Technical notes
and field tutorials.

A mature field needs a mature organization — not a Geology–Reservoir–Production hierarchy, but a light matrix of three value streams: base management (surveillance, workovers, bypassed pay, water handling), development & well delivery (infill, recompletions, fracturing), and modeling, planning & asset development (models, reserves, FDP). Plus one integrated forecast, a governance rhythm, and the accountability boundaries that keep opportunities from falling between the boxes.

An exploration well is drilled once, at great cost. How reservoir and petroleum engineers shape it across the lifecycle — well objectives and a data-acquisition plan, pre-drill P90/P50/P10 volumetrics and the pressure prognosis behind well design, the data event of drilling and well testing (kh, skin, deliverability), then reserves under PRMS, concept selection, production forecasts and the Field Development Plan.

Reading a basin's story in plain language — the petroleum system, basin types from plate tectonics, stratigraphy and unconformities, structural traps, and how to read a cross-section. No degree required.

An ultrasound of the earth. The echo principle, acoustic impedance, two-way time versus depth, stacking and migration in plain words, and how to read horizons, faults, and bright spots on a section.

In-situ stress, pore pressure and rock strength through four dimensions: the 1D mechanical earth model and its mud weight window, the 2D section and stress polygon, the 3D volume with fault slip tendency, and 4D coupling where depletion moves the stress path — driving compaction, subsidence, permeability loss and induced seismicity as one mechanism seen six ways.

The methods are not interchangeable: each assumes a mechanism. Disequilibrium compaction versus fluid expansion, normal compaction trends and Eaton's exponent, the Bowers loading and unloading curves that explain velocity reversals — with a worked case where ignoring unloading underestimates pressure by 1750 psi — plus the centroid in dipping reservoirs, a hierarchy of calibration evidence, and why the estimate narrows but never becomes a number.

Two different failures with one symptom: pay that hides because the log lacks contrast, and pay that hides because the rock cannot flow. Saturation models, pore-throat classification, capillarity, and the test that separates them.

Building a Python, agentic AI, Excel, and dashboard framework for modern reservoir characterization. Six methods, one platform — Winland R35, FZI/HFU, PGS, Modified Lorenz, Leverett J-function, and ML on logs. Where the calculations end and the interpretation begins.

How to estimate OGIP and identify drive mechanism from pressure decline. P/Z plot, Cole plot, Roach plot, aquifer modeling across four classical models, and three worked numerical examples spanning volumetric, water drive, and geopressured reservoirs.

Original oil in place and drive mechanism from pressure decline — Havlena-Odeh, gas cap and water-drive diagnostics, drive indices, and three worked examples whose true answer is known by construction.

The whole path: conservation equations and how they are discretised, grid design and numerical dispersion, the Peaceman well model and why block pressure is not flowing pressure, where uncertainty actually concentrates, history matching as an ill-posed inverse problem with published proof that a good match can forecast badly, and how forecasting differs between a green field and a mature one.

Thermal (steam flooding, SAGD, in-situ combustion), gas and miscible (CO₂, N₂, hydrocarbon), and chemical (polymer, surfactant, ASP) methods on one map — recovery mechanisms, screening criteria, and when each method actually pays.

What CO₂ does to oil and how minimum miscibility pressure governs it, then continuous flooding versus cyclic single-well injection: sweep and conformance against soak time and declining cycles — with a decision table, screening criteria, what actually kills projects, and the associated-storage angle.

Free injectant, punishing miscibility pressure: the MMP ladder against the fracture limit, gravity-stable crestal injection, pressure maintenance and chase-gas duty, cyclic N₂ where repressurization does the work instead of dissolution, and the back-end costs — compression horsepower and nitrogen rejection.

The complete chemical catalogue organised by the two failures it cures — poor sweep and trapped residual oil — with capillary desaturation, optimal salinity and the Huh relation, polymer retention and temperature limits, ASP complexity, and cyclic surfactant treatment driven by spontaneous imbibition in tight rock.

A practitioner's guide to nodal analysis. Where IPR meets VLP, why the operating point is never decided at any single component, and how to translate seventy years of correlations into something that runs on your laptop. Equations kept to what's useful, references kept honest, opinions kept brief.

Producing stacked pay zones singly or together, compared from both sides: differential depletion, crossflow and thief zones, blurred pressure and material balance on the reservoir side; composite IPR by horizontal summation, well productivity, artificial lift, flow assurance and allocation on the production side — with a decision table and the intelligent-completion middle ground.

Gas wells don't fail because the reservoir runs dry — they fail because liquids start winning the upward race. A practical guide to the equation, the physics behind it, and how to read the safety ratio in the field.

When to use exponential, hyperbolic, or harmonic. How to read the b-factor as reservoir physics. Window selection, terminal decline switch, and common pitfalls when forecasting EUR.

When Arps hyperbolic fits over-extrapolate by 40% or more, the reciprocal rate plot offers a theoretically rigorous second opinion. Plot 1/q vs N_p/q, fit a line, take the inverse slope. Worked example, log-log cross-check, and the factor-of-2 caveat for gas wells.

Mature waterfloods need a different forecasting tool than rate-time decline. Ershaghi's log-WC vs cumulative oil linearization extracts ultimate recovery directly from data the operator already has. Worked example, window selection, and engineering override.

Coning, channeling, or near-wellbore problem? Chan's 1995 diagnostic plot turns water-oil ratio time series into a mechanism identification tool. Read it correctly and the workover plan writes itself.

Most disappointing treatments were correctly executed on the wrong problem. Water-production problems ranked by how treatable they are, why coning and normal sweep are not shut-off candidates, mechanical versus chemical options, disproportionate permeability reduction and where RPMs actually work, gel placement in linear versus radial flow, and the failure modes decided before the truck arrives.

Choosing the right well and the right intervention: a physics-first, multidisciplinary workflow from production gap to risked-value ranking — integrating G&G, reservoir, and production engineering. Includes nodal analysis, a diagnosis-to-intervention map, and deep dives on artificial-lift selection, sand control, and water shut-off screening — with a full end-to-end workflow chart from team and data through methodology to well-program type.

Four questions kept separate: is this well a candidate, what geometry and conductivity should the fracture have, did the pumping do what was designed, and what did the reservoir actually get. With the PKN and KGD width models, a conductivity curve computed from the Cinco-Ley pseudo-skin relation showing why the “optimum” depends on what is held fixed, the Nolte-Smith pressure diagnostic, and the gap between created, propped and effective length.

The complete, systematic framework under SPE-PRMS 2018 — the two-axis classification matrix, Reserves vs Contingent vs Prospective Resources, 1P/2P/3P and P90/P50/P10, project maturity sub-classes, the commerciality test, probabilistic aggregation, and how PRMS relates to SEC and UNFC.

Why P50 isn't enough — and how Monte Carlo Simulation transforms volumetric reserves estimation. Probabilistic OGIP/OOIP, choosing the right distribution per parameter, multi-zone aggregation, tornado analysis, and the upcoming MCVR tool.

A complete, systematic tour across the upstream value chain — seismic fault-detection CNNs, facies classification, surrogate reservoir models, drilling analytics, production forecasting and ESP failure prediction — plus the physics-informed turn, the real MLOps workflow, and the subsurface-specific pitfalls (leakage, tiny data, extrapolation) that sink naive models.

From observed field behavior back to mechanism: inverse problems and history matching, experimental design (factorial, LHS, response surfaces), ML surrogates and explainability, and the tool-assisted agentic AI loop — plus a step-by-step Python stack and the anatomy of an exception-based surveillance dashboard, all under physics-first governance.

RCAL and SCAL end to end: the four filters that make a plug dataset optimistic, Klinkenberg and net confining stress, the cleaning step that destroys wettability before SCAL begins, capillary pressure and its conversion to reservoir conditions, the capillary end effect that manufactures residual oil, and why a saturation exponent from cleaned core overstates hydrocarbon in place.

OOIP divides by B_oi; material balance stands on B_o, R_s and B_g. Fluid classification, sampling and representativity, the four laboratory studies and what each one actually measures, the differential-to-field-basis conversion nobody checks, Y-function and K-value consistency tests, and where correlations stop being defensible.

A decline forecast that cannot be re-run is not a forecast. What has to be true of the code that replaces the workbook: judgements as declared parameters, a bounded Arps fit that raises rather than guesses, and a validation harness that reports blind-test error instead of a single R².

Four ways a machine-learning result on production data can be wrong while every validation check reports success — a baseline nobody asked, a model that structurally cannot extrapolate, a score that measured the choosing rather than the model, and a feature that contained its own answer. Each one measured on a five-well gas history, with the five questions that catch them.

When the product is heat, not fluid. How petroleum reservoir engineering transfers to geothermal — energy-in-place, material balance, reinjection and thermal breakthrough, and why a renewable resource can still decline.