In situ and remote sensing platforms offer the potential to estimate precipitation rates at fine temporal and spatial resolutions. A wealth of research has identified meteorological scenarios in which a given sensor may yield inaccurate estimates of rainfall or snowfall. For example, rain gauges underestimate rainfall rates when wind speeds are high. Being an in situ instrument, they can also misrepresent rainfall fields that often exhibit high spatial variability. More recently, algorithms have been developed to utilize multiple data sources in a single precipitation product. This multisensor approach has been refined and developed into a real-time algorithm.

The precipitation algorithm presented here is called Quantitative Precipitation Estimation and Segregation Using Multiple Sensors (QPE SUMS). It ingests and utilizes level II WSR-88D radar data, infrared satellite data, rain gauge observations, lightning flashes, and RUC-2 model output to provide rainfall and snow-water equivalent estimates every 5 minutes on 1x1 km grid cells. The weight applied to each sensor varies with the character of the meteorological conditions under observation. For example, a bright band identification algorithm has been devised to detect artificially high reflectivity associated with melting layers in stratiform precipitation. This detection will then prohibit radar reflectivity observed in these regions from being used in the final product. QPE SUMS uniquely and adaptively calibrates infrared satellite data using uncontaminated radar data (ie. measured below the bright band). This calibrated product is used to “fill in gaps” in regions where estimates from radar alone are known to be inaccurate.

This presentation will describe each component of the algorithm. In addition, examples will be shown how the multisensor approach yields rainfall estimates that are different and more accurate than radar-only products in three separate domains covering the states of Arizona, Oklahoma, and North/South Carolina.