Geospatial Dataset of Forest Canopy Height and Structural Regionalization in La Malinche, Mexico: 2019 Random Forest–Derived Canopy-Height and Structural Products and Structural Products Derived from the 2020 Global Canopy Height Map

La Malinche Canopy Height and Structural Regionalization (LMCHSR) is a public geospatial dataset of forest canopy height and structural regionalization for an area of interest encompassing the La Montaña Malinche o Matlalcuéyatl Federal Protected Natural Area, Mexico, and the adjacent peripheral area included in the analysis. The 2019 canopy-height raster was produced by integrating three Random Forest models corresponding to Sentinel-2 Level-2A acquisitions on 22 January, 28 March, and 12 April 2019. Each model used the red (B4), red-edge (B5), and shortwave-infrared (B12) bands, together with elevation and slope, as predictors. Canopy heights from CHMv2 (Brandt et al., 2026) served as the response reference after spatial harmonization to the 10 m modeling grid. The date-specific predictions were integrated into a multitemporal canopy-height surface. Structural products were derived from this RF surface on a 20 m analytical grid. They include structural objects, aggregated regions, five canopy-height classes (S01–S05), and connected class patches. An independent set of 20 m structural products was derived from the 2020 Global Canopy Height Map of Lang et al. (2023), with separately determined objects, regions, and five classes (E01–E05). The harmonized source ETH canopy-height raster is not redistributed. The release contains 18 geospatial assets: one area-of-interest boundary, the 2019 RF canopy-height raster (RF 2019), eight RF-derived structural products, and eight ETH2020-derived structural products. Raster and vector files are supplied as GeoTIFF and GeoPackage files, respectively, using the WGS 84 / UTM zone 14N coordinate reference system (CRS; EPSG:32614). The release also includes metadata, dictionaries, a class legend, and Secure Hash Algorithm 256-bit (SHA-256) checksums. The two structural classifications were generated independently; matching class codes do not imply identical class definitions or direct temporal forest change.