Microwave Radiometer PWV, LWP and Brightness Temperature

First translate the observation, not the colourful chart

A ground-based microwave radiometer measures incoming microwave radiation from the sky and reports calibrated brightness temperatures at specified frequencies. It does not directly collect raindrops, weigh cloud droplets, or read the temperature of the whole atmosphere with an ordinary thermometer. The instrument senses emission and absorption along a line of sight. A retrieval algorithm then uses those observations, physical radiative-transfer relationships and sometimes supporting meteorological inputs to estimate integrated water vapour and cloud liquid water. Understanding this chain prevents some of the most consequential mistakes in translating weather-science numbers.

Precipitable water vapour, abbreviated PWV, is the liquid-equivalent depth that would result if all water vapour in an atmospheric column were condensed and collected over its horizontal area under the definition’s ideal accounting. Integrated water vapour, sometimes IWV, is ordinarily presented as kilograms per square metre, and has a simple numerical equivalence to PWV in millimetres of liquid water. Neither metric is rain already falling. They describe atmospheric water vapour inventory in a column, subject to retrieval assumptions.

Liquid water path, or LWP, is the column integral of liquid cloud water content, ordinarily reported in grams per square metre or kilograms per square metre. It excludes water vapour because vapour and liquid are different phases, and instrument algorithms may distinguish cloud liquid from larger precipitation drops imperfectly. A chart reporting 150 g m^-2 LWP describes one-and-a-half tenths of a kilogram of suspended liquid per square metre along the sampled column. It does not mean the ground received 150 millimetres of rain.

Brightness temperature Tb uses kelvin because radiation intensity is expressed as the temperature a blackbody would have to yield the corresponding radiance at that frequency under the relevant convention. Atmospheric emission depends on frequency, water vapour, oxygen, cloud liquid and physical temperature along the path. Thus a 40 K brightness temperature is not a claim that air or cloud droplets have physical temperature 40 K. It is a calibrated radiometric variable whose interpretation requires transfer physics.

These distinctions matter beyond specialist meteorology. A news report that turns PWV into a coming rainfall total, a lesson that calls LWP ‘cloud thickness’, or a multilingual caption that converts Tb to Celsius as though it were ambient air temperature loses important meaning. The transferable method is to identify what the instrument directly measured, what was reconstructed using a model, the spatial and temporal scale of the reconstruction, and independent observations that could validate it.

The spectrum: why different channels respond differently

Microwave frequencies interact differently with atmospheric gases and condensed water. Ground instruments often observe channels near the 22 GHz water-vapour absorption feature and the atmospheric window near 31 GHz, where cloud liquid water has useful influence. Individual observing systems can use channels around 23.8 and 31.4 GHz, but those precise frequencies are instrument-specific rather than universal. Some radiometers add a channel close to 89 or 90 GHz, providing extra sensitivity to clouds under suitable conditions, while multi-channel profilers also observe oxygen-related frequencies to constrain temperature structure.

Water vapour affects radiation through molecular absorption and emission, whereas small cloud droplets contribute a more broadband liquid-water absorption signal in microwave windows. With more than one well-calibrated frequency, the relative spectral signatures can help estimate separate column vapour and liquid components. In a single-frequency design, those components can be more difficult to distinguish without auxiliary temperature, humidity or model information. A statistically trained retrieval may still be useful, but its predictions remain conditional on its training domain and inputs.

A radiometer points at a region of sky with finite angular width, and can scan different elevations. At lower elevation, the atmospheric path is longer, so more emission may be received even if the local vertical cloud column is unchanged. This geometry can help constrain opacity but also adds complications if the atmosphere varies horizontally, the beam intersects different clouds, or rain wets the observing window. A value from a slant look cannot automatically be assigned to the vertical atmospheric column without a retrieval accounting for that geometry.

Radiative transfer relates brightness temperature to emission, absorption, source temperatures and any background radiation. In a simplified non-scattering isothermal slab model with opacity τ and physical temperature T, Tb can be sketched as T(1−e^-τ) plus attenuated background. This equation is a teaching approximation, not an endorsed universal operational inversion. Real atmospheric profiles are not uniform-temperature slabs; clouds and gases are layered, scattering may matter and instruments require calibration.

As opacity increases, Tb may rise towards an effective emitting-layer temperature and respond less linearly to extra absorber. A channel that is nearly opaque can have poor sensitivity to additional column water even though its brightness looks high. Conversely, a thin cloud may change some frequency channels only by a small number of kelvins, near calibration and model uncertainties. Brightness-temperature trends therefore need channel labels and retrieval sensitivity rather than a universal rule that brighter means wetter.

How to read PWV and IWV units without inventing rain

PWV and IWV express the integrated mass of water vapour over an atmospheric column. A column with IWV=25 kg m^-2 corresponds to 25 mm of liquid-equivalent water because a water layer one millimetre deep over one square metre has volume 0.001 m³ and mass approximately one kilogram. This simple unit conversion is exact at the conventional water-density approximation used for millimetre water equivalent. It is a conversion of inventory units, not a forecast of when or where rain will fall.

For a worked area translation, suppose IWV is 20 kg m^-2 across a hypothetical square kilometre of horizontally uniform column. The area is one million square metres, so total vapour mass represented over that region is 20 million kilograms. A reader must resist the conclusion that twenty thousand tonnes of rain will necessarily fall there. Atmospheric water vapour is continually transported, evaporated, condensed and advected. A storm may import additional vapour, export most of it or produce no precipitation at that specific location.

A high PWV value can coexist with no low-level cloud and no immediate rain. Condensation requires thermodynamic and dynamical conditions; precipitation also requires growth processes and a favourable path from cloud to ground. An air mass may hold substantial vapour while remaining unsaturated in much of its column. PWV is therefore relevant to potential moisture supply but insufficient as a deterministic rain warning. Humidity structure, lift, instability, cloud development and storm motion must be considered.

Different atmospheric conditions can produce similar column-integrated PWV but different vertical moisture profiles. For example, one atmosphere might have a moist boundary layer and comparatively dry upper air, while another has moisture distributed deeper into the troposphere. The integral may be equal, yet cloud-base conditions, convective potential and infrared structure differ. Basic microwave channel retrievals cannot uniquely identify the full vertical humidity profile from one or two integrated observables. A sounding or a well-characterised profiling method provides additional constraints.

Time changes in PWV are also not equal to rainfall during that interval. A decline of five millimetres of column vapour could result from horizontal advection, subsidence or redistribution as well as precipitation, and simultaneous evaporation and transport can obscure relationships. Water conservation only gives useful precipitation estimates when the relevant mass-flux terms and boundaries are included. Avoid deriving a surface rain depth from the difference between two PWV snapshots without a proper moisture budget.

Reading liquid water path and cloud structure

LWP is the integral through a column of liquid-water content, often LWC in grams per cubic metre. If one imagines a uniform 1000-metre-deep liquid cloud with LWC=0.15 g m^-3, the column contains 150 g m^-2 and has LWP of 150 g m^-2. The calculation is LWC multiplied by depth, with metres cancelling appropriately. This is a hypothetical vertically uniform example and does not imply the radiometer observed the depth or a constant profile.

Using liquid-equivalent units, 150 g m^-2 equals 0.150 kg m^-2 or 0.150 mm of water. The conversion is straightforward, but the resulting millimetres are not a surface rain-gauge total. Cloud water can remain suspended, re-evaporate, flow horizontally, collect into drizzle, or grow through processes replenished by continuing condensation. Rainfall accumulation over an hour can exceed or fall below the instantaneous liquid water path depending on the dynamics. Treat LWP as inventory, not delivered rainfall.

A cloudy column with a very small positive retrieved LWP may be within the noise and uncertainty of the instrument and retrieval. Some software suppresses negative raw retrieval values to zero, but clipping changes the distribution and can bias averages near zero. Inspect product documentation for clear-sky offsets, noise thresholds and quality controls before interpreting a few grams per square metre as a meaningful cloud change. Differences should be weighed against known calibration and algorithm uncertainties.

Radiometers integrating along a path generally have limited direct information on where within that path cloud liquid is located. The same LWP could arise from a thick cloud with low average LWC or a thin cloud with high average LWC. A ceilometer or cloud radar can constrain layer locations and boundaries, and a radiosonde or temperature-profile retrieval can supply environmental information. Combining instruments can yield more useful vertical interpretations but also introduces alignment, model and sampling assumptions.

Some cloud and drizzle retrievals require attention to scattering by larger droplets. Atmospheric System Research reports show that neglecting drizzle-drop scattering in a relevant marine cloud analysis can bias total retrieved liquid amounts. The point is not that all LWP products always fail in drizzle. It is that a retrieval model built primarily around cloud-liquid absorption may need additional physics when precipitation-size particles contribute appreciably. State whether a cited method addresses such particles before generalising.

Clouds containing ice or mixed phases complicate a water-path label. LWP conventionally concerns liquid rather than ice water path, and a radiometer’s sensitivity may not unambiguously separate all hydrometeor components without supporting observations. A cold cloud can contain supercooled liquid as well as ice. Avoid silently treating the whole cloud’s condensed water inventory as liquid because a single retrieval field happens to carry the name LWP.

Calibration, rain on the window and retrieval uncertainty

A microwave radiometer’s radome or protective window can become wet during rain. Liquid water on the instrument enclosure emits and absorbs microwave radiation and changes observed Tb independently of the intended atmospheric path. A rain sensor may flag the event, but a radome may remain damp after the rain stops. Recent instrument-quality research documents spectral inconsistencies and residual brightness-temperature biases associated with wet radomes. A suspicious retrieval after a shower may therefore be an instrument-enclosure effect.

Suppose a radiometer shows a sudden several-kelvin increase across channels immediately after heavy rain while nearby temperature and cloud observations do not support a matching atmospheric change. Investigate the rain flag, radome condition, spectral residuals and recovery pattern. Do not automatically turn the extra Tb into tens of grams per square metre of cloud liquid. An artefact and a physical atmospheric change can both raise some channels, so quality assurance uses multiple measurements and time behaviour.

Calibration also matters in dry conditions. Biases in hot- and cold-load referencing, receiver stability and absolute radiometric scales can propagate through retrieval coefficients into PWV or LWP. Small brightness errors can have different consequences at different frequencies and atmospheric states. A correction must be documented for the exact instrument and configuration. It is not scientifically sound to use a universal ‘one kelvin equals one millimetre of water’ formula for all channels.

Atmospheric retrieval algorithms depend on assumptions about temperature profiles, humidity distributions, water-vapour continuum absorption, cloud microphysics and the geographical conditions represented by training data. A machine-learning or statistical regression may perform well on its validation sample but be less accurate under climates or cloud regimes that differ from training. A physically based inversion also relies on forward-model correctness and regularisation. Both approaches should report validation metrics and the domain in which they were evaluated.

Finally, uncertainty should be associated with the derived quantity and conditions, not borrowed from an unrelated instrument. A published result of an 89 GHz single-channel retrieval reported particular percentage errors on its study samples and noted climate dependence. Those percentages are not generic specifications for every two- or three-channel radiometer. The scientifically useful reading is that retrieval accuracy depends on sensor design, auxiliary predictors, atmospheric state and assessment method.

Checked worked examples: interpreting radiometer quantities

Worked example 1 starts with precipitable water vapour PWV=28 mm. In water-equivalent mass units this corresponds to integrated water vapour IWV≈28 kg m^-2, because one millimetre of water over a square metre has about one kilogram of mass. The conversion does not change the physical atmospheric vapour inventory. It does not state that twenty-eight millimetres of rainfall is expected, because moisture convergence, condensation, storm transport and evaporation remain separate processes.

Worked example 2 considers a horizontally uniform atmospheric column with IWV=18 kg m^-2 across a hypothetical two-square-kilometre area. Two square kilometres equal two million square metres. Multiplication gives 36 million kilograms of vapour in the columns above that region under the assumed spatial uniformity. Dividing by the area returns 18 kg m^-2. This is a mass-inventory exercise, not a claim that all 36 million kilograms will fall inside the area during a weather event.

Worked example 3 gives liquid water path LWP=120 g m^-2. Convert grams to kilograms: 120/1000=0.120 kg m^-2. The equivalent liquid water depth is about 0.120 mm. Writing 120 mm would overstate the equivalent depth one thousandfold. The value is an integrated cloud-liquid inventory; without a precipitation flux history it cannot be transformed into an accumulated rain-gauge reading.

Worked example 4 asks for mean cloud liquid water content from a stated LWP=200 g m^-2 and an independently established 500-metre cloud thickness. Under the simplifying assumption that liquid content is uniform, mean LWC=200/500=0.4 g m^-3. That average does not show the profile’s shape. A thin high-LWC layer and a thick low-LWC layer can yield the same path integral. If the height were guessed rather than measured, the calculated content would inherit that uncertainty.

Worked example 5 reverses the calculation. Imagine a 750-metre cloud layer with constant liquid water content 0.2 g m^-3. Its LWP is 0.2×750=150 g m^-2, equal to 0.15 kg m^-2 or 0.15 mm of liquid-equivalent water. The result is meaningful only within the stated uniform-layer model. Real water contents can vary with height and across the radiometer’s finite beam, so it is not a complete retrieval recipe.

Worked example 6 compares PWV=30 mm with LWP=300 g m^-2 in the same column. Converting LWP gives 0.3 kg m^-2 or 0.3 mm water equivalent, while PWV represents 30 kg m^-2 of vapour. The vapour inventory is one hundred times larger by mass in this hypothetical scenario. The comparison does not mean that only one percent of the moisture is capable of becoming precipitation; the system is open to moisture transport and continued condensation.

Worked example 7 interprets a time change. PWV falls from 36 to 31 mm between two quality-controlled retrievals. The observed column-integrated vapour difference is five millimetres water equivalent. It would be unjustified to announce that five millimetres fell as rain: dry-air advection, changing vertical winds, surface evaporation and condensation all affect the atmospheric budget. A full diagnostic needs winds, moisture flux convergence, precipitation and possibly storage-change terms.

Worked example 8 shows why brightness temperature cannot be read like air temperature. A channel reports Tb=65 K while a surface thermometer reads 299 K. The values need not contradict because Tb is a radiance-equivalent temperature determined by a weighted atmospheric path, absorption and background radiation, whereas the thermometer senses local physical temperature. Subtracting them yields a numerical difference of 234 K but not a scientifically meaningful correction to PWV without a radiative-transfer model.

Worked example 9 uses a simple non-scattering isothermal slab solely to illustrate opacity. Take atmospheric effective temperature T=280 K, negligible background and opacity τ=0.2. Then Tb≈280(1−e^-0.2)≈50.76 K. For τ=0.4, Tb≈92.31 K, less than twice 50.76 K despite doubled opacity. The nonlinearity matters, and real vertically varying temperatures invalidate a universal inversion from one channel. This is a toy calculation, not a site-specific retrieval.

Worked example 10 examines approaching saturation. With the same simplified slab T=280 K, opacity τ=1 gives Tb≈176.99 K, while τ=2 gives Tb≈242.11 K. The additional unit of opacity increases brightness by about 65.12 K, less than the first unit’s 176.99 K from τ=0. The response depends on starting opacity. This illustrates why highly opaque channels may carry less sensitivity to further increases in absorber column.

Worked example 11 identifies frequency dependence. One channel responds strongly to water vapour, another is additionally sensitive to liquid cloud water. If both channels brighten, possible explanations include more vapour, more liquid or temperature and viewing-geometry changes. With only two raw Tb values one should not solve for two unknowns by arbitrary subtraction. A legitimate retrieval has calibrated response functions and environmental assumptions, and reports its performance.

Worked example 12 considers a wet radome. A dry sky measurement is followed immediately after rainfall by a 5 K rise in several channels, although other observations show limited atmospheric change. The jump could be liquid on the instrument window. The analyst should inspect window heaters, rain sensors, spectral residuals, maintenance records and how long the signal takes to recover. A five-kelvin change is an observed radiometric difference, not automatically five millimetres of precipitable water.

Worked example 13 addresses a very small retrieved LWP=8 g m^-2 with a documented uncertainty of, hypothetically, 20 g m^-2 for the specific product and conditions. Since the apparent signal is smaller than that uncertainty, the data do not resolve a confident nonzero liquid amount. The correct report would avoid asserting that a cloud newly formed or that exactly eight grams were present. The 20 g m^-2 assumption is illustrative and must not be used as a universal sensor specification.

Worked example 14 asks what happens if all slightly negative LWP estimates are reset to zero. For three raw estimates -10, 0 and +10 g m^-2, their original arithmetic mean is zero; the clipped values 0, 0 and 10 have mean 3.33 g m^-2. The numerical distribution and average change even without new atmospheric evidence. Whether clipping is appropriate for a particular product is a methodological choice, but an analyst must disclose it and avoid treating the adjusted mean as raw measurement.

Worked example 15 concerns a cloud layer detected by ceilometer between 600 and 1000 metres above ground. The independently estimated thickness is 400 metres. With LWP=160 g m^-2, the uniform-layer average LWC would be 160/400=0.4 g m^-3. This does not establish that liquid water fills the entire layer equally. The ceilometer cloud boundaries may themselves be uncertain in fog, precipitation or optically thick conditions, so both instruments contribute to uncertainty.

Worked example 16 contrasts column retrieval with rainfall amount. An LWP of 0.2 mm water equivalent persists while a rain gauge collects 4 mm over one hour. There is no necessary contradiction: liquid can be replenished by condensation and transported through the cloud during that hour, and falling drizzle or rain can involve additional liquid components. Instantaneous inventory and integrated downward flux are different physical quantities. A complete water budget would account for supply, removal and storage changes.

Worked example 17 considers a retrieval using one 89 GHz channel plus near-surface weather and climatological priors. The derived PWV might look similar to a multi-channel result, but the dependence on auxiliary predictors differs. One cannot infer that the single-frequency radiance uniquely separated vapour and liquid by itself. Evaluate performance against independent profiles, the range of training climates, and whether supplementary model information influenced the answer.

Worked example 18 explores time-series representativeness. A radiometer averages brightness temperatures over one minute while a radiosonde launched nearby samples a moving balloon’s profile over many minutes. PWV values may differ because the instruments sample different positions and times, not only because one calibration is wrong. To assess agreement, record averaging windows, radiosonde launch time, balloon drift, retrieval version and any rain flags. A numerical difference without this context is not a calibration diagnosis.

Quality assurance across instruments and algorithms

A skyward-looking microwave radiometer, a tipping-bucket rain gauge and a radar can report different numerical values during one shower because their measured quantities differ. The radiometer senses atmospheric microwave emission along a path, the gauge accumulates water at a point, and the radar observes scattered power in an elevated volume. A meaningful comparison first chooses a common physical target, then aligns timing, spatial support, phase and retrieval method. Direct numerical agreement between incompatible quantities is not expected.

A radiosonde provides an independently sampled humidity and temperature profile that can be integrated to estimate column vapour. Agreement with radiometer PWV is informative, but radiosonde humidity sensors can have their own biases and temporal differences. A cloud radar or ceilometer can reveal where cloud boundaries lie, though neither instrument automatically measures liquid water path. Cross-instrument verification is a chain of conditional constraints, not a contest where one sensor is declared unquestioningly correct.

Multi-channel retrievals may use statistical regression, neural networks or iterative radiative-transfer fitting. Training and validation data can include atmospheric profiles generated from radiosondes or numerical weather prediction. When a model trained primarily in midlatitude conditions is applied in very humid tropical conditions, its error characteristics may change. Validation on the original test dataset does not prove universal geographic transfer. Report the training regime and independent local evaluation where available.

Data-quality flags are part of the measurement and should not be discarded when exporting a chart. Rain contamination, missing calibration, receiver problems and out-of-domain input conditions can invalidate or degrade individual retrieval samples. It is better to show an honest gap than interpolate a fictitious cloud signal across a wet-radome period without explicit justification. Any gap filling should be labelled as reconstructed data.

Spectral consistency checks compare observed brightness temperatures with predictions from plausible atmospheric states or other channels. Large residuals concentrated after rainfall can suggest a wet radome or sensor issue. Such diagnostics are especially valuable because a simple rain flag may clear before the instrument window dries. A persistent bias should prompt examination of maintenance, hydrophobic coating, heating, rain timing and ambient conditions rather than only editing atmospheric-retrieval coefficients.

Independent transfer problems with answer explanations

Transfer A: A dashboard labels a field ‘IWV 32 kg m^-2’. Express the same column inventory as precipitable water. Answer: approximately 32 mm of liquid-equivalent water, using the conventional one-kilogram-per-square-metre to one-millimetre conversion. The correct explanation adds that this is vapour in the air, not thirty-two millimetres measured by a rain gauge. Precipitation depends on processes beyond the column total.

Transfer B: An LWP retrieval gives 75 g m^-2. Convert to kilogram per square metre and liquid equivalent depth. Answer: 0.075 kg m^-2 and about 0.075 mm. A translation saying seventy-five millimetres has missed the factor of one thousand between grams and kilograms. State which original product uses g m^-2, because another dataset may publish LWP directly in kg m^-2.

Transfer C: A hypothetical cloud has average LWC 0.3 g m^-3 and occupies a vertical layer 800 metres thick. What is its LWP? Answer: 0.3×800=240 g m^-2. This assumes a spatially uniform concentration along the line of sight and that the whole stated thickness is liquid cloud. In real conditions a retrieval of total path does not establish the uniformity or the exact vertical boundaries.

Transfer D: A 300 g m^-2 liquid water path is spread evenly through a 1.5-kilometre layer. Find mean liquid content. Answer: 300/1500=0.2 g m^-3. The layer length must be converted to metres before division. A different vertical distribution could have the same path integral, so the result is only the average under the specified thickness.

Transfer E: A student claims a radiometer Tb of 75 K means the cloud temperature is minus 198.15 degrees Celsius. Explain the problem. Answer: the arithmetic conversion from kelvin is irrelevant to the claimed interpretation, because brightness temperature is the blackbody-equivalent microwave radiance, not a direct thermometer measurement of one cloud level. Atmospheric emission and absorption over the complete path control the reported radiance.

Transfer F: Two channel brightness temperatures rise by 4 K and 2 K. Can their numerical difference directly yield LWP in millimetres? Answer: no. Channel responses depend on radiative transfer, calibration, absorber spectra and atmospheric temperature. A proper retrieval may use a multi-frequency model, but the raw Kelvin differences do not carry the units of column liquid mass. The analyst should ask for coefficients, domain and uncertainty.

Transfer G: PWV grows from 22 to 27 mm over one hour while no rain is observed. Is the increase impossible? Answer: no. Moisture advection and local evaporation can add vapour without condensation or precipitation. An atmospheric column is not a sealed container. A rigorous explanation considers wind-driven moisture convergence and changes to vertical humidity, rather than treating every increase in PWV as a pending rainfall event.

Transfer H: A modelled LWP changes from 5 to 15 g m^-2 while the product’s stated uncertainty is ±25 g m^-2 in those conditions. Is the increase established? Answer: the change is ten grams per square metre, smaller than the cited scale of uncertainty, so the evidence may not resolve it robustly. The uncertainty should be treated according to its actual definition and correlation; one cannot infer statistical significance merely by comparing raw numbers.

Transfer I: A radome remains wet after a storm, but the rain sensor has turned off. Can the retrieved LWP immediately be treated as reliable? Answer: not automatically. Water on the protective window may alter brightness temperatures after rain ends. Inspect spectral residuals, temporal recovery, window condition and product quality flags. The rain sensor’s absence of current detection is not proof the optical or microwave path has dried completely.

Transfer J: A clear sky is reported with PWV=45 mm and LWP near zero. Are these observations contradictory? Answer: no. Water vapour can be abundant in unsaturated air without condensed liquid cloud. A high total column vapour inventory does not require a cloud at the observation time. Verify that the retrievals are quality controlled, but do not infer a physical impossibility from the difference between vapour and liquid paths.

Transfer K: A cloudy column has LWP=200 g m^-2 while a rain gauge measures 6 mm over an hour. Does conservation fail? Answer: no. Instantaneous cloud liquid inventory is only 0.2 mm equivalent, whereas precipitation is a cumulative flux through time and can draw on continuing moisture supply, condensation, growth and transport. A true budget needs storage change and fluxes, not a comparison of one snapshot with an hourly total.

Transfer L: One radiometer estimates PWV using channels around 23 and 31 GHz, while another uses 89 GHz and surface-weather inputs. Can the products be assumed identically accurate? Answer: no. Instrument channels have different sensitivities and the models use different auxiliary information. Compare independent validation and climate range. The same final unit does not mean the measurements contain the same information.

Transfer M: A colleague multiplies LWP by horizontal cloud area to obtain total liquid mass. What must be checked? Answer: use compatible units, confirm LWP is spatially representative and integrate spatial variations where possible. If LWP=100 g m^-2 uniformly over 1000 m², total suspended liquid mass is 100,000 g or 100 kg. This is an atmospheric inventory estimate, not a rain total, and the uniformity assumption may be strong.

Transfer N: A one-minute PWV record is compared with a radiosonde humidity integral derived from a balloon that drifted several kilometres over half an hour. Why may values differ? Answer: the instruments sample different space-time volumes and may use different calibration and retrieval assumptions. Align times as closely as possible, record the balloon path, examine atmospheric heterogeneity, and evaluate known sensor biases before declaring either result wrong.

Transfer O: A graph shows negative unfiltered LWP values near clear-sky periods. Must the atmosphere contain physically negative cloud water? Answer: no. Negative estimates can arise from retrieval noise, calibration baseline, inversion unconstrained by positivity or model error. Whether to clip them for reporting depends on the statistical purpose and product guidance. Simply replacing every negative value with zero can bias means upward.

Transfer P: A thin cloudy layer is located by a ceilometer but its retrieved LWP changes while cloud base stays fixed. Does the cloud necessarily become geometrically thicker? Answer: no. Cloud water content may increase at roughly unchanged depth, the cloud top may change, or retrieval errors may be involved. The ceilometer constrains some vertical boundaries, not every liquid-water profile. A multi-instrument interpretation asks which physical variable actually changed.

Transfer Q: A publication reports a ten-percent relative error for PWV in one dataset. Is ten percent a universal uncertainty for all radiometers in all climates? Answer: no. The figure belongs to a particular instrument, retrieval method, atmospheric sample and evaluation metric. Humidity regime, rain contamination and predictor choice can change performance. A trustworthy translation states where and how the uncertainty was established.

Transfer R: A graph labels water vapour path as millimetres and liquid water path as grams per square metre. A reader adds 30 and 120 and reports 150 millimetres. Repair the calculation and interpretation. Answer: convert 120 g m^-2 LWP to 0.12 mm liquid equivalent, giving a numerical combined atmospheric water mass equivalent of 30.12 mm under consistent assumptions, but even that sum crosses vapour/liquid inventories and is not a rainfall forecast. The wrong direct sum had incompatible units.

Frequently asked questions about microwave water-path charts

What does a microwave radiometer physically measure? It detects naturally emitted microwave radiation from the observed sky direction, calibrated as brightness temperature at known frequency channels. PWV and LWP are estimated with physical or statistical retrievals. Calling both direct measurements obscures assumptions and model uncertainty. Always retain instrument frequency, viewing geometry and processing method when the result matters.

Does 40 mm PWV mean 40 mm of rain is expected today? No. PWV is a water-vapour inventory, not a surface rainfall forecast. Moisture can remain unsaturated, move horizontally, evaporate or condense elsewhere. The amount of rain reaching one place can be smaller or greater than the initial local vapour inventory because atmospheric transport and ongoing source terms matter.

Why is LWP sometimes expressed as g m^-2 rather than mm? Column liquid content can be small compared with familiar rain depths, making grams per square metre convenient. The conversion is 1000 g m^-2 equals roughly 1 mm of liquid-equivalent water. The choice of scale should not change the physical concept. LWP represents suspended liquid cloud water, not the total surface water received during an interval.

Can a radiometer alone identify the height of a cloud? Limited microwave channels primarily constrain path-integrated radiative information and usually cannot uniquely locate every liquid layer. Combining a ceilometer or cloud radar with temperature profiles can help infer layer structure, although each additional observation brings sampling and model limitations. Do not infer cloud thickness from LWP alone without an independently supported profile assumption.

Why do wet windows matter? Water on the radome modifies emitted and transmitted microwave radiation near the sensor and can mimic atmospheric signals. A window may remain wet after a precipitation flag stops. Quality control uses rain history, channel consistency, maintenance information and recovery behavior. The most important correction is often rejecting contaminated intervals rather than fabricating an exact atmospheric water path.

Does a higher brightness temperature always mean more cloud liquid? No. Different channels respond to vapour, cloud liquid, physical atmospheric temperature and viewing geometry. Some responses saturate and some can be altered by rain or radome contamination. A multi-channel radiative-transfer or statistically validated retrieval is needed. The numerical Tb value should not be translated directly into a rain amount.

What is the central skill to retain? Separate measured radiation in kelvin, retrieved water-vapour column in mm or kg m^-2, retrieved liquid-water column in g m^-2, and rainfall accumulated at the ground in mm. Keep the sampling direction, instrument-quality flags and validation evidence attached to each quantity. Scientific translation means making the inference trail clear enough that another reader can reproduce or challenge it.