Rainfall Estimation Using Artificial Neuron Networks (ANNs) and FY-2E Satellite Image in Thailand

Abstract

The objectives of this research were to estimate the brightness temperature from remotely sensed data as FY-2E satellite, and to evaluate rainfall amount by using Artificial Neural Networks (ANNs) model that require brightness temperature for assessment. The percentage error and root of the mean square error (RMSE) were employed to calibration and validation approach. The data were compiled during 1 July to 31 December 2010 and bands resolution within FY-2E were used as IR1, IR2, WV and IR4 for find out brightness temperature.

The results showed that the average monthly brightness temperature indicated that maximum in winter season and minimum in rainy season due to the wetness of land cover affected to decrease surface temperature. The relationship between brightness temperature and rainfall amount were no significant in every region of Thailand due to the fact that rainfall amount were caused by various factors and atmospheric condition. The estimated rainfall amount from ANNs model were found that the highest appeared in August while the lowest was showed in December in every region of Thailand. The ranged of correlation between rainfall estimation and observation was moderated to high as 58 – 99 percentage in which high correlation was found in heavy rainy period and moderately showed in dry period. The model calibration and validation were found that the percentage error showed a little bit low between 29.96 to 40.72 and the root mean square was similar pattern between 29.07 to 45.62.

Keywords: Artificial Neuron Networks, Satellite, FY-2E, Rainfall Estimation