Water Supply: Demand, Sources, Reservoirs, Distribution, Treatment and Quality

Fire demand is a large rate for a short time. It decides pipe diameters and the fire reserve in the reservoir, not the yearly volume. Empirical formulas relate it to the population P in thousands. Always read the unit: four give a flow in litres per minute, the Indian formula gives kilolitres per day.

Worked example: a town of 64,000 people has P = 64, √P = 8. Kuichling gives Q = 3182 × 8 = 25,456 L/min; Freeman gives 1136 × (6.4 + 10) = 18,630 L/min; the Indian formula gives 100 × 8 = 800 kL/day. A fire storage in the reservoir is then the fire rate × the duration the authority adopts. The mistake to avoid is putting P = 64,000 instead of 64, which inflates Kuichling’s answer about thirty-fold.

The design period is the future year up to which a component is sized. The rule is economic: works that are costly and disruptive to duplicate, large civil works and buried pipes, are sized for 30 years or more (dams and large storage longer, up to about 50 years); mechanical and electrical plant and treatment units, which wear out, can be added in modules and are cheaper to extend, are sized for shorter periods of about 15 years. Sizing for the future needs a population forecast.

• Arithmetic increase: P_n = P₀ + n·x̄, x̄ the mean decadal increase; fits old, large, near-saturated cities.
• Geometric increase: P_n = P₀ (1 + r)ⁿ, r the mean decadal growth rate as a fraction; fits young, growing towns, but overestimates when room runs out.
• Incremental increase: P_n = P₀ + n·x̄ + n(n+1)/2 · ȳ, ȳ the mean of the second differences; it lets the increment itself change.
• Also: decreasing rate of growth, the logistic (S-curve), graphical extension and comparison with similar cities, the ratio method (the town’s share of a larger region’s forecast), and employment-based forecasts for a planned new town.

Worked example: counts of 40,000, 48,000 and 56,000 at successive decades give x̄ = 8,000. Three decades after the last count, P = 56,000 + 3 × 8,000 = 80,000. At 135 lpcd the average demand is 10.8 ML/d, the maximum day is 1.8 × 10.8 = 19.44 ML/d, and the maximum hour runs at 2.7 × 10.8 = 29.16 ML/d, which is 337.5 L/s.

Sources are surface water (rivers, lakes, impounding reservoirs, canals), groundwater (open wells, tube wells, springs, infiltration galleries), harvested rain, and, on the coast or in scarcity, desalinated or reclaimed water. A source is chosen on four tests: it must give the design quantity in the driest year, of a quality that treatment can correct at acceptable cost, at an elevation and distance that keep pumping and conveyance affordable, with legal right to the water.

An intake is the structure that draws raw water from the source. Its design aims at the best quality water, at all levels of the source, with little silt and debris.

• Reservoir or lake intake (intake tower): ports at several levels let the operator draw from the best layer, avoiding surface algae and bottom sediment.
• River intake: a submerged intake pipe or a well (jack-well) on the bank linked by inlet pipes; screens and low entry velocity keep out debris and fish. The inlet sits below the lowest water level but clear of the bed.
• Infiltration gallery or well: a perforated conduit in the river bed or bank takes water already filtered by the sand.
• Spring: a protected spring box over the eye of the spring, with a fenced catchment.

An aquifer is a saturated formation that stores and yields usable water (sand, gravel, fractured rock). An aquiclude (clay) holds water but does not transmit it; an aquitard transmits it poorly. In an unconfined aquifer the top of the saturated zone is the water table, open to the atmosphere. A confined (artesian) aquifer lies between impermeable layers, the water is under pressure, and its piezometric surface can stand above the ground, so a well flows by itself. Porosity n is the void fraction; the part that drains by gravity is the specific yield S_y, the part held back is the specific retention S_r, and n = S_y + S_r. Clay has high porosity but tiny specific yield, which is why it is a poor aquifer.

Darcy’s law governs flow through the aquifer: v = K·i, so Q = K·i·A, where K is the hydraulic conductivity, i = dh/dl the hydraulic gradient, and A the whole cross-section of flow. v is the Darcy velocity, discharge per unit gross area; the water in the pores really moves at the faster seepage velocity v/n. The product T = K·b (b the aquifer thickness) is the transmissivity. Example: a confined sand aquifer 10 m thick and 400 m wide, K = 30 m/day, gradient 0.002, carries Q = 30 × 0.002 × (10 × 400) = 240 m³/day.

Specific capacity is yield per unit drawdown, Q/s, and measures how productive a well is. Yield cannot be taken for ever: the safe yield is the extraction that recharge can replace. Natural recharge comes from rain infiltrating through pervious ground and seepage from rivers, tanks and irrigated fields. Over-extraction lowers the water table, dries wells, raises pumping cost, causes land subsidence, and draws saltwater into coastal aquifers. Artificial recharge uses percolation tanks, check dams, recharge pits, trenches and shafts, spreading basins, and injection wells for confined aquifers. For a planner the implication is spatial: impervious cover over recharge zones cuts recharge, so such land is protected, and roof and road runoff is led into recharge structures.

A service (distribution) reservoir stores treated water close to the demand. It does three jobs: it balances the constant rate at which water is pumped or treated against the varying rate at which it is used; it holds pressure in the network, especially when elevated; and it keeps a fire and emergency reserve. Types are ground-level or underground tanks (cheap, but the network then needs pumping), and elevated reservoirs (overhead tanks, giving gravity head). The best site is high ground at the centre of the zone served, which shortens mains and evens out pressure.

Worked example: daily demand 4,800 m³. In four six-hour blocks the draw is 10%, 40%, 30% and 20% of the day’s demand, while pumping gives 25% in each block. Cumulative (inflow − demand) is +15%, 0, −5%, 0. The largest surplus is 15% and the largest deficit is 5%, so balancing storage = 15% + 5% = 20% of 4,800 = 960 m³. Add a 400 m³ fire reserve and an emergency reserve of 10% of daily demand (480 m³): capacity = 960 + 400 + 480 = 1,840 m³.

Pumps lift water or add head. The centrifugal pump (impeller, smooth flow, wide range, compact) is the usual choice in water supply; a submersible pump, with its motor below water, serves tube wells; reciprocating pumps give high head at small, pulsating flow; jet pumps suit shallow wells; the hydraulic ram runs on falling water with no power. The power needed is P (kW) = ρ g Q H / (1000 η) = 9.81 Q H / η, with Q in m³/s, H the total head in m (static lift plus friction and minor losses) and η the overall efficiency. For Q = 0.04 m³/s, H = 25 m and η = 0.5, P = 9.81 × 0.04 × 25 / 0.5 = 19.62 kW. A pump works where its head–discharge curve cuts the system curve H = H_static + K Q²; two pumps in parallel add discharge at the same head, in series they add head at the same discharge.

The distribution system takes treated water to the consumer at adequate pressure. By source of energy it is a gravity system (source above the town), a pumping system, or a dual system, pumped to a service reservoir and fed by gravity from it, which is the commonest and gives the steadiest pressure. A continuous (24 × 7) supply keeps the pipes pressurised so that polluted water cannot be sucked in through a leak, which is why intermittent supply is a public-health risk as well as a convenience problem. Large towns are divided into zones or metered areas to control pressure and measure loss.

Network design rests on a head-loss law. The Hazen–Williams equation is empirical, for water at ordinary temperature in turbulent flow: in SI, V = 0.849 C R^0.63 S^0.54 (V in m/s, R = D/4 in m for a full pipe, S = h_f/L), and h_f = 10.67 L Q^1.852 / (C^1.852 D^4.87) (L in m, Q in m³/s, D in m). C is the roughness coefficient, larger for smoother pipe, smaller for old tuberculated pipe. The Darcy–Weisbach equation, h_f = f L V² / (2 g D), is general for any fluid and regime, with f from the Moody chart or f = 64/Re in laminar flow. Because h_f falls as D^−4.87, doubling a diameter at the same flow cuts head loss about 2^4.87 ≈ 29 times: small diameters are expensive in head. Example: D = 0.3 m, C = 100, S = 0.002 gives R = 0.075 m and V = 0.849 × 100 × 0.075^0.63 × 0.002^0.54 = 0.58 m/s.

In a looped network the flows are not known directly, so the Hardy Cross method finds them by successive correction. The two laws are continuity at every junction (inflow equals outflow) and energy around every closed loop (the algebraic sum of head losses is zero). The steps: (1) assume flows in every pipe that satisfy continuity at each node, clockwise positive; (2) for each loop compute h = K Q |Q|^(n−1) with its sign, n = 1.852 for Hazen–Williams and 2 for Darcy–Weisbach; (3) compute ΔQ = − Σh / (n Σ|h/Q|); (4) add ΔQ to every pipe of the loop, and to a pipe shared by two loops add both corrections, the neighbour’s with opposite sign; (5) repeat until Σh is within tolerance.

Treatment makes raw water safe, clear and palatable. The train depends on the source: groundwater may need only aeration and disinfection, or a specific removal of iron, fluoride or arsenic, while turbid surface water needs the full sequence below.

A grain settles by Stokes’ law: v_s = g (ρ_s − ρ_w) d² / (18 μ), valid for small spheres in laminar flow (Reynolds number below 1). Settling speed rises with the square of the diameter, so tiny colloids, which would take days to years to settle, must first be clumped by coagulation.

The ideal settling tank (Hazen, Camp) has four zones, inlet, settling, sludge and outlet, with horizontal flow and no mixing. Its key quantity is the overflow rate (surface loading) v₀ = Q / A_s, with A_s the plan area. A particle whose settling speed v_s ≥ v₀ is removed completely; one with v_s < v₀ is removed in the ratio v_s / v₀ (if it is spread evenly over the inlet depth). The depth does not appear: removal depends on plan area, not depth. Depth only fixes the detention time t = V/Q = H / v₀. Example: Q = 12,000 m³/d and v₀ = 30 m³/m²/d give A_s = 12,000/30 = 400 m². With t = 2.4 h = 0.1 day, V = 12,000 × 0.1 = 1,200 m³ and H = V/A_s = 3.0 m. A rectangular tank is typically long, with length to width of about 3:1 to 5:1.

Colloids carry like charges and repel each other, so they stay suspended. Coagulation destabilises them. Alum, Al₂(SO₄)₃·18H₂O, is the commonest coagulant: it reacts with the alkalinity of the water (bicarbonate) to form the gelatinous Al(OH)₃ floc, so a water short of natural alkalinity needs lime or soda ash as well. Ferrous sulphate with lime, ferric chloride and polyaluminium chloride are alternatives, and polyelectrolytes serve as coagulant aids. The dose and pH are found by the jar test: a row of beakers is dosed differently, mixed fast, then slowly, allowed to settle, and the one with the least turbidity gives the optimum. Rapid mixing lasts seconds to disperse the chemical; flocculation is slow paddling for about half an hour, gentle enough that the growing floc is not sheared apart.

Disinfection kills or inactivates pathogens; it is not sterilisation. Chlorine in water forms hypochlorous acid, Cl₂ + H₂O → HOCl + HCl, which ionises, HOCl ⇌ H⁺ + OCl⁻. HOCl is the far stronger disinfectant, and low pH favours it, so chlorination works better below about pH 7.5 than at pH 9. The chlorine demand is the dose minus the residual. The residual is either free (HOCl and OCl⁻) or combined (chloramines, formed with ammonia, longer-lived but weaker). In break-point chlorination the residual first rises, then falls as ammonia is oxidised, and beyond the break-point every extra dose appears as free chlorine. Bleaching powder, CaOCl₂, carries a quarter to a third of its weight as available chlorine; the chlorine mass needed is dose (mg/L) × flow (m³/d) in g/d, divided by the available fraction to get the powder. Example: 10 ML/d at 2 mg/L needs 10,000 m³/d × 2 g/m³ = 20 kg/d of chlorine, or 20/0.25 = 80 kg/d of powder at 25% available chlorine.

Quality is judged by physical, chemical and biological parameters against the drinking-water specification, IS 10500, whose concept is a two-level standard: an acceptable limit, which consumers accept without objection, and a permissible limit in the absence of an alternate source, beyond which water is rejected. The headline figures (as in the 2012 edition) are below.

Hardness is the soap-consuming power of Ca²⁺ and Mg²⁺, expressed as mg/L of CaCO₃: each ion is scaled by 50 over its equivalent weight (Ca 20, Mg 12), so hardness = 2.5 Ca + 4.1 Mg approximately. For Ca = 60 mg/L and Mg = 12 mg/L: 60 × 50/20 + 12 × 50/12 = 150 + 50 = 200 mg/L as CaCO₃. Carbonate (temporary) hardness is the part matched by alkalinity and removed by boiling or lime; non-carbonate (permanent) hardness from sulphates and chlorides is removed by lime–soda or ion exchange. With total hardness 200 and alkalinity 120 mg/L as CaCO₃, carbonate hardness is 120 and non-carbonate is 80.

BOD (biochemical oxygen demand) is the oxygen micro-organisms use to oxidise the biodegradable organic matter in water in 5 days at 20 °C, written BOD₅; it follows a first-order law, BOD_t = L₀ (1 − e^(−kt)), with L₀ the ultimate demand. COD (chemical oxygen demand) is the oxygen equivalent of the matter oxidised by a strong chemical oxidant (dichromate), so COD is at least BOD, and a high COD/BOD ratio signals organic matter that bacteria cannot break down. Both measure organic pollution, mostly in rivers and sewage rather than in treated drinking water, where dissolved oxygen is the other key indicator. Water is tested for indicator organisms, coliforms and E. coli, because pathogens are too varied and scarce to test for one by one: their presence signals faecal contamination.

The conventional approach treats water as three separate pipe systems: bring supply in, take sewage out, and carry storm water away as fast as possible. Water-sensitive urban planning (water-sensitive urban design) treats supply, storm water, wastewater and groundwater as one urban water cycle and plans it together with land use, street layout and open space. The aim is a city that conserves water, recharges its aquifer, cleans its runoff and floods less, and that keeps its post-development runoff close to what the land gave before.

• Conserve potable water: metering, leakage control, efficient fixtures, demand management.
• Fit-for-purpose supply: use greywater, treated sewage or harvested rain for flushing, gardens and cooling, and keep potable water for drinking and cooking.
• Manage runoff at source: infiltrate, detain and retain on the site so post-development peak and volume stay near pre-development values.
• Protect the natural water system: lakes, wetlands, streams, floodplains and recharge zones form a blue–green network that master plans keep free of development.
• Clean runoff where it falls: swales, rain gardens and wetlands remove sediment and pollutants before the water reaches a drain.