River Mapping from Satellite Radar Altimeter

River Mapping from Satellite Radar Altimeter Sigma0
(1)
Bramer, S.M.S.(1), Berry, P.A.M. (1)
EAPRS Lab, De Montfort University, The Gateway, Leicester LE1 9BH, United Kingdom ([email protected])
ABSTRACT
The ERS-1 Geodetic Mission yielded a unique, closely spaced mesh of measurements over the earth’s land surface.
This unique dataset contains echoes obtained over inland water, with very dense spatial sampling along river courses.
However, interpretation of these data is problematic, since environmental events also cause ephemeral bright targets to
appear on the earth’s surface; also the sigma0 response over rivers is extremely variable, affected by surface roughness,
off-ranging and the spatial extent of the flooded surface. Using an expert system approach to re-calculate sigma0 and
screen the echo shapes, a methodology has been designed to create detailed sigma0 maps from these data, supplemented
by ERS-2 data. Whilst this approach was initially developed to generate accurate sigma0 maps of desert regions for
calibration purposes, the techniques for removal of environmental contamination have proved so effective that it has
been possible to apply this technique over major river basins, obtaining precise maps of these river system locations.
This work therefore demonstrates the extent to which pulse limited altimeters can retrieve inland water signatures over
river networks and gives a valuable insight into the untapped potential of radar altimetry for river monitoring.
METHOD
The radar altimeter data used in this study have been reprocessed using a rule-based expert system [1] to optimise the
recovery of the range-to-surface. This works by selecting one of eleven retracking algorithms based on the waveform
characteristics. The system has been under development for several years and has been tuned for ERS-1/2, Envisat,
TOPEX and Jason-1.
After retracking, an automated method was used to remove environmental contamination from the processed sigma0
data, as part of a research programme designed to create accurate sigma0 models of selected desert areas for
calibration/validation of multi-mission altimeter sigma0 [2,3]. In all cases the lowest values at any point are chosen as
most closely representing a ‘dry earth’ value. Values from all tracks are reconciled in an iterative process, and the
models are then created by use of triangulation and bilinear interpolation.
The technique was then tested by modelling progressively wetter regions, and proved extremely robust, to the extent of
providing remarkable levels of detail even over wet terrain such as rainforests.
THE AMAZON
To investigate the performance of this technique in difficult circumstances, the entire Amazon basin was used as a test
area. The ERS-1 Geodetic Mission dataset was retracked, and the technique used for desert regions [3] was applied to
the data, to produce a model. Whilst the environmental filter has screened out some data, a surprisingly high proportion
of sigma0 data was retained, which allowed the successful calculation of a sigma0 model for the entire Amazon basin
(Fig. 1.). The path of smaller tributaries within the Amazon river network are clearly revealed in greater detail using
this technique than in the GLCC mapping of the region shown in Fig. 2. As a quantitative indicator of the internal
consistency of the model, the sigma0 values at all locations where two altimeter tracks cross were calculated both
before and after the track reconciliation and cleaning process. Crossover differences (given in Table 1.) do show that
there is still a great deal of noise in the model, with standard deviations of crossover sigma0 differences in the region of
7dB. In contrast, typical values are in the region of 1dB over desert calibration zones. This does not, however, detract
from the remarkable clarity of the mapping.
Fig. 1. Altimeter Sigma0 (dB), Amazon Region 0-10S 55-70W
Fig. 2. GLCC Mapping of Amazon Region, 0-10S 55-70W [4]
Table 1. Sigma0 Crossover Statistics for Amazon Region 0-10S 55-70W
Sigma0 Crossover Statistics
Standard Deviation of Crossover
Differences dB
Number of Crossovers
GM prior to modification
7.483
1045703
35-day crossovers with GM prior to
modification
7.893
1689232
GM after cleaning
6.447
957260
35-day after cleaning
7.272
1599084
THE CONGO
Following the successful application of this automated technique to the Amazon basin, the methodology was applied to
the Congo basin. Within the Congo region the sigma0 plot (Fig. 3.) reveals fascinating detail. The river courses are
more clearly defined than in the Google Earth plot of the region shown in Fig. 4. Orthometric heights derived from the
ERS-1 Geodetic Mission and plotted as Fig. 5. show good correspondence with the river courses revealed in the sigma0
map. Raw crossover statistics (see Table 2.) are broadly comparable with those of the Amazon region, although the
final results of the modelling process are marginally better.
Fig. 3. Altimeter Sigma0 (dB), Congo Region, 4S-4N 16-26E
Fig. 4. Google Earth View of Congo Region, 4S-4N 16-26E
Fig. 5. Altimeter Orthometric Heights (m), Congo Region, 4S-4N 16-26E
Table 2. Sigma0 Crossover Statistics for the Congo Region 4S-4N 16-26E
Sigma0 Crossover Statistics
Standard Deviation of Crossover
Differences dB
Number of Crossovers
GM prior to modification
7.465
484490
35-day crossovers with GM prior to
modification
7.947
733574
GM after cleaning
5.992
430955
35-day after cleaning
6.998
709992
LUNDA REGION
To investigate the use of this technique in a more complex environment with varying terrain, the Lunda region of Africa
was then modelled. The northerly flow of rivers in the Lunda region is clear in Fig, 6. and corresponds well with the
orthometric height model as depicted in Fig. 7. Crossover statistics in Table 3. show,, marginally less noise, prior to
adjusting, than is found in the rainforest regions, and a slightly better final result..
Fig. 6. Altimeter Sigma0 (dB), Lunda Region, 5010S 15-25E
Table 3. Sigma0 Crossover Statistics for the Lunda Region 5-10S 15-25E
Sigma0 Crossover Statistics
Standard Deviation of Crossover
Differences dB
Number of Crossovers
GM prior to modification
7.133
249074
35-day crossovers with GM prior to
modification
7.533
483697
GM after cleaning
5.639
210230
35-day after cleaning
6.842
472777
Fig. 7. Altimeter Orthometric Heights (m), Lunda Region, 5-10S 15-25E
DISCUSSION
The technique developed originally for modelling altimeter sigma0 in desert regions for calibration purposes has been
shown to be very successful at modelling wetter regions. Within these plots, river systems, which have high sigma0
values, are found to show up very clearly, and it is evident that mapping the course of river systems is possible using
this technique. The models generated show finer scale detail of the river systems than is available in global datasets
such as the GLCC database. Additionally, floodplains show up clearly. This technique will now be applied to a larger
selection of river systems, to determine the limits of applicability, and to estimate the potential contribution of this
method to the mapping of river systems.
ACKNOWLEDGMENTS:
The authors would like to thank the European Space Agency for providing the ERS data used in this study.
REFERENCES:
[1]
P.A.M. Berry, A. Jasper, H. Bracke, “Retracking ERS-1 altimeter waveforms overland for topographic height
determination: an expert system approach,” ESA Pub. SP414, Vol. 1, 403-408., 1997
[2]
S.M.S. Bramer, P.A.M. Berry, “Cross Calibration of Multi-Mission Altimeter and TRMM PR Sigma0 Over
Natural Land Targets”, Proceedings of the Symposium on 15 Years Progress in Radar Altimetry, Venice, Italy,
2006
[3]
S.M.S. Bramer, P.A.M. Berry, J.A. Freeman, B. Rommen, ” Global Analysis of Envisat Ku and S Band Sigma0
over all Surfaces”, Proceedings, ESA: ENVISAT Symposium, Montreux, Switzerland, 2007
[4]
T.R. Loveland et.al., “Development of a Global Land Cover Characteristics Database and IGBP DISCover from
1-km AVHRR Data”, International Journal of Remote Sensing, v. 21, no. 6/7, p. 1,303-1,330, 2000