A METHODOLOGICAL APPROACH FOR THE AGRICULTURAL STATISTICS BY REMOTE SENSING.

摘要

The development of the agricultural activity requires a system of periodic evaluation, reliable and efficient, constituting a coherent instrument of observation on agriculture. The study should lead to the evaluation of the land and to the outputs of main cultivations. Facing this objective and regarding the failure of the classic methodologies of statistical collection, a new technology based on geographic studies, which is remote sensing, was used. The study was carried out on the region of Oum El Bouaghi (Algeria), the area of Ain Kercha has also been choseen, to carry out this research. Cereal cultivations have been studied. They represent the biggest part in the region. The development of a methodology of collection on the agricultural statistics let us work on different spatial scales locating, in the first time, the agricultural zones and in the second the cultivated zones. The investigation was done on high resolution SPOT data. A digital image processing methodology was developped in order to discriminate the different cultivated species. The stratification of the region into homogenous zones let us improve the results precision. On the zone of study, the setting up of a survey plan was done and the different parameters of its application were discussed. In order to improve slightly the estimations, the satellite data processed were compared to those issued from investigation by means of a regression estimator. A number of methods were tested, they are based on the combination of satellite data and terrain informations. The results obtained told us that there were over estimations (5,88 % for irrigated areas) and under estimations (0,86 % for cereal cultivation class with low efficiency). In fact, the approach developped gave us a result for which the error margin over all the sampled segments about 2,53 %.