Radar-based rainfall thresholds for debris flow warning: A review of opportunities, effect of estimation uncertainties, and assessment of key challenges

The increasing availability of weather radar precipitation products provides new opportunities to improve upon existing methods for debris flow warning. The aim of this work is to examine how different characteristics of precipitation products, derived either from raingauges or from weather radar, may impact on the identification and use of precipitation thresholds that are used for debris flow warning. Precipitation exhibits space and time variability at all scales leading to high uncertainty in raingauge-based rain estimation. One distinct feature of the precipitation estimation problem for raingauge-based threshold relationship identification and use, is that the triggering precipitation to be estimated at the debris flow location exceeds an actual threshold which is likely not to be exceeded at the measuring raingauges. Recent results has shown that these characteristics may lead to biased precipitation threshold identification and low warning efficiency. Weather radar monitoring represent an interesting alternative for precipitation threshold identification, overcoming the sampling problem of point measurements. However, despite long-standing efforts, radar derived estimates are still affected by considerable uncertainties, particularly in the rough topography terrain typical of debris flows. It is therefore important to understand how uncertainties due to either rainfall sampling (typical of raingauges) or to rainfall estimation (typical of weather radar) propagates through the precipitation threshold identification methodology. Results are presented for a set of 10 high intensity debris-flow triggering storms that impacted the Southern Tyrol Region (Eastern Italian Alps) during the last decade. The region is characterized by rough orography, with elevation ranging from 300 to 4000 m asl, and it is monitored by a raingauge network with an average density of 1/70 km2 and a well calibrated and maintained C-band Doppler radar. High quality radar rainfall estimations are obtained taking into account both vertical (VPR) and radial (attenuation, screening) sources of error, and are adjusted with raingauges to provide reference rainfall estimates at the ground. Radar- and raingauge- based precipitation thresholds are identified and are compared, showing that a bias arises when using raingauge measurements for threshold assessment. The bias is related to the spatial variability characteristics of the considered storms and to the relative geometry of raingauges and debris flows. Even though the weather radar estimation uncertainty also impact the precipitation threshold identification methodology, no bias is reported for this last case. This provides a basis to identify opportunities in the use of radar-based estimates for debris flow warning in alpine regions. Key challenges are also identified, including the requirement for high quality, high resolution radar-based precipitation reanalyses and problems in areas with mixed or poor radar-coverage.