Key points are not available for this paper at this time.
While DataCube is a concept of data storage in space (x, y or lon, lat) and time (t) dimensions. The concept accelerates the understanding and the usage of data as a defined space and time extent and is enormously helpful for geospatial analysis and modeling. Peter et al. (2018) describes this concept that generalizes the concept of seamless map from 2-D to n-D, and enables spatial analysis and multi-temporal analysis simultaneously. It is more trivial to realize in raster data due to its gridded structure, but not in the case of vector data, which is tabular data. In the cloud world, vector data can be tiled and subsequently be reconstructed into a specific DataCube. Apache Parquet provides vector data with partitions that flexibly define space and time extent and create cloud objects that allow HTTP range requests. This research took two huge datasets as input, which are GEDI Level 2 version 2 A and B (~200 TB) and ICESat-2 ATL08 version 6 (~20TB). The data are stored in time-series sequence, from 2018-04-18 to 2023-03-26 and 2018-10-14 to 2023-06-21 respectively. Our approach took space and time partitions, given 1x1-degree tile and monthly interval. It reconstructs the data into a Vector DataCube that has a minimal unit of 1x1 degree and a single month (e.g. 044E10N-201910). Furthemore, due to the support of HTTP range requests in Parquet, users of these datasets are exempted from reading the whole dataset and subsetting the small portion. A simple metadata search has been developed, supporting direct access to data subsets without the computational overhead of opening and navigating through a massive data object simply to retrieve a small subset of bytes within it. This Vector DataCube is crucial for data providers to handle large datasets on the cloud and then offer serverless computing. The structure of Vector DataCube allows users to navigate to their area of interest, cache the metadata and extract the data for subsequent analysis or modeling. The Vector DataCube created by partitioning offers an explicit manner to organize data that developers can further build upon to achieve higher level work, especially to support global mapping projects in the field of terrain modeling and canopy height mapping. The data of our work will be available as open data (CC-BY license) and analysis-ready and cloud-optimized (ARCO), and the presentation will also address the preliminary test of mapping capabilities of short vegetation height and production of the Ensemble Digital Terrain Model (EDTM).
Ho et al. (Sat,) studied this question.