Content uploaded by Simon Scheper
Author content
All content in this area was uploaded by Simon Scheper on Feb 19, 2019
Content may be subject to copyright.
Living planet symposium 2019, Milan 13-17 May 2019
Derivation of near-surface soil parameters from
Sentinel-2 data in combination to field survey
Schmidt, Simon1; Frei, Michaela1; Iwan, Lukas1; Waldmann, Frank2
1Federal Institute for Geosciences and Natural Resources BGR, Stilleweg 2,
30655 Hannover, Germany
2State Office for Geology, Resources and Mining, Albertstraße 5, 79104
Freiburg, Germany
So far the most state-of-the-art soil map (soil map 1:50’000) of the German
federal state Baden-Wuerttemberg is based on conventional data sources as
soil surveys, forest mapping, geological and topographical maps, and digital
elevation models. However, the integration of remote sensing products as a
complementary data source has attained little attention over the years due to
various reasons (e.g., high costs for commercial products, low spatial and/ or
temporal resolution). Nowadays with Sentinel-2 data, obtained within the
Copernicus Mission, new multispectral images of the earth at a relatively high
spatial resolution (up to 10 m) and a repetition rate of 5 days are made available
free of charge. As such, Sentinel-2 data can be used as an additional database
that delivers continuous information about the soil surface since 2015.
The aim is to elaborate the relationships of near-surface soil parameter (NSSP)
and their spectral characteristics by fusing the multispectral Sentinel-2 data with
analyzed soil samples (n=164) of arable lands in Baden-Wuerttemberg and
hyperspectral field and laboratory spectrometer measurements. The soil
sampling network is relatively dense with a mean distribution of 5.7 km x 5.7 km
on arable land. Such NSSP are the contents of sand, silt, clay, organic carbon,
and the presence of carbonates.
A correlation of measured NSSP and spectral characteristics in the spectrum of
Sentinel-2 is to be expected. Based on the identified relationships of NSSP and
spectral signatures, a proposed regression equation should be applicable to the
total arable land within Sentinel-2 scenes to model the spatial distribution of
NSSP. Mixed pixels first need to be unmixed to provide a clearer and
unambiguous spectral signal. Additionally, the high repetition rate of Sentinel-2
enables to capture the influence of temporal dynamics as soil moisture on the
reflectance and spectral signatures on NSSP. That information facilitates the
interpretation of spectra regarding the overruling effect of water saturated NSSP
that reduce the actual soil parameter signal.
Other soil data like the Land Use/Cover Area frame statistical Survey (LUCAS)
can serve as an independent dataset to validate the derived NSSP of
Sentinel-2. The objective is to integrate the obtained additional information
about NSSP into the existing soil map of Baden-Wuerttemberg.
The results may demonstrate that the combination of traditional field surveys
with remote sensing data can lead to a better understanding of soil properties
and their interactions. Furthermore, the future integration of remote sensing
data in soil maps enables to keep soil maps up-to-date and integrate additional
information cost-effectively and time-saving.
Living planet symposium 2019, Milan 13-17 May 2019
---------
Planned submission to:
A5.05 Earth Observation for Soils
Regular Contributed
Description: Soils are critical for the human existence as they have essential
ecosystems functions. Soils and litter represent the largest terrestrial carbon
pool of the Earth, acts as a climate regulator, influences biodiversity and is the
driver for biomass production and thus secures humans food production. Earth
observation provides information about the status and evolution of soils. With
the accessibility of local to global-scale high spatial resolution Earth observation
data new opportunities and concepts have been developed that enable the
retrieval of soil status and development of soil compositional distribution maps
at variable spatial scale. Additionally, new spaceborne plattforms such as
imaging spectrometer will be launched in the near future that can further
enhance the derivation of soil properties. The goal of this session is to bring
together experts working on new concepts, algorithms and products using EO
data for soil sciences. Methodical papers as well as applied disciplines are
welcome that present the use of EO data for several disciplines related to soil
sciences such as soil survey, soil spectral libraries, digital soil mapping,
precision agriculture and environmental monitoring, soil mapping and
monitoring, large-scale soil maps, etc.
Convenors: Uta Heiden (DLR)