Weather predictions could be used to give decision support guidelines for the agricultural management. The agricultural yield productivity depends on crop protection and its effective management, which could be increased by avoiding losses due to the effect of low and high-temperature damage. In Maharashtra state, the seasonal changes affect potential losses of susceptible crops and livestocks, due to variations in temperature. Therefore, it is essential to give effective decisions which are used to avoid damages from the extreme weather conditions. The main goal of the study is to give predictive weather analytics in agro-climatic scarcity region using the iterative approach. In this study, predictive weather analytics approach is proposed based on iterative linear regression and polynomial regression predicting methods. The study region of research falls in the agro-climatic scarcity region having inadequate rainfall, variation in temperature, and dry land. So, the proposed iterative approach based on linear methods are applied and designed to predict future conditions. Both models provide findings that are useful for the future farming and management. Also, comparative study and plots are depicted for the actual and estimated maximum and minimum values of temperature, humidity, and rainfall results of the proposed approach. The effectiveness and performance are tested using the statistical tests and error measure statistics. The prediction accuracy of the maximum and minimum temperature, humidity, and rainfall at the 95% confidence level is tested. The estimates using iterative polynomial approach is much better compared to iterative linear regression approach. The statistical measures and error estimators, such as mean, variance, median, standard deviation, kurtosis, mean squared error, root-mean-squared error and correlation coefficient are evaluated and discussed in detail. Lastly, the statistical hypothesis testing methods such as ANOVA, t test, f test and z test are used in this study to test the actual and predicted data of the model. Also, the frost conditions and occurrence probability in the estimated year 2013 (20% in the month May and 38.33% in December–January) are noted. The results of proposed approach demonstrate the visualization, accuracy, and suitability for the managing decisions in the agricultural products and livestock.