Rainfall is one of the most difficult elements of hydrologic cycle tomeasure and forecast. This is due to the tremendous range of variability it displays over a wide range of scales both in space and time. Weather radar constitutes an attractive possibility for improving the description of rainfall fields. Radar emits electromagnetic energy in narrow bands.From the reflected energy that returns to the transmitter it is possible to obtain measurements of the rainfall field. In practice, the exclusive use of radar is yet to be achieved and rain gauge or other punctual systems are required to calibrate radar. The identification of the relationshipbetween radar reflectivity factor and rainfall intensity is crucial for both thecalibration and operational phases. In this framework, the aim ofthe paper is twofold:(1) to develop a non-parametric approach which is more flexible and offersmore generalisation capabilities than the Marshall and Palmer relationship, and(2) to use a vector of all 11 reflectivity factors at different CAPPI levels.For this purposes, three kinds of neural networks (NNs) are developed:the multi-layer perceptron (MLP), radial basis function networks (RBFNs)and Bayesian networks (BNs) Models are trained and tested using a real data set of reflectivity observedby the Monte Grande weather radar (Teolo, Italy) and rainfall intensitymeasured at five rain gauges in the Cortina d'Ampezzo area (NorthernItalian Dolomites), during the June 12, 1997 storm event(from 11.15am to 12.00pm).
Neural network identification of Z-R relationships / Orlandini, Stefano; Morlini, Isabella. - STAMPA. - (2000), pp. ---. (Intervento presentato al convegno EGS XXV General Assembly tenutosi a Nizza, Francia nel 25-29 Aprile 2000).