Forecasting of precipitation patterns in the Agra district in the context of anthropogenic climate change (2022-2031)
B.S. GARDNER, N.L. CHAUHAN, SH. GARDNER
St. John’s College, Dr. Bhimrao Ambedkar University, Agra, India
Keywords: precipitation trends, augmented Dickey-Fuller test, seasonal autoregressive integrated moving average, time series analysis, model approval, water resources management
Abstract
Precipitation is a key indicator for determining the climate of a region. However, erratic precipitation patterns are detrimental to urban infrastructure. The objective of this study is to forecast precipitation patterns using historical time series data. The study used precipitation data for a century (1922-2022) to project future precipitation trends for the semi-arid region of the Agra district of Uttar Pradesh, India. A seasonal autoregressive integrated moving average (SARIMA) model was employed for the same purpose. Through data decomposition and the augmented Dickey-Fuller test, the study confirms the model’s stationarity prerequisites, with the SARIMA (2, 1, 1) (0, 1, 1)12 configuration proving to be the most accurate for the study. The model accuracy was further validated through residual checks and performance assessment against established metrics, such as the Akaike information criterion, Bayesian information criterion and Hannan-Quinn information criterion. Forecasts were generated at 80 % and 95 % confidence intervals using both the “pm.auto_arima” function of the “pmdarima” package in the Python library and manual assessment for further adjustment and verification. According to the forecasts, 2022-2031 is a decade of reduced precipitation, with an average of 590 mm. This forecasted average is lower than the historical average of 643,96 mm and the median of 665,92 mm.
|
