**5. Mapping hydrothermal alterations**

The date of **Figure 2** corresponds to the dry season, where vegetation influence is minimized; the scene is devoid of clouds. Landsat 8 images contain 11 bands; in this processing, we used coastal/aerosol, blue, green, red, NIR, and 2 of the SWIR bands [12].

 Atmospheric correction was performed with the Histogram Minimum Method (e.g., [13]). The radiometric correction was performed resampling the image's digital numbers to fit the 8-bit radiometric resolution; the spatial resolution was not modified. The image with these corrections was presented above in **Figure 2**.

To map regions of hydrothermal alteration, we focus on enhancing the spectral response of their typical mineralogical contents. Three major groups characterize hydrothermal alterations (e.g., [14]), hydroxyls (clays and micas), iron minerals (hematite, goethite, and jarosite), and hydrated sulfates (chalk and alunite). Their spectral signatures appear in **Figure 20**.

The identification of hydrothermal alterations is approached with four methods, to be discussed below, applied to the satellite image: Band Ratio, Principal Component Analysis (PCA), the Crosta Technique (Crosta), and the Fraser Technique (Fraser). Results are compared and discussed to optimize the hydrothermal identification.

#### **5.1 Band ratio**

Enhancement of hydrothermal alterations is performed making the ratio of the satellite bands that better characterize them. Bands B4/B2 for the oxides, and bands B6/B7 for the hydroxyls. The ratio B5/B4 is included to represent vegetation in the color composition representation of these components, as shown in **Figure 21**.

The digital numbers resulting from the band ratios are rescaled to cover the digital values from 0 to 255 for each ratio. A better visualization is obtained when the false color image is classified; we obtain the percentage of pixels belonging to each class: oxides, hydroxyls, and vegetation, as well as their combinations, and a marine water class, which define eight classes. K-means clustering [16] was used to partition the n-observations into k-clusters, in which each observation belongs to

the cluster with the nearest mean, resulting in a prototype of the cluster. The result of applying this algorithm to the image in **Figure 21** is shown in **Figure 22**, where the classes have been color coded.

**Figure 20.**  *Spectral signatures of hydrothermal alteration minerals. (a) Oxides and (b) hydroxyls. Adapted from [15].* 

#### **Figure 21.**

*RGB composition of the band ratio process: vegetation (B5/B4, red), oxides (B4/B2, green), and hydroxyls (B6/B7, blue). Vegetation pixels and oxide pixels are abundant, while a few cyan colors are observed, which represent the combined response of oxides and hydroxyls to hydrothermal alterations. The fuzzy appearance is due to the scale used, which begins to show individual pixels.* 

*Radiometric Mapping of Hydrothermal Alterations in Isla Isabel, Mexico DOI: http://dx.doi.org/10.5772/intechopen.80530* 

#### **Figure 22.**

*Classification resulting of applying the K-means algorithm to the band ratio and RGB image in* **Figure 21***.* 


#### **Table 3.**

*Type, class, and the number of pixels in each class in the classified image of* **Figure 22***, obtained with the command "Area" in IDRISI.* 

As can be appreciated from these results, oxides and hydroxyls (Class 8) are confused with the water of Lago Crater. From **Table 3**, we can see no pixels associated with vegetation or oxides, indicating the need for a better class identification scheme.

#### **5.2 Principal component analysis**

 Principal component analysis (PCA) is a procedure to decorrelate a set of original variables by means of orthogonal transformations [17]. The principal components are linear combinations of the original variables, and it is expected that only the first ones contain the largest variability, obtaining a decrement in the data dimensions. IDRISI [18] contains a module to calculate the resulting matrix; each band of the cropped image (**Figure 2**) is loaded into the program. **Table 4** shows the result of the calculation. As an example of the use of the matrix, we recall the band ratio for the oxides: B4/B2. Across B2 and B4, we select the two most distant values from zero value (positive and negative), finding this condition in C4, B2, and B4; thus {B4/B2} = 0.100/−0.159. For hydroxyls, we used band ratio B6/B7, finding the above


*Rows show the original bands and columns show the principal components. Blue, green, and red identify selections.* 

#### **Table 4.**

*Weight matrix of the principal component analysis.* 


#### **Table 5.**

*Classification of pixel type obtained with the K-means algorithm for the PCA.* 

condition along C6. This criterion is not fulfilled for the ratio B5/B4 representing vegetation; to obtain a false color image for PCA, we introduce it separately.

A PCA representation of vegetation, oxides, and hydroxyls (RGB) similar to that in **Figure 22** is obtained (not shown). The K-means algorithm was also applied to that image, and the results appear in **Table 5**. Vegetation and oxide pixels are now present although the combined vegetation & oxides class could not locate any. The ocean classes show a new, large area NE of the island that is tentatively identified with the existence of basalt flows at the shallow ocean bottom. The classification result is shown in **Figure 23**.

#### **5.3 Crosta technique**

A variant of the PCA is the oriented principal components (OPC), also known as the Crosta Technique [19]. This technique consists of subjecting the specific bands of a given type, and bands not associated with it, to a PCA. A new PCA was calculated with the bands representative of vegetation (B4 and B5) and the band showing the least possible correlation, which in this case is Band 7. **Table 6** is similar to **Table 4**, but with only three components; the band ratio (B5/B4) can be

*Radiometric Mapping of Hydrothermal Alterations in Isla Isabel, Mexico DOI: http://dx.doi.org/10.5772/intechopen.80530* 

#### **Figure 23.**

*The K-means algorithm was applied to the PCA. The class ocean (light brown) is observed near shore and at a large lobe on the NE portion of the figure corresponding to the rim of the island where basalts are exposed at the shallow bottom.* 


#### **Table 6.**

*Weight matrix for the Crosta technique applied to vegetation.* 


#### **Table 7.**

*Weight matrix for the Crosta technique applied to oxides.* 


#### **Table 8.**

*Weight matrix for the Crosta technique applied to hydroxyls.* 

#### **Figure 24.**

*The K-means algorithm was applied to the results of the Crosta calculation to obtain the number of pixels in each type. Two classes of ocean pixels are distinguished here, the one with the lighter tone corresponding better with the area of the clearly visible submerged basalts.* 


**Table 9.** 

*Classification of pixel type obtained with the K-means algorithm for the PCA of the Crosta technique results (***Table 6***).* 

obtained from column C2. Additionally, two independent analyses are performed for oxides and hydroxyls; for oxides, we take bands 2, 4, 5, and 7 to avoid mapping hydroxyls. The corresponding weight matrix appears in **Table 7**. For the hydroxyls, we used bands 2, 5, 6, and 7 to avoid mapping oxides. The corresponding weight matrix is shown in **Table 8**.

 A RGB color composition can now be made with the principal components that represent hydroxyls, oxides, and vegetation (not shown), as previously done with the band ratios in **Figure 24**. The K-means algorithm was applied to it; the result is shown in **Table 9**.

Pixel identification resulting from the Crosta Technique for vegetation and the PCA results for oxides and hydroxyls show that hydroxyl identification improved with respect to the PCA results. We applied the K-means algorithm to this procedure to define the number of pixels that correspond to each type. Results appear in **Table 9** and **Figure 24**. This result shows class ocean (light brown) mainly

*Radiometric Mapping of Hydrothermal Alterations in Isla Isabel, Mexico DOI: http://dx.doi.org/10.5772/intechopen.80530* 

 around the island and in a large lobe NE of the island, interpreted as shallow-depth (≤10 m) regions that contain visible basalt flows at the bottom. With this algorithm, we obtained a larger number of pixels of oxides and hydroxyls than those obtained with the band ratio method. However, it still cannot properly identify regions of vegetation with oxides or hydroxyls.

 Although this is an improved version of the band ratio method, we also observe that the Crosta algorithm has not been able to fully identify the different pixel types, as shown in **Table 7**. Nonetheless, this algorithm shows improvements with respect to the former since it has distinguished more pixels of the oxide-hydroxyl mixture (Classes 5, 6), which is the objective of this work. Additionally, it shows some yellow pixels representing the mixed type of oxides and vegetation, only present in the band ratio and in the Fraser technique, to be discussed later.

## **5.4 Fraser technique**

 This technique uses input bands as the band ratios that highlight the spectral characteristics of the materials of interest [20]. The idea is to separate the spectral differences of the materials, accomplished via the two resulting eigen vectors. In this process, a PCA analysis must be performed twice, one involves the band ratio


#### **Table 10.**

*Weight matrix for the Fraser technique applied to vegetation and oxides.* 


#### **Table 11.**

*Weight matrix for the Fraser technique applied to vegetation and hydroxyls.* 

for vegetation and the band ratio for the oxides. The second involves the band ratio for vegetation and the band ratio for hydroxyls. **Tables 10** and **11** show the corresponding results. To continue with the analysis, the associated RGB color composition image was prepared (**Figure 25**). Subsequently, the K-means algorithm is applied to the RGB color composition of the Fraser technique (**Figure 26**).

#### **5.5 Lineament extraction**

In geological structures, surface lineaments are often generated by deeply seated processes; such is often the case of faults and some mineral deposits. From satellite images, one can extract surface lineaments that may be associated with a given problem. In the present case, we can explore the association between surface lineaments and the hydrothermal alterations already mapped. To this end, we used Band 6 of the cropped Landsat image (**Figure 2**) and applied the four directional filters: N-S, E-W, NE-SW, and NW-SE, which represent the Freeman Code of Eight Directions (FCCE) [21]. The traces of the obtained lineaments are superposed to

#### **Figure 25.**

*RGB color composition image for the Fraser technique: vegetation (red), oxides (green), and hydroxyls (blue) including a set of lineaments (black lines) obtained from band 6 of the image, by means of directional filters.* 


#### **Table 12.**

*The K-means algorithm is applied to the color composite image of the Fraser technique (***Figure 25***).* 

#### **Figure 26.**

*The K-means algorithm is applied to the RGB color composition image of the Fraser technique (***Figure 25***), with the indicated classes.* **Table 12** *shows the number of pixels identified in each class.* 

*Radiometric Mapping of Hydrothermal Alterations in Isla Isabel, Mexico DOI: http://dx.doi.org/10.5772/intechopen.80530* 

 the RGB color composition of the Fraser technique in **Figure 25**. A tendency is observed for the lineaments to follow a predominantly NW-SE direction, aligned with the main body of the island and with the orientation of the underlying laccolith [2], as well as with the main alignment of reported explosion craters [1]. The residual Bouguer anomaly of the island [3] also follows this direction suggesting that regional tectonic mechanisms favor this orientation. A more comprehensive discussion of the relation between lineaments and geophysical properties is beyond the scope of this study.
