Analytics for Rainfall Predictions

Objective

To provide the residents of Peterborough with a mobile app where they can enter journey information and that predicts the rainfall along the course of that journey. Aim to be to increase the number of residents using sustainable transport solutions (cycling, bus, walking) within the City by removing one of the main objections, namely concern over the weather.

Details

This project is funded as part of an InnovateUK Smart City competition and is delivered in partnership with Anglia Ruskin and Loughborough Universities and PECT a Peterborough based Community Action Group.

This project will generate individual, very accurate short-term rainfall forecasts specific to a location or planned journey. By using a combination of radar-based real-time rainfall data combined with real-time wind direction and wind speed data and local rain gauge data, this project will deliver much more accurate short-term forecasts than available through existing weather services and apps. By knowing exactly what rain is actually falling at a moment in time, and it’s location, this service looks to track the progress of individual cells that are ‘downwind’ of Peterborough and which can be predicted to arrive over the City within the next hour. By using existing real-time wind direction and speed data, available through the Peterborough Living Portal for 25 weather stations installed around the City, a prediction when each ‘Rain event’ will arrive over Peterborough will be calculated. These predictions will be updated every 5 minutes using the latest rainfall and wind data so the predictions will become more and more accurate as they approach the current time. These predictions will then be married to the actual starting location of the user within Peterborough and the planned final destination so that the forecast is specific to that user and to that particular journey.

Additional rainfall data from rain gauges located around the Peterborough region will be used to help calibrate or ‘ground truth’ the radar rainfall data to further increase confidence levels associated with the predictions. Historic rainfall data will also be used to automatically categorize different types of rain events using variables such as wind direction and intensity. These categories will then be used to further enhance the predictions by automatically categorizing each rain event as it happens and, based on historical observations, make predictions on the expected rainfall intensity. I.e. for a rain event category X the historical observation is that rain intensity reduces by 25% every 10km. Users will access these personalized and highly accurate rainfall predictions via the mobile web application.

A comprehensive community-based programme will engage with local residents and retailers to implement a rigorous user centred design methodology to determine the exact requirements for the system and determine the best way to interact with the users and maximise the benefits they receive from the project. The output will be a real-time personalized rainfall forecast service for Peterborough that can be accessed via a free mobile application as well as a rainfall related coupon scheme to engage with local retailers.

Benefits of using MAP

MAP makes it easier to combine different data streams and acts as a Smart City platform into which a broad range of data can be added. The calculation power and flexibility of MAP makes it possible to develop and apply complex models and apply them to these fused datasets in real time so ensuring the information is available to residents when they need it. By providing the entire back end analytics system we can spend our time concentrating on developing and testing the underlying models rather than focusing on the mechanics of merging datasets and delivering robust, fast and adaptable calculations.