With the exponentially growing earth observation data of specific sensor-determined resolutions and update frequencies, earth observation has irreversibly arrived in the Big Data era, enabling new insights in science and engineering. With great opportunity comes great challenges regarding efficient and effective data management because earth observation data is of different scales and characterized by complexity in spatial relationships related to the real world. To overcome the challenges is crucial for, for instance, data mining, land surveying, and especially emergency mapping for disaster response. To improve the querying efficiency of big earth observation data, we proposed a novel data management approach: Global Multi-scale Grid Integer Coding and Spatial Indexing. Among our contributions are: (1) proposing Global Multi-scale Grid Integer Coding Model (GMGICM), which presents clustering property in both the scale dimension and spatial dimension, and theoretically facilitates an efficient querying; (2) deliberately applying GMGICM on multi-scale earth observation data spatial indexing, which results in one-dimensional data index, which can be queried using simple B-tree, inversion, and other one-dimensional indexes; (3) designing a strategy to assure the completeness of spatial querying, which is not well solved by existing grid-based coding models. The advantages of our proposed approach have been demonstrated with both simulated and real remote sensing data, with spatial operation 20 times as fast as Geohash and spatial querying 10 times as fast as Oracle Spatial on average. The proposed approach can be easily adapted for three or higher-dimensional earth observation data and bring potential benefit to all big earth observation data analytic projects.

随着特定传感器确定的分辨率和更新频率的地球观测数据呈指数级增长，地球观测已不可逆转地进入了大数据时代，从而在科学和工程学方面有了新的见解。由于地球观测数据的规模不同并且具有与现实世界相关的空间关系的复杂性，因此在有效管理数据方面面临着巨大的挑战。克服这些挑战对于例如数据挖掘，土地测量，尤其是灾难响应的紧急制图至关重要。为了提高大地球观测数据的查询效率，我们提出了一种新颖的数据管理方法：全局多尺度网格整数编码和空间索引。我们的贡献包括：（1）提出了全局多尺度网格整数编码模型（GMGICM），该模型在尺度维度和空间维度上都表现出聚类性质，并且在理论上促进了有效的查询；（2）故意将GMGICM应用于多尺度地球观测数据空间索引，产生一维数据索引，可以使用简单的B树，反演等一维索引进行查询；（3）设计一种确保空间查询完整性的策略，而现有的基于网格的编码模型并不能很好地解决这一问题。我们的方法的优势已在模拟和真实的遥感数据中得到了证明，其空间操作速度平均是Geohash的20倍，而空间查询的速度平均是Oracle Spatial的10倍。