The Caerfanell Valley

 (abridged version of 'Talybont - the Legacy of the Ice Age')

This study is the product of nearly forty years of geographical fieldwork at Talybont-on-Usk in South Wales. The Afon Caerfanell is a tributary of the River Usk draining the eastern end of the main Brecon Beacons ridge, and is a popular venue for visitors to the Beacons National Park. We will concentrate here on the physical geography of the valley; its economic and historical geography will be summarised later. 

           Sketch Map of the Caerfanell Basin.    

The River Caerfanell has many surprising features: 

The Upper Caerfanell

1. Vertical aerial photograph of the Upper Caerfanell basin.                                                                               Craig y fan ddu is the narrow ridge in the centre, separating the deeper and larger Blaen y glyn headwater valley to the right (east) of it from the smaller Blaen Caerfanell/Bwrefwr stream just to the west.  They join up on the edge of the forested slopes near the southern edge of the image, before curving sharply east and north-east into the Glyn trough. The valley in the south-west corner is that of the Taf fechan. Note the 650 metre high glacially scoured rocks and peat hags in the north, flanking the cirque headwall of Cwm Oergwm, and the hanging valley and elbow of capture of the Blaen Caerfanell just north of Craig y fan ddu. This the south-eastern end of the main Brecon Beacons ridge.                                                                                                                         

The Afon Caerfanell proper has twin sources near the scarp crest of the Brecon Beacons: GR (SO) 060 205 at 750 metres above sea level, source of the Blaen-y-glyn, and 050 200 at 735 m, source of the Blaen Caerfanell. Both rise among peat hags which are being rapidly eroded, exposing the underlying glacial rock pavement. They are flowing southwards following the ancient regional tilt of the Old Red Sandstone rocks on the north flank of the South Wales syncline (structural basin). There is a subsidiary source 1 km short of the ridge (at about 670 metres OD) shown on the profile below, the Nant Bwrefwr.

2. Long profile of the Afon Caerfanell.                                                                                                             NB:  The profile is drawn to scale in two equal portions, which join conveniently at the reservoir delta.

The Blaen-y-glyn branch quickly descends the head of a deep U-shaped glaciated valley bounded by the scar of a periglacial landslip on the eastern side (Cwar y Gigfran) and nivation and solifluction features on the west (Graig Fan Las). There is no obvious recent cirque at its head, partly because it faces south, and partly because of its position at the extreme eastern (leeward) end of the main scarp, both of which would have reduced snow and ice accumulation. However, the scale and shape of the valley strongly suggest glacial origins, perhaps in an earlier period than the Late Devensian Glaciation and Mountain Readvance that ‘freshened up’ most of the features that today characterise the Brecon Beacons. There are hollows with ice-steepened back walls like glacial cirques, and these have arcuate drift features below them that have been described as snow moraines or protalus ramparts. The present stream is plainly a misfit: it is far too small to have entirely produced such a large valley. It has barely cut through the superficial deposits on the valley floor to the bedrock below, and its tributaries are little more than rills and gullies coursing straight down smooth valley sides mantled with scree and periglacial debris. Even if the valley has not been recently occupied by a glacier, it certainly has been shaped and possibly even initiated by moving ice, and then modified by nivation under snow patches and solifluction on frozen ground (permafrost).

The Blaen Caerfanell branch, on the other hand, flows for some way on the gentle five degree dip slope (or ‘back slope’) of the Beacons escarpment. It runs partly over a glacially scoured bedrock surface (tough jointed ‘grey grits’ showing fossil wave ripples and worm casts, and occasional striations (glacial scratches)), and partly wanders indeterminately through blanket bog and peat hags – the eroded remains of post-glacial birch forests. There are also exposed outcrops and boulders showing evidence of frost action: repeated freeze and thaw cycles producing frost heave, riving and shattering of rocks, terracettes and turf rolls, and solifluction of boulders down slope.

Having flowed gently southwards thus for about a kilometre, the Blaen Caerfanell suddenly turns through 90 degrees into the deep Blaen-y-glyn valley to the east (at GR 050 192). It becomes a hanging valley with a staircase of waterfalls totalling more than 200 metres – each step marking a resistant layer in the grits and the underlying ‘brownstone’ sequence. It is spectacular in wet weather, but the water often freezes in winter or evaporates in summer before reaching the bottom. Strong anabatic (upslope) winds can sometimes cause a plume from the top like a fountain.

This hanging valley is an interesting example of river capture. Before the last Ice Age the Blaen Caerfanell stream almost certainly followed the ancient dip slope of the Brecon Beacons all the way southwards, probably as a headwater of the ancestral River Taff. But during the last cold period (the Mountain Readvance) nivation (a combination of ice, frost and snow action) widened and deepened the larger valley parallel to it on its eastern side – the precursor of the Blaen y Glyn. The steep valley edge of the latter was weathered and eroded backwards, possibly where a shaded and sheltered gulley allowed snow and frost to linger, eating into the basin of its smaller, higher neighbour. This diverted the upper part of the Blaen Caerfanell eastwards at what is called an ‘elbow of capture’, to join the Blaen y Glyn stream at GR 058 188. The latter was rejuvenated as a result of the capture of its neighbour’s headwaters, contributing to the very dramatic erosion downstream. At the confluence with the Blaen-y-glyn branch the combined streams become the Afon (River) Caerfanell.

3a. The upper glaciated valley of the Caerfanell.              3b. The Blaen Caerfanell source on the Beacons ridge. The Blaen y glyn flows down the headwall on the                 The  stream is flowing due south towards us, then right, and the Blaen Caerfanell tumbles down the              it suddenly turns east (right) in the foreground, steep gulley from the high back slope on the left.                  and plunges steeply down the gulley shown in          This is the site of the river capture described in the            photo 3a.  This is an 'elbow of capture'; before ice     text.                                                                                                                            action enlarged the Blaen-y-glyn valley the Blaen                                                                                                                                         Caerfanell continued down the dip slope to the left.   

Meanwhile, you can trace the peat-filled depression through which the Blaen Caerfanell used to flow continuing southward for two or three hundred metres down the original Beacons back slope – now a small ‘wind gap’ only two or three metres higher than the present stream at the elbow. From here springs and pools in the boggy peat rise to become the modern Nant Bwrefwr (another misfit stream, this time the result of river capture).
The Nant Bwrefwr headwater flows for about a kilometre almost due south from its source towards the steepening slopes of Torpantau. There is hardly a recognisable channel through the rapidly eroding peat on the gentle Beacons dipslope nearest the source, but the gradient progressively steepens to a waterfall as it descends the head of the glacial section of its valley via a distinct notch in the headwall. The gradient lessens noticeably on reaching the floor of the valley; it widens quickly to about 400m, with steep sides and the beginnings of a flood plain. This too is a hanging valley, similar to the Blaen Caerfanell headwater, but on a smaller scale.

4a. Data collection in the upper Nant Bwrefwr.                   4b. The Bwrefwr begins its steep descent to Torpantau. One km below its source near the elbow of capture (beyond the skyline in 4a) the beheaded stream incises itself increasingly within this typical upland valley, before plunging steeply down a dramatic staircase of waterfalls (4b) to Torpantau and Blaen y glyn.

The river channel, though well-defined now, is small and variable in shape and composition: details of channel and valley form in this upper section of the Caerfanell are governed by the thickness and composition of the superficial mantle. The bed is sometimes smooth sandstone, sometimes stony or shingly, and intermittently stepped; flow is often obstructed by boulders that have soliflucted down the valley sides. The banks are low and often overhung, partly from undercutting by the stream, and partly from soil creep and slopewash. Bedload is generally coarse and angular, since it has not had enough time in the river to become rounded and comminuted by abrasion. The depth is small, friction is locally high, and consequently there are wide variations in velocity across the channel, especially in low-flow conditions.

However, in flood conditions cross-sectional area, velocity, and discharge (area x velocity), all increase very rapidly. As energy in a stream increases approximately as the square of its velocity, and greater depth produces relatively less friction, these apparently small Caerfanell headwaters have a surprising capacity for sediment transport and erosion when at bankfull. Where these streams meet downstream, and discharge is further increased by runoff and groundwater springs, the effects are multiplied many times. We will consider the effects of this rapidly escalating energy in the next sections.

At Torpantau, one kilometre from its beheaded source, the gradient of the Nant Bwrefwr steepens again as it begins its 150 m cascade to join the main River Caerfanell at Blaen y glyn. It becomes increasingly incised into the periglacial slopes, producing a distinct V-shaped notch within the broader U-shaped glacial valley. All the upper south-facing slopes are mantled with boulder clay and stony material that soliflucted (sludged) down from the summits during summer melt towards the end of the Ice Age. There is a frost-riven boulder field on the flanks of the col used by the road from Talybont to Taf fechan. On the colder north-facing slope of Pant y Creigiau it is possible to make out a `staircase` of altiplanation terraces – these are frost weathered benches trimmed by nivation and veneered with solifluction deposits, not related in any obvious way to variations in rock resistance.

5a. Altiplanation terraces in the Torpantau col.        5b. Heavily eroded footpath alongside the Bwrefwr.               These result from freeze-thaw and nivation.               This is where Wilsons School did their survey; it is now The Glyn is to the left, the Taf  fechan valley to          largely pitched with stones (see text below).                           the right, linked by a road and a disused rail-                                                                                                                                                   way tunnel.                                                      

The Torpantau col was probably formed by ice spilling over from the large glacier that filled the adjoining Taf fechan valley earlier in the Devensian glacial period. This ice would have been instrumental in producing the very steep gradients at Blaen y glyn, and was probably the main source of the glacier that produced the considerable modifications to the middle Caerfanell valley we shall describe later. However, we shall look first at a more recent but equally potent erosion process that has affected these slopes. 

 Hill Walkers:  The path leading to Craig y fan ddu from the Torpantau car park has been pitched with stones to protect it from erosion by the boots of hill walkers, which has become a serious problem in many parts of the Brecon Beacons in recent years. The erosion process is as follows:

walkers` boots trample the plants, killing them and compacting the exposed soil >

rainwater is unable to percolate the compacted soil, so it runs down the path >

the path is eroded by the water into a gulley, and stones are dislodged >

walkers increasingly avoid the gulley, so the path widens as well as deepens >

the process continues, spreads, and accelerates, facilitated by the readiness with which the stones in the unconsolidated glacial and periglacial regolith are dislodged.

In the early 1990s students from Wilsons School compared the rates of down-cutting on footpaths with those in local streams by calculating from measurements made directly at Torpantau. The depth of the footpath below the adjoining land surface can be measured directly, and if its age can be determined or inferred from secondary sources (in most cases it is less than 100 years up in the hills), simple division will produce the mean amount of footpath erosion per year.

To arrive at an estimate of stream erosion rates the assumption first needs to be made that the pronounced V-shaped notch incised into the boulder clay material that lines the bottom of the glacial valleys in the Brecon Beacons has been produced since the ice and permafrost melted at the end of the last glacial episode, about 10,000 years ago. The depth of this ‘valley in valley’ notch can be measured and calculated using simple trigonometry (it varies between about 5 to 10 metres at Torpantau). This is then divided by 10,000 to give the mean annual rate of downcutting by the stream.

Wilsons’ results showed that the main footpath from Blaen y glyn to Craig y fan ddu was being eroded approximately seven times faster than the Nant Bwrefwr stream alongside it. During the 20th century an ugly scar up to 10 metres wide and one and a half metres deep, and visible from many miles away, had developed on the hillside, most noticeably in the last thirty years or so. All the additional soil and runoff from this erosion goes quickly into the streams on these steep gradients, and the implications for bedload and flood levels in the middle section of the Caerfanell and for the rate of silting of the Talybont Reservoir are examined later.

                Ice and water at Blaen y glyn

This part of the basin has experienced rapid rejuvenation resulting from the combination of river capture, glacial scour, gulleying, and footpath erosion we have described above. The increasingly steep and irregular gradients, rapid surface runoff, and high drainage density have produced high energy, rapid downcutting, and unstable slopes. A striking series of waterfalls has resulted, particularly where the process has been temporarily arrested by resistant rock bands such as the ’12 foot Sandstone'.

6a. The main Blaen y glyn Waterfall.                                               6b. Helical (corkscrew) flow - extreme downcutting The water descends an 8 metre step in the valley caused      results from sudden narrowing and steepen- by the tough ’12 foot Sandstone’. The back of the fall is          ing of the channel. It is only 60cm wide and 10cm not quite vertical, but the water has already built up            deep, so it speeds up and gains enormous energy. considerable momentum from the convex slopes above,      Note the tube-like shape of the chute below it,       on smooth rocks regularly swept clean of debris.                        and the undercut rocks, soil and vegetation. It  Note the thinner layers at the top which are cutting              may have begun as a riverbed pothole.                         back more quickly, the undercutting at the sides produc-                                                                                                                       ing a gorge, and the plunge pool and potholes at the base.

 Here we can appreciate the different means by which water erodes resistant rocks:-           1. speed and weight of water (hydraulic action);                                                                                2. explosive compression of air in fissures (cavitation);                                                                    3. stones bouncing (saltation) and grinding (traction) along the bed during floods;             4. potholes drilled where pebbles spin round in joints and fissures (abrasion and attrition); 5. undercutting of the valley sides along bedding planes; and                                                     6. spring sapping where water issues from joints on the sides of the gorge (assisted by the prising action of growing tree roots, and freeze-thaw weathering).                                    

Erosion is rapid, especially when the river is in flood, as we have seen. It is likely that the entire gorge at the waterfall (about 50 metres in length) has been formed in the 10 000 years since the end of the Ice Age. This is consonant with measured and calculated rates of erosion such as those at Torpantau above. There would of course have been little runoff when the ground was frozen, and any valley existing before the last glacial period is likely to have been filled with glacial and periglacial debris.

The vegetation is important around the waterfalls, as plant colonisation has had a role in valley and gorge development. Plants grow in zones either side of the waterfall according to variations in the amount of available sunlight, water and nutrient – algae on rocks frequently under water, next mosses and liverworts where humidity is high, with lichens on drier rocks, then ferns and grasses in cracks where organic matter has begun to accumulate, flowering plants where there is more light, soil and shelter, and finally alder, hawthorn and birch trees anchored in joints. This spatial arrangement also reflects the probable sequence of plant colonisation after the retreat of the ice. Biotic weathering assists in the chemical decomposition and mechanical disintegration of rocks, producing peaty soils.

Contrast the dark, evergreen spruce and fir plantations on the slopes around Blaen y glyn with the more open natural oak and alder woodland near the waterfall. The conifers were planted on steep slopes partly with the intention of reducing the rate of runoff and soil wash into the rivers and the reservoir, but regular planting in rows and lack of an under storey reduces any such benefit, and must be set against the disadvant-age of a reduction in habitat and soil quality that seems to be irreversible, at least in the short term.

300 m below Blaen y glyn waterfall, past the confluence of the Nant Bwrefwr with the Caerfanell, is another hard rock band. The river narrows down suddenly from 3 metres wide just upstream to less than 1 metre here, forcing it to speed up dramatically. It then corkscrews down a three metre chute, undercutting the resistant rocks on either side to produce a tube-like channel. Features like this can be seen today emerging from underneath melting Alpine glaciers, and it is possible that this was initiated in a similar way. It is a torrent under normal conditions, but when the river is in spate this helical flow acts like a water turbine, producing some of the most rapid erosion of all. Below the chute is an attractive curtain waterfall, and near here landslips have narrowed the main river channel and diverted two small tributaries downstream. Instability is very much the norm at Blaen y glyn.

 7a. Bouldery channel upstream of Pont Blaen y glyn.     7b. Bedload sampling and channel measurements.   

      Glyn Collwn: the Middle Caerfanell Valley

Below Blaen y glyn the river and its valley noticeably change character (figures 8 & 9). There are no more waterfalls, but there is a distinctly bouldery section of river channel. Some of the boulders are too bulky to have been brought by the modern river; they probably originated as material moved by ice and meltwater, or were delivered to the river down the sides of the valley by solifluction, landslips and other gravity movements on slopes steepened by ice action.






8. Glyn Collwn (Caerfanell Valley) from the southern watershed at Bryn Melyn. Blaen y glyn and the Torpantau col are at the head of the valley, the summit ridge of the Brecon Beacons is off to the right. Note the delta where the Caerfanell enters Talybont Reservoir, the wide, almost  flat valley floor, and the curve of the valley. Ynys poste and Abercynafon (see below) are in the centre of the picture, just upstream of the delta.


In 'the Glyn' (Glyn Collwn) the gradient lessens, and the river no longer fills the bottom of the valley. This has widened into a U-shaped trough typical of a valley carved by a powerful glacier. The southern valley side is particularly steep here, and the Dyffryn Crawnon watershed above it is very narrow in places, suggesting an arete (a knife-edged ridge separating two glaciated valleys). Opposite the site of Ynys poste farmhouse is the most southerly point on the river. Here the river has been directed right up against the valley side, undercutting it under the influence of the strong southward dip of the rocks. River erosion is obviously still active here, as the pronounced slope-foot scars demonstrate, but it is directed laterally more than vertically. This uniclinal shifting would have been particularly potent in and just after the Ice Age, when firstly glaciers and then their meltwater swept down the tributary valleys from the north, such as the Nant Cynafon and the Blaen y Glyn. The combination of active ice erosion in the past and recent uniclinal shifting by the river is probably the explanation not only for the particularly steep valley slopes on the south side, but also for the asymmetrical shape of the whole Caerfanell basin we commented on earlier. Projected into the future, continuation of these processes could lead to migration and lowering of the watershed above, and since the Glyn is deeper than Dyffryn Crawnon, eventually to expansion of the Caerfanell catchment area at the expense of its neighbour. This potential river capture is likely to be facilitated by the line of prominent springs high up near the watershed, which are partly fed by highly permeable limestone rocks which were quarried until recently at the head of Dyffryn Crawnon.

At Abercynafon the Caerfanell springs a new surprise. The river that has gained so much energy, discharge and momentum during its steep 250 metre descent from above Torpantau suddenly loses much of it again. Stream measurements here often show a decrease in discharge in summer and autumn compared with Blaen y glyn upstream, such that sometimes the river bed is completely dry. Yet there is also evidence of occasional flooding. The impermeable parent rock and rainy climate have not changed, so why does discharge decrease overall, and the river level fluctuate so much?

9. The Afon Caerfanell at Abercynafon.                                                                                                                                                                Note the much reduced gradient and discharge in this stretch, about a mile below Blaen y glyn; much of its water has been absorbed by the permeable glacial debris making up the valley floor here. The river is flowing against the left-hand (southern) side of the valley, following the dip of the rock strata, with the edge of a broadening flood plain evident to the right; in the trees upstream it is actively undercutting the valley side. The channel is still very bouldery, and it can obviously only move most of this coarse bedload when the water table is much higher, and the river is in spate.

The first clue lies in the bouldery layer visible in the bed and banks of the river, which is highly permeable. Boreholes have shown this bouldery layer is several metres thick, and is a product of glaciation. During the Pleistocene the valley was widened and deepened by a fast-moving valley glacier. Some of this ice had spilled over at Torpantau from the glacier occupying the Taf fechan valley, some gathered on the high ground to the north-west and north, where heavy snowfall was whipped by strong westerly winds, and some accumulated in nearby hollows facing north and east and shaded from the sun, like those at Darren fach and Pant y creigiau. Fuelled by rapid accumulation and steep gradients, lubricated by meltwater, and armed with large quantities of coarse, angular sandstone fragments plucked and scraped from the steep slopes, the ice coalesced, thickened and flowed sufficiently quickly to gouge out the floor of the Caerfanell valley. Then it would have melted (ablated) equally quickly as it approached the low ground of the Usk valley, the position of its front fluctuating widely according to prevailing temperatures. For a time the overdeepened part would have been a lake (– a precursor of the modern reservoir?).

The debris the glacier was carrying filled up the overdeepened middle part of the valley when the ice melted back, and this is what produced our bouldery layer. Meltwater from ice in the Nant Cynafon valley also swept frost-shattered stones, sand and silt southwards into the lake in the main valley before it drained away. All this material allows the river water to drain into the interstices between the stones, and explains the overall reduction in discharge. But to explain the short-term variations in river level we need more.

The presence of the Talybont Reservoir less than one kilometre downstream provides our second clue. The construction of the dam and the subsequent inundation of three kilometres of the valley bottom raised the level of the water table in the valley, and so affected the river and its channel processes both upstream and downstream. Most importantly for our argument the reservoir level fluctuates by several metres during a normal year: any rise is communicated some way upstream on the comparatively gentle gradients here. A reservoir does not behave like a natural lake, where changes in level are usually small, long term and slow: it may fluctuate quite widely over a period of months or even weeks. When the reservoir is full, the water table is high, and some of the reservoir water moves up river; the voids in the river bed are quickly filled, along with the river channel. When the reservoir level is falling the river lengthens and the water table progressively lowers. This allows the river to drain increasingly into its bed until it may disappear altogether, only reappearing as springs at the delta and under the reservoir.

The variable river level and lower gradient of the channel above the reservoir reduce the energy available in the stream and so affect channel processes. The river has a smaller capacity to carry sediment and a lesser competence to move the larger glacially-derived stones, which means that sediment transport and the erosion resulting from it are reduced. The long wetted perimeter in a bouldery stream at low water levels increases friction, slows the river, and causes it to drop even more of its load, hence the very stony channel here.

The reservoir has an effect on the upper Caerfanell similar to that of sea level at the coast: it acts as a local base level below which the river cannot cut. So any excess energy tends to be used for lateral erosion on the outside of bends rather than down-cutting, while sediment is deposited on the banks when in flood, in the channel as point bars, and on the inside of bends, levelling and aggrading the valley floor and beginning to form a flood plain. The channel sometimes chokes with debris and becomes braided, with evidence of abandoned former channels and cut-offs, but no true meanders. All of this, ironically, is reminiscent of meltwater streams at the snout of a melting glacier. It is in marked contrast to the valley incision we have observed upstream, and to the situation existing before the dam was built, when the Caerfanell`s base level was effectively the River Usk and the valley floor would have had a continuous gradient.

The finer sediment is more readily carried onwards to the reservoir, where it has produced a marked delta which is visibly extending year by year, so reducing the reservoir’s capacity. The worry is that this process is accelerating as footpath, peat and gulley erosion increase in the Beacons – part of the reason for the remedial and conservation work on the upper Caerfanell footpaths. At low water levels it is possible to inspect the sides of the channel approaching the delta and to identify thin layers of alternating coarser and finer sediments. These graphically represent the seasonal and annual variations in discharge and water level since the reservoir filled in 1938.

10a. The reservoir delta of the Caerfanell at low water.   10b.View of the Caerfanell valley from Talybont Dam.  In  the 1976 Drought, Talybont Reservoir was one of very few in England and Wales that never dried up. In photo 10b the spur that narrows the valley (normally it is below water level) is part of a former glacial moraine; it was the original site of the dam, but the water was too shallow, and it often drained into the  bouldery deposit in the floor beneath, so the present larger dam was built 300 yards down the valley.

The construction of the Dam and Reservoir has affected the River Caerfanell downstream quite considerably. Firstly, water is abstracted and piped down valley to Newport, so discharge is now reduced overall. The river rarely, if ever, floods as much or as frequently as it used to before the dam was built. Secondly, discharge is not just dependent upon local climatic and edaphic (ground) conditions, but also on demand at Newport, fifty kilometres down the Usk. While there is still a seasonal rhythm to the river’s regime, short term variations are evened out at the dam, especially when the reservoir is less than full. Water is required by law always to be released into the river to keep it flowing downstream, even in very dry periods, but the amount of this compensation water varies seasonally:-

Nov-Apr: 5.5m gallons/day, May-Jul: 4m gals/day, Aug-Oct: 3m gals/day*.

Thirdly, the reservoir acts as a settling tank for sediment brought down by the Caerfanell. There is little load other than that small and light enough to be stirred up and carried in suspension, and with no bedload little mechanical erosion can take place. Stones in the river bed are therefore mainly ‘fossil’ bedload and the new river regime is rarely able to move them. Fourthly, the stabilised regime and clear, well-oxygenated water makes for a more favourable biotic environment, so there is a wider range of aquatic plants and animals here.

* Imperial figures published by Welsh Water. To convert to more useful metric flow measurements,
1 million gallons per day is approximately 0.06 cubic metres per second (cumecs).

      The Lower Caerfanell – Deposition by Ice and Water:

The view downstream from the dam is quite different from that upstream, and not just because there is no reservoir (see figures 5,6 and 7). The valley is lower and the river is hidden from view, incised some 10 metres or more in a narrow V-shaped valley. The uniform trough so characteristic of the middle Caerfanell loses its steep western slope and becomes noticeably less symmetrical. A large tributary (the Nant Clydach) comes into the Caerfanell here from the slopes of Bryn, at the end of the main Beacons escarpment. The eastern side of the valley remains as steep and uniform as ever, and the land use consists as before of conifers. However the gentler slopes opposite and much of the Clydach valley are occupied by farms and hedgerows, with the roadside hamlet of Aber (which means river mouth or confluence) at the valley junction.

These gentler slopes are composed of fluvio-glacial material deposited near the farthest point reached by the Caerfanell ice during the Late Devensian glaciation. The ice was stagnating – depositing rather than eroding – from here onwards, and this seems to be supported by the fact that the valley has lost its clear cut glacial trough shape. Analysis of the deposits shows that some stones have come from rocks many miles to the west and north of Talybont. This ‘erratic’ material suggests that ice from the Usk glacier penetrated the lower Caerfanell valley when the Caerfanell ice was on the retreat, possibly reaching as far as the Nant Tarthwynni. The river is cutting down into that material, leaving the prominent terrace followed by the road from the Dam to Cross Oak.

11a. Lower Caerfanell valley: Vertical air photo.         11b. Sketch map of lower Caerfanell at similar scale.
Talybont Dam is near the south-west corner, with the Afon Caerfanell flowing almost due north from it for a mile, before curving north-east around the prominent wooded slopes of Dan y wenallt. The river then passes through the village of Talybont-on-Usk near the centre, and continues north-east for 3/4 of a mile to its confluence with the River Usk. The trees marking the edge of the kame terrace in photo 13 are 1/2 mile due north of the village, with the Mon and Brec Canal obliged to wind partly round, and partly through, the obstacle.                

The lower part of the Caerfanell is effectively a new river, with the Talybont Reservoir as its source. It starts with a moderate and fairly even discharge with little or no sediment, rather like a stream emerging from an underground cave system. It is flowing in a narrow valley incised into marked benches of glacial and fluvioglacial debris in the bottom of the glacial trough. This trough now becomes wider and less symmetrical. The gradient has increased again (as a result of the rejuvenation mentioned in the previous section), and both discharge and sediment load increase quickly as springs and tributaries join in number. The Afon Clydach is the first, a stream now relatively much more important to the behaviour of the lower Caerfanell than it was before the dam was built. The Clydach has a high density branching network draining the boggy hillsides on the flanks of Bryn, and as a result it has a ‘flashy’ regime (rising and subsiding rapidly after heavy rain), and a stony channel. The Cui and Coity bach streams that join nearer Talybont behave in a similar fashion.

 
12a. Caerfanell valley below Talybont Dam.                         12b. The incised Afon Caerfanell ( in trees on the right). Note the steep far side and the gentler near side.            Note the steep edge of the fluvio-glacial terrace (left).

Approaching Talybont and the Usk Valley the main river and the lower reaches of its tributaries were rejuvenated when the Usk glacier finally disappeared. They began actively eroding again, with alternating ‘riffle and pool’ sections that often expose the smooth sandstone bedrock. There is evidence of bank undercutting, particularly on the eastern side (uniclinal shifting again), with marked fluvio-glacial terraces on the west, but there are no real meanders, and little flood plain. The valley as a whole is still asymmetrical, with the river mostly hugging the eastern (outer) side of the curving valley. The river is quite fast-flowing still as a result of the steepening of gradient below the dam, and the relatively smooth channel, but it normally does not seem to be carrying much sediment, for reasons we have seen; the exception is when the river and its three main tributaries are at bankfull.

As the Caerfanell valley begins to open northwards into the Usk there is an Ice Age feature that is important not only for its effects on the river and its valley, but also for the site and morphology of Talybont village. At the junction of the two valleys the fluvio-glacial terrace on the west side of the river widens out into a prominent sloping triangle of outwash and partly sorted boulder clay. The base of the triangle abuts the main valley side between Coity mawr and Cross Oak; its attenuated apex juts out sharply into the middle of the Usk flood plain at Gilestone. It has caused the Caerfanell to turn north-eastwards, and the Usk to swing against the northern side of its valley. This fluvio-glacial feature is a kame, deposited where meltwater in the Caerfanell valley was ponded up as a pro-glacial lake against the much larger glacier that occupied the valley of the Usk. Near the village it is used for pasture, but at Gilestone it is a stony and infertile spur, picked out by trees planted where nothing much else would grow.

13. The edge of the Gilestone kame terrace in Talybont-on-Usk village, jutting out into the Usk flood plain.        Note the contrast in land use between the former productive meadow land in the foreground (now used for recreation), and the less fertile wooded ridge leading directly out for half a mile towards the River Usk, with Gilestone Farm located at its tip. The kame causes the Usk to swing against the north side of its valley, producing the steep slopes of Allt yr esgair in the background.
 

Meanwhile, the river Caerfanell is still actively downcutting in order to reach the Usk floodplain. The Usk is lower because the much larger Usk glacier deepened its valley more than did the Caerfanell. The latter passes under an aqueduct carrying the Canal, and  Talybont’s eponymous road bridge,before finally reaching the Usk floodplain near the former Talybont railway station. It then meanders north-east to its confluence near Llansantffraed, some 600 metres lower than, and 15 kilometers distant from, its source. Here for the first time are classic river flood plain features, but these are part of the Usk flood plain rather than that of the Caerfanell.

 

14. The Caerfanell Valley meets the Usk flood plain at Talybont (viewed from Allt yr esgair to the north).            The  Brecon Beacons are to the right (west); the Glyn trough approaches us from the south, the Usk flows from right to left across the bottom of the picture. On the right of the picture is a kame terrace (a triangle of stony  material deposited by the meltwater from the Usk glacier). The trees in the right foreground mark the tip of the kame, pushing the River Usk to this side of its broad  valley, and deflecting the confluence of the Caerfanell with the Usk just out of the photo to the left. In the centre of the picture, the kame terrace also provided a firm, dry site for Talybont village, the old Usk road through it, and the Monmouth and Brecon Canal.

 Conclusions:   (please refer to the 'Sketch map of the Caerfanell Basin' at the top of the page)

From the dam to the Usk the Lower Caerfanell runs almost due south to north, against the dip of the rocks, and against the regional geological and physiographic trend. It is more or less parallel to streams that drain the scarp face of the Brecon Beacons further west, and shares some of their characteristics. However, its source is effectively Talybont Dam, which largely governs its present discharge, and its gradient and valley shape reflect its proximity to the Usk, whose ice advanced up to three miles into the Caerfanell valley, and was responsible directly and indirectly for much of the material into which the Caerfanell is incised.

It is diametrically opposed to its own headwaters, which flow as misfit streams from north to south in a distinctive glacial and periglacial landscape, following the southward dip of the rocks. At Blaen y glyn in particular, active and dramatic downcutting is the main fluvial process, accompanied by intense weathering and slope instability. 

It is also at odds with the west to east curve of the steep-sided Glyn which connects the upper and lower sections. Here in its middle section the discharge of the stream and gradient of the valley is at its lowest, and the fluvial processes are partly governed by the deep layer of permeable glacial material covering the floor of the overdeepened trough, and partly by the water level in the Reservoir. It may be significant that it is here close to the Dyffryn Crawnon, which follows the line of an ancient, but still occasionally active, deep-lying fault zone aligned west-south-west to east-north-east across South Wales from Swansea Bay to beyond the English border.

We have seen that the 'surprising' features of the Caerfanell Basin outlined at the start are mainly a result of the varying effects of glaciation. When the character and orientation of the upper, middle and lower parts of the Caerfanell are compared, it seems as if we have not one river, but three separate ones that have become joined up.  If this is the case, then it would seem likely that glaciation had a major part to play in that too.