Abstract
Anomalous concentrations of natural gas in the soil may be sourced from leaking underground gas pipelines or from natural microseepages. Due to the explosive nature of hydrocarbon gases, early detection of these gases is essential to avoid dangerous situations. It is known that natural gas in the soil affects vegetation health, which may be detected through analysis of reflectance spectra. This thesis characterizes the effects of underground gas leakage on plant and canopy development and reflectance. Based on the reflectance properties, a general gas leak detection method is proposed.
It was assumed that natural gas displaces the soil air and that oxygen shortage is the cause of changes in vegetation growth and reflectance; however it was not known whether the hydrocarbon gases have an additional effect on the vegetation. Therefore two experiments were performed to compare the effects of small gas leaks (without oxygen shortage) with large leaks (with oxygen shortage) on plant growth and reflectance. The small gas leaks were simulated by delivering natural gas, methane and ethane to pots with maize (Zea mays) and wheat (Triticum aestivum) plants. The large natural gas leak was simulated by delivering 2200 1 of gas per day to 2 by 2 m maize and wheat canopy plots. Whereas in several studies a decrease in vegetation chlorophyll was one of the main indicators of large gas leaks, this study showed that leaf area is a better indicator of gas leakage. Moreover, it was shown that when the ethane concentration in the soil reaches 0.75%, plant growth is not only affected by oxygen shortage but also by the gas itself.
The leaf reflectance of the plants was analysed using continuum removal of the blue (400-550 nm), red (550-750 nm) and two water absorption features (1370-1570 nm and 1870-2170 nm). The analysis showed that ethane caused an initial increase of 10% in reflectance between 560 and 590 nm, followed by a decrease during the course of the experiment. All gases caused an increase in reflectance in the water absorption bands. The physiological reflectance index PRI, which has previously linked water stress to photosynthetic activity, suggested that the hydrocarbon gases (particularly ethane) decreased the photosynthetic activity of the plants. The combination of reduced band depths in the chlorophyll and water absorption regions and the increased PRI suggests that ethane gas in the soil hampered a normal water uptake by maize plants in an early stage of their growth.
Since gas leaks are often accompanied by elevated carbon dioxide concentrations due to bacterial methane oxidation, an additional experiment was performed to study the effects of CO2 gas in concentrations ranging from 2% to 50% on vegetation reflectance. The red edge position in combination with a new index named the 'yellow edge position' showed that an increasing CO2 concentration corresponded to decreasing leaf chlorophyll. Two water absorption features at 1400 and 1900 nm indicated that a concentration of 50% COi decreased leaf water content. However, since other aspects such as the effects of oxygen shortage and ethane may either diminish or enhance the effects of CO2 on plant reflectance, the effects should be separated to know whether CO2 affects vegetation reflectance in a larger leak.
To define the effects of oxygen shortage on vegetation reflectance, measured oxygen concentrations were correlated with reflectance indices. Statistical covariance analysis indicated that at 29 days after oxygen shortage occurred, the reflectance indices had the highest correlation with oxygen concentrations in the soil, for both species. The effect was consistent within species, but the absolute values varied between the species. Normalization between species resulted in significant linear regression models for six reflectance indices. However, the performance of each index varied in time, resulting in the best predictions 29 days after gassing started, which was halfway through the growth cycle of the plants. When the same normalization procedure was applied to differences in time, the Vogelmann index (Vogl; R740IR12Q) could predict oxygen concentrations under the maize canopy at any timestep. Oxygen concentrations predicted from the Lichenthaler index (LIC3; R440/R740) based on the wheat canopy were less reliable (R2 of 0.55) due to patchy growth on the control plots. When these indices were used to predict the oxygen concentration in the soil in order to detect the gas leak, the maximum distance to the gas source at which the oxygen concentration was reduced was just 0.5 m. However, when a filter is applied that searches for the round shape of the leak, the chance of detecting a gas leak is increased significantly. Leak detection should optimally take place when canopy cover is between 40 and 80%, or when the NDVI is between 0.6 and 0.9. LIC3 should be used at NDVI values below 0.75, while Vogl will be a better predictor under canopies of higher NDVI.
In a large leak, ethane, CO2 and oxygen shortage all occur together. The effect of oxygen shortage in the simulated gas leak however was so strong that the additional effect of ethane could not be measured. The actual CO2 concentrations in the leak were calculated using a model that incorporates bacterial oxidising activity and continuous replenishment of the soil with methane and air. Soil CO2 concentrations did not reach values that are harmful to plants. It is therefore concluded that in a large leak, low oxygen concentrations are the main cause for changes in reflectance.
Finally, the results of the simulated gas leaks were tested in the field in a natural hydrocarbon seepage area. Even though the hyperspectral imagery dated from several years earlier, the patterns observed in the field were also observed in the image using the indices proposed earlier. This supports the conclusion that the selected indices are not season dependent and can be used at any moment in time. Moreover, using indices that are related to biomass makes the method generally applicable. The disadvantage is that such indices lead to large number of false anomalies, however, including a filter in the analysis reduced the number of false anomalies remarkably and should therefore be an integral part of the detection process.
In conclusion, this thesis has shown that gas leaks cause changes in vegetation reflectance as early as 2 weeks after gas leakage starts. If the spectral resolution of the sensor is high in the visible and NIR and the spatial resolution as high as possible (preferably as high as 1 m), vegetation reflectance can be used as indicator of gas leakage.
Original language | English |
---|---|
Qualification | Doctor of Philosophy |
Awarding Institution |
|
Supervisors/Advisors |
|
Award date | 22 Jun 2007 |
Place of Publication | [S.l.] |
Print ISBNs | 9789085046714 |
DOIs | |
Publication status | Published - 22 Jun 2007 |
Keywords
- natural gas
- soil chemistry
- seepage
- impact
- vegetation
- spectral analysis
- reflection
- remote sensing