Accurate forecasting of the appropriate pollination period in apple flowers remains one of the most significant problems in modern pome fruit growing. This research proposes a new approach using deep learning (DL) model using CNN-LSTM to identify the best time for apple flower pollination from temporal image frames. This model treats eight time points of the apple flowers imagery sequentially, where the ith image is related to the growth stage importance metrics and the flower class information. This model successfully and significantly predicted flower pollination status using a dataset of flower cluster images and their pollination status labels, predicted accuracies depended on the chosen threshold ranging from 0.1 to 0.3. To identify the best days of pollination, the research uses a convolutional neural network for the spatial feature extraction and the long short-term memory network for temporal pattern recognition. The heatmap illustrations indicated differences in temporal trends in pollination importance, where peak values occur in specific developmental phases. This model was useful in classifying flowers into pollination ready and non-ready states, making it easy for the orchardist to time the pollination rightly. This study presents an automated approach which provides a leap forward in terms of precision agriculture, whereby the ability to better time pollination can help increase fruit yield. The results of our study prove that for the prediction of dynamic pollination status in apple orchards, for this DL is a valid option and can aid in better orchard management. Highlights