Artificial Neural Network for Markov Chaining of Rainfall Over India

Artificial Neural Network for Markov Chaining of Rainfall Over India

Kavita Pabreja (Maharaja Surajmal Institute, GGSIP University, India)
Copyright: © 2020 |Pages: 14
DOI: 10.4018/IJBAN.2020070105
OnDemand PDF Download:
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Rainfall forecasting plays a significant role in water management for agriculture in a country like India where the economy depends heavily upon agriculture. In this paper, a feed forward artificial neural network (ANN) and a multiple linear regression model has been utilized for lagged time series data of monthly rainfall. The data for 23 years from 1990 to 2012 over Indian region has been used in this study. Convincing values of root mean squared error between actual monthly rainfall and that predicted by ANN has been found. It has been found that during monsoon months, rainfall of every n+3rd month can be predicted using last three months' (n, n+1, n+2) rainfall data with an excellent correlation coefficient that is more than 0.9 between actual and predicted rainfall. The probabilities of dry seasonal month, wet seasonal month for monsoon and non-monsoon months have been found.
Article Preview
Top

Background

Synoptic, numerical and statistical methods have been used for forecasting of rainfall by various authors viz. (Tyagi et al., 2011); (Chatterjee et al., 2009); (Iyengar & Basak,1994). Time series technique which is one of the important statistical techniques has been used by many authors as stated in (Sengupta & Basak, 1998); (Iyengaer, 1991).

Markov model, a statistical approach, has been applied to provide forecasts of weather states (dry or wet day) for some future time depending upon values of weather variables given by current and some previous states. Markov chains specify the state of each day as wet or dry and generate a relation between the state of current day and the states of preceding days. The order of Markov chain depends upon the number of preceding days under consideration. Sorup et al. (2011) observed that the first order Markov chain is quite important whereas the second order Markov Chain is even more significant. The authors also found that that there is no noticeable difference between the model parameters of first and second order. It has been observed that the wet day of previous two time periods affect positively the wet day of current time period in the rainy season as compared to the dry day of previous two time period as explained by Hossain and Anam (2012). A 3-state Markov chain model with five independent variables for rainfall forecasting in Haryana, India, has been applied by Aneja and Srivastava (1999). Dash (2012) has found that the first order Markov chain is able to provide value for the precipitation occurrence for all months in Odisha, India.

Complete Article List

Search this Journal:
Reset
Volume 9: 4 Issues (2022): 1 Released, 3 Forthcoming
Volume 8: 4 Issues (2021)
Volume 7: 4 Issues (2020)
Volume 6: 4 Issues (2019)
Volume 5: 4 Issues (2018)
Volume 4: 4 Issues (2017)
Volume 3: 4 Issues (2016)
Volume 2: 4 Issues (2015)
Volume 1: 4 Issues (2014)
View Complete Journal Contents Listing