Environ. Sci. Technol. 2003, 37, 4293-4298
Monitoring of Landfill Leachate Dispersion Using Reflectance Spectroscopy and Ground-Penetrating Radar T. SPLAJT, G. FERRIER,* AND L. E. FROSTICK Department of Geography, University of Hull, Hull HU6 7RX, England
The utility of ground-penetrating radar and reflectance spectroscopy in the monitoring of landfill sites has been investigated. Strong correlations between red edge inflection position and chlorophyll and heavy metal concentrations have been demonstrated from grassland species affected by leachate contamination of the soil adjacent to the landfill test site. This study demonstrated that reflectance spectroscopy can identify vegetation affected by leachatecontaminated soil at a range of spatial resolutions. To identify the vegetation affected by leachate contamination, the spectroradiometer must have contiguous bands at sufficient spectral resolution over the critical wave range that measures chlorophyll absorption and the red edge (between 650 and 750 nm). The utility of ground-penetrating radar data to identify leachate escaping from breakout points in the contaminant wall has also been demonstrated. An integrated approach using these techniques, combined with field and borehole sampling and contaminant migration modeling, offers a possible cost-effective monitoring approach for landfill sites.
Introduction The environmental impact of landfill sites is of major concern in developed countries due to their increasing development adjacent to urban areas (1). Abandoned landfill sites where the type and volume of the fill and the nature of the site boundaries may be unknown are also of concern (1). The generation and dispersion of leachate from landfills are slow, unsteady, nonuniform, and sometimes discontinuous depending on the degree of compaction of the fill, seasonal changes in the water supply to the system, and changes in the capping and contaminant walls (2). Continuous monitoring, both at the surface and underground, is therefore required to satisfy the risk assessment of landfill leachate (3). The assessment of the actual or potential degradation of groundwater resources at contaminated sites involves a combination of investigations that are both time-consuming and costly (4). There is therefore a requirement for rapid, noninvasive, cost-effective methodologies to aid landfill managers to position additional boreholes and to focus field sampling surveys at areas most affected by leachate migration (5) and as inputs to leachate migration modeling (6). The influence of contaminate leachate on a wide variety of vegetation species has been analyzed both in the laboratory * Corresponding author phone: +44-1482-466060; fax: +44-148244340; e-mail:
[email protected]. 10.1021/es020133f CCC: $25.00 Published on Web 08/12/2003
2003 American Chemical Society
and on landfills (7-10). The dry weight of grassland species decreased significantly when the degree of leachate contamination was increased (8, 10). Strong correlations exist between the concentration of many biochemicals within vegetation canopies and their reflectance spectra (11). Derivative analysis of reflectance spectra can identify the point of maximum slope at wavelengths between 690 and 740 nm. This point known as the red edge inflection position (REIP) has been widely used as an indicator of foliar chlorophyll concentration (12-15). Calculation of the REIP depends on the number and spectral resolution of bands within the 650-750 nm wave range and the smoothing and polynomial approximation algorithms (16, 17). Healthy foliage normally has a REIP greater than 0.715 µm, while foliage experiencing loss of chlorophyll tends to have REIP values below 0.710 µm (18). If the landfill area is very large and/or the leachate dispersion extends quite far from the landfill site, then a ground-based spectroscopy approach becomes prohibitively time-consuming and expensive. Airborne-based spectroscopy offers the potential of overcoming these constraints and providing a cost-effective method for repeated monitoring of large areas. An empirical approach to determining areas of anomalous vegetation stress has been carried out in this study. Additional research using semiempirical and fully numerical approaches such as radiative transfer modeling (19), numerical leaf models such as PROSPECT (20), and bidirectional reflectance modeling (15) to determine if additional and more quantitative vegetation physiological information can be retrieved is currently under way. In most situations shallow seismic reflection and groundpenetrating radar (GPR) methods have been demonstrated to be more useful than magnetometric (21) and geoelectrical (1) methods as landfill-mapping tools (22-24). GPRs provide high-resolution images of the dielectric properties of the top few tens of meters of the earth which can be used to detect liquid organic contaminants (25), obtain models of the largescale architecture of the subsurface, and assist in estimating hydrogeological properties such as water content, porosity, and permeability (26).
Materials and Methods Study Area. A landfill site, located in eastern England, was selected for this study because of its restricted size, 40 ha, and because prolonged leachate escape from sections of the landfill had already been identified. The site is subdivided into 15 cells (Figure 1). Cells 1-6 are unlined and capped, cells 7-10 are lined and capped, and cells 11-15 are working cells. The landfill site under study had grass (Lolium perenne) planted on the containment walls to reduce erosion and dust. Leachate had been observed escaping continuously from cells 5, 10, and 11 since 1997 with a marked increase noted after rainfall. An analysis of aerial photographs from 1992 to 1999, all acquired during the summer, combined with preliminary field investigations was carried out. This study identified the development of sporadic patches of stressed and dying vegetation growing up to 10 m from the edges of cells 5, 10, and 11 from 1997 onward. The restricted area of this landfill, the nonnatural source of the vegetation (giving an almost homogeneous vegetation cover), and the constant soil and drainage conditions allow patches of unhealthy vegetation to be used as indicators of vegetation stress caused by leachate contamination of the soil with a high degree of confidence. Arrangement of Survey Lines. After discussion with the managers of the landfill site, analysis of aerial photographs, VOL. 37, NO. 18, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4293
FIGURE 1. Map of a landfill site indicating cell and borehole locations and GPR and spectroscopy survey lines. Transects 1-5 are the combined GPR and spectroscopy survey lines, the direction of survey indicated by the arrow. Transects A-C are the chlorophyll survey lines, and the S, W & G survey line is the soil, water, and grass heavy metal concentration survey line. and preliminary field investigations, the section adjacent to cells 2 and 3 was selected as being the most likely area representative of vegetation unaffected by the leachate contamination of the soil, shown as area H in Figure 1. The section adjacent to cells 5, 10, and 11 was selected as being the most likely area representative of areas affected by leachate, shown in the close-up section of Figure 1. Five GPR and field spectroscopy survey lines were arranged parallel to the edges of the two representative sections separated by approximately 1.5 m (Figure 1). Three survey transects sampling chlorophyll and heavy metal concentrations were carried out away from the landfill edge at cells 5 and 10 for a distance of 10 m (shown as transects A-C in Figure 1). One survey line sampling the heavy metal concentration of the soil and the grass was carried out away from the landfill edge at cell 5 for a distance of 30 m, while a second survey line sampling the heavy metal concentration of the surface water was carried out for a distance of 50 m, both shown by the “S, W & G”-titled survey line in Figure 1. Chlorophyll a concentrations in the foliar samples were determined using acetone extraction and analysis by spectrophotometry. Field spectral and GPR surveys along all the planned transects, with corresponding vegetation samples, were acquired in August 1999. The spectral surveys in April 1999 and August 2000 were reduced to a limited number of point samples due to severe restrictions on sampling time caused by limited instrument availability and poor weather conditions. Field Spectroscopy. Two field spectroradiometers, the Geophysical and Environmental Research Corp. (GER) 1500 and 3700, were used in this study. The GER1500 has a spectral range of 300-1100 nm with a spectral sampling of 1.5 nm, while the GER3700 has a spectral range of 350-2500 nm and a spectral sampling of 1.5 nm (350-1050 nm), 6.2 nm (10501900 nm), and 9.5 nm (1900-2500 nm). Prior to each measurement, reference spectra from a calibrated Spectralon tablet were collected to convert final measurements to absolute percent reflectance. For each spectral sample location, three replicate spectra were recorded under clear skies around local noon. The field of view of the spectroradiometers was set at 8° and the sensor head located 1 m above the target. The first derivative spectra were derived using polynomial functions fitted by least squares over a 6 nm interval (27, 28). From these, the location of the REIP was determined. 4294
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 18, 2003
Airborne Imaging Spectroscopy Data. Airborne imaging spectroscopy data were acquired by the compact airborne spectrographic imager (casi) in April and September 1999. Data were acquired between 407 and 944 nm using both 13-discrete-band and 72-contiguous-band configurations. The spatial resolutions of all the casi images were constant at 1 m. The spectral resolution of the 13-band casi data was 10 nm, while that of the 72-band casi data was 8 nm. Image Data Preprocessing and Processing. Apparent surface reflectance was retrieved using the empirical line atmospheric correction technique utilizing ground spectra acquired coincident with the casi data acquisition (29). A large mound of gypsum and roads and car parks within the landfill site gave a number of ground targets of variable reflectance. The casi image data were roll corrected using the navigation data to remove most significant aircraft motion effects from the imagery. A preliminary analysis of the image dataset, using a minimum noise fraction (MNF) transform implemented in the IDL/ENVI software package (30), was carried out to identify the spectrally distinct landcover end members. The principal components derived from the MNF transform when plotted in two and three dimensions and related to their spatial distribution have been found to be useful in differentiating landcover end members (31). The first-derivative spectra of the 72-band casi data were derived using polynomial functions fitted by least squares over a 4-band interval. From these, the location of the REIP was determined. To assess the sensitivity of the casi imagery in identifying vegetation health, both the reflectance and derivative spectra of the field spectroscopy data were convolved to the 13- and 72-band casi configurations. Ground-Penetrating Radar. Pulse EKKO 100 and 1000 GPRs, produced by Sensor and Software Ltd., were used in this study. The surveys were conducted under sunny and dry conditions using 50, 100, 200, 225, and 450 MHz antennas, a “reflection” mode of operation, and a fixed offset profile mode. Calibration of the unprocessed radar data was carried out using the EKKO and Slicer 3D software packages. The accuracy of calculating the attenuation, skin depth, and vertical resolution was calculated using the methodologies described in refs 32-35. The GPR image depths were validated to 0.06 m/ns using geological profiles from the preliminary study and drilling logs from boreholes 13, 5a, and 6a.
FIGURE 2. Heavy metal concentrations in soil and grass from transect S, W & G (location shown in Figure 1).
FIGURE 4. Heavy metal concentration and chlorophyll concentration (µg/g) from transects A-C (locations shown in Figure 1).
FIGURE 3. Heavy metal concentrations in surface water from transect S, W & G (location shown in Figure 1). Interpretation of the images was carried out according to the procedures outlined in ref 36.
Results Laboratory Analysis. Soil, water, and vegetation samples from both study sections were analyzed. Vegetation samples from
the section unaffected by leachate contamination (area H in Figure 1) showed mimimum, mean, and maximum chlorophyll concentrations of 814, 844, and 931 mg/kg, respectively. In the section affected by leachate contamination anomalously high concentrations of a number of heavy metals were present in the vegetation, soil, and surface water up to 10 m from the edge of the landfill adjacent to cells 5 and 10 (Figures 2-4). Field and Airborne Spectroscopy. The analysis of the reflectance spectra from both of the studied sections of the landfill could not resolve any distinct spectral classes. The REIP positions of the same dataset however were spread over a range of wavelengths with two distinct spectral groupings identified. The first grouping, whose samples all VOL. 37, NO. 18, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4295
FIGURE 5. REIP values from transects 1-5 (locations shown in Figure 1).
FIGURE 7. First derivative of stressed and healthy vegetation reflectance spectra convolved to the 72-band casi configurations. FIGURE 6. Chlorophyll concentration (µg/g) versus REIP.
TABLE 1. Correlation between Chlorophyll Content and Heavy Metal Concentration along Transects A-C from 1 to 7 m from the Landfill Edge (Figure 5) transect
heavy metal
correlation coefficient (R 2)
formula
A B C
Ti Cr Zn
0.844 0.833 0.9028
y ) -34.283x + 51381 y ) -30.116x + 63183 y ) -43.501x + 73080
came from the uncontaminated section (section H in Figure 5), had high reflectance values above 800 nm and REIP positions located between 725 and 730 nm (Figure 5). The second grouping had generally lower reflectance values above 800 nm and REIP positions located between 695 and 708 nm (Figure 5). The second grouping consisted of samples acquired from the contaminated section between 8 and 17 m along the transects (Figure 5). A strong correlation was found between REIP and laboratory-measured chlorophyll content (R2 ) 0.748, y ) 11.962x - 8154.2, shown in Figure 6). Strong correlations between the heavy metal and chlorophyll concentrations were found from 1 to 7 m from the landfill edge (Table 1, Figure 7). These results, combined with previous research (6, 37-39), strongly support the validity of using spectroscopy to identify the chlorophyll concentration. From 7 to 11 m 4296
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 18, 2003
from the landfill the heavy metal concentrations decrease linearly; however, there is a distinct drop in the chlorophyll content in all the samples acquired 9 m from the landfill edge. These chlorophyll concentrations appear to be anomalous and may have been caused by the presence of an unidentified additional inhibitor to vegetation growth, possibly a buried road or ditch. As the drainage, illumination and climate conditions, and soil and vegetation types are virtually homogeneous over the whole landfill site, an empirically derived relationship between the chlorophyll concentration, and hence REIP, and the degree of soil contamination from leachate can be proposed with a high degree of certainty. Using both these results and previous research (6, 37-39), the first grouping of REIP data points can be considered representative of “healthy” vegetation while the second grouping can be considered representative of “stressed” vegetation. The field and derivative spectra convolved to the casi band configurations show that the 72-band images have sufficient spectral resolution to clearly differentiate the stressed and healthy vegetation types (Figure 7). The bandwidth configuration used for the 13-band casi images in this study meant that the 13-band casi images could not differentiate the vegetation types, showing only that the healthy vegetation types had a derivative value between 735 and 745 nm, higher than that of the stressed vegetation. When the derivative images from the 72-band casi data were classified to highlight
FIGURE 8. Distribution of pixels with (a) REIP values between 695 and 710 nm (colored gray and marked by the letter C) and (b) REIP values between 735 and 745 nm (colored white and marked by the letter H) derived from the 72-band casi data from September 1999 with the landfill site plan overlain (shown in Figure 1).
FIGURE 9. GPR image (100 MHz) from transect 1. The anomalous feature is circled. the pixels with REIP values between 690 and 710 nm, a number of patches appeared adjacent to cells 5, 10, and 11 indicating regions of high probability of stressed vegetation (colored gray and identified with the letter C in Figure 8). When the derivative images from the 72-band casi data were classified to highlight the pixels with REIP values between 735 and 745 nm, a number of patches appeared adjacent to cells 1 and 2 indicating regions of high probability of healthy vegetation (colored white and identified with the letter H in Figure 8).
GPR Surveys. Transects 1-5, shown in Figure 1, identified an anomalous feature, 2 m wide and 12 m in depth, in the 50, 100, and 200 MHz GPR images. The 50 and 100 MHz images identified the same depth and extent of the anomalous feature; however, the 100 MHz image gave much higher detail (Figure 9). The anomalous feature was weak and poorly defined in the 200 MHz images. The strength of the anomalous feature was noticeably greater in the 100 MHz image in transect 5 than in transect 1. The survey area soil consisted of very dry sandy soil which the radar signal could VOL. 37, NO. 18, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4297
penetrate to a considerable depth. At the anomalous feature the radar signal does not penetrate the soil smoothly but is seen to be absorbed in a distinct vertical strip approximately 12 m in depth and 2 m wide. This effect is most probably due to the pore water conductivity, increasing due to the presence of leachate contaminant from the landfill, causing the radar signals to be more strongly absorbed and hence the signal penetration to decrease. After comparison with similar surveys from published research (40) and discussions with domain experts (41), this anomalous feature has been interpreted as most likely to be a leachate breakout point in the containment cutoff wall.
Discussion This study has demonstrated the potential of reflectance spectroscopy to identify vegetation affected by leachatecontaminated soil at a range of spatial resolutions. The spectroradiometer must have contiguous bands at sufficient spectral resolution over the critical wave range that measures chlorophyll absorption and the red edge (between 650 and 750 nm) to achieve this. Reflectance spectroscopy has considerable advantages over algorithms using band ratios to determine vegetation health, such as the normalized difference vegetation index (NDVI). NDVI relates the density of vegetation to the relative health of the vegetation, whereas reflectance spectroscopy directly observes the health of the vegetation. In addition, because the spectral response of vegetation is very different from that of soil, particularly in the 650-750 nm wave range, vegetation health can be determined even at relatively low levels of vegetation cover and biomass. The utility of GPR to identify leachate breakout points in the contaminant cutoff wall has also been demonstrated. An integrated approach using both these techniques, combined with field and borehole sampling and contaminant migration modeling, offers a possible monitoring approach for landfill sites. Preliminary and repeat surveys using airborne- or satellite-based spectroradiometers could provide first-pass observations of large areas which could be followed up using ground-based spectroradiometers which could further focus field sampling and GPR surveys on the most affected areas in real time. This study has presented results from only one study site. A much more comprehensive validation of this monitoring approach is required. A number of landfill sites of different ages, with different geological, hydrogeological, and climatic settings, need to be studied before this approach can be considered for operational use.
Acknowledgments This work was supported by equipment loans from the NERC Equipment Pool for Field Spectroscopy and the NERC Geophysical Equipment Pool. The airborne remote sensing data were supplied by the NERC Airborne Remote Sensing Facility. T.S.’s studentship was funded by Landfill Tax Credits through Enventure Northern. We thank the editor and reviewers for their helpful comments.
Literature Cited (1) Meju, M. A. J. Appl. Geophys. 2000, 44 (2-3), 115-150. (2) Fang, H. Y. In Waste Disposal by Landfill, Green’93; Sarsby, R. W., Ed.; A. A. Balkema: Rotterdam, The Netherlands, pp 419426. (3) Butt, T. E.; Oduyemi, K. O. K. Environ. Int. 2003, 28, 597-608. (4) Asante-Duah, D. K. Managing Contaminated Sites; WileyInterscience: Chichester, U.K., 1996. (5) Abbaspour, K.; Matta, V.; Huggenberger, P.; Johnson, C. A. J. Contam. Hydrol. 2000, 40, 365-380.
4298
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 18, 2003
(6) Fatta, D.; Naoum, D.; Loizidou, M. J. Environ. Manage. 2002, 64, 4, 333-343. (7) Cureton, P. M.; Groenevelt; P. H.; McBride, R. A. J. Environ. Qual. 1991, 20, 1, 17-24. (8) Adarve, M. J.; Hernandez, A. J.; Gil, A.; Pastor, J. J. Environ. Qual. 1998, 27, 6, 1286-1293. (9) Chan, Y. S. G.; Wong, M. H.; Whitton, B. A. Water, Air, Soil Pollut. 1999, 111 (1-4), 29-40. (10) Hernandez, A. J.; Adarve, M. J.; Gil, A.; Pastor, J. Chemosphere 1999, 38, 7, 1693-1711. (11) Curran, P. J.; Kupiec, J. A.; Smith, G. M. IEEE Trans. Geosci. Remote Sens. 1997, 35, 415-420. (12) Jago, R. A.; Cutler, M. E. J.; Curran, P. J. Remote Sens. Environ. 1999, 68, 217-224. (13) Curran, P. J.; Dungan, J. L.; Gholz, H. L. Tree Physiol. 1990, 7, 33-48. (14) Curran, P. J.; Dungan, J. L.; Macler, B. A.; Plummer, S. E. Remote Sens. Environ. 1991, 35, 69-76. (15) Curran, P. J.; Windham, W. R.; Gholz, H. L. Tree Physiol. 1995, 15, 203-206. (16) Tsai, F.; Philpot, W. Remote Sens. Environ. 1998, 66, 41-51. (17) Dawson, T. P.; Curran, P. J. Int. J. Remote Sens. 1998, 19, 21332139. (18) Bubier, J. L.; Rock, B. N.; Crill, P. M. J. Geophys. Res. 1997, D24, 29483-29494. (19) Weiss, M.; Troufleau, D.; Baret, F.; Chauki, H.; Prevot, L.; Olioso, A.; Bruguier, N.; Brisson, N. Agric. For. Meteorol. 2001, 108, 2, 113-128. (20) Zarco-Tejada, P. J.; Miller, J. R.; Mohammed, G. H.; Noland, T. L.; Sampson, P. H. J. Environ. Qual. 2002, 31, 5, 1433-1441. (21) Hinze, W. J.; Roberts, R. L.; Leap, D. I. Geotechnical and Environmental Geophysics; Ward, S. H., Ed.; Society of Exploration Geophysicists: 1990; Vol. 2, pp 267-272. (22) Ali, J. W.; Hill, I. A. Reflection seismics for shallow geological investigationssa case study from Central England. J. Geol. Soc. (London) 1991, 148, 219-222. (23) De Iaco, R.; Green, A. G.; Maurer, H. R.; Horstmeyer, H. J. Appl. Phys. 2003, 1408, 1-18. (24) Peters, L.; Daniels, J. J.; Young, J. D. Proc. IEEE 1994, 82, 12, 1802-1822. (25) Nobes, D. C. Surv. Geophys. 1996, 17, 4, 393-454. (26) Knight, R. Annu. Rev. Earth Planet. Sci. 2001, 29, 229-255. (27) Savitzsky, A.; Golay, M. J. E. Anal. Chem. 1964, 36, 1627-1639. (28) Tsai, F.; Philpot, W. D. IEEE Trans. Geosci. Remote Sens. 2002, 40, 2, 416-425. (29) Ferrier, G. Int. J. Remote Sens. 1996, 17, 2881-2889. (30) Green, A. A.; Berman, M.; Switzer, P.; Craig, M. D. IEEE Trans. Geosci. Remote Sens. 1988, 26, 65-74. (31) Ferrier, G. Remote Sens. Environ. 1999, 68, 125-137. (32) Reynolds, J. M.; Taylorl, D. I. Proceedings of the 2nd International Conference on Polluted and Marginal Land, London, 1992; pp 121-131. (33) Padaraze, I.; Forde, M. C. Proceedings of the 3rd International Conference on Polluted and Marginal Land 1994; pp 397-401. (34) Davis, J. L.; Annan, A. P. Geophys. Prospect. 1989, 37, 531-551. (35) Forde, M.C. Proceedings of the 4th International Conference on Polluted and Marginal Land; London, 1996; pp 443-450. (36) Annan, A. P.; Cosway, S. W. Proceedings of the Symposium on the Application of Geophysics to Environmental and Engineering Problems (SAGEEP), 1992; pp 191-208. (37) Miller, J. R.; Hare, E. W.; Wu, J. Int. J. Remote Sens. 1990, 11, 1755-1773. (38) Horler, D. N. H.; Dockray, M.; Barber, J. Int. J. Remote Sens. 1983, 4, 279-289. (39) Curran, P. J.; Hay, A. M. Photogramm. Eng. Remote Sens. 1986, 52, 229-241. (40) Liu, G.; Yonggang, J.; Liu, H.; Qiu, H.; Qiu, D., Shan, H. Environ. Sci. Technol. 2002, 36, 1077-1085. (41) Annan, A. P., Sensor and Software Ltd. Personal communication, 2000.
Received for review July 4, 2002. Revised manuscript received June 27, 2003. Accepted June 27, 2003. ES020133F
