Using Self-Organizing Maps to Identify Coherent Precipitation Regions: Application to Extreme Precipitation Mechanisms

Tuesday, December 11, 2018 - 13:40
Add to Calendar

Identifying areas with similar precipitation characteristics has long been done by consensus and intuition. A common way of dividing the CONUS typically uses eight or nine groups of states. Given that there is no inherent meteorological reason that precipitation must change at political boundaries an objective method to define regions with similar precipitation characteristics is developed. We train self-organizing maps on the annual cycle of precipitation for each grid point to find coherent regions with matching annual cycles. As expected these precipitation regions align with large scale geographical features and are readily interpretable in an intuitive way. The correct number of regions seems to be between a dozen and a half dozen. The number is dependent on whether emphasis is placed on the ease of making composites of events within a region or on the consistency of the regions between independent time periods of data respectively. Extreme six-hour precipitation in each region is examined. Our examination focuses on the mechanisms which create extreme precipitation such as: fronts, convection, and cut-off-lows, etc. Events often spring from a mixture of these mechanisms, so a weighting system to measure the relative magnitudes of these mechanisms on each event is created. Some regional statistics will be shown as well.

Link for More Information: