The oldest method: build a temporary structure under the bridge, cast or assemble the permanent one on it, remove the temporary one. Falsework is economical where the ground is firm, the height is modest and nothing below must stay open, which describes most short viaducts and river bridges under 40–50 m span.

Falsework is also where construction accidents concentrate. It is designed by someone other than the bridge designer, often to a tighter budget, and it carries the full weight of wet concrete before the concrete has any strength. Injaka (South Africa, 1998) and the many scaffold collapses in the Failures library are the case for treating temporary works as real structures with their own reviews, and for the AASHTO Guide Design Specifications for Bridge Temporary Works.

Precast prestressed girders arrive by truck or barge and are lifted by crane, one or two mobile cranes for a road overpass, a floating crane for a river, then stabilised with temporary bracing until the diaphragms and deck tie them together. The critical moments are the lift itself (a slender girder can roll and buckle if picked at the wrong points) and the interval before bracing, when a wind gust has toppled more than one line of girders.

Where cranes cannot reach, a launching gantry straddles the piers and places girders span by span, moving itself forward on the completed deck, the method of long low viaducts and of urban metros. Erection tolerances, bearing seating and the sequence of deck pours (to control cracking over piers in continuous decks) are the details that make or break the schedule.

Start at a pier with a short pier-table segment. Hang a form traveller off each end. Cast a 3–5 m segment on each side, stress the cantilever tendons through the top slab, move the travellers, repeat. The two arms grow in balance so the pier is never seriously out of equilibrium, and the river, road or valley below is never touched. When the arms from adjacent piers meet, a closure segment joins them and continuity tendons in the bottom slab make the span a continuous beam.

Free-cantilever construction (FCM) built most of the world's concrete box-girder bridges between 70 and 300 m. Its engineering is in the stages: every segment changes the structure, the creep of concrete cast last month differs from concrete cast last year, and the deflection of the tip must be predicted so that the two arms meet at the right height. Unbalanced loads, one traveller ahead, a wind, an accidental drop, are checked with temporary pier fixity or props.

Cast the deck in 15–30 m segments in a fixed casting yard behind one abutment, and push the growing deck out over the piers on temporary sliding bearings. A light steel launching nose on the front reduces the cantilever moment as the deck reaches for the next pier. Every section of the deck passes over every pier, so the whole deck must carry both sagging and hogging moment during launching; it is prestressed centrally for the launch and then given its final tendons afterwards.

ILM needs a constant cross-section and a constant curvature in plan and profile (straight, circular or a single helix), spans typically 30–60 m, and a site where the casting yard can sit. In return it builds high viaducts and railway bridges over sensitive ground with almost nothing below the deck. Friction on the launching bearings sets the jack size; the launching nose and the pier-top guidance are the temporary works that matter.

An MSS is falsework that moves itself. A steel girder system, either under the deck or over it, carries the formwork for one full span, is supported on brackets at the piers, and after the span is cast and stressed it launches itself forward to the next. It suits long viaducts of 30–60 m spans with repetitive geometry, the standard-gauge railway viaducts of Europe and Asia and the elevated roads of the Gulf, where its high initial cost is spread over dozens of identical spans.

The deck can be cast span-by-span with a construction joint at about a fifth of the span past the pier, where the moment is near zero. The MSS itself, and the pier brackets it rests on, are the temporary works to review; a system launching in wind or with an uneven load has failed more than once.

Cast the box girder in 2–4 m segments in a yard, each against the previous one (match casting) so that the joints fit exactly, and assemble them on site with epoxy in the joint and post-tensioning through the segments. Erection is either span-by-span, a gantry holds all the segments of one span and they are stressed together, for viaducts of 30–50 m, or by balanced cantilever with a crane or gantry placing segments on each side of the pier for spans to about 150 m.

The method separates the slow part (casting) from the critical-path part (erection) and produces a factory-quality product; it built the Florida Keys, the Bangkok expressways, Dubai's and Doha's elevated roads, and much of the Hong Kong and Singapore rail network. Its risks are geometry control in the casting yard, where an error repeated over a hundred segments becomes a curve, and the durability of the joints and internal tendons.

Cable-stayed decks are built by cantilevering from the tower: each segment (steel, precast concrete or cast in a traveller) is added, its stay is installed and stressed, and the next follows. The stay forces are adjusted in stages so that the finished bridge has the intended geometry and force distribution, a problem solved by "backward analysis" from the target state. Side spans are often built on falsework first to hold the tower.

Suspension bridges go in order: towers, catwalks, main cable (aerial spinning wire by wire, or PPWS strands hoisted whole), cable bands and hangers, then deck segments lifted from barges by gantries riding on the cable, starting at midspan. The cable's shape changes as the deck is hung, so the sequence is planned so that the deck's curvature stays within what the joints can tolerate.

Arches are erected by cantilevering the ribs out on temporary stays from towers behind the abutments, or by building half-arches vertically and rotating them down; long-span steel arches have also been built by rotation in plan, cast the whole bridge parallel to a railway, then swing it 90° into place in a night. The Methods animations show each of these.

A pier in water is built in two stages that have little to do with the bridge above it: first get plant to the pier, then make a dry place to work. In shallow water, up to about 5–6 m, the usual answer to the first question is a temporary trestle: a working bridge of steel pipe piles and girders built alongside the alignment, from which crawler cranes drive the permanent piles as if they were on land. Deeper water, large tides and navigation channels call for floating plant: spud barges, jack-up platforms that stand on their own legs above the waves, and crane vessels. The choice is mostly about weather downtime: a jack-up keeps working in a swell that stops a barge.

The dry place is usually a cofferdam. Steel sheet piles are driven around the footing, and the inside is excavated and pumped down in stages, with a level of walings and struts installed at each stage before the next is dug. In sand or gravel the water will not stay out from below: the head difference drives seepage up through the base, and if the upward gradient approaches the buoyant weight of the soil the base boils and the excavation floods. The standard cure is a tremie seal, a slab of concrete placed underwater through pipes whose tips stay buried in the fresh concrete, which is then pumped dry and becomes the working floor for the pile cap.

Two rules come from cofferdams that failed. Design for a flood that can realistically occur during the construction period, agreed with the river authority, and have a plan for a larger one: openings or valves to flood the cofferdam deliberately before it overtops, so that the walls never carry a full head of water with an empty box behind them. And remember that the cofferdam narrows the channel, which raises the flood level upstream and the scour around the cofferdam itself.

Where the water is too deep for a braced cofferdam, the foundation becomes its own cofferdam. An open caisson, the well foundation of the great Indian and Chinese rivers, is a concrete or steel shell sunk by dredging inside it until its cutting edge reaches a firm stratum below the scour depth, then plugged with tremie concrete. A floating caisson is built in a dock, towed out, sunk onto a prepared bed and filled; the main towers of the Akashi Kaikyō Bridge stand on steel caissons about 80 m in diameter placed this way. On sea viaducts with many identical piers, precast pile cap shells are lowered over the driven piles and sealed, becoming formwork and cofferdam in one piece, which takes the most weather-sensitive work off the water.

A low pier is cast in one or two lifts in crane-handled formwork. Above 15–20 m the formwork has to stand on what it has just built. Climbing formwork (jump form) is anchored into the previous lift, carries its own working platforms, and is raised to the next position either by crane or, in self-climbing systems, by hydraulic rams that leave the crane free for reinforcement. A lift of 4–5 m every four to seven days is normal, and the construction joint between lifts is a planned detail with a roughened surface and lapped bars. Slipform removes the joints altogether: a form about a metre tall is jacked upwards continuously on climbing rods while concrete is placed around the clock, so that the concrete leaving the bottom of the form is just stiff enough to stand. It gains a few metres a day, but it suits constant or gently tapering sections, needs a guaranteed concrete supply, and does not forgive a stop.

Tall piers and pylons bring three problems that low piers do not. The first is geometry. Every lift is surveyed, because a small error repeated over 200 m becomes a large one, and surveyed at the same time of day: sun on one face of a tall concrete pylon bends it measurably, so readings are taken early in the morning or corrected for the measured temperature difference. The second is the temporary structural state. The inclined legs of an A-shaped or inverted-Y pylon cantilever outwards as they rise and bend under their own weight; temporary struts are installed between them at intervals and jacked to a set force, so that the legs stay on their theoretical line and the base sections are not overstressed before the crossbeam or apex closes the frame. The third is the stay anchorage zone, the most congested concrete on the bridge, where anchor pipes or steel anchor boxes must be set to a few millimetres because each one fixes the working point of a stay. Many recent pylons use prefabricated steel anchor boxes, acting compositely with the concrete walls, which carry the horizontal stay forces and take much of the congestion out of the formwork.

Footings, pylon bases and thick pier heads are mass concrete: the heat of hydration cannot escape, the core gets hot while the surface cools, and the difference cracks the surface. Specifications commonly follow ACI 301 in limiting the peak temperature to 70 °C (158 °F), above which delayed ettringite formation can damage the concrete years later, and the core-to-surface difference to about 20 °C (35 °F). The tools are low-heat binders with slag or fly ash, chilled water or ice in the mix, cooling pipes cast into the pour, insulated forms, and thermocouples that show the engineer what is happening inside the block.

The pier head is where the substructure meets the deck programme. A hammerhead cap cantilevering several metres each side needs heavy brackets or falsework at height; on urban viaducts and in accelerated construction the cap, and sometimes the whole pier, is precast in segments and lifted on, joined by grouted reinforcing bars in ducts or by post-tensioning.

Steel arrives in shop-welded pieces 12–30 m long, limited by what a road or a barge can carry, so steel erection is mostly about holding unstable pieces in the right place until they are joined into a stable whole. Where the ground below is reachable, girders are set by crawler or mobile cranes, with temporary towers under the field splices where a span is longer than one piece. Over water or deep valleys they are launched from an abutment, lifted as complete units by strand jacks or floating cranes, or built out by cantilever with derricks or lifting frames on the completed structure.

Stability governs. A single I-girder is a slender beam with a weak axis. Lifted at the wrong points, or left standing unbraced, it can buckle laterally under its own weight and a modest wind. Standard practice is to hold the first girder with the crane, or tie and guy it, until the second girder is set and enough cross-frames are bolted to make a stable pair, and to choose pick points and spreader beams so that the girder's buckling capacity in the lift is checked, not assumed. A braced pair is not automatically safe either. When the girder spacing is small relative to the span, the whole braced unit can twist and buckle together, a global (system) buckling mode that cross-frames do not prevent, and which may govern for narrow twin-girder units and single tub girders before the deck is composite. In 2002 a single-span steel tub girder footbridge at Marcy, New York, collapsed while its deck was being poured, killing one worker; the research that followed focused on the lateral-torsional stability of tub girders and their bracing before the deck becomes composite.

The deck pour is the critical load case. The non-composite steel carries the whole weight of wet concrete, forms and finishing machine. AASHTO LRFD Article 6.10.3 (constructibility) checks the flange stresses during the pour, including lateral bending of the exterior girder flanges caused by the overhang brackets that hang eccentrically from the girder, and web bend-buckling. On continuous girders the pour sequence, typically the positive-moment regions first and the pier regions last, is part of the design, because it changes both the stresses and the deflections. On curved and skewed bridges the cross-frames are detailed for a chosen fit condition, no-load fit, steel dead load fit or total dead load fit, which decides whether the webs end up plumb and how much force the erector needs to connect the frames.

Launching suits steel because it is light and strong in bending: spans of 60–100 m are launched without temporary piers, longer ones with them. Every point of the bottom flange passes over every roller, so the web must resist the concentrated roller reaction at any position, the patch loading check of EN 1993-1-5 Section 6, which often sets the web thickness or calls for longitudinal stiffeners near the bottom flange. The steel deck of the Millau Viaduct was launched from both abutments over its concrete piers, with temporary towers midway between them that halved the 342 m spans during the launch.

On a busy motorway or railway, the cost of building a bridge is often smaller than the cost of the disruption it causes. Accelerated bridge construction (ABC), in the FHWA's definition, uses planning, design, materials and construction methods to reduce the on-site construction time and the mobility impact of building or replacing a bridge. In practice it means building the bridge, or large parts of it, somewhere harmless and moving it into place in a short closure. FHWA's Every Day Counts programme spread the techniques widely in the 2010s.

• Self-propelled modular transporters (SPMTs) are computer-controlled multi-axle platforms that drive under a span built on temporary supports nearby, lift it hydraulically, carry it at walking pace and lower it onto the bearings, usually overnight. The axle hydraulics are grouped into three support zones, so the span sits on a statically determinate three-point support and a dip in the road does not twist it.
• Lateral slide (slide-in): the new span is built beside the old one, the old one is demolished in the closure, and the new one is pushed or pulled sideways on PTFE pads or rollers over temporary and permanent supports.
• Strand jacks lift or lower thousands of tonnes from a gantry on the structure: suspension bridge deck sections from barges, arch segments, whole spans.
• Floating cranes place complete spans where the water is deep and calm enough. The crane vessel Svanen, built to place the 110 m concrete spans of the Great Belt West Bridge in the early 1990s, later set the main girders of Canada's Confederation Bridge, single pieces of up to about 7,500 t.

The engineering is the same in every case: the move is a structural load case. A span carried by SPMTs is supported inboard of its bearings, so a prestressed or composite deck may see hogging moments it was never designed for and may need temporary strongbacks or extra tendons. A slide has a break-away friction higher than the sliding friction, and if the two ends do not move together the span racks; jacks are synchronised and the position is monitored continuously. Every lift has a dynamic allowance on the rigging and lifting points, a landing sequence, and a defined weather limit. The permanent design must anticipate all of this: lifting points and pick-up zones are cast in, the deck is checked for the transport support condition, and the closure joints between prefabricated elements (often grouted or filled with ultra-high-performance concrete) are detailed to be made quickly and inspected.

The measure of success is not tonnes moved but hours of closure. The FHWA's 2007 manual on SPMT bridge moves and its later slide-in guidance describe replacements in which a road lost a weekend instead of a construction season. The price is the temporary works, the staging area and the risk concentrated into one night, which is why every move is rehearsed, with a trial lift or a short trial slide, and has a written fallback for each failure that can happen halfway.

Every bridge is designed for its finished state, but it has to survive dozens of intermediate states on the way, each with a different structural system. Construction engineering is the work of making the chosen erection method safe and making the finished bridge arrive at the right shape and the right internal forces. AASHTO LRFD Article 2.5.3 requires that where a design assumes a particular construction sequence, that sequence is shown in the contract documents; in practice the designer demonstrates one feasible method and the contractor's erection engineer develops the real one.

Stage analysis builds the structure in the computer in the order it is built on site: each segment cast or erected, each tendon stressed, each support added or removed, each stay installed and adjusted, with the age of every piece of concrete tracked so that creep and shrinkage can be computed step by step. Two results matter. The first is the locked-in force state. A balanced cantilever bridge is built as a set of cantilevers and becomes a continuous beam at closure; creep then slowly redistributes the moments towards those of a beam built in one piece on falsework, increasing the sagging moment near midspan for years after opening. The second is the deflected shape at every stage, from which the setting levels are derived.

Camber is the reverse of every deflection still to come. A worked example for a 40 m simply supported composite plate girder, 1.8 m deep, with a non-composite steel moment of inertia of about 0.028 m⁴:

Each line is 5wL⁴/(384EI). The fabricator cuts the web plate to this curve; the first check on site is that the erected steel, before the pour, has lost about 26 mm of its camber and not 50. On a continuous girder the same sum is made at tenth points for the actual pour sequence.

Geometry control on a cantilever bridge is a loop: predict, set, measure, correct. Each segment is set at a level equal to the final profile plus the reverse of all later deflections, including creep to the end of the bridge's life and the deflection of the traveller under the wet segment. After casting and stressing, the tip levels are surveyed at a fixed time, early morning, before the sun puts a temperature gradient through the deck. The difference from prediction is fed back into the model, adjusting the concrete modulus or the creep parameters, so the next setting level is corrected and the errors do not accumulate. On cable-stayed bridges the stay forces are part of the loop. They are measured with load cells or by lift-off at the anchorage, or from the stay's vibration: for a taut string the tension is T = 4mL²f₁², where m is the mass per metre, L the length and f₁ the fundamental frequency, a quick check that needs only an accelerometer.

The common mistakes: a stage analysis not updated after the method changes, cambers computed for the design sequence rather than the real one, and a geometry error found at closure that could have been corrected ten segments earlier.

Temporary works are everything built in order to build the bridge and then removed: falsework, formwork, cofferdams, trestles, crane platforms, temporary towers, travellers, launching noses and gantries. They carry real loads, yet they are designed quickly, by someone other than the bridge designer, for a structure that will stand for weeks. That combination is why so many of the deaths in bridge construction have happened in a temporary state.

Britain turned this into a procedure after a run of falsework and box girder failures around 1970. The Bragg committee's final report (1975) led to BS 5975, now the code of practice for temporary works procedures and for the design of falsework. Its framework has spread well beyond the UK:

• A named temporary works coordinator (TWC) on every project, responsible for a register of all temporary works, for seeing that each has a design brief, a design and a check, and for issuing a permit to load before concrete is placed or a structure is loaded, and a permit to strike or dismantle afterwards.
• Temporary works supervisors on site, and a designated individual in the contractor's organisation who owns the system.
• Design check categories 0 to 3, rising with complexity: from a standard solution checked within the site team (category 0) to an independent check by a separate organisation for complex or novel schemes (category 3).

Design rules for falsework are in BS 5975 and EN 12812; in the United States the AASHTO Guide Design Specifications for Bridge Temporary Works and ASCE/SEI 37 (design loads on structures during construction) cover the same ground. The loads that are forgotten are the lateral ones: falsework designed only for gravity has nothing to resist wind, the surge of a concrete pump, a slightly out-of-plumb prop or an impact. BS 5975 therefore requires a minimum horizontal load, 2.5% of the vertical load, whatever the calculated lateral loads may be.

The pattern across these cases is consistent. The temporary or intermediate state, not the finished bridge, failed; something changed after the design without being re-analysed; and a warning (deflection, distortion, cracking) was seen and explained away. The defences are equally consistent: a single person accountable for temporary works, an independent check proportionate to the risk, permits that stop work until the check is done, and authority on site for anyone to stop the job.

A construction method is chosen, not discovered, and the choice usually comes down to five questions. What is below: firm dry ground, a road or railway that must stay open, a river with floods, the sea? How high is the deck? How long and how repetitive is the bridge, since the fixed cost of a gantry, an MSS or a casting yard must be shared among enough identical spans? What can be transported to the site, and what plant and skills are available locally? And how much time does the programme allow? Cost per square metre comes last, because it follows from the answers to the first four.

A worked example, with illustrative figures. A 1.2 km viaduct of 30 spans of 40 m, deck 16 m wide and 10–14 m above a dry wadi that floods a few days a year, straight and of constant section, with a road alongside for access. Four methods are feasible:

Score each option from 1 (poor) to 5 (good) against weighted criteria agreed with the owner: here cost 40%, programme 25%, construction risk 20% and durability and maintenance 15%.

• A: cost 5, programme 4, risk 4, durability 3 → 0.40×5 + 0.25×4 + 0.20×4 + 0.15×3 = 4.25
• B: cost 3, programme 3, risk 3, durability 5 → 3.30
• C: cost 2, programme 4, risk 4, durability 4 → 3.20
• D: cost 4, programme 2, risk 2, durability 5 → 3.25

Precast girders win comfortably, as they do on most low, accessible viaducts of this length. The value of the exercise is in the sensitivity, not the totals. Make the viaduct 8 km long and the segmental yard's fixed cost is spread over 200 spans: C becomes the cheapest and fastest, which is why the elevated roads and metros of Bangkok, Dubai and Doha were built that way. Put the deck 40 m above a river and A loses its crane access, B or incremental launching takes over. Shorten the programme and give the closure of the road below a cost per hour, and prefabrication moves up every time. A matrix only records judgement; the judgement is in the scores.

Three mistakes recur in method selection: choosing the method the contractor used last time rather than the one that fits the site; comparing methods on permanent quantities while ignoring the temporary works, which on a long-span or marine bridge can be a large share of the total cost; and fixing the method in the design so rigidly that a contractor with better plant cannot offer an alternative.