Monthly Rainfall Prediction Using SARIMA models for 63 Mandals of Anantapur District, Andhra Pradesh, India
N. Ashokkumar
*
Department of Irrigation and Drainage Engineering, Dr. NTR College of Agricultural Engineering, Bapatla-522101, Acharya N.G. Ranga Agricultural University Andhra Pradesh, India.
M. V. Ramana
Administrative Office, Acharya N.G. Ranga Agricultural University, Lam- 522 034, Guntur (Dist), Andhra Pradesh, India.
M. Raghu Babu
Department of Irrigation and Drainage Engineering, College of Agricultural Engineering, Madakasira-515301, Acharya N.G. Ranga Agricultural University, India.
P. Prasuna Rani
Agricultural College, Bapatla-522101, Acharya N.G. Ranga Agricultural University Andhra Pradesh, India.
B. Ravindra Reddy
Department of Statistics, Agricultural College, Bapatla-522101, Acharya N.G. Ranga Agricultural University Andhra Pradesh, India.
*Author to whom correspondence should be addressed.
Abstract
Anantapur district lies between 13º40’ to 15º15’N latitude and 76º50’ to 78º30’E longitude. The geographical area of the district is 19,130 km2. Anantapur has a semi-arid climate, with hot and dry conditions for most of the year. Monthly rainfall forecasting was done using Seasonal autoregressive moving average (SARIMA) model by testing 267 combinations using SPSS 26 software. Monthly rainfall forecasting was done using Seasonal autoregressive moving average (SARIMA) model by testing 267 combinations using SPSS 26 software. SARIMA models for prediction of monthly rainfall using SPSS 18 software. The best predicted model has been selected based on the maximum coefficient of determination (R2), minimum Bayesian Information Criterion (BIC) and minimum Mean Absolute Error (MAE). In Cluster 1, the SARIMA (0,1,0) (1,1,0) ₁₂ model exhibit strong reliability, notably for Uravakonda, which attained the highest overall R² (0.945) and a comparatively low Normalized BIC (7.656). Within cluster 2, Narpala emerges as the most accurately modeled Mandal (SARIMA (0,0,0) (1,0,1) ₁₂) in this cohort (R² = 0.933), accompanied by the lowest MAE (12.214) and the lowest Normalized BIC (6.671), indicating a high-quality fit. In cluster 3, the highest-performing models within this set are represented. Yellanur attained the best overall outcomes, with an average of 0.944, the lowest MAE at 10.122, and the lowest Normalized BIC at 6.543. Although most Mandals performed well, Hindupur exhibited the lowest accuracy in this cluster. In Cluster 4, SARIMA (0,1,0) (1,1,1)12 model exhibit moderate to high consistency. This cluster was led by Kundurpi, whereas Kambadur displayed the lowest fit. Notably, Pamidi achieved a highly efficient fit, evidenced by a low BIC of 7.033, despite its moderate MAE. Cluster 5 contains the most varied results. Chenne Kothapalle is a standout, exhibiting a high score of 0.791 and an exceptionally low normalized BIC of 6.490, which indicates a highly parsimonious and accurate model (SARIMA (0,1,0) (1,1,1)12).
Applicability of the SARIMA model can be used with fair accuracy for the real time forecasting of monthly rainfall. The results of the SARIMA model will be very useful for planning agriculture and plantations in Anantapur district which is highly dependent of rainfall. Prediction of climatological parameters such as rainfall is very important. Since rainfall if the main source of water availability on the earth. Using time series forecasting model such as SARIMA is a powerful tool to investigate the future rainfall pattern over the regions. This future forecasting may be estimated as policy driven paper to government and local bodies.
Keywords: SPSS, SRIMA, cluster, mandals and rainfall