The HazWx project is one of the most fulfilling projects I’ve ever had the honor of being a part of, the feeling of building something great from the ground up with a small dedicated, passionate team was unlike any other.
As the UI developer, my role in this project was to build front-end functionality and light up the user experience. In 2015, HazWx was acquired by AccuWeather, Inc. Take a look at our work with a narrated video and image gallery below.
The rapid development and coordination is remarkable when you consider that only two team members, Brandon and myself, lived in the same state. We worked online with frequent web meetings and online hangouts while planning each new functionality.
Built on top of a WordPress membership site foundation for handing subscriptions and member access, the primary interface is Google Maps, on which we built our own data displays.
I selected and customized the JQuery UI framework to drive the user experience.
A user steps through each section in a menu to select and generate a forecast map overlay, beginning with the model selection – choosing the algorithm from which to generate a forecast. Essentially the style is a vertical paged wizard.
After making a selection, we make an AJAX call to our custom API written by Robert Dewey, which provides all the available options for each selection in context, and the next menu opens automatically.
Here the user selects the initialization time, or the particular start time of a model run. This way the forecaster can compare various runs of the same model to see how they differ.
Each model has various data fields to choose from, and our API provides a user the list of currently available data based on their earlier selections.
Within each field, the user can further choose specific attributes that can be plotted on the map.
In addition to the data metrics, a user can choose their desired valid time of the data, or the target time the model represents. For example, you can see on the legend below this is the HRRR model run started at 3pm, showing surface temperature as it was predicted to be at 6 hours in the future from that time, or 9pm.
Some models, such as this Hi Res GFS model, can generate forecasts for up to 192 hours into the future and allow the user to choose to view any in between, or pre-load the data and animate the forecast frame-to-frame.
A dual view allows the user to view two sets of data, or the same set for different target validity times, side by side.
Additional panels allow users to adjust the transparency mix of any overlay layer, or add additional layers such as polygons for NOAA watches and warnings, as well as SPC mesoscale discussions.
Radar overlay and visual spectrum GOES satellite feeds can also be mixed into view.
This example shows visual satellite and radar display together. This was late evening as indicated by the dark image in the eastern field of view.
JJ added some features post-release, including these to bring in SPC predictions and annotations.
Google Map traffic data and GIS outlines were also added.
Available options include adjusting times to local PC time instead of zulu time, preserving map settings in cookies, and preloading forecast hours.
Media package allows users to customize the overlay logo, and drawing tools provide editable shapes, polygons, pushpin points, and text boxes to be annotated on the map.
Share a map with the world with our implementation of sharing links and social posts, complete with all the post metadata you expect.
Finally, integrated UX with support for live tile updates (Windows 8 and 10), as well as taskbar integration for Windows 7 and higher, keeps users coming back with zero friction, and staying updated in between visits.