In situ laser-induced fluorescence (LIF) analysis of petroleum product-contaminated soil samples

Abstract

General considerations of the calibrations of in situ measurements are presented and the concept of using an “average oil" with average analysability for calibration purposes is introduced. The in situ analysis of 30 petroleum product-contaminated soil samples with laser-induced fluorescence (LIF) spectroscopy was performed. Compared to an uncontaminated laboratory reference (LR) soil, 23 soil samples exhibited significantly higher LIF signals, so that these soil samples were classified as contaminated. The repeatability and reproducibility of the in situ LIF analysis were investigated. For the calibration of the LIF data, two LR oils (a fuel oil and a crude oil) were employed. The degree of soil contamination with petroleum products ranged from the limit of detection (LOD) for LIF analysis (ca. 100 ppm), or below, to more than 10 000 ppm. The petroleum product concentrations determined with in situ LIF analysis reveal a reasonable correlation with the results of standard IR analysis after extraction of the contaminated soils.

---

Introduction

Laser-induced fluorescence (LIF) spectroscopy is excellently suited for the in situ analysis of fluorescent organic compounds in the environment. It has been shown that both stationary and time-resolved fluorescence measurements can be employed to determine, e.g., petroleum product concentrations in water or soils with limits of detection (LOD) in the range relevant for the surveillance of regulation limits or clean-up decisions.^1–8 For petroleum products in soils, a concentration of 5000 ppm can be regarded as a characteristic reference value for regulation purposes (see below). The major advantage of in situ LIF analysis is the lack of any sampling and clean-up procedure preceding the analysis and the resulting potential for field screening applications.⁹ In spite of these advantages, only the “detect/non-detect" or at most the semiquantitative field screening capabilities of the method have been used in recent application reports.^3,4 The caution with respect to the reliability of the quantitative results of LIF analysis is mainly related to the considerable difficulties in establishing a reliable calibration procedure. If the analysis of petroleum product contamination of soils is considered, the widely varying physicochemical and photophysical properties of analytes and matrices strongly influence the fluorescence signal obtained.

Calibration is usually accomplished by analysing a set of calibration standards which are prepared from certified reference (CR) materials or standard analytical materials.¹⁰ As every individual soil contamination is a combination of a specific soil and a specific petroleum product, standard analytical materials are usually not available for the matrix or the analyte. This lack of well-defined calibration standards for the calibration of in situ LIF measurements can be overcome using two different approaches: (i) a stringent calibration procedure can be established using well-characterized laboratory reference materials which model as close as possible the properties of the contamination and the matrix on a specific site; (ii) calibration is performed with the standard analytical materials available and the influences on the measured fluorescence signal are defined, characterized and taken into account for the analysis of unknown samples. Throughout this paper, both concepts will be discussed with regard to in situ LIF analysis of petroleum products in soils. The first approach is discussed in detail with respect to the results of the in situ LIF analysis of 30 petroleum product-contaminated soil samples. The results of these in situ analyses are compared with the results of a standard method (according to German industrial norm DIN 38409-H18¹¹) obtained in the laboratory. The results presented here are related to our earlier reports on LIF investigations of contaminated soils which focused mainly on laboratory-doped (spiked) natural and model soils.^5,6,12,13 The influences of the physicochemical and photophysical parameters of the analyte (fluorescence efficiency) and the matrix (reflectivity) on the fluorescence signal have been described in Refs. 6 and 12 and will only briefly be summarized.

Experimental section

Excitation of the fluorescence was performed with a nitrogen laser (MSG800-TD, LTB, Berlin, Germany; excitation wavelength λ_ex = 337 nm; pulse energy E_P = 400 µJ before and E_P > 100 µJ at the end of the optical fibre). The fluorescence signal was detected with a polychromator/CCD (charged coupled device) camera combination (Model 01-001, Amko, Tornesch, Germany, grating with 150 lines mm⁻¹; and PicoStar, LaVision, Göttingen, Germany, 576 × 384 pixel, spectral resolution 0.6 nm channel⁻¹, maxum sensitivity 1 photon count⁻¹). In comparison with our previous results for petroleum product-contaminated soils,⁵ which were obtained with a different detection system [SIT (silicon intensified target) camera: SIT500, B&M Spektronik, Olching, Germany, 500 × 400 pixel, spectral resolution 0.8 nm channel⁻¹, maximum sensitivity 15 photons count⁻¹], better LOD were achieved with the CCD camera detection system (typically, LOD = 2200 ppm with the SIT camera and LOD ≅ 100 ppm with the CCD camera were obtained for the fuel oil) due to the different inherent sensitivities as well as the different signal-to-noise ratios of the detection devices.

For the preparation of the calibration standards, we used laboratory reference (LR) materials: a well-characterized soil [LR soil: A_h horizon, total organic carbon (TOC) = 4%, humidity approximately 14%¹⁴] and several petroleum products (LR oils: fuel oil, heavy gasoil, diesel fuel, Statfjord, Brent and German crude oil; all oils were received from Deutsche Shell AG, Hamburg, Germany). All LR oils were extensively characterized with regard to their photophysical properties (see below and Ref. 13). The LR materials were chosen to match the most common properties expected for real world samples. The chosen soil type (A_h horizon) is in most cases the top soil layer from which real world samples are usually obtained. The properties of this soil will obviously not match the properties of other soils for which different LR soils would have to be used. The calibration standards were prepared by carefully mixing 10 g of the soil with an aliquot of the petroleum product under investigation dissolved in 10 ml of n-pentane, and removing the solvent by application of a slight vacuum (p ≅ 1 hPa). We have previously shown that linear calibrations with very good quality (given, e.g., by the correlation coefficients r = 0.991 and 0.987 for fuel and crude oil, respectively) are obtained in the concentration range c = 0–25 000 ppm for the reference oils.⁶ Therefore, quick calibrations for reference were sometimes limited to a concentration range of c = 1000–5000 ppm.

The petroleum product-contaminated soil samples were made available within an interlaboratory study¹⁵ (the strict conditions of a ring test with respect to the exact agreement of analytical and storage procedures as well as identical times of analysis have not been met). Thirty different samples were taken from the top soil layer (A_h horizon) at eight different sites (filling stations and surroundings of fuel or oil storage tanks) in Hamburg, Germany (the origin of different sites is indicated by roman numbers I–VIII). At all sites, samples were taken from different locations (indicated by the letters following the site number, e.g. VIIIa–VIIIk). The samples were characterized by means of their soil parameters (total carbon content, TOC, humidity, particle size fractions, mineral contents, etc.) by A. Baermann.¹⁵ The age of the contamination is not known for any sample; the physical properties range from rather dry (water content for most samples between 3–5% by weight¹⁵) fine-grained soils to wet (water content 15%) clumpy samples.

All soil samples were placed in Pyrex glass tubes without further preparation and sealed. The fluorescence was detected from six different locations directly on the surface of the tubes using a multifurcated optical fibre (no separation of the fluorescence signal from the six locations was performed). The integrated fluorescence intensity in the wavelength range λ = 370–524 nm was used as the concentration-dependent fluorescence signal. During the overall analysis time for all samples, an external standard (perylene in cyclohexane, concentration c ≈ 1 × 10⁻⁴ M) was measured repeatedly to determine possible temporal variations of the experimental setup.

Two sets of measurements were performed with the petroleum product-contaminated soils. The samples were analysed five times in one day and the standard deviation σ for the five measurements was calculated and used to describe the uncertainty of the LIF measurement. It is noted that this standard deviation describes a combination of the inhomogeneity of the contamination in the soil and the uncertainty of the fluorescence measurement. The fluctuation found for the fluorescence signal of the external standard during these measurements was approximately 8% and the deviations for the soil samples were quite similar. Therefore, the fluorescence signals were not corrected with respect to the external standard for this set. In the second set of measurements, a week later, the petroleum product-contaminated soil samples were analysed again and the calibration standards were measured. Compared with the first set of analyses, the fluorescence signal of the external standard changed significantly [first set: I_F(standard) ± σ(standard) = (4830 ± 391) counts; second set: I_F(standard) ± σ(standard) = (3330 ± 140) counts] and therefore a correction with respect to the external standard was applied (see inset of Fig. 2 and Results and Discussion).

As a standard method, the IR analysis of a 1,1,2-trichloro-2,2,1-trifluoroethane (freon 112) extract of the soil samples was used in analogy with the procedure described in the German industrial norm DIN 38409-H18.¹¹ All analyses were performed by H. Borsdorf and J. Flachowsky at the UFZ-Centre for Environmental Research Leipzig.¹⁵ Samples were prepared for IR analysis by extracting 10 g of the contaminated soil for 30 min with 25 ml of freon 112 using a shaker. The resulting extracts were centrifuged at 3000 rpm for 5 min; 5 ml of the extracts were cleaned using a column of 8 g of aluminium oxide, in order to remove polar compounds. The petroleum hydrocarbons were eluted from the column with 20 ml of freon 112 and the eluate was filled up to 25 ml. The sample pretreatment is described in detail in Ref. 15, and the procedure for IR analysis can be found in Ref. 11. The results for the analysis with the standard method depend strongly on the sample pretreatment (e.g.c = 1368 ppm, c = 1827 ppm and c = 2753 ppm were found when the untreated soil was extracted with freon 112, the dried soil was extracted with freon 112 or supercritical fluid extraction with CO₂ was applied, respectively). Therefore we have used the results for the sample pretreatment which is as close as possible to in situ measurements (samples were not dried before extraction) for comparison.

Results and discussion

Calibration of in situ measurements: general considerations

In situ analysis requires that the substances under investigation (analytes) remain in their original environment, i.e. are not removed by extraction from the matrix. The capability of quantitative in situ analysis of petroleum product-contaminated soils holds high promise, but also encounters considerable experimental challenges and analytical difficulties. The latter are evidently due to the fact that petroleum products per se are very complex, ill-defined analytes, and that soils are extremely complex matrices. Thus, the in situ LIF investigation of contaminated soils requires a specific analytical strategy which will be discussed first. In general terms, four characteristic situations, ranging from simple analytes in simple matrices (case A), through simple analytes in complex matrices (B), and vice versa (C), to complex analytes in complex matrices (D), will be distinguished as limiting cases (Fig. 1). The most important step in quantitative analytical procedures is the calibration of the measurement, i.e. the establishment of a correlation between the instrumental response and the concentration of the analytes. This is usually accomplished by measuring a set of calibration standards of known concentration and obtaining the (preferably linear) calibration function. It is recommended to prepare the calibration standards from standard analytical materials, e.g. chemical standards or CR materials.¹⁰

Fig. 1 Schematic representation of useful calibration procedures for the in situ analysis of multicomponent analytes in heterogeneous matrices.

For in situ analysis, this procedure is straightforward only if two conditions are met (case A in Fig. 1): (i) the analytes are known or identifiable and are available as standards (simple analytes), and (ii) the matrix is also available as a standard material (simple matrix). Obviously, problems related to the matrix can be overcome by extraction and subsequent quantification of the analytes (bringing us from case B to case A in Fig. 1). However, the conventional ex situ analysis, which requires sampling and pretreatment of the samples, often introduces new uncertainties into the analytical procedure.

In environmental and, e.g., biological in situ analysis, the matrices encountered, e.g. natural waters, soils, tissues, body fluids, etc., are usually very complex and heterogeneous and can thus affect the properties even of simple analytes and the measurement process unpredictably. Moreover, such matrices are usually not available as standard analytical materials. Hence, for the calibration of in situ measurements in case B, there are two alternatives: (i) LR materials can be used as matrices if they match the real matrices in the most relevant properties; this requires the ability to comprehensively classify and characterize the matrices under investigation; and (ii) standard addition of the analytes can be applied as calibration procedure.

Many analytes under investigation are so complex in nature that a complete qualitative analysis of the large number of constituents is not feasible or not useful (complex analytes). This also holds for petroleum products which are represented by a wide range of fuels and oils and can consist of hundreds of compounds in highly varying compositions. The complexity of petroleum products inherently excludes the use of standard analytical materials for calibration. Petroleum products in the environment, e.g. under the influence of weathering, microbial transformation, etc., exhibit strongly changed properties with respect to the original material, so that even the availability of the original contaminant does not improve the analysis of an altered contamination. However, under certain conditions, the extraction of an altered contamination and the use of the extract for the preparation of calibration standards may be possible.

Essentially, there are three strategies to analyse and calibrate complex analytes. (i) Sum parameters, e.g. the total petroleum hydrocarbon (TPH) or total recoverable petroleum hydrocarbon (TRPH) content, can be determined. The lack of adequate calibration materials even for this procedure is often circumvented in laboratory analysis by calibrating with a single or a few compounds not directly related to the analyte. The drawback of this method is obviously that it is difficult to find substances with properties representative of complex analytes. For example, the calibration of IR analysis according to the German industrial norm DIN 38409-H18 is usually performed with the triterpene squalane as a standard (using the CH₂ and CH₃ stretching bands). However, squalane contains no aromatic hydrogen–carbon bonds and therefore the IR spectroscopic quantification of petroleum products with a high content of aromatic molecules is doubtful. (ii) Alternatively, single individual constituents of complex analytes, e.g. selected polycyclic aromatic compounds (PAC) from petroleum products, after extraction and separation (bringing us from case D via case C to case A in Fig. 1) can be identified and quantified. However, usually a relationship between the quality and quantity of these “indicator compounds" and the properties of a complex analyte is not obvious. (iii) Finally, it can be attempted to find appropriate LR materials for the complex analytes. Although a complete qualitative analysis, or at least an unambiguous identification of petroleum products, is usually not possible, we have shown previously that a classification of petroleum products by means of their photophysical parameters can be obtained.^5,6 We were able to differentiate between various refined and crude oils and thus to create a database for different petroleum products which can serve as LR materials. Of course, it is essential to comprehensively characterize the analyte under investigation in order to find a suitable LR material with resemblance of the most important properties.

It is important to note that the lack of suitable calibration standards, the problems related to the interpretation of “indicator compounds" and the finding of suitable LR materials for complex analytes hold for both ex situ measurements in simple matrices, e.g. of extracts in solution (case C), and for in situ analysis in complex matrices, e.g. in soils (case D).

A different approach for the calibration of in situ measurements can be applied if the dependence of the instrumental response on the matrix and analyte properties, relevant for the analytical method used, is well characterized. For example, the fluorescence intensity I_F(λ_em) at a distinct emission wavelength λ_em can approximately be described by the following equation:

Here const. includes constant instrumental and experimental parameters, such as detection geometry, detector sensitivity, optical interaction length, etc., I₀(λ_ex) is the excitation intensity at the excitation wavelength λ_ex and c is the concentration of the analyte. The influence of analyte (A) and matrix (M) specific parameters on the measured fluorescence signal is described by the functions f(A) and f(M), respectively.

We have shown previously that f(A) is mainly characterized by the effective extinction coefficient and the fluorescence efficiency (which both depend on λ_ex) of the petroleum product under investigation. We have measured these parameters for our LR oils in cyclohexane solution, and were able to distinguish between crude oils and refined petroleum products with respect to these photophysical parameters. A summary of these and other photophysical parameters of the LR oils in use is given in Ref. 13. The influence of these parameters on the analysability of the different LR oils in soils has been discussed in detail.^5,6

Many matrix parameters are expected to influence the fluorescence intensity and therefore the function f(M) obtained for a given analyte on a soil surface, e.g. the moisture content, the total carbon, the humic substance content and the grain size of the soil. Since optical measurements are applied, the main influence is obviously defined by the optical parameters of the soil, e.g. the absorptivity and reflectivity of the matrix. Recently, we have shown that, from diffuse reflectance measurements of the soil under investigation, reliable parameters can be extracted to quantitatively describe the influence of different matrices on the fluorescence signal and to determine the function f(M) in a first approximation.¹²

If both functions f(A) and f(M) are sufficiently well known, the calibration procedure can be performed with standard analytical materials, and the deviations for different analytes or matrices found on a specific site can be calculated from these functions. This concept has been discussed in more detail with respect to different oils in Ref. 6 and for different soils in Ref. 12.

Analysis of petroleum product-contaminated soil samples

The measured LIF intensities of the petroleum product-contaminated soil samples are shown in Fig. 2 and are compared with the intensity of the background signal detected from the LR soil [I_F(reference) ± σ(reference) = (110 ± 10) counts]. The background signal arises mainly from the humic substances which are themselves fluorescent. However, the fluorescence efficiency of humic substances in general, and therefore the overall fluorescence intensity detected for the LR soil, is rather low and background subtraction was not necessary. For 23 of the 30 petroleum product-contaminated soil samples, a fluorescence intensity significantly higher [i.e.I_F(sample) ± σ(sample) > I_F(reference) ± σ(reference)] than that of the uncontaminated LR soil was obtained. The largest LIF signal obtained from sample VIIId was I_F(VIIId) = (3375 ± 350) counts, i.e. 30-fold higher than the LR soil background signal. Only one contaminated sample (VIa) showed an average fluorescence intensity which was lower than that of the LR soil, but was within the standard deviation. This demonstrates clearly that the chosen LR soil is an excellent reference material for the contaminated soil samples under investigation. The repeatability of the method, which is described by the standard deviation for the set of five replicate measurements, was found to be below 20% for all samples and even below 10% for 21 samples. This clearly underlines the high reliability of LIF analysis even under in situ conditions. To determine the reproducibility, we measured the contaminated samples a week later and corrected the LIF signals for changes in the experimental setup by means of the external fluorescence standard (for details, see Experimental section). The inset of Fig. 2 shows a scatter plot of the LIF intensities (x axis) of the first five measurements versus the LIF intensity of the sixth measurement performed a week later (y axis). That the values are evenly distributed around the diagonal (bisector) indicates the absence of systematic deviations between these two sets of experiments. For the 23 samples exhibiting higher fluorescence intensities than the reference, a significant contamination with petroleum products can be assumed. In conclusion, the results presented in Fig. 2 demonstrate that even the uncalibrated LIF data can be used—validated by one routine laboratory analysis, if necessary—for the screening of contaminated sites, for the location of the contamination boundary on a specific site or for the control of on-site or off-site remediation processes.

Fig. 2 Fluorescence intensities of petroleum product-contaminated soil samples measured directly from the surface [Ia–VIIIk: real samples from different sites (I–VIII) and from different locations at the same site (a–k); Ref.: uncontaminated laboratory reference material; λ_ex = 337 nm, emission was detected with the CCD detection system and integrated in the range λ_em = 370–524 nm]. The inset shows a scatter plot of the LIF intensities (x axis) of the first five measurements versus the LIF intensity of the sixth measurement performed a week later (y axis).

In order to calibrate the LIF data obtained for the contaminated soils, we investigated two sets of calibration standards for which the LR soil was mixed with the fuel oil or the Brent crude oil. These LR oils were chosen from the group of refined (fuel oil) and crude (Brent) oils and represent within their group medium analysability. The idea behind this concept was to identify the type of contamination by means of the fluorescence data obtained from the contaminated samples and their comparison with the known data of the LR oils. Fig. 3 shows the fluorescence spectra and fluorescence decay functions of several contaminated soil samples measured directly from the surface. The comparison of these data for all samples under investigation reveals that the distinction between samples from different sites is possible by means of their fluorescence properties. This is additionally emphasized by the fact that the fluorescence spectra and decay curves are rather similar for samples taken from the same site (e.g. VIIIa–k, data not shown here). Of course, a direct correlation between the calibration standards (fuel oil or Brent crude oil in the LR soil) and distinct contaminations on the basis of fluorescence properties was not possible. We have therefore used the regression parameters of a combined data set obtained with the fuel and crude oil calibration standards for the calibration of the LIF intensities. This procedure was chosen because the selected reference oils are representative of poor and excellent analysability on the reference soil. Analysability, which under constant experimental conditions is defined by the photophysical properties of the oil under investigation, can be described by the sensitivity, i.e. the slope, of the calibration curve S. The sensitivities have been determined to be S = 0.38 counts ppm⁻¹ for the crude oil and S = 0.20 counts ppm⁻¹ for the fuel oil. It is hoped that, by combining the data sets of two oils with poor and excellent analysability and thereby introducing the fluorescence properties of an “average oil" with an average sensitivity of S ≈ (0.3 ± 0.1) counts ppm⁻¹ into the calibration, the properties of the unknown contamination of the soil samples can be matched. It is noteworthy that the uncertainty introduced by this calibration procedure is small compared with the deviations found, e.g., in interlaboratory comparisons dealing with the analysis of contaminated soil samples using standard analytical methods in the laboratory (see discussion below and Refs. 16, 17).

Fig. 3 Fluorescence spectra (upper) and fluorescence decay curves (lower) of contaminated soil samples from different sites directly measured on the soil surface [measurements with CCD detection system, λ_ex = 337 nm, for the fluorescence spectra a high-pass filter (cut-off wavelength 368 nm) was placed in the emission path, parameters for measurement of fluorescence decay curves were emission wavelength range λ_em = 508–564 nm, gate width t_G = 500 ps, temporal increment Δt = 200 ps].

After calibration, the concentrations determined with in situ LIF analysis can be compared with the results obtained with a standard method. It must be noted that the specific analytes which are accessible with the analytical methods compared here are different. Whereas the analysis according to DIN 38409-H18 is optimized for the quantitative analysis of hydrocarbons (the method is similar to US EPA Method 418.1¹⁸ for the detection of TRPH), only fluorescent compounds independent of their chemical structure are analysed if LIF spectroscopy is used.

The petroleum product concentrations determined with the different analytical methods for the contaminated soil samples under investigation are shown in Fig. 4. It was found that the degree of soil contamination with petroleum products ranged from the LOD for LIF analysis (ca. 100 ppm), or below, to more than 10 000 ppm. The double-logarithmic plot reveals a fair correlation between the concentrations found with the different methods over a wide concentration range (more than two orders of magnitude). The regular scattering of the data points around the line of exact agreement (dashed diagonal line in Fig. 4) suggests that no systematic deviations occurred. Only six of the 30 samples showed concentrations which were below the LOD for the in situ LIF method (LOD ≈ 100 ppm). However, these samples, where the concentration cannot be determined using in situ LIF analysis, are only slightly contaminated (concentrations with the standard method found to be below c_DIN = 400 ppm). If, e.g., the I value (intervention value) of the so-called Dutch list^19,20 for petroleum products (I = 5000 ppm) is considered, probably no environmental risk arises from these samples. The LOD found for our experimental setup (see also Ref. 12) is in good agreement with the values reported by Lieberman et al.²¹ for cone penetrometer systems utilizing LIF for the detection of subsurface petroleum product contaminations.

Fig. 4 Petroleum product concentrations of the contaminated soil samples measured with in situ LIF analysis and with IR spectroscopy after extraction in accordance with DIN 38409-H18. Also indicated is the limit of detection (LOD) for the LIF technique.

The relative standard deviations (RSD), obtained by dividing the standard deviation σ with the average of the concentrations measured with the different methods, are in the range 5–124% and therefore in reasonable agreement with the deviations found in other, more sophisticated, interlaboratory comparisons. For the analysis of eight PAC in a natural sediment or a sewage sludge, deviations between Δc = 29–58% and Δc = 37–149%, respectively, were found in an interlaboratory comparison where 23 laboratories participated and either GC-MS with flame ionization detection or HPLC analysis with UV detection was used for the analysis.¹⁷ Similar results were found for the analysis of soil samples from a former gaswork: for the 16 PAC defined as priority pollutants by the US EPA, the RSD lies between 54.7–154.6% for a certain sample analysed by nine laboratories using also GC-MS and HPLC.¹⁶ Although the statistical database for the calculation of the RSD in our approach is only small, and therefore the reliability of the results must be considered with caution, it should be pointed out that the results obtained in the cited interlaboratory comparisons were achieved after extraction of the matrices with highly sophisticated analytical procedures. In contrast, the results in this work were obtained with a method which can obviously be performed in situ without any sample pretreatment, and therefore is capable of rapid and reliable measurements with high temporal and spatial resolution.

Acknowledgements

This work was performed as part of the research programme Optimierung von Analysenverfahren zur umweltverträglichen Analytik von Mineralölkohlenwasserstoffen im Boden (Optimization of Analytical Procedures for the Sustainable Analysis of Petroleum Products in Soils) in cooperation with Dr H. Borsdorf and Dr J. Flachowsky, Umweltforschungszentrum Leipzig-Halle (optimization of extraction procedures for analysis according to DIN 38409-H18) and Prof. Dr G. Matz and Dipl.-Geol. J. Kübler, Mobilab MM1 Umweltanalytik GmbH, Hamburg (application of mobile GC-MS analysis). The research programme was financially supported by the Deutsche Bundesstiftung Umwelt (DBU), Osnabrück. More experimental details and the results of the other groups are included in the final reports available from the DBU (in German).

References

1. R. Kotzick and R. Niessner, Fresenius' J. Anal. Chem., 1996, 354, 72.
2. D. S. Knowles S. H. Lieberman, in Environmental Monitoring and Hazardous Waste Site Remediation, ed. T. Vo-Dinh and R. Niessner, Proc. SPIE vol. 2504, Munich, 1995, pp. 297–307..
3. G. Bujewski B. Rutherford, The site characterization and analysis penetrometer system (SCAPS) laser-induced fluorescence (LIF) sensor and support system, EPA Rep. EPA/600/R-97/520, Environmental Protection Agency, Washington DC, 1997..
4. G. Bujewski B. Rutherford, The rapid optical screening tool (ROST^TM) laser-induced fluorescence (LIF) system for screening of petroleum hydrocarbons in sub-surface soils, EPA Rep. EPA/600/R-97/020, Environmental Protection Agency, Washington DC, 1997..
5. H.-G. Löhmannsröben Th. Roch, in Analytiker Taschenbuch, ed. H. Günzler, A. M. Bahadir, R. Borsdorf, K. Danzer, W. Fresenius, R. Galensa, W. Huber, M. Linscheid, I. Lüderwald, G. Schwedt, G. Tölg and H. Wisser, Springer, Berlin, Heidelberg, 1997, vol. 15, pp. 217–253..
6. H.-G. Löhmannsröben Th. Roch, in Advanced Technologies for Environmental Monitoring and Remediation, ed. T. Vo-Dinh, Proc. SPIE vol. 2835, Munich, 1996, pp. 128–134..
7. M. U. Kumke, H.-G. Löhmannsröben and Th. Roch, J. Fluoresc., 1995, 5, 139.
8. W. Schade and J. Bublitz, Environ. Sci. Technol., 1996, 30, 1451.
9. Field analytical and site characterization technologies: summary of applications, EPA Rep. EPA/542/R-97/011, Environmental Protection Agency, Washington DC, 1997..
10. A. R. Castro, in Encyclopedia of Analytical Science, ed. A. Townsend, Academic Press Harcourt Brace & Company, London, 1995, vol. 1, pp. 433–447..
11. Deutsche Einheitsverfahren zur Wasser-, Abwasser- und Schlammuntersuchung, Beuth Verlag, Berlin, 1981..
12. H.-G. Löhmannsröben and L. Schober, Appl. Opt., 1999, 38, 1404.
13. H.-G. Löhmannsröben C. Kauffmann Th. Roch, in Remote Sensing of Vegetation and Water, and Standardization of Remote Sensing Methods, ed. G. Cecchi, T. Lamp, R. Reuter and K. Weber, Proc. SPIE vol. 3107, Munich, 1997, pp. 305–314..
14. K. Czeschka, Einfluß von Starterkulturen und Tensiden auf den Abbau von Kohlenwasserstoffen in Bodenfestbettreaktoren, Dissertation, TU Braunschweig, 1995..
15. Optimierung umweltverträglicher Analysenverfahren für Mineralölkohlenwasserstoffe im Boden, ed. H. Borsdorf and J. Flachowsky, Umweltforschungszentrum Leipzig-Halle GmbH, UFZ-Bericht, Leipzig, 12/ 1997..
16. I. Blankenhorn, D. Meijer and R. J. van Delft, Fresenius' J. Anal. Chem., 1992, 343, 497.
17. P. de Voogt, J. Hinschberger, E. A. Maier, B. Griepink, H. Muntau and J. Jacob, Fresenius' J. Anal. Chem., 1996, 356, 41.
18. Methods for chemical analysis of water and wastes, EPA Rep. EPA 4.600479020, Environmental Protection Agency, Washington DC, 1983..
19. Bodenschutz, ed. D. Rosenkranz, G. Einsele and H.-M. Harreß, E. Schmidt, Berlin, 1988, pp. 1–27..
20. Bodenschutz, ed. D. Rosenkranz, G. Einsele and H.-M. Harreß, E. Schmidt, Berlin, 1995, vol. 2, pp. 1–4..
21. S. H. Lieberman G. A. Theriault S. S. Cooper P. G. Malone R. S. Olsen P. W. Lurk, in Proceedings of the Second International Symposium on Field Screening Methods of Hazardous Wastes and Toxic Chemicals, Air and Waste Management Association, Pittsburgh, PA, 1991, pp. 57–63..

---

Footnote

† Present address: Deutsche Post AG, SNL IT-Systeme Brief Kommunikation, AE Produktion Brief 1, Konrad-Adenauer-Platz 1, 40210 Düsseldorf, Germany.

---