I love graphics. The information density of a good graphic is tremendous. Unfortunately that information density, and frequent use of terms-of-art can easily preclude a meaningful understanding of a graphic, be it a bar chart or a map. I've been there, you look at a map and think to yourself, "well, that's neat", but you don't really gain from it. I know I'll be guilty of this. In particular, I love maps, but I'm horribly rusty at making them look nice. Essentially, this paragraph is my disclaimer. My maps and figures may look confusing at first, but I'll attempt to thoroughly explain them and improve from one set to the next. I'll also take any requests for data or more specific maps.
That said, this first map is in awful, awful violation of those principles.
The above map is immediately colorful, has an incredibly weird set of breaks, too many breaks, a typo in the title, a somewhat inscrutable title, and is kind of an all-around train-wreck.
What data went into this map?
- Monthly electricity consumption data for each state
- Population of each state
- Shapefile of the states
What happened to the data to result in that abomination of a title?
- The electricity consumption for each month in each state was divided by the state's population in 2011.
- Using this data a national average of consumption was created for each month.
- Each state's consumption per capita per month was put in terms of the national average as a % difference.
What this map is showing is how much electricity is consumer per capita in a state each month compared to the national average. Except it's every state at once, so it's frustrating to track and distracting to look at.
While the EIA also has monthly reports of cleaned up data, I wanted to practice working with text in Matlab and matching together several values while looping through the data to do some basic calculations and output a decent matrix to join with the database file of the states layer I was using in Arcmap. The generation data reported by utilities to the EIA comes in a monthly format. By linking each utility to a specific state I tracked electricity generation month-to-month for the United States, and mapped the change between each month. If January's generation for Texas was 100GWh, and February's was 75GWH, then Texas would have a -25% change for January-February, and appear as a light green on the map.
The trends evident on the map are largely seasonal, the demand for air conditioning is readily apparent in the eastern and southern halves of the country. New Hampshire and Vermont present an interesting anomaly in October. In 2011, the east coast experienced a massive snowstorm, putting a significant chunk of generation offline. This change is reflected in the map.
Consumption follows a similar pattern.
Hopefully I can come back after midterms and offer a little more analysis for these maps. Honestly, they're not meant to convey a startling new concept, or make you think differently. These maps are simply a graphical reflection of annual electricity use that I hadn't seen presented before.
My basic workflow on these maps goes:
- Excel: Glance at the data, check for gaps, make sure multiple datasets will agree in Matlab. Honestly, I'll go back and forth between Excel and Matlab a few times in the beginning to make sure things are lining up. I'm not so confident that I don't check at least part of my Matlab calculations by "hand" in Excel.
- Matlab: Any modifications to or combinations of my data happen in Matlab. I also use it to make any figures that are not maps, although you can draw maps from longitude and latitude data if you really want to. I'll also find the min and the max of any dataset I'm using in Arcmap.
- Excel: Back to Excel if I'm going to be making something in Arcmap. I use Excel to make sure column headers match if I'm going to join with the .dbf file of a layer, and eyeball the data again for mistakes like unit problems or column/row misalignment in Matlab.
- Arcmap: If I'm following the process for the maps above I'll join the numbers crunched through in Matlab to the .dbf file associated with the layer in question. I normally try to join on standardized codes, like FIPS, rather than names, which can be slightly misspelled or just different (particularly for countries). The min and the max I find above come into play here when choosing a series of breaks in order to display the data with chloropleth symbology. I'll set the symbology in the column (month in this case) with the highest value, then apply it to each other column to remain consistent. Too many breaks is confusing and boring, too few smooths the curves and presents far less information. Teasing out decent breaks is something of an art, one which I admit I need work at. Normally I use a mix of manual choice and natural breaks (Jenks) depending on the data. In the end, I apply the symbology form the most extreme column to every time-set of data, and export each map accordingly.
- Photoshop: Finally, after exporting the maps I need from Arcmap as .PNGs I load them into a stack in Photoshop, create an animation, then export it as a .gif.