Reservoir engineering: material balance and decline curves

In short

Material balance treats a reservoir as a single tank and applies conservation of mass (expressed as volumes at reservoir conditions) to find original hydrocarbons in place, the active drive mechanism and the aquifer influx, using only production, pressure and PVT data. Decline curve analysis works on the production rate history alone and extrapolates it to estimate future rate, remaining reserves and economic life. Together they are the two classical "dynamic" reserve-estimation methods that complement volumetric estimates and numerical simulation. The topic carries heavy weight in the GATE Petroleum Engineering paper (PE) and in university reservoir engineering papers, and is directly used by reservoir engineers at ONGC, Oil India, Cairn (Vedanta), Reliance, Shell, Halliburton, SLB and Baker Hughes, and in reserve statements audited to Indian/SPE-PRMS practice for DGH.

Explained

Basis and the tank model

• Assumptions: one pressure for the whole tank (zero-dimensional), instantaneous equilibrium, uniform PVT, no change in pore volume except compaction, production and injection known accurately, rock and fluid properties representative.
• Mass balance in volume terms: underground withdrawal (fluid removed, measured at reservoir conditions) = expansion of oil + expansion of gas cap + expansion of connate water and reduction of pore volume + net water influx + injection.
• Data needed: cumulative oil Np, gas Gp, water Wp; static reservoir pressures from build-ups; PVT (Bo, Bg, Rs, Bw, cw); initial saturations and rock compressibility.
• Why it matters: it is an *independent* check on volumetric OOIP (A·h·φ·(1−Sw)/Boi) and it reveals the drive mechanism, which fixes the expected recovery factor and the need for pressure maintenance.

Forms of the material balance equation (MBE)

• General (Schilthuis/Tarner) form for oil: N[(Bo−Boi) + (Rsi−Rs)Bg + m Boi(Bg/Bgi − 1) + (1+m)Boi(cwSw + cf)Δp/(1−Sw)] + We + Winj Bw + Ginj Bginj = Np[Bo + (Rp−Rs)Bg] + Wp Bw.
• Terms identified: (Bo−Boi)+(Rsi−Rs)Bg = oil + liberated-solution-gas expansion; m Boi(Bg/Bgi−1) = gas-cap expansion, m = gas-cap volume / oil-zone volume; the cw, cf group = connate-water expansion plus pore compaction; We = cumulative water influx.
• Undersaturated, above bubble point, no water influx: Np Bo = N Bo i co e Δp effectively → N = Np Bo / [Boi(co,e)Δp] with effective compressibility co,e = (coSo + cwSw + cf)/(1−Sw). Recovery is small (typically 1-10%) because liquid compressibility is small.
• Saturated, solution-gas drive, no water influx, no gas cap: N = Np[Bo + (Rp−Rs)Bg] / [(Bo−Boi) + (Rsi−Rs)Bg].
• Havlena-Odeh straight-line form: write F = N(Eo + mEg + Ef,w) + We, where
  • F = Np[Bo+(Rp−Rs)Bg] + WpBw (underground withdrawal),
  • Eo = (Bo−Boi) + (Rsi−Rs)Bg, Eg = Boi(Bg/Bgi − 1), Ef,w = (1+m)Boi(cwSw+cf)Δp/(1−Sw).
  • Plot F vs Eo (no gas cap, no influx) → straight line through origin of slope N. If it curves upward, water influx or a gas cap is being ignored; plot F/Eo vs We/Eo (slope 1, intercept N) or F vs (Eo+mEg) to solve for m.
• Drive indices (sum = 1): DDI (depletion) = N Eo/F, SDI (segregation/gas cap) = N m Eg/F, WDI (water) = We/F (with F defined to include WpBw; if F is taken as Np[Bo+(Rp−Rs)Bg] only, WDI = (We−WpBw)/F), CDI (compaction/connate water) = N Ef,w/F. The largest index names the dominant drive.

Gas reservoirs: p/z straight line

• Volumetric (no influx) gas MBE: p/z = (pi/zi)(1 − Gp/G). A plot of p/z against Gp is a straight line; the intercept at Gp = 0 gives pi/zi and the x-intercept (p/z = 0) gives G, the original gas in place.
• Interpretation of departures: a line bending *concave upward* (p/z falling more slowly, flattening) indicates water influx or gas influx from a connected reservoir — extrapolating the early line then *overestimates* G; an abnormally-pressured (geopressured) reservoir gives two slopes as compaction support ends.
• Water-drive gas MBE: G Bgi = Gp Bg + (We − Wp Bw) → water drive keeps pressure up but traps gas behind the advancing water, so recovery (typically 45-70%) is often *less* than volumetric depletion drive (80-90%+).
• Gas condensate needs a two-phase z-factor (z₂ϕ) once pressure falls below the dew point, because liquid drops out in the reservoir.

Drive mechanisms and expected recovery

• Solution-gas (depletion) drive: pressure falls fast, GOR rises then falls, no water production; RF typically 5-30%.
• Gas-cap (segregation) drive: slow pressure decline, rising GOR in upper completions; RF 20-40%; favoured by high dip, good vertical permeability and low production rates.
• Water drive: pressure nearly maintained, increasing water cut; RF 35-60% (edge vs bottom water matters); strong aquifers give a near-flat pressure history.
• Rock and connate-water expansion / compaction drive: significant in overpressured and in unconsolidated or chalky reservoirs; RF small on its own.
• Gravity drainage: effective in steeply dipping, high-permeability reservoirs; can give the highest primary recovery (up to 60%+) at low offtake rates.
• Combination drive: most real fields; drive indices change with time.

Aquifer (water influx) models

• Pot (small, steady-state) aquifer: We = cw+f · Wi · Δp — instantaneous response, for a limited aquifer.
• Schilthuis steady state: dWe/dt = k(pi − p), so We = k ∫(pi−p)dt.
• van Everdingen-Hurst unsteady state: We = B Σ Δp · WeD(tD), with B = 1.119 φ ct re² h f (field units, bbl/psi) and dimensionless time tD = 6.33×10⁻³ k t /(φ μ ct rw²) type expressions; needs superposition over pressure steps.
• Fetkovich pseudo-steady state: an aquifer "material balance" with productivity index J, We computed stepwise — much simpler than van Everdingen-Hurst and widely used.
• Carter-Tracy: avoids superposition by an incremental formulation.

Decline curve analysis (Arps)

• Decline analysis needs boundary-dominated flow at (roughly) constant drawdown/BHP and unchanged operating conditions — adding compression, a frac job, or shutting wells in breaks the trend.
• Loss ratio and decline exponent: nominal decline D = −(1/q)(dq/dt); b = d/dt(1/D) = −d/dt[q/(dq/dt)] defines the Arps family.
• Exponential (b = 0): constant fractional decline; q vs t is a straight line on semi-log; most conservative; typical of solution-gas drive and many tight wells late in life.
• Hyperbolic (0 < b < 1): decline rate decreases with time; typical b values 0.2-0.5 for conventional oil, 0.5-1.4+ for shale/tight gas (b > 1 gives an infinite reserve if not truncated — hence a terminal decline switch to exponential at 5-10%/yr).
• Harmonic (b = 1): the limiting case; straight line on a log q vs Np plot; often seen in high-viscosity/water-drive wells.
• Nominal vs effective decline: effective (secant) annual decline de = 1 − e^−D for exponential; exam problems often give "10% per year" as effective.
• Diagnostic plots: log q vs t (exponential = straight), q vs Np (exponential = straight), log q vs log t, D vs t, and Np vs t; modern practice adds rate-transient (Fetkovich/Blasingame/Agarwal-Gardner) type curves which handle transient flow that Arps cannot.
• Reserves/EUR: integrate the rate to the economic rate qa (the rate at which revenue equals operating cost); remaining reserves = Np(qa) − Np(now). Economic limit depends on oil price, operating cost, royalty and cess — in India, royalty plus OID cess (nomination acreage) and PSC/HELP-DSF terms change qa materially.

Limitations, comparisons and Indian practice

• MBE limitations: needs good static pressures (hard in wells that never shut in), averages away pressure gradients, ignores geometry and sweep, sensitive to PVT errors and to small denominators early in depletion (use data only after meaningful Δp), and cannot distinguish a small reservoir with influx from a larger depleting one without independent data.
• DCA limitations: purely empirical, no physics, no pressure or PVT input; invalid in transient flow, after any change in operating conditions, and for multi-well/commingled data; b is non-unique and often over-fitted.
• Comparison of methods: volumetric (static, usable before production), material balance (dynamic, needs pressure and production), DCA (needs only rate history), simulation (needs all of it, gives distribution in space and time).
• Indian regulatory practice: reserves are reported under SPE/WPC/AAPG/SPEE PRMS categories (1P/2P/3P; proved/probable/possible; reserves vs contingent resources) to the DGH under the Oilfields (Regulation and Development) Act, 1948 and the Petroleum and Natural Gas Rules, 1959 (amended; the ORD Act was amended in 2025 to broaden "mineral oils" and introduce petroleum leases — check the current notified text for details). Field development plans and reserve revisions submitted to the DGH/Management Committee rest on material balance and DCA alongside simulation.
• Gas sales and pricing (APM/ceiling price notifications by PPAC/MoPNG) and the Indian practice of reporting in metric units (MMm³, MMt, kg/cm², MMSCMD) sit alongside oilfield units (bbl, psi, scf) — exam answers should state the unit system used.

Key points

• Material balance - volumetric conservation on a tank model; gives N or G, drive indices and We from production, pressure and PVT.
• Underground withdrawal F - Np[Bo+(Rp−Rs)Bg] + WpBw, all at reservoir conditions.
• Eo, Eg, Ef,w - oil+solution-gas, gas-cap, and formation/connate-water expansion terms per unit N.
• m - ratio of initial gas-cap reservoir volume to initial oil-zone reservoir volume.
• Havlena-Odeh - recast the MBE as a straight line; the plot that *is* straight identifies the correct drive model.
• Drive indices - DDI + SDI + WDI + CDI = 1; the dominant index names the mechanism and the likely recovery factor.
• p/z plot - p/z = (pi/zi)(1 − Gp/G); straight for volumetric gas, flattening for water drive.
• Water-drive gas recovery - usually *lower* than depletion drive because gas is trapped at high pressure behind the water front.
• Effective oil compressibility - co,e = (coSo + cwSw + cf)/(1−Sw); governs undersaturated recovery, which is only a few percent.
• Aquifer models - pot, Schilthuis (steady state), van Everdingen-Hurst (unsteady), Fetkovich and Carter-Tracy (pseudo-steady).
• Arps decline - q = qi(1 + b D i t)^−1/b; b = 0 exponential, 0 < b < 1 hyperbolic, b = 1 harmonic.
• Exponential decline - straight on semi-log q-t and on linear q-Np; D constant; most conservative EUR.
• Hyperbolic b > 1 - must be truncated with a terminal exponential decline or reserves become unbounded.
• Nominal vs effective decline - de = 1 − e^−D; never substitute one for the other in the formula.
• Boundary-dominated flow - the precondition for Arps; transient data needs rate-transient (Blasingame/Agarwal-Gardner) analysis.
• Economic limit rate qa - where revenue = operating cost; it sets reserves, not EUR of the reservoir.
• Reserve classes - PRMS 1P/2P/3P reported to DGH; MBE and DCA are the dynamic basis of those numbers.
• Error sources - bad static pressures, unrepresentative PVT, unmetered water or gas, and early-life data with tiny Δp.

Formulas and examples

Example 1 - Undersaturated oil, OOIP from MBE. A reservoir above bubble point: pi = 4,000 psia, p = 3,500 psia, Np = 1.2 MMSTB, Bo = 1.310, Boi = 1.300 rb/STB, Sw = 0.25, co = 12×10⁻⁶, cw = 3×10⁻⁶, cf = 4×10⁻⁶ /psi, no water influx or production.

• So = 1 − 0.25 = 0.75. co,e = (12×0.75 + 3×0.25 + 4)/0.75 ×10⁻⁶ = (9 + 0.75 + 4)/0.75 = 13.75/0.75 = 18.33×10⁻⁶ /psi.
• Δp = 500 psi. Denominator = Boi·co,e·Δp = 1.300 × 18.33×10⁻⁶ × 500 = 0.011917 rb/STB.
• F = Np·Bo = 1.2×10⁶ × 1.310 = 1.572×10⁶ rb.
• N = 1.572×10⁶ / 0.011917 = 1.319×10⁸ STB ≈ 132 MMSTB.
• Recovery so far = 1.2/131.9 = 0.91% — typical of undersaturated depletion.

Example 2 - Solution-gas drive, N and drive index. Np = 8 MMSTB, Gp = 9.6 Bscf, Bo = 1.25, Boi = 1.42 rb/STB, Rs = 500, Rsi = 900 scf/STB, Bg = 0.00120 rb/scf, no gas cap, no water influx (ignore cf, cw).

• Rp = 9.6×10⁹/8×10⁶ = 1,200 scf/STB.
• F = 8×10⁶[1.25 + (1,200 − 500)(0.00120)] = 8×10⁶[1.25 + 0.84] = 8×10⁶ × 2.09 = 1.672×10⁷ rb.
• Eo = (1.25 − 1.42) + (900 − 500)(0.00120) = −0.17 + 0.48 = 0.31 rb/STB.
• N = F/Eo = 1.672×10⁷/0.31 = 5.394×10⁷ STB ≈ 53.9 MMSTB; RF to date = 8/53.9 = 14.8%.
• DDI = N·Eo/F = 1.0 → pure depletion drive, as assumed. Note the oil shrinks (Bo fell); all the expansion came from liberated gas.

Example 3 - p/z plot for a gas reservoir. pi = 3,200 psia, zi = 0.88. At Gp = 12 Bscf the measured p = 2,400 psia, z = 0.90.

• (p/z)i = 3,200/0.88 = 3,636.4 psia; p/z now = 2,400/0.90 = 2,666.7 psia.
• From p/z = (pi/zi)(1 − Gp/G): 2,666.7 = 3,636.4(1 − 12/G) → 1 − 12/G = 0.73333 → 12/G = 0.26667 → G = 45.0 Bscf.
• Abandonment at p/z = 400 psia: Gp = G[1 − 400/3,636.4] = 45.0 × 0.8900 = 40.05 Bscf → RF ≈ 89% (volumetric depletion).
• If instead water influx were supporting pressure, the early line would be too flat and G would be overstated — check for water production before trusting the extrapolation.

Example 4 - Exponential decline: rate, time and reserves. A well declines exponentially from qi = 500 STB/d; one year later q = 410 STB/d. Economic limit qa = 40 STB/d.

• D = ln(500/410)/1 yr = ln(1.21951) = 0.19845 /yr (effective de = 1 − e^−0.19845 = 18.0%/yr).
• Rate after 5 years: q = 500 e^{−0.19845×5} = 500 e^−0.99225 = 500 × 0.37072 = 185.4 STB/d.
• Time to economic limit: t = ln(500/40)/0.19845 = ln(12.5)/0.19845 = 2.52573/0.19845 = 12.73 yr.
• EUR from now = (qi − qa)/D = (500 − 40)/0.19845 = 2,318.0 STB·d/yr... convert: ×365 = 846,100 STB ≈ 0.85 MMSTB. (Using D in 1/day: D = 0.19845/365 = 5.437×10⁻⁴ /d → (500−40)/5.437×10⁻⁴ = 846,100 STB — same answer.)

Example 5 - Hyperbolic vs exponential EUR for the same well. qi = 500 STB/d, Di = 0.19845/yr, qa = 40 STB/d, b = 0.5.

• Hyperbolic Np = qi/[Di(1−b)] ·[1 − (qa/qi)^1−b] = 500/[0.19845×0.5] × [1 − (0.08)^0.5] = 500/0.099225 × (1 − 0.28284) = 5,039.1 × 0.71716 = 3,614.0 STB·d/yr → ×365 = 1,319,100 STB.
• Time: t = [(qi/qa)^b − 1]/(b·Di) = [(12.5)^0.5 − 1]/(0.5×0.19845) = (3.53553 − 1)/0.099225 = 2.53553/0.099225 = 25.55 yr.
• So the same early data with b = 0.5 gives 1.32 MMSTB over 25.6 years against 0.85 MMSTB over 12.7 years for b = 0 — a 56% higher EUR. This sensitivity to b is why b must be justified, truncated by a terminal decline, and reconciled with material balance.

Where it is used

• Reservoir engineering at E&P operators: ONGC, Oil India, Cairn Oil & Gas (Vedanta), Reliance-BP, ONGC Videsh — annual reserve estimation, drive-mechanism studies and field development plans submitted to the DGH.
• Reserve audit and consultancy: PRMS 1P/2P/3P certification by third parties for listed companies, bank financing and farm-in valuation; MBE and DCA are the standard independent checks on simulation output.
• Production engineering and surveillance: monthly decline fits to flag wells needing workover, artificial lift or stimulation; aquifer/water-drive diagnosis before an injection project.
• Gas field operations and contracting: p/z analysis to confirm OGIP and to plan compression timing and MMSCMD commitments in gas sales agreements (KG basin, Cambay, Assam, Rajasthan fields).
• CBM, tight gas and DSF/HELP blocks: hyperbolic decline with terminal decline is the main EUR tool for small fields with little pressure data.
• Job roles: reservoir engineer, production/surveillance engineer, reserves analyst, petroleum data analyst (Python/Spotfire decline fitting), and service-company roles at SLB, Halliburton, Baker Hughes and Weatherford.

Interview questions

• Q: What does the material balance equation actually conserve? A: Mass, expressed as reservoir volumes — the volume voided by production equals the expansion of the remaining fluids and rock plus any influx or injection.
• Q: Why is the Havlena-Odeh form preferred over solving the MBE for N at each point? A: It turns the equation into a straight line, so the plot's shape tests the assumed drive model instead of merely returning a number that scatters when the model is wrong.
• Q: What does m represent and how do you find it? A: The ratio of initial gas-cap reservoir volume to initial oil-zone volume; from mapping/logs, or by regression on a Havlena-Odeh plot of F against (Eo + mEg).
• Q: A p/z plot curves upward (flattens) with cumulative production. What is happening? A: Pressure is being supported, most likely by water influx or inflow from a connected reservoir, so the straight-line extrapolation overestimates OGIP.
• Q: Why can water drive give *lower* gas recovery than depletion drive? A: Water traps residual gas behind the advancing front while reservoir pressure stays high, so much gas is left at abandonment.
• Q: Why does Bo sometimes *decrease* during solution-gas drive while recovery continues? A: Below bubble point, gas leaves solution and the oil shrinks; the expansion driving production comes from the liberated gas, not from the oil.
• Q: What flow regime must exist for Arps decline analysis to be valid? A: Boundary-dominated (depletion) flow at essentially constant operating conditions; transient data needs rate-transient analysis instead.
• Q: Which Arps model gives the most conservative reserves, and why? A: Exponential (b = 0), because the fractional decline stays constant instead of flattening with time.
• Q: Why is b > 1 a problem? A: It implies an unbounded cumulative, so practice truncates it with a terminal exponential decline, typically 5-10% per year.
• Q: What is the difference between nominal and effective decline? A: Nominal D is the instantaneous fractional decline used in the formulae; effective de is the drop over a period, de = 1 − e^−D for exponential decline.
• Q: How do material balance and decline curves cross-check each other? A: MBE gives in-place volumes and the drive, which bound a plausible recovery factor; DCA gives EUR directly — if DCA EUR exceeds N × a credible RF, the decline fit or b is wrong.
• Q: Early in a field's life the MBE gives wildly scattered N. Why? A: Δp and the expansion terms are tiny, so small errors in pressure and PVT dominate; wait for meaningful depletion before trusting the solution.

Common mistakes

• Mixing surface and reservoir volumes — forgetting to multiply Np by Bo, or using Rp = Rs.
• Using Rs instead of Rp = Gp/Np in the withdrawal term, which silently discards produced free gas.
• Dropping the connate-water and formation compressibility term in undersaturated or overpressured reservoirs, where it is the *only* energy source.
• Confusing nominal D with effective annual decline, or mismatching time units between D and t (per year vs per day).
• Extrapolating a decline fit taken from transient (early) flow, or across a frac job, compressor installation or long shut-in.
• Extrapolating a p/z line without checking water production, and reporting an inflated OGIP.
• Reporting EUR rather than reserves — ignoring the economic limit rate, royalty, cess and operating cost — or quoting reserves without a PRMS category.

Related topics

• Reservoir rock and fluid properties · Enhanced oil recovery · Well testing and pressure transient analysis · Multiphase flow in porous media and relative permeability · Reservoir simulation and history matching · Petroleum economics and reserve classification (PRMS).