Should EU land use and land cover data be managed with a NOSQL document store?

Abstract

Land cover (LC) is a scientific landscape classification based on physical properties of earth materials. This information is usually retrieved through remote sensing techniques (e.g. forest cover, urban, clay content, among others). In contrast, Land use (LU) is defined from an anthropocentric point of view. It describes how a specific area is used (e.g. it is usual to indicate whether a territory supports an intensive, extensive use or it is unused). Both geospatial layers are essential inputs in many socio-economic and environmental studies. The INSPIRE directive provides technical data specifications for harmonization and sharing of voluminous LU/ LC datasets across all countries of the EU. The INSPIRE initiative proposes Object-Oriented Modelling as a data modelling methodology. However, the most used Geographic Information Systems (GIS) are built upon relational databases. This may jeopardize LU/LC data usability, since GIS practitioners will eventually face the object-relational impedance mismatch. In this paper, the authors introduce the SIOSE database (Spanish Land Cover and Land Use Information System), which was the first implementation of an object-oriented land cover and Land-use datamodel, in line with the recommendation of the INSPIRE Directive, separating both themes. SIOSE data can be downloaded as relational database files, where information describing each single LU/LC object is divided among several related tables, so database queries can be complex and time consuming. The authors show these technical complexities through a computational experience, comparing SQL and NoSQL databases for querying spatial data downloaded from SIOSE. Finally, the authors conclude that NoSQL geodatabases deserve to be further explored because they could scale for LU/LC data, both horizontally and vertically, better than relational geodatabases, improving usability and making the most of the EU harmonization efforts. Keywords: Land Use, Land Cover, document store, SIOSE, geodatabase, PostgreSQL, jsonb LAND USE AND LAND COVER DATABASES IN THE EULand is a limited resource and its mismanagement is one of the main drivers of global change, with significant effects on ecosystem functions, goods and services [1]. There are complex environmental problems such as the over-exploitation of natural resources, biodiversity loss or climate change that require a long-term management perspective of natural resources. Many studies agree that these problems can be aggravated by land-use changes, so it is mandatory to monitor and apply long-term management policies at different scales [2].Land use (LU) and land cover (LC) information at national and regional level has been historically recorded in many EU Member States because of their environmental and territorial management's needs and requirements. In addition to Corine Land Cover (CLC) 1990 databases, many EU countries have been producing LC databases to manage and satisfy their requirements on environmental, agricultural, forest and land planning issues. As a conse...

Efficient storage of heterogeneous geospatial data in spatial databases

“…Here, the table-based layout is able to insert the test-data more than twice as fast as the jsonb-layout. These findings contradicts similar studies found in literature [18,23], which report that NoSQL document stores or data types outperform relational layouts.…”

“…Another important aspect of a database used for managing open geospatial data is usability. Navarro-Carrión et al [18] noted that the query syntax used for the PostgreSQL JSON data type is "somewhat convoluted", an assessment we find to hold true (see Fig. 1 for an example).…”

“…A preliminary study by Navarro-Carrión et al [18] examined the feasibility of using a NoSQL document-store to store EU land cover and land use data. In their experimental set-up, they used two PostgreSQL/PostGIS instances.…”

Leveraging Container Technologies in a GIScience Project: A Perspective from Open Reproducible Research

“…As mentioned above, the SIOSE database is designed according to an object-oriented model, but due to practical reasons, it is adopted and implemented within an object-relational database, which is a clear example of the object-relational impedance mismatch [23]. The usability problem investigated by the SIOSE-INNOVA project has to deal with the fact that regular users of this type of information layers usually work with ESRI Shapefiles and these are generally not used to work with such information in any of the original distributions of SIOSE (corporate databases).…”

“…A sophisticated computational experience using SIOSE data was described in detail in [23]. In this case, developed in 2016, an experiment was carried out in which it was verified that specific queries to the SIOSE database could be expedited if the data were stored in a document-oriented database (document store).…”