Analysis of variable modification in the Keetch-Byram drought index using artificial neural network

• Published in: AIP conference proceedings Vol. 2508; no. 1
• Main Authors: Sukmono, Yudi, Suyatno, Addy, Buana, Gumelar Ardhi
• Format: Journal Article
• Language: English
• Published: 09.03.2023
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0126498

The variables in the original Keetch-Byram drought index calculation method (without modification) consisted of the average annual rainfall, the difference in humidity, and the maximum temperature. In Indonesia, drought is not only caused by reduced rainfall and high maximum temperatures. Several other variables also have the potential to strengthen the accuracy of the drought index calculation, namely wind speed, atmospheric pressure, and duration of solar irradiation. The problem is whether the addition of a new variable will change the accuracy of the calculation? If the new variable has a positive impact, which new variable has a significant effect on the calculation of the drought index. The analysis of the addition of variables in this study uses the method of training and testing artificial neural networks, namely the backpropagation algorithm. The training was carried out on many original variable data, then compared with testing using new variables. The results showed that the addition of a wind speed variable and replacing one of the variables with a wind speed variable showed an increase in prediction accuracy. Prediction accuracy using the three original variables only reached 86.67%, while with the addition of one wind speed variable the accuracy increased to 91.67%. This proves that the wind speed variable is a new candidate variable that has the potential to be one of the variables that play a role in the Keetch-Byram drought index calculation model in the future.