TDR Soil Moisture, Permittivity and Calibration

A soil-moisture reading begins as a travelling electrical signal

Time-domain reflectometry, usually shortened to TDR, is a method for estimating soil water content from how an electrical pulse travels along metal rods embedded in the ground. The main output that readers want is often volumetric water content, the volume of water divided by the bulk volume of the soil. Yet the instrument does not directly count water molecules or extract a water sample. It observes a waveform, estimates the propagation characteristics of an electromagnetic signal and converts those properties into a water-content estimate using a calibration relationship.

A TDR soil-moisture report can include apparent dielectric permittivity, pulse travel time, electrical conductivity, volumetric water content, probe length, measurement depth and quality diagnostics. Each term answers a different question. Permittivity relates to how the soil mixture responds to an electric field; electrical conductivity relates to its ability to carry current; water content describes how much of the bulk soil volume is occupied by water. Translating a plot correctly means retaining the distinction between these linked but nonidentical quantities.

The practical appeal of TDR is that liquid water has a much larger relative dielectric permittivity than air or many dry mineral constituents at commonly used measurement frequencies. Adding liquid water to the space around the rods often slows the electrical signal, allowing moisture changes to be inferred. The relationship nevertheless depends on probe geometry, the soil matrix, salinity, temperature and contact conditions. This article develops the physics, numerical units, calibration choices, errors and transfer tests with illustrative calculations rather than pretending that one universal reading fits every field.

What a TDR waveform actually represents

An instrument launches a rapid electrical step or pulse down a cable and along a probe that acts as a waveguide. Changes in impedance produce reflections, much as a wave changes when it reaches a boundary between two different transmission-line conditions. The instrument records returning signal amplitude against elapsed time. Analysts identify positions in the trace corresponding to the start and end of the probe section, or use model-based waveform analysis to infer effective propagation time. Those choices can strongly affect a computed water-content number.

A simple schematic may show a first reflection where the cable meets the probe head and a later reflection near the rod ends. The separation between these features is related to round-trip travel along the effective probe length. Real waveforms may have broad or distorted features, multiple reflections, loss along the cable, and electrically conductive attenuation. A tidy diagram in a textbook is a teaching aid; it is not proof that every field trace provides two sharp, unambiguous endpoints.

If a field logger already reports volumetric water content, the underlying waveform analysis and calibration have usually been performed for the reader. That convenience does not remove the need to know the firmware settings and physical conditions. A sudden reported jump can result from genuine irrigation, altered electrical contact, a wet cable connection, changing salinity or an invalid endpoint selection. The raw or diagnostic waveform, when available, can help distinguish the possibilities.

Travel time, path length and the factor of two

For a simplified homogeneous waveguide, the signal’s effective velocity is approximately the speed of light in vacuum divided by the square root of apparent relative dielectric permittivity. If L is the physical one-way rod length and t is the measured round-trip time along that rod section, a simple formula is apparent permittivity equals the square of c times t divided by two L. Here c is about 0.300 metres per nanosecond when using that unit system. The factor of two accounts for travel down and back.

Suppose the effective rod length is 0.150 metres and the corrected round-trip travel time is 4.00 nanoseconds. Multiplying 0.300 metres per nanosecond by 4.00 nanoseconds gives 1.20 metres. Dividing by twice the rod length, 0.300 metres, gives four. Squaring gives apparent relative permittivity sixteen. This example is a controlled arithmetic illustration; real instruments may define the displayed travel length or trace interval differently. The user must not apply a round-trip formula again if the instrument has already converted the waveform to a one-way effective length.

Units are an excellent diagnostic. Nanoseconds and metres must be converted consistently; one nanosecond is one billionth of a second. If a spreadsheet treats four nanoseconds as four seconds, the computed permittivity becomes fantastically large. If the cable delay is included in the rod travel interval, the inferred soil permittivity can also be inflated. Before interpreting an impossible value, check the timing definition, effective length, cable correction and probe calibration rather than inventing an exotic soil.

Why apparent permittivity increases when water enters the soil

Dry soil contains solid mineral grains and air spaces. When those spaces hold more liquid water, the mixture’s effective electromagnetic response usually changes considerably. In ordinary mineral soils, that response often creates a useful monotonic relationship between apparent relative permittivity and volumetric water content. The response is not solely a property of liquid water molecules: the geometry of soil pores, electrical conductivity, bound water and instrument operating frequency also influence the effective signal.

Think of permittivity as an intermediate measurement rather than a synonym for wetness. If two sensors show apparent permittivity of eighteen and twelve, it is tempting to announce that the first soil contains fifty percent more water because eighteen divided by twelve is one and a half. That conclusion is not supported. The conversion from permittivity to volumetric water content is nonlinear and calibration dependent. Ratios of intermediate signals do not automatically equal ratios of the physical quantity eventually reported.

The same principle explains why a moisture sensor may provide several channels. Apparent permittivity indicates an electromagnetic property, bulk electrical conductivity provides information on losses or conductive behaviour, temperature offers environmental context, and a derived water-content field is the actual volumetric estimate. These fields should be interpreted together. A rising permittivity accompanied by unusual conductivity and a distorted trace might require examination before being treated as a clean irrigation signal.

The Topp relationship is empirical, not a conservation law

A widely cited calibration for many mineral soils is the Topp polynomial, which expresses volumetric water content as a third-degree function of apparent dielectric permittivity. Using dimensionless permittivity K, one form is theta equals minus 0.053 plus 0.0292 K minus 0.00055 K squared plus 0.0000043 K cubed. Theta is in cubic metres of water per cubic metre of bulk soil and is therefore a volume fraction. Multiplying by one hundred converts that fraction into volumetric percent.

At K equal to sixteen, the terms are minus 0.053, plus 0.4672, minus 0.1408 and plus 0.0176128. Their sum is about 0.2910 cubic metres per cubic metre, or 29.1 percent volumetric water content. Check the result: it is neither sixteen percent nor sixteen cubic metres of water. The polynomial is a calibration rule applied to the measured intermediate property, and it does not transform the permittivity reading into an exact universal water inventory.

This calibration is convenient because it frequently works reasonably for mineral soils. But organic soils, volcanic materials, strongly saline soils, some fine textures or unusual porosities may require site-specific calibration and manufacturer-specific processing. Some modern probes use additional conductivity corrections, firmware models or other dielectric methods. The technically faithful wording is water content estimated from apparent permittivity using the specified model, not water content directly measured with no assumptions.

Volumetric and gravimetric water content use different denominators

Volumetric water content theta-v means water volume divided by total bulk soil volume, with units often reported as cubic metres per cubic metre. Gravimetric water content theta-g means water mass divided by oven-dry soil mass, commonly kilograms per kilogram. They can be related if dry bulk density and an appropriate water density are known. Confusing them is especially easy because both may be displayed as percentages even though their denominators are different.

For a simple soil sample, suppose gravimetric water content is 0.20 kilograms of water per kilogram of dry soil and dry bulk density is 1.40 megagrams per cubic metre. Taking liquid water density as approximately 1.00 megagram per cubic metre, volumetric water content is 0.20 times 1.40 divided by 1.00, or 0.28 cubic metres per cubic metre. A field report that calls the gravimetric twenty percent equal to volumetric twenty percent has silently omitted a bulk-density conversion.

This distinction also matters when comparing TDR with the laboratory. A destructive soil core provides a mass-based comparison after weighing and drying, but turning that measurement into a volumetric calibration target requires an accurate sample volume and dry bulk density. Stones, roots, shrinking or swelling, uneven cores and drying-method choices can introduce uncertainty. A disagreement between the TDR and a laboratory number may be a denominator problem before it is evidence that the electromagnetic method failed.

Convert water content into a layer water depth

Volumetric water content is a fraction, not yet a depth of water stored in a profile. For a uniform layer of thickness D, equivalent water depth is theta-v multiplied by D. If theta-v equals 0.28 and the layer is 0.30 metres thick, the equivalent liquid-water storage is 0.084 metres, or 84 millimetres. This means that the volume of water in that hypothetical soil layer equals a water sheet 84 millimetres deep spread over the same horizontal footprint.

Now suppose a 20-centimetre layer increases from theta-v 0.22 to 0.27 after a rainfall event. The change is 0.05 of the layer volume. Multiplying 0.05 by 200 millimetres gives a ten-millimetre increase in equivalent stored water. That number does not prove ten millimetres of rainfall fell: water could also have entered from irrigation or upward flow, and some incoming rain may have run off, drained deeper or evaporated during the interval.

A single TDR probe samples a local volume around its rods. Extrapolating a sensor to an entire 30-centimetre layer assumes the sampled moisture represents that layer. In a layered soil, moisture may be very different at ten centimetres and thirty centimetres, and a probe mounted horizontally at one depth cannot describe the vertical profile alone. Profiles need multiple depths or a justified sampling design before a layer-integrated water storage is credible.

Probe geometry affects the physical sampling volume

Rod length, diameter, spacing and installation orientation influence the geometry of the electromagnetic field and the physical soil region contributing to the measurement. A short probe may resolve a shallower or smaller local volume than a long installation, but no simple cylinder of soil is sampled uniformly by every sensor. Sensitivity is often stronger near the metal rods, and the exact field distribution depends on the probe design and surrounding material.

Consider two sensors installed at the same nominal depth. One has rods embedded parallel to the soil surface; the other is pushed downward through several distinct layers. The first may sample more nearly one horizon while the second crosses materials with different moisture properties. Their displayed water-content numbers should not be compared as if the probes sampled identical volumes merely because both are labelled depth twenty centimetres. The orientation and full rod extent are part of the metadata.

Probe spacing and rod contact matter during installation. Large air gaps around metal rods change the local dielectric environment and can lead to an unrepresentative reading, often biased toward a drier response. Forcing a probe through stony soil can bend rods or create voids. A repaired field protocol should document how the probe was inserted, whether pilot holes were used appropriately, how disturbed soil was handled and whether initial readings were compared with nearby independent samples.

Why salinity can corrupt a clean-looking moisture trend

Water containing dissolved ions conducts electricity. Highly conductive soil solutions can attenuate the travelling pulse and distort the reflected waveform, making the effective travel time harder to identify. A TDR probe may also report bulk electrical conductivity from the same measurement system, but this quantity differs from the electrical conductivity of extracted pore water or a saturated soil paste. Those variables need their own definitions and calibration before being interpreted interchangeably.

Imagine an irrigated field where water content is fairly stable but fertilizer changes the ionic strength of the pore water. The waveforms may become more attenuated even though the liquid-water volume has not increased by an equivalent amount. If the user reads any conductivity-related change as wetness, the interpretation becomes misleading. Checking the conductivity output, waveform quality, irrigation chemistry and independent mass-based samples helps diagnose whether the apparent moisture trend is physically plausible.

Manufacturer limits are relevant, not just catalogue decoration. TDR probes designed for typical soils may have specified maximum conductivities beyond which their readings become unreliable, whereas shorter or specialised probes may be suited to more conductive environments. The reader should locate specifications for the exact model, not transfer a limit from a different device family. A claimed 0.001 precision in a laboratory does not override the possibility of major site-specific bias in salty or poorly contacted soil.

Temperature, freezing and water state complicate the story

Temperature affects the dielectric behaviour of water and the conductivity of the soil solution. Sensor response can therefore vary with temperature even if the actual liquid-water volume changes little. Instrument firmware may compensate for parts of this influence under specified conditions. A careful comparison between day and night or between seasons checks whether the correction is appropriate and whether the reported moisture change is larger than the expected temperature-linked measurement effect.

In freezing conditions, TDR can respond strongly to the reduction in unfrozen liquid water because ice and liquid water behave differently electromagnetically. A falling inferred liquid-water content during freezing is not the same as the disappearance of all water from the soil. The water may remain as ice while its contribution to the measured permittivity changes. If an article calls every winter decline in TDR moisture evaporation, it has conflated water phase with water loss.

Organic soils and wetland peats also deserve caution. Their composition, bulk density, pore structure and bound-water environment can differ markedly from those of the mineral soils used to establish common calibration equations. Some instruments offer alternative calibration forms, but the presence of a special preset does not prove it fits a specific site. Independent calibration points across the expected moisture range are more convincing than a single agreement check on one unusually wet day.

Make a calibration with matched samples, not imagined precision

A site-specific calibration relates sensor observables, often apparent permittivity, to independently measured volumetric water content. A robust plan collects samples across dry, intermediate and wet states representative of actual use, with known core volumes and carefully measured dry mass. The samples should be closely matched in space and time to the sensing volume. A handful of points all taken immediately after irrigation cannot establish how the instrument performs during prolonged drydown.

Suppose calibration pairs are obtained around theta-v 0.12, 0.25 and 0.38 with corresponding processed permittivities that are monotonically increasing. One can fit a simple suitable function and compare residuals on withheld observations. A lower residual on training data is not sufficient evidence of better future prediction if the polynomial was excessively flexible. Validation should include separate times, probes or conditions when possible, and error statistics should retain the units of water content.

A practical uncertainty statement may report that a calibrated sensor’s water-content error on independent checks was within plus or minus 0.02 cubic metres per cubic metre over the tested range, if that is what real tests showed. This is not equivalent to plus or minus two percent of the reported reading; it means two volumetric percentage points. At a reading of 0.20, an absolute 0.02 spans 0.18 to 0.22, whereas a relative two percent error spans only 0.196 to 0.204. Denominators and error definitions matter.

Read an irrigation pulse without skipping the water balance

A common TDR chart shows a sudden upward step after irrigation followed by a gradual decline. The upward step may represent infiltration into the sensed zone, but its timing depends on depth, preferential flow and sensor placement. The later decline could reflect evapotranspiration, redistribution and drainage. The shape alone does not identify the fraction assigned to each process. Soil hydraulic properties, boundary flows and weather measurements are needed to separate those pathways.

Imagine a profile with sensors at ten, thirty and sixty centimetres. After rain, the upper sensor responds first, followed by deeper sensors. That timing is consistent with a propagating wetting front, but the profile is still a collection of local observations rather than a moving camera of every pore. If the deepest sensor remains unchanged, rainfall may have failed to reach that depth, passed along preferential paths outside the local sensitivity zone or reached it too briefly for the logging interval.

A careful narrative can estimate storage change from depth-weighted profile values and compare it with measured rainfall, irrigation, runoff and drainage. It should not treat one probe’s change as total water used by crops. Even a physically reasonable storage estimate can include uncertainties from spatial heterogeneity, layer boundaries, calibration and missing intervals. State the sampling design before making a whole-field water budget.

Find impossible values before explaining them

If a derived theta-v value is negative, exceeds physically credible porosity, or jumps abruptly without any supporting hydrological event, inspect the data processing first. Problems can arise from an incorrect calibration polynomial, voltage or timing issue, cable damage, poor contact, changing conductivity or inappropriate treatment of missing-data codes. There is no universal rule that every value above 0.5 must be bad; peats and highly porous materials can have much larger water contents than compact mineral soils.

Porosity offers one plausibility check. If soil porosity is independently estimated at 0.42 cubic metres of pore volume per cubic metre of bulk soil, an unfrozen soil cannot contain 0.55 cubic metres of liquid water per cubic metre under that same volume definition. The inconsistency demands an explanation about porosity estimation, swelling, sample representativeness, sensor error or units. It is not a reason to silently cap the measured series at the supposed porosity without documenting the correction.

Another frequent spreadsheet failure is mistaking percentages for fractions. A reported 31 percent volumetric content should be represented as 0.31 when multiplied by a layer thickness, not 31. Multiplying thirty-one by 200 millimetres yields an impossible 6,200 millimetres of water stored in a 200-millimetre layer. The legitimate calculation is 0.31 times 200, or 62 millimetres. A unit check exposes the mistake before any scientific speculation is required.

A complete worked soil-storage calculation

Assume two independently calibrated sensors represent soil layers from zero to twenty centimetres and twenty to fifty centimetres. At the beginning, the respective theta-v readings are 0.18 and 0.24. At the end, they are 0.26 and 0.28. The top layer is 200 millimetres thick; the lower layer is 300 millimetres. For this exercise assume each layer is uniform and the sensor represents it, while recognising that field observations would need to justify those assumptions.

Initial storage is 0.18 times 200 plus 0.24 times 300, giving 36 plus 72, or 108 millimetres. Final storage is 0.26 times 200 plus 0.28 times 300, giving 52 plus 84, or 136 millimetres. The estimated increase is 28 millimetres of equivalent water across the 50-centimetre profile. Notice that the arithmetic uses different layer thicknesses; simply averaging four sensor percentages would not reproduce this storage change.

If 35 millimetres of precipitation fell during the interval, the 28-millimetre increase does not imply exactly seven millimetres evaporated. The difference could also include runoff, deeper drainage, interception, lateral flow and measurement error. A defensible statement is that the estimated profile storage rose by 28 millimetres under the given layer assumptions, compared with 35 millimetres measured as rain. Any separate process attribution requires further measurements or a documented water-balance model.

Transfer problem: reconstruct a waveform measurement

A probe has an effective physical rod length of 0.20 metres, and the properly isolated probe round-trip time is 5.0 nanoseconds. Assume a simple homogeneous waveguide and the speed of light 0.300 metres per nanosecond. Calculate the apparent relative permittivity. First multiply the speed by travel time: 0.300 times 5.0 equals 1.50 metres. Divide by twice the rod length, 0.40 metres, giving 3.75. Squaring produces K approximately 14.06. The quantity is dimensionless and should not be labelled fourteen percent moisture.

A tempting wrong answer is 56.25, obtained by dividing by the one-way rod length before squaring even though the observed time is round trip. That error produces a dramatically wetter-looking permittivity and possibly an implausible water-content estimate. The repair is to sketch the path down and back, confirm whether the displayed travel time contains both legs, and include the factor of two exactly once. The same disciplined drawing prevents cable transit time from being mistaken for soil transit.

Now imagine the logger already reports K equal to 14.06 rather than raw travel time. In that case, the user should not square it again. Apply the relevant calibration to K directly, or check whether the logger’s water-content field has already applied one. The important general lesson is to locate each variable in the observation-to-estimate chain and perform only the transformations not already completed by instrument software.

Transfer problem: calculate a Topp estimate with an uncertainty range

Use the Topp equation theta-v equals minus 0.053 plus 0.0292 K minus 0.00055 K squared plus 0.0000043 K cubed at an illustrative K of ten. The linear term is 0.2920, the quadratic contribution is minus 0.0550 and the cubic contribution is plus 0.0043. After the constant minus 0.053, the estimated volumetric water content is 0.1883 cubic metres per cubic metre, or 18.83 percent by volume. Keep at least a few intermediate digits to avoid a misleadingly neat final figure.

For a sensitivity check, suppose apparent permittivity might reasonably lie between nine and eleven under a stated measurement uncertainty. Substituting nine gives approximately 0.1686 and substituting eleven gives approximately 0.2068. These are endpoints of a simple input-sensitivity calculation for the assumed polynomial, not a formal confidence interval that incorporates calibration error, salinity, probe contact or spatial variability. Their value is to show how an uncertainty in the intermediate signal transfers into the derived water-content estimate.

If a laboratory comparison suggests an additional systematic offset, the numerical interval may need to widen or shift. It would be incorrect to declare the true soil moisture known to four decimal places because the polynomial calculator prints four decimals. Precision of arithmetic is not accuracy of field measurement. An acceptable answer describes the resulting estimate, the calibration used, and which important uncertainties have or have not been considered.

Transfer problem: distinguish rainfall from storage and drainage

A soil profile is estimated to store 90 millimetres before a storm and 112 millimetres afterward. Rainfall measured nearby is 30 millimetres. The storage gain is 22 millimetres; eight millimetres is the difference between the measured rainfall and the estimated increase. Does that prove eight millimetres drained below the profile? No. The difference may be distributed among runoff, intercepted water, evaporation, lateral movement, deeper percolation and estimation error.

A correct worked response begins with the control volume and all its terms. A simplified profile balance is storage change equals precipitation plus irrigation plus upward or lateral inflows minus evaporation, transpiration, runoff, drainage and relevant outward lateral fluxes. If terms are missing, the difference is a residual rather than one named process. Only when the other inputs and outputs have been constrained can the residual be attributed credibly to drainage.

Try a variation: the rainfall gauge is known to undercatch during windy weather, so the true precipitation may exceed the reported 30 millimetres. The apparent residual becomes even more uncertain. The soil sensor and gauge answer different questions and have independent errors. The reader should avoid forcing the soil storage estimate to equal the rain gauge or forcing the gauge to equal the soil storage; compare evidence with its own sampling and calibration limitations.

Diagnose an implausible response after installing a new probe

Imagine that an established probe reads 0.30 while a newly inserted probe at a nearby location reads 0.14, both described as measuring the same depth. Before declaring the field heterogeneous, check whether the rod insertion created air gaps, whether probes have the same geometry and calibration, and whether the second probe crosses a drier horizon. Compare raw permittivity and conductivity rather than only the firmware-generated percentage. The two displayed moisture values may already be based on different equations.

A reasonable diagnostic sequence is to inspect the installation record, cable integrity, temperature, probe orientation and waveform or quality fields. Next compare nearby carefully collected gravimetric samples and repeat during at least one moisture change. If the difference persists with clear contact and independent soil evidence, spatial variability becomes a stronger interpretation. If the value changes after correcting an air gap, the original anomaly was likely installation related.

Do not automatically overwrite early readings with the later preferred value. A transparent dataset can mark the suspect period, record the maintenance and preserve the uncorrected series for traceability. If a bias correction is applied, specify the evidence and dates. Readers need to distinguish a physical wetting event from a change in instrument conditions; a time series can contain both, and the plot alone cannot always reveal which event occurred.

Diagnose conductivity interference without confusing two EC values

A report may include bulk electrical conductivity measured by the probe and electrical conductivity of the soil solution measured from an extraction procedure. Bulk conductivity depends on the conducting paths through water and solid phases, contact geometry and water content, whereas extracted-water conductivity refers to a prepared solution under defined laboratory conditions. Their units may both use decisiemens per metre, but matching units do not make them interchangeable observations.

Suppose a fertilizer event causes the probe’s conductivity channel to rise while moisture content from well-calibrated independent samples remains nearly stable. The right question is whether ionic concentration and waveform attenuation affected the TDR estimate, not whether the sensor has discovered a sudden doubling of water. A comparative diagnosis includes soil solution chemistry, electrical contact, temperature and the manufacturer’s conductivity operating envelope. Such checks help distinguish salinity artefacts from genuine infiltration.

An especially subtle spreadsheet error occurs when dS/m values are compared directly with mS/cm figures without noting that one dS/m equals one mS/cm. The numbers can be equal under that unit conversion, but other units such as mS/m differ by a factor of ten. Write the units beside every observation and conversion. Conductivity describes ionic transport, not a direct count of water molecules; the moisture inference needs the dielectric model and quality checks.

FAQs: what a TDR record can and cannot promise

Does an apparent permittivity of twenty mean twenty percent moisture? No. Permittivity is a dimensionless electromagnetic property; the water-content fraction is inferred through calibration. For typical mineral soils, the Topp polynomial may be a useful first approximation, but the reader must confirm the instrument model and soil applicability. Two datasets can share the same permittivity values yet use different calibration coefficients and therefore report different water-content estimates.

Can one sensor prove the whole field has enough water for crops? No. A sensor represents a limited volume and depth. Irrigation suitability depends on root distribution, soil hydraulic properties, plant demand, rainfall, irrigation scheduling and spatial variation. Volumetric water content alone does not tell us the matric potential or how easily roots can extract the remaining water. A related soil-retention curve is useful for interpreting availability, but its relationship to TDR data is a further model and measurement task.

Does a perfectly smooth moisture curve mean the instrument is highly accurate? Not necessarily. Strong smoothing, long logging intervals or firmware filtering can hide short wetting events and numerical noise while leaving systematic calibration error untouched. Conversely, a noisy response may reflect real spatially localized changes or an unstable contact condition. Judge the sensor by calibration, diagnostics, resolution and independent checks, rather than by visual attractiveness of a chart.

Translate a TDR paper into a defensible public explanation

Begin with the sensor model and exact measured variable: travel time, apparent permittivity, conductivity or already converted water content. State the physical insertion depth and full probe extent. Explain the calibration and how a derived theta-v relates to the mass or volume of water. Describe any temperature, salinity or freezing conditions that could invalidate a simple interpretation. Keep the area and layer denominators visible whenever the narrative turns a fraction into millimetres of profile storage.

A robust explanation also separates local measurement from watershed-wide inference. If the study integrates multiple depths and locations to estimate profile water storage, tell the reader how those values were weighted. If a weather event changed the series, name the observed change first and interpret rainfall, drainage or evapotranspiration only with supporting evidence. When calibration uncertainty was not measured, an honest limitation is better than a made-up percentage.

The final test is whether another person could reproduce the numerical chain from the paper: pulse travel time, relative permittivity, empirical conversion, volumetric fraction and storage depth. Correct units and explicit assumptions make that chain easy to inspect. The purpose of translation is not to remove every technical word, but to give each technical word a stable meaning so the reader can reason beyond the example rather than memorize a solitary moisture percentage.