Magnetics & Remanence

The uses of magnetics

Many of the volcanism-related geological systems have magnetic signals. These signals can be measured and interpreted to help us better understand how new oceanic crust was formed, what the plumbing system of a volcano looks like, where mineral deposits are, etc. I have collected a few examples of magnetic anomalies from different geological setting in Figure 1. The first example in Figure 1(a) comes from the TAMU massif where we can clearly see the linear magnetic anomalies (LMA), based on which Sager et al. (2019) propose that oceanic plateau at Shatsky Rise formed by seafloor spreading volcanism, i.e., the same process responsible for the formation of oceanic crust, as opposed to centralized eruptions. The authors were also able to reconstruct the formation history of the TAMU massif based on the spatial patterns of these LMA.

The magnetic anomalies in Figure 1(b) are from the intrusive bodies in the Auca Mahuida volcano located east of the Andean thrust front in the Neuquen basin (Argentina). These anomalies are caused by both reversely magnetized sources and the normal polarity sources (Pain et al., 2016). Magnetic modeling helps define the geometry of the shallow plumbing system of volcanoes with remanently magnetized sources, and estimate the depth and geometry of potential oil reservoirs in volcanic areas.

Figure 1(c) shows the magnetic anomalies due to volcanic units in a sedimentary basin in Songliao Basin in Northeast China (Li et al., 2012). These volcanic units are the targets for natural gas exploration. Inversion of magnetic amplitude data helps identify the volcanic units for subsequent targeting and delineation through a drilling program.

The magnetic anomalies in Figure 1(d) are mainly caused by igneous intrusions in a Precambrian basement in Northwest Iowa Intrusive Complex near the Midcontinent Rift system (MRS). Interpretation of the magnetic data (Drenth et al., 2015) and 3D modeling (Sun et al., 2020) have resulted in a better characterization of the structures and compositions of the basement, and provide guidance for future drilling activities and follow-on geophysical data acquisition.

Another example comes from an iron-oxide-copper-gold (IOCG) deposit in the Carajás Mineral Province, Brazil, as shown in Figure 1(e). 3D magnetic inversion (Leão-Santos et al., 2015) helps identify the mineralized zones for future drilling.

Magnetic data are also useful for diamond explorations. The magnetic anomalies in Figure 1(f) are from DO-27/DO-18 kimberlites from the Tli Kwi Cho (TKC) kimberlite complex in the Northwest Territories of Canada (Devriese et al., 2015).

Challenges of interpreting magnetic data

Magnetic data contain so much useful, and most often, unique information that they are heavily relied on by geophysicists and geoscientists to help understand many intriguing subsurface geological systems. However, interpreting magnetic data is not a trivial task. It often requires highly advanced and sometimes sophisticated mathematical tools.

The difficult is largely due to (1) the fact that most of the interesting and useful magnetic signals are caused by remanent magnetizations, instead of induced magnetizations, and (2) the fact that, more often than not, we have very limited knowledge, usually zero knowledge, of the directions of the remanent magnetizations.

Magnetic materials of a rock were magnetized when exposed to an external magnetic field. We use the term 'magnetization' to describe such a state of a rock. Magnetization is a vector and has a direction and a magnitude. Usually, magnetization changes its direction and strength when the applied magnetic field changes its directions and strength. This type of magnetization is called induced magnetization. However, if a rock contains ferromagnetic materials (such as basalt lava flows), it turns out that such rocks can preserve the direction and intensity of the magnetization regardless of the changes in the external magnetic data. This type of magnetization is called remanent magnetization or remanence. Therefore, remanence reflects the state of a rock (and the Earth's then magnetic field) when it was magnetized millions of years ago.

Many of the magnetic data processing techniques require an accurate knowledge of the remanent magnetizations. The notable examples include reduction-to-pole and pseudo-gravity transform. The commonly used 3D susceptibility inversion methods also assume weak induced magnetization and ignore the remanence. Numerous research works haves shown that, when remanence is not accounted for properly, all subsequent interpretations become questionable and the inverted anomalies bodies will have incorrect depths, shapes and susceptibility values.

Some of my work on interpreting magnetic data complicated by remanence

Together with my co-author Dr. Yaoguo Li, I have developed a new 3D inversion method that directly invert the magnetic measures (usually, total-field anomalies) for a 3D distribution of magnetization vectors. This is a highly under-determined problem because we are trying to recover three 3D functions from a 2D data set. Additional information and constraint need to impose on the inversion. Otherwise, the inversion solution would not be geological meaningful.

Our basic philosophy is that magnetization directions should show a certain degree of spatial coherence or consistency. That is, we expect two magnetization vectors to show highly similar directions if they are spatially close to each other.

To promote region-wise consistency among the recovered magnetization directions, Li and Sun (2016) develop a highly constrained magnetization vector inversion (MVI) by combining an unsupervised machine learning algorithm, fuzzy c-means (FCM) clustering, with the generic magnetization inversion. This method, termed magnetization clustering inversion (MCI) henceforth, limits the number of magnetization directions that the numerous model cells can take to only a few possibilities through the use of FCM clustering. With additional help from the smoothing regularization used in the inversion, MCI achieves a greater level of spatial consistency among the recovered directions without the need for any site-specific information.

The greatest advantage of our method is that, the region-wise consistency constraint is a generally applicable one that can be applied to any field site. It eliminates the need for any site-specific information. This is especially convenient for areas where no much prior information is available.

To illustrate our method, we have designed a synthetic example with variable magnetizations. The model consists of two magnetic bodies with differing geometry and magnetization embedded in a nonmagnetic background (Figure 2a). Figure 2b shows the total-field anomaly under an inducing field in the direction (I,D)=(65∘,−25∘). The negative anomaly to the north and rotation of the trough away from D=−25∘ in the southern anomaly indicate that magnetization directions of both source bodies are significantly different from the inducing field direction. We used magnetization directions of(Im,Dm)=(−60∘,75∘) and (Im,Dm)=(25∘,45∘)(Im,Dm)=(25°,45°), respectively, for the two causative bodies. These parameters are chosen to emulate the presence of strong remanent magnetization.

We have inverted the magnetic data set in Figure 2(b) using the MCI method by assuming two differing magnetization directions. The recovered magnetization directions are displayed in Figure 3(a), and they exhibit two tight clusters that are well-separated from each other, corresponding to the two anomalous bodies. The recovered magnetization at the depth of 150 m in the 3D model domain is shown in Figure 3(b). We observe that significant magnetizations are spatially coincident with the two anomalous bodies and are dominated by two spatially coherent directions that are consistent with the true magnetization directions.

Magnetics for diamond exploration

We have applied the MCI method to a set of field data acquired on Victoria Island, Northwest Territories, Canada for diamond exploration (Figure 4). The inducing field in the study area is in the direction of I=86.7° and D=26.3°, yet the total-field anomaly exhibits a nearly symmetric negative anomaly in the center surrounded by several small-scale positive anomalies. The source of the central negative anomaly is a kimberlite dike and clearly has a strong remanent magnetization.

For the field data, it is not immediately obvious how many clusters (or, dominant directions) there are in the magnetic source bodies. We have performed several MCI by assuming different number of clusters. The recovered magnetization models are shown in Figure 5. The recovered magnetization directions when assuming 2 clusters are summarized in Figure 5(a) where we observe two well-defined clusters. In the 3D view in (b), we observe that the magnetization vectors point upward in the central region directly below the negative total-field anomaly, whereas the magnetization vectors corresponding to the smaller positive anomalies point downward with a groupwise consistency within the model.

The four different magnetization vector models in Figure 5 all reproduce the observed magnetic anomalies equally well. Without other independent information, we cannot rule out any one of them as a possible geological scenario. In other words, given the limited information we have (except the magnetic data and some limited geology information), these four models are equally valid.

This actually provides us a great opportunity to assess the uncertainties of the recovered magnetization directions. Figure 6 shows a 3D view of the high-confidence magnetic source bodies. The recovered magnetization directions of these bodies are much less affected by the assumed number of clusters. When viewed from a plan view, these high-confidence bodies are highly correlated with the closures of the magnetic data contours, as shown in Figure 7.

Magnetics for oceanic plateau studies

I have also extended this method to the study of oceanic plateaus which are massive underwater volcanoes that often rise several kilometers above the surrounding seafloor with areas as large as millions of square kilometer. However, the formation and evolution of oceanic plateaus are poorly understood, because oceanic plateaus are typically found many kilometers underwater and thousands of kilometers away from land, making it difficult to sample these massive oceanic features and collect geoscientific data. Magnetic data are among the limited types of data that are available for oceanic plateau studies.

Figure 8 shows the interpolated marine magnetic data over Ori Massif in the northwest Pacific Ocean (left) and the effective susceptibilities recovered from MCI when assuming two dominant magnetization directions. I am currently working with my colleague Dr. Will Sager to refine the magnetic inversions and to develop an understanding of the formation of Ori Massif.