of the foredeep depozone as the foldthrust belt advanced eastward. The absence of unambiguous growth strata in the upper part of the Camargo Formation suggests that the wedge-top depozone was not preserved in the Camargo syncline. Palaeocurrent data support the interpretation of a migrating foreland basin system, indicating: (1) an eastern sediment source for the Santa Lucia Formation, presumably cratonic cover rocks exposed to the east in Palaeocene time; (2) an eastern sediment source for the Cayara Formation, possibly a forebulge developed to the east in roughly mid Tertiary 50 time; and (3) a western sediment source for the Camargo Formation, clearly the fold-thrust belt active during mid Tertiary time in the central and western areas of the Eastern Cordillera (Sempere et al., 1997.; DeCelles & Horton, 1999.). Although a well-de®ned chronostratigraphy exists for the lower interval of this Cretaceous±Tertiary succession, the age of the upper part is poorly constrained (Fig. 5.). The El Molino Formation is de®ned as Maastrichtian to early Palaeocene based on invertebrate and mammal fossils, magnetostratigraphy, and volcanic tuffs dated in correlative strata in the Altiplano to the west (Sempere et al., 1997.). The Santa Lucia Formation contains mammal fossils and magnetostratigraphic data that limit it to 60±58 Ma (Sempere et al., 1997.). The ages of the Impora Formation and younger deposits, however, remain poorly known. Whereas Sempere et al. (1997.) argue for a late Palaeocene age for the Impora through Cayara Formations, new palynological age data for a probable age equivalent of the Camargo Formation, the Potoco Formation of the Altiplano (< 200 km to the west; Horton et al., in press)., suggest that these units may be signi®cantly younger. Based on these new data, and on early Miocene tuff ages (Kennan et al., 1995.) for the Mondragon and Bolivar Formations (probable upper Camargo Formation equivalents 100±200 km to the north; Sempere et al., 1997.), we tentatively assign a late Eocene to Miocene age for the Camargo Formation. We report here on the sedimentology of the Camargo Formation that comprises the majority of the Cretaceous±Tertiary succession in the Camargo syncline. SEDIMENTOLOGY OF THE CAMARGO FORMATION Methods A measured stratigraphic section in the Camargo syncline includes the Santa Lucia through Camargo Formations. The lower part of the section, Santa Lucia through Cayara strata, was measured along the gently dipping east limb of the syncline. The overlying Camargo interval (. Fig. 6.) was measured along the nearly complete exposure afforded by cliff faces comprising the Cerro Tonka peaks (. Fig. 7.), two massive topographic features with up to 1500 m of relief that ¯ank an intervening river, the Rio Tumusla (Fig. 4.). Due to inaccessibility of the steep east face of the southern Cerro Tonka peak (at 3824 m, the higher of the two peaks), we measured the upper part of the section along the west face (Fig. 4.). This upper interval is correlated with the lower interval by tracing laterally continuous sandstone bodies from the east face to the west face, a distance of about 400 m. The complete, laterally continuous exposure of the Camargo interval allowed collection of detailed lithofacies data and observation of bed geometries over lateral distances of hundreds of metres. Palaeocurrent data sets were measured throughout the # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia stratigraphic section at localities consisting of a single channel sandstone composed of trough-cross strata. The average trough-axis orientation for each locality was calculated from 15 to 25 measured trough limbs (method I of DeCelles et al., 1983.). Additional measurements included primary current lineations, erosional furrows and clast imbrications. More than 360 palaeocurrent indicators were measured at 24 localities in the Camargo syncline (Fig. 6.). These data indicate generally westward palaeo¯ow for the Santa Lucia and Cayara Formations and eastward palaeo¯ow for the Camargo section (Fig. 6.). In general, the Camargo Formation coarsens and thickens upward from a lower interval of thin-bedded siltstone and lenticular sandstone to an upper interval of very thick-bedded sandstone and conglomerate (Figs 6 and 7.). The Camargo succession is divisible into ®ve lithofacies described below and summarized in T . able 2.. Overbank facies Description The lower 250 m of the Camargo succession is dominated by facies 1, brick-red, mottled, structureless, noncalcareous siltstone (. Fig. 8A.). The thin (0.25±2.5 m), laterally continuous beds of this facies commonly exhibit gradational basal contacts with underlying sandstone facies, contain spheroidal siltstone aggregates (peds) coated by clay skins and slickensided surfaces, and display gypsum veins and nodules. Root traces, mudcracks, carbonate nodules and thin calcretes are rare. Facies 2 is more common in the middle Camargo interval. It is composed of light-red to grey, laminated, calcareous, clayey siltstone (Fig. 8A,B.). Beds of this facies are 0.05±0.20 m thick, ®nely laminated, typically overlie siltstones of facies 1, and nearly always underlie sandstones of facies 4 (Fig. 8B,C.). In several places, beds of facies 2 ®ll lenticular scours up to 3 m deep by several tens of metres wide (Fig. 8A.). Facies 3, common throughout the lower Camargo Formation, is composed of laterally continuous beds of red, ®ne- to medium-grained sandstone. Beds are 0.20±0.80 m thick by several tens to hundreds of metres wide and exhibit nonerosional bases, horizontal strati®cation, ripple cross-lamination, primary current lineations and bioturbation. These sandstones are interbedded with siltstones of facies 1 and 2. Although individual sandstone beds do not display upward grain-size trends, the beds are commonly organized into upward thickening, 1±5-m-thick packages that underlie sandstone beds of facies 3 and 4. Interpretation We interpret facies 1±3 as nonchannellized, ¯uvial overbank deposits (Table 2.). Variegated, mottled siltstones of facies 1 are interpreted as palaeosols in ¯oodplain settings; laminated siltstones of facies 2 are # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 interpreted as suspension fallout deposits in local marshy environments; and thin sandstone beds of facies 3 are attributed to crevasse splays in overbank areas. Palaeosol units (up to 2.5 m thick) developed in the siltstones of facies 1 indicate that some areas of the ¯oodplain sustained intervals of nondeposition suf®cient for soil development. These palaeosols generally lack carbonate material, suggesting intense leaching under well-drained conditions (Mack, 1993.) and/or a lack of carbonate in the sediment source area. The thin, laterally continuous, nonerosional sandstones of facies 3 are clear indicators of uncon®ned ¯ow on the ¯oodplain, probably crevasse splays formed by overtopping of channel banks or breaching of levees during ¯ood-stage conditions. Upward-thickening packages composed of facies 3 may represent progradation of splay complexes (Morozova & Smith, 2000.) and/or increased proximity to active channels. Paludal deposition of facies 2 represents local ponding of water in the topographically low areas of the ¯oodplain. In nearly every case, the paludal siltstones are abruptly overlain by a channel sandstone (facies 4) with limited relief on the basal scour surface (Fig. 8B,C.). We attribute this stratigraphic pattern to avulsion of channels into the lowest areas of the ¯oodplain (e.g. DeCelles et al., 1998.). A lack of well-de®ned epsilon cross-bedding for channel facies (described below) further indicates that channel mobility involved rapid, avulsive changes rather than continuous channel migration and point-bar deposition. Several cases of lenticular channel scours ®lled with siltstone (Fig. 8A.) attest to the importance of rapid abandonment of channels and subsequent plugging by overbank facies. Facies 1±3 all lack evidence for widespread, prolonged desiccation. In addition, most palaeosols are thoroughly leached of carbonate. These features may suggest an overall wet and perennial depositional system. Channel facies Description Facies 4 dominates the Camargo succession and is de®ned by lenticular, 0.5±6.0-m-thick beds of red-brown, medium- to coarse-grained sandstone that typically ®ne upward from a trough cross-strati®ed base to a horizontal- and ripple-strati®ed top (Fig. 8C.). The base of this facies is erosive and commonly contains siltstone intraclasts. Pebbles are locally present as clasts ¯oating within a sandstone matrix or as lenticular, clast-supported beds 2±10 cm thick. Dewatering structures, bioturbation and gypsum veins are present locally. In the lower Camargo Formation, this facies commonly forms the base of ®ning-upward packages that are overlain by siltstones of facies 1 and 2. Upsection, in the middle to upper Camargo Formation, these beds are commonly organized into larger multistorey sandstone bodies that are up to 50 m thick and persist laterally for 200±500 m (Figs 7 and 8D.). 51 B. K. Horton and P. G. DeCelles Fig. 6. Measured stratigraphic section of the Camargo Formation in the Camargo syncline (see Fig. 4.for location). An overlying < 800-m-thick interval of coarse-grained, uppermost Camargo strata was not measured due to inaccessibility of the highest cliffs of the Cerro Tonka peaks. Note the upsection decrease in palaeosols and siltstones and the upsection increase in grain size and bed thickness. 52 # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia Fig. 6. (continued). # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 53 B. K. Horton and P. G. DeCelles Fig. 7. View to the west of Camargo Formation exposures in the Camargo syncline. About 1500 m of relief exists between the basal Camargo contact (de®ned as the base of the lower ridge) along the vegetated Rio Tumusla (lower right) and the crest of the upper ridge (Cerro Tonka peak, 3824 m elevation, upper left). Note the upsection increase in frequency and thickness of laterally continuous sheets of lower Camargo sandstone (lower ridge) and middle±upper Camargo conglomerate (upper ridge). Facies 5 is composed of very thick-bedded, wellorganized, buff-red sandy conglomerate and is only found in the upper Camargo Formation (.Fig. 9.). Individual, broadly lenticular beds are 0.5±10.0 m thick and form large amalgamated bodies up to 80 m thick and over 400 m in lateral extent (Figs 7 and 9A.). These bodies have erosional basal contacts and commonly ®ne upward (Fig. 9B.). Sedimentary structures include moderately developed trough cross-strati®cation and subhorizontal strati®cation with imbricated clasts (Fig. 9C.). This facies is interbedded exclusively with sandstone beds of facies 4. Slingerland, 1998.). These values are similar to estimates of ancient ¯uvial systems in the Himalayan foreland basin system (Willis, 1993.; DeCelles et al., 1998.). Facies 4 and 5 lack epsilon cross-bedding, suggesting that the channels were not characterized by substantial lateral migration of point bars. In fact, the common pattern of paludal siltstones (facies 2) abruptly overlain by channel sandstones or conglomerates (facies 4 or 5) suggests that the ¯uvial channels avulsed, rather than migrated, into low areas of the ¯oodplain. Interpretation DEPOSITIONAL MODEL AND COMPARISON WITH MODERN FLUVIAL MEGAFANS We attribute facies 4 and 5 of the Camargo succession to deposition in large, bedload-dominated ¯uvial channels (Table 2.). Deposition was primarily by migrating sand dunes, bars and gravel sheets. Most individual sandstone and conglomerate beds are 0.5±10.0 m thick, ®ne upward and display sedimentary structures indicative of decreasing ¯ow velocity. We therefore interpret these units as the deposits of individual ¯ood events characterized by initial dune or bar migration followed by lower ¯ow velocities and plane-bed and ripple conditions. Observed channel dimensions (up to 10 m deep by 200±500 m wide), grain size (0.125±200 mm) and sedimentary structures indicate bank-full discharges in the range of 500±8000 m3 sx1 (calculation according to method outlined by Robinson & 54 The Tertiary Camargo Formation of the Camargo syncline forms a well-de®ned, upward-coarsening, upward-thickening succession attributable to progradation. An upsection increase in the dimensions and average grain size of channel bodies is coupled with an upsection decrease in the amount of ®ne-grained ¯oodplain deposits (Figs 6 and 7.). Palaeocurrent data indicate that Camargo strata were derived from a western sediment source area within the fold-thrust belt (Fig. 6.), whereas the underlying lower Tertiary deposits (about 335 m thick) composed of the Santa Lucia, Impora and Cayara Formations were derived from the east (Sempere et al., # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia Table 2.0Summary of lithofacies in Camargo Formation. Facies Description Stratigraphic occurrence Interpretation 1 Mottled, brick-red siltstone; generally non-calcareous; beds 0.25±2.5 m thick; no primary sedimentary structures; basal contacts gradational with underlying unit (commonly facies 3 and 4); local occurrences of soil aggregates (peds) 0.5±2 cm in diameter that exhibit clay-skin coatings and slickensided surfaces; gypsum nodules 1.0±3.0 cm in diameter and gypsum veins are common in lower 150 m; minor root traces, mudcracks, carbonate nodules and thin calcretes. Laminated, light-red to grey, clayey siltstone; calcareous; beds 0.05±0.20 m thick; characteristically occurs on top of facies 1 and below facies 4; uncommon channel plugs up to 3 m thick Thin, laterally continuous, red, ®ne- to medium-grained sandstone; beds 0.20±0.80 m thick; non-erosional basal contacts; ripple cross-strati®ed and horizontally strati®ed; primary current lineations; bioturbation; interbedded with facies 1 and 2; commonly occurs in upward-thickening, 1±5-m-thick packages that underlie facies 4. 1±50-m-thick bodies of lenticular, red-brown, mediumto coarse-grained or conglomeratic sandstone; beds 0.6±6.0 m thick; erosional basal contacts (commonly lined by siltstone intraclasts); upward ®ning; trough cross-strati®cation; plane-parallel lamination; ripple lamination; dewatering structures; local bioturbation; gypsum veins common in lower 250 m; pebble clasts include Palaeocene sandstone, Cretaceous limestone and sandstone, and Ordovician phyllite; commonly occurs as basal unit of ®ning-upwad package capped by units of facies 1 and 2. 2±80-m-thick bodies of well-organized, sandy, pebble to small boulder conglomerate; beds 0.5±10.0 m thick; erosional basal contacts; upward ®ning; cross-strati®ed or subhorizontally strati®ed and imbricated; clasts include Cretaceous limestone, sandstone and volcanics, and Ordovician sandstone, quartzite and phyllite. Present throughout Camargo interval, dominant in lower 250 m of Camargo Formation; bed thickness decreases upsection. Palaeosols in overbank areas. General lack of pedogenic carbonate suggests thorough leaching of soils and/or lack of carbonate in source area. Paludal deposits in overbank areas. 2 3 4 5 1997.; DeCelles & Horton, 1999.). We attribute this reversal in sediment dispersal direction and the upwardcoarsening and thickening trend to long-term, eastward progradation of the Camargo depositional system. Progradation may have been related to eastward propagation of the fold-thrust belt during mid Tertiary time (DeCelles & Horton, 1999.). The interpreted upsection progression from distal to proximal facies provides the basis for a Camargo depositional model that replicates the main features of the modern ¯uvial megafans of the Chaco Plain (.Fig. 10.). The Camargo depositional system is similar to the modern ¯uvial megafans in terms of channel and overbank facies, avulsive channel behaviour, proximal-to-distal trends in grain size and channel size, and approximate plan-view dimensions. First, similar facies attributes in the Chaco Plain ¯uvial megafans and # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Present mainly in middle Camargo Formation. Present mainly in lower± middle Camargo Formation. Crevasse splay deposits. Dominant facies of Camargo interval; unit thickness increases upsection: lower Camargo contains 1±5-m-thick units; middle±upper Camargo composed of packages of stacked, multistorey standstones forming units up to 50 m thick. Present in middle±upper Camargo Formation; thicknesses of multistorey conglomerate bodies increase upsection (up to 80 m thick). Fluvial channel deposits on megafan. Lower Camargo contains small (few m) channels of distal megafan; middle± upper Camargo contains large (<10 m) channels of proximal megafan. Large ¯uvial channel deposits on proximal megafan; channels up to 10 m deep. Camargo system include comparable-sized channels ®lled by sand and/or gravel and ¯oodplain settings exhibiting crevasse splay, paludal and pedogenic processes. These depositional systems are distinguished from smaller alluvial fans by the presence of ¯oodplain environments and the lack of sediment gravity ¯ows. These systems are distinguished from meandering and anastomosed rivers by a lack of point bars, scroll-bar topography, epsilon cross-bedding and permanent vegetated islands. Second, recent abandonment of channels and avulsive activity of the Chaco Plain megafans provides a modern analogue for the Camargo system, particularly in the case of laminated ¯oodplain siltstones abruptly overlain by newly established sandstone channels. Third, the decrease in both channel size and grain size from proximal to distal facies of the Camargo progradational succession matches the downslope trends 55 B. K. Horton and P. G. DeCelles Fig. 8. Representative mudstone and sandstone facies of the lower Camargo Formation. (A) Mottled siltstone palaeosols (visible in lower right) overlain by trough cross-strati®ed sandstone (0.3 m thick at right, 3 m thick at left), an interval of laminated siltstone (5 m thick to right, 3 m thick at left) and overlying section of multistorey sandstone bodies. Laminated siltstone interval has basal erosive scour (arrows) and is interpreted as the ®ne-grained plug of an abandoned channel. (B) Basal, massive to laminated siltstone overlain by low-angle cross-strati®ed, medium-grained sandstone that has a cap displaying poor horizontal and ripple cross-lamination. Note the limited scour relief (a few cm) along the sharp base of the sandstone. Knife (15 cm) for scale. (C) Ledges of 1±6-m-thick, lenticular, cross-strati®ed sandstones overlying laminated siltstone interval. Strata dip 60u to left. Shrub at lower left (2 m tall) for scale. (D) Cliff face composed of multistorey sandstone bodies roughly 5±50 m thick. Exposed thickness is < 250 m. in the modern ¯uvial megafans. Although a downslope reduction in grain size would be expected in nearly any ¯uvial system, most continental-scale rivers have contributary ¯ow patterns that will cause channel size (and presumably water discharge) to increase, not decrease, downstream. The Camargo depositional system is more compatible with ¯uvial megafans in which discharge diminishes downslope (e.g. Friend, 1978.). Such a decrease in relative discharge from proximal to distal settings in the Camargo system may best be attributed to distributary ¯ow in which the main feeder channel (relatively high discharge) at the exit point of the mountain belt bifurcates into numerous smaller channels (relatively low discharge) downstream (Fig. 10.). Finally, 56 although the original area of the Camargo depositional system cannot be con®dently estimated, continuous exposure of these strata in a ¯ow-normal direction provides an estimate of megafan width. The 25 km outcrop length of upper Camargo conglomerate in a north-trending ridge (including Cerro Tonka peaks) that is erosionally truncated on its northern and southern margins (Figs 4 and 7.) represents a minimum width across the proximal part of the megafan. This value is similar to the widths of the proximal parts of the Chaco Plain megafans (Fig. 2.). # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia Fig. 9. Representative conglomeratic facies of the Camargo Formation. (A) Coarsening- and thickening-upward interval in the middle Camargo Formation. Laterally continuous sheets of multistorey sandstone and sandy conglomerate are 10±80 m thick. Total exposed thickness of subhorizontal strata is < 400 m. (B) A lower, ®ning- and thinning-upward interval of cliff-forming conglomerate and slope-forming sandstone that is overlain by a cliff-forming conglomerate displaying an erosive basal contact (arrow). Total exposed thickness is < 40 m (wingspan of condor in foreground is < 3 m). (C) Close-up view of sandy conglomerate facies showing sandy matrix, rounded clasts, imbrication and subhorizontal strati®cation. Hammer (30 cm) at top for scale (arrow). # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 57 B. K. Horton and P. G. DeCelles Fig. 10. Fluvial megafan depositional model for the Camargo Formation. Key elements include an extensive drainage network, a large feeder canyon, gravel (G) deposition at the apex of the ¯uvial megafan, sand (S) and mud (M) deposition in middle to lower megafan areas, and an overall distributary channel pattern. Proximal to distal trends include a progressive increase in the number of channels and associated decrease in the size (width and depth) of channels. The model shows the approximate geomorphic con®guration of the region during deposition of the Camargo Formation, including the mountain front roughly 20 km to the west, and a major drainage network approximately similar in scale to the present-day network de®ned by the Rio Cotagaita and Rio Tumusla catchment areas. D I S C U S S I O N : A N C I E N T FL U V I A L MEGAFANS AND IMPLICATIONS FOR DRAINAGE NETWORK EVOLUTION Identi®cation of ¯uvial megafans in the ancient record may provide valuable information about the history of erosional denudation in a mountain belt. First, the existence or nonexistence of ¯uvial megafan deposits in an ancient foreland basin system provides constraints on the size of the source drainage network. Although in¯uenced by tectonic, climatic, lithological and geomorphological parameters unique to a given mountain belt, modern ¯uvial megafans in the central Andes (this study) and the Himalayan foreland (DeCelles & Cavazza, 1999.) suggest a minimum required drainage area of roughly 104 km2 to generate a ¯uvial megafan. Thus, megafans are genetically linked to relatively large drainage networks, suggesting their deposits may house the most direct record of largescale erosion in adjacent mountain belts (e.g. Schlunegger et al., 1997., 1998.; DeCelles et al., 1998.). In contrast, sediment gravity ¯ow-dominated and stream-dominated alluvial fans of the foreland basin will provide only local erosional histories of the frontal-most structures of the 58 orogenic belt. Furthermore, nonmegafan river systems (such as major axial rivers) in the foreland will contain erosional records that are subject to complicating factors such as mixing of multiple drainage networks and possible cratonic sources. For an incipient orogenic wedge, initial small drainage networks will develop on the foreland-sloping topographic surface of the fold-thrust belt (Beaumont et al., 1992.). Assuming no large antecedent drainages, ¯uvial megafans would not be expected in the early stratigraphic record of a foreland basin. Drainage networks would grow as the orogenic wedge advanced (widened), ultimately reaching a critical size for the production of ¯uvial megafans. As the larger drainage systems continually incorporate smaller watersheds through stream capture (Koons, 1995.), there would be some point when nearly all surface water in the mountain belt would be focused into these large drainage systems (e.g. Oberlander, 1965.). At this stage, the majority of sediment eroded from the mountain belt would be funnelled through a few outlet points at the orogenic front. Owing to their high stream power, these major drainage networks would be ®rmly imprinted on the evolving landscape, preventing sig# 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia ni®cant changes in drainage networks through stream diversion or ponding upslope of growing frontal structures (e.g. Burbank et al., 1996.; Gupta, 1997.; Schlunegger et al., 1998.). This stage could represent a sort of dynamic equilibrium or steady-state condition in which tectonic input to topography may be balanced by erosional out¯ux. We present a simple geometric model of drainage network evolution that highlights these concepts (.Fig. 11.). In the initial stages of fold-thrust belt evolution, numerous small streams feed sediment to small-scale alluvial fans ¯anking the mountain front (Fig. 11A.). The limited topography and areal extent of the orogen preclude integrated drainage networks, favouring smaller isolated catchments. The trunk streams within these small catchments have limited stream power, and are therefore more likely to be defeated by growing structures downslope (e.g. Burbank et al., 1996.). Such structural damming would lead to a greater proportion of closed drainage basins (including piggyback basins) in the early stages of mountain building. With advance of the foldthrust belt, individual drainage networks expand in area Fig. 11. Schematic model for the evolution of drainage networks and ¯uvial megafans during long-term advance of a fold-thrust belt. Three successive phases are depicted in which the structural front of the mountain belt propagates toward the foreland (bottom). Different stippled and shaded patterns for subaerial fans represent different relative fan areas and corresponding catchment areas. Note that the hinterland drainage divide (top) is arbitrarily ®xed throughout mountain building. (A) Early mountain building: small, poorly integrated drainage networks sourcing numerous alluvial fans (stippled patterns), large number of closed drainage basins due to topographic damming, and no ¯uvial megafans. (B) Intermediate stage of mountain building: headward growth and stream capture leads to larger, better integrated drainage networks and initial ¯uvial megafans (shaded patterns). (C) Late stage of mountain building: well integrated drainage networks sourcing a few, very large ¯uvial megafans. # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 59 B. K. Horton and P. G. DeCelles as tributary and trunk streams grow headward (Fig. 11B.). This elongation of drainage networks (Fraser & DeCelles, 1992.) is augmented by events of stream capture in which drainage basins expand their catchment area instantaneously. By this stage of mountain building, major trunk rivers in large drainage networks have become established. These rivers have more or less straight planform morphologies, are clearly antecedent to deformation in the frontal parts of the fold-thrust belt and deposit ¯uvial megafans in the proximal foreland basin system. Alluvial fans remain an important element at the mountain front. Continued propagation of the fold-thrust belt leads to further integration of drainage networks and the production of very large ¯uvial megafans (Fig. 11C.). Stream capture and headward erosion yield a simple drainage morphology in which the fold-thrust belt is ef®ciently drained by a few very large drainage networks (i.e. there are little to no areas of closed drainage within the orogen). At this point, a few large trunk rivers transport the majority of water and synorogenic sediment. The power of these rivers exceeds the uplift rate of any growing structure in the frontal part of the fold-thrust belt, preventing any substantial topographic damming and ponding of sediment. Although these rivers may owe their origin to the early stages of mountain building, they are clearly antecedent to the majority of fold-thrust structures in the orogen. It should be noted that this model considers the case where the hinterland topographic divide is ®xed in space and does not account for possible migration of the divide and establishment of hinterland plateaus. The model also ignores complexities related to variations in lithology, climate, local tectonics and geomorphic thresholds. Despite its simplicity, this model proposes a link between drainage network evolution and foreland-basin sedimentation that can be tested and modi®ed through further study of ancient foreland basin systems. The existence of distinct drainage outlets along the orogenic front would produce a foreland basin dominated by a few major point sources of sediment. This would generate large, along-strike variations in sediment ¯ux, foreland-basin stratigraphic architecture, and degree of sediment loading. Furthermore, the spacing of large drainage outlet points may maintain a uniform geometric relationship such that the ratio of orogen width to outlet spacing is < 2 throughout the growth of the orogenic wedge (see Hovius, 1996.). This ratio applies reasonably well in southern Bolivia where the three modern outlets are spaced < 135 km apart and the width of the foldthrust belt from the hinterland drainage divide to the thrust front is < 300 km. If this ratio is maintained during orogenesis, early stages of mountain building would display numerous closely spaced drainage outlets whereas later stages (with greater orogen width) would exhibit more widely spaced outlets (see Fig. 11.). Therefore, thrust-front advance and widening of the mountain belts must be accompanied by lateral expansion of mountainous drainage networks through sideward erosion and/or capture of smaller, laterally adjacent networks. 60 On a more local scale, the history of a particular drainage catchment may be assessed through identi®cation of ancient ¯uvial megafans in the stratigraphic record. Questions about river course and antecedence provide good examples. Simply stated, the existence of an ancient ¯uvial megafan deposit in a particular area of a foreland basin system implies that the area was receiving sediment near the outlet point of a relatively large drainage network in the ancient mountain belt. For the Camargo syncline of the Eastern Cordillera, we suggest that the location of the coarsest and thickest deposits of the interpreted Tertiary ¯uvial megafan (preserved in the resistant conglomeratic masses that form the Cerro Tonka peaks) directly north and south of the modern Rio Tumusla is not fortuitous. Rather, we infer that this modern river system represents a large ancestral river that dates back several tens of millions of years. Roughly 20 km upstream of the Cerro Tonka peaks, two trunk rivers (the Rio Tumusla and Rio Cotagaita) merge (Fig. 10.). The total area of the present-day drainage network upstream of the point where the Rio Tumusla enters the Camargo syncline is roughly 14 000 km2 (Fig. 10.), greater than the area of the modern Rio Parapeti network. We suggest that a precursor to this modern Tumusla±Cotagaita drainage system, similar in geometry and area, formed the source catchment for the late Eocene to Miocene ¯uvial megafan deposits in the Camargo syncline. In this interpretation, the mid Tertiary mountain front and Camargo megafan were located directly downstream of the con¯uence of the Tumusla and Cotagaita trunk rivers (Fig. 10.). The combined power of these rivers ensured the long-term existence of a notch in the mountain front through which large volumes of sediment were transported to the east. We conclude that the Tumusla±Cotagaita catchment, now a part of the drainage network that feeds the modern Pilcomayo ¯uvial megafan of the Chaco Plain (Fig. 1.), is largely inherited from an ancestral drainage system that fed the Camargo ¯uvial megafan during Tertiary time. We speculate that the persistence of such drainage networks, and their incorporation into larger integrated networks, is a natural progression in the evolution of mountain river systems in fold-thrust belts (Fig. 11.). Because topographic lows are dif®cult to destroy, it is possible that such drainage systems may even outlive the tectonic processes that produced them. CONCLUSIONS 1. Fluvial megafans form ®rst-order geomorphic elements of the modern foreland basin system adjacent to the central Andes. On the Chaco Plain of southern Bolivia, ¯uvial megafans are large (5800±22 600 km2), fan-shaped masses of sediment deposited where major rivers emanate from the fold-thrust belt at ®xed outlet points along the Subandean topographic front. Fluvial megafans contribute the greatest volume of sediment to the modern foreland basin, much more than the # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia stream¯ow-dominated alluvial fans derived from the frontal-most structures. Most erosion of the fold-thrust belt presently takes place in large mountain catchments such as the Rio Pilcomayo drainage network (81 300 km2). Megafan deposits therefore provide one of the most direct records of the erosional history of a mountain belt. 2. An ancient ¯uvial megafan system is recognized in the well-exposed, 2300-m-thick, Camargo Formation in the Camargo syncline of the Eastern Cordillera. This depositional system is similar to active ¯uvial megafans of the Chaco Plain in terms of approximate dimensions, lithofacies and depositional processes. The Camargo system had a ¯ow-normal (north±south) minimum width of 25 km, comparable to modern megafans. Ancient facies include thick bodies of lenticular channel sandstone and conglomerate as well as overbank deposits composed of mottled siltstone palaeosols, laminated paludal siltstones and crevasse-splay sandstones. Similar channels and overbank areas exist on the modern megafans. In both the modern and the ancient examples, channels exhibit little stability over time, constantly avulsing into new areas and abandoning previously active channels. 3. The upward-coarsening, upward-thickening Camargo succession documents the progradation of a ¯uvial megafan as the fold-thrust belt advanced eastward. The coarsest and thickest strata in the upper Camargo Formation spatially coincide with the location of a modern drainage outlet for the present-day Rio Tumusla ± Rio Cotagaita drainage network (14 000 km2). This network may represent an antecedent catchment that initially provided a point source of sediment to the Camargo megafan and has been inherited by the presentday Pilcomayo drainage system. 4. A simple geometric model for the development of drainage networks in advancing fold-thrust belts suggests that drainage networks become larger and better integrated over time. A given network must reach a critical size (roughly 104 km2) before ¯uvial megafans will be produced in the proximal foreland basin. Expansion of drainage networks leads to a situation whereby most sediment eroded from the orogenic belt is delivered by point sources (¯uvial megafans) to the foreland. Mainstem rivers in the larger drainage networks will have high stream power and their courses should not be affected by growing frontal structures. ACKNOWLEDGMENTS This research was supported by NSF grant EAR9804680. We are grateful to Richard Fink for ®eld assistance and Sohrab Tawackoli, Reinhard Roessling, Carlos Riera and Juan Huachani of Sergeomin (La Paz) for logistical assistance. We have bene®ted from discus# 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 sions with Thierry Sempere, Justin Wilkinson and Nadine McQuarrie. Constructive reviews by Rebecca Dorsey, Fritz Schlunegger and Gerilyn Soreghan signi®cantly improved the manuscript. REFERENCES BAKER, V.R. (1978) Adjustment of ¯uvial systems to climate and source terrain in tropical and subtropical environments. In: Fluvial Sedimentology (ed. by A.D. Miall), Mem. Can. Soc. Petrol. Geol., 5, 211±230. BEAUMONT, C., FULLSACK, P. & HAMILTON, J. (1992) Erosional control of active compressional orogens. In: Thrust Tectonics (ed. by K. R. McClay), pp. 1±19. Chapman & Hall, London. BEER, J.A., ALLMENDINGER, R.W., FIGUEROA, D.E. & JORDAN, T.E. (1990) Seismic stratigraphy of a Neogene piggyback basin, Argentina. Am. Ass. Petrol. Geol. Bull., 74, 1183±1202. BLAIR, T.C. & MCPHERSON, J.G. (1994) Alluvial fans and their natural distinction from rivers based on morphology, hydraulic processes, sedimentary processes, and facies assemblages. J. Sedim. Res., A64, 450±489. BURBANK, D., MEIGS, A. & BROZOVIC, N. (1996) Interactions of growing folds and coeval depositional systems. Basin Res., 8, 199±223. BURBANK, D.W. & RAYNOLDS, R.G.H. (1988) Stratigraphic keys to the timing of thrusting in terrestrial foreland basins: Applications to the northwestern Himalaya. In: New Perspectives in Basin Analysis (ed. by K. L. Kleinspehn & C. Paola), pp. 331±351. Springer-Verlag, New York. DAMANTI, J.F. (1993) Geomorphic and structural controls on facies patterns and sediment composition in a modern foreland basin. In: Alluvial Sedimentation (ed. by M. Marzo & C. PuigdefaÁbregas), Spec. Publ. Int. Ass. Sediment., 17, 221±233. DECELLES, P.G. & CAVAZZA, W. (1999) A comparison of ¯uvial megafans in the Cordilleran (Upper Cretaceous) and modern Himalayan foreland basin systems. Geol. Soc. Am. Bull., 111, 1315±1334. DECELLES, P.G., GEHRELS, G.E., QUADE, J., OJHA, T.P., KAPP, P.A. & UPRETI, B.N. (1998) Neogene foreland basin deposits, erosional unroo®ng, and the kinematic history of the Himalayan fold-thrust belt, western Nepal. Geol. Soc. Am. Bull., 110, 2±21. DECELLES, P.G. & GILES, K.N. (1996) Foreland basin systems. Basin Res., 8, 105±123. DECELLES, P.G. & HORTON, B.K. (1999) Implications of early Tertiary foreland basin development for orogenesis in the central Andes. Eos, Transactions, Am. Geophys. Union, 80, 1052. DECELLES, P.G., LANGFORD, R.P. & SCHWARTZ, R.K. (1983) Two new methods of paleocurrent determination from trough cross-strati®cation. J. Sedim. Petrol., 53, 629±642. FLEMINGS, P.B. & JORDAN, T.E. (1989) A synthetic stratigraphic model of foreland basin development. J. Geophys. Res., 94, 3851±3866. FRASER, G.S. & DECELLES, P.G. (1992) Geomorphic controls on sediment accumulation at margins of foreland basins. Basin Res., 4, 233±252. FRIEND, P.F. (1978) Distinctive features of some ancient river 61 B. K. Horton and P. G. DeCelles systems. In: Fluvial Sedimentology (ed. by A. D. Miall), Mem. Can. Soc. Petrol. Geol., 5, 531±542. GEDDES, A. (1960) The alluvial morphology of the Indo-Gangetic plain: Its mapping and geographical signi®cance. Transactions and Papers. Inst. Br. Geographers, 28, 253±276. GOHAIN, K. & PARKASH, B. (1990) Morphology of the Kosi megafan. In: Alluvial Fans: a Field Approach (ed. by A. H. Rachocki & M. Church), pp. 151±178. John Wiley and Sons, Chichester, UK. GUBBELS, T.L., ISACKS, B.L. & FARRAR, E. (1993) High-level surfaces, plateau uplift, and foreland development, Bolivian Central Andes. Geology, 21, 695±698. GUPTA, S. (1997) Himalayan drainage patterns and the origin of ¯uvial megafans in the Ganges foreland basin. Geology, 25, 11±14. GUYOT, J.L., CALLE, H., CORTES, J. & PEREIRA, M. (1990) Transport of suspended sediment and dissolved material from the Andes to the Rio Plata by the Bolivian tributaries of the Rio Paraguay (Rios Pilcomayo and Bermejo). Hydrol. Sci. J., 35, 653±665. HELLER, P.L., ANGEVINE, C.L., WINSLOW, N.S. & PAOLA, C. (1988) Two-phase stratigraphic model of foreland-basin sequences. Geology, 16, 501±504. HORTON, B.K. (1998) Sediment accumulation on top of the Andean orogenic wedge: Oligocene to late Miocene basins of the Eastern Cordillera, southern Bolivia. Geol. Soc. Am. Bull., 110, 1174±1192, 1513. HORTON, B.K. & DECELLES, P.G. (1997) The modern foreland basin system adjacent to the Central Andes. Geology, 25, 895±898. HORTON, B.K., HAMPTON, B.A. & WAANDERS, G.L. (2001) Paleogene synorogenic sedimentation in the Altiplano plateau and implications for initial mountain building in the central Andes. Geol. Soc. Am. Bull., 113, in press. HOVIUS, N. (1996) Regular spacing of drainage outlets from linear mountain belts. Basin Res., 8, 29±44. IGM. (1972). Mapa Topogra®co: Vallegrande. 1: 250,000, Hoja SE 20±10. 1st edn. Instituto Geogra®co Militar, La Paz. IGM. (1974) Mapa Topogra®co: Charagua. 1:50,000, Hoja 6934 I, 1st edn. Instituto Geogra®co Militar, La Paz. IGM (1980a) Mapa Topogra®co: Cerro Capitan Ustarez. 1:250,000, Hoja SE 20±15, 1st edn. Instituto Geogra®co Militar, La Paz. IGM. (1980b) Mapa Topogra®co: Charagua. 1:250,000, Hoja SE 20±14, 1st edn. Instituto Geogra®co Militar, La Paz. IGM. (1981) Mapa Topogra®co: Villamontes. 1:250,000, Hoja SF 20±6, 1st edn. Instituto Geogra®co Militar, La Paz. IGM. (1998) Mapa Fisico de Bolivia. 1:1,000,000, 3rd edn. Instituto Geogra®co Militar, La Paz. IRIONDO, M.H. (1984) The Quaternary of northeastern Argentina. In: Quaternary of South America and Antarctic Peninsula, Vol. 2 (ed. by A. A. Balkema), pp. 51±78. Balkema, Boston, MA. IRIONDO, M.H. (1993) Geomorphology and late Quaternary of the Chaco (South America). Geomorphology, 7, 289±303. IRIONDO, M.H. (1997) Models of deposition of loess and loessoids in the upper Quaternary of South America. J. S. Am. Earth Sci., 10, 71±79. ISACKS, B.L. (1988) Uplift of the central Andean plateau and bending of the Bolivian orocline. J. Geophys. Res., 93, 3211±3231. 62 JORDAN, T.E. (1995) Retroarc foreland and related basins. In: Tectonics of Sedimentary Basins (ed. by C. J. Busby & R. V. Ingersoll), pp. 331±362. Blackwell Science, Cambridge, MA. JORDAN, T.E. & ALONSO, R.N. (1987) Cenozoic stratigraphy and basin tectonics of the Andes Mountains, 20ux28u south latitude. Am. Ass. Petrol. Geol. Bull., 71, 49±64. JORDAN, T.E., REYNOLDS, J.H. & ERIKSON, J.P. (1997) Variability in age of initial shortening and uplift in the central Andes, 16±33u30k S. In: Tectonic Uplift and Climate Change (ed. by W. F. Ruddiman), pp. 41±61. Plenum Press, New York. KELLY, S.B. & OLSEN, H. (1993) Terminal fans ± A review with reference to Devonian examples. Sediment. Geol., 85, 339±374. KENNAN, L., LAMB, S. & RUNDLE, C. (1995) K±Ar dates from the Altiplano and Cordillera Oriental of Bolivia: implications for Cenozoic stratigraphy and tectonics. J. S. Am. Earth Sci., 8, 163±186. KOONS, P.O. (1995) Modeling the topographic evolution of collisional belts. Annu. Rev. Earth Planet. Sci., 23, 375±408. MACK, G.H. (1993) Paleosols as an indicator of climatic change at the Early±Late Cretaceous boundary, southwestern New Mexico. J. Sedim. Petrol., 62, 483±494. MOROZOVA, G.S. & SMITH, N.D. (2000) Holocene avulsion styles and sedimentation patterns of the Saskatchewan River, Cumberland Marshes, Canada. Sediment. Geol., 130, 81±105. NEWELL, A.J., TVERDOKHLEBOV, V.P. & BENTON, M.J. (1999) Interplay of tectonics and climate on a transverse ¯uvial system, Upper Permian, southern Uralian foreland basin, Russia. Sediment. Geol., 127, 11±29. NICHOLS, G.J. (1987) Structural controls on ¯uvial distributary systems ± The Luna system northern Spain. In: Recent Developments in Fluvial Sedimentology (ed. by F. G. Ethridge, R. M. Flores & M. D. Harvey), Spec. Publ. Soc. Econ. Paleont. Miner., 5, 269±277. OAS. (1971) Cuenca del RõÂo de la Plata: Estudio para su plani®cacioÂn y desarrollo; Inventario y anaÂlisis de la informacioÂn baÂsica sobre recursos naturales. Organization of American States, Of®ce of Regional Development, Washington, DC. OBERLANDER, T. (1965) The Zagros Streams, Syracuse Geographical Series, no. 1. Syracuse University, New York. ROBINSON, R.A.J. & SLINGERLAND, R.L. (1998) Grain-size trends, basin subsidence and sediment supply in the Campanian Castlegate Sandstone and equivalent conglomerates of central Utah. Basin Res., 10, 109±127. RAÈSAÈNEN, M., NELLER, R., SALO, J. & JUNGER, H. (1992) Recent and ancient ¯uvial deposition in the Amazonian foreland basin, Peru. Geol. Mag., 129, 293±306. SCHLUNEGGER, F., JORDAN, T.E. & KLAPER, E.M. (1997) Controls of erosional denudation in the orogen on foreland basin evolution: The Oligocene central Swiss Molasse Basin as an example. Tectonics, 16, 823±840. SCHLUNEGGER, F., SLINGERLAND, R. & MATTER, A. (1998) Crustal thickening and crustal extension as controls on the evolution of the drainage network of the central Swiss Alps between 30 Ma and the present: constraints from the stratigraphy of the North Alpine Foreland Basin and the structural evolution of the Alps. Basin Res., 10, 197±212. SEMPERE, T., BUTLER, R.F., RICHARDS, D.R., MARSHALL, L.G., SHARP, W. & SWISHER, C.C. (1997) Stratigraphy and chronology of Late Cretaceous-early Paleogene strata in # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 Fluvial megafans in the central Andes, Bolivia Bolivia and northwest Argentina. Geol. Soc. Am. Bull., 109, 709±727. SEMPERE, T., HERAIL, G., OLLER, J. & BONHOMME, M.G. (1990) Late Oligocene ± early Miocene major tectonic crisis and related basins in Bolivia. Geology, 18, 946±949. SINHA, R. & FRIEND, P.F. (1994) River systems and their sediment ¯ux, Indo-Gangetic plains, northern Bihar, India. Sedimentology, 41, 825±845. STANISTREET, I.G. & MCCARTHY, T.S. (1993) The Okavango fan # 2001 Blackwell Science Ltd, Basin Research, 13, 43±63 and the classi®cation of subaerial fan systems. Sediment. Geol., 85, 115±133. TUCKER, G.E. & SLINGERLAND, R. (1996) Predicting sediment ¯ux from fold and thrust belts. Basin Res., 8, 329±349. WILLIS, B. (1993) Ancient river systems in the Himalayan foredeep, Chinji Village area, northern Pakistan. Sediment. Geol., 88, 1±76. Received 12 April 2000; revision accepted 6 November 2000 63