@Article{GuimarãesNarGalSilCam:2018:RaApFl,
author = "Guimar{\~a}es, Ulisses Silva and Narvaes, Igor da Silva and Galo,
Maria de Lourdes Bueno Trindade and Silva, Arnaldo de Queiroz da
and Camargo, Paulo de Oliveira",
affiliation = "{Centro Gestor e Operacional do Sistema de Prote{\c{c}}{\~a}o da
Amaz{\^o}nia/Centro Regional de Bel{\'e}m
(CENSIPAM/CR-Bel{\'e}m)} and {Instituto Nacional de Pesquisas
Espaciais (INPE)} and {Universidade Estadual Paulista (UNESP)} and
{Universidade Federal do Par{\'a} (UFPA)} and {Universidade
Estadual Paulista (UNESP)}",
title = "Radargrammetric approaches to the flat relief of the amazon coast
using COSMO-SkyMed and TerraSAR-X datasets",
journal = "ISPRS Journal of Photogrammetry and Remote Sensing",
year = "2018",
volume = "145",
number = "B",
pages = "284--296",
month = "Nov.",
keywords = "Synthetic aperture radar, Radargrammetry, Amazon coastal
environments, Digital surface models.",
abstract = "The Amazonian coast consists of extensive flood plains and
plateaus characterized by a high discharge of water and sediment
from the Amazon River. This wide landscape occurs under a tropical
climate with heavy rains and high cloud cover, making it
unsuitable for conventional mapping based on optical images.
Additionally, the flat relief and vegetation structure of the
Brazilian Amazon coast define an incoherent to partially coherent
behavior for the microwave signal, rendering radargrammetric
models more suitable for the three-dimensional mapping of its
surface. This study aimed to assess the digital surface models
(DSMs) provided by Cosmo-SkyMed (CSK) and TerraSAR-X (TSX)
Stripmap datasets throughout the radargrammetric models from
SARscape and Toutin. The DSMs were generated from SAR (synthetic
aperture radar) data with an acquisition geometry that addressed
the need for a compromise between the intersection angles and low
temporal decorrelation. The radargrammetric SARscape and Toutin's
models were developed from different amounts of stereo ground
control points (SGCP). The generated DSMs were evaluated
considering a set of 40 independent checkpoints (ICP) measured by
GNSS in the field, in their entirety and disaggregated by coastal
environment. The vertical accuracy was based on the estimation of
the discrepancies, bias and precision (standard deviation and root
mean square error RMSE), and the Taylor and Target diagrams were
used for a more comprehensive comparison. In the vertical accuracy
analysis using all ICPs measured in situ, the DSM obtained by the
SARscape's model from the CSK SAR data resulted in the lowest RMSE
(4.34 m) and mean discrepancy (0.05 m), but Toutin's model had the
lowest standard deviation (2.58 m) of the discrepancies. The
Taylor and Target diagrams showed fluctuations in accuracy that
alternated the DSMs generated from the two types of SAR data,
indicating that TSX produced more stable models and CSK produced
better vertical accuracy. The Amazon Coastal Plateau and Fluvial
Marine Terrace environments defined three-dimensional
representations with lower RMSEs (better than 7.8 and 8.9 m,
respectively), regardless of the type of SAR data or the
radargrammetric model used. The worst performance, which was for
the Fluvial Marine Plain, was influenced by the specific
characteristics of this coastal environment, such as the structure
of the mangrove vegetation and the shoreline. In general, the high
resolution and good ability to revisit the SAR data used, together
with the radargrammetric models, allowed for the accurate mapping
of the flat relief of the Amazon coastal environments, providing
detailed spatial information that can be acquired in severe
rainfall conditions in a region of intense morphological
dynamics.",
doi = "10.1016/j.isprsjprs.2018.09.001",
url = "http://dx.doi.org/10.1016/j.isprsjprs.2018.09.001",
issn = "0924-2716",
language = "en",
targetfile = "guimaraes_radargammetric.pdf",
urlaccessdate = "20 maio 2024"
}