Towards a better use of the time dimension and periodic revisits of near real-time products

It is 17:30. You ask to display the radar rainfall accumulation from 16:00 to 17:00, a product that was generated at 17:00. Yet, since that time, new information has arrived: additional non-radar information is now available, as well as radar-derived information from after 17:00. Both shed new light that could affect the computation of past rainfall accumulations, or of any other radar product. But we do not use such information, because we always act as if data had only been collected at time T, and not before or since, a practice inherited from times when computing resources were much more limited.

This is a call for renewing the way radar products are generated by taking into account information available before the time of interest as well as since the time of interest. In this work, we first try to evaluate how past and future information can be used to improve the cleaning of radar data (e.g., clutter suppression, attenuation correction, etc) as well as the computation of hydrological products such as rainfall accumulation (e.g., improved estimate of bright band height). Then, we define the path to follow to implement this vision.