Conference Agenda

B33: EO Day
Wednesday, 08/July/2020:
11:00am - 11:45am

Session Chair: Dirk Tiede
Location: B - Sessions


Modeling observations and objects for spatio-temporal Earth observation data

Martin Sudmanns

University of Salzburg, Österreich

Modeling observations and objects for spatio-temporal Earth observation data

Using spatial concepts to integrate data and information from various sources for a knowledge-based assessment of impervious surfaces

Thomas Strasser, Dirk Tiede

University of Salzburg, Österreich

In this study, we present a concept for the assessment of impervious surfaces integrating VHR satellite data and a priori information from additional datasets. Spatial concepts like neighbourhood and region, distance, spatial dependence or spatial variability are adapted in a knowledge-based approach using an object-based image analysis model to accumulate evidence from different sources. We look at constraints for timely and comprehensive VHR optical data acquisition that covers larger areas with adequate image characteristics (sensor family, seasonality, sensor viewing angles and sun inclination). For a study area covering the municipality of Hallein (Austria), we discuss preliminary results with a focus on real-world object characterisation (including surface material, spectral reflectivity, object size and shape) and on building a knowledge-base for the classification of real-world objects. We also assess image characteristics and effects on image analysis. The knowledge on real-world object characteristics and image object statistics will be used to develop an integrated approach that aims for transferability to larger areas.

OBIA4RTM: Object-based plant parameter retrieval using radiative transfer modelling of vegetation

Lukas Graf

University of Salzburg, Department of Geoinformatics Z_GIS, Austria

Knowledge about the physiological development of plants is of outstanding importance for precision farming applications. The leaf area index (LAI) is an important plant physiological parameter that can be derived from optical satellite data using radiative transfer models (RTMs). Previously used approaches to derive LAI are mostly pixel-based and do not use spatial information. For this reason the OBIA4RTM tool was recently introduced, which combines RTMs with object-based image analysis (OBIA).

This work applied the ProSAIL RTM to Sentinel-2 data to derive the LAI for a test area in Southern Germany (Munich-Isar-North) in 2017 and an entire catchment in Upper Austria (Gurtenbach catchment) in 2018 for three different crop types: corn, winter triticale and winter wheat.

For the Munich-Isar-North test site, in-situ data were available for the year 2017, which provide information on the validity of the object-based LAI derivation: When compared to in-situ measurements of LAI in winter wheat (one field), winter triticale and maize fields (two fields, each) over the growing period in 2017, deviations between Sentinel-2 and in-situ measured object-based LAI values were observed with relative errors of 67% (N=11) in case of maize and 36% (N=22) in case of the winter cereals.

The OBIA4RTM approach therefore offers promising prospects, but also highlights the current challenges including the choice of the underlying spatial objects and the object-based validation of the results.