A method to quantitatively estimate rainfall rate above water surface based on different spectrum shapes generated by rainfall

Rainfall measurements at sea are very difficult to make because rainfall is sharply discontinuous in both time and space.Distinct ambient sound spectra for various rainfall rates can be identified through a series of discrimination processes or vice versa.In this paper,a method combined with the nonlinear least-square regression to quantitatively estimate rainfall rate over water surface from the different spectrum shapes generated by rainfall in some frequency bands is presented.About 2000 min spectrum data generated by rainfall have been collected in an open-lake in Xiamen city.As a result,the spectrum shape in 15~30 kHz is fitted to log-linear law which allows classification of rain into three categories from the spectrum slope:no rain,drizzle (rainfall rate ＜10 mm/h) and heavy rain (rainfall rate ＞10 mm/h). Then,rainfall detection is made and rainfall rate is quantified in drizzle time with Logistic model in the frequency bands of 2~15 kHz,and the rainfall rate in heavy rainfall is also quantified with log-linear regression.Finally,all measured spectrum data are used to calculate rainfall rate with the algorithm.The results show that it is comparably coincidente with the synchronously measured ones.The average error of accumulative rainfall per min is only 1%~4%.