An observation-well drawdown derivative helps you interpret a pumping test by showing how quickly drawdown changes with logarithmic time. A flat, sustained derivative can support radial-flow interpretation; a rising late-time derivative may indicate a barrier boundary, while a falling response can suggest a recharge influence or leakage. These are diagnostic hypotheses, not automatic geological identifications. To read a derivative responsibly, check the pumping history, measurement quality, distance, aquifer setting and independent wells.
The key distinction is between a water-level curve, which records how much hydraulic head changed, and its derivative, which exposes changes in the shape of that response. A derivative can sharpen an otherwise subtle feature, but it can also amplify noisy measurements, clock mistakes and abrupt pumping-rate changes. This guide explains the physics, a transparent calculation, competing explanations, meaningful checks and a way to report uncertainty without claiming that a graph has proved a buried boundary.
Imagine a constant-rate test in an aquifer and a separate, instrumented observation well. The measured drawdown is the pretest hydraulic head minus the head observed during pumping, expressed in metres, with the sign convention declared. Plot drawdown against elapsed time and then evaluate a derivative with respect to natural logarithmic time. The result remains in metres because the logarithm of a time ratio is dimensionless; it is not a pumping rate, velocity or direct measure of permeability.
Start with the physical experiment
Separate pumping well from observation well
The pumping well includes entrance losses, changing water columns and wellbore storage, whereas a nearby observation well is intended to register the aquifer’s pressure diffusion with less local hydraulic disturbance. If a pattern appears only inside the pumped well, screen losses or pump behaviour may dominate before an aquifer boundary becomes plausible.
An independent reader can challenge this interpretation as follows: Compare two independent water-level locations, their screen intervals and the actual pump-discharge log. The error to avoid is concrete: Treating a pumped-well curve and a remote observation curve as interchangeable hides the reason the second well exists. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Define the hydraulic reference
The field log should identify the measuring point, surveyed elevation, sensor position and pretest water level. A transducer measures pressure near the instrument, not automatically elevation above a common vertical datum. Drawdown needs a consistent reference at each time, so a shift in measuring point or logger depth can imitate a hydrologic step.
An independent reader can challenge this interpretation as follows: Recalculate one manual dip reading from the surveyed reference and compare it with the corresponding electronic reading. The error to avoid is concrete: Subtracting unlike elevations creates a precise-looking derivative from an inconsistent head series. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Control the clock
Elapsed time starts at the actual onset of the hydraulic stress, not when an operator intended to turn on the pump or when the logger file was opened. An initial timing offset distorts logarithmic spacing most severely at early time, precisely where wellbore effects are already strong.
An independent reader can challenge this interpretation as follows: Match pump-controller timestamps, discharge telemetry and transducer clocks before differentiating. The error to avoid is concrete: Plotting nominal times without synchronization can move apparent transitions across decades of time. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Measure actual discharge
Even a test described as constant-rate has transient startup, electrical interruptions and valve adjustments. The analytical constant-Q solution assumes a stable volume per unit time. Drawdown responds to the complete discharge history, so an increase in Q can look like a change in aquifer confinement.
An independent reader can challenge this interpretation as follows: Graph measured Q alongside drawdown and model known rate steps using superposition when necessary. The error to avoid is concrete: Calling every late-time slope change a boundary while ignoring changing discharge is a category error. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Specify well geometry
Screen length, depth, separation and aquifer thickness set the geometry in which flow occurs. Partial penetration can produce vertical gradients before radial assumptions become suitable. A derivative change at one depth may trace vertical flow redistribution rather than a distant lateral obstruction.
An independent reader can challenge this interpretation as follows: Compare shallow and deep observation intervals and note the length of each open section. The error to avoid is concrete: A map-view radial conceptual model is incomplete when the actual test samples different vertical layers. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Preserve a baseline
A pretest baseline should cover barometric movement, background pumping, weather and instrument drift. A flat-looking hour is not necessarily a stationary hydraulic reference. Subtracting a drifting baseline can either invent a late-time rise or erase a real gradual response.
An independent reader can challenge this interpretation as follows: Use an unaffected control well, logged regional withdrawals and a clearly documented baseline correction. The error to avoid is concrete: A perfectly smoothed baseline is not evidence if its correction was chosen after seeing the wanted result. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Check instrument resolution
A pressure transducer that resolves centimetres may be adequate for metre-scale drawdown yet marginal when late derivative changes are a few millimetres. Differentiation increases sensitivity to measurement scatter because it estimates change from neighboring observations.
An independent reader can challenge this interpretation as follows: Compare derivative magnitude with noise measured in the pretest time series and vary the smoothing window. The error to avoid is concrete: Publishing fine diagnostic detail below measurement precision confuses computational digits with information. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Test the datum convention
Water depth below a casing mark increases during drawdown, while hydraulic head elevation decreases. Different vendors may export either sign. A derivative can appear to rise, fall or cross zero purely because of sign and plot choices.
An independent reader can challenge this interpretation as follows: Write the sign convention in the figure caption and verify two manually calculated points. The error to avoid is concrete: Interpreting a flipped curve as artesian recharge reverses the physical story without changing the raw sensor data. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Differentiate without inventing information
Understand logarithmic time
The derivative commonly plotted in pumping diagnostics is ds/dln(t), where s is drawdown and t is elapsed time. Logarithmic scaling gives comparable attention to early and late orders of magnitude. A difference in drawdown over a tenfold time interval must be divided by ln(10), approximately 2.302585, to obtain the natural-log derivative.
An independent reader can challenge this interpretation as follows: Compute a simple three-point comparison with a calculator before trusting software output. The error to avoid is concrete: Calling a derivative per time decade the same numerical quantity as ds/dln(t) introduces a factor-of-2.303 error. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Use a reproducible numerical example
Suppose a synthetic observation well has 0.25 metre drawdown at ten minutes and 0.85 metre at one hundred minutes during a steady pumping test. The secant slope is 0.60 metre per decade, or 0.60 divided by ln(10), approximately 0.2606 metre per natural-log time.
An independent reader can challenge this interpretation as follows: Substitute the same two head differences into both versions of the expression and check that the units remain metres. The error to avoid is concrete: A single secant is illustrative rather than proof of a truly flat derivative over a wide interval. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Translate derivative into transmissivity
For the late-time Cooper–Jacob approximation under its radial-flow assumptions, ds/dln(t) equals Q divided by 4πT. Transmissivity T has area per time units. At Q = 240 cubic metres per day and derivative 0.2606 metre, T is 240/(4π×0.2606), about 73.3 square metres per day.
An independent reader can challenge this interpretation as follows: Recalculate with the per-decade slope using T = 2.303Q/(4πΔs) and confirm the same result. The error to avoid is concrete: Using Q in litres per second with T reported in square metres per day without conversion gives an enormous false discrepancy. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Distinguish derivative from increment
A drawdown difference of 0.10 metre between two readings does not define a derivative unless the corresponding elapsed times and logarithmic interval are known. Near a logarithmic sampling gap the denominator may differ greatly from a neighbouring pair despite identical drawdown increments.
An independent reader can challenge this interpretation as follows: For each displayed point, retain the two timestamps and denominator used by the numerical operator. The error to avoid is concrete: A plot of successive differences alone cannot justify flow-regime labels. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Avoid zero logarithms
Logarithms of elapsed time require positive time measured since the stress begins. Startup samples stamped at zero are not valid arguments for ln(t). Numerical software may silently replace zero with a tiny positive number and manufacture an extreme early derivative spike.
An independent reader can challenge this interpretation as follows: Document startup exclusion and keep the physical time-origin uncertainty visible. The error to avoid is concrete: An arbitrarily chosen epsilon is a computational patch, not a measured aquifer property. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Control smoothing choices
A moving-window regression across log time can reduce instrument noise, but a wider window also hides short transitions and shifts their apparent timing. Different smoothing widths can make a rising boundary signature look horizontal or create an overshoot near a discharge change.
An independent reader can challenge this interpretation as follows: Show raw drawdown, derivative with at least two sensible windows and the window definition. The error to avoid is concrete: A smooth curve is not automatically the most truthful curve when its shape depends on a chosen filter. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Recognize edge effects
Numerical derivatives at the first and last samples have less information on one side, making them sensitive to extrapolation or one-sided differences. An alleged final-time boundary that uses only the last two points is especially fragile.
An independent reader can challenge this interpretation as follows: Extend monitoring beyond the suspected transition and compare interior-window estimates. The error to avoid is concrete: The most dramatic point on the plot may be the point with the least defensible derivative. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Check time units
Taking ln(t in minutes) and ln(t in days) produces derivatives that are identical because their natural logarithms differ by a constant offset. Absolute elapsed-time labels matter to match geometry and diffusivity, but converting units does not change a correctly computed ds/dln(t).
An independent reader can challenge this interpretation as follows: Repeat a derivative using seconds and minutes and check equality to rounding. The error to avoid is concrete: If a numerical derivative changes by a factor of sixty solely from conversion, the code differentiated with respect to linear time instead. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Carry uncertainty
Measurement noise, pumping-rate variability and baseline error affect the derivative and any transmissivity estimate inferred from it. Rather than give T = 73.3000 square metres per day, acknowledge that a coarse 0.60-metre slope and rate estimate warrant rounded reporting.
An independent reader can challenge this interpretation as follows: Vary Q and derivative within documented uncertainty and present a reasonable range. The error to avoid is concrete: Four decimal places of precision cannot compensate for omitted field uncertainty. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Read diagnostic shapes as hypotheses
Establish an early regime
At the start, the pumped well may deliver water from its own casing storage, and near-well skin or turbulent losses may influence its measured levels. A rising derivative during startup may say more about the measurement system than about regional hydraulic structure.
An independent reader can challenge this interpretation as follows: Wait for an interpretable aquifer-controlled interval and examine an independent observation well. The error to avoid is concrete: Labeling the first visible bend ‘the fault boundary’ assigns distant geology to local apparatus behaviour. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Search for radial-flow support
A sufficiently sustained, approximately level derivative under constant-rate confined radial-flow conditions is compatible with the Theis family of responses. This helps estimate transmissivity, but one straight stretch is neither a certification of a homogeneous aquifer nor a measurement of storage.
An independent reader can challenge this interpretation as follows: Compare Q-normalized curves from multiple observation distances and check the underlying semi-log drawdown slope. The error to avoid is concrete: A horizontal derivative generated by aggressive smoothing is weaker evidence than one supported across wells. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Consider a no-flow barrier
An approximately impermeable lateral boundary can be represented in an idealized model by a pumping image well on the same side of the boundary. At late enough times its added drawdown can steepen the derivative; a roughly doubled ideal response can occur in a simple geometry.
An independent reader can challenge this interpretation as follows: Look for a consistent transition lag across observation wells and compare geological evidence of a barrier. The error to avoid is concrete: Rising derivative alone cannot distinguish a fault from unlogged pumping changes or an aquifer property transition. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Consider recharge boundary
An ideal constant-head boundary may be modelled using a negative image well, representing support that limits increasing drawdown. An affected late-time curve can flatten in drawdown and yield a derivative trending toward smaller values in a suitable ideal geometry.
An independent reader can challenge this interpretation as follows: Inspect streams, lakes, hydraulic connection and head measurements at the putative boundary. The error to avoid is concrete: A flattening late derivative does not prove that a nearby river actually supplies the pumped aquifer. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Separate leakage from boundaries
Vertical leakage through an aquitard can moderate late drawdown in a leaky confined aquifer even when no lateral recharge boundary exists. The time evolution may resemble a discharge-supporting lateral feature in one observation well.
An independent reader can challenge this interpretation as follows: Fit plausible Hantush–Jacob alternatives and examine nested piezometers above and below the pumped layer. The error to avoid is concrete: Naming the water source from one derivative curve alone exceeds what the measurement identifies. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Recognize delayed yield
In unconfined settings, water-table drainage and vertical flow can cause a multi-stage drawdown response that differs from a simple confined radial-flow signature. A derivative dip or transitional segment may reflect delayed gravity drainage rather than a river or geological fault.
An independent reader can challenge this interpretation as follows: Use screen-depth information, aquifer saturation and an unconfined-aquifer model as a competing explanation. The error to avoid is concrete: Forcing a confined formula onto a water-table aquifer can make its natural response appear anomalous. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Beware composite transmissivity
A pumping disturbance spreading from a permeable region into a less transmissive region can produce a changing derivative even without a discrete no-flow wall. The spatial volume sampled by the test grows with time, so the effective hydraulic resistance can evolve.
An independent reader can challenge this interpretation as follows: Compare lithologic maps, distant wells and hydrostratigraphic changes with candidate composite-aquifer models. The error to avoid is concrete: A continuous property gradient should not automatically be translated as a single sharp buried wall. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Investigate fracture-controlled flow
Some fractured-rock responses are not well represented by homogeneous radial Darcy flow, especially during transitions between fractures and rock matrix. Flow dimensionality and connected pathways can produce distinctive slope changes unrelated to the simplest image-well constructions.
An independent reader can challenge this interpretation as follows: Compare fracture orientation and packer or televiewer evidence with the hydraulic data. The error to avoid is concrete: A diagnostic plot without a geological conceptual model is a pattern-recognition exercise, not a reliable fracture map. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Differentiate nearby pumping
A second production well starting halfway through the test adds a new hydraulic stress to the same observation point. Its pressure signal may create rising drawdown and derivative features that resemble a late impermeable boundary.
An independent reader can challenge this interpretation as follows: Check metered withdrawals at neighbouring wells, regional telemetry and stepwise superposition calculations. The error to avoid is concrete: Treating the only recorded pump as the only active stress is an untested assumption. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Watch pump shutdown
An unnoticed stop, generator switch or abrupt flow reduction can flatten the drawdown series and lower its derivative. The response after shutdown is recovery, not evidence that the aquifer encountered a recharge feature.
An independent reader can challenge this interpretation as follows: Inspect discharge records and use recovery-time analysis for the affected portion. The error to avoid is concrete: A visually appealing late-time plateau should not survive quality control when the pump was no longer operating. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Diagnose sensor saturation
A water-level sensor reaching its measurement range may show a stable top or bottom value despite a moving water column. Such artificial clipping can resemble a stabilised drawdown or constant-head boundary.
An independent reader can challenge this interpretation as follows: Examine raw pressure limits, instrument specifications and manual verification readings. The error to avoid is concrete: Interpreting a hard electronic limit as aquifer physics is preventable with basic instrument checks. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Separate background recharge
Rainfall, irrigation or managed recharge during a test may alter hydraulic head independently of the pumped well. Its influence can be spatially uneven and temporally lagged, confusing the derivative as a boundary signal.
An independent reader can challenge this interpretation as follows: Compare unaffected control locations, precipitation intensity, infiltration delay and field notes. The error to avoid is concrete: Subtracting rain timing mechanically without a causal infiltration model can be as misleading as ignoring it. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Use multiple wells and geometry to test interpretations
Compare transitions at two radii
A disturbance diffuses outward on a timescale related approximately to length squared divided by hydraulic diffusivity, subject to model geometry. Two wells at different distances should not simply have identical boundary-onset times by assumption.
An independent reader can challenge this interpretation as follows: Review distance from stress and candidate boundary and test predicted transition timing with a forward model. The error to avoid is concrete: Copying a late-time diagnostic label from one well to another ignores the spatial experiment. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Resolve directional response
Observation wells north and south of a pumping well may see different effects when aquifer boundaries, permeability or recharge are directionally asymmetric. A directional contrast is potentially informative, but different screens and instrument histories can also cause it.
An independent reader can challenge this interpretation as follows: Standardize screen intervals, datum and discharge history before comparing orientations. The error to avoid is concrete: Inferring anisotropy from raw curves before ruling out construction effects conflates causes. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Estimate diffusivity carefully
Hydraulic diffusivity in a simplified confined framework is T/S, with dimensions square metres per day when T is area per day and S is dimensionless storativity. A transient response time and distance jointly inform diffusion, while the derivative amplitude mostly constrains T in the ideal radial regime.
An independent reader can challenge this interpretation as follows: Check that a fitted storativity is plausible for the hydrogeology and data geometry. The error to avoid is concrete: A single observation-well derivative typically cannot isolate every geometric and storage parameter uniquely. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Identify the information radius
A pumping test does not have a single sharp radius of investigation; hydraulic disturbances attenuate continuously in space and time. Phrases such as ‘the test saw everything within 300 metres’ overstate the precision of a diffusive, threshold-dependent process.
An independent reader can challenge this interpretation as follows: Use response thresholds and model predictions to describe which distances are meaningfully constrained. The error to avoid is concrete: Treating a heuristic influence radius as a mapped boundary creates unwarranted certainty. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Use geological ground truth
Core, lithology, structural mapping, borehole images and water chemistry provide different lines of evidence for hydraulic boundaries. A fault mapped independently can make a hydraulic hypothesis more plausible, while disagreement can prompt re-examination.
An independent reader can challenge this interpretation as follows: Document how each dataset bears on the model and whether it could have been influenced by the same assumption. The error to avoid is concrete: Repeatedly restating one derivative curve in different words is not independent corroboration. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Cross-check measured hydraulic gradients
If a supposed recharge boundary is connected to a stream, the difference between stream stage and adjacent groundwater head matters. A stream may be losing or gaining at different times; proximity alone says nothing about actual flux direction.
An independent reader can challenge this interpretation as follows: Measure contemporaneous stream stage and well heads, while acknowledging uncertain bed conductance. The error to avoid is concrete: A visually nearby river cannot be declared the recharge source without hydraulic connection evidence. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Test a boundary with superposition
Ideal boundary solutions add conceptual image wells, and total drawdown comes from superposing their effects in a linear framework. The method predicts a whole time series and spatial pattern, not only the direction of a derivative bend.
An independent reader can challenge this interpretation as follows: Plot observed and calculated drawdown plus residuals at multiple locations. The error to avoid is concrete: A model that matches one late point but fails early and intermediate behaviour has not explained the full test. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Model aquifer thickness
If saturated thickness changes appreciably in an unconfined aquifer, transmissivity may depend on drawdown rather than remain constant. Derivative changes can result from losing saturated conductive thickness as the water table falls.
An independent reader can challenge this interpretation as follows: Check saturated thickness and whether the discharge is small enough for a constant-transmissivity approximation. The error to avoid is concrete: A thickness-dependent material response should not be recast as a lateral barrier merely to preserve a simple model. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Check well interference across depths
Observation wells screened at distinct depths can display lagged or muted responses when aquitards restrict vertical transfer. These variations can illuminate interlayer connectivity, but also invalidate a single-layer radial model.
An independent reader can challenge this interpretation as follows: Plot depth-specific heads, hydraulic separation and the stress history together. The error to avoid is concrete: Combining wells across an aquitard into one derivative trend erases an important physical distinction. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Recheck after pumping ends
A recovery record offers a new stress history that can be predicted using superposition of pumping and stopping contributions. A genuine conceptual model should have explanatory value beyond the pumping-only curve.
An independent reader can challenge this interpretation as follows: Compare recovery residuals and test whether the same parameters fit both stages. The error to avoid is concrete: Fitting a boundary only during pumping and ignoring contradictory recovery evidence invites overinterpretation. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Independent verification and communication
Separate model fit from proof
Many hydraulic parameter combinations can create similar drawdown records over a limited monitoring window. An acceptable residual alone does not prove uniqueness of the model or the identity of a remote boundary.
An independent reader can challenge this interpretation as follows: Compare at least one competing conceptual explanation and discuss what observation would separate them. The error to avoid is concrete: A low numerical error value is not a geological certificate. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Report measured and inferred values
Measured quantities include time, pressure, water-level elevation and pumping rate; inferred quantities include transmissivity, storativity and boundary location. Mixing these categories makes a parameter estimate sound like a direct field reading.
An independent reader can challenge this interpretation as follows: Use a report table with units, measurement methods, assumptions and uncertainty for each item. The error to avoid is concrete: A derived transmissivity should never be described as a sensor’s direct output. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Use residual patterns
Residuals are observed minus predicted drawdown at each time. Structure remaining in residuals reveals model inadequacy that average error can hide. Systematic positive residuals late in the test can signal missing processes even when the mean residual is near zero.
An independent reader can challenge this interpretation as follows: Plot residuals against log time for every observation well, not just a global fit statistic. The error to avoid is concrete: Averages can cancel positive and negative errors while a key physical misspecification persists. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
State the data window
An interpretation valid for the middle of a test may be unsupported at startup or late times, when constraints differ. Restrict claims to the interval in which the proposed response is actually observed.
An independent reader can challenge this interpretation as follows: Mark excluded points, time cutoffs and reasons on plots. The error to avoid is concrete: Removing inconvenient measurements without explaining them undermines otherwise careful interpretation. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Validate with hold-out data
A conceptual model calibrated to all wells simultaneously can still be tested against an independently held-out well or later event. Successful prediction provides stronger support than re-fitting a new parameter set every time.
An independent reader can challenge this interpretation as follows: Reserve a record for checking where the design and data volume permit it. The error to avoid is concrete: Calling an in-sample match an independent check mistakes calibration for validation. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Perform a transfer test
Suppose a second well has a late derivative decline but discharge falls from 240 to 180 cubic metres per day at the same moment. A 25 percent rate reduction offers an immediate alternative explanation before invoking leakage or constant-head support.
An independent reader can challenge this interpretation as follows: Normalise or simulate the rate history and see whether the decline remains after correction. The error to avoid is concrete: Naming a recharge boundary first is premature because the forcing itself changed. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Recalculate sensitivity
A derivative calculated as 0.26 metre with reasonable window choices might range from 0.23 to 0.29 metre. At 240 cubic metres per day, the ideal radial estimate spans about 83 to 66 square metres per day rather than one exact transmissivity.
An independent reader can challenge this interpretation as follows: Compute both endpoints with T=Q/(4πd) and carry the assumptions into the reported range. The error to avoid is concrete: Using more decimals does not make the model less sensitive to its input derivative. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Choose exact words
The term ‘indicates’ should be reserved for evidence of a measured feature; ‘is consistent with’ correctly expresses a nonunique conceptual interpretation. English precision matters because ‘proves a fault’ and ‘supports a possible barrier’ communicate very different levels of certainty.
An independent reader can challenge this interpretation as follows: Ask a reader to distinguish the observed derivative trend from the cause proposed in the sentence. The error to avoid is concrete: Technical translation succeeds only when the strength of the language matches the strength of evidence. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Use a diagnostic sequence
Begin with raw data integrity, then identify response regimes, then test physical models, and only then attach geological interpretation. This order prevents a preferred story from dictating which data to accept and how much to smooth.
An independent reader can challenge this interpretation as follows: Have a second analyst reproduce the time origin, derivative and chosen diagnostic windows. The error to avoid is concrete: Starting with a boundary name and forcing the graph to fit it reverses the direction of sound inference. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Write an auditable conclusion
A responsible conclusion identifies a measured derivative transition, a plausible mechanism, alternative explanations and a discriminating follow-up test. For example, ‘Late drawdown steepens after stable radial response; an impermeable boundary is compatible, but rate changes and property contrasts require checks’ is defensible.
An independent reader can challenge this interpretation as follows: Attach the actual Q history, observation coordinates, raw levels and analysis script or calculation steps. The error to avoid is concrete: Readers should be able to see which part of the conclusion is data and which is hypothesis. In a proper test report, retain the underlying readings and state what this single diagnostic feature can and cannot establish. Ask whether an alternative conceptual model would produce the same visual signature, because a curve is a measurement of response rather than a photograph of subsurface conditions.
Independent worked test: compare two derivative claims
A synthetic test pumps 240 cubic metres per day at a steady rate. Drawdown in an observation well increases from 0.25 metre at 10 minutes to 0.85 metre at 100 minutes. The change is 0.60 metre over one factor of ten in time, so ds/dln(t) is 0.60/2.302585 = 0.2606 metre. Under the eligible late-time radial approximation, T = 240/(4π × 0.2606) = about 73.3 square metres per day. This arithmetic can be checked without a plotting program, but it does not by itself prove the radial approximation holds.
Now imagine that a later period has a derivative of 0.52 metre. This is twice the early value, yet the data alone cannot identify a no-flow boundary: changing Q, a transition to lower transmissivity and instrument issues also matter. A model with an image pumping well is one candidate. The investigator must compare competing forward curves across multiple observation wells, inspect mapped structure and ensure Q was constant. A label on a single diagnostic plot is not an independent geological observation.
Another synthetic record shows a derivative declining after a long stable period. The tempting headline is ‘recharge boundary’. Before accepting it, check whether pumping rate decreased, a second well stopped pumping, rainfall affected baseline head, or vertical leakage supplies the stressed aquifer. A correct explanation should predict both the sign and timing of changes at more than one distance. The independent check is to retain the original rate data and compute predicted changes from competing mechanisms.
Transfer problem one: drawdown changes by 0.40 metre from 5 to 50 minutes. The per-decade slope is 0.40 metre, but ds/dln(t) is 0.40/2.302585 = about 0.1737 metre. Transfer problem two: if Q is 180 cubic metres per day and derivative 0.30 metre, ideal radial T = 180/(4π×0.30) = approximately 47.7 square metres per day. Answers are contingent on the flow regime and matching units, not literal field values.
Transfer problem three: two curves both flatten late, but one coincides with a measured discharge drop and the other coincides with stable Q and a consistent head response near a mapped stream. The second is a stronger candidate for boundary investigation because one competing explanation has been checked, though the stream connection still needs testing. Transfer problem four: a boundary interpretation matches the pumped well and fails every observation well. That mismatch favours re-examining well losses or the conceptual model rather than declaring validation.
FAQ: Can I calculate storativity from one flat derivative? Not reliably from the derivative magnitude alone; its radial plateau mainly constrains transmissivity, while storage requires timing, geometry, distance and model assumptions. FAQ: Does a barrier always double the slope? No. Doubling is a particular idealised asymptotic image-well result, not a universal field rule. FAQ: Can I read a derivative without a discharge log? You can inspect it descriptively but should not attribute features confidently. FAQ: Should I smooth until it looks clean? No; justify and disclose smoothing, and check the raw record.
Sources and connected learning
Primary methodology: USGS Halford and Kuniansky, aquifer-test and slug-test analysis; USGS pumping test methods and boundaries; Singh (2002), boundary identification. Compare the Cooper–Jacob constant-rate interpretation and recovery interpretation. For interpreting qualifiers and domain words, see eduKate vocabulary, words and meaning, and English, context and transfer.