Analysis of optical proprieties of the water reservoir Rodolfo Costa e Silva – Itaara, RS, Brazil, with field spectral data and orbital multispectral images

  • Fábio Marcelo Breunig SERE - INPE
  • Flávio Wachholz UFSM
  • Waterloo Pereira Filho UFSM
  • Conrado de Moraes Rudorff INPE


An evaluation of the discrimination of water classes using continuum removal technique applied over spectral data obtained in field and multispectral images classification is presented. The study area was the Rodolfo Costa e Silva water reservoir, located in central region of Rio Grande do Sul (RS) State, in Southern region of Brazil. The methodology was based on in situ data collection of: total suspended solids, chlorophyll (a, b and c), water transparency, and bidirectional spectral reflectance. These data were collected in 21 point (samples) in May 16, 2006. The continuum removal technique was applied on the spectral data over 4 absorption bands: 400-550nm, 610-640nm, 650-680nm e 580-700nm. The continuum removal parameters analyzed for each absorption band were: depth, area and width. The multispectral images used were CBERS-2/CCD and Landsat 5/TM. The images were acquired in a date nearest to field work and with appropriate weather conditions. These images were corrected by removing atmospheric effects and then classified. According to the results obtained from the continuum removal technique, it was verified that band depth, area and width did not present a good potential to separate different water classes. Digital classification results did not show significant correlations with the limnological parameters collected in field and, therefore, could not be used to characterize spectrally different water classes or compartments. The main problem of establishing relationships between spectral reflectance and water quality parameters was due to the low variability of optical components in the water of Rodolfo Costa e Silva Reservoir. In this case the spectral analyses (considering both techniques) were not sensitive to the relative small variations observed in field data.