March 12 Notice: The CISA IOCC has issued an Awareness Message regarding a malicious website masquerading as a live map for Coronavirus COVID-19 Global Cases by Johns Hopkins University is circulating on the Internet to entice unwitting users to visit. Visiting the website infects the user with the AZORult trojan, an information stealing program which can exfiltrate a variety of sensitive data. The URL for the malicious site is (corona-virus-map[dot]com).
As updates to the Coronavirus (COVID-19) are posted, the common tool that citizens, state and local officials, and researchers are using to track the virus and communicate its impact are maps. Maps have been used for centuries to respond to threats of disease and viruses. This Story Map provides several examples including one of Fillippo Ariette's maps of quarantine zones in Bari, Italy, which was created in 1691. And, for many epidemiologists and geographers, the Cholera epidemic in London during the 1850's and the efforts by Dr. John Snow illuminate the power of maps to solve a community health threat.
The Coronavirus presents another opportunity for researchers, GIS professionals, and organizations (health, news, etc.) to develop their own "Coronavirus maps" and web applications in an attempt to visualize the extent, spread, and impact of the virus. And while UCSF has provided a dedicated page of resources which includes a web map from John Hopkins CSSE, there are an increasing number of maps and apps that can be found online.
Below is a compilation of maps from organizations and websites discovered as of March 3rd, 2020. Some are done very well and others, not so much. Understanding the principles behind cartography and map design is important if we want to help the general public understand the threat. Esri's Kenneth Field has a blog post on how to map the coronavirus and some pitfalls, many of which will be evident in the examples. By reviewing the maps circulating and being promoted online, we can better understand what methods and approaches work to best communicate pertinent information, and how we might deploy them in future incidents.
The main point of this article, which I hope will resonate with readers, is that while both mapping software and the geographic data are readily available to anyone the study and learned skill of cartographic principles remain highly relevant in order to communicate responsiblly through maps.
Want to learn more about map design techniques and methods? Scroll to the bottom for suggested courses available to all UCSF staff and students.
Esri, the world leader in mapping technology launched a Hub site that aggregates a lot of different maps that use their technology as well as common datasets. Included is the John Hopkins dashboard and data that many others are using to build their own maps. The Hub site includes:
The John Hopkins CSSE dashboard is fairly simple with three statistics and three graphs. Clicking a location in a list will zoom to the location and update the three statistics (to note, only certain countries are further broken down by province or state such as China and the US). Clicking the map symbol will open a pop-up window with statistics in the top-left of the map which is nice as it doesn't interfere with the map itself.
The map has been updated cartographically to make it easier to see the proportional symbols under the Hubei symbol, creating a less cluttered look that was found in previous versions. Typically, a reader will interpret red to indicate danger or death and in this case red indicates a confirmed case while green is recovered and white is death. It may be more effective to use orange to denote a confirmed case color and a level of concern or danger while to reserving red or purple for death. Overall, the map has a gamified feel using the dark gray basemap with bright red map symbols which presents a more heightened threat to the reader.
The Early-Alert dashboard includes the basic statistics of cases and deaths but adds travel health notices, travel advisory notices, and air restrictions and cancellations.
The WHO dashboard:
Their Story Map provides a journalistic approach to engage the audience on the topic and includes dynamic maps which add context and additional information.
And Esri China has provided a List of Coronavirus Dashboards from governments in Asia including Hong Kong (English | Chinese), China, Japan, Macau, South Korea, Singapore, Taiwan, Indonesia, Philippines, Thailand, and others like Italy and Isreal.
The SF Chronicle has one of the few maps showing the impact in California by county. The circle size represents the number of cases but it does not appear that the size of the circle represent a proportional value. Using a proportional symbol would likely be a more effective choice to inform the reader of the total number of cases. The legend might be useful but it does not indicate a value associated with the size forcing the reader to guess the amount of cases per circle. It also includes a circle much larger than any symbol on the map - it's a useless and confusing detail which is part of a very important aspect of the map for the reader to be able to decipher.
This map from CNBC is poorly designed. The number of color classes, at nine, is to far too many for most humans to discern the change in color, and the data uses absolute totals which can trick the reader. An experienced cartographer can usually identify a novice map designer based on few things: the chosen projection (or lack of), whether certain quantative data has been normalized, and how map elements such as north arrows and legends are applied.
Because each province is a different size with a different amount of people the only correct way to compare one province to another or one country to another (apples to apples) is to normalize the data. By displaying case totals by province the reader has no idea how many people are in that province (10,000? 1 million? 10 million?) and no context to understand the density of the virus' impact. Normalizing by the population would allow the author to display values such as 100 cases per 100,000 in population, which is close to the current infection rate in Hubei versus nearby provinces which are less than 5 per 100,000. Lastly, the inset map is too small and doesn't contain any data for mainland China. In this case, they should just create a new map for a world view.
This map from the Washington Post uses the correct proportional symbols on the China map. While the US map is useful, I would be in favor of using the same proportional symbols to demonstrate impact.
The New York Times China map also uses proportional symbols, and also includes the statistics for surrounding countries. The US map follows the same method maintaining consistency for the reader.
This Coronavirus.app map uses OpenLayers and Charts.js to produce a functional web map in terms of components. The obvious lack of a legend leave users with no way to evaluate the impact, thus making the symbols almost useless. Clicking on statistics will change the results in the map and clicking a location in the list or map will display a graph of the same statistics in the top-right of the application. Perhaps a more effective approach would be to display these location specific graphs as part of the existing interface instead of having to expose them with a click.
Emergent Epidemics Lab web map includes a layer of days since confirmation, which might be useful if there was a method to filter the data - otherwise the number of categories, at ten, makes it difficult for the reader to interpret the data. In addition, the red-to-blue color ramp seems to imply the longer a confirmed case the more dangerous; neutral colors would be more effective here, along with categorizing the length of confirmation. This map also includes a density layer highlighting the "hot spots" of confirmed cases but no information is included to understand the method used to create it. A density layer is an interpolated surface of point values. The problem here is that the location of the points matter. If all the points are centered on the province then a density layer is practically useless. The filter is fairly limited but includes "Travel History," which appears to show where the individual may have travelled - but there is no explanation.
HealthMap's web map is fairly simple with a single layer of cases. It appears to be out of date even though the last update time is yesterday at the time of writing this article. Wuhan, the epicenter is devoid of any cases which is odd. This may have been useful in the first days of the outbreak but the authors appear to have neglected its upkeep.
The EipRisk map models is similar to the Coronavirus.app in using models to predict the risk to populations. The design and color choices are captivating and the histogram of relative risk to each outgoing location is a nice addition. The use of "spider-lines" to connect the locations from the basin adds the visual touch to focus the reader. The form to filter by select basins is a bit awkward, but otherwise provides interesting variables with which to interact. This app also makes it easy to download the dataset or map.
Tableau is known for data visualization using graphs and charts but the product also includes a basic level of mapping capabilities. Although I did not find any official dashboard from the Tableau map team there were a few dashboards available publicly from users. This Tableau dashboard is mostly a graph-heavy visualization but it is one of the better dashboards I encountered. The map contains no legend or supportive text about the value the colors represent or what the numbers labelling the countries represent.
Another public Tableau dashboard. The graduated symbol pie charts make it both difficult to read the values and difficult to interact with at this world zoom level. A user can zoom-in to areas of the map but the data in the dashboard does not update based on the extent. The color ramp indicates cases from 1 to 21,461 but there's no indication what each color category is.
Tableau's charts and graphs capabilities are well known for data visualization but the tools available for mapping are not nearly as strong as Esri's ArcGIS Online Smart Mapping, ArcGIS Insights, or other vendor map services available to users. Admittedly, the Tableau map team has made great strides in the past year with new functionality and we look forward to future enhancements to enable users to create better maps.
More news outlet maps...
This Foreign Policy map uses absolute values to color code the country. The choice in colors for the categories is interesting, it suggests that countries with less than one hundred cases are somehow more safe (using muted tones) than those with greater than 100 cases (red hue). Again, without normalizing the data I'm not sure the reader can understand the impact of the virus. And, I do not understand why only some countries are labeled.
This NBC map follows the same approach as many other news outlets using number of cases grouped by category and then a color ramp to suggest intensity. The choice of class breaks for number of cases seems to be a favorite for many of the maps examined. The vast difference in cases between China and all other countries is staggering yet this map suggests to the reader that countries with 500-1000 cases are impacted almost as much as China. I don't think this is the message to send to readers. Also, the authors chose to use a projection that noticeably distorts the actual size of the countries.
This BBC map is strange because it simply does not categorize China with any confirmed cases. Admittedly, the map is titled as cases "outside" of China but is China's total too scary to display? Did they run out of colors in the hue of red? And the choice to color China as a dark gray much like countries with no confirmed cases is bizarre. Then, dark gray is not even shown in the legend. Otherwise, it is another example of authors using absolute numbers and not normalizing the data.
The BBC also published a story regarding a map incorrectly cited and used by various news outlets. I have to believe that either the authors didn't bother to perform due diligence, or knew this posting would generate clicks.
In this Bloomberg map I thought the removal of all interior country borders for countries with no confirmed case was an interesting choice. As a reader it is feels a bit disorienting, was the design team overthinking things? In addition, like many others, the use of absolute numbers and not normalizing the data along with the class breaks make this map difficult to understand the virus' impact.
NPR's US map is basic, as it highlights only the states with confirmed cases, and uses * to denote those states with repatriated individuals. The label size is a bit large and the placement of the New England state labels should not obstruct the map; further, NH and RI are abbreviated to two letters yet Mass. is not.
And, their world map of travel advisories of which, on first look appear that the colors don't quite match. I assumed the yellow countries are places of confirmed cases but that's not stated at the top of the map, rather it's at the bottom of the map -- eitherway, it's distracting and confusing.
This Financial Times map is a bit overwhelming with the background color and using a darker color for countries without a confirmed case. The darker color will stand out to readers by suggesting something "more" is occurring there. At least they used proportional symbols correctly.
Prevent Epidemics offers a map of 'Current International Inbound Flights from Countries with Confirmed Cases'. The flight paths and country centroids are visually attractive the underlining country colors are not described. Does Green for the US and Canada mean it's safer? And China is not colored at all. I didn't find any information, time-stamp, or source regarding the 3033 flights.
This web map created by Mapbox, a web mapping service, is a lot to absorb on first view. It displays both a color-coded layer of provinces and a hotspot layer showing confirmed cases and both layers use absolute numbers. The color choices for the hotspot layer is completely distracting and we previously discussed the issue of using this type of data representation. The font is difficult to read. The province/state boundaries beyond China should use a much lighter color to denote the lack of importance which would make it less distracting, also there is no distinctive country boundary. And, this is the first time I have seen "suspected cases", a click on the "About this map" link does not offer any clues to the source of this data.
In early February Carto, a web mapping service, posted an article on their blog with three maps - this one is focused on the Hubei Province. This is a good example for a novice to critique and learn from as it has a lot of poor map design choices. The use of log scale values for confirmed and cured cases is particularly strange.
As with the Cholera epidemic in London, the spatial component is critical to understanding the transmission patterns and rates to effectively update response plans and to communicate to citizens the impact of a disease. Today, it's easier than ever to create maps, and while many organizations will publish accurate and useful maps, many more will likely create, and have created, misleading or poorly designed maps for their constituents.
Bay Area based Cartographer, Darin Jensen, who moonlights as President of the nonprofit Guerilla Cartography and the former staff cartographer in UC Berkeley's Geography Department, provides us with this thought regarding the ease-of-use with mapmaking, "...the democratization of mapping (through apps and software) is revolutionary and also potentially dangerous".
Those seeking a beginners guide to undertanding the potential impact of cartographic design may be interested in Mark Monmonier's How To Lie With Maps.
UCSF staff and students can access Esri's training academy for free courses which includes several on the topic of map design such as the Cartography MOOC, Cartographic Creations in ArcGIS Pro, and Map Design Fundamentals. Request an ArcGIS Online account to get started.
Lastly, our fascination with the spread of disease can be explored more safely in the confines of our livingroom, self-quarentined or not. For those who enjoy boardgames the collaborative game, Pandemic by Z-Man Games is an enjoyable 45 minutes of group strategy, and some luck, to hold back the spread of four viruses that spread across the map and threaten to collapse our civilization.
Until next time, happy mapping.
