Previous PageTable Of ContentsNext Page

SPATIAL ESTIMATION OF RAINFALL USING ARTIFICIAL NEURAL NETWORK

D. Midya2, A. K. S. Huda2 and W. Zhou3

2School of Agriculture and Rural Development, 3School of Food Sciences
University of Western Sydney, Hawkesbury, Richmond NSW 2753

An artificial neural network (ANN) model was developed for spatial estimation of rainfall. Annual rainfall data (1901-1975) for 87 stations in the Hunter Valley were accessed from MetAccess and Australian Rainman. ANN model was trained using above 87 data points. The model was validated by estimating rainfall for 20 independent stations.

RESULTS AND DISCUSSION: Latitude, longitude and elevation for the stations used for model development (87 stations marked by ┼) and validation (20 stations marked by •) are given in Fig. 1. Trained ANN model explained 99% of total variation observed within 87 data points used for model development. The model predicted annual rainfall well for 19 of 20 independent stations (Fig. 2).

Fig. 3 compares the actual mean annual rainfall of eight stations with the estimated values using ANN and spline (Hutchinson, 1983) methods. Based on the preliminary results, it is concluded that the ANN approach appears to be promising for further investigations.

Figure 2. Validation of ANN model

Figure 3. Comparison of ANN with spline

REFERENCE: Hutchinson, M. F. 1983. A new method for estimating the spatial distribution of mean seasonal and annual rainfall applied to the Hunter Valley, New South Wales. Aust. Met. Mag. 31(1983) 1979-184.

Previous PageTop Of PageNext Page

...

Search Site