A cosmic-ray neutron soil-moisture sensor can sit above a field and infer water content over a landscape much larger than the tiny volume sampled by an ordinary soil probe. Its number may look like a direct moisture measurement, but the instrument actually counts secondary neutrons whose abundance is influenced by hydrogen and environmental conditions. Correct translation requires following the chain from neutron counts through atmospheric corrections, local calibration, dynamic sensing depth and other hydrogen pools.
This guide explains cosmic-ray neutron sensing, usually abbreviated CRNS, from the underlying physics to field validation. We separate raw counts from corrected count rates and calibrated volumetric water content, show checked examples of counting uncertainty and weighted ground sampling, and examine pressure, incoming cosmic-ray intensity, humidity, biomass, snow, roads, water bodies and footprint variation. We then test the method through independent problems rather than presenting a one-size-fits-all formula.
The headline distinction is that a single CRNS value is not the moisture of every soil parcel within a fixed 200-metre circle. Its effective footprint and depth change with environmental hydrogen, detector setup and surrounding landscape. A reliable interpretation tells the reader what was counted, what was corrected, how soil calibration was obtained and which land-cover or water-storage features may influence the estimate.
Where the detectable neutrons come from
High-energy cosmic radiation interacts with atoms in the atmosphere and initiates cascades of secondary particles. Some secondary neutrons reach the land surface and interact with soil, air, vegetation and water. A cosmic-ray neutron sensor detects neutrons in instrument-specific energy ranges, often emphasising epithermal or related fast neutron components for environmental water estimation. It uses naturally produced radiation rather than placing a conventional radioactive source in a soil access tube.
Neutrons transfer energy in collisions with atomic nuclei. Hydrogen is particularly effective at moderating neutrons because its nuclear mass is comparable to that of a neutron. Water contains abundant hydrogen, so increasing environmental water tends to reduce the detected population of particular above-ground epithermal neutrons. The relation is inverse: after appropriate corrections, a lower detected intensity often corresponds to more hydrogen and therefore, in a calibrated soil setting, greater water content.
The sensor cannot identify whether every moderating hydrogen atom came from liquid soil water. Water in plants, snow, ponds, litter, organic matter and even atmospheric vapour can affect the signal. The conversion from counts to soil moisture is therefore an inverse interpretation conditional on other hydrogen sources and their treatment, not a direct laboratory chemical analysis of soil water alone.
Do not confuse CRNS with a borehole neutron log
A conventional borehole neutron log typically involves a deployed source or specialised neutron instrument and interprets neutron responses near a borehole, often in terms of hydrogen index and formation properties. A cosmic-ray neutron station passively counts naturally generated secondary neutrons across a broad environmental footprint. The sources of neutrons, observation geometries, calibration procedures and spatial support are fundamentally different.
Two plots may both show neutron counts decreasing as hydrogen rises, but that common mechanism does not make their formulas interchangeable. Borehole measurements are influenced by lithology, borehole conditions and tool response. An aboveground CRNS station also responds to atmospheric pressure, incoming cosmic-ray flux, humidity, vegetation biomass and heterogeneous soil water across a landscape.
The already published eduKate guide on borehole neutron logs and water-filled porosity treats the downhole context separately: https://edukatesg.com/2026/10/27/how-to-translate-easily-borehole-neutron-hydrogen-index-water-filled-porosity/. The responsible reader first identifies which instrument produced the data before borrowing an interpretation from another neutron-based method.
What is actually counted
A CRNS data file may contain raw counts accumulated during a time interval, a count rate converted to counts per hour, diagnostic channels or a count series corrected for changing environmental influences. These are different variables. If the instrument records 1800 counts in an hour, its mean rate is 1800 counts per hour. If it records 900 counts in half an hour, the corresponding mean rate is also 1800 per hour, but the counting uncertainty differs because the accumulated total differs.
Raw count rates can change even when soil water does not. Atmospheric pressure modifies the atmospheric shielding of cosmic rays. Solar and geomagnetic effects influence incoming cosmic radiation. Atmospheric vapour contributes hydrogen above the soil. Changes in vegetation biomass or snow can further moderate neutrons. A reader should therefore avoid interpreting every downward movement in raw counts as rainfall infiltration.
The first data-audit step is to inspect variable names, integration time, energy channel, detector status and correction level. Counts are dimensionless events, while a rate adds a time denominator. A calibrated soil-water estimate often has units of cubic metres of water per cubic metre of bulk soil. Silently treating all three as the same sort of number makes otherwise careful plots misleading.
Why counting statistics matter
Neutron detections occur as discrete events. Under a simplified independent Poisson counting model, the standard deviation of the total count N is approximately the square root of N. The relative standard uncertainty is therefore about 1/sqrt(N). For N = 1800 accumulated counts, sqrt(1800) is approximately 42.4 counts, giving a relative counting uncertainty of 42.4/1800, or about 2.36 percent.
If a site accumulates four times as many independent counts under otherwise stable conditions, N becomes 7200, square root approximately 84.9 and relative counting uncertainty about 1.18 percent. The relative noise improves roughly as the inverse square root of counting time when rate and assumptions remain stable. Longer averaging is not free, however: it can blur short rainfall pulses and changing environmental states.
The Poisson calculation describes an ideal statistical component of the raw detections. Real corrected soil-moisture products inherit additional uncertainty from calibration, pressure, humidity, incoming-flux adjustment, background changes and spatial sampling. The standard uncertainty of a count rate is not automatically the standard uncertainty of volumetric water content. The nonlinear conversion can amplify or attenuate changes depending on moisture and calibration.
Worked count-rate example
Suppose two stations both report an average 1800 counts per hour. Station A counted 1800 events in a single hour. Station B accumulated 7200 events over four hours and divided by four to report the hourly rate. If environmental conditions were stable and individual counts were approximately independent, Station A’s standard count-rate uncertainty is sqrt(1800)/1 hour, approximately 42.4 counts per hour.
For Station B the total count standard deviation is sqrt(7200), about 84.9 events. Dividing by four hours gives a rate uncertainty of about 21.2 counts per hour. Its relative uncertainty is about 1.18 percent, half the approximately 2.36 percent of Station A. The two station values have equal nominal rates but different statistical precision because their integration times differ.
The example does not mean the four-hour average is always the better environmental measurement. During a strong rainfall event, soil moisture might change substantially over those four hours, so the average can conceal timing. A reader should select temporal averaging to match the scientific question and report the trade-off between counting precision and environmental time resolution.
Pressure corrections before soil moisture
Atmospheric pressure alters the column of air through which cosmic-ray secondaries travel and affects detected neutron intensity. A typical CRNS processing chain applies a pressure correction using measured pressure and a method-specific coefficient or response model. The correct sign and reference level depend on how the correction factor is defined. Rather than copying a formula blindly, inspect the instrument or network documentation and state whether the product is raw or normalised to reference pressure.
Pressure is normally reported in hectopascals or equivalent units, and sensor pressure should be time-aligned with neutron counts. A faulty pressure sensor can create false moisture variability. A pressure correction calibrated for one detector or environment may not transfer exactly to another without justification. If raw counts drop while pressure rises, the decrease could be atmospheric rather than caused by wetter soil.
To translate a study, write the actual correction convention and its reference conditions in words before applying it. An adjustment intended to remove a pressure-related count change must not be confused with a change in soil water. The final water estimate depends on the corrected count rate and site calibration, not on atmospheric pressure alone.
Incoming cosmic-ray intensity also changes
The natural cosmic-ray supply varies through time. Monitoring networks can use reference neutron monitors or other correction schemes to account for changes in incoming intensity. A local station’s counts may therefore move with a broad regional or global cosmic-ray signal even when its soil and vegetation remain nearly unchanged. This is particularly important when comparing long time series or sites at different geographical positions.
A correction factor should be defined by the product’s convention. If the factor converts raw local counts toward a specified incoming-flux reference, its numerical direction matters. Applying the reciprocal by mistake can amplify rather than remove the unwanted variation. Always identify the reference data stream, temporal alignment, data gaps and whether the source was itself quality controlled.
Even after correction, the sensor still responds to soil moisture and environmental hydrogen, but there can be residual atmospheric and instrumental uncertainty. A credible paper does not declare every remaining change purely hydrological. It reports how incoming cosmic-ray variation was handled and how large that correction was during the study period.
Humidity is another hydrogen pool
Water vapour in air contains hydrogen and affects the moderation of secondary neutrons. Correcting for atmospheric humidity is therefore part of many CRNS methods. Air temperature and relative humidity may be used to estimate an absolute humidity variable, but the calculation and units should be specified. Relative humidity alone is not the same as the mass concentration of water vapour in air.
A hot day with moderate relative humidity can contain more atmospheric water vapour than a cooler day with the same relative humidity. If a processor substitutes one for another without the required conversion, it can misattribute weather-driven count changes to soil moisture. Sensor placement and time synchronisation are again important: a station’s atmospheric measurement should represent conditions during the neutron integration interval.
The term humidity correction should name which humidity quantity, reference and model were used. Treat the correction as a physical adjustment, not a generic smoothing filter. A corrected series may look calmer, but its credibility depends on the appropriateness of the adjustment and on independent calibration tests.
A simple correction-factor demonstration
Suppose a teaching example defines three dimensionless correction multipliers for pressure, incoming intensity and humidity as fP = 1.03, fI = 0.98 and fH = 1.01. The raw rate is 900 counts per hour. Under this hypothetical convention, N_corrected = N_raw × fP × fI × fH, so 900 × 1.03 × 0.98 × 1.01 = 917.54 corrected counts per hour.
The important result is that the corrected rate is about 917.5 counts per hour under the explicitly stated factor convention. Those factors have been invented for arithmetic practice, not extracted from an actual neutron monitoring network. Their numerical directions must not be used as a recipe for a real instrument. Different processing implementations may define factors differently or work with reference ratios rather than this simple product.
The arithmetic illustrates a second important point: correction factors multiply counts, not soil volumetric fractions directly. Applying fP, fI and fH to a final moisture percentage would generally be a different, unjustified operation because the count-to-water relation is nonlinear and site calibrated. Always adjust the observation in the proper stage of the processing chain.
What the measurement footprint means
The footprint is the region around a sensor from which environmental hydrogen meaningfully influences its neutron signal. It is not a uniformly weighted circle drawn with a permanent radius. Monte Carlo transport studies have found characteristic radial extents on the order of hundreds of metres, with estimates varying with soil moisture, air humidity and vegetation. A commonly discussed range is roughly 130 to 240 metres for particular definitions and conditions, not a universal boundary.
The sensor is often most sensitive to nearby soil, and radial weighting changes with water and the surrounding environment. A stream close to a station may matter differently from a stream near a nominal outer radius. Roads, buildings, crop patches and slopes can all contribute to spatial heterogeneity. Reporting a single radius without the weighting convention hides the fact that the station integrates a spatially structured signal.
For calibration, ground samples should reflect the footprint’s weighting rather than simply average every sampling location equally. Studies may use radial and depth weighting schemes informed by the physics and local conditions. The spatial support of a CRNS estimate is broad, but it is neither a tiny point nor a perfectly uniform area mean.
Sensing depth is also dynamic
Neutrons are influenced by hydrogen in near-surface soils, and their effective sensitivity decreases with depth. The depth range depends on moisture and soil properties. Published estimates can extend from a few tens of centimetres toward deeper values in dry conditions; one often cited simulation study reported approximate depth ranges of 15 to 83 centimetres across its conditions. Those figures define a method-dependent effective support, not an excavated cylinder of known constant depth.
As soil becomes wetter, neutrons may be moderated more strongly nearer the surface, often reducing effective sensing depth. After a shallow rainfall pulse, the station can respond before deeper probes change much. Conversely, a deep root-zone deficit may be partly invisible if the effective support is dominated by shallower material. A fixed-depth conversion that ignores these dynamics can mislead irrigation interpretations.
When a paper compares CRNS with point sensors at 10, 30 and 60 centimetres, check the actual weighting method. An unweighted arithmetic mean of those three measurements may not represent the neutron response. The relevant depths and weights can change across events, so dynamic comparison is often more defensible than a single permanent depth.
Ground calibration converts counts to water content
A neutron count rate is not itself a volumetric water fraction. A calibration procedure relates corrected counts to independently measured environmental water content while accounting for soil and non-soil hydrogen sources. Different empirical or model-based transfer functions are used, sometimes including a reference count N0 and additional parameters. The resulting relationship is typically nonlinear, so a fixed change of 100 counts need not equal the same moisture change everywhere on the curve.
Calibration requires well-designed field sampling over the support volume. Soil cores may be collected at multiple distances and depths, with gravimetric water content, bulk density and relevant chemical or organic matter information measured. Local vegetation, roads and water bodies can affect the appropriate sampling design. Reusing an N0 parameter from an unrelated field without validation can yield convincing time-series shapes but biased absolute moisture.
The correct interpretation reports the calibration date, soil states sampled, weighting scheme, reference measurements and uncertainty. A sensor that responds well to wetting and drying can be valuable even if absolute calibration remains uncertain, but the distinction between relative trend and reliable volumetric estimate should be explicit.
Gravimetric and volumetric moisture are not interchangeable
Gravimetric water content is mass of water divided by dry mass of soil, commonly expressed as kilograms per kilogram. Volumetric water content is volume of water divided by total bulk volume of soil, commonly cubic metres per cubic metre. They are related through dry bulk density and water density when units are consistent. Treating a gravimetric percentage as a volumetric percentage without density can introduce material error.
As a worked example, a dry soil has gravimetric water content 0.12 kilogram per kilogram and dry bulk density 1.45 grams per cubic centimetre. Taking water density approximately 1.00 gram per cubic centimetre, the volumetric fraction is 0.12 × 1.45/1.00 = 0.174 cubic metre per cubic metre, or 17.4 percent by volume. The result is not 12 percent by volume.
Field sampling must consider stones, coarse fragments, changing bulk density and sample preparation. A laboratory core representing fine earth alone may not match the bulk volume used in a footprint-scale water balance. A sound calibration states its mass and volume conventions instead of relying on the ambiguous label soil moisture percent.
A toy conversion curve with correct limitations
For teaching only, imagine a local calibration table giving corrected neutron count rates and corresponding volumetric fractions as follows: 1200 counts per hour corresponds to 0.15 cubic metre per cubic metre, 900 corresponds to 0.25, and 700 corresponds to 0.35. The inverse ordering captures the physical tendency for more water to moderate neutrons. The table is fictional and its simple intervals must not be treated as a field-calibrated transfer function.
Suppose the corrected observation is 1050 counts per hour and we use linear interpolation only between the first two table points. It lies halfway between 1200 and 900, so the estimated moisture is halfway between 0.15 and 0.25, namely 0.20 cubic metre per cubic metre. The interpolation describes the toy table and nothing beyond it.
For a corrected count rate of 1000, the fraction of the interval from 1200 toward 900 is 200/300, about two-thirds. The toy interpolation therefore gives 0.15 + (2/3) × 0.10 = approximately 0.2167 cubic metre per cubic metre. A real CRNS calibration may curve significantly, and extrapolation outside its calibration range should be avoided without validated physics or empirical support.
Why a weighted soil sample matters
Imagine two zones in a hypothetical footprint. The near zone has volumetric water content 0.12 and the far zone 0.32. If a validated sensitivity scheme assigns 75 percent of the relevant effective weight to the near zone and 25 percent to the far zone, the weighted mean is 0.75 × 0.12 + 0.25 × 0.32 = 0.17 cubic metre per cubic metre.
The simple arithmetic mean of the two values is (0.12 + 0.32)/2 = 0.22, considerably wetter. The difference is not an error in algebra; it comes from defining different spatial averages. In reality, CRNS weighting may be nonlinear and depends on three-dimensional hydrogen distribution, so the 75/25 model is a teaching device rather than a universal physical footprint rule.
This example explains why calibrating a broad-area neutron sensor with one nearby probe can fail. A point sensor measures conditions at its installation; a neutron station integrates heterogeneous influences. Agreement between the two during one rainfall event does not demonstrate they measure identical support volumes or will agree during all seasons.
Vegetation water and biomass corrections
Plants store hydrogen in water and organic tissue, and a changing crop canopy alters the neutron environment. As biomass develops, neutron intensity can decline independently of soil wetting. If a calibration was established during bare soil and used through rapid crop growth without accounting for changing aboveground hydrogen, the resulting soil-moisture estimate may drift wetter even when soil water follows a different trajectory.
Some research applies empirical biomass corrections or uses additional neutron energy information to help characterise vegetation effects. The appropriate method depends on detector, crop and local calibration. A biomass correction should not be copied as a universal percentage per kilogram without checking the study’s assumptions and units. Dry biomass and biomass water equivalent are not the same quantity.
When comparing a growing season, inspect the crop development stage, harvest dates, canopy water content and whether vegetation correction data were available. A sudden count increase after harvesting may not imply that soil dried in one hour. It may reflect removal of a significant hydrogen pool above the ground.
Snow, interception and surface water
Snow stores water and can change the neutron signal strongly. It also changes the shielding and spatial distribution of hydrogen around the detector. Wet litter, canopy-intercepted rainfall and temporary ponding may contribute hydrogen that is not part of the mineral soil’s pore-water fraction. A station used in snow conditions may need an application-specific interpretation or correction rather than a standard rain-season soil-moisture conversion.
A nearby pond or a river can affect the environmental neutron field. Its influence depends on location and weighting, not simply on the percentage of a large circular footprint that is water. Seasonal flooding can change the physical scene without the same change in subsurface water across agricultural soil. Similarly, the sensor may respond to a wet vegetation layer after rain before water has infiltrated deeply.
A good methods section describes surrounding water features, snow episodes, canopy interception and land-cover changes. If a sudden moisture increase is interpreted as infiltration, compare timing with precipitation, soil probes and surface wetness observations. The neutron signal sees environmental hydrogen; deciding its location requires additional evidence.
Roads, infrastructure and local anomalies
Roads, foundations and buildings may contain little soil water but can influence spatial weighting, scattering or the mixture of materials within the sensing area. Vehicles and storage tanks near a detector can introduce temporary changes in hydrogen or shielding. Even a station mounted near an unusual surface differs from a station above uniform open soil. Calibration assumptions should match the actual setting rather than an idealised field.
If a count series shows a sudden step after site maintenance, investigate changes in detector height, nearby objects, power, electronic settings and environmental shielding before attributing the step to hydrology. Installation stability and metadata are part of the measurement chain. An apparently long record may contain instrument configuration changes that require separate processing regimes.
For cross-site comparison, report detector type, height, local vegetation and major infrastructure. Soil moisture estimates should not be ranked as if all sites occupied identical hydrogen environments. Network processing can standardise many variables, but it cannot erase every local physical influence.
Counts during storms: timing and filtering
A storm can wet the soil rapidly and introduce canopy interception, surface ponding and humidity changes almost simultaneously. Raw neutron counts may fall, but the cause is a mixture of effects until corrections and independent observations are considered. If hourly integration contains only a modest number of events, counting noise may obscure the exact onset. Longer smoothing reduces noise but can shift apparent response timing.
Consider a site with 1800 counts in a typical hour and a genuine wetting response of only a few percent. The raw counting standard uncertainty is about 2.36 percent. A small one-hour shift near that scale should not be overinterpreted from a single interval. Multiple consecutive observations, appropriate correction and comparison with rainfall timing can strengthen the evidence. Rainfall also affects the atmosphere, so the raw signal alone is insufficient.
A useful hydrograph report shows precipitation, raw and corrected counts, volumetric moisture and processing flags on aligned time axes. It states whether gaps were interpolated and how event timing was defined. A narrow rainfall spike and broad neutron response do not necessarily imply slow infiltration; integration, footprint heterogeneity and corrections can also shape the response.
Comparing CRNS with TDR probes
Time-domain reflectometry, TDR, estimates soil water content from the dielectric response of soil near installed rods. Its support volume is local and strongly influenced by probe geometry, calibration and soil properties. CRNS integrates hydrogen across a much broader environment. Both may be converted to volumetric water fraction, yet equal units do not mean equal spatial representation. This is a common source of confusion in validation figures.
A TDR station at 10 centimetres depth might respond sharply to a small local infiltration pulse, while a neutron station integrates that pulse across a large heterogeneous area. Conversely, a broad storm may produce a coherent CRNS response even while one point probe sits in a dry patch. Neither disagreement automatically means one sensor is broken. The first step is to examine the support difference and expected water distribution.
The existing eduKate TDR guide explains waveform and permittivity calibration separately: https://edukatesg.com/2026/10/28/how-to-translate-easily-soil-moisture-tdr-permittivity-waveform-calibration/. Combining both instruments is useful when the sampling and weighting design is explicit. The comparison should test whether their observations can be reconciled under a physically credible soil-water distribution.
How to read an uncertainty budget
A CRNS moisture product inherits counting noise, uncertainty in atmospheric corrections, local calibration, soil density and organic matter, vegetation effects and footprint representation. Some are random; others persist as systematic bias across hours or seasons. Averaging many hourly values can reduce independent counting noise but will not remove an incorrect biomass correction or an inaccurate calibration parameter.
A laboratory determination of gravimetric water may be precise, yet spatial sampling can dominate calibration uncertainty. A few convenient soil cores cannot reveal all landscape heterogeneity. The footprint’s changing depth also means that validation uncertainty is state dependent. Error should therefore be described at the scale and moisture conditions relevant to the claim.
A scientifically careful confidence statement may say that the record resolves major seasonal wetting and drying while individual hourly absolute fractions have larger uncertainty. This is more useful than reporting four decimal places of volumetric water from a noisy and spatially mixed observation. Numerical display precision must not exceed evidence quality.
A reproducible processing workflow
Begin with detector identity, energy channels, integration interval, time standard and installation details. Preserve raw counts before applying corrections. Document pressure, humidity, incoming neutron intensity data and any detector-specific background treatments. Flag missing or implausible meteorological values rather than silently filling them with defaults that may change the soil-water estimate.
Next compute corrected counts using documented reference conditions and coefficients. Check correction magnitudes and discontinuities. Calibrate the count-to-moisture relation with spatially and vertically representative ground sampling, including bulk density and relevant non-soil hydrogen adjustments. Keep a record of sampling dates, soil conditions and uncertainty. Verify whether a calibration equation is being extrapolated beyond the observed water range.
Finally compare moisture estimates with independent point probes, precipitation, water-balance expectations and known land-cover changes. Inspect storms, snow, harvest and maintenance separately. Publish metadata explaining sample support and correction version. A quality-controlled time series should retain enough provenance that another researcher can reconstruct where each number came from.
Independent problem one: Poisson counting
A CRNS station measures 2500 counts during one hour. Under the simplified independent Poisson model, calculate the standard uncertainty of the count total, relative uncertainty and hourly rate. Then imagine the same stable rate measured over four hours and compute the relative counting uncertainty for the four-hour total.
For 2500 events, sqrt(2500) = 50 counts. Relative uncertainty is 50/2500 = 0.02, or 2.0 percent. The rate is 2500 counts per hour with counting standard uncertainty 50 counts per hour for the one-hour interval. Over four hours at the same true rate, the total is 10,000 counts, square root 100, relative uncertainty 100/10,000 = 1.0 percent.
The four-hour average has better relative counting precision under stable conditions, but it sacrifices temporal resolution. A correct answer also states that pressure, humidity, incoming intensity and calibration uncertainties have not been included. The Poisson calculation does not justify claiming that volumetric water content is accurate to one percent.
Independent problem two: moisture units
A soil calibration sample has wet mass 1.28 kilograms and dry mass 1.10 kilograms. Its dry bulk density is 1.35 grams per cubic centimetre. Calculate gravimetric water content and approximate volumetric water content using water density 1.00 gram per cubic centimetre. Distinguish both denominators.
Water mass is 1.28 − 1.10 = 0.18 kilogram. Gravimetric content is 0.18/1.10 = about 0.1636 kilogram per kilogram. Volumetric content is 0.1636 × 1.35/1.00, approximately 0.2209 cubic metre per cubic metre, or 22.1 percent by volume. The gravimetric value would be 16.4 percent by dry mass.
If a paper states only 22 percent moisture without specifying whether it is by mass or volume, the reader cannot safely compare it with another dataset. The repair is to demand units, density and sample preparation conventions. For rocky or organic soils, additional treatment may be necessary to represent the relevant bulk volume and hydrogen contribution.
Independent problem three: the apparent rain response
A field sensor reports a 5 percent reduction in raw neutron counts during a humid storm. Air pressure rose, atmospheric moisture increased, and leaf surfaces became wet. The user claims that mineral soil water content increased by exactly 5 percent. Explain why this conclusion is invalid and propose a testable interpretation.
The count change is not numerically equal to water content change, because the transfer function is nonlinear and calibrated in different units. Pressure, incoming cosmic-ray intensity, atmospheric humidity and intercepted water may affect the raw signal. A corrected count series and site-specific conversion are needed, and even those may retain uncertainty about where the extra hydrogen is located.
Compare corrected counts with rainfall, canopy wetness and shallow and deep soil probes. Inspect processing flags and detector condition. An honest interim conclusion is that the neutron environment changed during the storm, consistent with multiple water-related processes, rather than asserting a precise soil-water increment from raw counts alone.
Independent problem four: a weighted footprint
A simplified survey has three zones with volumetric water contents 0.10, 0.20 and 0.35 cubic metre per cubic metre. Suppose a validated weighting method for this exercise assigns fractions 0.50, 0.30 and 0.20 to them. Compute the weighted mean and compare it with the arithmetic mean. Explain why these numerical weights are not universal CRNS constants.
The weighted mean is 0.50 × 0.10 + 0.30 × 0.20 + 0.20 × 0.35 = 0.05 + 0.06 + 0.07 = 0.18 cubic metre per cubic metre. The arithmetic mean is (0.10 + 0.20 + 0.35)/3 = approximately 0.2167. Both are valid averages under different weighting definitions, but they answer different spatial questions.
Actual neutron sensitivity depends on radial distance, depth, moisture, humidity, vegetation and geometry, not arbitrary three-zone weights. A reader should ask whether sampling locations and weights were derived from a physically appropriate scheme. Calibrating a landscape-integrating instrument with unweighted convenience samples can introduce a persistent bias.
Independent problem five: a harvest discontinuity
A CRNS moisture series appears to become abruptly drier on the day a dense crop is harvested, while nearby point soil sensors and rainfall records show little immediate soil change. The canopy water and dry matter were removed from the sensing environment. Formulate two competing explanations and identify how the records can distinguish them.
One explanation is genuine soil drying, perhaps through rapid evaporation or drainage; another is that removal of aboveground hydrogen increased detected neutron counts without the implied soil-water loss. Check raw counts, atmospheric correction, harvest timing, biomass estimates and independent soil water observations. If the step aligns exactly with crop removal and soil probes remain stable, a vegetation-related explanation becomes credible.
The conclusion should still consider possible irrigation or soil disturbance during harvest. A properly corrected and calibrated record may require a new biomass state or adjustment rather than treating the whole step as hydrology. The example shows why land management metadata can be as important as meteorological data in environmental sensing.
FAQ on cosmic-ray soil moisture
Does the station contain a radioactive source? Standard passive CRNS uses naturally produced secondary cosmic-ray neutrons, unlike many conventional active or source-based neutron probes. Does it measure every soil point equally within a circle? No. Is its sensing depth always 80 centimetres? No; effective depth depends on wetness and other conditions. Can lower raw counts indicate wetter soil? Often, but only after checking atmospheric and other hydrogen influences.
Do raw counts equal water content? No. Can one local TDR probe validate a broad CRNS footprint perfectly? No. Must calibration use bulk density? It is commonly needed when gravimetric core water is converted to volumetric fraction, with attention to site conditions. Does a lower count mean a fixed increase of 0.05 cubic metre per cubic metre? No; the count-to-moisture relation is calibrated and nonlinear.
Can CRNS be useful for irrigation and hydrological modelling? Yes, especially for intermediate spatial scales not sampled by a single point probe, when correction, calibration and application match. Can snow or biomass masquerade as soil moisture? Yes, and these influences must be considered explicitly. Should all network products use identical settings? Standardisation helps comparison, but local corrections and metadata remain necessary.
How to write a trustworthy interpretation
Start by naming the sensor and quantity: raw neutron detections over a defined interval, corrected neutron count rate or calibrated volumetric soil water content. State the reference conditions used in pressure, incoming intensity and humidity corrections. Identify any treatment of vegetation, snow, water bodies and infrastructure. Tell readers how calibration samples were collected and weighted, not merely that calibration was performed.
Next describe the footprint and its environmental dependence. A kilometre-scale weather map, a metre-scale TDR probe and a hectometre-scale neutron sensor provide observations at different supports. Their disagreement can be informative about spatial heterogeneity. If the comparison assumes uniform depth or a perfectly circular equal-weighted footprint, name the approximation and test its sensitivity when possible.
The defensible conclusion connects the original count observation to the soil-water estimate through documented physical corrections and independent field validation. It does not convert a coloured map, a smooth time series or a high number of digits into certainty. Cosmic-ray neutron sensing is powerful precisely when its broad environmental sensitivity is understood rather than hidden.
Original research, network documentation and related guides
UK Centre for Ecology & Hydrology COSMOS: how cosmic-ray soil moisture is measured
UK CEH COSMOS: measured and derived data variables
Köhli et al. (2015), Water Resources Research: dynamic cosmic-ray neutron footprint
Jakobi et al. (2022), Water Resources Research: dynamic biomass corrections