Space Apps Challenge 2018

Weather to Launch

Rocket launches are supremely powerful events. Loud, rumbling, bone-shaking noise can be felt even miles away from the launch site. There is truly nothing that compares to witnessing the power of a rocket in person. Unfortunately, there are only a few sites around the world where one can watch a launch. To make things even harder, some launches end up being scrubbed the day-of due to factors such as inclement weather or mechanical problems. Wouldn’t it be nice to know how likely it is that a launch is going to happen when it is supposed to?

WeatherToLaunch.com was created to help answer this question. Whether you are you planning a trip to Cape Kennedy to see a launch in person or just wondering if you should head home to watch the launch on your big screen, you can use WeatherToLaunch to make an informed decision. WeatherToLaunch is the only site that combines information about upcoming rocket launches, weather forecasts, and technical launch criteria to show you key launch indicators and their likelihood of contributing to a scrubbed launch.

Technical Approach:

Here's how our website works! A detailed description can also be found on our website github and app github

First, upcoming launches are collected using the “Launch Library”, a free to use API and repository of rocket launch information created for the community by Pete Riesett and Benjamin Higginbotham in 2011. This library contains not just time and location of launches but also interesting mission information, vehicle information and lots of nice-to have features.

Next, the location and time of upcoming launches are fed into the Aeris Weather forecasting api, collecting forecasted weather information such as wind speed, thunderstorm chance, and cloud coverage. The Aeris API provides forecasts up to 14 days out, containing a wealth of information. Unfortunately, the Aeris product is not free, WeatherToLaunch must maintain an API subscription to remain active.

Finally, the weather forecast information is compared to Key Launch Indicators (KLI) discovered by the WeatherToLaunch team. The goal of the KLI are to represent the most important factors contributing to a successful launch. Contributing sources include NASA publications concerning Launch Commit Criteria (https://www.nasa.gov/centers/kennedy/news/releases/2003/release-20030128.html and https://www.nasa.gov/pdf/649911main_051612_falcon9_weather_criteria.pdf). During the research phase, it became clear that different launch vehicles have different launch criteria due to unique vehicle attributes and even launch sites. For example, the Falcon 9 rocket cannot be launched in over 30 knots of wind where as the atlas V can be launched up to 33 knots. Additionally, each country’s space agency seems to have their own requirements for launch. As a proof of concept, WeatherToLaunch applies the Falcon 9 criteria to every launch. This is an excellent opportunity for improvement in the future.

Once the KLI have been determined, the result is stored in the WeatherToLaunch datastore on Google’s Firebase platform. Firebase provides an array of development tools which helped get the project online quickly. Firebase Functions, serverless functions written in Typescript and running on Node.Js, were used to collect information from the various APIs, combine the data and serve it to the user interface. These functions are available on the project Github for inspection. The end result of the WeatherToLaunch process is an API endpoint available at https://us-central1-whatsgoingup-spaceapschallenge.cloudfunctions.net/LoadLaunches. This endpoint serves JSON of all the upcoming launches (matching the Launch Library schema) as well as the KLI in a node named “WGUPred” at the top level of each launch. The WGUPred node combines launch date and location along with an array of predictors.

Each launch predictor provides the name (eg. “Wind Speed (KTS)”), the forecasted value (“Value”), the threshold for a launch (“Threshold”) and a Boolean of “HighValue” where a true represents a high threshold (value cannot go over the threshold) and a false represents a low threshold (value cannot go below the threshold). From here, our Key Launch Indicator gauges are born and the only remaining task is to create a UI so that users can easily use this information.

User Interfaces:

WeatherToLaunch includes two user interfaces: WeatherToLaunch.com and the WeatherToLaunch Android app. Both interfaces utilize the “GetLaunchData” API endpoint including information about rocket launches as well as the forecasted Key Launch Indicators (KLI).

WeatherToLaunch.com is a single-page website written in HTML, JavaScript, and CSS in the Bootstrap style. The data is read from the API endpoint using a Jquery AJAX call and presented by templatizing HTML. Each of the KLI are presented in a dial gauge format with color coded bounds for the threshold. This visual management style is borrowed from Lean Manufacturing practices, designed to give a clear indication of launch parameters. The gauges are created dynamically using the Gauge.JS library. The design ethic of the website focuses on delivering the KLI as simply and clearly as possible. The website source code is available on the project github: https://github.com/codegenesis/whatsgoingup.

In addition to the website, the “Weather to Launch” app is going to offer a mobile friendly interface to provide users with information about upcoming rocket launches on the go! The app will show the same detailed information such as weather conditions, launch windows, launch locations, and information about the rockets’ mission. The application will also provide notifications of upcoming launches so the user never misses a launch! Currently we are building the Android application; in the future we wish to expand this into IOS users as well. The WeatherToLaunch app documentation is available on the app github.

Future Goals:

WeatherToLaunch was developed over two days during the Space Apps Hackathon by a small team. As such, many original goals were not incorporated into the product as it stands. This section details possible improvements for the WeatherToLaunch platform and the technical challenges that must be overcome for each of them.

• Launch Prediction: WeatherToLaunch provides Key Launch Indicators based on forecasted weather and launch window. The original goal was to provide a mathematical prediction of launch or scrub based on these KLIs. This would require trend analysis of past launches to create a statistical model for future launches. Unfortunately, historical data on weather, launch windows and launch success/scrub reasons could not be located in the time allowed. It is possible that no publicly available repository of this information exists. However, the WeatherToLaunch system was set up with the goal of producing such a repository over time. Given 30 launches, enough data will be available to build the required model and provide the user a exclusive percentage of scrub.
• Friendlier User Experience: Weather to Launch was inspired by it's users and was created with them in mind. Because of this we would like to include several functions in the future such a:
  • Sorting our website by Launch Site, Country, Provider, and Rocket to provide the user with an organized informative experience.
  • Notable viewing locations around the launch site, and in the future a way for users to contribute their own locations or comment on ones provided with helpful reviews.
  • Live Social Media stream. The best way to know if the rockets are being delayed is to listen to the community! We are looking forward to getting our twitter API request approved so we can incorporate current tweets into the design.
• Better Website: The WeatherToLaunch.com website was designed as a single page HTML file with javascript. All of the data retrieval is done via API endpoint from the Firebase Functions, which are quite slow. We plan to switch to Angular for the website, to take advantage of native FireStore integration to pull the data directly out of the store without needing the Firebase Functions. This would greatly improve loading time as well as allow for “live data” (updating forecasts automatically for the user without needing to refresh the page).
• Additional KLI: Currently, only four indicators are presented to the user: Wind Speed, Thunderstorm change, cloud coverage, and launch window. There are several other possible indicators available on the NASA Launch Commit Criteria which could also be presented to the user. These indicators were chosen based on the ease of finding forecast data and personal experience with scrubbed launches. Adding additional indicators would help the user better judge the chance of launch, however, adding more data may make the site less user-friendly. Indicators will be re-evaluated after more statistical analysis of launch scrub data is collected.
• Vehicle Data: The KLI shown are judged against the criteria for the Falcon 9 rocket launched by NASA. There are many other vehicles and many other launch providers around the world with different criteria for launch. Ultimately, it will be necessary to collect the various criteria and refine the KLI based on both the vehicle and the launch provider. Additionally, we wish to expand into vehicle reliability, based off of their individual scrub record and incorporate this into our calculations and create it's own KLI gauge.
• Forecast Analysis: WeatherToLaunch relies heavily on the Aeris weather forecasting API to produce KLIs starting at 14 days pre-launch. This API does not include any information about expected error/variance of the forecast. Therefore we are unable to provide any indication of how confident we are about the KLI’s accuracy. As launches get closer, we expect the forecast data to become more accurate but currently there is no way to know for sure. By analyzing how the forecast change over time, we should be able to calculate a level of expected “error” in the KLI ourselves, without relying on the weather API to do so for us. Additionally, being able to track the trend in the forecast as the launch window nears should provide an indication of whether the KLI are getting closer to the threshold or further away from the threshold as the launch nears. This will be presented nicely in stock graphs for the user.