Development of a neural network based algorithm for rainfall estimation from radar observations

Rainfall estimation based on radar measurements has been an important topic in radar meteorology for more than four decades. This research problem has been addressed using two approaches, namely a) parametric estimates using reflectivity-rainfall relation (Z-R relation) or equations using multiparameter radar measurements such as reflectivity, differential reflectivity, and specific propagation phase, and b) relations obtained by matching probability distribution functions of radar based estimates and ground observations of rainfall. In this paper the authors introduce a neural network based approach to address this problem by taking into account the three-dimensional (3D) structure of precipitation. A three-layer perceptron neural network is developed for rainfall estimation from radar measurements. The neural network is trained using the radar measurements as the input and the ground raingage measurements as the target output. The neural network based estimates are evaluated using data collected during the Convection and Precipitation Electrification (CaPE) experiment conducted over central Florida in 1991. The results of the evaluation show that the neural network can be successfully applied to obtain rainfall estimates on the ground based on radar observations. The rainfall estimates obtained from neural network are shown to be better than those obtained from several existing techniques. The neural network based rainfall estimate offers an alternate approach to the rainfall estimation problem, and it can be implemented easily in operational weather radar systems.

[1]  V. N. Bringi,et al.  Differential reflectivity and differential phase shift: Applications in radar meteorology , 1978 .

[2]  V. N. Bringi,et al.  Potential Use of Radar Differential Reflectivity Measurements at Orthogonal Polarizations for Measuring Precipitation , 1976 .

[3]  Francesco Palmieri,et al.  MEKA-a fast, local algorithm for training feedforward neural networks , 1990, 1990 IJCNN International Joint Conference on Neural Networks.

[4]  Eugenio Gorgucci,et al.  Radar and Surface Measurement of Rainfall during CaPE: 26 July 1991 Case Study , 1995 .

[5]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .

[6]  Mahmood R. Azimi-Sadjadi,et al.  Fast learning process of multilayer neural networks using recursive least squares method , 1992, IEEE Trans. Signal Process..

[7]  M. Sachidananda,et al.  Rain Rate Estimates from Differential Polarization Measurements , 1987 .

[8]  Louis J. Battan,et al.  Radar Observation of the Atmosphere , 1973 .

[9]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[10]  Kultegin Aydin,et al.  Compared With Ground-based Rain Gauges During MAYPOLE'84 , 1990 .

[11]  Ken-ichi Funahashi,et al.  On the approximate realization of continuous mappings by neural networks , 1989, Neural Networks.

[12]  Richard P. Lippmann,et al.  An introduction to computing with neural nets , 1987 .

[13]  J. Marshall,et al.  THE DISTRIBUTION OF RAINDROPS WITH SIZE , 1948 .

[14]  R. Lippmann,et al.  An introduction to computing with neural nets , 1987, IEEE ASSP Magazine.

[15]  Kultegin Aydin,et al.  Polarimetric Radar Measurements of Rainfall Compared With Ground-Based Rain Gauges , 1990 .