Skip to content Skip to sidebar Skip to footer

Research and Collaboration Stay of Susana Lagüela López at Delft University of Technology (The Netherlands)

Dr. Susana Lagüela López developed a research stay at TU Delft from January to April 2016.

The stay consisted on the collaboration with H2020 project SIM (Smart Irrigation from soil Moisture forecast using satellite and hydro-meteo modelling) and ESA project Tiger for “Assessment of climate effect on crop water productivity using Earth observation data: case study of Doukkala-Western Morocco”.

For these projects, Susana has focused on multispectral image processing, particularly on the fusion of images from different satellites, in order to include the TIR (Thermal InfraRed) band in the bands analysed by satellites that miss it. In particular, the TIR band of the CBERS4 – IRM sensor, will be combined with the bands from SENTINEL 2, which miss the TIR band. The following steps are required, according to the Unmixing technique:

  • Analysis of the presence of cloud coverage in the images, which is performed through the analysis of their Gaussian curves. If the coverage percentage is low, the mean value of the images is subtracted in order to generate a smooth transition between images. If the cloud coverage is almost complete, the image is eliminated from the process.
  • Generation of TIR mosaic for the images covering the area under study. The mosaic is required in order to extend the coverage of the images to larger terrain areas, and facilitate the correspondence between images of different satellites, thus with different coverage. Next, overlapping between CBERS and Sentinel 2 images is computed, in order to compute the Sentinel 2 pixels inside each CBERS2 pixels.
  • Evaluation of the behaviour of the TIR band for the measurement of evapotranspiration phenomena. The results show that each soil use presents different behaviour under sun presence, with different reflectance values. As an example, in the vineyard region of Mendoza (Argentina), three different land uses are detected as a function of their reflectances: soil, uncovered rock, and vegetation (vineyards), with reflectance values of 75, 105 and 130, respectively.
  • Pixel classification is performed, using the conversion from RGB to LAB colour space, which allows the elimination of the luminosity effect (L) in order to make the process more stable. The process is performed for images of both satellites, in order to compute the values of the new pixels created for CBERS images as a function of the pixels contained in them from Sentinel images.

The work will continue to be developed during the months of October 2016 – January 2017, so more news are to come.