Metadata of Spectral Data Collections

Metadata is important for the interpretation of scientific data, quality assessment and long term usability of data sets. The sharing of spectral data collections among research groups is uncommon and one of the reasons for this is the missing standardisation of the sampling process. Appropriate metadata serves the purpose of detailing the sampling procedure and the surrounding conditions during data capture, thus providing necessary information for data sharing. Reliable data retrieval requires the organised storage of spectral and metadata. To this means RSL developed the SPECCHIO system which is based on a relational database and provides data input, query and output mechanisms that strive to minimize the manual data capture. SPECCHIO serves as a non-redundant repository and source for spectral signatures which can be retrieved by metadata queries. The system will be used in the level 2/3 processing of the APEX (Airborne Prism Experiment) product generation to support the classification of natural and manmade materials and landcovers. Posted at the Zurich Open Repository and Archive, University of Zurich ZORA URL: https://doi.org/10.5167/uzh-77970 Published Version Originally published at: Hueni, Andreas; Nieke, Jens; Schopfer, Jürg; Kneubühler, Mathias; Itten, Klaus I (2007). Metadata of spectral data collections. In: 5th Workshop on Imaging Spectroscopy, Bruges (B), 23 April 2007 25 April 2007. Proceedings 5 EARSeL Workshop on Imaging Spectroscopy. Bruges, Belgium, April 23-25 2007 1 METADATA OF SPECTRAL DATA COLLECTIONS Andreas Hüni, Jens Nieke, Jürg Schopfer, Mathias Kneubühler and Klaus I. Itten 1. University of Zürich, Department of Geography, RSL, Zürich, Switzerland; [email protected] ABSTRACT Metadata is important for the interpretation of scientific data, quality assessment and long term usability of data sets. The sharing of spectral data collections among research groups is uncommon and one of the reasons for this is the missing standardisation of the sampling process. Appropriate metadata serves the purpose of detailing the sampling procedure and the surrounding conditions during data capture, thus providing necessary information for data sharing. Reliable data retrieval requires the organised storage of spectral and metadata. To this means RSL developed the SPECCHIO system which is based on a relational database and provides data input, query and output mechanisms that strive to minimize the manual data capture. SPECCHIO serves as a nonredundant repository and source for spectral signatures which can be retrieved by metadata queries. The system will be used in the level 2/3 processing of the APEX (Airborne Prism Experiment) product generation to support the classification of natural and manmade materials and landcovers.Metadata is important for the interpretation of scientific data, quality assessment and long term usability of data sets. The sharing of spectral data collections among research groups is uncommon and one of the reasons for this is the missing standardisation of the sampling process. Appropriate metadata serves the purpose of detailing the sampling procedure and the surrounding conditions during data capture, thus providing necessary information for data sharing. Reliable data retrieval requires the organised storage of spectral and metadata. To this means RSL developed the SPECCHIO system which is based on a relational database and provides data input, query and output mechanisms that strive to minimize the manual data capture. SPECCHIO serves as a nonredundant repository and source for spectral signatures which can be retrieved by metadata queries. The system will be used in the level 2/3 processing of the APEX (Airborne Prism Experiment) product generation to support the classification of natural and manmade materials and landcovers. INTRODUCTION Ground based hyperspectral signatures are collected for (a) calibration and validation of remote sensing imagery and its data products, (b) feasibility studies for airborne/spaceborne missions, (c) basic investigation of the relationship between physical or biochemical properties and the electromagnetic reflectance of objects and (d) definition of directional dependence of the reflectance of objects on the illumination and viewing geometry. Since the advent of field spectroscopy with the first specifically built portable field instrument appearing in the late 1980’s, e.g. PIDAS (34), a lot of research on the spectral properties in the VIS/NIR electromagnetic spectrum of natural and manmade objects has been carried out. At the same time considerably less effort has been spent on the issue of standardisation of the measurement process itself and the systematic collection and interpretation of ancillary data, the so called metadata. The comparison of spectral signatures between studies is complicated by the many different techniques for the capturing of spectral field data (19). Utilizing data from other studies requires an assessment of the data quality and suitability of the data set for the given task. Milton et al (20) state that accuracy depends on a clear definition of what is being measured and on the conditions under which is being measured, i.e. the description of the sampling experiment and of the sampling environment is of importance if the data quality is to be assessed. The factors that influence the spectral measurements taken in the field are detailed in (19), i.e. for a traceability of the measurement process these factors should be recorded and stored as metadata. Metadata support broad and long-term use and interpretation of scientific data (18). The lack of metadata can render previously collected data useless for new applications (6). Spectral libraries are data collections that provide reference spectra for a number of procedures in remote sensing, e.g. spectral unmixing based on endmember spectra, landcover classification or atmospheric correction by the empirical line method (27). A number of public spectral libraries exist, e.g. the USGS spectral library (2), that contain high quality spectra of numerous targets but are mainly focused on minerals. Such libraries usually only contain first order statistical information, i.e. only one representative spectrum per target. This poses a serious restriction on the use as the variation described by second order statistics needed for e.g. classifications is not available (16). There is a need to include such information in spectral libraries to increase the matching accuracy of field spectra against library spectra (26). Furthermore, such libraries do often not account for the Proceedings 5 EARSeL Workshop on Imaging Spectroscopy. Bruges, Belgium, April 23-25 2007 2 spatiotemporal variability of objects, e.g. plant phenology or intra species variability (23). Thus there is still a need to build customised spectral libraries that account for local variables affecting the spectral reflectance of objects. A limited number of studies explicitly mention the building of a spectral library containing ground reflectances of targets which are subsequently used to derive products from imagery (11; 15; 25) while numerous studies based on field spectroradiometer data have been carried out but do not explicitly consider their spectral collections as libraries, e.g. (29; 31; 32; 33). However, most studies focus on the spectral characteristics of the targets and while the acquisition of field data is described, not much detail is given about the organisation and storage of these data and associated metadata in most cases. Given the scenario outlined above, an organised and non redundant storage of spectral data and associated metadata is an important step towards better data quality, long term usability and the possibility of data sharing between researchers. A relational database with appropriate interfacing software seems a natural choice of technology in this respect. RSL has implemented the SPECCHIO system which acts as a repository for spectral field campaign and reference signatures plus metadata (1; 12). A recent redesign of the data model and user interface has been based on an analysis of the metadata space and minimizes the needed user actions during data capture while offering added value to the researcher. SPECCHIO is planned to be an integral part of the APEX (Airborne Prism Experiment) (22) level 2/3 processing chain and is foreseen to be used for pre and post classification of hyperspectral image cubes (28). In this paper we describe the general concept of metadata space, its application to field spectroradiometer metadata, the metadata set implemented in SPECCHIO, the conceptual integration of SPECCHIO into the APEX level 2/3 processing and the user interfaces of the SPECCHIO system.