|
|
vol. 5, nr. 1 (2005)
|
Evaluation of artificial neural networks application possibility for mean annual rainfall erosivity factor value estimation
|
|
|
Paweł Licznar
|
|
| (get PDF )
|
|
|
Institute of Building and Landscape Architecture, University of Agriculture pl. Gruwaldzki 24, 50-363 Wroclaw |
|
|
vol. 5 (2005), nr. 1,
pp. 65-74
|
|
|
abstract:
Rainfall erosivity factor is one of the most poorly assessed parameters of Universal Soil Loss Equation (USLE) in Poland. Evaluation of artificial neural networks application possibility for mean annual rainfall erosivity factor value estimation only on the basis of known mean monthly precipitation was the main aim of the research. The research was based on the database from 90 gauging stations from the area of Poland and Germany. Within the frame of the conducted evaluation, the application possibility of single or double hidden layer perceptron and radial base networks for a chosen aim realization was examined. It was proved that satisfactory results of mean annual rainfall erosivity factor values estimation were obtained on the basis of known mean monthly precipitation by means of perceptron artificial neural networks of one or two hidden layers and radial base networks. At the same time, the research showed a failure of Modified Fournier Index application for this purpose.
|
|
keywords:
rainfall erosivity factor, artificial neural networks, Modified Fournier Index
|
|
original in:
Polish
|
|
|
|
|
|
