Hydrocarbon microseepage theory establishes a series of cause-effect relationships describing the formation of special surface anomalies that result from hydrocarbon microseepage and related mineral alteration (Wu et al., 2014). The theory suggests that CO2, H2O, inert gases, methane and other hydrocarbon materials enclosed in hydrocarbon traps can penetrate through the overlying, compact rock, reach the earth's surface, and become responsible for the presence of "red-fading", "clay-mineralization", and "carbonatization" enrichment phenomenon.
In the hydrocarbon microseepage model, the main reactions initiated by hydrocarbons (CH) are
The reaction in Eq. (1) causes "carbonatization" and "red-fading", and "clay-mineralization" and botanical anomalies attributed to generated H2S gas. Carbonate is promoted by the chemical oxidation and bacteriologic conversion in the hydrocarbon microseepage process. Carbonates can be detected by their spectral absorption feature at 2.34 μm. The "red-fading" phenomenon can be created by the reaction in equation (2) where ferric ion (Fe3+) is converted to ferrous ion (Fe2+) in rocks and soils caused by the action of acid/reducing solutions. This phenomenon can change spectral features of ferric Fe in the visible (VIS) and near infrared (NIR) region. Hematite, goethite, and limonite can be detected in these spectral ranges (Lammoglia and de Souza Filho, 2013). The clay minerals of kaolinite, illite, and chlorite generates by feldspars in acid environment caused by CO2 and H2S near surface. Spectral features of clay minerals can be detected in soils near 1.40 μm and near 2.2 μm.
The reactions in Eqs. (3)-(6) reflect hydrocarbon related reductions in different forms. The products from these alterations exhibit detectable spectral absorption characteristics that are the foundations of hyperspectral detection of hydrocarbon microleakage.
Near infrared wavelengths from 1.3 to 2.5 μm could be a significant remote-sensing tool for clay mineral detection due to hydroxyl-group vibration (Post and Crawford, 2014). Hydroxyl is generally bound to Mg or Al. The first overtones of OH stretches occur at approximately 1.4 μm and the combinations of the H-O-H bend with OH stretches are found near 1.9 μm. Those near 2.2 μm are characteristic of combinations involving OH-stretching and A1OH-bending modes, but 1.4 and 1.9 μm are also water absorption positions. The location and relative intensity of groups in band at 2.2 μm is characteristic of specific clay minerals. Carbonates also appear in the diagnostic vibrational absorption band due to the CO32− ion at 2.30-2.35 μm (van der Meer, 2004). Most minerals contain Fe2+ and Fe3+. The charge-transfer of this forms weak absorption between 0.40-0.55 μm. Most Fe2+ has deeper and broader absorption between 0.8 and 1.2 μm relative to samples with the least Fe2+ (Lammoglia and de Souza Filho, 2013). The electron transition of Fe3+ forms a strong absorption band at 0.89 μm.
The mineral and chemical composition of samples can be quantitatively estimated using spectral absorption features in the visible and near-infrared wavelengths from field and laboratory hyperspectral reflectance data (van der Meer, 2004). We used the spectral absorption parameters P, H, W, S and A to describe spectral absorption features after continuum removal (Fig. 1). In turn, we used continuum removal to normalize measured reflectance spectra and isolate individual absorption features of interest from a common baseline. This continuum is the overall albedo of the reflectance curve as shown in Fig. 1. The continuum removed spectrum is the ratio of the observed spectrum and continuum. The P is the spectral position of deepest absorption position. The H is the difference between unity and the reflectance of the spectral absorption position. The W is the full width at half maximum in the spectral absorption position. The A is full absorption area in the spectral absorption position. The spectral absorption symmetry (S) is ratio between absorption area on the left side (A1) and full absorption area (A), as S=A1/A.
The Xifeng Oilfield, is located in the Loess Plateau of eastern Gansu Province, China. Geologically, this oilfield is developed in the Ordos Basin, the second largest sedimentary basin in China. In the Middle and Late Triassic periods, the warm and humid climate made the Ordos Lake gradually evolve into a typical freshwater lake. The Upper Triassic Yanchang Formation becomes an important oil bearing formation (Dou et al., 2017). The elevation of the plateau ranges from 1 050 to 1 460 m. The surface is entirely covered by loess and there are relatively abundant water and natural resources within the basin, such as coal, natural gas, petroleum and halite.
Field work was conducted in August 2010. Reﬂectance spectra were measured using an ASD FieldSpec Pro ﬁeld portable spectrometer with the spectral range of 350-2 500 nm and spectral resolution 3 nm for the range 350-1 000 and 10 nm for 1 000-2 500 nm. In total, 14 soil spectra were measured at 13 field locations. For the locations where spectral measurements were taken, fresh loess soil was sampled at a depth of 5-10 cm below the land surface. The line of sight was perpendicular to scale.
Due to instrument instability, variation in illumination conditions, and atmospheric water vapor interference, large reflectance fluctuations exist in spectral ranges 1 350-1 420, 1 800-1 970, and 2 308-2 500 nm. To accurately characterize soil spectral characteristics, we conducted noise-suppressed processing on the measured reflectance spectra. We applied high-order polynomial curve fitting to smooth the whole curve. The order is 50 (Fig. 2).
We used X-Ray Diffraction (XRD) to analyze 14 ﬁeld samples collected from the study area for calcite, dolomite, kaolinite, illite and mixed-layer illite/smectite content. With an atomic absorption spectrometer, the content of CaCO3, CO32-, Fe3+, and Fe2+ were also measured (Table 1). In sites G1971, G1035, G1345, G1765 and G1866, soil sample composition test analysis does not indicate clay-mineralization compared with other soil samples. In sites G0724, G1449, G1660, G1866, soil sample composition reveals carbonatization compared with other samples. In sites G0313, G0622, G0828 and G1971, soil sample composition test analysis indicates a "red-fading" phenomenon compared with other soil samples. This analysis suggests that "carbonatization", "red-fading", and "clay-mineralization" phenomena do not necessarily appear in single field soil samples.
Table 1. Content of minerals and iron ions in soil samples. Light gray shadow represents samples where the component content is higher than the sum of mean and deviation. Dark gray shadow represents samples where the component content is lower than an average (plus standard deviation) indicator
Image data used in this study are the Hyperion L1R product acquired on January 20, 2003. Hyperion is one of three sensors onboard the earth observation satellite EO-1 launched by NASA on January 21, 2000. The sensor obtains hyperspectral images in the VNIR (400-1 000 nm) and SWIR (900-2 500 nm) regions using the push-broom method. Atmospheric correction and band removal were performed during preprocessing. The Hyperion VNIR sensor has 70 bands and the SWIR has 172 bands, however we used only 198 bands and set the others to zero values.
The Hyperion spectral bands in the ranges 121-127, 167-178 and 224 are affected by water vapor (Zhang et al., 2015). In this study, we removed the ranges 1-7, 58-78, 121-127, 167-178 and 224-240. The bad line is caused by sensor calibration and interpolated by immediate left and right neighbors. The fast line-of-sight atmospheric analysis of spectral hypercubes (FLAASH) was applied on Hyperion L1R data for atmospheric correction.
1.1.1. Hydrocarbon microseepage theory
1.1.2. Diagnostic spectral features of alterations
1.2.1. Study area and soil samples measurements
1.2.2. Soil composition test analysis data
1.2.3. Hyperspectral image data
After continuum removal, the absorption features of carbonate minerals in the spectral region 2 250-2 380 nm and the absorption features of clay minerals at 2 150-2 280 nm were highlighted. In the case of Fe3+ and Fe2+, there is an obvious absorption position between 350 and 550 nm in the soil spectra (Fig. 3). The spectral absorption parameters P, H, W, S and A in 350-550 and 2 250-2 380 nm are calculated to make regression analysis to investigate the relationship between absorption parameters and mineral contents as well as the density of Fe3+ and Fe2+ in soil samples.
This relationship was evaluated with R2 and significance (Sig). R2 values close to 1 indicate a high correlation between the dependent and independent variables. Sig represents a test of significance with the Sig value less than 0.05 indicating that the regression equation passes the test and accurately represents a correlation between the dependent and independent variables. Compared with other regression models, cubic curve estimation gave rise to the highest R2 value (Table 2). We applied multiple stepwise regression analysis to select the statistically significant variables. The correlation of five spectral absorption features generally sort as follow: H, A, S, W, P. Total carbonate is the sum of calcite, dolomite and calcium carbonate. Calcite has the highest correlation (R2=0.8) with five spectral absorption features. The calcium carbonate has the lowest correlation with five spectral absorption features. Total clay and all clay mineral including illite/smectite, illite and kaolinite are not correlated with five spectral absorption features. The five spectral absorption features are highly positively correlated with Fe2+ and negatively correlated with Fe3+.
Correlation R2 Sig Regression analysis Total carbonate H > A > S 0.530 0.048 Y=30.44-3.392A+336.561H-17.163S Calcite H > A > S 0.824 0.000 Y=10.914-1.706A+171.41H-6.307S H > A > S > W 0.836 0.001 Y=7.776-2.101A+185.222H-4.814S+0.44W H > A > S > W > P 0.846 0.004 Y=45.453-2.099A+182.186H-3.832S+0.045W-0.016P Dolomite H > A > S > P 0.591 0.065 Y= -7.661-0.055A+6.308H-1.878S+0.004P Calcium carbonate H > A > S > W > P 0.376 0.493 Y=771.176-10.808A+71.049H-368.836S-0.307W-0.313P Total clay H > A > S > W > P 0.214 0.812 Y= -125.334-0.046A-31.107H+2.423S+0.0025W+0.068P Illite/smectite H > A > S > W > P 0.288 0.672 Y= -183.693+0.364A-86.156H+1.623S-0.009W+0.089P Illite H > A > S > W > P 0.114 0.712 Y=62.124-0.332A-40.629H+1.709S-0.013W-0.025P Kaolinite H > A > S 0.101 0.774 Y=5.392-0.095A+14.432H-0.842S Fe2+ H > A > S > W (+) 0.816 0.008 Y=31.328+0.319A-55.304H+1.008S-1.67W-0.000 09P H > A > S > W (+) 0.816 0.002 Y=31.401+0.32A-55.49H+0.997S-0.167W Fe3+ H > A > W > S (-) 0.640 0.039 Y= -40.328-0.302A+51.144H-0.283S+0.23W H > A > W > S (-) 0.634 0.015 Y= -41.411-0.273A+45.219H+0.235W
Table 2. Correlation between minerals and spectral absorption features. The spectral absorption position (P), depth (H), width (W), symmetry (S), and absorption area (A) are used to multiple stepwise regression analysis
Among the five absorption parameters, the H resulted in the highest correlation to the mineral content, Fe3+ and Fe2+ using cubic curve regression (Table 2). The H is highly negatively correlated with Fe3+ and positive correlated with Fe2+ (Table 2). In this case, the H is not significantly correlated with mineral clay.
To find an effective and fast way to investigate the hydrocarbon microseepage anomalies in a loess covered area using Hyperion data, we developed an indicator. The spectral absorption depth (H) provided this opportunity by automatically detecting the "red-fading", "clay-mineralization", and "carbonatization" enrichment phenomenon of Hyperion image data.
Our absorption feature regression indicates that absorption depth and mineral content were highly correlated. Although hydrocarbon microseepage anomalies can be delineated only using absorption depth for individual minerals, we proposed an index for effectively detecting hydrocarbon microseepage anomalies in this study, where two kinds of alteration (clay-mineralization and carbonatization) are combined
H1, H2 and FI represent the absorption depth of clay-mineralization, carbonatization and fast index, respectively. In the altered area, the sum of H1 and H2 will be larger than that where no alteration area exists and the absolute difference of H1 and H2 will be smaller than that of where alteration area exists. In this case, the value of FI will also be larger. In the no alteration area, H1+H2 will be relatively small and |H1-H2| will be relatively large. In comparison, the value of FI will be small. Therefore, the area of oil and gas can have been determined by the size of the FI. In such cases though, the Hyperion spectral bands at 350-420 nm have no data and cannot be used to calculate the absorption depth of iron ions.
After atmospheric calibration with FLAASH, we used the continuum removal algorithm to transform Hyperion image data. This enhanced the contrast of individual absorption features prior to calculating the spectral absorption depth (H) of clay mineralization and carbonatization. We used Eq. (7) to calculate FI for each of the 14 soil samples based on their spectral absorption depths H1 (2 213 nm, clay) and H2 (2 335 nm, carbonate) (Table 3). With the exception of G0313, all samples in the known oil and gas area have FI values higher than the sum of mean and standard deviation of all FI values collected. Except sample G0313 and G0828, all samples near the oil wells have FI values higher than the sum of mean and standard deviation of all FI values. There are 5 samples, namely G1449, G1453, G1660, G1971 and G2075, which are located within the known oil and gas region and close to oil wells. We used the mean and standard deviation of FI of these 5 samples as a threshold to detect anomalous clay-mineralization and carbonatization in the study area. We applied this threshold to detect clay-mineralization and carbonatization information in Hyperion image data in our study area (Table 3). The absorption depth (H) and FI of carbonate and clay minerals. All soil samples are displayed with geographic information that belongs to known oil and gas areas or near oil wells. Geological mapping in this area determined the oil and gas area.
(2 335 nm)
(2 213 nm)
FI Oil and gas point
Near oil wells
G0313 0.054 5 0.050 1 23 Y Y G0622 0.026 2 0.072 8 2.3 N N G0724 0.029 6 0.071 7 2.4 N N G0828 0.054 4 0.046 3 12 N Y G0931 0.033 3 0.063 0 25 N N G1035 0.029 4 0.047 5 4.2 N N G1345 0.051 1 0.451 0 16 N N G1449 0.161 9 0.165 3 81 Y Y G1453 0.098 8 0.096 3 78 Y Y G1660 0.047 1 0.046 5 147 Y Y G1765 0.040 7 0.045 4 18 N N G1866 0.064 5 0.067 2 48 N N G1971 0.036 9 0.038 1 63 Y Y G2075 0.066 4 0.068 7 58 Y Y
Table 3. The absorption depth (H) and FI of carbonate and clay minerals. All soil samples are displayed with geographic information that belongs to known oil and gas areas or near oil wells. Geological mapping in this area determined the oil and gas area
A calculation suggests that FI values above 90 indicate possible hydrocarbon microseepage anomalies. Figure 4 shows the anomalies plotted on a Hyperion image of the study area. The blue line indicates the boundary of the oil and gas region adopted from a geological map.
Figure 4. Hydrocarbon microseepage anomalies information detected by applying the developed fast index to a Hyperion image of the study area.
The comparison between the fast index map and the geological map of this area indicates that most of hydrocarbon microseepage abnormal locations match oil and gas areas. From Fig. 4 we see that the anomalous microseepage region well covers the known oil and gas region, and the anomaly intensity in the oil and gas region is in general greater than that outside it. Therefore, we conclude that our method can be used to extract hydrocarbon microseepage from the hyperspectral satellite image of the study area.