| Popis: |
Precipitation is instrumental for the existence of life forms on the globe. The agricultural sector, the economic mainstay of the country, etc., depend majorly on the rainfall. Hence, the rainfall forecasting models hold an unprecedented significance for the same. This work’s main objective is to create an efficient, easy-to-implement rainfall forecast system that can promptly warn and alert the people of any unexpected downpours. As the attempt is to implement a simple system capable of giving noticeable results, the work starts with the assessment of rudimentary techniques like baseline/Naive forecast, seasonal persistence algorithm, autoregression. The observed improvement in the performance as the complexity of models increased led to the evaluation of a comprehensive model like autoregressive integrated moving average (ARIMA) that imbibes some of the positive traits of the above models. This work does not present a novel forecasting technique but instead conducts a comparative study of univariate analysis using various statistical modeling techniques. The models are assessed on their performances using the root mean square error (RMSE) values calculated by comparing the predicted rainfall intensity values of the model with that of the actual rainfall intensity values present in the dataset. The research and analysis undertaken for this work are to predict the rainfall intensity of Coonoor in Nilgiris district, Tamil Nadu using a comprehensive statistical approach. Python was used as primary platform for the implementation of the above models. The prediction model developed using the ARIMA technique with an RMSE score of 11.9604 proved to be a better and more efficient model than the other statistical models considered in the paper (Toth in J Hydrol 239(1–4):132–147, 2000 [1]). |
