Figure 3
From: High-resolution precipitation monitoring with a dense seismic nodal array
Seismic precipitation measurements compared to rain gauge data. (a) Seismic power spectral density at the location of the rain gauge for a precipitation event on 17 April 2016. The profile is obtained by the weighted average seismic PSD sampled at several nearby stations. Red sections mark five individual precipitation sub-events, and are divided by black dotted lines. (b) Similar to (a), but smoothed. The gray shadow shows one standard deviation due to station averaging. All smoothing for panels in this figure done by convolving a Gaussian (10 min half-width). (c) The blue line shows smoothed precipitation rate from the rain gauge. Red sections are converted precipitation rate from seismic PSD by fitting each red PSD section in (b). to the rain gauge measurements in c. through linear regression after both are converted to log-scales. (d) The blue line shows raw rain gauge records, which often appear discretized due to the time intervals between bucket tips. Red lines show converted precipitation rate using the unsmoothed PSD in (a) and the regression relationship in (c). The gray line shows instantaneous precipitation rate from the ground-based radar. (e) Smoothed seismic PSD versus smoothed rain gauge precipitation rate in log-scale. Here, each line represents the evolution during an event. Different colors for events on different days (Fig. S1, each event may consist of multiple sub-events as separated lines). The black line shows the fitted linear relationship between the PSD and the precipitation rate using all sub-events except two dashed line outliers, and the 95% prediction interval for the fitting is characterized by the gray shadow. Fitted relationship and the 95% interval width are shown at the top left corner. Events 7 and 8 are not shown since they did not pass the rain gauge.
