Lake sediments serve as archives that reflect biological and chemical conditions of lakes. Carbon, nitrogen and phosphorus are determined in sediments and in suspended material in lakes, termed seston, for a number of purposes. These include: assessment of sediment quality, identification of the trophic level of lakes, and study of contaminants. The use of near-infrared spectroscopy provides a rapid and cost-effective alternative for routine analysis of large numbers of samples. Calibration equations developed from spectral data and the results of conventional chemical analysis on an initial set of representative samples are used to predict constituents in future unknown samples of the same type. In sediment samples and seston, carbon, nitrogen and phosphorus were generally predicted very successfully by near-infrared spectroscopy. The best samples for analyzing sediment quality are generally those from the deepest part of a lake. Spatial variability in sediment quality of a lake was successfully explored by this method, suggesting that near-infrared spectroscopy is a potential complementary tool to standard methods for analyzing and characterizing sediments.

Introduction

Aquatic ecosystems, at this time in history, are subjected to a diversity and scale of human-induced influence previously unseen. These impacts include enrichment with nutrients, over-harvesting of fish and invertebrates, oil spills, climate change, loss of habitat, and others. On the positive side, acidification is one serious impact on freshwater systems whose influence has been successfully identified and widely reversed.

Nevertheless, there is widespread contamination of sediments from heavy metals (Allen and Nriagu, 1933), organochlorines (Muir and Norstrom, 1994) and excess phosphorus. Critical to early detection of risk is knowledge of background conditions in lakes, other water bodies and marine areas, to provide reliable baseline data against which to measure human-induced changes and to identify environmental goals for remediation.

Methods

Chemical analysis of liquid, solid or biological aquatic samples is key to understanding and monitoring pristine lakes for baseline data and evidence of pollution. Often these analyses are costly, time-consuming and generally destroy the samples. Near-infrared spectroscopy (NIRS) (Williams and Norris, 2001; Burns and Ciurczak, 1992), is used widely in agriculture and has its longest history for analysis of soil since the 1960s (Malley, 1998). It is used in many other industries, including medical diagnostics, and is a growing alternative technology for the rapid, non-destructive, and cost-effective analysis of some types of samples from aquatic ecosystems, including living organisms (Malley and Nilsson, 1995, 1996). Intact samples can be analyzed in a relatively undisturbed state (Malley and Nilsson, 1996), thereby allowing the determination of functional as well as compositional parameters. Multiple analyses may be conducted on the same samples simultaneously. Furthermore, NIRS can be operated in the field, a mobile laboratory, remote laboratory, or onboard ship for facilitating environmental monitoring. This can result in not only lower costs for research and monitoring, but, more significantly, on-site early detection of environmental changes and early opportunity to respond to, mitigate, or even prevent unfavourable environmental changes.

Near-infrared spectroscopy utilizes the absorbance of overtones and combinations of vibrations from molecular absorptions in the mid-infrared region of the electromagnetic spectrum. The near-infrared region lies between 700 and 2500 nm, i.e. between the visible range with shorter wavelengths and the infrared range with longer wavelengths. Chemical bonds absorbing in this region frequently include C-H, N-H and O-H. Most frequently, materials absorbing in this region are organic compounds, often associated with living organisms. Also, water is a major absorber in the NIR range.

Quantitative analysis using NIRS depends entirely on the process of calibration (Williams, 2001). A calibration is based on spectral data from a set of representative samples (Figure 1) that range significantly in concentration in the parameters to be predicted. These spectral data and analytical data obtained on the samples by conventional analytical methods are modeled using multivariate statistical analysis methods to produce calibrations. A separate calibration is required for each constituent or property to be determined. Numerous calibrations may be uploaded to an NIR spectrometer at one time to predict multiple constituents or properties simultaneously. Subsequently, when an unknown sample is scanned by the instrument, the constituents for which calibrations have been entered into the NIR instrument are displayed immediately.

Figure 1.

Representative spectra of sediment samples from lakes at the Experiment Lakes Area. The Y-axis is absorbance.

Figure 1.

Representative spectra of sediment samples from lakes at the Experiment Lakes Area. The Y-axis is absorbance.

Figure 1 shows spectra, in the reduced range of 1100 to 2500 nm to remove areas of noise. These spectra have little character and few distinct peaks. Sets of multiple wavelengths are identified in the calibration process for each constituent or property. The spectral data may be subject to mathematical treatments, such as transformation to derivatives as part of calibration development.

Figures 2a–d, shows the results of calibration development for carbon (C), carbonate, nitrogen (N) and phosphorus (P) in which the NIR-predicted values for each sample on the Y-axis are plotted against the chemically-determined values for the same samples. The coefficient of determination, r2, used here is one of the key numerical measures of success of a calibration. Based on r2, the best calibration was developed for P, followed by C and N, and lastly for carbonate. The calibrations are judged to be excellent (r2 > 0.95) for P, successful (r2 = 0.90 to 0.95) for C and N, and moderately successful (r2 = 0.8 to 0.9) for carbonate using evaluation guidelines proposed by Malley et al. (2004).

Figure 2.

Linear relationships between NIR-predicted values (Y-axis) and the measured values (X-axis) for 11 sediment core samples. Figure reprinted from Malley et al. (1999).

Figure 2.

Linear relationships between NIR-predicted values (Y-axis) and the measured values (X-axis) for 11 sediment core samples. Figure reprinted from Malley et al. (1999).

Results

Analysis of nutrients in, and predicting lake water pH from, freshwater sediments

Swedish researchers lead the way in exploring the use of NIR spectra from surface sediments of lakes to predict water quality. Nilsson et al. (1996) collected surface sediment samples (0–1 cm thick) from the deepest point in 58 lakes in southern Sweden. This 1 cm sediment thickness represents between 2 and 5 years of sediment accumulation. The lakes were 10–30 ha in surface area, 5–20 m deep, and ranged from severely acidified to eutrophic lakes. Near-infrared spectroscopy (NIRS) predicted total P in the lake water from the lake sediments with an r2 of 0.83 between the NIR predicted values and chemical analytical values. For lake pH, r2 was 0.85, and for total organic carbon (TOC) in 25 of the lakes was r2 = 0.68. Although the TOC prediction was not very successful, the authors conclude that lake sediments contain information about TP, pH, and TOC conditions in the water of lakes in which the sediments are formed. These relationships could be explained by TP, pH and TOC being major factors controlling the composition of lake microbiota and the particulate matter of the water and, therefore, sediment properties. The need for understanding the course of acidification and eutrophication in Europe and North America led to the search for methods, such as NIRS, to reconstruct baseline conditions and long-term temporal trends.

A study by Malley et al. (1996a) combined sediments collected in 80 cores from Roddy Lake, 290 ha in surface area, at the Experimental Lakes Area, northwestern Ontario, with those from Colville Lake in the Northwest Territories. One 10-cm long core was taken in Colville Lake and sliced to give 10 1-cm thick slices. Sediment samples were freeze-dried, passed through a sieve to remove large particles, and ground to 125 μm mean particle size. For sediments combined from the two lakes, NIRS spectroscopy predicted both total C and N with r2 of 0.99, and P with r2 of 0.97.

Sediment samples (n = 153) were studied from four Canadian lakes with locations across more than 19 degrees of latitude, where the lakes varied in surface area from 1205 ha to Lake Winnipeg with 2.38 × 106 ha (Malley et al., 2000). For total C, r2 was 0.99, CO32− was 0.98, organic carbon was 0.98, for N was 0.97. For P, r2 was a less favourable 0.53. Profiles for each constituent and lake showed generally good agreement between the chemically measured values and those predicted by NIRS. Re-analysis of P in the outlying P sediment samples by analytical chemistry showed closer overall agreement, with r2 of 0.70. This demonstrated a secondary role of NIRS in identifying errors in results from conventional chemical analysis.

An early study of NIRS analysis of surface sediments from the deepest part of 21 forest lakes in southern Sweden with pH ranging from 4.4 to 7.0 resulted in r2 of 0.69 (Korsman et al., 1992). Excluding only one lake, the r2 improved to 0.96. Two lakes with pH 4.6 and 4.8, respectively, were cored down to 20 to 25 cm of sediment depth. The r2 between predicted historical pH down the sediment cores based on diatom inferred pH was 0.79 for one lake, and 0.76 for the second.

In a study on 73 lakes in southern Sweden, surface sediment samples were collected in triplicate from the deepest parts of the lakes (Dabakk et al., 1999a). Lakes were 10 to 30 ha in area, 5 to 20 m deep, and ranged from acidified to eutrophic. Measured lake water pH data, averaged over a 4-year period, was sufficient in 33 lakes to predict lake water pH accurately from the NIR spectra of sediments. Within each lake, r2 values from predicting lake pH ranged from 0.89 to 0.97. This study determined that sampling the sediments from the deepest point in the lake gives the most reliable results.

Eutrophic Lake Arendsee situated in an agricultural area of North Germany was cored at the deepest point in the lake in September 1993 (Malley et al., 1999). The lake has a surface area of 5.14 × 106 m2 and maximum depth of 48.7 m. A single core of 47-cm length was sectioned into 47 sub-sections, 34 of which were used in this study. Samples were scanned with NIRS without grinding. For total C, r2 was 0.93; for carbonate, 0.81; for N, 0.94; for P 0.99; and for N:P ratio, 0.93. Three species of diatoms were also analyzed in the core slices to examine the possibility that NIRS may reveal biological records in sediments as well as chemical. The three species of diatoms in the sediment cores, Stephanodiscus binatus, Cyclotella rossii and Fragilaria crotonensis each had a unique distribution with depth and together comprised 62.5% of the total diatoms in the core. Calibrations for the three species of diatoms separately produced r2 from 0.55 to 0.69.

Analysis of heavy metal concentrations in lake sediments

At the Experimental Lakes Area, northwestern Ontario, Lake 382 was subjected to the experimental addition of Cd as CdCl2 up to, and not exceeding, the Canadian Water Quality Guideline (Malley and Williams, 1997). Other metals, Fe, Mn, Zn, Cu, Pb, and Ni were present at normal background levels for a pristine lake at ELA. Lake 382 is 36.9 ha in surface area and 13.1 m deep. During the ice-free seasons of 1987 to 1992, Lake 382 received a total addition of 6.7 kg of Cd, more than 95% of which was retained in the sediments. Sampling consisted of 4 to 6 cores taken at each of three sites on 7–8 September 1994. In addition to the metals, total C and total N were determined. The sites varied in organic matter content from <1–<3% to 25–69%. Total C ranged from 2–13 mg g−1 to 230–425 mg g−1. Percent organic matter and total C were highly correlated (r = 0.97). Carbonate was <2% to 4.15% of total C. Total N ranged from 0.4 mg g−1 dw to 11.0 mg g−1 dw. Total P ranged from 0.1 mg g−1 dw to 0.24 mg g−1 dw. Metal concentrations ranged from several to 20-fold over the sample set. The r2 between the NIR predicted and the measured values was 0.86 for Fe, 0.92 for Mn, 0.93 for Zn, 0.91 for Cu, 0.81 for Pb, 0.88 for Ni, and 0.63 for Cd. This shows it is feasible to determine background concentrations of the first six heavy metals in sediments using NIRS. The display of latent variables showed that the prediction of heavy metals by NIRS was facilitated by their association with the cellulose in the sediments. Cadmium was experimentally elevated in the samples from the upper 5–9 cm in the sediment cores, yet it was present in the lowest concentrations of all the metals. Nevertheless, it is speculated that the poorer predictions for Cd are due to the shorter time that the added Cd had to bind to the ligands in the sediment. The NIRS technique is based on the absorption of light by organic materials. The metals themselves do not absorb NIRS wavelengths. If the added Cd had not reached stability in binding with the ligands, this may account for the poorer performance of Cd in the NIR prediction. This study demonstrates that the prediction of resident heavy metals concentrations in freshwater sediment by NIRS is feasible.

Analysis of spatial variability in lake sediments

Paleolimnological methods are used to study temporal and spatial environmental impacts including acidification or eutrophication. The sediments arise from material produced within the lake itself or derived from the catchment or airborne sources. Korsman et al. (1999) used NIRS analysis of sediment samples in an attempt to relate variation in sediment composition to water depth, amount of organic matter, and catchment sources. A mesotrophic, humic lake in northern Sweden 0.5 km2 in area with a pH of 5.8 was the subject of the study. The top 1-cm of sediment was sampled at 165 coring sites. Water depth and organic matter content accounted for 20% and 16%, respectively, of the variance in the NIR data. The influence of inlets and clear-cutting of forests were more important. These were associated not only with variability in the amount of organic matter, but more importantly with the quality of the organic matter. Significant in its effect on the variability of organic matter was allochthonous organic matter brought in by ditches from a farm and dwellings, that probably brought wastewater from animals and people. This organic matter was different in chemical composition from that in the lake, a situation that is readily detected by NIRS.

Analysis of suspended material in lake water

Analysis of particulate C, N and P in freshwaters contributes to the understanding of C, N and P cycles. Suspended particles, termed seston, are commonly collected on Whatman GF/C filters which are air-dried and then analyzed by chemical methods for total C, total N, and total P. In a study using NIRS, 53 filters containing seston from two lakes at the Experimental Lakes Area, northwestern Ontario, were prepared by filtering water from water depths of 1 to 29 m in one lake of surface area, 56.1 ha, and depth, 30.4 m, and 1 to 12 m in the other of 36.9 ha area and 13.1 m depth (Malley et al., 1993). Seston collected on filters represented 420 to 2900 μg l−1 of lake water. The filters were air-dried and scanned from 400–1800 nm using NIRS. The r2 between the NIR predicted and the chemical values for the concentration C, N and P in the lake water of the two lakes combined were 0.98 for C, 0.99 for N and 0.97 for P.

In another study, ten lakes within the Experimental Lakes Area, northwestern Ontario were sampled for seston (Malley et al., 1996c). These lakes varied in surface area from 5.0 to 26.9 ha and maximum depth from 4.1 to 20.8 m. They ranged from oligotrophic to eutrophic, circum-neutral in pH to acidic, and from pristine to receiving metal additions. Dissolved inorganic C (DIC), DOC, suspended C, suspended N and suspended P were sampled in the ice-free season of 1992 in all lakes, except for P, which was sampled in four lakes. The r2 between the NIR predicted and the chemical values for the concentration in the seston for C ranged from 0.89 to 0.95; for N ranged from 0.91 to 0.97, and for P ranged from 0.88 to 0.92.

A study conducted in Lake 382 at the Experimental Lakes Area, receiving additions of cadmium, predicted C, N, P, and Cd in the 0.1–3 μm size fraction from the water column (Malley et al., 1996b). Aliquots of lake water of 1.5 mL were filtered through Whatman GF/C glass fibre filters that were dried and scanned. The r2 between the NIR predicted and the chemical values were 0.92 for C, 0.85 for N, 0.87 for P and 0.75 for Cd. NIRS is useful for analyzing organic matter composition in this size fraction, including the organic matter that binds this metal, predominantly algae <3 μm in size.

Dabakk et al. (1999b) sampled 261 lakes in northern Sweden at 1 m water depth. In addition to attempting to predict total phosphorus (TP), total organic carbon (TOC), pH and colour, several variables were examined including the type of filter, suction applied during filtration, filter area, and time between water collection and filtration. Measured pH in the lakes varied from 4.74 to 7.87. The r2 for predicting pH in the 261 lakes was 0.51, and for 256 lakes with pH >6 was 0.34. Colour was predicted with r2 of 0.66. For TOC in lake water as μg l−1, r2 was 0.59; for TP in lake water as μg l−1, r2 was 0.34. Prediction of TP on the filter, as μg filter−1, was more successful with r2 of 0.85. It was concluded that pH, P and water colour affect the properties of the seston, and that these properties can be described by NIR spectra. The type of filter can affect the results and best results were obtained with glass fibre filters. A larger diameter filter produces better results than a smaller diameter filter. The authors concluded that NIR spectra from seston of lakes can be monitored and lead to action when deviations from the normal are detected.

Discussion and conclusions

Near-infrared spectroscopy does not entirely replace conventional analytical methods since calibrations for each constituent and brand of instrument have to be developed for each sample type. In addition, in routine operation, a small proportion of samples are subjected to conventional analytical analysis to confirm that the calibrations are predicting accurately over time. These new analytical results and the spectra with which they are associated are periodically added to update the calibration data base. Calibrations are recalculated and made more robust.

Capital costs are associated with the acquisition of NIR instruments and training of operators. Training personnel to operate instruments takes some hours. Training personnel for the development, installation, evaluation, and updating of calibrations is a somewhat longer process. Near-infrared spectroscopy is optimally used for the analysis of constituents and properties in situations where large numbers of samples will be analyzed over significant periods of time. Its capacity to analyse samples in real-time, as in food processing lines, and its safety in analyzing live samples, such as fish or crustaceans makes NIRS a valuable analytical tool.

The in-field or on-ship portability, rapidity of the analyses, the resulting cost-savings, and the simultaneous analysis of multiple constituents or properties, place NIRS technology in a position to contribute greatly to routine monitoring of aquatic ecosystems, including oceans, to enable early detection of environmental problems, and documenting environmental recovery as it occurs. The spectra collected and samples that are dried or frozen and archived can serve as an on-going reference to the quality of sediments in selected places and times. Moreover, as capabilities develop and additional constituents or properties can be determined by NIRS, the archives can serve as a data source for time-dated analyses in the past, including over geological time (Rosen, 2005; Rosen et al., 2000, 2001).

References

Allen, R. J. , Nriagu, J. O. (Eds.),
1993
.
Heavy Metals in the Environment, 9th International Conference on Heavy Metals in the Environment, Toronto
,
1993 September
, Vol
1
,
CEP Consultants Ltd.
,
Edinburgh
.
Burns, D. A. , Ciurczak, E. M. (Eds.),
1992
. Handbook of Near-Infrared Analysis. Practical Spectroscopy Series No. 13
,
Marcel Dekker, Inc.
,
New York, Basel
,
Hong Kong
.
Dabakk, E. , Nilsson, M. , Geladi, P. , Wold, S. , Renberg, I.,
1999a
.
Sampling reproducibility and error estimation in near infrared calibration of lake sediments for water quality monitoring. J
.
Near Infrared Spectroscopy
7
,
241
250
.
Dabakk, E. , Nilsson, M. , Geladi, P. , Wold, S. , Renberg, I.,
1999b
.
Inferring lake water chemistry from filtered seston using NIR spectrometry
.
Wat. Res.
34
(
5
),
1666
1672
.
Korsman, T. , Nilsson, M. , Ohman, J. , Renberg, I.,
1992
.
Near-infrared reflectance spectroscopy of sediments: A potential method to infer the past pH of lakes
.
Environ. Sci. Tech.
26
,
2122
2126
.
Korsman, T. , Nilsson, M. B. , Landgren, K. , Renberg, I.,
1999
.
Spatial variability in surface sediment composition characterized by near-infrared (NIR) spectroscopy
.
Journal of Paleolimnology
21
,
61
71
.
Malley, D. F.,
1998
.
Near-infrared spectroscopy as a potential method for routine sediment analysis to improve rapidity and efficiency
.
Wat. Sci. Tech.
37
, No.
6–7
,
181
188
.
Malley, D. F. , Nilsson, M.,
1995
.
Environmental applications of near infrared spectroscopy: Seeing the environment in a different light
.
Spectroscopy Europe
7
(
6
),
8
16
.
Malley, D. F., Nilsson, M.,
1996
.
Environmental applications: seeing the environment in a different light
.
NIR news
,
7
No.
1
,
10
13
.
Malley, D. F. , Williams, P. C.,
1997
.
Use of near-infrared reflectance spectroscopy in prediction of heavy metals in freshwater sediment by their association with organic matter. Environ, Sci
.
Technol.
31
,
3461
3467
.
Malley, D. F. , Williams, P. C. , Stainton, M. P. , Hauser, B. W.,
1993
.
Application of near-infrared reflectance spectroscopy in the measurement of carbon, nitrogen, and phosphorus in seston from oligotrophic lakes
.
Can. J. Fish. Aquat. Sci.
50
,
1779
1785
.
Malley, D. F. , Hauser, B. W. , Williams, P. C. , Hall, J.,
1996a
.
Prediction of organic carbon, nitrogen and phosphorus in freshwater sediments using near infrared reflectance spectroscopy
. In: A.M.C. Davies and P. Williams (Eds.),
Near Infrared Spectroscopy: The Future Waves
, pp.
691
699
, NIR Publications, Chichester, UK.
Malley, D. F. , Lawrence, S. G. , Holoka, M. H. , Williams, P. C.,
1996b
.
Applying near-infrared reflectance spectroscopy to predict carbon, nitrogen, phosphorus, and organic-bound cadium in lake picoplankton. J
.
Aquatic Ecosystem Health
5
,
135
147
.
Malley, D. F. , Williams, P. C. , Stainton, M. P.,
1996c
.
Rapid measurement of suspended C, N, and P from Precambrian Shield lakes using near-infrared reflectance spectroscopy
.
Wat. Res.
30
(
6
),
1325
1332
.
Malley, D. F. , Ronicke, H. , Findlay, D. L. , Zippel, B.,
1999
.
Feasibility of using near-infrared reflectance spectroscopy for the analysis of C, N, P, and diatoms in lake sediments
.
J. Paleolimnology
21
,
295
306
.
Malley, D. F. , Lockhart, L. , Wilkinson, P. , Hauser, B.,
2000
.
Determination of carbon, carbonate, nitrogen, and phosphorus in freshwater sediments by near-infrared reflectance spectroscopy: rapid analysis and a check on conventional analytical methods
.
Journal of Paleolimnology
24
,
415
425
.
Malley, D. F. , Martin, P. D. , Ben-Dor, E.,
2004
.
Application in analysis of soils
. In: C.A. Roberts , J. Workman, Jr. , J.B. Reeves, III (Eds.),
Near-infrared Spectroscopy in Agriculture, Agronomy Monograph 44, Chapter 26
, pp.
729
784
.
American Society of Agronomy, Crop Science Society of America, Soil Science Society of America
, Madison, WI.
Muir, D. C. G. , Norstrom, R.J.,
1994
.
Persistent organic contaminants in Arctic marine and freshwater ecosystems
.
Arctic Res. United States
,
8
,
136
146
.
Nilsson, M. B. , Dabakk, E. , Korsman, T. , Renberg, I.,
1996
.
Quantifying relationships between near-infrared reflectance spectra of lake sediments and water chemistry
.
Environ. Sci. Technol.
30
,
2586
2590
.
Rosen, P.,
2005
.
Total organic carbon (TOC) of lake water during the Holocene inferred from lake sediments and near-infrared spectroscopy (NIRS) in eight lakes from northern Sweden
.
Biogeochemistry
76
,
503
516
.
Rosen, P. , Dabakk, E. , Renberg, I. , Nilsson, M. , Hall, R.,
2000
.
Near-infrared spectroscopy (NIRS): a new tool for inferring past climatic changes from lake sediments
.
The Holocene
10
(
2
),
161
166
.
Rosen, P. , Segerestropm, U. , Eriksson, L. , Renberg, I. , Birks, H. J. B.,
2001
.
Holocene climatic change reconstructed from diatoms, chironmids, pollen and near-infrared spectroscopy at an alpine lake (Sjjuodjijaure) in northern Sweden
.
The Holocene
11
(
5
),
551
562
.
Williams, P.,
2001
.
Implementation of Near-Infrared Technology
. In: P. Williams and K. Norris (Eds.),
Near-Infrared Technology in the Agricultural and Food Industries. 2nd Edition
. pp
145
169
.
American Association of Cereal Chemists, Inc.
,
St. Paul, Minnesota
.
Williams, P. , Norris, K. (Eds.),
2001
.
Near-Infrared Technology in the Agricultural and Food Industries. 2nd Edition
.
American Association of Cereal Chemists, Inc.
,
St. Paul, Minnesota
.