Query goals

Please add your ideas and wishes for what questions our weather Oracle should be able to answer. This structuring into three types of queries I was talking about and which I also noted down here can be discussed/discarded/extended. Even if keeping that structure we do not have to implement all types and should prioritize something for the presentation (If possible, please note down what that would be for you). Something additional: If you have ideas for a good name for our project (other than bccn_prog_weather_2016), please share them!

Instant:

Simple queries that can be answered in real time, maybe accessible from the GUI

• What is the weather like {today, tomorrow, on the weekend}? ( for a given location) --> all parameters
• What is the weather like {today, tomorrow, on the weekend}? ( for Germany --> heat map) --> one parameter(?)
• Define a parameter and a date and see a heatmap of Germany with the values of this parameter visualized (i.e. historical)
• Prediction data: given a parameter, a data source(website), and a date in the future, visualize this over Germany --> daily, will use the nearest possible forecast
• Hourly: given a location, and a data source, show the predicted weather forecast for the available future (graphs with values parameter over time
• Hourly/daily map: given a data source (or taking mean of the sources) and a time interval, show one parameter on a map of Germany as an animation of hourly/daily maps.
• Visualize two parameters with one being the function of the other. This can include temperature dependent on date, but also rain dependent on teperature etc. The second parameter (date in the first case, temperature in the second) can be bounded appropriately.
• See whether there is a weekend bias

PDF Reports:

Bundles of queries, only few open parameters that can take a while to process, output will be in a PDF file

• Show the evolution of the monthly/yearly mean for one weather station for a time interval as big as possible, is there a statistically relevant increase?
• Given a certain time interval/number of time points e.g. the weekend: How much was which weather provider off? E.g. plot the mean error on the y axis and the time distance on the x axis

Application to more complex questions:

Use our query engine to investigate in something, e.g. comparison with health data, reality-check of weather sayings,...

• From our data, can we prove climate change and if yes, how much in which location?
• Summarizing and weighting the prediction quality for one day and one provider: Are there patterns? (e.g. weekend?)
• Compare accuracy of providers for different time spans and parameters.
• Find more or less funny correlations (health data, unemployment data, ice cream sales, ...)
• What does it mean that it will rain X%? Decrypting this based on the actual outcomes.
• Given that it rained for x days, how likely is it that it rains the following y days? We can plot the graph of tis function for x, y in (0,4) or so as a 3d surface ;).
• Visualize the (conditional) distributions of parameters (for example temperature or temperature given the time of the year or temperature given the rain amunt etc...)

Names:

Oracle

Greek gods of weather

Weather gods (wikipedia)