Visual Variables J. Bertin: Sémiologie Graphique, 1968
Total Page:16
File Type:pdf, Size:1020Kb
Geoinformation in Environmental Modelling Communication by maps Jussi Nikander ENY-C2005 13.2.2019 Slides by Paula Ahonen-Rainio and Jaakko Madetoja Topics today • Communication by maps – Information & aesthetics • For whom, what, and why? • How: elements to be designed • Visual variables and how they link with different types of data • Visual hierarchy and colour design Examples of potential exam questions relating to this lecture (in English) • There is a poorly composed thematic map in the attachment. Which three mistakes and/or missing elements would you correct in the first place? What other mistakes there are in the map? • Explain what it means that data to be presented in a map has to be normalized. When is the normalization necessary, and how can it be done? • Name the visual variables that can be used on map symbols (and in information visualisation in general). How does the measurement scale (nominal, ordinal, interval, ratio) of attribute data rule which visual variables should be used? Same questions in Finnish • Liitteenä on huonosti toteutettu teemakartta. Mitkä kolme virhettä ja/tai puutetta ensisijaisesti korjaisit kartassa? Mitä muita virheitä kartassa on? • Selitä mitä tarkoittaa, että kartassa esitettävä tieto on normalisoitava. Milloin normalisointi on tarpeen, ja miten se voidaan tehdä? • Nimeä visuaaliset muuttujat, joita voidaan käyttää karttamerkeissä (ja yleensäkin tiedon visualisoinnissa). Miten esitettävän ominaisuustiedon mitta-asteikko (nominaalinen, ordinaalinen, intervalli-, suhdeluku-) vaikuttaa siihen, mitä visuaalisia muuttujia tulisi käyttää? Communication by maps Communication process: SENDER encoding MESSAGE decoding RECEIVER content form (visualization) … Communication by maps • Visual presentation can reveal something that numbers don’t tell us • An image is worth a thousand words, a map perhaps even more Visual patterns on a map represent: spatial relationships, distributions, trends, outliers,… http://ilmatieteenlaitos.fi/kuukausitilastot … Communication by maps “The purpose of visualization is insight, not pictures” J. Bertin Aims of map design • Informativeness For example, attributes of a building signify… – Discrimination (residental, administrative, industrial, office) – Order (buildings of 1-2; 3-5; over 5 floors) – Emphasis (buildings with geothermal heating) Clarity, visual hierarchy, logics There are no neutral • Aesthetics choices – a factor of usability: ”satisfaction” in map desing – connotations: eg style up-to-dateness Affect on map reading and interpretation For whom, what, and why? Phenomenon & data User Purpose - visual perception - specific...generic - profession, age,… - big picture…detailed study - cultural context - convincing, provoking, - motivation, expectations, exploring,… tiredness,… Important terms Topographic map Large scale map Small scale map Thematic map A bad example Study with you neighbour the map example Can you understand its message? What is wrong in it? Kantasolusiirre = Stem cell transplant (blood cancer treatment) Red = export % Blue = import % Verestä elämää 1/2017 How? – Elements to be designed Decisions on and design of… 1. Data to be included (and excluded) – Normalization, classification 2. Mapping method – Different symbolization, colours 3. Area of the map & presentation scale – Generalization of geometry 4. Background for the theme – Relevant content, visual hierarcy 5. Key for interpretation: title and legend 6. Metadata 1. Data to be presented • What data to include, and what to exclude – Object classes and their attributes – One or more attributes per object (how many we can visualize) OR – Spatially continuous variable(s) • Data analysis – Which measurement scale (nominal, ordinal, interval, ratio) – Range and distribution of values – Counts or proportions? Need for normalization? – Unclassified or classified? – Need for preprocessing of data? Why normalization Example of population distribution • Think of a grid with 500 people in each grid cell; that means constant population density across the whole study area • Then divide the study area into regions of different sizes so that each grid belongs to one area • Then count the number of population of each region by summing up the population in the grid cells; as a result, the number of population is the larger the larger the region, and vice versa • If we now present this information, the number of population, in a map – take one colour hue and variate it from light to dark, darker meaning a higher population – what do we get? • We have a map that communicates, that larger areas have more population than smaller ones, but not that the population density actually is the same in every region. Not very useful! Instead, eliminate the disturbing effect… …Why normalization 5 5 5 5 5 5 Example: 5 objects in each grid cell. 5 5 5 5 5 5 4 regions of varying size. 5 5 5 5 5 5 5 5 5 5 5 5 Map of non-normalized values: Map of normalized values reveals The larger the region the higher the value. the constant density across the area. 20 20/4=5 50 50/10=5 20 30 30/6=5 20/4=5 Normalization of data …eliminate the effect of varying size of the enumeration areas - or some other distorting effects: spatial or other dependencies Relevant with ratio scale values only • Normalization – per area • e.g. population density: number of people per area • e.g. productivity: tons of wheat harvested per area (yield) – per another variable • e.g. proportion of Swedish speaking people per population • e.g. birth rate per women of child-bearing age Example: normalization of data Consider what is the message of the map, what makes sense. Classification of data • Classification of data – First of all, classified or unclassified why do we classify? – How to classify: one of the profound choices in map design • Systematic grouping of data based on one or more attributes • For a clearer map, even if this map image is generalised – “Less is more” principle • The classification, and the resulting map, should reflect those patterns or structures that are characteristic for the mapped phenomenon – And each class should contain its share of the observed values Example: classification of data The map should display the characteristics of the phenomenon …Classification of data Aim: Homogeneous classes + contrast between the classes What is the suitable number of classes depends on... • map users: the lower the number of classes, the easier to interpret the map – but not too low, either what is the sufficient level of information We can perceive 5 without counting, or keep in working memory about 7 meaningful chunks at a time • characteristics of the phenomenon: present essential behaviour – Avoid deciding the number of classes before studying the data • the number of objects in the map (if very small, avoid too many classes) • visualisation of classes: what we can perceive & how it is presented For example, bipolar data (diverging from a meaningful turning point, such as 0 for temperature) affords more classes: use two hues that vary in lightness What is the suitable classification method See the examples Example: classification of data Here measured values of 20 data objects Classified into five classes Classification by equal intervals: Classification by quantiles (equal count): Classification by maximum breaks: Example: classification of data Searching natural breaks visually by a cumulative frequency diagram 20 16 12 8 4 2. Mapping method Nature of data, first of all, determines what is appropriate map type • Spatially: points, lines, areas (or grid), continuous field • Attribute values: – Absolute ratios (or interval) – Relative ratios Quantitative data – Ordinal – Nominal Qualitative data – Cyclic: Quantitative values on a scale that, at a certain point, is back to zero • direction of wind, time of day Notice the use of visual variables • Intuitive interpretation • Variables in different map types Visual variables J. Bertin: Sémiologie graphique, 1968 Mapping of data variables to visual variables to encode data values visually to points, lines and areas Kraak & Brown 2001 Web Cartography Later additions, such as transparency, crispness, perspective hight, … Example: Relative ratio value per area / classified: ordinal data Choropleth map Notice the projection (Lambert, conical): north direction and scale vary across the map. no north arrow, no scale bar Display grid lines instead. Notice the legend: visual presentation of the class intervals. eea.europa.eu …Choropleth map However, even if the map is correctly made: • in the interpretation: bigger areas gain more attention • intuitive assumption: map suggests that areas are homogeneous – that is seldom true – Dasymetric map corrects this drawback: If further data is available, form homogeneous areas instead of enumeration units that may have remarkable internal variation. E.g. exclude waters, mountains; divide urban and rural areas, etc. Example: Nominal value per area Chorochromatic map Don’t follow this example! Colour here suggests and order. Provinces in Finland – commons.wikimedia.org Example: Absolute ratio value per area Dot density map & Proportional point symbols e-education.psu.edu / Geoff Hatchard Example: Quantitive or ordinal data per point or area Proportional point symbols Two attributes (bivariable map): population presented with size, its change with colour in the same symbol. Two attributes: Example: Nominal value per point type of service (colour) and Point symbol map accessibility (ring) http://www.hel.fi/palvelukartta/ Example: Qualitative