GEO-IMAGERY MANAGEMENT AND STATISTICAL PROCESSING
IN A REGIONAL CONTEXT USING OPEN DATA CUBE
U.Otamendi, I.Azpiroz, M.Quartulli,I.Olaizola
Vicomtech Foundation
Basque Research and Technology Alliance (BRTA)
Donostia-San Sebasti´
an, 20009, Spain
F.J.Perez, D.Alda, X.Garitano
HAZI Foundation
Granja Modelo
Arkaute, 01192, Spain
ABSTRACT
We propose a methodology to manage and process remote
sensing and geo-imagery data for non-expert users. The pro-
posed system provides automated data ingestion and manip-
ulation capability for analytical data-driven purposes. In this
paper, we describe the technological basis of the proposed
method in addition to describing the tool architecture, the in-
herent data flow, and its operation in a specific use case to
provide statistical summaries of Sentinel-2 regions of inter-
est corresponding to the cultivation polygonal areas located
in the Basque Country (ES).
Index Terms—Open Data Cube, Sentinel, data exploita-
tion, management, storage
1. INTRODUCTION
The use of high spatial resolution optical imagery for land use
and land cover change mapping [1, 2, 3] has generated a high
demand for efficient geo-imagery storage, processing, and
management infrastructures [4]. In response to that request,
The Committee on Earth Observation Satellites (CEOS) has
founded the Open Data Cube (ODC) initiative [5], publish-
ing a free and open geospatial exploitation tool [6]. Large
Spatio-temporal Earth Observation (EO) data volumes are
rapidly processed through metadata indexing and ingestion
procedures, providing an efficient tool to query remote sens-
ing data. The main drawback of implementing an ODC-based
environment for geo-imagery management and analytics is
that it requires an investment in hardware, as well as an
initial effort on configuring the system with the considered
metadata and product descriptions. Several contributions [6]
address data governance issues in terms of the use of cloud
environment tools such as Google Earth Engine (GEE) [7].
GEE provides a cloud environment, where the analysis of
georeferenced data (Earth observation satellites, weather, and
climate data) is possible with limited data management ef-
fort. This has resulted in an efficient and widely used tool
for tasks that range from querying and synchronizing cli-
mate reanalysis datasets to the exploitation of georeferenced
measurements for e.g. land use analysis [8] or to improve
existent plant phenology models [9]. A possible alternative
to GEE lies in initiatives such as ODC, which allow local
institutions to undertake geo-imagery data management and
analysis directly.
The ability to rapidly generate local statistics for a time
series of remote sensing images represents a valuable asset
for geo-imagery exploitation for rapid mapping. The pos-
sibility to routinely perform this generation in a completely
automated manner and in terms that are familiar to a domain
expert such as a forester or an agronomist represents an added
value point. In particular, the capability to translate the mea-
surements available in remote sensing image products in the
format of a multi-resolution tile pyramid allows the informa-
tion to be queried by simply specifying a spatial region and
a temporal interval of interest and a set of collections to con-
sider as sources. The possibility for the domain expert analyst
to deal with pre-processed, pre-organized, and pre-tiled infor-
mation content, allows them to focus on the intended applica-
tion without having to cope with specificities of the original
data that stem from the way those were collected.
In this sense, the main goal of the current contribution is
to describe and introduce a geo-imagery data management,
processing, and exploitation service. This service integrates a
methodology intended to provide statistical summaries of re-
gions of interest corresponding to geo-polygonal data. The
service is oriented to geospatial data analysts with limited
knowledge of remote sensing technology.
In what follows, we present in detail the problem to solve,
focusing on the processing of geo-polygons in ODC for the
extraction of geo-spatial statistics. We describe the imple-
mented architecture, detailing its main components. We pro-
ceed by presenting the imagery characteristics for static and
time-dependent products. In addition, we have verified the
performance of the service using Sentinel-2 imagery to ana-
lyze an approximated quantity of 20000 areas of interest lo-
cated in the Basque Country (ES).
arXiv:2210.01470v1 [cs.CV] 4 Oct 2022