Choropleth & Graduated/Proportional Symbol Mapping

Data classification appears to be an ongoing theme here- and with good reason.  Simply put: it is not simple, nor easy to do.  Choropleth maps sound very complicated, and most people are not familiar with the term "choropleth," and as such it is a word I like to throw around in casual discussion of my classes. It makes the work I'm doing sound very difficult and technical (which it is, to some degree).  The truth is that it's a term used for a map that displays some characteristic within administrative boundaries, or enumeration units.  Graduated or proportional symbols are another method of displaying some characteristic, with their placement on a map coinciding with their incident geographic location.

Choropleth maps are, as the one here, often used to display population characteristics based on administrative units, such as countries, states, cities, etc.  The population density in the uppermost frame is shown in shades of green, with the highest density values being the darker shades, and the values decreasing with the lightness of the color.  The bottom frames, with the population percentages by gender, follow this pattern as well- darker shade means higher value, lighter means lower.  With this type of map population patterns are easily revealed, as the top map clearly shows higher population density in western Europe in comparison with the countries in the east.  Again, as we found last week, the classification of the data values is of paramount importance here.  We want to display on the map what is actually true in the real-world, and thus must take care to place each country in a class with members as alike to each other as possible, and as different to values in other classes as we can.  The top map had a few statistical outliers in the population density measure, as there are countries in Europe that are very, very small (Vatican City, Malta, etc.) and have very, very high density values.  These anomalous nations are not visible in a map of this scale, and so are placed in a class with a few countries with much lower values, but easier visibility on the map.  The wine consumption, in liters per capita, is displayed by circles that increase in magnitude as the per capita consumption increases.  Again, placing the countries in classes of different value ranges helps depict a visual pattern, with consumption generally greater in western Europe.  I employed the natural breaks method of classification for that, as I feel it is an accurate way to group the countries by that value.  

All in all, mapping statistics isn't the simplest thing to do, but, like anything else, it becomes easier with practice.  This week's maps seemed daunting at first, but through creating them I can honestly say I have, once again, broadened my understanding of this subject.  I will probably still occasionally obliquely refer to the complicated nature of the choropleth map in casual conversation though, as opportunities to successfully slip such arcane and technical terms into a discussion will always be somewhat gratifying.

Classifying Data: for a Map that Makes Sense

Appropriate data classification is one of those things that, though we are mostly unconscious of, can really mean the difference between a map that logically displays data accurately, and one that is imprecise and misleading.  When one creates a map that is meant to display some characteristic phenomena, like population percentage of a specified age, by division of some administrative boundary, like state, county, or census tract, one must choose carefully the categories that each administrative unit is placed in.  The technical term for this type of map is choropleth, and there are several different ways the data categories, or classes, can be decided.  

Through my creation of the above I personally discovered some of the challenges in deciding which classification method is most appropriate for accurate depiction of the characteristic phenomena being mapped.  The 4 maps are all displaying the same raw data- census information on the population percentage of adults 65 and older in Escambia County, by census tract area.  They differ in how they display that data, namely how the different percent values for each tract are grouped and shaded the same, or are classified, to create a map that accurately shows the viewer a general idea of the distribution of this population characteristic throughout the county.  The 2 bottom maps classify the data with the quantile and equal interval methods, which divide the values without consideration of where the natural groupings of values occur within the range.  The upper right map classifies values according to their distance from the mean percentage, and the upper left uses algorithms to determine the most accurate "natural breaks" in values, in order to minimize the difference between values in the same class, and maximize the difference between values in different classes.  Upon close inspection, in my opinion, there aren't many huge differences between the maps, and yet I was asked, as part of the assignment, to choose, and defend my choice of, which is the superior method of accurately displaying the actual data.  This was a bit of a challenge.  I concluded the superior method is the natural breaks, but a decent case could be made for any of the other 3.  This type of challenge becomes especially politically, and emotionally, loaded when one is mapping some characteristic like race, political affiliation, or crime statistics, as the decision to place areas in one category or another can be a cause of consternation for some.  One also has the ability, in certain situations, to present the raw data in a misleading fashion according to how it is categorized- in which situation certain causes can be championed, etc.  All in all, it's a very important lesson for a nascent cartographer to learn, as these issues may not be immediately evident in the process of map creation or assessment. 

Map Projections, la deuxième partie

I seem to recall saying something last week about finally truly understanding the various concepts of map projection, and how projections and coordinate systems are handled in a GIS.  This week I learned I was wrong.  The bravado caused by my hubris was a preemptive celebration, and, though I came to a much, much better understanding of projections and coordinate systems with this week's work, I now know that I've only seen the tip of the iceberg.  I also had the opportunity this week to search for and download some public GIS data, in the form of shapfiles and aerial photos, which was something I'd only ever dabbled in before.

This map is comprised of 2 ortho quads (each made with 4 quarter sections), created by the USGS, and 3 different vector datasets- Florida county boundaries, major roads, and USGS quad boundaries.  The vector line data was in a different coordinate system than the aerial photos, and had to be reprojected in ArcMap to the State Plane Florida North system.  The files would have appeared correctly together in the map document if added together in different coordinate systems, but there would be a loss of functionality- which is something I have personally encountered in ArcMap, and never knew how to "fix" it.  It isn't enough to merely describe the coordinate system in the metadata, ArcMap has to actually change and export the file in a different projection.  The process itself is easy enough, but the theory behind it was a definite challenge for me.  I did, after many hours of work, finally get a decent grasp on the general concepts and theory that exist, and run in the background, to accurately display the 3 dimensional earth on a 2 dimensional surface.  Additionally, I gained some experience in adding x,y data coordinates from a tabular source to create features on a map- which are the storage tank contamination monitoring sites of Escambia county in the map above.  Another procedure that may seem simple and intuitive, but is, in reality, a bit of a complicated process.  

Map Projections (and a long overdue epiphany)

The concept of a map projection isn't difficult to grasp- it's something students in an undergraduate introductory geography class learn as a matter of course.  It's the basic procedure of taking the 3-dimensional, (roughly) spherical surface of the earth, and transcribing it to a 2-dimensional surface to make a map.  In practice, though, there are a myriad of ways this can be done, and all produce results that differ in a myriad of different respects.  To put it simply, different maps, layered together in GIS, will not line up exactly in most cases if they have differing map projections.  The kinds of basic GIS functions I've used in jobs I've held thus far haven't necessitated learning much about how this works within the GIS, and it's been a topic I've long felt both a fear of not understanding, and a trepidation about attempting to learn, lest it be beyond my comprehension.  

With the creation of the above I can say I finally understand exactly how map projections work both conceptually, and within the use of a GIS.  The three different maps of Florida are all done with different map projections, and the difference in area each gives for the 4 selected counties is listed below.  Going through the steps of using GIS to re-project these maps as required, and exploring the inherent spatial differences in each, has (finally) illuminated, for me, the arcane knowledge of the map projection.  Yes, it is somewhat complicated; no, it isn't beyond my ability to comprehend.  I only wish I would've realized this sooner, but I suppose there is a time and place for everything- the truth in which is evident in really any well made map.

Statistics... of the *Spatial* Variety

I know the sentiment probably isn't popular amongst my classmates, but I have always rather enjoyed statistics.  The idea of describing and predicting phenomena, be it natural or man-made, with math is fascinating to me.  Spatial statistics are an especially important facet of geographic science and GIS; describing the "why of the where," as a professor I once had put it, is one of the discipline's prime directives, and cannot be done without statistical analysis.  

The above is a map, created in ArcGIS Desktop, of weather monitoring stations in Western Europe.  This map of points presents a convenient set of data with which to perform some basic spatial analysis of central tendency- such as the calculation of the mean and median centers of station distribution.  The green shaded oval shows the general trend of distribution, using the number of stations within one standard deviation of the mean.  These analyses were relatively simple to perform with GIS, and undoubtedly represent only a small fraction of what the program can do with statistics.  My exposure to these capabilities, up to this point, has been nonexistent, and I greatly look forward to exploring them as I progress in this program.    

Typography and Labeling (the good kind)

Labels are an essential element of maps that we probably take for granted.  The font, color, size, position, etc. of the map labels can mean the difference between immediate recognition of the map's meaning and intended purpose, and the unpleasant task of squinting and muttering unpleasant phrases by the map reader attempting to decipher the map content.  We actually got to read up a bit on typography this week as well, which is an interesting subject to an amateur graphic designer such as myself.  The discipline of cartography, as one might imagine, has many specific and proprietary design conventions, including those involving text and labels.  

Effectively and clearly labeling a map of one of the Florida Keys is the perfect exercise in both cartographic labeling skills, and beginner's frustration-threatening-to-become-blind rage at the intricacies and idiosyncrasies of Corel Draw.  Suffice it to say I spent many an hour creating and making minute changes to the above map, and (hopefully) gained some decent skills in effective design and placement of map labels. 

I wanted also to add interest to the design of the map itself, without distracting too much from the actual content, and so decided to make the map's background a gradient of white and blue.  I went with a clear and readable sans serif type for the letters on the map's labels, and added a white halo behind some of them to allow them to stand out a little better.  I used differently styled labels for the Keys themselves, the cities and water bodies, which ideally allows the map reader to instantly recognize which is which.  Corel Draw remains a bit of a challenge for me, but I do notice an improvement in my ability to wield it effectively each time I use it. 

ArcGIS Online and Map Packages

ArcGIS, like most modern software, is moving increasingly in the direction of web-based applications and solutions.  This development and integration of online GIS was nascent in the version of the program I used last in about 2009, but has evolved into an exceedingly effective and pragmatic component of ESRI's ArcGIS.  This week finds us exploring a couple of ESRI's tutorials, and practicing the creation and optimization of map packages.

A map and/or tile package is a convenient means of sharing map data over the internet, as it allows users to share GIS maps without the usual networked database connections to access data sources.  Above is a screenshot of a map package (MPK), created to share on the internet.  The above MPK can be combined with a tile package (TPK) of basemap data, and would present a creative solution for an organization to share map data among members.

The above is a screenshot of an optimized map package.  Optimization is the process of selecting the most appropriate map data to share, using the most appropriate symbology.  Any and all practice in making these decisions is invaluable to those practicing to become able GIS professionals- that much I know personally to be true, even with my limited experience.  I honestly hope that this course makes use of these online resources further at some point, as those particular areas of expertise will likely only become more lucrative in the future.