RIVER WATER LEVEL PREDICTION USING AMSR-E AND SMOS DATA
C.Vittucci (1), L.Guerriero (1), P. Ferrazzoli(1), R. Rahmoune(1), V. Barraza (2), F.Grings (2)
(1) Tor Vergata University, DICII, Roma, Italy, email:vittucci@disp.uniroma2.it
(2) Instituto de Astronomía y Física del Espacio (IAFE), Buenos Aires, Argentina
ABSTRACT
The sensitivity of the microwave brightness temperature (TB ) and emissivity (e) to rainfall and flooding has been
already investigated by several scientists. It has been demonstrated that passive microwave sensors are able to detect
soil conditions to some centimeter depth, depending on the instrument wavelength and that the combination of different
frequencies, polarizations and incidence angles can significantly improve the information content. In this research, the
effectiveness of the satellite data in the study of hydrological processes will be pointed out exploiting AMSR-E data and
presenting also some recently available SMOS L band data.
The river basin selected for this study is Bermejo, in northern Argentina. This river is seasonally affected by severe
flooding events in the lower part, mostly due to rains occurring in the upper basin, that produce sediment loadings (8
kg/m3) flushing down along the lower basin and changing the watercourse. A dataset collection of ground and satellite
data, covering the 2010 and 2011 time frame, has been analyzed for the lower Bermejo Basin where seasonal flooding
events often occur during the rainy season, due to heavy rains. Depending on the year, floods can occur on a short time
of 3-4 days, to several days. This means that a reliable system able to simultaneously check the entire basin hydrologic
balance is requested, although using a moderate resolution. Passive microwave sensors acquire data at high temporal
frequency (1-2 times daily) with near-global coverage. These characteristics represent the major attraction of this type
of data in the monitoring of hydrological cycle. The lower resolution can be considered as the principal weakness of this
kind of instruments, however, this problem becomes manageable when the considered phenomena are related to soil
saturation conditions over large areas.
This study confirms, first, the capability of passive remote sensing instruments to record brightness temperature
variations due to rainfall and floods occurred near river edges under different seasonal conditions. Then it supports the
effectiveness of microwave radiometers at Ka, X, C bands (provided by AQUA AMSR-E) as a monitoring tool, and
highlights the better sensitivity of L band data (made recently available thanks to SMOS-MIRAS) over moderately and
densely vegetated areas. The analysis of the brightness temperature and emissivity trends reports a correlation with
rainfall and flooding events, since these phenomena generate saturation conditions in the soil on extended areas.
The obtained results suggested the idea to develop an adaptive model for river water level forecasting, using TB (or e)
variations together with rainfall rates. The model assumes a linear correlation between water level and input values, but
it is able to perform a dynamic adjustment of the weight parameters as soon as new measurements are available as time
flows. The study presented here shows that the combined use of satellite data and traditional ground‐based rainfall and
hydrometric observations, acquired in the same period, offers advantages in the surveillance of flooding phenomena
compared with conventional monitoring techniques.