Intercomparison of Gauge Performance during Harvey's Extreme Rains

Hurricane Harvey dumped historic amounts of rain onto a heavily-populated portion of Texas and Louisiana. The unusually large density of rain gauges from conventional and flood-control networks, combined with large numbers of backyard-type gauges operated by the general public, presents a unique opportunity to compare the performance of different rain gauge networks during a single real-world rainfall event. We include data from Weather Underground and Earth Networks as well as from members of the public who voluntarily submitted their own observations to the Office of the State Climatologist of Texas. We find that systematic network biases are generally small when compared to a common background precipitation analysis (Stage IV), except for voluntary observations which are biased high (perhaps because they are not a random sample). Variance within networks is higher for nontraditional networks than for traditional networks. Few systematic differences are found between tipping bucket and cylinder gauges. Across all gauge types, the gauge-radar comparison implies that the Stage IV analysis underestimates measured rainfall totals of 800 mm or more.