possibility of employing pattern recognition to predict freezing occurrence

abstract accurate prediction of hourly minimum temperature is required for various crop models which simulate photosynthesis and transpiration. such data can be used for crop protection and reducing the crops injuries due to freezing as well. our objective of this study is employing trigonometric and pattern recognition (k-nn) approaches to evaluate their potential in prediction of hourly temperature for the whole 24 hours and also minimum temperature time occurrence. our observed data contain every 3 hours minimum temperature data for 16 years of synoptic mashhad climate station. various scenarios were employed to predict the minimum temperature for first of aban and first of ordibehesht using, 1 day, 7 days, 110 days and 315 days observed data for next day minimum temperature prediction. our results showed that if there is no full access or partly access to the minimum temperature data then the trigonometric function including sine function is able to reproduce the required data. k-nn approach showed that as the distance of data to target data decreased the accuracy of prediction increased.