Applying Time Series and a Non-Parametric Approach to Predict Pattern, Variability, and Number of Rainy Days Per Month
Alamgir Khalil1, Subhan Ullah2, Sajjad Ahmad Khan3, Sadaf Manzoor4, Asma Gul5, Muhammad Shafiq6
More details
Hide details
1Department of Statistics University of Peshawar
2M.Phil Scholar Department of Statistics Allama Iqbal Open University Islamabad
3Department of Statistics Abdul Wali Khan University Mardan
4Department of Statistics Islamia College University Peshawar
5Department of Statistics Shaheed Benazir Bhutto Women University Peshawar
6Department of Economics, Kohat University of Science and Technology, Kohat
Submission date: 2016-02-11
Final revision date: 2016-09-06
Acceptance date: 2016-09-08
Online publication date: 2017-03-22
Publication date: 2017-03-22
Pol. J. Environ. Stud. 2017;26(2):635–642
In the past 100 years, the annual global temperature has increased by almost 0.5ºC and is expected to increase further with time. This increase in temperature negatively affects the management of water resources globally as well as locally. Rain is an important phenomenon for agriculture, particularly in hilly areas where there is no feasible irrigation system. The present study is concerned with the analysis and modeling of the rain pattern, its variability, and prediction of monthly number of rainy days for the Abbottabad District, which is considered to be one of the greenest and most beautiful areas of Khyber Pakhtunkhwa, Pakistan, by incorporating both parametric and nonparametric techniques. Non-parametric statistical techniques are used for movement detection and significance testing; in this context, statistical tests were incorporated for inspection of homogeneity of rainy days among successive periods. A time series data for the period 1971-2013 was analyzed. Box Jenkins methodology and time series decomposition were applied for fitting the selected model, which was assessed for forecasting the monthly number of rainy days for 2015-2020. In this study several time series parametric and non-parametric approaches were applied to model rainfall data. The results showed that SARIMA (1, 0, 1) (0, 1, 1) was a better choice in predicting the monthly number of rainy days. Further analysis of the data suggests that January, March, May, July, and December have a considerable declining tendency in the number of rainy days.