The abutment ends the bridge and starts the road. It holds the last span, retains the approach fill, and, through the joint and bearings at its seat, absorbs the movement of the whole deck. Seat-type abutments with a backwall and wingwalls are the traditional answer; stub abutments on piles sit at the top of a fill slope; mechanically stabilised earth (MSE) walls carry the fill while a separate footing carries the bridge.

seat-typeintegralstub on MSE wall

The trend of the last thirty years is the integral abutment: no joint, no bearings; the girders are cast into the abutment and the whole structure moves as one, the piles flexing with it. It removes the two components that cause most maintenance, at the cost of a deck length limit (roughly 100–200 m depending on climate and material) and careful design of the approach slab. Semi-integral versions keep bearings but drop the joint.

Loads: earth pressure (active, at-rest or passive depending on how the wall moves), live-load surcharge, the bridge reactions, and during earthquakes the inertia of the backfill. Stability is checked for sliding, overturning and bearing.

A pier is chosen for its site as much as its load. Multi-column bents suit wide decks on land; single hammerhead piers keep a small footprint in medians and water; wall piers resist stream flow and ship impact; tall hollow piers carry viaducts across valleys. The cap distributes the bearing reactions to the columns and is designed as a deep beam or by strut-and-tie.

multi-column benthammerheadwall pierhollow pier

Columns are compression members under bending in two directions, with slenderness effects that become significant above height-to-width ratios around 10, and with ductility requirements in seismic zones that dictate spiral or hoop confinement, lap-splice locations, and the amount of longitudinal steel. Interaction diagrams, axial load against moment, are the design tool, and the site's pier tool draws them from real bar layouts.

Vehicle collision on piers next to roadways and vessel collision on piers in navigable water are extreme-event checks that often govern pier dimensions and require protection systems rather than strength.

For vehicle collision, AASHTO 3.6.5 requires piers within 9.1 m of the edge of the roadway either to be protected by an embankment or a suitable barrier, or to resist an equivalent static force of 2,670 kN acting 1.5 m above the ground, in any direction between 0 and 15° to the edge of the pavement. Slender columns and open bents often cannot take this unprotected, so the force frequently decides between a wall pier, a crash wall between the columns, or a barrier in front of them.

Where competent soil or rock is within a few metres, a spread footing is the cheapest foundation there is: a concrete pad wide enough that the bearing pressure is below what the ground can carry, and stiff enough to spread the column load. The checks are bearing resistance (from the soil's shear strength, with φ around 0.45–0.55), settlement at service load (usually the governing check on soil), eccentricity (the resultant must stay within the middle of the footing under strength loads), and sliding.

Footings on erodible soil in rivers are the classic scour victim; AASHTO requires the footing bottom below the check-flood scour depth. Structural design of the footing itself is one-way and two-way shear and bending as a cantilever from the column face.

When the good ground is deep, the load goes down to it through driven piles, steel H-sections, pipe piles, precast prestressed concrete, timber, or drilled shafts (bored piles), cast in a hole that may be 1–3 m in diameter and 30–80 m deep. Driven piles are cheap and their capacity is confirmed as they are driven (by dynamic testing and driving formulas); shafts carry huge loads on a single element, suit noise-sensitive sites and large lateral loads, and are the foundation of most long-span towers.

driven pile group: side friction + end bearingdrilled shaft with reinforcing cage

Axial capacity is side friction plus end bearing, estimated from soil strength by the α (total stress, clay) and β (effective stress, sand) methods or from rock strength for sockets, then verified by static or dynamic load tests. Lateral capacity governs piers under seismic or ship-impact loads and is analysed with p–y curves. Group effects reduce capacity when piles are close, and negative skin friction (downdrag) adds load when the surrounding soil settles.

Moving water removes the bed around a pier. Three mechanisms add up: long-term degradation of the channel; contraction scour where the bridge narrows the flow; and local scour, the horseshoe vortex that digs a hole at the pier itself. Local scour depth scales with pier width and with flow depth and velocity; HEC-18 gives the equations, and the site's tool applies them.

Scour is the leading cause of bridge collapse in the United States and among the leading causes everywhere. The Schoharie Creek bridge (New York, 1987) fell when a flood undermined a spread footing that inspectors could not see; the responses were the National Scour Program, scour-critical ratings, and the rule that foundations are designed for the 100-year scour and checked for the 500-year. Countermeasures are riprap and collars for existing bridges and deep foundations for new ones.

Bridges survive earthquakes by design of the substructure. The philosophy of capacity design: decide where the structure will yield (the column ends), make those regions ductile with closely spaced confinement, and make everything else, cap, footing, joint, shear, stronger than the overstrength moment the hinge can deliver, so that the chosen hinge forms and nothing brittle fails first.

Then keep the deck on: seat widths sized for the displacement demand plus a margin, restrainers and shear keys where needed, and joints and bearings that can move or fail without dropping the span. Isolation bearings (lead-rubber, friction pendulum) lengthen the period and cut the demand, and are common on retrofits and on important bridges. Liquefaction of loose saturated sand is the foundation hazard: it removes the ground's support and can move whole abutments, and is addressed by ground improvement or by designing piles to survive it.

In numbers: US practice under the AASHTO Guide Specifications for LRFD Seismic Bridge Design (2009, second edition 2011) places each bridge in a Seismic Design Category A to D by its one-second spectral acceleration S_D1 (below 0.15, 0.15–0.30, 0.30–0.50, and 0.50 or more), and in categories B to D checks displacement capacity against demand rather than relying on force-reduction factors. The displacement ductility demand is limited to about 5 for single-column bents and 6 for multi-column bents, and the longitudinal steel of ductile columns is usually 1–4% of the gross area. The plastic hinge length that turns curvature into rotation is commonly taken as L_p = 0.08L + 0.022f_yed_bl ≥ 0.044f_yed_bl (mm, MPa), with L the distance from the hinge to the point of contraflexure: for an 8 m cantilever column with 32 mm bars of expected yield strength 470 MPa, L_p = 640 + 330 ≈ 970 mm. Capacity-protected elements are designed for the overstrength moment, taken as 1.2 M_p for ASTM A706 bars.

Seat width is the last line of defence. AASHTO 4.7.4.4 gives the minimum support length as N = (200 + 0.0017L + 0.0067H)(1 + 0.000125S²) mm, with L the length of deck to the next expansion joint and H the pier height, both in mm, and S the skew in degrees, applied at 75–150% depending on the seismic zone. For L = 100 m, H = 10 m and a 30° skew, N = 437 × 1.11 ≈ 486 mm.

A single-column pier stands on a spread footing on dense sand. The footing is 6.0 m long in the direction of the moment, 5.0 m wide and 1.5 m thick, with its base 2.5 m below ground. The geotechnical report gives a nominal bearing resistance of 1,200 kPa, calculated for the effective footing dimensions, and a friction angle of 35° for the sand. The loads below act at the base of the footing and include the footing and the soil above it.

1. Eccentricity. e = M/V = 1.0 m in the maximum case and 8,000/6,500 = 1.23 m in the minimum one; the lighter case is the more eccentric, which is why both must be run. Both are inside L/4 = 1.5 m, the limit that earlier AASHTO editions used for footings on soil; current editions (10.6.3.3) allow the resultant anywhere in the middle two-thirds of the base, e ≤ L/3 = 2.0 m.

2. Bearing. On soil AASHTO uses Meyerhof's effective area: the load is taken as uniform over a reduced footing L′ = L − 2e, centred on the resultant. In the maximum case L′ = 4.0 m, A′ = 20 m² and q = 9,000/20 = 450 kPa, against φq_n = 0.45 × 1,200 = 540 kPa, a ratio of 0.83. The nominal resistance must itself be calculated for the effective dimensions, since bearing capacity depends on them. On rock the pressure is taken as linear instead, q_max = (V/A)(1 + 6e/L), which here would be 300 × 2.0 = 600 kPa; mixing the two methods gives either a false pass or a false fail.

3. Sliding. The resistance is friction on the base, V tan δ, with δ equal to the soil's friction angle for concrete cast directly on sand, so tan δ = 0.70. It must be calculated with the minimum vertical load: 6,500 × 0.70 = 4,550 kN, times φ_τ = 0.80, gives 3,640 kN against a demand of 1,200 kN. Using the maximum vertical load here is a classic error, and an unconservative one. Passive pressure on the front of the footing is usually ignored, since it needs movement to develop and the soil in front may be excavated or scoured away.

4. Settlement. At service the average pressure is 6,000/30 = 200 kPa, and a Schmertmann or Hough calculation from the SPT or CPT profile gives, say, 20 mm, most of it immediate on sand. AASHTO's commentary (C10.5.2.2) cites angular distortions of 0.004 for continuous spans and 0.008 for simple spans as tolerable, which for 40 m continuous spans is 160 mm of differential settlement between neighbouring piers; joints, bearings, ride quality and drainage usually set a much tighter project limit. On soil this check, not bearing, sizes the footing more often than not.

5. Structure. Finally the footing is designed as a reinforced concrete cantilever from the column face under the factored net soil pressure: one-way shear at d_v from the face, punching shear on a perimeter d_v/2 from the column, flexure with bottom steel both ways, and development of the column bars. In seismic regions it is capacity-protected, designed for the column's overstrength moment.

Every way of predicting pile capacity has its own uncertainty, and AASHTO prices it into the resistance factor. The static formulas of section 7.4, side friction plus end bearing from soil strength, are indispensable at the design stage for choosing pile type and length, but measured capacities scatter widely around their predictions, and their resistance factors are low. Field methods measure the pile that was actually installed, and earn higher ones.

What it means in practice. Suppose a 500 mm square prestressed pile must carry a factored axial load of 1,800 kN. It is driven until it demonstrates a nominal resistance of 1,800/φ, where φ belongs to the field method used to confirm it. With the Gates formula the target is 4,500 kN; with wave equation analysis alone, 3,600 kN; with dynamic testing and signal matching, 2,770 kN; with a static load test backed by dynamic tests, 2,250 kN. Same pile, same load, and a factor of two in the resistance it has to prove, which shows up as longer piles, more piles or a bigger hammer. The Engineering News formula would ask for 18,000 kN, which explains why it survives only in old specifications. On large projects a test programme usually pays for itself.

The tests. A static load test jacks a pile against reaction piles or kentledge and measures load and settlement directly. A bidirectional test casts a sacrificial jack (the Osterberg cell is the best-known) low in a drilled shaft and pushes the upper part up against the lower part down, so no reaction system is needed and loads of several tens of meganewtons become testable. Dynamic testing instruments the pile with strain gauges and accelerometers during driving or restrike, and signal matching separates shaft from toe resistance. Drilled shafts are checked for integrity by crosshole sonic logging or thermal integrity profiling, and have their own resistance factors (AASHTO Table 10.5.5.2.4-1), rising to 0.70 when a static load test is made.

Time matters. Driving remoulds the soil around a pile. In clays and silts the excess pore pressure dissipates over days and the resistance grows, sometimes doubling; this is set-up. In dense fine sands and some shales the opposite can happen, relaxation. A restrike test a few days after driving, compared with the end-of-driving record, measures the effect, and set-up confirmed by restrike is often the cheapest capacity a project will find.

Braking, wind, the thermal movement of integral bridges, earth pressure, ship impact and earthquakes all push sideways on foundations, and piles resist by bending against the soil. The standard model treats the pile as a beam on a bed of nonlinear springs: at each depth a p–y curve relates the soil reaction p, a force per unit length of pile, to the lateral deflection y. The curves depend on soil type, depth, pile width and whether the loading is static or cyclic. The classic families are Matlock's for soft clay (1970), those of Reese and co-workers for sand and for stiff clay from the mid-1970s, and the API curves used offshore; the beam-on-springs problem is then solved numerically with programs such as LPILE, GROUP or FB-MultiPier.

Where the action is. A laterally loaded pile is decided near the surface. For intuition, treat the soil as linear. A 600 mm steel pipe pile with a 12.7 mm wall (EI ≈ 200,000 kN·m²) in medium dense sand, with a rough soil modulus of 20 MPa, has a characteristic length 1/λ = (4EI/k)^1/4 of about 2.5 m. Free-headed and loaded by 100 kN of shear, it deflects about 4 mm at the top and has its largest moment, about 80 kN·m, some 2 m down; below about 8 m it barely notices the load. Fixed into a rigid cap, the same pile deflects half as much but takes its largest moment, about 125 kN·m, at the junction with the cap, which is where its reinforcement or wall thickness must be. A p–y analysis refines these numbers but rarely changes the picture.

What takes the soil away. Anything that removes or weakens the upper soil changes the answer more than any refinement of the curves. Scour lowers the ground line exactly where the resistance was, so river piers are analysed with the design-flood scour removed at the strength and service limit states and the check-flood scour at the extreme event (AASHTO 3.7.5). Group shadowing: piles in trailing rows push into soil already loaded by the row in front, and AASHTO (10.7.2.4) reduces their resistance with p-multipliers which, at a centre-to-centre spacing of three diameters, fall from about 0.7 for the leading row to about 0.35 for the third and later rows, and recover to 1.0 and 0.7 at five diameters. Cyclic loading softens clays and opens a gap behind the pile. Liquefaction removes most of the resistance of the liquefied layer and, where the ground slopes toward a river, lets a crust of unliquefied soil spread sideways and push the piles; that is how the Showa Bridge in Niigata lost its spans in 1964.

Battered piles are efficient against static horizontal load because they turn it into axial force, but they are stiff, attract large forces in earthquakes and are bent by settling soil, so many seismic codes and owners discourage them.

A pile is normally helped by the soil around it, which holds it up through side friction. When the soil settles more than the pile, the friction reverses and the soil hangs on the pile instead. That reversed friction is negative skin friction, the force it adds is the drag load, and the settlement it drags the pile into is downdrag. The usual causes at bridges are a new approach embankment on soft clay, recent fill still consolidating, a lowered water table, and ground settlement after liquefaction.

The neutral plane. Going down the pile, the soil's settlement decreases while the pile's stays nearly constant. At some depth the two are equal: the neutral plane. Above it the soil moves down relative to the pile and friction acts downward; below it the pile moves down relative to the soil and friction acts upward, together with the toe resistance. The axial force in the pile is greatest at the neutral plane, where it equals the permanent load plus the drag load. Because pile and soil move together there, the pile head settles as much as the soil does at that depth, plus the shortening of the pile. In Fellenius's unified method this is the central point: downdrag is mainly a settlement problem, and transient live load is not added to the drag load, because a live load pushes the pile down relative to the soil and temporarily removes the negative friction.

A quick number. A 600 mm steel pipe pile passes through 12 m of soft clay with an undrained strength of 25 kPa, under a new 5 m embankment, and ends in dense sand at 30 m. If the whole clay layer settles past the pile, the drag load is roughly perimeter × thickness × unit friction = 1.885 m × 12 m × 25 kPa ≈ 570 kN. AASHTO treats this as a load, DD, with a maximum load factor of 1.4 when it is computed by the α-method (1.05 by the λ-method and 1.25 for drilled shafts by the O'Neill and Reese method), and the factored resistance below the downdrag zone must carry it together with the other factored loads. On a pile designed for 1,500 kN of permanent load, 570 kN of drag is significant but not dramatic. The settlement that comes with it may be the larger problem, because a pier that follows its piles down 50 mm takes the deck with it.

What to do about it. The best cure is to deal with the soil before the piles go in: build the embankment early, surcharge it and speed its consolidation with wick drains, so that most of the settlement has happened before the piles are driven; or replace heavy fill with lightweight material such as EPS geofoam or foamed concrete. Where that is not possible, a slip coating of bitumen over the length of pile in the settling layer, or a sleeve around it, cuts the negative friction substantially, and a deeper toe in firmer ground reduces the settlement. At abutments there is a second, related danger: soft clay under a new embankment squeezes sideways as well as down, bending the abutment piles and pushing the abutment toward the span.

Section 7.1 explained why integral abutments are popular; this section is about what makes them work. The deck expands and contracts with temperature, and with no joint the movement goes into the abutments, their piles and the soil behind them. For a 120 m concrete bridge with α = 10.8 × 10⁻⁶/°C and a design range of 39 °C (AASHTO Procedure A for a moderate climate, −12 to 27 °C), the total change in length is 120,000 mm × 10.8 × 10⁻⁶ × 39 ≈ 50 mm, about 25 mm at each end of a symmetrical bridge, plus a one-way shortening of 10–20 mm at each end from shrinkage and creep. A steel bridge of the same length, with its wider temperature range, moves nearly twice as much.

Piles. The abutment sits on a single row of flexible piles, so that it can translate and rotate without attracting large forces. In US practice these are usually steel H-piles, often oriented for weak-axis bending, sometimes driven through pre-bored holes 2.5–3 m deep filled with loose sand to soften the top. They are designed for axial load combined with the bending from the imposed movement; some owners accept yielding at the pile head and treat it as a low-cycle fatigue problem governed by the annual temperature cycle.

Soil behind the wall. When the bridge expands in summer it pushes the abutment into the backfill and mobilises pressures well above at-rest; when it contracts in winter the backfill slumps a little into the gap, and the next summer the wall pushes into denser soil. Over the years this ratcheting raises the earth pressure. UK practice (PD 6694-1, which replaced BA 42/96) gives an enhanced pressure coefficient K* that depends on the wall movement; US agencies use passive pressure or a fraction of it. A compressible geofoam layer behind the wall limits the pressure.

The approach slab and the bump. The fill behind an abutment almost always settles relative to a pile-supported abutment: the foundation soil consolidates under the embankment, fill placed in the cramped zone behind the wall is hard to compact, water through leaking joints erodes it, and an integral abutment's movement loosens the zone nearest the wall. The result is the bump at the end of the bridge, the subject of NCHRP Synthesis 234 (1997). An approach slab, typically 6–9 m long and around 300 mm thick, hinged to the abutment and resting on a sleeper slab at its far end, bridges the settling zone and turns a step into a gentle slope. With an integral abutment the approach slab moves with the bridge, so the movement joint moves to the far end of the slab, between it and the pavement, where it has to be designed as carefully as any bridge joint.

The cures are unglamorous and effective: select granular backfill compacted in thin lifts; drainage behind the wall; preloading or wick drains under the embankment where the subsoil is soft; an approach slab long enough to span the zone that will settle; and, when settlement still comes, slab-jacking with grout or polyurethane. Most agencies also limit integral abutments in skew, often to about 30°, and in length, and tie the approach slab to the abutment so that it moves with it.

On 9 May 1980 the freighter Summit Venture, caught in a squall, struck an anchor pier of the Sunshine Skyway Bridge across Tampa Bay. More than 300 m of the southbound bridge fell, and 35 people died. The response was the AASHTO Guide Specification and Commentary for Vessel Collision Design of Highway Bridges (1991, second edition 2009), whose method is now part of AASHTO LRFD Article 3.14. Its central idea is that ship collision is a matter of risk, not of a single design load.

Annual frequency of collapse. For every pier and span within reach of shipping, Method II computes AF = N × PA × PG × PC × PF. N is the annual number of vessels, by type, size, direction and loading; PA the probability that one goes off course, the aberrancy rate, from the accident record of the waterway; PG the geometric probability that an aberrant vessel hits that pier; PC the probability that the bridge collapses if it does, from the ratio of the pier's strength to the impact force; and PF an adjustment for any protection. Summed over the bridge, AF must not exceed 0.0001 per year (one in 10,000) for critical or essential bridges and 0.001 for typical ones.

Impact force. AASHTO's equivalent static force for a ship striking a pier head-on is P_s = 1.2 × 10⁵ V√DWT newtons, with V in m/s and the deadweight tonnage DWT in tonnes. A 40,000 DWT bulk carrier at 5 knots (2.57 m/s) gives 1.2 × 10⁵ × 2.57 × 200 ≈ 62 MN. For comparison, a highway pier beside a road is designed for a vehicle collision force of 2,670 kN, some twenty-three times less. Forces of this size make it uneconomic to strengthen most piers to survive a direct hit by the largest design vessel; the answer is to keep the ship away from the pier.

Protection, in order of preference: put the piers where the design vessel cannot reach them, in water too shallow for it, by lengthening the main span; build dolphins (large cellular sheet-pile or pile-group structures) or artificial islands that stop or ground a ship before it reaches the pier; and, for small craft and barges, fender systems that absorb energy. The replacement Sunshine Skyway (1987) has a far longer main span than the old one, and its main piers are guarded by islands and dolphins.

The lesson was taught again on 26 March 2024, when the container ship Dali, having lost power, struck a main pier of the Francis Scott Key Bridge in Baltimore, opened in 1977, before the guide specification existed. The continuous truss collapsed and six members of a road-repair crew were killed. The NTSB later found that the bridge's annual frequency of collapse was almost 30 times the AASHTO threshold for critical bridges, and in March 2025 asked the owners of 68 bridges in 19 states to carry out Method II vulnerability assessments. Ships have grown faster than the bridges built to resist them, so old assessments must be repeated with today's traffic. In Eurocode practice ship impact is covered by EN 1991-1-7 (accidental actions).

A new bridge is designed to stand with the full computed scour around its foundations, and FHWA's HEC-18 rules out riprap or other armour as a substitute for founding new piers below the local scour depth. Countermeasures are for the many existing bridges whose foundations were not designed for scour and which analysis or inspection has rated scour critical. FHWA's HEC-23, Bridge Scour and Stream Instability Countermeasures (third edition, 2009), is the standard catalogue, in four families.

• Hydraulic: river training works that change the flow, such as guide banks (spur dikes) at abutments that move the scour hole away from the abutment, spurs and bendway weirs that push the current off an eroding bank, and check dams that stop a channel from degrading.
• Structural: changes to the foundation itself, such as underpinning, deepening footings, adding piles or a new pile cap, and streamlining a blunt pier.
• Armouring: protection of the bed with riprap, articulated concrete blocks, gabion mattresses, grout-filled bags or partially grouted riprap.
• Monitoring: fixed instruments (sonar, tilt sensors, sliding collars, float-out devices), portable instruments and underwater inspection after floods, tied to a plan of action.

Sizing pier riprap. HEC-23 sizes the median stone diameter as D₅₀ = 0.692(KV)²/[(S_s − 1)2g], where V is the approach velocity, K a pier-shape factor (1.5 for a round nose, 1.7 for a square one) further adjusted for the pier's position in the channel, and S_s the specific gravity of the stone, about 2.65. For a round-nosed pier in a 2.5 m/s current, D₅₀ = 0.692 × 3.75²/(1.65 × 19.62) ≈ 0.30 m; a square-nosed pier needs 0.39 m. The layer is placed with its top level with the bed, not mounded up, extends at least two pier widths from the pier face in every direction, is typically about three D₅₀ thick, and sits on a geotextile or granular filter. The filter matters: without it the flow draws fine bed material out through the voids, winnowing, and the stone sinks. Riprap also fails when stones are carried away, when scour at the edge of the apron undermines it, and when migrating dunes in sand-bed rivers pass beneath it.

Monitoring and the plan of action. Under the US National Bridge Inspection Standards every scour-critical bridge must have a plan of action: who watches which gauge, at what river stage or flow the bridge is closed, how it is inspected after the flood, and which countermeasure is scheduled. Two collapses show why maintenance matters as much as design. At Schoharie Creek (1987) the riprap around the footings had been depleted over the years and not replaced. At Entre-os-Rios in Portugal (2001) the Hintze Ribeiro Bridge fell when a pier, standing in a riverbed lowered by years of sand extraction, was undermined in a flood; a bus and three cars fell into the Douro and 59 people died.