Programme OS4d Remote sensing, DSS and
Monitoring of surface water quality in large rivers with satellite imagery
Author(s): Application to the Amazon basin
Author(s): Jean Michel Martinez, Jean-Loup Guyot, Gérard Cochonneau
Institut de Recherche pour le Développement
Keyword(s): Monitoring, water quality,
remote sensing, sediment
Session: OS4d Remote sensing, DSS and
The use of satellite images to
monitor inland water quality has been commonly reduced to one shot studies because of the lack of appropriated
remote sensing sensors offering suitable spatial resolution and adequate revisit frequency. Satellite sensors measure
water colour that can be related to the presence of three parameters of interest : inorganic particular material
(sediment), organic particular material (alga) and organic dissolved matter. The monitoring of these elements
encompasses a wide range of application from contaminant transport (trace elements, nutrients, hydrophobic organic
compounds), water quality trends, reservoir sedimentation, lake eutrophication or soil erosion. Over inland waters
the challenge is to retrieve the parameters of interest when the water optical properties are affected at the same time
by the presence of mineral and organic matter at high concentrations.
The present decade has seen the launch
of sensors that propose new tradeoff between spatial resolution and time revisit. In this study we show that recent
sensors such as MODIS and MERIS, by joining fine spectral resolution, high temporal resolution (up to 2 images a
day) and medium spatial resolution (250 to 500 meters) may be used for the monitoring of the water quality of the
largest rivers of the world. In particular we will present and analyze the use of MODIS sensor for the monitoring of
suspended sediment concentration in rivers.
In this presentation we will briefly present the main steps of the
methodology (datamining tools, image feature extraction and generalised inversion techniques) allowing to process
remote sensing images and to produce time series of water quality parameters at different locations in the watershed.
Very high resolution images from SPOT 5 are also investigated in order to assess the potential of the next generation
sensor that will combine such fine resolution with high revisit frequency. The images are analysed over the Amazon
basin that presents a wide variety of hydrologic and geomorphologic conditions and where a large database of
measurements is available from the HYBAM measurement network maintained by the french research Institute IRD
and the local national water agencies.
We will show results on the monitoring of surface suspended sediment
fluxes of the Amazon river over the whole MODIS archive : 2000-2008. Comparisons with in situ data show an
absolute accuracy between 33 and 90 mg/l (about 45 % of relative accuracy). This accuracy is satisfying considering
the tremendous spatial scale difference between in situ measurements (250-millimeter samples) and the satellite data
(assessed from buffers containing hundreds of pixels) and the different time sampling. Impact of some geophysical
and hydrological factors of interest are analysed such as river width, water cycle variability, sediment properties.
Also, the impact of the remote sensing sensor configuration is tested as a function of spatial resolution, radiometric
accuracy and time sampling. These results show that current medium resolution sensors can be used operationnaly
for the monitoring of sediment fluxes in large rivers and that next generation sensors will expand this potential to
medium size rivers and other parameters of interest such as alga concentration.