How statistical modeling and data visualization help judges, juries, law enforcement and environmental agencies understand the evidence needed to bring arsonists to justice and to hold accountable those responsible for environmental contamination.
Challenge
Arson and environmental contamination cases are notoriously difficult to prove. For arson cases, much of the evidence is destroyed and masked by the fire. For forensic investigations, years of historical, environmental and geospatial data need to be sifted through and analyzed to identify the responsible party.
Solution
Experts at Chemistry Matters use a variety of statistical tools in JMP® to build a chemical profile from environmental samples. The platform’s interactive data visualization capabilities help communicate key findings to lawyers, judges, juries and environmental regulators so that they fully understand the nature and meaning of the forensic evidence.
Results
Chemistry Matters brings scientific and statistical rigor to legal and regulatory hearings, where an evidence-based approach is critical to ensuring that responsibility is appropriately allocated.
Every year, thousands of wildfires ravage Western Canada. Hundreds are set intentionally, but the arsonists who start them may evade prosecution.
“Arson is very hard to prove,” says forensic chemist Court Sandau, PhD. Compared with other crimes, it can be difficult for prosecutors in arson cases to definitively link evidence to an individual perpetrator, he explains. Accelerants like gasoline are readily available to the public, and tracing residue from a crime scene to the site of purchase can be challenging. However, advances in forensic chemistry are now tipping the scales in the investigators’ favor.
“We are developing a new chemometric approach to automatically analyze fire debris samples to determine what type of accelerant was used and identify a likely source,” Sandau explains. He and his colleagues at Mount Royal University and Chemistry Matters, a research consultancy Sandau founded in 2011, are now using this method to help catch and convict arsonists.
Sandau’s innovation lies in combining comprehensive two-dimensional gas chromatography (GC×GC) with time-of-flight mass spectrometry (TOFMS). This approach provides far better rates of detection than previously available methods for arson cases in which crime scene samples contain low concentrations of signal. Whereas traditional accelerant analysis looks at between 30 and 50 organic compounds per sample, Sandau’s GC×GC-TOFMS method creates a profile of up to a thousand unique chemical compounds. “It’s like DNA,” he explains. “I can fingerprint something precisely with that much information.”
That fingerprint, Sandau says, can then be matched to possible sources with statistical analysis – identifying, for example, which gas station an accelerant was purchased from. Such findings are essential to police investigators who then use this information to identify suspects from a store’s purchase records and surveillance footage. Ultimately, that means more arsonists are brought to justice.
In environmental forensics, chemistry really matters
Sandau got his start at an Environment Canada lab in Ottawa studying the industrial contaminant polychlorinated biphenyls (PCBs) in polar bears, seals, bald eagles and humans. Later, he began using big data to study human exposure to environmental contaminants for the US Centers for Disease Control (CDC) in Atlanta. After returning to Western Canada in the early 2000s, Sandau threw his energy into forensic chemistry consulting and Chemistry Matters.
Now with a dozen chemists and data scientists on staff, Chemistry Matters offers research, data science and consulting services in geoforensics, environmental forensics, arson investigations and biomonitoring. The firm also provides scientific support and expert witness services in environmental litigation. “We take big data and chemistry data sets, do multigrade statistics and then communicate [those findings] to the court,” Sandau explains.
“Our clients bring us in to develop an experiment, figure out what kind of measurements we need to prove or disprove [the matter under investigation], and then take those measurements and try to make sense of them.”
By bringing statistical and scientific rigor to civil and criminal investigations, the firm aims to promote environmental stewardship and reduce risks to human and wildlife health. A big piece of that mission is the team’s work on the cleanup of United States Superfund sites.
Upholding the ‘polluter pays’ principle
First established in 1980 and administered by the Environmental Protection Agency, the United States Superfund program is a federal effort to systematically investigate and remediate land contaminated by hazardous waste. Under the Comprehensive Environmental Response, Compensation and Liability Act, any company responsible for environmental contamination must pay for its cleanup – a policy that has come to be known as the “polluter pays principle.” Identifying the responsible party, however, can be a long process involving years of research and court proceedings.
That is where Chemistry Matters’ expertise is crucial. In the firm’s Superfund projects, Sandau and his team may gain access to data dating back 30 years or more. “We have to compile all this data, examine the chemistry behind it, see how things changed over time, assess the current state of contamination and figure out who is responsible,” he says.
For example, if landowners believe their wells have been contaminated by fracking – or if there is an oil spill – both parties to the dispute would turn to chemical data to prove or disprove the claim in court. Chemistry Matters helps its clients not only assess and evaluate historical data, but also identify areas where they need additional data to create a more complete picture of environmental contamination and its source. In such cases, the firm then provides guidance to clients as to what data they needs to collect, and how and where to collect samples.
When the historical records, aerial photography, witness statements and chemistry data agree, Sandau says, they’ve likely uncovered the truth. “Chemistry data shows us when a [contamination] process started. It’s an historical archive of what went on at that site…. If these companies are still around, they can be brought to trial with the polluter pays principle. They caused it, so they have to pay for the cleanup.”
Statistical models identify those responsible for contamination
Chemistry Matters’ source apportionment process uses statistical modeling to determine who is responsible for different contaminants found in the environment. This approach begins with data review. “I'm a big proponent of looking at your data right at the start, because if there's something wrong with it, it will stick out like a sore thumb,” he says.
A longtime SAS® user, Sandau began using JMP® for data exploration, visualization and analysis after recognizing its potential as a replacement for graphical features in Excel. “It’s visual and intuitive,” he explains. “The true power of JMP lies in data integration and being able to learn from the data set.”
For example, a standard CDC data set called NHANES provides an evaluation of contaminants in the US population. It’s a data set that, in order to produce the analysis Sandau’s clients require, must be normalized for the number of people each sample represents using population statistics and techniques like jackknifing that are not possible in Excel.
In most projects, the Chemistry Matters team looks for geospatial associations between samples using a dataset of GPS coordinates. They import data into JMP to analyze it for distributions and outliers, and then apply various statistical treatments. Once the data has been cleaned and the team has a map of samples, they’re ready to explore what the data is telling them.
“We’ll do a principal component analysis or a hierarchical cluster analysis and start color-coding it,” Sandau explains, adding that these applications allow environmental chemists to look for evidence of contamination without any preconceived knowledge of why contaminants are where they are. “That’s my favorite thing about JMP – the interactivity. We truly do learn about our data through JMP.”
Divergent data treatments may obfuscate the truth
While complex statistical modeling may be crucial to discovering the truth, however, proving it in court is another matter. Sandau explains that when facing judges and juries, experts should lead with the story – not the statistics.
“All the lines of evidence should agree. There shouldn’t be any loose ends, and that goes beyond just the statistics,” he says, explaining that in litigation cases, both sides have access to the data. But that doesn’t mean they always get the same results; data can be presented in different and sometimes misleading ways. It is therefore crucial that Sandau and his team understand how alternative treatments of the data arrived at a different conclusion so that they can find where the opposing interpretation of the data is wrong.
“Data can be displayed however you want,” Sandau says. “But I have to prove that I’m doing it in a scientific fashion. And I have to be transparent about how I came to my conclusions by communicating visually.
“That’s where I get my clients involved with JMP: to be able to load my data into JMP and click on a spot and show them where it is and how [we arrived at our] interpretation. That’s the power of visualization: communicating the data, learning from the data and allowing our clients to see where the data comes from.”
Communicating data insights to judges and jury
More often than not, environmental cases hinge on scientific evidence. Yet the lawyers tasked with arguing these cases in court may lack the scientific expertise to sufficiently explain the data. Extensive preparation with experts like Sandau helps them to not only prepare their argument, but also to pose informed, intelligent questions of the opposing counsel’s expert witnesses – and demonstrate to judges and juries any gaps of logic in the opposing viewpoint. That’s why it’s so important, Sandau explains, to teach clients about what the data shows.
“We use JMP as a teaching tool because it enables us to show chemical fingerprinting in ways that [those without a background in chemistry] understand,” he says. “JMP allows us to communicate to lay persons the connectivity of information so that they understand our conclusions…. And because I can do this in real time and interact with the data [in platforms like Graph Builder], it's much more convincing. They start to think, ‘I may not understand what he's doing, but I certainly see how he has connected all those dots.’ Now I’m telling a story rather than trying to explain multigrade statistics.”
Even with these advances, the court system has a way to go, Sandau explains. While legal teams have the advantage of using statistical tools like JMP to prepare their arguments outside of the courtroom, judges and juries may still have to flip through pages and pages of legal filings where data tables are disconnected from witness testimony. And it can be difficult for the lawyers to discuss and dispute opposing interpretations of the data since they cannot interrogate it in real time.
“The Achilles' heel of the courtroom is always clarity on what we're talking about,” Sandau says. “[Judges and juries] usually only see a sample here or there, and the data is always [static]…. If I had my druthers, I would have a data analysis tool right there in the courtroom. It would be much better to say, ‘OK, you have a question about a data point? Let’s pull it up. Let’s have a look at it.’
“It’s my goal to get there [with our legal system]. There’s such power in being able to show visually what the data is telling you.”