forked from mapschool/mapschool
/
index.html
133 lines (132 loc) · 21.9 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<title>map school</title>
<link href='https://www.mapbox.com/base/latest/base.css' rel='stylesheet' />
<link rel='stylesheet' type='text/css' href='css/site.css' />
<meta http-equiv='content-type' content='text/html; charset=utf-8' />
<meta name='viewport' content='initial-scale=1.0 maximum-scale=1.0'>
<meta name="description" content="a brief introduction to map and geospatial concepts">
</head>
<body>
<div class='header'>
<h1><a href='./'>map school</a></h1>
<p>a free introduction to geo.</p>
<div class='pill'>
<a class='icon github button' href='https://github.com/tmcw/mapschool'>code</a><a
class='icon tooltip button' href='https://tmcw.wufoo.com/forms/mapschool-feedback/'>feedback</a><a
class='icon book button' href='https://raw.github.com/tmcw/mapschool/gh-pages/mapschool.epub'>ebook</a>
</div>
</div>
<div class='wrap'>
<div class='content'>
<h1 id="-map-school-section-map-school-html-"><a href="section-map-school.html">map school</a></h1>
<p>What is a map? Examples of maps are common, interspersed in driving directions, visualizations, and political boundary disputes. Let's look deeper and think about the fundamental elements of maps from the eye of the creator.</p>
<p>Maps are fundamentally composed of data. Data is in the abstract, composed of billions of points or just a few polygons, or a photo-like recording of colors and temperature. It is important that data is not specific to a certain usage.</p>
<p>From data, we make numbers, pictures, and decisions. Most commonly we make pictures, in a process called 'symbolization' - deciding what visual elements will represent what different parts of a dataset. We analyze data, which means we transform, aggregate, and summarize it to give different answers and represent different aspects of knowledge. The two tasks of symbolization and analysis are often combined, with symbolization forming the limits of what we can represent and analysis defining the aspects of data we focus on.</p>
<h1 id="-data-section-data-html-"><a href="section-data.html">Data</a></h1>
<p>Fundamentally, geographical data is either <strong>raster</strong> or <strong>vector</strong> - composed of pixels, or of geometry. The two formats are commonly used together, like in the case of satellite images as raster data with vector data roads overlaid.</p>
<h2 id="-raster-section-raster-html-"><a href="section-raster.html">Raster</a></h2>
<p><img src="img/raster.png" alt=""></p>
<p><strong>Raster</strong> data is like a picture that you would take with a digital camera: at the lowest level of abstraction, it is a list of pixels with values. When you 'zoom in' and look closer at raster data, at some point you'll see these discrete pixels, and it will look pixelated.</p>
<p>Raster data is used in pictures of the earth, like those taken by satellites - but that is just the beginning. Pixels don't need to have colors - instead, each pixel can have a number that represents height and the raster data as a whole stores elevation data. Or pixels can store temperature or reflection data and be useful for environmental work.</p>
<h5 id="-raster-bands-section-raster-bands-html-"><a href="section-raster-bands.html">Raster Bands</a></h5>
<p>The pixels in raster data are not necessarily just filled color: we call its contents 'bands'. A normal image has three familiar bands: Red, Green, and Blue. Combined, they make a picture we're familiar with. Some raster data can have fewer bands, like just one for Elevation, or some can have a lot more - not just visible colors, but wavelengths we can't see, like infrared and ultraviolet. When raster data is analyzed and displayed, you can combine and pick bands to use to suit what you're looking for.</p>
<h5 id="-raster-formats-section-raster-formats-html-"><a href="section-raster-formats.html">Raster Formats</a></h5>
<p>Raster formats aim to compact data and make it accessible fast for analysis and display. Some of them are geospatially-enabled versions of common image formats, <a href="http://trac.osgeo.org/geotiff/">GeoTIFF</a> and JPEG2000.</p>
<p>Internally, raster data formats manage two tasks - packing data into pixels, and then storing the relationship between those pixels and actual places on the globe - the 'extent' of the data.</p>
<h2 id="-vector-section-vector-html-"><a href="section-vector.html">Vector</a></h2>
<p><img src="img/vector_types.png" alt=""></p>
<p><strong>Vector</strong> data stores basic geometries rather than pixel data. No matter how much you 'zoom in' on vector data, you won't see pixels: the data stored is composed of geometric points and lines, and only converted into an image when necessary.</p>
<p>Vector data is used to store roads, buildings, points of interest, and other things that have some place in the world.</p>
<h5 id="-vector-formats-section-vector-formats-html-"><a href="section-vector-formats.html">Vector Formats</a></h5>
<p>The most established vector format is the <a href="http://en.wikipedia.org/wiki/Shapefile">Shapefile</a> - a simple, file-based format that awkwardly spreads the necessary data between four separate file - <code>.shp</code>, <code>.prj</code>, <code>.shx</code>, and <code>.dbf</code>. Shapefiles are binary data, so opening them in a text editor won't show anything accessible.</p>
<p><a href="http://geojson.org/">GeoJSON</a>, <a href="https://github.com/mbostock/topojson">TopoJSON</a>, and <a href="http://developers.google.com/kml">KML</a> are newer formats based on <a href="http://www.json.org/">JSON</a> and <a href="http://en.wikipedia.org/wiki/XML">XML</a> text encoding, respectively. Being text-based, they tend to be simpler to implement in software than Shapefiles, and combined with additional flexibility and features, they have become the standard in new web software.</p>
<h3 id="-topology-section-topology-html-"><a href="section-topology.html">Topology</a></h3>
<p>In addition to storing places and shapes, some vector data keeps track of topology, the relationships between different shapes that touch. For instance, political borders often touch - you can stand with one foot in Arizona and another in New Mexico. A lot of geospatial data, though, will have one shape that represents Arizona and another that represents New Mexico, with two borders that precisely overlap, but have no other association.</p>
<p>This gets tricky when you want to do something like ask 'which states touch?' or simplify the shapes of objects while preserving borders lined up. We use the concept of topology: some vector data, instead of storing a shape of Arizona and another of New Mexico, saves a bunch of lines and keeps track of which ones form the boundary of which object. So, the border between Arizona and New Mexico is a single line that's used to draw the border of both states, and if you have the line, it changes the shape of both states.</p>
<h2 id="-geocoding-section-geocoding-html-"><a href="section-geocoding.html">Geocoding</a></h2>
<p>A great deal of geographic data is none of the above - instead of being composed of the numbers that computers understand, it is stored as text data - references to placenames, streets, addresses, and so on.</p>
<p>Unfortunately, this isn't directly mappable. Due to variance in naming and the many definitions of place, there's an indirect and often inaccurate process involved in transforming the words 'United States' into the point <code>-120, 40</code>. This process is what we call <strong>geocoding</strong>.</p>
<p>The opposite process is <strong>reverse geocoding</strong>, where the readily available data is precise geographic positions, and the objective is a human description, like <code>United States</code> or <code>1714 14th Street</code>. Likewise, this isn't always accurate - one place on earth can be inside of overlapping and conflicting boundaries or between address points.</p>
<p>Geocoding and reverse geocoding can be very tricky: coordinate position errors, poorly formatted address data, and an ever-changing grid of streets and buildings contribute to the difficulty in turning addresses into coordinates, or vice versa.</p>
<h2 id="-data-collection-section-data-collection-html-"><a href="section-data-collection.html">Data Collection</a></h2>
<p>Map data has been collected in countless ways through the years - everything from sailors' logs to geocoded Tweets. Currently, there are a few major sources that merit discussion:</p>
<p><strong>GPS</strong>, the satellite constellation that gives your cell phone a blue dot on the map, is the foundation of collecting accurate vector data. Surveyors will drive with highly accurate GPS units and combine their results into something trustworthy.</p>
<p>Observational <strong>satellites and airplanes</strong> collect most raster data we have today, constantly taking photos from different altitudes and combining them into something that looks a little like a picture of the world. The same sensors also capture what we call 'non-visible spectrums', like infrared light that's useful for mapping agriculture. Some tricked-out rigs include <a href="http://en.wikipedia.org/wiki/Lidar">LiDAR</a>, a kind of laser sensor that measures altitude and yields us raster altitude data.</p>
<p><strong>Corporations, governments, and communities</strong> maintain different world maps of varying detail. For instance, Google and <a href="http://www.openstreetmap.org/">OpenStreetMap</a> focus on mapping all roads and details about them, and sources like the <a href="https://www.cia.gov/library/publications/the-world-factbook/">CIA World Factbook</a> and <a href="http://www.naturalearthdata.com/">Natural Earth</a> keep track of political borders.</p>
<h3 id="-storage-section-storage-html-"><a href="section-storage.html">Storage</a></h3>
<p>There are many ways to store geographic data. Data can be stored as printouts, but recently it is more popular to store it in a way easily accessed and shared, digitally.</p>
<p>There are many different filetypes and conventions for storing geographical data, and thus a variety of tools to translate amongst different representations. Data is stored in databases or in individual files, but the practical difference between the two is not necessarily important.</p>
<h1 id="-information-section-information-html-"><a href="section-information.html">Information</a></h1>
<p>What is a map, graphically</p>
<h2 id="-latitude-longitude-section-latitude-longitude-html-"><a href="section-latitude-longitude.html">Latitude & Longitude</a></h2>
<p>The most common way to store places on earth is with latitude and longitude values. Historically latitude and longitude are sometimes represented in sexagesimal notation, like <code>38° 32'</code>, but the new standard is to represent them as simple numbers, like <code>38.2</code>, that are easy for computers to understand.</p>
<p><img src="img/latlon.png" alt=""></p>
<p><strong>Latitude</strong> values range from -90 at the South Pole to 90 at the North Pole. All along the equator the latitude value is 0.</p>
<p><strong>Longitude</strong> start at 0 at the Prime Meridian which cuts through Africa and Greenwich, and extends to -180 and 180 at the antemeridian.</p>
<p>The combination of latitude and longitude is usually called a coordinate, and can be represented as 'latitude, longitude', or 'longitude, latitude': historically, the former was standard, but 'longitude, latitude' has recently gained popularity because it mirrors the 'X, Y' ordering of coordinates in math's euclidean space.</p>
<p>Coordinate ordering can cause some confusion, as browser based mapping software often expects 'latitude, longitude', whereas many wire formats specify 'longitude, latitude'.</p>
<p>Sometimes more than just the latitude, longitude position is recorded as data: altitude can also be included, as well as time of capture and other factors. In the case of including altitude, it's usually stored as a third coordinate, like 'longitude, latitude, altitude.'</p>
<h2 id="-projection-section-projection-html-"><a href="section-projection.html">Projection</a></h2>
<p>Projections are what we call the mathematical equations that do the trick of turning the world, which is a sphere, into some flat shape that fits on a printout or a computer screen. It's a messy task to do, this transformation - there's no way to smoosh the world onto a screen without distorting it in some way. You either lose direction, or relative size, or come out with something very weird looking.</p>
<h2 id="-symbolization-section-symbolization-html-"><a href="section-symbolization.html">Symbolization</a></h2>
<p>Symbolization is a fancy word for the particular ways that data is transformed into graphics in the world of maps.</p>
<p>Fundamentally, data doesn't look like anything: a list of pixel values or road lines is just as well represented in a spreadsheet or a chart as it is on a map. Thus to 'convert' is not the right word for what we do with data: at the most basic level, everything is a decision.</p>
<p>Symbolic techniques include anything representable in graphics or even 3D, so let's only look at a few:</p>
<h2 id="-sequential-categorical-section-sequential-categorical-html-"><a href="section-sequential-categorical.html">Sequential & Categorical</a></h2>
<p>Symbolization tends to highlight two different characteristics of data: sequential and categorical. Sequential, or continuous data could also be called linear - it tends to be number values within a set range, like graduation rates between 0 and 100. Categorical, or discrete data is, instead, one of a set number of values - like 'true', 'false', 'democrat' or 'republican'.</p>
<p>This division between data is one of the main concerns for symbolization - a sequential datasource would fit with a scaled point map whereas categorical data would work with multiple symbols for markers, and so on.</p>
<h3 id="-choropleth-section-choropleth-html-"><a href="section-choropleth.html">Choropleth</a></h3>
<p>Choropleth maps preserve the existing boundaries and shapes of places and represent data by changing their colors, patterns or textures. A familiar example of this kind of map is in election results or demographic makeup, in which data is percentage values for some fixed piece of land - a value per voting precinct or census area.</p>
<p>Choropleth maps are a natural fit for data like rates, densities, or percentages. They aren't recommended for absolute values: since they keep the area of shapes the same, they tend to over-emphasize large features. Also, since choropleth maps rely on color differences to represent information, it's important that the colors are well-chosen to be colorblind-safe, understandable, and consistent.</p>
<h5 id="-choropleths-in-depth-section-choropleths-in-depth-html-"><a href="section-choropleths-in-depth.html">Choropleths in Depth</a></h5>
<p>The selection of colors in a choropleth map is generally called the 'palette' or the 'color ramp', especially when the colors are sequential. <a href="http://colorbrewer2.org/">ColorBrewer</a> is a well-researched set of colors that are understandable and clear.</p>
<h3 id="-point-section-point-html-"><a href="section-point.html">Point</a></h3>
<p>Point maps are a better alternative for absolute values - the only geometry that they preserve is a single point for each feature.</p>
<p>The specific point or marker used in this style varies tremendously - Coloring points based on their sequential or categorical value can be useful, but points can also be scaled to different sizes to show their relative value. These scaled symbols can be any shape or image, such as circles, squares, or pictures of what they represent. In cases where there are multiple values that total up, scaled pie charts can be a terrific way to visualize what would otherwise be a complex dataset. </p>
<p>Care must be taken to not show too many points at once, as this will make a map difficult to read. In cases where there are too many points, a choropleth with aggregated values from the points may be a good alternative. Another alternative is to use clustering, where crowded points are grouped together until the map is zoomed in closer.</p>
<h2 id="-publishing-section-publishing-html-"><a href="section-publishing.html">Publishing</a></h2>
<h2 id="-analysis-section-analysis-html-"><a href="section-analysis.html">Analysis</a></h2>
<p>Raster & vector analysis as aggregation & transformation</p>
<h3 id="-vector-to-raster-section-vector-to-raster-html-"><a href="section-vector-to-raster.html">Vector to Raster</a></h3>
<p>It's possible to move between the two types of map data, as disparate as they might seem, and it's actually quite common: it's just not direct.</p>
<p>Typically vector data always becomes raster - what we call 'rasterized' or 'rendered' when it is displayed: computer screens and printers operate on the level of pixels, not lines or shapes. This conversion is imperfect: remember that vector data is not pixel-based, so you can never zoom in and see fuzzy features. Thus, generally, when vector data is converted into a raster format, you can't transform that raster representation back into exactly the source.</p>
<p>Sometimes we convert vector data to raster ahead of time - in order to run some kinds of analysis, it's easier to do the math on a pixel basis.</p>
<h3 id="-raster-to-vector-section-raster-to-vector-html-"><a href="section-raster-to-vector.html">Raster to Vector</a></h3>
<p>Likewise, raster data can be vectorized in a number of ways. On raster satellite images of the earth, people draw, or 'trace', lines for streets, points for houses, or polygons for buildings. This lets us have a version of the data upon which we can do more things - you can figure out driving routes from vector data of streets, but can't do that with a satellite image of streets.</p>
<h3 id="-simulation-section-simulation-html-"><a href="section-simulation.html">Simulation</a></h3>
<p>With geographic data, it's possible to simulate certain natural processes, and this simulation is a big part of what working mappers do. Given elevation data for a mountain range, it's possible to simulate highlights and shadows for those mountains in light, in a process called hillshading.</p>
<p>More complex processes are also possible, like determining where water will collect after rainfall, called a 'watershed', or determining everywhere on the map that will be visible if you're standing at a mountaintop, called a 'viewshed'.</p>
<h3 id="-aggregation-section-aggregation-html-"><a href="section-aggregation.html">Aggregation</a></h3>
<p>The most common form of aggregation is the idea of a sum - given a large group of numbers, you can add them together to get an idea of them all at once. For instance, the GDP of a country is more immediately informative than simply listing every individual contribution to an economy.</p>
<p>Aggregation in maps is similarly used. Given granular data, like millions of individual household incomes, you can use mapping algorithms to sum or average the values inside of specific areas of the world, to show average income per town or city.</p>
<p>Aggregation is also used along with a technique called <strong>binning</strong>: given many discrete points, you can draw arbitrary shapes on the map, like squares or hexagons, and total all of the points that fall in each shape. That way, instead of having millions of tiny points that are difficult to grasp at a glance, you can style the map as a choropleth.</p>
<h3 id="-interpolation-section-interpolation-html-"><a href="section-interpolation.html">Interpolation</a></h3>
<p>Whereas aggregation takes lots of data and distills it into something simpler to analyze and visualize, interpolation 'fills in the blanks' between values. Interpolation is often used for datasets like elevation, where you have raster data values that records the height of every inch of a mountain, but there are a few places missing from the data - what programmers call <code>null</code> values.</p>
<p>Interpolation looks at the values around the 'blanks' and assumes that the missing values are basically similar to what's around them - a missing pixel at the top of a mountain will be assumed to be pretty high, while one in a valley is assumed to be low.</p>
<p>Interpolation can happen on a vector data level as well, through the Voronoi algorithm, that takes a set of points and turns them into polygons of all of the area around them.</p>
<h2 id="-conceptual-review-section-conceptual-review-html-"><a href="section-conceptual-review.html">Conceptual review</a></h2>
<p>The unity of data</p>
<p>Edges blending into math and art</p>
<h3 id="-license-section-license-html-"><a href="section-license.html">License</a></h3>
<p><a href="http://creativecommons.org/licenses/by/3.0/deed.en_US">Creative Commons Attribution 3.0 Unported</a></p>
<h4 id="-report-errata-https-tmcw-wufoo-com-forms-mapschool-feedback-section-report-erratahttpstmcwwufoocomformsmapschool-feedback-html-"><a href="section-report-erratahttpstmcwwufoocomformsmapschool-feedback.html"><a href="https://tmcw.wufoo.com/forms/mapschool-feedback/">Report Errata</a>.</a></h4>
</div>
</div>
<script type="text/javascript">
var _gauges = _gauges || [];
(function() {
var t = document.createElement('script');
t.type = 'text/javascript';
t.async = true;
t.id = 'gauges-tracker';
t.setAttribute('data-site-id', '528a6fcff5a1f56dc400009f');
t.src = '//secure.gaug.es/track.js';
var s = document.getElementsByTagName('script')[0];
s.parentNode.insertBefore(t, s);
})();
</script>
</body>
</html>