Purpose and Scope

This report is intended to assist USGS personnel and others in collecting field-based fluorescence data that target DOM, wastewater, and hydrocarbons that are consistent and comparable across studies and sites and to ensure that data meet high-quality standards by following tested and established protocols. For those who have selected fluorometers for continuous data collection in the field, this report provides guidance on deployment, maintenance and field operations, quality-assurance procedures, and data reporting. Additionally, appendix 1 provides further guidance on fDOM sensors, such as calibration criteria and procedures, guidance on correcting for interferences, and parameter and method codes. The goal is for additional appendixes to be added over time for additional sensor types. This report supplements, but does not replace, the manufacturers’ user manuals for individual sensors.

Related Information

This report is designed to supplement other USGS documents that describe the collection and reporting of water-quality data, including the deployment and use of continuous water-quality monitors. The following list briefly summarizes related USGS policies and publications, ordered from general to specific in terms of the guidance provided:

Future USGS policy updates or changes should be sought out before beginning new data collection efforts in order to ensure that the references used are up to date and that all current policies are being followed.

Principles of Fluorescence

Fluorescence occurs when an electron is in an excited state and absorbs light, then emits light as the electron returns to its former state. This emission of light can then be measured by a fluorometer. The intensity of emitted light as measured by a fluorometer in a water medium is directly proportional to the concentration of the constituent, according to the Beer-Lambert law (Lakowicz, 2006). The transmission of light through water is determined by the physical and chemical properties of the constituents in the water sample (fig. 1). Scattering is a process whereby light energy passing through the water column is redirected in multiple directions, primarily by particles. Scattering forms the basis for measurements such as turbidity, which will only be discussed in this report in the context of interferences. Absorption is a process by which light energy is converted into nonradiant energy by exciting an electron in a substance dissolved or suspended in the water column, such as DOM, from the ground state to an excited state. Absorption of light (attenuation) by particles or by solutes (commonly DOM) can occur, and either can be the dominant effect. Absorbed energy is subsequently released through multiple competing pathways that include thermal pathways, such as vibrational relaxation and nonradiative decay, and the near instantaneous release of light energy in the form of fluorescence. Fluoresced light is always re-emitted with less energy and therefore at longer wavelengths than the original light that was absorbed, a phenomenon known as the Stokes shift (fig. 2) (Stokes, 1852). For more detail on the chemistry and physics of fluorescence and absorbance, numerous reviews include those by Lakowicz (2006) and Coble and others (2014).

The overall mixture of substances dissolved in water determines the fluorescent response of excitation-emission pairs across the UV and visible spectra, and this information is used to generate three-dimensional EEM plots (fig. 3). DOM, hydrocarbons, wastewater indicators such as optical brighteners, and even some dyes (that is, fluorescein) all fluoresce in the shorter wavelengths (excitation 240–500 nanometers [nm] and emission 250–600 nm). Because this document focuses on sensors that target these substances, the figures presented henceforth focus on the area defined by these wavelength ranges (fig. 3). Notably, commonly used red dyes, such as rhodamine, do not fluoresce within this region of focus.

Application of Fluorescence Spectrometry to Water-Quality Studies

The ability of a fluorophore to convert absorbed light into fluorescence depends on the molecular structure of the dissolved substances in the water, including the composition of the dissolved fraction of the natural organic matter (Aiken, 2014). The resulting fluorescence EEM for a water sample represents the complex mixture of fluorophores within the water, representing the origin and processing of the dissolved organic materials (Hansen and others, 2016, 2018). These regions of the EEMs are commonly denoted by letters referred to as fluorescence “peaks” within the fluorescence literature, as shown in figure 4. The characteristics of the identified EEM regions are described in table 1 and expressed in Raman-normalized intensity units (RU; Murphy and others, 2010).

Factors Affecting Fluorescence Measurements

Several physical and chemical properties of water and its environment can affect fluorescence measurements. Several properties of water that affect how much light and what wavelengths of light are able to pass through from the source to detector, such as suspended particles and absorbing substances like DOM, interfere with the fluorescence measurement because of “matrix effects” (figs. 5 and 6). Interferences from matrix effects are correctable as long as the signal-to-noise ratio remains sufficiently high to meet the sensitivity requirements for the site’s conditions. Other water properties, such as temperature, pH, and the presence of some dissolved metals, can quench fluorescence at the molecular level. Quenching from temperature effects is correctable, whereas the effects of pH and various metals are either not correctable or have not been sufficiently tested to determine whether corrections may be possible. The collective extent of these effects can reduce fluorescence measurements from a few percent to over 90 percent in the most extreme cases, but the loss of signal in most systems will typically be less than 20 percent (Downing and others, 2012). Further descriptions of these factors are provided next.

Wavelength Array

The excitation and emission bandpass filters define what fluorophores a sensor targets within the full array of excitation and emission wavelengths. For the purpose of this discussion, the range of wavelengths for a particular sensor is defined as the wavelength array. Manufacturers have developed a variety of sensors for specific applications, often balancing the importance of low detection limit with the need for specificity of fluorophores. For example, many sensors used in the ocean sciences often have large wavelength arrays to increase the fluorescence signal to lower detection limits, whereas sensors used for estuarine to riverine studies often have sufficient fluorescence to target specific fluorophore regions of the full excitation-emission array. Choosing the proper sensor for the intended application is critical to successful measurement in the field. The general “families” of sensors are summarized next. It is important to note that manufacturer specifications are often reported using different conventions and that specifications of bandpass and (or) filters may not reflect the actual range of effective excitation wavelengths. To ensure the most precise estimate of a sensor’s wavelength array, it is recommended that the user confirms the actual effective wavelength array with the manufacturers’ engineers or product managers.

Humic DOM Sensors

In the field fluorometer vernacular, fDOM (sometimes also written as FDOM or fdom) has most commonly been associated with a relatively small region of the EEM spectra near excitation 370 emission 460, which is most closely related to terrestrial, humic-like organic matter (McKnight and others, 2001;Stedmon and others, 2003;Coble and others, 2014). However, this region is not always the convention used across manufacturers for the reasons mentioned earlier, despite the similar use of the term “fDOM” or “CDOM” as a sensor name. Therefore, we recommend the adoption of a new naming convention that captures more specific information with respect to the region of the fluorescence spectra where the field fluorometers collect their measurements. For the greatest clarity and consistency, the parameter name will include the wavelength array described as the full-width half-maximum range of the excitation-emission channels. For example, the YSI EXO fDOM sensor is commonly used throughout the USGS community and the parameter name is “Fluorescence, water, in situ, excitation 360–370, emission 440–520, micrograms per liter as quinine sulfate equivalents (QSE).”

For data to be comparable, the same wavelength array must be used. To illustrate the variability, a laboratory benchtop instrument (an Aqualog 800) was used to compare the various wavelength arrays used by various field fluorometers in two standard reference materials (SRMs) and 76 environmental samples (collected from multiples States, including Pennsylvania, Colorado, Texas, California, Florida, and Iowa). Comparing the different wavelength arrays using the Aqualog 800 allowed the comparisons to be made with the potential interferences removed (Fleck and others, 2022). The first SRM was the International Humic Substances Society 2R101N equivalent to 4.0 milligrams per liter (mg/L). The second SRM was Pure Leaf brand unsweetened black tea diluted to 1 percent concentration (Hansen and others, 2018). The Aqualog 800 readings are reported in RU and then converted to QSE (Fleck and others, 2022). The International Humic Substances Society 2R101N, equivalent to 4.0 mg/L, was analyzed 15 times on the Aqualog 800 for each wavelength array, and the equivalent field fluorometer outputs ranged from 26.1 to 82.3 QSE. The tea SRM was analyzed 47 times on the Aqualog 800, and readings for the different wavelength arrays ranged from 4.2 to 8 QSE.

Additionally, the Aqualog 800 took readings of the 76 environmental samples using the equivalent wavelength arrays from field fluorometers and the output values for some arrays were more than double the output values for other arrays, illustrating the potential differences among sensor outputs (table 2).

Table 2.    

Comparisons from an Aqualog 800 standardized to quinine sulfate equivalents (QSE) with excitation (ex) and emissions (em) equivalent to fluorescence of dissolved organic matter (fDOM) sensors from multiple manufacturers.

[Data summarized from Fleck and others (2022). Data are provided for two standard reference materials (SRMs) and the mean, minimum, and maximum from 76 environmental samples. No replicates were collected for environmental samples; therefore, the relative standard deviation is not available. IHSS, International Humic Substances Society; mg/L, milligram per liter; DOC, dissolved organic carbon; N, sample size; %, percent; CDOM, colored dissolved organic matter]

Table 2.    Comparisons from an Aqualog 800 standardized to quinine sulfate equivalents (QSE) with excitation (ex) and emissions (em) equivalent to fluorescence of dissolved organic matter (fDOM) sensors from multiple manufacturers.

Equivalent wavelength arrays set on the Aqualog 800 to simulate the wavelength arrays from various field fluorometers (equivalent sensor name in parentheses)	Aqualog 800 mean value in QSE, IHSS 2R101N equivalent to 4.0 mg/L DOC; N=14	Relative standard deviation (percent)	Aqualog 800, mean value in QSE, tea SRM (Hansen and others, 2018); N=51	Relative standard deviation (percent)	Aqualog 800 value in QSE, environmental samples; N=76
					Mean	Minimum	Maximum
Fluorescence, field, excitation 360–370 and emission 440–520 (YSI EXO fDOM)	48.4	4.86	5.7	4.36	56.2	5.4	369.3
Fluorescence, field, excitation 355–375 and emission 440–500 (Turner Designs CDOM)	45.5	4.87	5.6	4.34	54	5.3	351.1
Fluorescence, field, excitation 365–375 and emission 400–520 (Sea-Bird/Wetlabs fDOM/CDOM)	55.6	4.87	6.3	4.97	71.3	7.2	446.7
Fluorescence, field, excitation 360–390 and emission 455–530 (in situ fDOM)	182.4	5.24	18.1	5.05	195.5	19.1	1632
Fluorescence, field, excitation 364–376 and emission 420–460 (Seapoint CDOM)	45.8	4.82	5.6	5.10	64.2	6.6	385.5
Fluorescence, field, excitation 240–270 and emission 422–478 (Chelsea Technologies Group CDOM)	226.3	4.55	24.3	5.53	234.8	23.7	2259.2

¹

Denotes the high for each statistic.

²

Denotes the low for each statistic.

Different manufacturers use a wide range of wavelength arrays under the fDOM and CDOM sensor names, but all measure various portions of the humic regions of the EEM spectra (fig. 7). The various arrays reflect the intended marketable application for each sensor rather than any physically or chemically based naming convention. Most manufacturers target the “peak C” region of fluorescence, while at least one specifically targets the “peak A” region (table 1, fig. 4). The relative contribution of different humic substances to the overall fluorescence of DOM can vary among locations and seasons, depending on the environmental setting. This variability contributes to variation among measurements performed by the different sensors, despite the sensors having a common name. The C:A ratio has even been used as an indicator of DOM composition and source tracing (table 3; Moran and others, 2000; Hansen and others, 2016), highlighting the need to specify wavelength arrays rather than using the sensor name in the parameter name. When selecting a sensor for deployment, consideration should be given to the location and specificity of the wavelength arrays and their relation to the project objectives.

Table 3.    

Commonly used excitation-emission matrix (EEM) ratios for dissolved organic matter characterization.

[ex, excitation; em, emission; DOM, dissolved organic matter; nm, nanometer; >, greater than]

Table 3.    Commonly used excitation-emission matrix (EEM) ratios for dissolved organic matter characterization.

Peak ratio	Calculation	Purpose	Reference
A:T	The ratio of peak A (ex260em450) to peak T (ex275em340) intensity	An indication of the amount of humic-like (recalcitrant) versus fresh-like (labile) fluorescence in a sample	Hansen and others (2016, 2018)
C:A	The ratio of peak C (ex340em440) to peak A (ex260em450) intensity	An indication of the relative amount of photosensitive humic-like DOM fluorescence in a sample	Moran and others (2000); Hansen and others (2016)
C:M	The ratio of peak C (ex340em440) to peak M (ex300em390) intensity	An indication of the amount of diagenetically altered (blue shifted) fluorescence in a sample	Coble (1996); Burdige and others (2004); Para and others (2010); Helms and others (2013)
C:T	The ratio of peak C (ex340em440) to peak T (ex275em340) intensity	An indication of the amount of humic-like (recalcitrant) versus fresh-like (labile) fluorescence in a sample	Baker and others (2008)
Fluorescence index (FI)	The ratio of fluorescence intensity at em wavelengths at 470 nm and 520 nm, obtained at ex 370 nm	Shown to identify the relative contribution of terrestrial and microbial sources to the DOM pool	McKnight and others (2001); Cory and others (2010)
Humification index (HIX)	The sum of fluorescence intensities from em wavelengths 435–480 nm divided by the sum of fluorescence intensities from em wavelengths 300–345nm and 435–480 nm obtained at ex 254 nm	An indicator of humic substance content or extent of humification. Higher values indicate an increasing degree of humification.	Ohno (2002)
Freshness index (b:a)	The fluorescence intensity at em 380 nm divided by the maximum fluorescence intensity between em 420 nm and 435 nm, obtained at ex 310 nm	An indicator of recently produced DOM, with higher values representing a higher proportion of fresh DOM	Parlanti and others (2000); Wilson and Xenopoulos (2009)
Biological index (BIX)	The ratio of fluorescence intensity at em 380 nm and 430 nm, obtained at ex 310 nm	An indicator of autotrophic productivity. High values (>1) correspond to recently produced DOM of autochthonous origin.	Huguet and others (2009)

Wastewater Sensors

Wastewater sensors typically target contributions from publicly owned treatment work effluents and (or) septic discharges and leaks rather than industrial wastewater effluents (addressed in the following section). These sensors typically target fluorescence of microbial byproducts that fluoresce near the T and B peaks (fig. 4) of the fluorescence spectra (excitation near 275 nm, emission 300–350 nm) or with the fluorescence associated with optical brighteners (for example, fluorescent whitening agents) commonly used in laundry detergents and the textile and paper/pulp industries (excitation 340–380 nm, emission 400–480 nm) (fig. 8).

Hydrocarbon Sensors

The hydrocarbon class of sensors typically target industrial effluents including hydrocarbons, oil spills, and solvents. The sensors used for petroleum and its various byproducts often use large wavelength arrays because they are typically used as a screening tool for spills and plume characterization and thus need a wide response range to detect a wide range of substances (fig. 9). With the expansion of hydraulic fracturing and more widespread use of hydraulic fluids in oil and gas exploration and recovery, more specific sensors are likely to be developed to discriminate sources and specific contaminants.

Configuration

The overall design of commercial fluorometers generally follows one of two physical configurations. The first is an open-face design in which optical windows are located on the outside of the sensor body, and fluorescence is measured in a sample volume external to the sensor housing. The second is a flow-through design, which uses a quartz tube within the sensor body to measure a fixed sample volume as it passes through the sensor (fig. 10). The optical windows of the open-face sensors protect the light source and photodetector elements and are made from materials (typically sapphire) that minimize abrasion and degradation of the sensing windows. Most sensors used for freshwater monitoring are of the open-face design, which allows for easy integration into existing monitors and the incorporation of wipers to reduce biofouling. Flow-through systems are a good option when water filtration is desired, boat-based measurements are made, or other pumped-water operations are necessary.

For open-face sensors, the angle at which the excitation light source and the emission detector are placed relative to each other (typically 90–140 degrees) determines the sensing volume, which is the geometrically defined volume where the excitation light beam intersects with the detector’s cone of reception (fig. 10). For closed sensors, the path length is defined by the geometry of the unit. As the acceptance angle or source beam cone increases, the length of the sensing volume shortens and attenuation along the sample path decreases.

Field fluorometers can be configured as a single fluorescence sensor, with multiple fluorescence sensors within a single housing, or as part of a sonde that contains a variety of sensor types. All field fluorometers are equipped with an electronic signal processor that converts the raw voltage from a photodetector to a useful measurement output. Instruments can use a system of analog and digital electronic circuits, or a specialized microcontroller known as a digital signal processor to detect, convert, and output signals at very high internal sample frequencies, which are then integrated to output high-frequency data (from 1 to 20 hertz) via an analog or digital (RS–232) signal output. A microcontroller also controls the modulation of the LED so that the system can detect and differentiate between fluorophores and background noise (for example, photodetector dark current).

Multisensor Monitors

Some multisensor monitors allow for the simultaneous deployment of multiple sensors, allowing for simultaneous measurements to be made in different regions of the EEM to better isolate specific substances from a general increase in fluorophore concentration across the EEM spectra (table 2). If the desired application benefits from fluorophore specificity and ratios among those fluorophore regions (table 3), it may be best to select a monitor that is capable of multiple sensor deployment. The codeployment of turbidity and temperature sensors is also required in many applications. If the sensor is not correcting for variations in temperature on the fluorescence itself, a temperature sensor must be deployed. If independent testing has not been conducted to verify that the fluorescence signal is temperature compensated internally, an independent temperature sensor should be deployed with the fluorescence sensor as a best practice until temperature compensation can be evaluated. If particle interference alters sensor readings by more than or equal to the calibration criteria (fDOM calibration criteria listed in appendix 1), then the deployment of a turbidity sensor is required. Multisensor monitors capable of integrating temperature and turbidity with fluorometers within the same monitor may perform better than units where turbidity is measured meters away from the fluorometers, as turbidity can differ substantially within a site at that scale. It should be noted that independent optical sensors located too close to one another can interfere with optical sensor measurements.

Materials

Component materials for the instrument and sensor body differ among manufacturers and among different models from the same manufacturer, depending on applications like fresh to saline waters and various depth ratings. Instrument housings can be made from a variety of materials (for example, acetal resin, 316L stainless steel, polyoxymethylene, or titanium) that affect their resistance to corrosion and fouling, as well as their pressure and temperature rating.

Spectral filters are made of glass or other materials that filter out unwanted wavelengths from the light source and from the emitted light returning to the detector. Bandpass filters used in most field fluorometers are commonly described by the wavelength with peak transmittance, the center wavelength (the average of two half-height wavelengths in a spectrum), and the full width at half maximum (the wavelength range at half of the maximum transmittance) (fig. 2).

The stability, lifetime, and quantum yield of LEDs vary by wavelength (λ), and LEDs exciting at wavelengths used for nonphytopigments (λ approximately 240–370 nm) typically drift more at the lower range (less than [<] 280 nm) as compared to the higher range (greater than [>] 320 nm) (Talbott and Clifford, 2013).

System Integration

Most field fluorometers are designed to be integrated with external data collection platforms that provide power to the sensors and a data logger controller to record the output of the fluorometer. Field fluorometers equipped with digital output in the form of serial communication (for example, RS–232, RS–485, USB, or SDI–12) are common and desirable compared to analog output fluorometers for most continuous-monitoring deployments. Although there are many other protocols for serial communication, the majority of data collection platforms currently in use are designed to directly interface to RS–232, SDI–12, and RS–485 serial protocols. Data loggers and controllers may be integrated in some field fluorometers, particularly when the sensors are developed to operate a component of a multiparameter sonde and to transmit continuous real-time data. Proprietary or generic data transfer protocols may be used, depending on the manufacturer, to control the instruments and collect and transmit data.

Performance

Field fluorometers are typically characterized by four data-quality specifications: (1) detection limit, (2) dynamic range, (3) signal resolution, and (4) temporal resolution. In addition, the accuracy of field fluorometers is the degree of agreement between the sensor measured and fluorescence intensity in the absence of interferences, which is primarily affected by two important sources of uncertainty: (1) instrument noise and (2) matrix interferences within environmental waters. Although instrument noise caused by the electronic components, the light source, and the detector resolution is often random (Skoog and others, 2007), a systematic error can result in good precision but poor accuracy (that is, bias). Therefore, instruments with low electronic noise and a greater tolerance for interferences from suspended particles or dissolved organic matter have greater accuracy. Although modern instruments typically have a very good signal-to-noise ratio, the noise can become an issue as the desired detection limit decreases, in which case the sensor selected should have increased sensitivity. Although it is important to recognize that these specifications relate to the fluorescence measurements themselves, they do not reflect the correlation of a measurement to a laboratory-measured constituent such as DOC concentration.

Manufacturer specifications are often determined under controlled operating conditions, averaged for a limited number of sensors, and by using reference materials in solutions that typically result in greater accuracy, greater precision, and lower detection limits than would be observed if measured in natural waters where matrix interferences are a factor. Prior to deploying an instrument in the field, it is important to independently verify instrument performance in a controlled laboratory environment with standards and spikes of substances that would potentially cause matrix interferences. The baseline performance data collected prior to deployment serves as a basis for evaluating instrument performance after deployments and during instrument servicing.

The detection limit for fluorometers can be defined as the lowest value measurable by a given sensor that can be distinguished from a blank value with some level of statistical confidence, usually 99 percent (U.S. Environmental Protection Agency, 2017). Detection limits for field deployable fluorometers are a function of a variety of factors, including the wavelength array, integration interval, signal amplification, size of the optical window, size of the wavelength array, and the instrument electronic noise (for example, “signal to noise”). The detection limits of the current generation of field deployable fluorometers such as those listed in table 4 are very low, and those detection limits are typically suitable for measuring fDOM in surface waters. Nevertheless, lower detection limits may be needed for making accurate measurements in environments such as groundwater, oligotrophic lakes, coastal waters, and more pristine rivers and creeks. Generally, specificity for “peak” fluorescence signal intensity targeting specific fluorophores is balanced with the detection limit. Repeated measurements or instrument-specific calibrations can be used to determine a true detection limit above or below that reported by the manufacturer if needed to meet project requirements.

The dynamic range is defined as the difference between the lowest and greatest concentrations that can be measured, although the range over which the instrument response is linear (namely, the linear dynamic range) is often of greater interest. Linearity is typically reported by the manufacturer as the coefficient of determination (r²) or can be determined by users from a calibration curve across the dynamic range. The maximum detectable value is particularly important for fluorometers and is determined by numerous factors, including the chemical composition of the target constituent and the instrument’s sampling volume. Some sensors are equipped with adjustable gain settings that can be used to scale the linear range, which may be useful for specific applications at the low or high end of the dynamic range. Although field fluorometers have a manufacturer-reported linearity (typically r² > 0.99) that covers the range of values expected in most natural environments, it is important to note that the range may be assessed in colorless laboratory standards. Therefore, the linear range in natural waters may be lower than observed in laboratory testing because of inner-filter effects (IFE), which are the attenuation of light by colored dissolved organic matter and other dissolved substances that cause a loss of linearity in the sensor response and result in an underestimate of the true value.

The term “signal resolution” is defined as the minimum difference between measured values reliably detected by the sensor. With the current (2022) generation of field fluorometers, the resolution can typically be measured to three significant figures of the calibration solution’s concentration units.

The term “temporal resolution” is defined as the amount of time between truly independent measured values. Most sensors average their data over an unspecified interval—one that can exceed the reporting interval. For some sensors, such as the YSI EXO, this interval is selectable. The test for this is to bias the sensor and then see how long it takes for the sensor to return to a baseline value. This test can be performed by altering the concentration of a standard or SRM and determining the time it takes for the sensor to stabilize.

Considerations/Checklist for Sensor Selection

The instrument selected for continuous monitoring will depend on several considerations, including the following:

Many of these questions can be answered with historical water-quality data, site documentation, and data-quality objectives determined for the project. The user can then evaluate instrument needs, relative to manufacturer-stated or user-defined specifications, and relevant publications. It is important to note that field fluorescence measurements might simply be impractical in some systems during times when color, turbidity, or both are extremely high.

Verification of Standards

Reference sensors and benchtop fluorometers are useful in ensuring that the standard solutions are not made in error and can guide calibration verifications to determine when to calibrate and when to verify the standard solution (table 5). A laboratory reference sensor (of the same model as the field fluorometer) is recommended for any project that is not using a benchtop fluorometer to quality assure the standard solution. A laboratory reference sensor serves no other purpose than to provide a calibrated reference measurement. A well-documented quality-assurance plan, like that used to track National Institute of Standards and Technology thermometers, would need to be maintained to ensure that the laboratory reference sensor retains its accuracy and that its calibration history can be traced.

A reference sensor should

Another approach to verifying successful calibration is to compare field measurements from the fluorometer with laboratory measurements on a benchtop spectrophotometer set to the same wavelength array as the field fluorometer. Comparisons between laboratory and field measurements can also ensure that there is not a systematic error in the standard preparation, such as one caused by expired chemicals or LED degradation.

Fouling Corrections

Fouling corrections for fluorescence sensors account for reductions in transmittance of the excitation or emission light signals caused by biological growth, staining, or sediment accumulation on the optical windows of the sonde. For most sites and deployment configurations, the sensors can be kept clean using antifouling measures, such as wipers and copper components, and fouling corrections might not be needed. When fouling corrections are needed, they may be applied routinely, or they may be limited to periods when a wiper failed or when extreme biological fouling occurred that was not easily removed by mechanical action.

Data from sensors of different makes and models are not equivalent in environmental water. The field reference monitor should be the same make and model as the site monitor. Finally, careful evaluation of the field reference monitor readings against site monitor readings is needed because of the many confounding factors discussed earlier in this document. Differences in readings between the field reference monitor and site monitor might not be due to changes in environmental conditions, but instead might be caused by more localized factors like differences in light or suspended particles. Because of this, the field reference monitor readings should be used with caution in determination of fouling corrections, and if other water-quality measurements are recorded concurrently, they will provide an additional basis to evaluate whether conditions are meeting the criteria for “rapidly changing conditions” as defined by TM1–D3 (Wagner and others, 2006). Equations and guidance for applying fouling corrections can be found in TM1–D3, and are applicable to time-series data, assuming the guidance in the previous paragraphs is considered in their application.

Fouling corrections often take the form of either a simple arithmetic shift or a percentage correction, both of which are described in TM1–D3. In either case, a fouling correction can be prorated to slowly increase over the time elapsed since the last cleaning. Fouling corrections can be prorated starting on an obvious fouling event or held constant depending on hydrographer judgement and knowledge of observed site conditions. In a simple arithmetic correction, values are shifted up or down according to the magnitude of the computed correction, and this can be prorated or held constant based on user judgement.

Percentage corrections can be useful to force a zero correction at a zero value, thus avoiding the creation of negative values that sometimes can arise from the use of a simple arithmetic correction. Percentage corrections also can be subject to extrapolation problems if the sensor signal increases but tend to be more appropriate for sensor readings near zero. Any time a step change is observed in the data stream, field protocols should be reviewed. If the sensor was swapped, the step change could indicate a lack of sensor intercalibration or a sensor quality issue.

Fouling typically is assumed to accumulate linearly over time starting from the previous sensor cleaning, but in some instances the onset of fouling coincides with a noteworthy event such as a storm runoff that carries a large, suspended sediment load. Sensor diagnostic data or data statistics, if available, sometimes can be used to identify the onset of fouling. If the onset of fouling is identified from such data, fouling corrections can be applied from that moment onward rather than from the date of the previous service visit.

Drift Corrections

Fluorescence sensors may experience some variation in their response over time if changes occur in the performance of their light sources or detectors; however, as noted previously, fluorometer drift is often minimal because of the stability of the solid-state components. Calibration drift can be evaluated using the difference between the expected sensor response and the measured response in at least three standards, where one is as a zero solution (often OBW or DIW) and multiple known concentrations of a calibration standards such as QS. The magnitude of the calibration drift error measured in this way follows the approach documented in TM1–D3 (Wagner and others, 2006). Only readings in calibration standards should be used to determine drift corrections; SRM readings should be used to determine when a full calibration check is needed but should not be directly used to determine a drift correction. SRM readings can be useful in determining when a drift correction should begin.

As with fouling corrections, calibration-drift corrections can be implemented as a simple datum shift, as a percentage correction, or as a multipoint correction when nonlinear drift is observed in a laboratory. The calibration drift correction is interpolated linearly over time from the date and time of one sensor performance check to the date and time of the next check unless specific information is available to indicate a different date and time for the initiation of sensor drift.

Temperature, Turbidity, and Inner-Filter Effect Corrections

The measured output from a field fluorometer can be affected by the temperature, turbidity, and color in sample water, as described previously (see the “Factors Affecting Fluorescence Measurements” section). The effects of temperature, turbidity, and colored water are not the same for all fluorometers or monitored locations; however, interferences still occur with most field fluorometers in turbid and (or) colored waters. All sensors should be corrected for temperature, turbidity, and IFE (unless it can be demonstrated that turbidity and [or] IFE corrections are not necessary) for measurements to be comparable among sites and through time. The corrections can be substantial relative to the final corrected values (fig. 13). In the example provided in figure 13, the IFE correction is the largest correction to the data; however, this is only an example. Some sites may be influenced to a greater degree by turbidity or even temperature corrections.

Although the relative magnitude of interferences and temperature effects are known in standard solutions (Downing and others, 2012), standard corrections for these effects have not been established by the USGS or instrument manufacturers for the range of instrument designs, sediment sizes, and DOM types found in natural waters (Saraceno and others, 2017). Therefore, the user will need to characterize the sensor response, apply the appropriate corrections, and determine acceptable limits for corrections beyond those for fouling and drift. It is important to note that as the effects of turbidity and IFE increase, the practical dynamic range of the sensor decreases and can decrease to the point that the sensor has little or no response to an actual change in concentration. The accuracy of corrected values should be viewed with caution when accounting for 50 percent or more signal loss caused by interferences. Additional details on how to assess the effects of turbidity and IFE on fDOM sensors are described in appendix 1.

Temperature corrections for the phenomenon of fluorescence are needed to ensure sensor consistency and comparability over a wide range of temperatures. A consistent reference temperature should be used for corrections; 25 °C is recommended for use as the reference temperature across the USGS. Temperature corrections are potentially instrument and site specific and should be verified using native water from the monitoring location.

Turbidity corrections for fluorescence increase the sensor signal to account for the excitation-emission light energy scattered away by suspended particles. The turbidity correction for fluorescence sensor data is dependent upon the manufacturer and model of the fluorescence and turbidity sensors used. For any change in either sensor model, the correction should be reevaluated. At extremely elevated turbidities, the practical dynamic range of the sensor decreases and can decrease to the point that the sensor has little or no response to an actual change in concentration. Using theoretical temperature and turbidity corrections based on experimental data in published reports is acceptable; however, the theoretical equations should always be verified for individual sensors and site-specific conditions.

IFE cause field measurements to differ from true values and can vary substantially among sites. Previous efforts used absorption spectra to derive corrections using laboratory measurements (Downing and others, 2012); however, serial dilutions can provide a simpler, more direct correction factor for field-based sensor measurements. Readings in serial dilutions of native water with OBW have been used to accurately determine the IFE correction for fDOM sensors.

One approach to verifying corrections is to compare fully corrected sensor readings to concurrently collected discrete samples analyzed on a benchtop spectrophotometer set to the same wavelength array as the field sensor. Benchtop spectrophotometer measurements are performed on water after necessary filtration; dilutions can be performed to eliminate interferences from particles and IFE as well as temperature variations. A combination of serial dilutions and validation with discrete samples for laboratory analysis is recommended. Although the collection of discrete samples requires additional demand on time and effort, this approach can provide further insights into whether or not other interferences, such as pH and salinity, are affecting the fluorescence measurements in the field. Additional guidance for correcting fDOM for temperature, turbidity, and IFE is available in appendix 1.

Final Data Evaluation and Approval

Final data evaluations and approval for fluorescence measurements follow the general guidelines of TM1–D3 (Wagner and others, 2006) for analyzing, approving, and auditing continuous water-quality data.

The maximum allowable limit for corrections—defined as the maximum value for the absolute sum of all data corrections where data are reported, stored in the NWIS database, or both (Wagner and others, 2006)—has important implications for field sensors. If the limit is too large, the ability to interpret results is compromised; if the limit is too low, the costs of instrument servicing and field visits can be excessive. As for other field sensors (Wagner and others, 2006; Pellerin and others, 2013), six times the calibration criteria have been suggested as the maximum allowable limit for fouling and drift corrections of field fluorometers. Temperature compensation and corrections for interferences are not included here as part of the maximum allowable limit, as they are theoretical corrections for the phenomenon of fluorescence. The point at which an interference correction is no longer justified is not yet known, but caution is warranted when corrections of greater than 50 percent are needed. Data requiring corrections large enough to be deemed unacceptable for publication or storage can be flagged in NWIS and archived according to USGS science center quality-assurance plans.

Corrections for drift and fouling are applied to fluorescence data prior to corrections for temperature, turbidity, and IFE. A final corrected time series is derived from the drift- and fouling-corrected time series, often using an equation that combines the corrections for temperature, turbidity, and IFE. Table 7 summarizes the different types of corrections.

It is reasonable that fluorescence sensor data would be initially classified as category 3 under USGS Water Mission Area continuous records processing guidelines (U.S. Geological Survey, 2010), providing additional time to characterize site-specific confounding effects not otherwise accounted for by the instrument. Associated data may need to be stored and archived in support of data review (and possible future model development), such as high-frequency burst data, instrument diagnostic data, and laboratory quality-assurance data for the sensor calibrations. A category 1 designation (approval within 150 days) may become appropriate when the site conditions are well understood and field operations become routine.