Super typhoon Haiyan makes landfall
Super typhoon Haiyan
Subscribe here to receive special images

VEGETATION - 2000

Lake Maggiore - Italy, 3-6 April 2000


Integration of VEGETATION and HRVIR data into yield estimation approach.

André HUSSON
SCOT
Robert FAIVRE INRA

Paper (pdf file, 164 k)

The possibility of obtaining reliable and relevant information on crop evolution and crop production is a key factor for decision making and defining strategy. Originally based nearly exclusively on meteorological data and on significant gathering of ground information, forecasting systems became rapidly reliant on statistical models which allows yields to be linked in a simple way to some explanatory variables. These "preliminary" tools proved to be limited in particular as regards "exceptional" years for which trends estimates have to be corrected of cyclical effects.

As earth observation data can monitor crop conditions of an on-going campaign during all the growing season, measure parameters related to plant functioning and improve the detection of spatial features of phenomena affecting crops, various studies have been carried out in these last years to investigate the contribution of remote sensing data in these fields. They have shown that using models seems to be a necessary step in obtaining information which, such as crop yields, cannot be directly derived from space technology.

This project, which fits into the general context of the VEGETATION Preparatory Programme for the International User Community (IUC), aims at improving the methodology for integrating remote sensing data into crop models. The proposed investigation, based on the performance of the VEGETATION instrument, was focused on the following main issues:

  • a methodological approach: to define a strategy for integrating remote sensing data into crop growth models.
  • a research approach: the combination and the complementarily between VEGETATION instrument and HRVIR instrument, and the capability of retrieving individual spectral response of each main crops on the site from VEGETATION signal unmixing.
  • an operational approach: the possible strategies, in the context of an operational programme, to combine Earth observation derived information and crop models in or
    der to forecast regional crop production.

As part of the " pre-launch phase " of the preparatory programme, a methodology for processing ‘coarse’ information (on a scale of 1 km²) on vegetation cover to derive more specific information on a given crop was tested on pseudo- Vegetation data, in fact a small number of Spot/HRV images which had been downgraded to Vegetation resolution.

The purpose of the second phase (post launch phase) is to validate the methodology on an actual series of Vegetation imagery. This validation was done for the Beauce Chartraine (an area of about 1 600 km² around Chartres).

The applied methodology is based on three main stages :

  1. Deconvolution
    • application of the method of estimating components in mixed data for each date and for each channel of interest (this disaggregation model requires prior knowledge of the land cover).
    • deducting a vegetation index (or a leaf area index) on the basis of the radiometric information thus obtained for each crop theme.
  2. Simulation :
    Use of the spatio-temporal evolution of this index to calibrate a crop simulation model pixel by pixel (as part of this work, the STICS model developed by INRA Avignon wi
    ll be used).
  3. Mapping of simulated yields for the studied region, and comparison with the official production statistics.

This approach should lead to forecasting of regional yields for the crop of interest (wheat) and its spatial distribution.