Ambient SO₂ ultraviolet-fluorescence data answer a carefully bounded question: what concentration of sulfur dioxide did a particular sampling and optical measurement system estimate for air entering its instrument during a defined interval? To translate the record into useful language, identify the sampling location, time window, units, calibration and quality flags before explaining a peak. The most important step is often deciding what the measurement cannot establish.
A sharp sulfur dioxide trace can represent a real ambient event, an introduced calibration gas, a sampling-system interruption or a transient optical interference. Those possibilities can produce deceptively similar graphs. A reader who knows only the highest number is not ready to identify a plume or an emitter. A reader who checks instrumentation and context can describe the evidence and uncertainty without either dismissing or exaggerating the observation.
This article teaches the physical principle, hydrocarbon interference, zero and span response, calibration traceability, unit conversions, aggregation, failure diagnosis and independent checks. Worked examples use invented teaching data explicitly, never claimed field observations. You will practise distinguishing the instrument’s response from the environmental interpretation and learn how to preserve that distinction when communicating a result.
What UV fluorescence actually detects
Sulfur dioxide molecules absorb ultraviolet energy and emit characteristic fluorescence. An ambient UV fluorescence analyser draws a sample through a measurement cell, illuminates it and measures emitted light through an optical detection system. Calibration relates the detector response to the concentration of a known sulfur dioxide reference gas. This is not a visual count of molecules, and the detector does not independently recognise the place or process that released them.
US regulation 40 CFR Part 50 Appendix A-1 specifies UV fluorescence as a continuous reference measurement principle. It describes typical excitation approximately between 190 and 230 nanometres and fluorescence detection in a broad band around 320 nanometres. These are method characteristics, not a recipe for deriving a concentration from light wavelength alone. Optical geometry, background subtraction, detection sensitivity and gas calibration remain necessary.
Do not confuse fluorescence with ultraviolet absorption. An absorption instrument measures how much selected incoming radiation has been removed; a fluorescence instrument detects light emitted after excitation. The distinction matters when diagnosing interference, validating a measurement and comparing methods. A particulate light-scattering sensor uses yet another interaction, and the shared word optical does not make their outputs interchangeable. Translate each instrument by its own sensing mechanism.
The sample path is part of the measurement
Follow the air from inlet to analyser. Tubing, filters, valves, manifolds, pumps, flow controls and the cell can all affect whether the sampled air represents the intended outdoor location. Inlet blockage, moisture, unexpected leaks or a changed tube connection can alter response while leaving the digital display apparently normal. Look up the actual station configuration, since different networks and analyser models implement different sampling trains.
Residence time and synchronization deserve attention. A change outside the building may appear at the instrument only after air moves through the sample path and is processed by the instrument. If a one-minute pollutant value is compared with a wind record that stamps the interval differently, apparent simultaneity may be illusory. Record whether timestamps denote interval beginnings, endings or midpoints and whether clock conventions are local or UTC.
Maintenance is not merely a note after the data. A filter replacement, probe change, calibration-gas cycle, pump service or relocation can divide a time series into comparability periods. When a step-change occurs near a maintenance event, test both possibilities: a real change in ambient air and a measurement response to the intervention. Source attribution becomes premature until the conditions of measurement are understood.
Hydrocarbon interference and scrubber logic
UV fluorescence is selective but not magically immune to the other molecules present in air. Aromatic hydrocarbons can interfere with the optical response, which is why reference-method descriptions discuss controls to reduce interfering species. Specific analysers use hydrocarbon-removal systems or other arrangements. Confirm the actual manufacturer’s design before writing that every station has the same scrubber or a particular scheduled replacement interval.
If elevated values diminish following replacement of a worn hydrocarbon scrubber, interference is a plausible hypothesis, not a proven account of every earlier point. Weather might have changed, gas checks may have occurred, or a technician may have altered other parts of the path. Compare operational logs, quality-check performance and collocated instruments around the intervention. The temporal coincidence is evidence worth testing rather than a complete causal demonstration.
Conversely, the existence of a recently serviced scrubber does not guarantee absolute interference-free readings. Maintenance records document actions, not universal performance. Ask what diagnostic checks confirm the device is functioning under observed conditions. Avoid the opposite mistake of declaring an entire high-SO₂ event false merely because an interference pathway exists. Measurement claims become stronger by testing rival explanations, not by choosing the most convenient one.
Zero readings and baseline drift
A zero check introduces appropriate zero air and measures the analyser’s baseline response. If an indicated zero concentration gives a nonzero instrument reading, this may reveal offset, contamination or another problem in the zero delivery or measurement chain. That discrepancy needs assessment against a documented quality plan. Automatically subtracting a constant from all ambient readings may be inappropriate because the discrepancy could vary with time or operating conditions.
Suppose daily zero checks in a teaching dataset are 0.0, −0.1, +0.1, +0.2 and +1.3 ppb. The last reading differs markedly from earlier ones. The relevant response is not to average the five into a seemingly benign offset. Check the zero-air source, valve state, tubing, flow, analyser diagnostics and maintenance. Decide what records are affected according to the specified procedure, and document the reason.
Negative ambient numbers can appear when a near-zero instrumental response is subjected to baseline subtraction or noise. The atmosphere does not have negative molecular concentration, but an indicated −0.3 ppb can be a legitimate raw or processed numerical outcome. Retain relevant flags and established network rules. Replacing every negative result with zero without justification biases low-concentration averages upward and erases evidence about baseline behaviour.
Span checks versus calibration
A span check introduces a defined nonzero reference concentration and tests whether the analyser responds sufficiently close to the expected value. Calibration establishes or updates a response relationship; verification checks whether an existing relationship remains acceptable. A passed check is evidence of response at a particular time and level. It cannot certify every earlier minute, every interfering species or all sample-path conditions.
For example, a nominal 40.0 ppb challenge yields 38.8 ppb. Signed error is 38.8 − 40.0 = −1.2 ppb, and relative error is (−1.2/40.0) × 100 = −3.0 percent. These are arithmetic properties of the teaching numbers, not an automatic pass or failure. The decision must follow the controlling acceptance criterion and whether that criterion applies to the specific concentration range.
Suppose a subsequent 40.0 ppb challenge reads 42.0 ppb. This is a +5.0 percent error rather than another negative deviation. Both magnitude and direction matter for drift diagnosis. Plot the signed responses over time and annotate maintenance and calibration events. An average absolute error can mask a systematic response trend; a long-term signed history helps distinguish gradual drift from isolated procedural disturbances.
Traceability reaches beyond the cylinder label. Delivery depends on the reference preparation, flow control, dilution where used, tubing integrity and stability. The regulatory method refers to NIST-traceable SO₂ standard concentrations. An accurately certified source does not automatically mean the gas delivered to the measurement cell has that concentration if a leak, incorrect dilution setting or contaminated line intervenes.
Reconstruct an illustrative monitoring hour
Consider twelve equal five-minute SO₂ observations in ppb: 2, 2, 3, 3, 4, 5, 8, 14, 11, 6, 4, 2. The total is 64 and the arithmetic mean is 64/12, or approximately 5.33 ppb, provided all twelve intervals are valid and equally weighted. The five-minute maximum is 14 ppb. Reporting 14 as an hourly average is a basic but consequential change in meaning.
If a gas-span operation occurred during the 14-ppb interval, that interval might not represent ambient measurement. Exclude it or otherwise dispose of it under the data-quality protocol, then calculate both the remaining mean and whether the official hour retains sufficient valid coverage. A mathematically calculated eleven-interval value does not automatically qualify as a compliant full-hour report.
A wind-direction change close to the spike suggests a transport hypothesis, not a uniquely identified source. Test whether wind speed, location, time synchronization and external measurements support the proposed interpretation. Even if the record proves a real ambient event, a single monitoring station cannot normally identify a particular source without additional independent evidence. A useful conclusion specifies that boundary.
Units and reference conditions
A parts-per-billion gas concentration and a microgram-per-cubic-metre concentration are not numerically equivalent. For ideal gas at about 25°C and one atmosphere, a commonly used conversion is SO₂ µg/m³ ≈ ppb × 64.07/24.45. Thus 10 ppb is about 26.2 µg/m³ under those stated conditions. The molar-volume denominator changes with reference temperature and pressure conventions.
If two reporting sites publish different mass concentrations for the same ppb measurement, inspect their conversion conditions before accusing either of a mistake. A network may use standard reference rather than actual inlet conditions. Conversion should be documented with units and sensible significant figures. A spreadsheet output of 26.204499 is not evidence that a field analyser measured to six decimal places.
For near-zero readings, examine reporting limits and uncertainty. Below a detection or reporting threshold does not mean there are no SO₂ molecules present. It means the signal cannot support a particular quantified claim under the specified method and conditions. Preserve censoring qualifiers rather than inventing precise zeros. If limits change over time, comparing substituted low values can distort seasonal trends.
Why averaging windows matter
An hour with twelve readings of 5 ppb and an hour alternating six zeros with six readings of 10 ppb have the same mean: 5 ppb. Their temporal patterns differ, and neither can be reconstructed from the hourly mean alone. Short peaks are valuable for studying episodic events, but response time, transient checks and low-end noise can affect short-window interpretation.
An official daily or longer average depends on coverage rules, time alignment and data validation. The arithmetic mean of whatever values survived may fail formal completeness requirements. Report whether an observation is validated, provisional or illustrative, and identify the period it represents. An attractive plotted curve is no substitute for the network’s data-quality specifications.
How to communicate a defensible result
An overconfident sentence says the SO₂ machine proved a factory created a plume. A precise sentence says the monitor estimated a five-minute SO₂ value of 14 ppb at a specified station, its quality checks indicated the value was or was not valid ambient data, and the source was not independently established. That phrasing is not weaker science; it tells the reader exactly what was observed and what remains uncertain.
Related instrument guides reinforce the need to distinguish methods: NO₂ chemiluminescence converter bias, ozone UV photometry and PM₂.₅ beta-attenuation monitoring answer different measurement questions. The transferable skill is to translate sensor physics, calibration, time and uncertainty into a clear claim that can be checked.
Further diagnosis: distinguish an instrument problem from a real change
Imagine a rise at two neighbouring monitors during the same afternoon. The first station records a substantial concentration increase, whereas the second reports a smaller but aligned change. Shared timing supports an environmental explanation, but it does not automatically rule out calibration differences or spatial gradients. Compare whether their observation windows actually overlap, whether each instrument meets independent checks, and how their inlet heights and distances from traffic or industrial activity differ.
Now imagine that only one instrument shows a sudden plateau, remaining almost unchanged through shifts in wind and temperature while its sample flow alarm occasionally activates. Such behaviour deserves attention, yet a flat line is not proof of malfunction: stable ambient conditions can also produce stable readings. Determine whether the plateau corresponds to a diagnostic alarm, communication freeze or value-repetition rule in the data acquisition system. A reliable analyst tests the instrument pathway and the plausible atmosphere side by side.
A particularly revealing pattern is an apparent event that repeats at precisely the same clock time each day. It could reflect a genuine scheduled emission or routine transport pattern, but it could also be an automated zero/span cycle inadvertently included in ambient observations. Check instrument mode logs, calibration schedules, inlet switching and time-zone handling before treating the recurrence as evidence of a human activity. A graph’s regularity is a question to investigate, not a mechanism by itself.
Maintenance records may state that the unit was cleaned, recalibrated and placed back in service at 13:00. This does not determine whether all data before that time should be invalidated. Review the nature of the defect, checks bracketing the affected period, the timestamps of any confirmed errors and the programme’s requirements. Some problems only affect a short interval. Other problems undermine the period since the last satisfactory independent check. The correct disposition follows evidence rather than the convenience of whole-day deletion.
Diagnostic case: zero, span and ambient traces together
Consider a teaching record with stable ambient readings near 3 ppb until noon, zero checks near 0.1 ppb and a 20 ppb span response of 20.1 ppb. At 12:10 ambient values suddenly rise to 18 ppb; the zero remains 0.1 and the span remains 20.0. These checks support stable response to two challenge conditions around the event, although they cannot rule out every interference or sample-path issue. Meteorology and independent stations now become important evidence.
In a contrasting record, the ambient trace rises steadily from 2 to 9 ppb, the zero response drifts from 0.1 to 3.4 ppb, and the span response also rises. The simultaneous instrumental trend weakens a straightforward interpretation as an environmental increase. Do not calculate a corrected ambient concentration by blindly subtracting the last zero; first establish the calibration model, affected periods, uncertainty and network rules for data correction or rejection.
In a third record, span response fails while zero response remains stable. This may indicate a change in sensitivity rather than a simple additive offset, but diagnosis requires more than the sign of one check. The challenge-gas delivery, optical path, contamination, calibration factors and reference standard each warrant investigation. A two-point response difference helps narrow possible errors; it does not automatically pinpoint the defective component.
Try a two-point numerical illustration: a zero source gives 0.4 indicated ppb and a 50.0 ppb span source gives 48.4 indicated ppb. The observed response difference is 48.0 units for a 50.0 ppb concentration difference, yielding an illustrative slope of 0.96 indicated ppb per true ppb. A single additive offset cannot fully describe that relationship. Actual instruments may use more elaborate corrections; the example teaches how offset and gain differ.
If an ambient indication is 10.0 under that simplified response model, subtracting the 0.4 offset and dividing by 0.96 gives an estimated 10.0 ppb. The convenient equality is an accident of these teaching numbers, not a general pattern. The more important exercise is to write the assumed equation, state the two calibration points and verify units before applying a correction to anything. An unjustified mathematical fix may make a compromised dataset look falsely authoritative.
Quality assurance as a chain of independent evidence
A quality control check differs from independent assurance. The same analyser can run a useful routine zero/span sequence, but an external performance audit or separately prepared reference can test aspects that the instrument’s own recurring routine might miss. The EPA’s air monitoring quality programmes distinguish quality assurance procedures, performance evaluations and technical systems audits. Read those categories as complementary tests of equipment, procedures and recorded data.
An audit report should clarify which reference was used, how the challenge was delivered and whether its procedure was independent of the routine calibration. If both routine and audit checks use the same flawed dilution arrangement, agreement between them may not reveal the shared error. Independence is a property of how evidence was generated, not merely a label attached to a second column.
A control chart can show whether check deviations remain within the programme’s prescribed boundaries over time. It should plot the signed response, time and relevant intervention notes. An unexplained change point deserves investigation even if isolated values remain below a formal threshold. However, control limits cannot be invented from one example or borrowed casually between instruments. Their meaning depends on the instrument, method and approved monitoring programme.
Treat data validation as a documented decision with reasons. A reviewer may flag a value as suspect, subsequently accept it after checking records, or invalidate it after finding a calibration mode or flow failure. Provisional and final data should be distinguishable. Silence in a data field should not be taken as proof that no review was needed. Retain raw values, flags and the history of changes so a later reader can reconstruct the judgment.
Practical completeness calculations
Suppose a one-hour record consists of twelve five-minute intervals but only nine are valid ambient data. The numerical coverage is 9/12 = 75 percent. Whether a reported hourly average qualifies depends on the governing programme’s completeness requirement. Without that requirement, do not call the hour regulatory-valid or invalid merely because 75 percent sounds sufficient. The reviewer can still calculate an exploratory average, properly labelled as incomplete.
For example, nine retained intervals sum to 45 ppb-observations. Their equal-weight mean is 45/9 = 5 ppb, which accurately describes the retained intervals, not necessarily the entire hour. If the three missing intervals were during an event, the full-hour mean might differ materially. A transparent report communicates both the numerical average and missing coverage rather than disguising the gaps within one displayed number.
If data are sampled every few seconds and then rolled up, count valid contributing samples under the instrument’s and network’s aggregation rules. A nominal five-minute observation can itself be incomplete due to communications loss, calibration or instrument downtime. Completeness therefore has layers: raw sampling, short averages and final reporting windows. Each layer needs a stated denominator and inclusion rule.
A station with fewer missing hours this year than last year may have a more complete record, but its annual mean can still be incomparable if the missing hours in the earlier year coincided with unusual emissions or meteorological conditions. Assess whether missingness is related to the outcome. Blindly comparing means from different coverage patterns may create the appearance of improvement or deterioration where confidence should be lower.
Interpret trends without overstating the precision
Suppose monthly average SO₂ values fall from 4.2 to 3.8 ppb. The raw difference is −0.4 ppb, about −9.5 percent relative to 4.2. Whether this supports a reliable environmental change depends on sampling coverage, instrument comparability, uncertainty, seasonal patterns and meteorological variation. Percentage change is easy arithmetic, but statistical and causal inference require more information than two rounded monthly values.
A high day in one year should not be compared with a yearly average in another year as though both represent the same exposure statistic. Preserve the averaging interval and summary definition, including whether the reported maximum is the highest valid hour, the highest short reading or a percentile under a particular regulatory convention. Select the statistic to answer a legitimate question, rather than choosing the most dramatic number.
When a site relocates a short distance, consider changes in height, building wake, shielding, road proximity and airflow. The instrument may be equally well calibrated before and after the move, yet it samples a different microenvironment. Trend continuity requires evidence about site representativeness as well as calibration. A location change belongs in the figure caption whenever it can materially affect the interpretation.
Gas concentration and atmospheric exposure are different questions
An ambient station offers a time-resolved measure at its location, while human exposure depends on where people move, how much time they spend in each microenvironment and conditions indoors and outdoors. A monitor cannot reconstruct a commuter’s inhaled concentration throughout a day from one station’s hourly average. It can still be valuable for environmental comparisons and episodes, provided its representativeness is described honestly.
Health or regulatory claims also require appropriate jurisdiction-specific limits and precise averaging conventions. This article does not supply a universal pass/fail threshold because no single value answers every local law, monitoring purpose or year of operation. When making a comparison, cite the applicable authority, pollutant form, period, completeness criterion and effective date. The words exceeds a standard demand more care than seems obvious.
Distinguish an instrumental peak from sustained exposure in a public communication. A brief spike may be important for understanding emission events, but a reading with a five-minute interval cannot be treated automatically as an eight-hour or daily concentration. Conversely, a modest daily mean can mask short higher values. Both scales may be useful; each answers a different question. Never silently swap scales to strengthen a claim.
Independent exercises to test understanding
Exercise one: zero response is +0.6 ppb, and an ambient reading is +0.5 ppb. Does that prove sulfur dioxide is absent? No. It indicates the ambient value is near the observed zero response, which calls for method-specific low-end interpretation and uncertainty. It may not support resolving a precise positive concentration, but physical absence is a stronger claim than the record provides.
Exercise two: a five-minute value is 22 ppb, and the other eleven five-minute values are 2 ppb each. The sum is 22 + 11×2 = 44; the equal-weight hourly mean is 44/12 ≈ 3.67 ppb. A report that says hourly SO₂ was 22 ppb would be wrong. A report that says the maximum five-minute interval was 22 ppb may be right if the 22-ppb interval is valid ambient data.
Exercise three: a 10.0 ppb zero-corrected span challenge is reported as 10.8 ppb. Signed error is +0.8 ppb and relative error is +8.0 percent. Without the actual acceptance limit, is it acceptable? The calculation alone cannot decide. Also examine reference concentration uncertainty, delivery quality and whether the measurement was stable during the check.
Exercise four: Station A reports one-minute SO₂ observations, Station B reports hourly means and Station C reports twenty-four-hour summaries. Can you rank their highest displayed points and call the ranking meaningful? No. Standardise comparable time windows, units and valid-data criteria first. Different display intervals alone can change the maxima and relative appearance of volatility.
Exercise five: an SO₂ increase correlates with a wind shift and with NO₂ at one station. Is it established that both gases came from one source? No. The observations support co-occurrence and a possible transport relationship, but source allocation needs additional evidence. Atmospheric processes, multiple emitting sources, separate instrument issues and local dispersion could contribute to the association.
Exercise six: three five-minute records were invalidated in a one-hour interval, leaving nine readings. If those nine sum to 63 ppb-observations, the mean of available data is 63/9 = 7 ppb and coverage is 9/12 = 75 percent. Report both figures, and reserve the label valid official hour for a result that passes the programme’s completeness rules.
Exercise seven: a dashboard shows 12 ppb SO₂ at about 25°C and one atmosphere. Apply the approximate factor 64.07/24.45 to obtain 31.45 µg/m³. The unit is mass per volume, not mass per mole, and the conversion is tied to reference temperature and pressure. If the station uses another reporting convention, re-evaluate the conversion rather than silently borrowing this one.
Exercise eight: a monthly mean falls from 6.0 to 5.4 ppb. The absolute change is −0.6 ppb and relative change is −10 percent. Does that prove an emission-control programme worked? No. Assess comparability, time variation, weather, missingness and independent evidence before claiming causation. A percentage difference is a description of two summaries, not a controlled attribution study.
Frequently asked questions that matter in practice
Why does the analyser sometimes show negative values? Baseline subtraction and electronic or optical noise can generate negative reported values near zero. They do not describe negative atmospheric concentration, but indiscriminately resetting them to zero alters statistical properties. Maintain the raw record, quality flags and accepted processing conventions, especially when calculating a low annual background.
Does a scrubber remove sulfur dioxide as well as interfering hydrocarbons? Its performance must be judged by the design and validation for the instrument. Do not infer that every scrubber either perfectly preserves SO₂ or necessarily removes it. Manufacturer instructions and suitable tests define intended behaviour. The relevant issue is how the actual sampling system influences the concentration reaching the measurement cell.
What if the daily zero and span checks both pass but a peak looks impossible? Check sample flow, inlet activity, diagnostic logs, the timing of the instrument mode and independent measurements. Passing checks strengthens confidence in defined aspects of response; it does not establish that every sample path, interference and data transformation was correct throughout the day.
What does traceable calibration mean? The reference concentration is connected through an appropriate documented chain to recognised standards, with the relevant measurement uncertainties and handling. It does not mean that all readings from the calibrated instrument are exact. Calibration quality, operating conditions, sample handling and data-processing integrity remain separate responsibilities.
Can I convert every SO₂ reading from ppb to µg/m³ using one number? Only when a common agreed reference condition makes that legitimate for the intended dataset. The conversion follows gas molar mass and molar volume, and the latter depends on temperature and pressure conventions. State the conditions and avoid false precision.
How should a journalist describe a one-hour SO₂ mean without knowing the source? Name the location, the hour’s interval, the pollutant, unit, reported concentration and validation state. If source attribution was not performed, say the reading alone cannot identify a specific emitter. Reporting limits, meteorology and instrument data may provide context without supporting a categorical causal claim.
When should readers suspect a formatting rather than a measurement error? When concentration units shift between tables, timestamps appear displaced, identical records repeat unexpectedly, or an hourly graph has more or fewer intervals than expected. Verify the raw file schema, time-zone conventions and transformations. A correct analyser can be misrepresented by a spreadsheet that silently converts date strings, drops missing rows or rounds values.
Turn the method into transferable learning
A method reader does not simply memorise that sulfur dioxide is detected by fluorescence. The reader asks a sequence of testable questions: Which physical interaction generated the signal? What reference defined the calibration scale? What other compounds or sampling conditions could alter it? What quality checks were made? What spatial and temporal claim does the valid result allow? This reasoning can be reused for many technical instruments.
Compare the teaching logic with the site’s optical PM₂.₅ sensor discussion. Light interaction, moisture, calibration and representativeness can all matter, yet the target physical quantity and calibration are different. Learning the differences builds understanding; treating all optical devices as one generic sensor obscures exactly what students need to know.
The strongest final explanation is sometimes deliberately modest. It identifies a validated SO₂ estimate, shows its units and averaging window, explains which operational checks support it, and marks its unresolved source and exposure questions. Such an explanation respects the instrument and the reader. It is also more useful for future comparison because someone else can check the inference rather than merely trust the author’s confidence.
A compact audit trail for one difficult result
For a disputed SO₂ value, prepare a short evidence packet rather than a screenshot alone. Include the station identifier, sample-inlet description, time-zone convention, start and end of the interval, original raw concentration and unit, analyser mode, zero/span checks around the event, operating diagnostics, maintenance notes and the data-validation decision. If any part is unavailable, mark it missing rather than treating the gap as support for a preferred explanation. This packet lets another reviewer reproduce how the number became a conclusion.
Next attach meteorological context as a separate analysis layer. Wind direction and speed, temperature, rain, boundary-layer conditions and nearby independently monitored gases can help develop and test hypotheses about transport. But environmental context should not erase an identified calibration or sampling defect. Keep the measurement-quality decision and the atmospheric interpretation visibly distinct so readers can challenge each independently.
Finally write two summaries: one technical sentence preserving interval, concentration, uncertainty and validity, and one public sentence explaining what that result means in ordinary language. If the public sentence claims a cause, a legal exceedance or a personal health outcome that the evidence packet cannot support, revise the sentence. The aim of translation is not to reduce complexity by removing essential conditions; it is to communicate those conditions without forcing the reader to become an instrument engineer.
Primary methodological sources
40 CFR Part 50 Appendix A-1 — SO₂ reference measurement by ultraviolet fluorescence.