Optical measurements using ultraviolet–visible (UV–VIS) spectrophotometric sensors and fluorescent dissolved organic matter (FDOM) sensors have recently been used as proxies of dissolved organic carbon (DOC) concentrations in streams and rivers at a high temporal resolution. Despite the merits of the sensors, temperature changes and particulate matter in water can interfere with the sensor readings, over- or underestimating DOC concentrations. However, little efforts have been made to compare responses of the two types of the sensors to critical interferences such as temperature and turbidity. The performance of a UV–VIS sensor and an FDOM sensor was compared in both laboratory experiments and in situ monitoring in a forest stream in Korea during three storm events. Although the UV–VIS sensor did not require temperature correction in laboratory experiments using the forest stream water, the deviations of its values from the DOC concentrations measured with a TOC analyzer increased linearly as turbidity increased. In contrast, the FDOM sensor outputs decreased significantly as temperature or turbidity increased, requiring temperature and turbidity correction for in situ monitoring of DOC concentrations. The results suggest that temperature correction is relatively straightforward but turbidity correction may not be simple because the attenuation of light by particles can significantly reduce the sensitivity of the sensors in highly turbid waters. Shifts in composition of fluorophores also need to be carefully tracked using periodically collected samples since light absorbance and fluorescence can vary as the concentrations of dominant fluorophores change.
Dissolved organic carbon (DOC), which is a dominant form of organic carbon in many streams and rivers, plays significant roles in aquatic systems. Riverine DOC is the energy source for heterotrophs (Raymond and Bauer, 2000), protects living organisms from UV light (Morris et al., 1995), and affects metal availability (Di Toro et al., 2001). High riverine DOC concentration ([DOC]) can also lower the quality of drinking water by increasing trihalomethane formation potential during water treatment (Hur et al., 2014; Xie, 2004). Thus, many studies on [DOC] have been conducted on a variety of spatial scales, such as streams draining from small watersheds to major rivers from large basins (Aitkenhead and McDowell, 2000; Jeong et al., 2012; Oh et al., 2013).
Studies on DOC release from forest ecosystems showed a close relationship
between carbon export and hydrology, indicating an important role of
discharge in DOC loads (Jeong et al., 2012; Pellerin et al., 2012; Raymond
and Saiers, 2010). Stream DOC load increased as water discharge increased,
and thus it was observed that DOC released during storm events accounted for
a substantial amount of total carbon export from an ecosystem (Hinton et al.,
1997; Raymond and Saiers, 2010; Yoon and Raymond, 2012). Considering that a
large variation in water discharge during heavy rainfall can result in
large variations in daily as well as annual DOC loads, monitoring stream
carbon concentrations with a high temporal resolution during storm events is
necessary (Jollymore et al., 2012). This is valid especially in Asian monsoon
regions, including South Korea (Kim et al., 2013), where more than 50 % of
annual precipitation (an average of 1320 mm from 1981 to 2010) is
concentrated during summer months (Korea Meteorological Administration,
Two types of optical sensors have been used frequently for this purpose; the ultraviolet–visible (UV–VIS) spectrophotometer (Etheridge et al., 2014; Jeong et al., 2012; Jollymore et al., 2012; Strohmeier et al., 2013) and the fluorescent dissolved organic matter (FDOM) sensor (Pellerin et al., 2012; Saraceno et al., 2009; Watras et al., 2011). UV–VIS sensors use the range of ultraviolet and visible light wavelengths (e.g., 220 to 720 nm) to rapidly scan the absorbance of UV–VIS light by molecules in the water and estimate the concentration of the molecules based on the Beer–Lambert law. Strong correlation between [DOC] and light absorption has been used to provide algorithms that convert UV–VIS absorbance to [DOC] (Jollymore et al., 2012). FDOM sensors measure the intensity of fluorophores, molecules absorbing UV light and reemitting light at longer wavelengths. Streams and rivers containing terrestrial DOC have many fluorophores, and, thus, FDOM sensors can be used as a proxy to monitor [DOC] in freshwater systems (Downing et al., 2012; Wilson et al., 2013).
Although the two types of sensors have been employed to monitor [DOC] in various systems, several factors, such as pH, turbidity, inorganic matter, and temperature, could limit the use of both sensors. While the effects of pH and inorganic materials (e.g., nitrate and iron) commonly observed in most natural watersheds are negligible (Weishaar et al., 2003), a change in water temperature and increased turbidity could reduce the accuracy of the sensor readings (Downing et al., 2012). Fluorescence decreases as temperature increases, which is known as thermal quenching (Watras et al., 2011), and particles significantly attenuate or interfere with the detection of UV–VIS and FDOM sensors (Downing et al., 2012; Jeong et al., 2012). While in situ fluorescence measurements in filtered stream water can provide a reliable proxy of stream [DOC] by overcoming interference due to particles, filter clogging has been reported to result in data loss during the later monitoring phase (Saraceno et al., 2009).
Although the UV–VIS and FDOM sensors have been used widely to estimate stream and river [DOC], to our knowledge, there is no study directly comparing the performance of the two types of sensors (Table 1). The sensors may have their own strengths and weaknesses as a proxy for monitoring stream [DOC], and, thus, the objective of this study is to compare the performance of UV–VIS and FDOM sensors as a proxy for [DOC] using laboratory experiments and in situ measurements in a temperate forest stream.
Previous studies using UV–VIS or FDOM sensors.
Laboratory experiments and in situ measurements were conducted with a
UV–VIS sensor (carbo::lyser™,
s::can Messtechnik GmbH, Austria) and an
FDOM sensor (cyclops-7, Turner Designs, USA). The UV–VIS sensor used in
this study has two beams for autocalibration and a 5 mm optical path length,
which is fitted to measurement ranges of 1–150 mg L
The FDOM sensor uses LED (light-emitting diode) as a light source, and the sensor uses the single excitation/emission pair, 325/470 nm, with 120 and 60 nm excitation/emission band pass, respectively. Fluorescence intensity was normalized with quinine sulfate standards and expressed as quinine sulfate equivalent (QSE) in parts per billion. Quinine sulfate standards from 0 to 100 ppb were prepared to calibrate the FDOM sensor by diluting 1000 ppm of quinine sulfate stock solution, which was made by dissolving 1.21 g of quinine sulfate dihydrates in 1 L of 0.5 M sulfuric acid.
A data logger (CR1000, Campbell science, USA) was used to collect the optical data of the sensors either every minute during laboratory experiments or every 5 min during in situ monitoring. Turbidity and temperature sensors were included in the UV–VIS sensor, and thus the water temperature and turbidity data were collected together with the proxy of [DOC]. The temperature and turbidity sensors inside the UV–VIS sensor were tested using an independently calibrated temperature sensor (HOBO U12 stainless temperature data logger, Onset Computer Corporation, USA) and a Hach 2100P Portable Turbidimeter (Hach Company, Loveland, USA).
In order to examine the feasibility of using the UV–VIS and FDOM
sensors as a proxy to estimate [DOC], three reference materials from the
International Humic Substances Society (IHSS,
During the laboratory experiments, UV–VIS and FDOM sensors were
submerged in a 10 L glass beaker containing 10 L of DI with black-cover books lying below it to minimize light reflection. The stock solution
prepared with the IHSS standards was added to the beaker so that the final
[DOC] of the solutions was within the range of 0–5.1 mg L
Artificial turbid stream water was also prepared by adding 270 g of soils
collected from the study site (see Sect. 2.3) to 10 L of the stream water
and extracting DOC from the soils for about 48 h to preclude additional
organic matter dissolved from the soil (Downing et al., 2012; Jeong et al.,
2012). The soils were collected from the study watershed at 0–15 cm depth and were air-dried and sieved (
Linear regression between UV–VIS sensor and temperature was used to
estimate the temperature correction factor for UV–VIS sensor outputs,
The temperature correction of the FDOM sensor was conducted following the
method of Watras et al. (2011).
The turbidity of the solution was measured using aliquots and the sensor
readings of the artificial turbid water were recorded every minute while the
solution was continuously stirred with a magnetic bar during the experiments.
The sensor outputs of the turbid water were compared with those of the
filtered water to calculate the
The UV–VIS, FDOM, and temperature sensors were deployed in a second-order
stream from a forested watershed, Bukmoongol Watershed (BW;
35.0319
Relationship between
The use of the sensors as a [DOC] proxy was examined during three storm events: 27–28 October 2012 (storm 1), 10–11 November 2012 (storm 2), and 23–24 April 2013 (storm 3). Both
sensors were submerged next to each other in the water in the ponding basin of
a U-shaped weir. UV–VIS sensor was deployed with the sensor head facing
the streambed to minimize the settling of particles, and compressed air cleaned
the sensor head right before the measurements to prevent sediment
accumulation. The FDOM sensor was deployed with its head directed at the streambed to
minimize light reflection. Two abnormal UV–VIS data points out of a
total of 1088 data points were filtered off when they remained larger than
mean
Plots of UV–VIS (RU: relative units), FDOM, and FDOM
During the in situ deployment of the sensors, discrete stream water samples
were collected every 1 to 4 h from the start to the end of each event.
Samples were frozen immediately after sampling and transported on ice to the
laboratory. Then, they were filtered through a GF
Laboratory experiments of UV–VIS and FDOM sensors on SRNOM, SRHA, and
SRFA exhibited strong linear relationships between the sensor signals and lab
DOC (
SUVA
Humic acids and fulvic acids are the major fraction of DOC in natural waters (Del Vecchio and Blough, 2004), covering about 60 % of aquatic DOC with a humic acid to fulvic acid ratio of 1 : 3 in the median freshwater (Perdue and Ritchie, 2014). Although DOC composition can remain relatively constant across seasons, slightly increased fluorescence per unit absorbance was reported in a forest stream in the northeastern US (Wilson et al., 2013). Since stream and riverine DOC composition can shift following storms (Fellman et al., 2009), a comparison of monitored sensor signals with lab DOC of periodically collected samples is warranted.
The UV–VIS sensor outputs showed little variability with temperature
change (slope: 0.009 to
In contrast, FDOM signals can be significantly affected by temperature
changes because the temperature increase is likely to return an excited
electron to its ground state by radiationless decay, resulting in a reduced
fluorescence emission intensity (Watras et al., 2011). We observed strong
negative correlations of the FDOM sensor with temperature in the reference
materials as well as in the whole range of DOC concentrations from 1.1 to
10.5 mg L
A study on fluorescence of wetland-dominated lakes demonstrated that slope of
the fluorescence against temperature increased as concentration decreased
(Watras et al., 2011), and the same pattern was observed in this study
(Fig. 2e). The temperature coefficient,
These results suggest that the FDOM sensor requires temperature
correction to correctly estimate [DOC], especially in streams where [DOC]
is relatively high and temperature varies a lot. After the temperature
correction, FDOM
Relationships between
The UV–VIS sensor outputs increased as turbidity increased, and, thus,
In contrast, FDOM outputs decreased exponentially, as turbidity increased
from 0 to
In situ UV–VIS and FDOM sensor outputs of raw data, corrected
for temperature at 20
Comparison between lab DOC of the three storm events (Fig. 4) and
Turbidity can increase to more than 1000 NTU during strong storms in upstream forested watersheds in South Korea although turbidity was lower than 1000 NTU throughout the year in most streams and rivers (Kim et al., 2013). Given the strong dependency of DOC estimation on turbidity in the UV–VIS sensor and the exponential decrease in FDOM outputs due to increased turbidity, correction for turbidity is a critical step for the sensors to be used as a proxy for [DOC]. This could be even more critical in streams with relatively high slopes under Asian monsoon climates. Since stream turbidity can be a function of size of particles and soil mineralogy of a watershed (Hur and Jung, 2009), site-specific correction for turbidity is necessary.
The UV–VIS and FDOM sensors followed the changes in [DOC] in the three
storms (Fig. 4) in which water temperature ranged from 8.2 to
13.8
While the UV–VIS signals did not change significantly after temperature
and turbidity corrections, the FDOM signals decreased after temperature
correction because water temperature was consistently lower than the
reference temperature of 20
The three storm events were not strong in terms of precipitation intensity and did not capture a large variation in temperature and turbidity in the field, and this is a limitation of this study. However, this can be also interpreted as meaning that the sensors can be employed to provide reliable, high-resolution data for base flow conditions. Although it has been demonstrated that the sensors can be corrected for temperature and turbidity to be used as a proxy of [DOC], there are several other factors that should be considered for successful application of the sensors in the field.
The sensors use absorbance and fluorescence of light by dissolved organic matter (DOM), and thus the
DOM with optical properties may not represent the entire DOM pool
although fluorophores correlate well with diverse, known compounds in other
riverine environments (Stubbins et al., 2014). The single excitation and
emission pair that the FDOM sensor used in this study (Table 1) estimates the
intensity of fluorescence of humic-like DOM (Stedmon and Markager, 2005). If
the dominant DOM composition of water samples reacts to different excitation
and emission pairs, for example, tryptophan-like components, which absorb at
280 nm and emit at 344 nm of wavelength (Stedmon and Markager, 2005), the
FDOM sensor may underestimate stream [DOC]. However, considering that
Although we tested two specific models of UV–VIS and FDOM sensors, multiple models are available and we did not address the variability of many sensors or variability within a model line. Sensor-specific as well as site-specific calibration would be necessary to use the sensors as a proxy of [DOC], considering that each sensor reacts differently to a range of temperatures and turbidities. For example, four types of FDOM sensors showed a different ratio of attenuation to changes in turbidity although they all showed increasing trends of attenuation with increased turbidity (Downing et al., 2012). FDOM sensors with an open path responded more strongly to turbidity changes than those with a closed path (Downing et al., 2012).
The inner filter effect (IFE) could be a problem in obtaining correct
fluorescence data of stream water if the stream has a relatively high [DOC]
with high aromaticity. However, a study highlighted that common rivers and
streams have minor IFE effects from dissolved organic matter (Downing et al.,
2012). It is unlikely that FDOM signals need to be corrected for IFE in the
forest stream, where [DOC] was less than 3 mg L
Maintaining clean surfaces on the light sources of the sensors during long-term monitoring is an important practical consideration in ensuring data quality since particles can cause light absorption or scattering (Etheridge et al., 2013). Algae which commonly occur in lakes or large rivers during summer in South Korea could interfere with light path of optical sensors. Although algae were not observed at the study site, the surface of the sensor needed to be cleaned periodically because it might still be coated with inorganic materials. The UV–VIS sensor uses air bubbles to prevent the accumulation of particles in the light beam path, and some advanced FDOM sensors have an auto-cleaning wiper. However, the frequency of field checks may still need to be decided depending on site characteristics.
A variety of organic compounds can absorb UV–Visible light and reemit light
at a longer wavelength, and this optical property can be used to monitor stream
[DOC] by UV–VIS and FDOM sensors. The credibility as well as continuity
of the field DOC data may improve significantly due to the recent advances in
sensor technology as well as wireless remote online connections if combined
with field-based calibration processes. Terrestrially derived humic materials
have many fluorophores, and thus UV–VIS and FDOM sensors have a strong
potential to be used for continuous monitoring of [DOC] in streams of
forested watersheds. However, the results shown in this study suggest that
temperature and turbidity correction using site-specific and
sensor-specific information is critical to reducing inaccurate sensor responses
to large temporal fluctuations in temperature and turbidity, particularly
during strong storm events, when turbidity can increase by a few orders of
magnitude. While the sensor correction for temperature is relatively
straightforward, that for turbidity is not simple because turbidity can be
affected by particle size and soil mineralogy. More than 80 % of light
can be attenuated at turbidity
We would like to thank Yera Shin, Eun-Byul Ko, and Young-Joon Jeon at Seoul National University for lab analyses and discussions. We also thank staff at Seoul National University Forest. This paper was supported by the Basic Science Research Program, through the National Research Foundation of Korea (NRF), funded by the Ministry of Education, Science and Technology (2011-0024706). It was also supported by the Korea Forest Service (500-20120415). Edited by: T. J. Battin