Water-table fluctuation recharge uses changes in an unconfined groundwater level to estimate the depth of water newly added to saturated storage. The working equation is recharge depth R = Sy × Δh, where Sy is dimensionless specific yield and Δh is the water-table rise attributed to recharge in metres. The multiplication is elementary; the hydrologic interpretation is not. A storm-time rise can include pumping recovery, nearby surface-water influence or an instrument artefact.
To read a well hydrograph responsibly, first check whether the vertical axis is water elevation or depth below a reference point, then identify the pre-event recession, mark the peak and estimate the water level that would have occurred without the event. The recharge-related rise is the observed water level minus that counterfactual baseline, not necessarily the difference between the first and last measurements.
This guide explains why specific yield differs from porosity, how to turn a rise into millimetres of recharge, how to test uncertainty and when not to use the method. All numerical cases below are illustrative teaching examples, not field measurements. The technique is especially useful for shallow unconfined aquifers with clear event responses, documented well construction and defensible estimates of drainable storage.
The storage story behind the hydrograph
Rainfall is not synonymous with recharge. A storm’s water may leave as surface runoff, evaporate, be intercepted by vegetation, fill a soil-water deficit, flow laterally above a tight horizon, or ultimately cross the water table. The hydrograph method estimates the last process indirectly by observing the response of saturated storage. It therefore needs a plausible causal connection from surface input to aquifer rise.
Specific yield is the gravity-drainable fraction of saturated aquifer volume under the relevant conditions. If Sy is 0.18, then a simplified vertical storage calculation assigns 0.18 m of additional water depth to each metre of recharge-attributed water-level rise. It does not mean that 18 percent of rainfall becomes recharge. It describes an aquifer property, not a universal conversion efficiency from atmospheric precipitation.
Porosity includes all void volume, including pores in which water remains retained by capillary forces. A fine sediment may have substantial porosity but relatively little readily drainable water. Confusing porosity with Sy inflates results. Delayed drainage, water-table depth and stratification also complicate a single fixed Sy, particularly over very short response intervals.
A monitoring well reads head within the interval hydraulically connected to its screen. Before naming that head a free water table, inspect well construction, screen depth, perched zones and any confinement. A confined aquifer may undergo elastic pressure change without the drainable-storage response assumed by R = Sy × Δh. Applying unconfined specific yield to that signal is physically wrong.
Decode the vertical axis
A graph of water-table elevation ordinarily rises upward when groundwater rises. A graph of depth to water ordinarily decreases numerically as groundwater rises. If depth falls from 6.20 m to 5.94 m below the same reference, the water level has risen 0.26 m. If someone instead treats 5.94 as a lower elevation, the conclusion is reversed.
A tape measurement from top of casing and a pressure-transducer reading cannot be mixed until reduced to a common datum. The top-of-casing height should be surveyed or otherwise reliably fixed. A logger shifted upward during cleaning can create a false water-level step. Any sudden transition coinciding with field maintenance needs reconciliation before interpretation.
Absolute pressure loggers may require atmospheric-pressure correction. Vented transducers handle this differently. A barometric-compensation error can imitate a change in hydraulic head, particularly when the actual event is small. Read the instrument processing notes and compare manual dip measurements with logger output rather than subtracting an assumed atmospheric signal without knowing the instrument type.
Check time zones. Rain may be logged in local time while a groundwater sensor records UTC. A real response must follow the input, allowing travel time. A graph that seems to show groundwater rising hours before rainfall may reflect a clock offset rather than surprising hydrology. Align the clocks and record the transformation in the analysis notes.
Recover the no-recharge level
Even without a storm, an unconfined aquifer can decline through discharge to streams, groundwater evapotranspiration or pumping. The hydrograph before the event helps construct a hypothetical recession continuation. The relevant rise is the observed peak above the predicted no-recharge level at the same time, not simply the peak above yesterday’s reading.
Suppose elevation is 106.20 m Monday, 106.16 m Tuesday and 106.12 m Wednesday. Over that short interval, the observed decline is 0.04 m per day. If comparable conditions continued, an illustrative Friday no-recharge estimate is 106.04 m. If Friday’s observed peak is 106.38 m, the recharge-attributed rise is 0.34 m, not the 0.26 m difference from Wednesday.
A linear recession is a transparent teaching approximation, not a universal hydrograph law. Long recessions often curve because discharge and hydraulic gradients change. Choosing a baseline fit should be informed by the shape and length of the dry-period record. Show its fitting interval and sensitivity rather than hiding it in a spreadsheet.
Short storms can overlap. If a second storm arrives before the first response recedes, separating individual contributions may require a model or may be impossible with the observations. It is preferable to estimate a combined-period recharge than to invent separate event-specific numbers that have no defensible baseline.
The selected peak deserves scrutiny. A pressure spike or single logging glitch should not define the maximum. A consistent crest over several observations is more credible, although excessive smoothing may remove real short-term variation. Keep the original trace alongside any filtered version and identify every data-editing rule.
From metres of head to millimetres of recharge
For Δh = 0.28 m and Sy = 0.15, R = 0.042 m. Multiplying by 1,000 gives 42 mm of recharge. Notice the equation converts a water-table rise to an equivalent water depth; it does not calculate water velocity or the time taken to percolate through the unsaturated zone.
Write specific yield as 0.15, not 15, when using this multiplication. Entering a percentage as if it were a fraction gives a result one hundred times too large. An explicit unit audit catches this error: Sy has no unit, Δh is metres, R remains metres before conversion to millimetres.
Rainfall provides a plausibility benchmark but not an identity. If the event received 70 mm of rainfall and the estimated recharge is 42 mm, the apparent ratio is 60 percent. It may be reasonable for some wet antecedent conditions but cannot be generalised across seasons or regions. Delayed water from earlier storms and focused recharge can also affect the comparison.
If Sy is uncertain between 0.10 and 0.20, the same 0.28 m rise yields 28–56 mm. Reporting 42.000 mm creates false certainty. A range signals the controlling uncertainty and directs attention toward the measurements that could improve the estimate.
A different recession estimate may alter Δh as strongly as Sy does. If the assumed no-recharge level is revised upward by 0.04 m, the rise falls from 0.28 to 0.24 m, giving 36 mm at Sy = 0.15. If it is revised downward by 0.04 m, recharge becomes 48 mm. Both inputs deserve sensitivity checks.
A complete event calculation
Consider a shallow well in an unconfined sand body with a consistent elevation datum. The observed event peak is 48.39 m. A pre-event recession model predicts 48.07 m at that instant in the absence of recharge. Their difference is Δh = 48.39 − 48.07 = 0.32 m.
An independent aquifer assessment supplies Sy = 0.16. The estimated event recharge is 0.16 × 0.32 = 0.0512 m = 51.2 mm. A clear report includes the peak date, the recession basis, specific-yield source, data quality and the input period. It does not simply announce that the aquifer received exactly 51.2 mm.
For a lower-bound combination of Sy = 0.12 and Δh = 0.28 m, recharge is 33.6 mm. For Sy = 0.20 and Δh = 0.36 m, it is 72.0 mm. These are illustrative scenario bounds rather than a formal 95 percent confidence interval. Correlated errors or another unrecognised driver can place the true value outside them.
If the measured rain total is 85 mm, a central estimate of 51.2 mm represents roughly 60 percent of that storm depth. The ratio alone cannot prove the allocation. The observed aquifer response might include recharge from previous rainfall or lateral groundwater redistribution. The numerical check asks whether the result is plausible, not whether it is correct by definition.
What a flat hydrograph can conceal
Suppose a well was falling 0.05 m per day before a storm. After two days its observed level equals the starting level. Naively the change is zero. But a comparable no-recharge trajectory would have fallen 0.10 m, so the observed stable level may be 0.10 m above the counterfactual. At Sy = 0.12, this corresponds to 12 mm of estimated recharge.
The calculation is only as valid as the recession assumption. If the earlier decline was temporary pumping that stopped during the storm, the continued-decline baseline is invalid. A physically interpretable no-recharge trajectory is indispensable. The correct conclusion may be that no reliable recharge estimate can be extracted from this interval.
A hydrograph that does not rise may still be compatible with infiltration crossing the water table when lateral drainage offsets it, the rise occurs between infrequent measurements or the signal falls below measurement accuracy. A single monitoring well is not a direct flow collector.
Why a spectacular rise may be misleading
A well next to a river can rise because the river stage rises and water moves into bank sediments. The stored water may later return to the river. Multiplying such a response by Sy as though it were vertical rainfall recharge confuses a hydraulic boundary effect with local input across the water table.
Pair the hydrograph with river-level records and other wells farther inland. A delayed, attenuated propagation from the river suggests bank-storage influence. The measured head change is real, but its source and water-balance meaning are not established by correlation with rain alone.
Stopping a nearby production pump can also generate rapid head recovery. Plot operating schedules and compare wells inside and outside the pumping influence. A rise beginning at pump shutdown before any rainfall is evidence against interpreting the whole segment as rain recharge.
Shallow groundwater can show daily vegetation-driven oscillations. Water levels may decline during active daytime evapotranspiration and recover at night. Counting each overnight rise as recharge systematically invents water input. Inspect diurnal timing, weather and seasonality before classifying events.
Quality control before any interpretation
Read the field history. A manual dip taken after pumping a well may not represent static groundwater elevation. A pressure logger raised for cleaning and reinstalled at a different depth can introduce an abrupt offset. If an apparent event begins exactly when equipment was handled, compare check measurements, logger depths and calibration notes before invoking recharge.
A record needs a stable measuring reference, a known time base and an adequate sample rate. Where an event lasts hours, one daily reading may miss the peak and underestimate a response. Conversely, high-frequency pressure spikes may exaggerate the maximum. Explain which observations were considered valid and how any smoothing was performed; never erase inconvenient readings without a documented reason.
Plot precipitation and groundwater on aligned clocks. Rainfall accumulation measured for an hour ending at 09:00 is not necessarily timed like an instantaneous groundwater measurement at 09:00. Distinguish interval totals from instantaneous observations. Verify dates, time zones and daylight-saving adjustments where applicable. A timing mistake can change a plausible causal sequence into a false one.
Inspect the well construction report. A shallow screen crossing the water table responds differently from a deep piezometer installed in a partly confined unit. A damaged screen or poorly sealed annulus may connect zones that normally behave separately. The label monitoring well does not establish that a record meets the assumptions of an unconfined storage model.
How the hydrograph responds to rainfall history
Antecedent moisture matters. When soil is dry, early rainfall may fill unsaturated storage without raising groundwater promptly. Later rain under wetter conditions may produce a larger or quicker head response. Two storms of equal measured depth can therefore yield different hydrograph signatures without any error in rainfall measurement.
In layered soils, perched water may form above a low-permeability horizon. A shallow piezometer can respond rapidly to this local saturation while a deeper water table changes slowly. Both measurements can be genuine but describe different storage systems. Do not merge their peaks into one recharge estimate unless their hydraulic relationship is established.
Travel time is not fixed. Infiltration can move through pores, roots, cracks or preferential pathways, and the groundwater response can integrate multiple events. Water-table fluctuations record a state change in saturated storage; they do not by themselves tell the age or precise path of individual water molecules reaching the aquifer.
A storm separated from the preceding one by only a few hours may produce one extended rise with shoulders. Selecting a local maximum for every small pulse would double-count storage. Mark combined events when the hydrograph does not allow unique separation, and retain the rainfall timeline as supporting evidence.
Seasonal changes in vegetation and groundwater depth can change the sensitivity of a well to rainfall. An estimate transferred from the wet season to the dry season may fail if evapotranspiration, initial soil storage or aquifer boundary conditions differ. Explain whether comparisons refer to similar conditions rather than treating every event as interchangeable.
Understanding specific yield in the real aquifer
The numerical sensitivity to Sy deserves a separate paragraph in any technical report because its uncertainty may dominate all instrument error. A logger capable of millimetre resolution does not make the recharge estimate millimetre-accurate if Sy is only known within a broad geological interval. Precision of observation and precision of interpretation are different.
Sediment samples may provide texture and porosity information, but a field-scale effective Sy incorporates heterogeneity and drainage processes that small samples can miss. Coarse sand, fine sand, silt, fractures and cemented layers may respond differently. In a layered unit the water table may cross materials with different drainable storage as it rises.
A pumping or aquifer test can constrain hydraulic properties, but examine which model was used. Delayed-yield behaviour and early elastic storage can affect the interpretation. If a derived storage quantity represents confined storativity rather than unconfined Sy, simply substituting it into a water-table fluctuation calculation is not a valid shortcut.
The Sy appropriate for a rapid rise might differ from the value inferred after several days of gravity drainage. Entrapped air can also change the transient water-level response. A water table moving through capillary-fringe material is not always governed by instantaneous release from fully equilibrated pores. Such effects are reasons to describe method limits, not reasons to abandon quantitative checking.
For a hypothetical rise of 0.45 m, three plausible Sy candidates of 0.06, 0.12 and 0.24 yield 27, 54 and 108 mm. The fourfold range arises before debating the recession. If the event rainfall was only 60 mm, the high-Sy scenario demands an explanation such as antecedent input, focused recharge, inappropriate parameter choice or another head driver.
Do not select Sy after seeing the recharge result merely to make it agree with rainfall. That would be circular reasoning. Estimate or justify storage properties independently, then compare results against precipitation and water balance. An uncomfortable mismatch is potentially informative evidence about assumptions or field conditions.
Identifying non-recharge signals
A monitoring well affected by nearby pumping can show a sudden recovery when the pump stops. That recovery may overlap rain, making attribution ambiguous. Compare nearby pumping schedules and reference wells. Where the hydrograph is dominated by pumping, a different analysis may be needed; applying the simple recharge equation to the whole rise would overstate incoming water.
River-level changes cause lateral exchange with connected aquifers. A stream flood can drive water into bank sediments and lift well levels independently of local vertical infiltration. Paired river stage and groundwater records reveal response lag and attenuation. Some bank storage subsequently drains back to the stream, so it must not automatically be counted as permanent recharge.
Atmospheric pressure can perturb both physical groundwater head and an uncompensated sensor reading. Check logger type and processing. The groundwater barometric-efficiency guide offers a related diagnostic workflow. Avoid treating pressure-correction artefacts as centimetres of newly replenished storage.
In coastal or other responsive aquifers, tides can produce recurrent oscillations. These may resemble a succession of small recharge events if only the rising limbs are counted. Compare periodicity and phase with known tidal forcing. The tidal response analysis shows how such signals have a separate hydrogeological meaning.
Evapotranspiration can produce daily head cycles. A shallow water table sometimes falls in daylight during strong plant water use and rises overnight when uptake diminishes. The upward part alone is not an atmospheric infiltration event. Temperature, light and vegetation records help identify these cycles.
Surface loading from heavy rainfall can in some settings induce short pressure or mechanical effects preceding the slower arrival of infiltrating water. Rapid head changes do not invariably imply the same rapid pore-water travel. The explanation needs to fit the aquifer type and the timing; avoid equating hydraulic pressure propagation with tracer travel.
Why the spatial scale matters
A single well does not represent every part of a catchment. Recharge may be focused in sandy depressions, fractured zones, infiltration basins or stream beds while nearby paved or clay-covered ground contributes little. Site selection determines what the method can say about broader areas.
Compare wells placed in distinct hydrogeologic settings. Similar event timing with different amplitudes can arise from different Sy, recharge flux, water-table depth or boundary conditions. The amplitudes alone cannot resolve those causes. A network creates useful constraints but still needs geological interpretation.
Imagine two representative landscape units, one of 3 km² with estimated recharge 40 mm and another of 7 km² with 100 mm. Their area-weighted mean is (3×40 + 7×100)/10 = 82 mm. The simple mean of 70 mm would not represent area-weighted depth. The calculation is legitimate only if the well estimates truly represent their assigned areas.
For a verified area-average recharge of 82 mm over 10 km², equivalent recharge volume is 0.082 m multiplied by 10,000,000 m², or 820,000 m³. This multiplication converts depth to volume; it does not validate the area-average assumption. An impressive large volume can still rest on weak spatial evidence.
Monthly and annual estimates
Summing event estimates requires careful separation. If an autumn rise has not returned to recession before the next event, the two peaks may represent one complex response. Count the net storage response without duplicating overlapping water. A ledger showing start, end, recession model and event classification is more reproducible than a list of isolated peaks.
An annual total must acknowledge missing periods. A failed logger during a wet month creates uncertainty that cannot be eliminated by averaging surviving dry-season events. Report observed-period recharge separately from any explicitly modelled gap estimate. Users need to know which portions are measured and which are inferred.
Long-term groundwater decline is not proof of zero recharge. The level may fall because pumping and natural discharge exceed replenishment. Conversely a long-term rise can follow reduced withdrawal rather than exceptional recharge. A net head trend is a balance of multiple fluxes and boundary effects.
Comparing years also requires attention to land-use changes. New drainage infrastructure, vegetation change, irrigation and artificial recharge can alter pathways and local water-table response. A method calibrated under one set of conditions should be retested if the hydrologic system has materially changed.
A series of 12 event estimates could total 180 mm, but that sum should not automatically be called annual recharge if the monitored window spans only eight months. State the actual observation interval and the events included. The time denominator changes the meaning of a summary number.
Independent methods and competing explanations
The USGS water-table fluctuation method explicitly rests on suitable well hydrographs, specific yield and an estimate of head rise relative to recession. Its limitations guidance addresses the difficulty of isolating recharge from other causes and the uncertainty in Sy. Use these methodological references to check assumptions, not merely to decorate a bibliography.
The chloride mass-balance recharge guide explains a different evidence stream. Chloride inputs and groundwater chloride concentrations can constrain recharge under appropriate assumptions, while the hydrograph method emphasises transient head and storage. Agreement is informative only when temporal and spatial scales match.
Suppose a set of suitable hydrographs suggests 140 mm per year, while an independent long-term chloride approach suggests 85 mm. The difference could reflect uncertain Sy, unrecognised chloride sources, focused infiltration, nonmatching time windows or incorrect attribution of level rises. It is not scientific to choose the higher or lower estimate simply because it suits a planning objective.
Soil-water monitoring, unsaturated-zone models, lysimeters and water-balance studies can add independent constraints. Recharge estimates are strongest when several observations support one explanation without simply reusing the same assumption. Two models based on the same uncertain Sy are not truly independent confirmations.
When evidence conflicts, state precisely what would discriminate among hypotheses. An additional reference well, reliable pumping log, better aquifer-test estimate or deeper hydrograph history may resolve more uncertainty than fitting the same event curve with another sophisticated algorithm.
Transfer problem: reading an inverted axis
A well is plotted as depth below ground surface. Depth falls from 5.40 m to 5.09 m at the peak. The water rose 0.31 m, because smaller depth means higher water. If the projected no-recharge depth at the peak is 5.48 m, the recharge-related rise is 5.48 − 5.09 = 0.39 m.
At Sy = 0.13 the recharge estimate is 0.0507 m, or 50.7 mm. Two errors to watch for are using 0.31 m without the recession correction and interpreting a decline in depth value as falling water. Explicitly draw the water surface relative to ground if the axis convention is confusing.
Transfer problem: percent and fraction
A report lists Sy as 18 percent and the hydrograph-derived rise as 0.22 m. Convert 18 percent to 0.18. Then R = 0.18 × 0.22 m = 0.0396 m = 39.6 mm. Multiplying by 18 instead gives 3.96 m, an obvious scale error.
If Sy is plausible only between 0.12 and 0.22, the range is 26.4–48.4 mm. Recording the input as 18 without a percent sign is a reproducibility hazard; writing 0.18 in the calculation table avoids ambiguity.
Transfer problem: recession correction
On day zero the water-table elevation is 22.50 m. The pre-event recession suggests it would reach 22.42 m by the peak time. The observed peak is 22.67 m. The observed gain from day zero is 0.17 m, but the rise attributable relative to recession is 0.25 m.
With Sy = 0.11 the result is 27.5 mm. Explain the extra 0.08 m: the aquifer was expected to decline before receiving the event contribution. A simple peak-minus-start calculation understates the head perturbation when recession is otherwise continuing.
Transfer problem: invalid causal order
A monitoring well rises at 06:00, pumping stops at 05:45 and recorded rain begins at 14:00. The first rise should not be attributed to the later rain. Examine recovery from pumping, logger handling and other earlier drivers. A rain event may produce a separate later response, but do not combine both rises without evidence.
There is no legitimate R = Sy × Δh answer for the first rise as rainfall recharge based on these observations. An inability to isolate the target process is a valid scientific finding and should be documented rather than filled with invented precision.
Transfer problem: overlapping storms
Rain event A produces a rising hydrograph, but event B begins before the level peaks. The analyst marks each rainfall time and assigns the entire later peak to event B. This is not justified because the aquifer may still be responding to A.
Instead estimate a combined-period change against a coherent counterfactual recession, or use a more detailed calibrated transient model if event separation is essential. The result should preserve the actual evidence resolution. Splitting a combined response into two precise numbers without independent constraints creates false detail.
Transfer problem: a spatial mean
Three independently characterised landscape units have recharge estimates of 30, 90 and 150 mm with assigned areas of 2, 3 and 5 km². The area-weighted total is (30×2 + 90×3 + 150×5)/10 = 108 mm. The arithmetic can be checked as 60+270+750=1080, divided by 10.
The simple unweighted mean is 90 mm and answers a different question. If the areas are not grounded in mapping and site conditions, even 108 mm is provisional. Describe how the wells represent these units before reporting a catchment value.
Frequently asked questions
Is every rise after a storm recharge?
No. Timing and physical mechanism must be examined. Pumping recovery, river rise, barometric effects, tides and measurement offsets can cause upward head movement. A plausible rain association is necessary in some settings but insufficient for volume attribution.
Why extrapolate a recession?
Without recharge, the water table may continue declining. The extrapolated level estimates this counterfactual baseline at the event peak. Subtracting it from the observed level isolates a response more relevant to added saturated storage than a simple before-and-after reading.
Does a flat groundwater line prove there was no recharge?
No. Recharge may offset a continuing decline, be too small to detect, arrive between measurements or be balanced by lateral outflow. The absence of an obvious rise limits what the selected hydrograph can prove rather than disproving all infiltration.
Is the formula valid in confined aquifers?
Not in its ordinary Sy-based form. Confined systems can show pressure changes controlled by elastic storage and different boundary responses. First establish that the observation well monitors an appropriate unconfined water table.
Can porosity replace specific yield?
Not automatically. Porosity includes retained and non-draining water. Specific yield is the relevant gravity-drainable storage fraction, subject to timescale and geologic conditions. Treating a porosity figure as Sy requires independent justification.
What should the final reported number look like?
State the event interval, water-level reference, fitted recession, derived Δh, Sy basis, calculated recharge in millimetres and uncertainty scenarios. Separate what was measured from what was assumed, modelled or inferred. Avoid unsupported decimal precision.
Does this method measure the groundwater recharge rate directly?
It estimates an event or period recharge depth from water-level behaviour. Dividing by a stated time interval can express an average rate, but temporal resolution and the distinction between recharge arrival and atmospheric rainfall matter. The instantaneous infiltration flux is not simply the derivative of every head rise.
How does learning this method improve technical reading?
It teaches the difference between a graph, a counterfactual assumption, a storage coefficient and an inferred flux. The hydrogeology introduction supplies broader context, and the worked-learning reference illustrates the value of methodical checking across subjects.
Advanced worked case: two recession models disagree
Suppose the event peak is 71.82 m, while a linear extrapolation places the no-recharge baseline at 71.49 m. A curved recession fit based on a longer pre-event period predicts 71.56 m instead. The two plausible rises are 0.33 m and 0.26 m. With Sy = 0.17, the corresponding recharge estimates are 56.1 mm and 44.2 mm. This spread is not sensor error; it is model-selection uncertainty that must accompany interpretation.
Which recession should be preferred? Inspect whether earlier declines were consistent, whether nearby pumping or river levels changed and how much extrapolation each fit requires. A very flexible model can imitate minor noise and generate a seductive baseline that has no independent hydrologic support. Choose a physically defensible recession, publish its fitting interval and test the competitor rather than pretending there is a unique mathematical answer.
An additional reference well outside the recharge-affected zone could help identify a regional background trend, although its geometry and aquifer connection must be considered. The purpose of independent evidence is not to force models to agree; it is to discriminate among assumptions. Where the record cannot distinguish competing baselines, the final result should remain an interval or a conditional estimate.
Advanced worked case: event detection threshold
A pressure logger reports variations of a few millimetres even during stable weather. A supposed recharge response of 3 mm of head should not receive the same confidence as a 300 mm rise merely because both are positive. Determine the measurement noise, datum stability and expected non-recharge variation before declaring an event. A signal-to-noise comparison and field context can prevent the false accumulation of dozens of tiny apparent recharge pulses.
For a hypothetical Sy of 0.15, a 0.003 m head fluctuation produces only 0.45 mm of calculated recharge. If the level uncertainty is ±0.006 m, the event is smaller than the uncertainty in its own observation. Reporting 0.45 mm as a verified input misstates the evidence. Document the detection limit and, where appropriate, treat that response as indistinguishable from background variation.
Setting a threshold is not an excuse to discard inconvenient data. It should be justified before examining desired totals and applied consistently to comparable periods. Retain rejected or ambiguous events in a separate audit table with reasons. A transparent classification lets another analyst see whether annual recharge depends on many marginal signals.
Advanced worked case: moving from head to volume safely
A 0.30 m recharge-related rise at Sy = 0.12 corresponds to 36 mm of water. If, after independent hydrogeologic mapping, that value is representative of a 0.75 km² zone, the corresponding volume is 0.036 m multiplied by 750,000 m², or 27,000 m³. This is a volume-equivalent storage calculation, not a measurement of individual groundwater paths.
Suppose adjacent mapped units have different lithology, water-table depth and response times. Applying the same 36 mm indiscriminately across them is not justified. The volume will scale exactly with the area entered, but numerical scaling does not repair an invalid assumption of uniform recharge. State which map polygons and observations support the extrapolation.
This is the broader technical lesson: water-table fluctuation analysis is a chain of linked interpretations. Check the raw measurement, build the no-recharge counterfactual, estimate storage fraction, perform the dimensional calculation, investigate rival causes and only then consider spatial or annual extrapolation. Any weak link can dominate the quality of the final recharge estimate.
Final interpretation rule
Describe what the hydrograph showed, why you attribute part of its behaviour to recharge, how Sy converts that rise into water depth and which observations could overturn your interpretation. A readable result is neither merely a formula nor merely a story. It is a traceable calculation embedded in a physically justified explanation.
Original methodological sources
USGS Water-Table Fluctuation Method.
USGS Issues and Limitations of the Water-Table Fluctuation Method.