A2-AthulanVijayaraghavan
From CS294-10 Visualization Fa07
Contents |
[edit] Introduction and Motivation
The visualizations developed for this assignment help understand the environmental effects of extracting and processing different metals. Some of my research involves analyzing the environmental impacts of specific manufacturing processes. Most manufacturing processes begin with an input material which has been extracted and processed into a usable form (for example, to make automobile engine blocks, you start with Aluminum alloy). The total energy consumption and environmental impact of the manufacturing process depends not just on what happens during the manufacturing process, but also on the embedded energy and environmental impacts of the materials which are processed. Hence, its important to understand the relative environmental impacts and embedded energy of different materials. In this report, I have created visualizations of the relative environmental impacts from the extraction and processing of different metals.
The data for this comes from the Ecoinvent Lifecycle Inventory database (link). This database has a particularly poor interface, and it consists of a set of very large Excel spreadsheets. To find the environmental impacts of a certain process, device, or material, the procedure is to open a series of the Excel worksheets and enter the ecoinvent code for the data point one is looking for. Ecoinvent then searches for the data from its many databases (which are also Excel worksheets) and displays it on the screen. This is an especially cumbersome way to search for data, especially when one is looking to compare several entries from the table. Since ecoinvent lacks a visual interface, comparison of data is very difficult. For any sort of meaningful comparison, data needs to be extracted into its own worksheet (if excel is being used) or into some other software package (the approach in this report).
[edit] Question being Answered
The basic question I am trying to answer from the data set is: What are the relative environmental impacts and resource consumptions of the extraction and processing of common metals? Since Ecoinvent has data on different ways of extracting metals, I would also like to know how different methods of extracting the same metals compare. Finally, since we also know from Ecoinvent the energy/impacts for processing operations for a few metals (processing happens after a metal has been extracted), we would like to know how this compares to the energy/impacts for their extraction.
[edit] Setup
Ecoinvent stores all its data in Excel worksheets, which makes it very convenient for use in Tableau and Spotfire. It lists the different ways of processing metals in rows and the different impacts as columns. The problem with this is that we need some sort of aggregation for the visualizations to be meaningful. The image below shows how the ecoinvent listing was broken down so that the data could be easily aggregated. This was applied and a total of 23 metals with a total of 67 extracting/processing steps were studied.
Similarly, ecoinvent lists multiple impacts in their own individual columns. These impacts are also unaggregated. We do not need aggregation in the columns as neither Tableau or Spotfire can distinguish categories and aggregation from different column headings. They both read data in the rows, so having unaggregated data in the columns was acceptable.
But for some of the analysis I wanted to have some aggregation of the columnar data. This was similarly aggregated and an example is shown in the table below.
But, as mentioned earlier, Tableau/Spotfire does not recognize aggregation in the column headings, so I transposed the excel table when I needed aggregation of the impacts. Hence, there were two sets of tables each with aggregation in the rows so that specific visualizations could be created. Each of these tables was used for different analysis. If only one table was used, then the analysis would not be as effective as some of the data would be missing - hence I went along with two.
[edit] Table 1
Table 1 was used to study a limited set of impacts of all the metals extraction/processing steps. The steps were classified by type (extraction or processing) and the metal name. The table listed the names of different metal extraction processes in the rows and the impacts in the columns. The first column was the name of the metal, the second noted if it was a processing or extraction step, and the third column listed the name of the process. The rest of the columns were data values of the impact for the operation. I did not use all the impacts from ecoinvent as this would have been excessive and unmanageable. The following impacts were listed in the columns:
- upper limit of net GWP
- abiotic stock resources
- emissions into air
- emissions into soil
- emissions into water
- total non-renewable usage
- total renewable usage
[edit] Table 2
Table 2 was used to study a broader set of impacts for one specific extraction step for each metal. The impacts were classified by type (emission or energy usage etc.,) The table listed the names of the various impacts classified by impact type. The first column was the type of impact being measured, and the second column was the name of the specific impact. The remainder of the columns was the data values pertaining to each metal. For these columns, I chose to not list all the metals, as the data would have been unreadable (both in the file and in the visuals) due to the lack of aggregation. I instead chose to only list the impacts from the extraction process which produced the most common type of that metal (this is called the Primary metal).
[edit] Analysis
Analysis was performed using both Tableau and Spotfire. It was easier to setup the analysis in Tableau, but dynamic queries were easier with Spotfire; hence both packages were used. The analysis was performed on both data tables, and is as follows.
[edit] Table 1
This table was used to see the relative energy consumption in processing and extracting different metals. The chart below lists the usage of Non-Renewable Energy (from fossil and nuclear fuels) for the extraction and processing of three common metals: Aluminum, Copper, and Steel. The colors correspond to the type they are processed into (note, not all metals go through all processes) and the shapes denote if its a extraction or processing step. We can see that Aluminum generally has the most energy intensive processes, while Steel has the least.
This chart was created in Spotfire (to make use of the dynamic lookups), and marking some data points populated a bar-graph which compared the energy usage more explicitly. The column colors correspond to the process names (same as in the scatter plot) and the columns are organized by metal.
Finally, a scatter plot also was populated with these data points, which compared the non-renewable energy usage with the Global Warming Potential (GWP) from these processes. A linear trend is expected, as fossil fuel usage corresponds linearly with process GWP, and this is seen in the chart.
The remainder of the analysis was done in Tableau. The next question was, can we look at more plots comparing fossil fuel usage and GWP for different processes. The chart below plots this relationship for extraction/processing of Aluminum, Copper, and Steel. We can see that there is a general linear trend, with Aluminum clearly dominating in most impacts. It is also interesting to note that all the extraction steps are generally more energy/GWP intensive than the processing steps, which is expected. Also, we can see that the processes for Steel consume markedly less energy than for Aluminum. A couple of processes for Copper break the linear trend, by having a larger GWP than the trend would suggest, and this is because of the specific gases expelled in extracting Copper.
Clearly the previous chart had too much data. The following multi-plot makes it easier to compare specific extraction processes for Iron. The plots look at fossil fuel (non-renewable) and hydroelectric (renewable) energy usage which goes in making steel using these different processes, and compares this to air emissions, soil emissions and GWP for these processes. In all the cases a linear trend is seen, and we can quickly identify that extracting Iron into Cast Iron is much more intensive than other processes. We can also see which processes have a greater impact on air emissions than soil emissions (Pig Iron) or vice versa (Iron Sulphates).

[edit] Table 2
We can now move to Table 2 where we can ask more detailed questions on resource consumption and impacts. The chart below shows relative fossil fuel consumption for extracting different metals using into their most common form (Primary Metal). Silicon and Mercury dominate these plots, as they have the highest embedded energy.
But we can do much more with the data as this analysis was possible with Table 1 itself. Then next chart shows a stacked-bar comparing the sources of energy for different metals. The bar colors indicate the source of energy. We can see that all metals make sparing use of renewable sources of energy, and use mostly fossil fuel energy and nuclear energy. It is interesting to note that nuclear energy figures so prominently in the split-up. This is because ecoinvent data comes from Europe, where nuclear power is more common.
We can also drill down deeper into the data and compare resource usage for extraction of different types of Primary Metals. The chart below shows the relative consumption of coal, oil, and natural gas for extracting different primary metals. Silicon consumes much more Natural Gas than the others, and hence dominates the chart.
The data can also be quickly searched for metals which have really large resource consumption. Note how in the chart below which compares resource use for Silicon, Platinum and Palladium, Silicon is completely dwarfed by the other two metals.
[edit] Summary
The analysis with the packages did not lead me to any non-obvious or hidden results (but I was not expecting this either) as the trends seen from the visuals have been well established in the research community. But, the visuals themselves are quite powerful, and as both packages allowed for some amount of dynamic querying (Spotfire more so, with the markings) it made the data really come alive. This kind of analysis (environmental number crunching) is usually quite boring, but the ease in creating these visuals made the analysis engaging, and also helped make the results more memorable. (As a side note, some of my lab-mates have been quite taken by these packages and we will most likely order an evaluation license for our lab.)
