Identifying　agricultural　machinery　operations　is　crucial　for　enhancing　agricultural　productivity　and　promoting　the　transition　to　data-driven　agriculture.　Current　research　focuses　solely　on　administrative　divisions,　overlooking　the　links　between　machinery　movement,　natural　spatial　patterns,　and　spatiotemporal　dependencies.　The　direct　clustering　of　GNSS　points　is　inefficient　and　incurs　substantial　computational　costs.　In　response　to　these　challenges,　we　introduce　an　unsupervised　clustering　method　based　on　multiscale　spatiotemporal　partitioning,　which　systematically　integrates　spatial　and　temporal　dimensions　to　analyze　GNSS　trajectory　data.　By　designing　multiscale　grids　and　temporal　partitions,　we　efficiently　processed　high-dimensional　trajectory　data　by　employing　t-SNE　and　K-means++　algorithms　for　dimensionality　reduction　and　clustering,　and　the　visualization　validated　the　clustering　effectiveness.　When　applied　to　GNSS　data　from　the　wheat　harvest　season　in　China,　the　results　revealed　distinct　patterns　of　harvester　movement,　including　trans-regional　movement　trends.　The　geogrids　are　clustered　into　four　groups,　each　of　which　exhibits　a　distinct　spatiotemporal　relationship.　A　combined　geogrid　analysis　with　administrative　regions　identified　Anhui　as　having　the　highest　flow　density,　whereas　Henan　had　the　most　concentrated　areas　of　trans-regional　harvester　flow.　These　findings　offer　valuable　insights　for　planning　harvester　operations,　particularly　in　trans-regional　harvester　management,　by　understanding　complex　spatiotemporal　dynamics.