1. PROBLEMATIC
Agriculture must be considered like having prioritv among the
developing politics. Or, how can one formulate a plan of
restoration, a politic of development without knowing the
productions and needs of a region? it is necessary to establish a
"diagnosis" before establishing a plan. More than evervwhere
elsewhere, the inventories and the statistics are therefore basis
tools for our politicians.
The programs of agricultural development require some
agricultural reliable statistics and updating regularly. However
these measurements must be in adequacy with the used tool.
which is Remote Sensing. Indeed, the repetitivity of image taking.
the spatial resolution and the spectral discrimination of some
images permit the acquirement of the spatial and temporal
information on some cultivated surfaces in appropriate time.
Currently some statistics are published. but the methods used for
collection do not allow to value the reliability of these data. The
Algerian statistical system, especiallv in matter of agricultural
and socioeconomic data, suffers from the following problems:
+ the bad quality of the data:
— long delavs which separate their collection and their final
disposition;
— their elevated costs with regard to their reduced utility
2. OBJECTIVES OF THE STUDY
The objective of this survey 1s to set up a methodology of
agricultural reliable inventory the results of which are provided in
appropriatc timc.
Indeed. the development of a methodology. which is based on the
combination of ground data and remote sensing. for better
statistic and cartography of agricultural space. constitutes a lone
term objective of a capable system
— distinguish. identifv. and. measure the more quickly. possible
and with the maximum of precision. the agricultural surfaces:
-— cvaluate the achieved agricultural productions. this task needs
the evaluation of the surfaces and measures of outputs:
= achieve some classes of harvest forecasting. which constitutes
the most delicate case, since it requires a precocious evaluation of
some surfaces and a forecast of the outputs. In the setung of this
survey. it is the first two points that arc to be developped The
methodology developed in the setting of this survey is based on
the following points:
I) the determination of a strategy of sampling of the soil using
arca investigations on the basis of ecomcetric segments
2) clarification of techniques of preprocessing and processing of
images (IHS” . CC”” . geomctric corrections. classifications.
vegetation indices, ctc.) For some classifications and some
vegetation indices. the combination of a certain number of
processings 1s imposed. Indeed. it should first mask all which is
«not agricultural» in order to avoid ambiguities between classes.
3) constitution of a data bank of the region of survey and setup of
a data base on the segments tested
This survey should lead to an evaluation of the surfaces and some
outputs of the main cultures
Intensité Hue and Saturation
** A >
Color Composite
3. PRESENTATION OF THE ZONE OF STUDY
We worked on a zone around the city of Ain Kercha, which is
located in the Wilaya of Oum El Bouaghi. The zone of survey is a
plain zone the mean altitude is of 900 meters surrounded of
mountains. Indeed, the plain extends over 80% of the surface of
the Wilaya. The Geo-ceological conditions of territory is
favorable on the whole to agriculture.
The results of the climatic survey allowed us to classify this
climate as an attenuated climate subdescrtic, presenting a humid
scason in winter.
We will limit ourselves to the cercals statistics on the region of
survey because they represent the majority theme to the level of
the Wilava. the intended carths for cercals represent 56,1494
| Benhamouda and Al. 94] of the surface of the Wilava.
The knowledge of the agricultural calendar of wheat, allows us to
make an optimal choice on key dates of images acquisition. The
analysis of different stages and some step mark of the life of
wheat allowed us to identify the phase or the chlorophyll activity
and the density of recovery arc the most important. We notice
that this phase it spreads from March 15 to April 15, it is for this
rcason that the images acquired on this period arc supposed to
give a maximum information concerning the cereals.
4. METHODOLOGICAL APPROACH ADOPTED
The approach developped in the course of this study is based on
the information from satellite data (SPOT) as well as on the
information collected on the ground. Indeed, a double approach
was conducted on the region of survey from some satellite data.
In a first ume thc space meshing in big categories of land cover
was achieved. In order to opumize this straufication and to make
it adaptable to the objectives fixed while taking. into. account
some limits of remote sensing. we decided upon the stratification
«Corine Land Cover» which was modified slightly for the
specificities of our country. The stratification realised allowed us
to get to the following nomenclaturc:
Nomenclature kept Stratification «corine land cover»
1) Cereals GY" = Arablc land.
2) Cereals W. Y = Arabic land.
3) Fallow land = Arable land.
4) Market garde = Cultural systems.
5) Course on bare soil = Natural lawns and grazings
6) Bare soil =
7) Forest =
8) Maquis =
9) Agglomeration =
Scattered vegetation.
Forests of resinous.
Vegetation (sclérophwll).
Urban continuous tissu.
The methodological approach is based on the photo
interpretation of images. for the stratification of the region of
survey in big homogeneous units, and on some technics of
automatic classification. either by pixel (classification by
maximum Likelihood or classification bv minimum of distance)
bv parce! (Fig.l). This approach refers to some others data as
agricultural statistics. climatic data. ctc.
Cercals good viclding
Cereals weak vielding
294 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII. Part 7, Budapest, 1998
Auxill
(Agriculti
a
The clas
which v
criterias
localizat
allowed
In order
donc acc
* à SO-
Likeliho
= an ap
For the |
on onc I
and on t
algorithr
Likeliho
Groun
Data
berm —
C
The rad
determin
previous
rcjected :
avoid the
