Abstract: Objectives:  Parcel-based crop classification using multi-temporal satellite images plays a vital role in precision agriculture. However, exploring the multi-scale spatial information for identification of crop types from remote sensing images is a significant challenge.   Methods:  In this study, a tile-slice-based spatial-temporal method was developed for parcel-scale crop (type) classification using multi-temporal Sentinel-2 images. Central to this approach is the combined use of tile-slice-based representation of parcels in images and a deep spatial-temporal convolutional network. Firstly, spatial tiles were sliced according to crop types and parcels, to produce a tile-slicebased training sample dataset and a tile-slice-based prediction dataset. Then, the deep spatial-temporal convolutional network was established to estimate crop-type probabilities for tile-slice-based prediction dataset. Finally, tile-slicebased probabilities were fused under parcel polygons, to generate the final crop type maps.   Results:  The proposed method is further discussed and validated through parcel-based time-series crop classifications in the France study area with multi-temporal Sentinel-2 images. The classification results demonstrated great improvements in accuracy scores (0.03, 0.02, 0.02 in overall accuracy, precision and F1, respectively) over comparison methods.   Conclusions:  Through experiments and discussions, we concluded that: (1) the proposed tile-slice-based spatial-temporal method is effective for parcel-based crop classification using Sentinel-2 image sequences, (2) the spatial division and slice in parcel-scale representation are benefit more for crop type identification of larger and spatially concentrated parcels, (3) the proposed TSST-Net is lightweight but powerful for crop classification. The above experiments and their conclusions will provide new ideas for parcel-based agricultural remote sensing time series analysis.