Thumbs Up Viz: Hurricane Matthew
Hurricane Matthew was the first major hurricane (category 3 or above) to hit the U.S. since 2005. In the wake of the storm several visualizations appeared putting the storm in historical context, comparing the forecast track with the hurricane’s actual position, and showing the rainfall and resultant flooding the storm brought to the Southeast.
Matthew in Context
The National Hurricane Center has a wonderful dataset showing the position and intensity of Alantic hurricanes dating back to 1851(!). Lazaro Gamio of the Washington Post plotted the past 100 years of tropical cyclone activity in and around Florida. This map matrix (yes, I still like small multiples) shows the relative quiet of the past decade, compared to the record-setting 2005 season, when four major hurricanes (category 3 or above) hit the mainland United States.
In addition to showing position, the maps encode storm intensity on the Saffir-Simpson scale with a perceptual sequential color palette and max wind speed with line width. There’s a discrete break in the colors when the storms aren’t strong enough to be classified as hurricanes — tropical depressions and tropical storms are shown in cool blues instead of the hotter oranges and reds of hurricanes. The combination of warm, saturated colors and thick line widths push the strongest storms into the foreground, drawing our attention to them.
In contrast, the map elements are subdued grays, so they fade into the background. Coastlines and state boundaries are just dark enough to be visible. The allows storm names and annotations to stand out alongside the storm tracks. I particularly like this note for 2012’s Hurricane Sandy: “Regained strength and pummeled the Northeast”.
One other detail stands out: the key (with integrated title) is at the top of the figure, rather than tucked underneath. It’s a nice way to give readers the information they need to interpret a map right up front.
Evaluating Forecasts
Hurricane forecasters have a tough job: they’re tasked with predicting something inherently unpredictable, and a failed forecast could result in many lives and millions of dollars lost. Ian Livingston (also at the Washington Post) made this map that shows just how good the National Hurricane Center predictions for Hurricane Matthew were:
Each mini-map (yes, small multiples again) compares the predicted location of Matthew at different intervals from the time of prediction, with the observed track. It’s interesting that the 96- and 120-hour predictions both wavered between two possible paths—one close to shore and the other further out in the Atlantic. The 72-hour predictions converged on a single path, but showed a “hook” from the Carolina coast back towards Florida that never developed. The 48-, 24-, and 12- hour predictions are spot on.
Like the 100-years maps, these draw attention to the most important data (predicted location) by using hot colors to bring them into the foreground. Each map is directly labeled, with clean typography and anti-aliased coastlines. As a point of comparison, here’s an example of an official National Hurricane Center forecast cone:
Assessing Impact
Although Matthew’s winds had died down a bit by the time it made landfall in South Carolina, it still carried a lot of moisture and delivered a lot of rain. A lot of rain.
This interactive map by the USGS Center for Integrated Data Analytics Data Science Team shows just how much:
It’s successful at illustrating the aftereffects of Hurricane Matthew because it’s intensely multivariate, showing position of the storm over time, location of rainfall over time, position of stream gauges, and stream flow over time—11 dimensions of data. Only superb design prevents this wealth of information from becoming an overwhelming mess.
Like the previous maps, color (a nice wet perceptual sequential scale) is used to separate foreground from background. The position of each stream gauge is linked to the flow data via interactivity, which even allows a viewer to drill down to the source data. The flow rates are normalized, so tiny streams aren’t obscured by the massive flows of coastal rivers.
Interesting patterns emerge, like the concentration of rainfall ahead of the storm’s center, and the quick rise and fall of small creeks vs. the slower buildup of water downstream.
The team described the intention of the visualization on the DataIsBeautiful subreddit:
Our goal was to communicate the impacts of Matthew on spatial precip patterns and then river/streamflow. We wanted to show the wave of water moving from south to north (as you see the different lags in the discharge sparklines on the right) while also highlighting the diverse responses among different waterbodies.
Mission accomplished.
BTW, the federal government sometimes seems allergic to crediting employees for their work, so I’d like to mention the contributors (listed in the metadata) by name: Alison Appling, Lindsay Carr, Laura DeCicco, Emily Read, Jordan Read, Jordan Walker, David Watkins, and Marty Wernimont. They’ve even done the service of publishing the source on GitHub.
It’s also worth noting that all three of these visualizations were created from publicly funded, freely available data, and that both the USGS and NOAA/National Weather Service are often under funding pressure, and need continued support.
