You've Got Geochemical Data—Now What?
Organizing The Data
Before you begin evaluating and analyzing your results, you will want to make sure that you have all your data organized and in final form. This process usually involves "cleaning up" and organizing the data in a spreadsheet. At this stage you may want to make sure that sample names are correct, data columns and rows are properly labeled, replicate analyses are averaged, that samples are sorted or grouped, and that other data (e.g., location, age, field data, petrographic data) are integrated with the geochemical data.
Checking Data Integrity
Once the data are in final form, it is a good idea to do a preliminary check of the results before importing them into a database. This is an important, but often overlooked step. It is especially important to do this if values have been hand-entered (as opposed to imported directly from an instrument). You should carefully check all hand-typed values against their original values. If you are working with major element data, you might check that the totals are within an acceptable error (e.g., ??? 1 weight percent) of 100%. All other values should be inspected to determine that they are "reasonable" given what you already know about the rock. For example, the reported values should be geologically reasonable. Many instruments report negative values when interfering element overlaps are present. These numbers should not be reported as "zero" but rather as "below detection" or some other code indicating that the element has been analyzed, but that it's abundance is below the detection limit of the instrumental method. Where values are not available for a particular sample, these should be left blank or labeled with some other code indicating "not analyzed." Some database applications require that all data fields contain a value. In this case, you may need to enter special numeric codes (e.g., 999.999 into empty fields to indicate missing data or below detection.
A good example of comparison of the quality (precision and accuracy) of geochemical data acquired using different analytical methods (ICP-MS, XRF, INAA) can be found in Appendix A: Huang, S. and Frey, F.A., 2003. Trace element abundances of Mauna Kea basalt from phase 2 of the Hawaii Scientific Drilling Project: Petrogenetic implications of correlations with major element content and isotopic ratios. Geochemistry, Geophysics, Geosystems, 4(6).
Using Database & Plotting Software
- IgPet
- GCDToolKit
- PetroGraph
- PetroPlot
Most geochemical data can also be readily evaluated using spreadsheet applications (e.g., Microsoft Excel). However, when working with large or complex data sets, specially-designed geochemical database and plotting software offer several significant advantages:
- database software are commonly designed to sort, select, and group samples using Boolean arguments,
- database software often permit the assignment of symbols to samples,
- geochemical plotting software often contain pre-defined standard plots for classification and tectonic discrimination that include pre-defined field boundaries, and
- some database applications also include integrated packages for modeling.
What's Your Hypothesis?
When we analyze rocks, we do so in order to answer a question or test a hypothesis. In most cases, the question or hypothesis is the motivating force for the entire study and determines which samples are collected, how they are collected, and the analytical techniques that are applied to them. Like the other stages of a study, the analysis of geochemical data is driven by the question that is being investigated. That's how it is supposed to work anyway. All too often individuals choose the shotgun approach that involves plotting every element and every element combination versus every other element combination. Although one can sometimes "discover" useful relations using this approach, it is highly inefficient. A better approach is to begin with a clear question or hypothesis. What question(s) do you wish to answer? What is your explanation (hypothesis) of the phenomena being investigated? The degree to which you can clarify these largely determines how well you will be able to complete the following tasks below, including: (1) developing a strategy, (2) assigning samples to groups and assigning symbols, (3) the selection of elements for plotting, and (4) the choice of the type of plot.
Develop a Strategy
Once you have one or more clearly stated hypotheses in mind, its time to begin developing a strategy for testing those hypotheses. This requires some general knowledge of the processes or sources being investigated, the chemical behavior of elements, and the samples themselves. One of the best ways to develop a strategy is to look at how other people have approached similar problems. You might consult a textbook, a special volume, or journal articles for ideas of the kinds of arguments that others have been used. Carefully consider what chemical evidence or relationships are most appropriate for testing your hypothesis. This may seem like a time-consuming step, and it is, but you will save yourself a lot of time later if you invest in it now!
Assigning Groups and Symbols
When we make a geochemical plot to test a hypothesis, the "answer" often comes in the form of data trends for which the significance depends on spatial, temporal, or other information inherent to the dataset. For example, when investigating mechanisms of crustal contamination, it is important to determine whether the effects of contamination correlate with the age of magmatism or the degree magmatic evolution. Importantly, the most primitive melts in a system may not be the earliest. In order to investigate these multiple working hypotheses, one needs a way of testing the chemical effects of contamination against the earliest magmas in a series, and against the most primitive magmas. In order to do this kind of testing over and over again while analyzing geochemical data, it is most efficient to plan how to divide samples into meaningful groups. These groups might be based on geographic distribution, geologic units, rock types, or stratigraphic height. Once the samples are divided into meaningful groups they can be assigned a symbol for plotting purposes.
Show Caption
HideThe choice of symbols is important because symbols can be used to convey information about the relationships between samples. Before you begin selecting symbols, consider the final output and use of your plot. Will it be in color or black and white? Will it be a computer monitor or PowerPoint presentation, or will it be a color poster. These considerations will affect your choice of symbol size, color, and type.
Next, you will want to consider what relationships among the samples are important to understanding the significance of the data. For example, in a study of a layered intrusion, the samples might be related to each other by stratigraphic height (meters of base of intrusion) as well as rock type (e.g., ultramafic, mafic, intermediate, and felsic compositions). Interpreting the significance of most geochemical plots will depend on having information about the compositional and stratigraphic relationships of these samples. Which variable do you think is more significant, stratigraphic height or rock composition? Is there a correlation between rock composition and stratigraphic height such that one variable is a proxy for the other? Are some samples more altered than others? Do different rocks exhibit different mineral textures that might be significant to the question being investigated? In any case, you will want to think carefully about the assignment of symbols. Using a combination of symbol shapes, colors, and shading one can often devise schemes that accurately convey the complex relationships among a group of samples. Also keep in mind that for samples that have no relationship to one another, you will want to choose symbols that are significantly different. For example, don't use filled circles and filled squares to distinguish two groups of samples. Instead, you could use filled circles and crosses or open triangles so that they cannot be confused.
All Elements Are Not Equal
- Element Mobility
- Recognizing Alteration Effects
Classification
Choosing the Right Elements for the Task
- Elemental Behavior
- Process Effects
- Recognizing Sources
Choosing A Plot Format
- Bivariate Plots
- Ternary Plots
- Using Elemental Ratios
- REE and Trace Element Diagrams
- Normalizing Values
- Choosing The Element Order
- Tectonic Discrimination Diagrams
- Working With Isotope Data
- Comparisons With Previous Work
- Comparisons With Numerical Models
Presenting Your Results