Evaluation of Specific Humidity over Nigeria using Artificial Neural Network

Abstract

Weather forecasting is the application of science and technology to predict the state of the weather for a future time at a given location using quantitative data of past or present experiences. In this paper neural network–based autoregressive moving average with exogenous inputs (NNARMAX) and autoregressive moving average with exogenous inputs (ARMAX) models were used to obtain specific humidity (q) from the meteorological parameters  obtained from the archives of Nigeria Meteorological Agency NIMET, Oshodi Lagos, Nigeria. The data which covers a ten year period (1999-2008) were the daily temperature and relative humidity data taken at 09:00 hour and 15:00 hour over sixteen stations evenly distributed across Nigeria. The results showed that the two models could be applied to predict specific humidity (q) at all the selected stations. The performance evaluation mean square error (MSE) for training and validation error (MSTE & MSVE) that were obtained at most of the stations  showed that the NNARMAX model yielded better performances than the ARMAX model for instance, at Lagos, the mean square validation error (MVE) for training at 09:00 hour are 0.0007 and 0.2396 for NNARMAX and ARMAX respectively.

Keywords: Weather Forecasting, Artificial Neural Networks, ARMAX model, time series.