Quantifying the response of any given system beyond their current condition to weather alone or together with other factors requires predicting realizations of future weather conditions. The predicted weather data should not only be sufficiently accurate, but also their time scale should be in accordance to the decision support system in which the studied system is being applied. Inclusion of predicted weather data with, for example, a crop process-based simulation model could provide valuable and timely information for evaluation of various management techniques to avoid potential losses or increase crop production and income. Weather analogue as a nonparametric approach is easy and accurate to use to achieve this goal. In this study a weather analogue modeling tool is presented for predicting daily weather data realizations that are based on a modification of the k-nearest neighbor approach. Our intent was to develop a tool to predict a realization of real-time daily weather data by introducing two different methodologies for the k-nearest neighbor approach. In the first approach ( k-mean), weather prediction for day t + 1 was assumed as the average of all days found as the k best match days for the target day. In the second approach we assumed that only a fraction of the observed data (target year) was available (e.g. 90, 120, and 150 days) and that the realization for the remainder of the year is of interest. Based on this approach, the model should recognize the most similar pattern to the available data of the target year among the same sequence of historical data. Daily weather data of the selected year as the best match would be considered for the remainder of the target year. Both approaches were compared with observed data from 16 locations in the USA, Europe, Africa, and Asia, representing different climatic regions. Employing the first approach ( k-mean), the k-NN model was quite promising and was able to recognize the pattern of the target year among the historical observed weather data for solar radiation, maximum and minimum temperature. However, the k-mean approach only reproduced the observed pattern of precipitation successfully when there was not a high variability in the pattern of precipitation occurrences. Using the second approach, as expected, a larger share of observed data in the target year beyond 90 days greatly improved the accuracy of prediction. However, after using 150 days both bias measures, e.g., MSD and MASE, slightly increased due to a change of the best match year. The results from this study showed that this weather analogue program could be a valuable tool for realization of any weather dependent function. There is also scope for incorporation of this tool with application of agricultural, ecological, and hydrological process-based simulation models.