A buried heat sensor is not the surface flux
A soil heat flux plate measures a transfer of thermal energy through soil at the depth where the plate is installed. An energy-balance diagram, however, commonly asks for the conductive heat flux at the ground surface. These are not automatically equal because the soil between the surface and the buried sensor can absorb or release energy as its temperature changes. To translate a chart properly, identify the depth, the sign convention, the averaging period and whether an above-plate heat-storage correction has been applied. The difference can matter most during rapid morning warming and evening cooling, precisely when a simple reading of the plate can look most convincing. This guide explains the underlying physics, a transparent correction, calibration risks, checked examples, and ways of diagnosing questionable energy-balance results.
Think of the upper soil as a thin thermal reservoir. Sunshine warms its surface. Some energy moves down and crosses a flux plate five centimetres below the surface, while some is still accumulating in the layer above that plate. If you read only the buried plate in that moment, you miss the warming of the upper soil. In the evening, the reservoir may release stored energy toward the atmosphere even while the deeper soil flux has a different direction. The instrument and the desired quantity refer to different boundaries. This is a conservation problem, not merely a correction imposed to make graphs align. An accurate statement distinguishes a heat flux through one horizontal plane from the net rate of energy storage in the layer separating that plane from the surface.
The basic energy accounting, with signs made visible
Adopt the common micrometeorological convention that ground heat flux G is positive downward, away from the surface and into the soil. Let Gp be the calibrated ground heat flux through the plate at its installation depth. Let S be the rate of thermal energy storage in the soil above the plate, positive when that layer gains heat. Then the surface ground heat flux is Gsurface = Gp + S. Both terms must have units of watts per square metre. This sign convention is consistent with the Earth Observing Laboratory’s explanation of soil heat flux in the surface energy account. Other instruments or textbooks may reverse the signs, so always inspect their documentation rather than carrying a memorised plus sign across incompatible definitions.
Imagine a plate at five centimetres reading plus 25 W/m² while the surface layer is gaining heat at 35 W/m². The estimated surface heat flux is plus 60 W/m². If the plate reads minus 10 W/m² and the upper layer is cooling at minus 20 W/m², the surface flux is minus 30 W/m²: heat is moving upward from the soil in the adopted convention. The physically useful quantity is not just 60 or minus 30. It is the direction of energy movement, the area normalisation, and the period to which the estimates refer. Heat is measured in joules, while a heat flux density in W/m² expresses joules per second per square metre. Confusing energy with power is a frequent source of incorrect translations.
Heat storage is a temperature-change rate
For a uniform upper layer of thickness z, volumetric heat capacity Cv and representative temperature change ΔT over time Δt, the storage term is approximately S = Cv × z × ΔT/Δt. Cv has units J/(m³ K), z is in metres, the temperature change is in kelvin and Δt in seconds. Multiplying gives J/(m² s), exactly W/m². For temperature differences, a change of one Celsius degree is numerically the same as a change of one kelvin. This expression assumes a sufficiently representative layer-average temperature and heat capacity. If the layer is thermally heterogeneous, a layered sum is more defensible than one average value.
Work a complete example. A five-centimetre soil layer has an illustrative volumetric heat capacity of 2.0 million J/(m³ K). Its representative temperature rises by 0.9 K over 30 minutes, or 1,800 seconds. The storage rate is 2,000,000 × 0.05 × 0.9 / 1,800 = 50 W/m². If the plate beneath records plus 18 W/m² after appropriate calibration, the inferred surface flux is 68 W/m² downward. We have not observed a real location; these are teaching values. Notice that omitting S would understate the rate at which the surface supplied energy to the soil by 50 W/m². The estimate is only as reliable as the representative temperature and the heat capacity used.
Why soil moisture changes the storage correction
Dry mineral soil, wet mineral soil, organic matter, air spaces, ice and roots differ in volumetric heat capacity and thermal conductivity. Water holds substantial thermal energy per unit volume. When a soil wets after rainfall, the same temperature change can correspond to a different storage rate than before the storm. A thermal model that treats heat capacity as fixed may gradually become biased across a drying cycle. The heat flux plate also perturbs the thermal pathway because its properties may not match the surrounding soil. Moisture can change both how much energy the layer stores and how effectively it transmits energy toward the buried sensor.
A useful decomposition for the sensible heat capacity of a porous soil treats its volumetric heat capacity as the contributions from soil solids, liquid water and other constituents. One simplified approximation uses Cv ≈ (1−porosity)Csolid + θwater Cwater + θair Cair, provided consistent volume fractions and temperature-dependent properties. For a typical field calculation, the air contribution is comparatively small, but it is not a licence to ignore ice or phase-change energy in a freezing soil. If volumetric water content is measured by a nearby TDR probe, the analyst can update the heat-capacity estimate rather than assuming yesterday’s water fraction remains valid. The moisture sensor itself must be calibrated and positioned to represent the storage layer.
Depth and temperature sensors define the accuracy ceiling
A plate at ten centimetres has a thicker soil layer above it than a plate at five centimetres, so the storage correction can be larger. If temperatures are sampled only at the buried plate’s depth, they may not adequately capture the temperature of the overlying soil. Better deployments measure temperature at one or several depths inside that layer and calculate a weighted storage term. Suppose the upper five centimetres are divided into two layers with different volumetric heat capacities and different warming rates. Their contributions should be calculated separately as Cv1 z1 ΔT1/Δt and Cv2 z2 ΔT2/Δt, then added. A single surface-temperature probe cannot replace that entire vertical profile without additional assumptions.
Time alignment matters just as much as spatial alignment. A heat flux plate may report five-minute averages, whereas soil temperatures may be recorded at the end of each hour. Calculating the hourly temperature difference and adding it to the instantaneous plate reading can mix different temporal supports. Choose consistent time windows and document whether the value is an average over the preceding interval or a point sample. The change in stored heat is best thought of as a finite difference between two representative states of the same layer. This is more defensible than using the warmest and coolest points in an interval simply because they produce a visually dramatic correction.
The plate calibration is not a universal soil property
A heat flux plate uses a measured temperature gradient or thermopile response across an element with known properties to infer heat passing through it. Calibration typically establishes a relationship between sensor output and flux under specified conditions. The plate may conduct heat differently from the surrounding medium, producing a disturbance in the local flux field. Published correction approaches account for plate geometry and the thermal conductivity of the medium. This is why a catalogue calibration factor is a necessary starting point rather than an absolute guarantee that the field value exactly equals undisturbed soil flux.
The Earth Observing Laboratory describes an application in which plate measurements are adjusted for thermal conductivity differences between calibration medium, sensor and surrounding soil, and then combined with storage above the plate. A reader should not copy its particular coefficients to a different plate without checking geometry and calibration history. Sensor units may be microvolts per W/m², millivolts per W/m², or a logger-produced flux value already transformed using a supplied coefficient. Multiplying twice by the calibration factor or applying a conductivity correction twice can produce plausible-looking but wrong results. Preserve the raw electrical output and each processing step so the result can be reconstructed.
Comparing a heat flux plate with temperature gradients
Fourier’s law expresses conductive heat flux as proportional to the negative temperature gradient through a medium. In a vertical coordinate with depth increasing downward and downward-positive flux, a carefully defined form is G = -k dT/dz when z increases upward, or G = k dT/dd when d increases downward, depending on coordinate and chosen signs. The equation must be interpreted with the actual coordinate convention, not copied without defining z. If temperature increases with depth, heat tends to conduct upward. If the surface is hotter than the deeper soil, heat tends to conduct downward. Thermal conductivity k has units W/(m K) and must represent the soil conditions in question.
Suppose the near-surface soil is at 27°C and a level 0.05 metre deeper is at 25°C. If we assume a steady linear gradient and conductivity k = 0.8 W/(m K), then downward heat flux is approximately 0.8 × (27−25)/0.05 = 32 W/m². The gradient estimate and the plate value need not agree exactly because the real profile may be curved, the soil properties variable, and the temperature observations not representative of the same layer and time. This comparison can reveal a problem, but it does not identify which measurement is wrong without further evidence. Transient soil warming also means that a single steady-gradient estimate may not capture storage.
Reading the morning transition without confusing energy stores
At dawn, net radiation often becomes positive, but the soil can still be losing energy upward or the buried plate can lag behind surface warming. As sunlight strengthens, the surface layer may begin warming quickly while deeper soil remains cool. The storage term can be strongly positive even when the plate signal is small. A graph that shows Gp near zero at sunrise does not mean the ground receives no thermal energy. It may instead mean that the heat has not yet crossed the buried sensor plane. A clear explanation refers to the changing temperature of the layer above the plate and the propagation of the thermal wave with depth.
If a researcher uses Gp alone in a surface energy balance, the apparent allocation of net radiation to sensible and latent heat can be distorted. The soil’s contribution may look too small in the morning and too large or wrongly signed in the evening. This distortion can affect estimates of evapotranspiration calculated as an energy-balance residual. It is important not to “fix” the remaining energy imbalance by arbitrarily forcing latent heat to absorb every missing watt. Flux towers, soil plates, radiation sensors and storage estimates each have their own footprint and uncertainties. The soil correction is one essential term, not a universal closure algorithm.
Nighttime cooling and flux reversal
After sunset, the upper soil usually cools, but its deeper layers may remain warm. In the downward-positive convention the surface ground heat flux can become negative, representing upward energy movement toward the cooler surface. The plate depth may respond later than the surface layer. If the heat-storage correction is ignored, the sign reversal could appear delayed or muted. The temperature series often gives useful physical context: a declining upper-layer temperature indicates negative storage even if a deeply buried plate is still showing a small positive flux for a short time.
A negative sign is not an error code. It is a direction label determined by the coordinate convention. When a graph plots ground heat flux with the opposite sign, the same night can appear as a positive value. Before comparing two datasets, re-express both in a common convention and check whether the plotted variable is Gp or Gsurface. A statement such as “ground heat flux increased at night” is ambiguous if the author does not define direction. Better wording says “the corrected ground flux became 25 W/m² upward from the soil,” or “Gsurface = −25 W/m² under our downward-positive convention.”
Soil-layer stratification and weighted storage
Consider a surface layer containing two distinct zones. The first two centimetres are dry and warm rapidly; the next three centimetres are wetter and change temperature more slowly. Applying one average heat capacity and one thermistor’s temperature trend may systematically understate or overstate storage. The layered method assigns each zone its own Cv, thickness and change in temperature and sums contributions. One can refine the vertical profile into several sublayers, but the additional detail is useful only if sensor placement and soil-property estimates support it. A model that divides five centimetres into fifty layers without extra measurements can create an illusion of precision.
For a two-layer numerical example, let upper-layer thickness be 0.02 m, Cv1 = 1.5 million J/(m³ K), and ΔT1 = 1.2 K over one hour. Its contribution is 10 W/m². Let lower thickness be 0.03 m, Cv2 = 2.4 million, and ΔT2 = 0.5 K over one hour. Its contribution is another 10 W/m². Total storage is 20 W/m². A plate reading 14 W/m² therefore gives Gsurface = 34 W/m². The equal ten-watt contributions arise for different combinations of thickness, heat capacity and temperature change. It is the product of all three, divided by time, that matters.
A precise distinction between watts and cumulative energy
The corrected Gsurface is a flux density expressed in watts per square metre. To find energy transferred per unit area over a period, integrate G through time. For a constant 40 W/m² over one hour, the energy is 40 × 3,600 = 144,000 J/m², or 0.144 MJ/m². Calling 40 W/m² “40 joules” is dimensionally wrong. An hourly average of 40 W/m² and an hourly cumulative amount of 0.144 MJ/m² express the same integrated transfer in different units. When converting an energy flux to an equivalent evaporation depth, additional information on latent heat is required; ground conductive flux itself should not automatically be called evaporation.
For a day with alternating positive and negative values, signed integration matters. Suppose Gsurface equals +50 W/m² for eight hours and −20 W/m² for sixteen hours. The signed daily energy is 50×8×3,600 minus 20×16×3,600 = 288,000 J/m², or 0.288 MJ/m² downward net. Taking absolute values would instead total a much larger gross movement and remove the physical cancellation. A positive daily net does not mean flux remained downward all day. Distinguish mean flux density, signed cumulative energy and gross magnitude when translating charts into plain language.
Soil heat flux in the full surface-energy balance
A simplified energy balance often writes Rn = H + LE + G, using net radiation positive toward the surface and turbulent sensible heat H and latent heat LE positive away from it, with downward-positive G. The expression is an accounting framework under chosen conventions and may require additional storage or advective terms for particular surfaces. If Rn is 400 W/m², H is 180 W/m² and corrected G is 60 W/m², a residual estimate gives LE = 160 W/m². If someone instead uses an uncorrected plate value of 20 W/m², the residual LE becomes 200 W/m², artificially larger by 40 W/m². Thus the soil correction affects a different reported flux, even though the sensor is buried in the ground.
This is not evidence that the latent-heat method is automatically wrong; it explains sensitivity to an input. The same energy balance may depart from closure because of footprint mismatch, sensor tilts, frequency losses, canopy storage, horizontal transport or turbulent flux uncertainties. A publication should distinguish a residual estimate from independently measured LE. It should also avoid claiming that closing the budget by changing one term proves that term’s correction is right. Compare methods, examine uncertainties, and identify the variables actually measured versus those inferred. In a teaching context, the energy balance is a powerful way to show why a buried thermal sensor affects water-flux interpretation.
Different surfaces demand different caution
Bare soil, closely cropped grass, deep litter, forest floor and irrigated agricultural plots differ in radiative shading, thermal diffusivity, water content and heat storage. In a forest, a plate installed beneath litter may not capture storage in the litter and canopy, even after correcting the mineral soil above the plate. In dry crusted soils, contact between the plate and surrounding medium can be poor. In wet soils, a change in water content shifts both conductivity and storage properties. The translation should describe the actual measurement environment. The plate is a local observation, not an automatic spatial average of a heterogeneous landscape.
Large surface stones and irregular roots make placement difficult. Installing a flat plate in a hole can disturb natural soil structure, and backfilling may produce artificial contact conditions for some time. A plate installed near an irrigation emitter can sample a wetter thermal environment than the bulk field. Comparing two neighbouring plates often helps expose spatial variation, but identical readings do not prove correct calibration if both share the same systematic error. Record depth, orientation, sensor model, soil texture, vegetation cover, water content and installation date. These descriptions belong in an analysis intended to support replication, not just in a laboratory notebook.
Freeze-thaw, phase changes and latent energy
The simple storage term Cv z ΔT/Δt describes sensible heat storage, but frozen or freezing soil introduces additional processes. Water changing phase can absorb or release latent heat while temperature remains near the transition. A small measured temperature change does not prove that energy storage is negligible during thaw. Liquid water and ice have different thermal properties, and moving water can also transport heat. Applying a constant Cv through freeze-thaw conditions without accounting for phase change may miss important energy terms. A sensible report notes this limitation rather than presenting a seemingly exact surface flux based on a formula outside its assumptions.
Similarly, infiltrating rainwater can carry sensible heat into the soil, and soil water evaporation may affect temperature and energy transport. Under ordinary dry, well-characterised conditions, the standard plate-plus-storage estimate can be useful. Under strong advective or phase-change conditions, more complete energy accounting may be needed. Readers should learn to recognise when the mechanism being measured has changed enough to challenge the method. An advanced model is not necessarily better if crucial input data are absent. Sometimes the most honest conclusion is that the standard surface flux correction is uncertain during a particular event.
Uncertainty from each part of the equation
Because Gsurface = Gp + S, its uncertainty depends on the uncertainty of both terms and their covariance. If their errors are independent, their standard uncertainties may be combined by the square root of the sum of squares. A plate estimate of 20 W/m² with uncertainty 5 and storage of 35 W/m² with uncertainty 10 gives Gsurface = 55 W/m² with combined standard uncertainty about sqrt(25+100) = 11.2 W/m². This simplification ignores systematic calibration error and covariance. A result rounded to 55 ± 11 W/m² is more defensible than a visually polished 55.0000 if the assumed uncertainties are plausible.
The storage uncertainty can be dominated by thermal capacity, vertical representativeness or the difference between two noisy temperature observations. Consider the relation S = Cv z ΔT/Δt: an uncertain layer thickness, sensor depth or temperature trend changes the outcome. The errors do not all behave identically. A fixed bias in Cv affects every positive and negative storage term proportionally, while random thermometer noise may become especially damaging at very short intervals. Longer averaging windows can reduce some noise but obscure transitions; the selection of time resolution should be justified by the signal and the intended scientific question. Present uncertainty in W/m² alongside an explanation of its origin.
A reproducible workflow from raw voltage to publication
Start by logging plate voltage or instrument output, calibrated response coefficient, installation depth and sign orientation. Verify whether the reported output has already been converted to W/m² and whether conductivity corrections are embedded. Record soil temperatures at depths sufficient to represent the layer above the plate and record moisture or heat-capacity information. Synchronise observations to common intervals. Calculate plate flux, then temperature-change-based storage, then their signed sum. Preserve separate columns for Gp, S and Gsurface rather than overwriting Gp. Where values are invalid because of sensor failure, frozen ground or disturbed contact, annotate the decision.
Before publishing, inspect a full diurnal cycle and the changing relation between surface heating and deeper soil response. Compare temperature gradients and check whether negative nighttime signs make physical sense. Revisit the radiation and turbulent-flux series on a common time base. Report the period, sensor location, soil conditions, area normalisation and uncertainty. A graph legend should say “soil heat flux at 5 cm,” “storage above plate” or “corrected surface G” rather than one ambiguous label “soil heat.” The point of translation is to make the distinction evident without hiding it inside a software script.
A worked morning and evening sequence
At 08:00, the plate reads 12 W/m² downward while upper-layer storage is 38 W/m². Surface G is 50 W/m². At noon, the plate reads 45 W/m² and storage is 25 W/m², so G is 70 W/m². At 18:00, the buried plate still reads 10 W/m² downward but the upper layer is cooling at S = −30 W/m²; corrected surface G is −20 W/m² upward. These are illustrative readings on a consistent downward-positive scale. The three samples show why one cannot assume that the sign of a buried measurement always reveals the instantaneous direction of heat transfer at the ground surface.
If the plate signals were plotted without storage correction, the evening point would appear to show heat still moving downward. A careless reader might claim the surface remained a heat sink, whereas the corrected surface estimate indicates heat emerging from the upper soil layer. The apparent contradiction disappears once the boundaries are separated. It is important not to infer what happened between those three samples without a continuous record. A line connecting widely separated data points may be an illustrative interpolation rather than an observed history. Explain that distinction when the graph is used for teaching.
Diagnostic practice cases for diverse field conditions
Case 1 — a five-centimetre plate. A sensor reads 24 W/m² downward and storage above it is +16 W/m². Surface G is +40 W/m². Explain the extra 16 as energy accumulating in the upper layer rather than a mysterious calibration offset. If the data are to be compared with net radiation, ensure the intervals refer to the same averaging period and that the sign convention of the radiation instrument is understood. The answer tests the conservation equation, not a method for estimating latent heat directly.
Case 2 — evening release. Gp is +8 W/m² and S is −28 W/m². Then Gsurface is −20 W/m². A beginner may announce that the plate and the calculated surface flux contradict one another. They need not: the layer above the plate is releasing stored heat faster than energy passes down at plate depth. The resulting upward surface flux is physically possible. Review the layer-temperature trend to establish whether the negative storage estimate is independently supported.
Case 3 — a one-hour temperature change. A four-centimetre upper layer has Cv = 2.5 million J/(m³ K) and warms by 0.36 K during 3,600 seconds. Storage is 2,500,000 × 0.04 × 0.36 / 3,600 = 10 W/m². If the plate flux is 18 W/m², surface G is 28 W/m². Converting the four centimetres to 4 metres would inflate the storage term one hundredfold. Keep the dimensional calculation visible: metres, joules per cubic metre, kelvin, and seconds.
Case 4 — different interval lengths. A temperature rise of 0.50 K in 1,800 seconds does not yield the same storage rate as the same rise over 3,600 seconds. For fixed layer thickness and heat capacity, the first rate is twice the second. The result is a power flux, meaning energy per unit time, not simply a stored energy amount. A chart comparing thirty-minute and sixty-minute calculations should normalise both correctly before discussing whether the soil became more or less conductive.
Case 5 — a multilayer profile. Two sublayers produce storage rates of +12 and +7 W/m². The total upper-layer S is +19 W/m², even if their temperature gradients have different shapes. Add the signed contributions rather than averaging them without weighting. If a plate reports +21 W/m², Gsurface is +40 W/m². What matters is each layer’s thickness and heat capacity as well as temperature trend. The sensor count alone does not tell the reader how much heat was stored.
Case 6 — a wetting event. Rain raises volumetric water content and likely changes bulk heat capacity. An analyst continues using a dry-soil value because the spreadsheet stores one constant for the whole season. The resulting storage correction may be biased even if temperature measurements are accurate. Check moisture measurements, estimate a suitable wet-soil heat capacity, and make the change in method transparent. If heat carried by infiltrating water is important, the simplified conductive-storage framework may need expansion.
Case 7 — a thermocouple at the wrong depth. A logger treats a probe installed at nine centimetres as if it were centred in the upper five centimetres. The inferred temperature trend may not represent that layer, particularly during rapid warming. Recalculation should use the actual geometry and available temperature profile. Do not invent a shallow temperature from a deeper one without stating a model and its uncertainty. Physical installation records are part of the evidence, not a bureaucratic appendix.
Case 8 — a reversed sensor lead. A plate expected to read positive for downward flux shows negative at midday, despite a hot surface and cooler deeper soil. This could be a real local condition, but it can also indicate reversed wiring or a logger sign setting. Check manufacturer polarity, raw signal and nearby temperature gradients before simply multiplying by minus one. Record the correction if verified. A persuasive scientific result demonstrates why the convention was changed.
Case 9 — a calibration factor used twice. A logger already converts microvolts to W/m² using the manufacturer’s plate coefficient, yet a downstream spreadsheet multiplies by the same conversion again. The resulting values may be absurdly large or deceptively plausible depending on units. Trace the original raw signal to the data logger program and the exported file. A unit audit is often more useful than advanced filtering. The correct final value should have a clear provenance from sensor output to energy flux.
Case 10 — plate contact failure. A buried plate develops a gap against surrounding soil after shrinkage. Thermal resistance changes and the response no longer represents the undisturbed medium in the intended way. A smooth trace cannot rule out contact bias. Compare nearby temperature gradients, soil moisture and maintenance records; avoid claiming exact recovery from a purely mathematical adjustment without physical evidence. If contact is doubtful, qualify or withhold the affected data.
Case 11 — frozen ground. A soil temperature remains near freezing while energy enters during thaw. Estimating storage from sensible heat capacity times a small temperature change may severely understate total energy uptake because latent heat of fusion matters. Identify whether ice and liquid water coexist, and whether the simple method remains applicable. The case illustrates a general limitation: when the physical state of the medium changes, ordinary temperature-based storage formulae may omit a major component.
Case 12 — moving precipitation. Warm rain infiltrates cool soil, carrying enthalpy with flowing water. A plate and local temperature sensors may not capture this as a simple one-dimensional conductive transfer plus static storage. The correct analysis depends on whether advective heat transport is significant for the question. Do not assert that the standard correction is exact during intense infiltration. Flag the interval and describe additional evidence needed, including precipitation temperature, infiltration depth and water-flow timing.
Case 13 — a forest floor. Soil plates beneath litter describe heat entering the mineral soil, while energy may be stored in the organic litter and canopy above. Correcting only mineral-soil storage to the ground surface can leave missing terms if the defined surface is above that litter layer. Define the energy-balance boundary before comparing instruments. A forest-wide estimate also involves spatial variation and above-ground biomass storage, not simply the buried plate beneath one tree.
Case 14 — the energy residual. A surface has Rn = 300 W/m², H = 120 W/m² and Gp = 15 W/m². Storage correction is +25 W/m². Then Gsurface is 40 W/m² and the simplified residual LE equals 140 W/m². Without S, a calculation would report LE = 165 W/m², a 25 W/m² overestimate. This is sensitivity arithmetic under a closed simplified balance; it does not prove that real measurements have perfect energy closure.
Case 15 — positive and negative sums. For one hour the corrected surface G is +30 W/m²; for the next hour it is −10 W/m². The signed energy total is (30−10) × 3,600 = 72,000 J/m², or 0.072 MJ/m². Reporting an average of +20 W/m² is wrong because the equal-duration mean is +10 W/m². Likewise, integrating absolute magnitudes would give a larger gross movement rather than the net transfer. Always specify whether the quantity is rate, mean rate or accumulated signed energy.
Case 16 — an apparent zero. Gp = +25 W/m² and S = −25 W/m² yield surface G = 0 W/m². This does not mean no thermal activity occurs inside the soil. Heat is crossing the buried sensor while the layer above releases an equal amount of stored energy. The surface balance happens to sum to zero over that period. A reader who knows the individual terms sees a richer physical story than one who sees only the final zero.
Case 17 — an area trap. One plate reports a heat flux density of 40 W/m² over its sensing location. Multiplying by a 100-square-metre plot area gives 4,000 W only if the plate’s local flux is representative of the whole plot, an additional assumption. The plate does not itself measure the integrated plot power. The word “per” in W/m² matters. Site heterogeneity and footprint design must be discussed before upscaling one sensor to a landscape total.
Case 18 — a thermal conductivity mismatch. A plate and surrounding soil conduct differently; a calibration done in one medium may require a geometry-and-conductivity correction for use in another. The appropriate factor depends on instrument design and actual soil thermal conditions. Copying a correction from a published research site without checking materials can create a new systematic error. Use manufacturer and method-specific guidance rather than a generic universal percentage.
Case 19 — cloudy midday. The net-radiation input falls rapidly under passing cloud while the soil continues storing or releasing heat at a lagged rate. Do not expect plate flux to track every cloud edge instantaneously. Thermal diffusion takes time, and the depth of installation filters fast variations. A delayed response is not necessarily a broken sensor; compare with temperature changes and the corrected surface estimate. Where time scales differ, use consistent averaging and identify the phase lag.
Case 20 — uncertainty propagation. Gp = 22 ± 4 W/m² and S = 18 ± 6 W/m², with independent standard uncertainties. Corrected surface G is 40 W/m² and combined standard uncertainty is sqrt(4²+6²), about 7.2 W/m². State that the independence assumption may fail and that systematic error is not fully captured. Reporting 40 ± 7 W/m² expresses the calculation honestly; claiming 40.000 W/m² from a spreadsheet display does not.
Case 21 — an unrepresentative thermometer. The upper layer has sharp temperature gradients, but a single probe sits next to a sunlit stone. Its temperature change may be greater than the spatial mean. Multiplying that change by the entire layer’s heat capacity overestimates storage. Improve sensor layout or quantify the representativeness error. More decimal digits from the same probe are not a substitute for appropriate sampling of the physical medium.
Case 22 — soil moisture mismatch. A volumetric water content is reported in percent, and a model expects a fraction. Entering 25 instead of 0.25 can generate impossible heat-capacity estimates. Check dimensionless units before calculating S. Likewise, gravimetric water content is not equal to volumetric water content without conversion using soil bulk density and water density. A correct physical formula fed the wrong type of water-content input will produce a wrong result.
Case 23 — quality of the nightly sign. Two different instruments show opposite nighttime G because one exports upward-positive values and the other downward-positive values. Compare metadata before proposing a scientific contradiction. Reorient one series into a common convention, then compare magnitude, depth and storage treatment. In a good report the sign convention is stated next to the graph or equation. Every reader should be able to tell which direction of heat transfer the plus sign means.
Case 24 — missing hours. A soil temperature logger fails during the morning warming period, so the storage term cannot be computed directly for those intervals. Do not silently set S = 0 and treat the resulting Gp as surface heat flux. The uncertainty is not zero merely because data are absent. A modelled temperature profile might offer a labelled estimate, or the affected periods may be withheld from a measured surface flux product. Distinguish observed, corrected and reconstructed data.
Independent transfer problems with answers
Problem A. A plate at 0.05 m reads +30 W/m². The soil above it has Cv = 2.4 × 10^6 J/(m³ K) and warms by 0.60 K in an hour. S = 2,400,000 × 0.05 × 0.60 / 3,600 = 20 W/m². Therefore Gsurface = +50 W/m² downward. The calculation depends on the layer temperature representing the full thickness. A learner should label the 50 as corrected surface ground heat flux, not the direct plate observation. Check units before substituting: the centimetres must become metres and the hour must become seconds.
Problem B. A plate reports −8 W/m² and a five-centimetre layer with Cv = 1.8 × 10^6 J/(m³ K) cools by 0.80 K over one hour. The storage term is 1,800,000 × 0.05 × (−0.80) / 3,600 = −20 W/m². Hence Gsurface = −28 W/m². Under downward-positive signs, the surface heat flux is 28 W/m² upward. If an answer reports plus 28 without explaining a reversed convention, the reader cannot know whether the physics or the sign has changed.
Problem C. Net radiation is +380 W/m², measured sensible heat is +170 W/m² and corrected ground heat flux is +50 W/m². If the simplified energy balance applies, residual latent heat flux is 380 − 170 − 50 = 160 W/m². Now suppose the ground flux had been calculated as +20 from the buried plate alone. The incorrect residual would be 190 W/m², 30 too large. This exercise reveals why a subsurface correction can bias a seemingly unrelated water-exchange estimate. Real energy balances may include additional storage or advection terms.
Problem D. A surface ground heat flux averages +25 W/m² for six hours. The signed energy per area is 25 × 6 × 3,600 = 540,000 J/m² = 0.540 MJ/m². If instead the flux were reported as +25 MJ/m², the unit and magnitude would be fundamentally different. The quantity W/m² is a rate; MJ/m² integrates that rate across time. A scientifically literate summary keeps the measurement unit and the time integration visible, even when simplifying the language.
Problem E. Two plates installed within a plot show +18 and +42 W/m² after comparable corrections, while their soil textures and moisture differ. Their arithmetic average is +30 W/m², but that is only a representative plot estimate if sampling and weighting justify treating the two locations equally. Ask what fraction of the plot each station represents and whether one sits near an irrigation emitter. A mean without a spatial design can be numerically correct yet environmentally misleading. This problem has no unique area-average answer from the supplied information alone.
Questions worth answering before trusting a chart
Can the plate alone measure surface G? Not ordinarily, because the upper layer stores and releases heat; whether a storage correction is negligible must be evaluated, not presumed. Why are plate values and surface values different? They refer to different boundaries separated by soil that can gain or lose thermal energy. Is negative flux necessarily wrong? No, a defined sign convention can represent heat moving upward. Can every plate use the same calibration correction? No. Geometry, material, installation and conductivity conditions matter. Does a temperature probe measure heat capacity? No. It measures temperature; Cv needs independent knowledge or estimation. What changes during thaw? Latent heat of phase change can matter beyond sensible temperature storage.
How does this connect to evapotranspiration? The surface energy budget allocates net radiation among sensible heat, latent heat, soil heat and sometimes other terms. An incorrect G can bias a residual LE estimate. What is the most valuable data table? One that stores raw plate output, calibrated Gp, layer temperatures, heat capacity, storage S, final Gsurface, sign convention, interval and quality flag. Can the method cover any surface? Not without considering conditions, especially forest litter, wetting events, freeze-thaw, advection and spatial variability. The best translation says both what a corrected flux tells us and what remains uncertain.
How to read neighbouring eduKateSG material
The four-component net-radiation article explains the radiative energy input that often drives soil warming. The Bowen-ratio article examines how sensible and latent heat are distinguished. The TDR soil-moisture guide helps interpret a property that influences bulk soil heat capacity and thermal behaviour. These are related measurements, but their physical meanings and units must not be swapped. When translating an instrument chart into English, preserve the boundary, sign, unit, processing step and uncertainty.
Source notes
NSF NCAR Earth Observing Laboratory: Calculation of Soil Heat Flux at the Surface explains the plate-plus-storage method and conductivity corrections. FAO Irrigation and Drainage Paper 56, Chapter 2 defines reference evapotranspiration and energy terms. Google Search Central people-first guidance informs the editorial approach but does not prescribe length. All numerical case measurements in this guide are explicitly invented examples used to teach the calculation, not measurements presented as field observations.