Climatic data reconstitution of North Algeria: Application of neural network method

In the present context of climate change and preservation of biodiversity, the appreciation of the vulnerability of the natural ecosystems and their capacity of adaptation appears among the main preoccupations to the world level (GIEC, 2007). This assessment of the ecosystems requires the availability of climatic data, what is often made difficult by the weak density or even the absence of meteorological stations notably, to the level of the mountains zones. In order to study the climate vegetation relationship in North Algeria, we use an automatic interpolation method, the neural network method, for the reconstitution of climatic data of the sampled sites, (1035 phytoecological samples), from the existing meteorological network (269 stations). This method is characterized by a great suppleness of non-linearity and by its capacity for reconstituting information from partial and not well-defined indications such as the case of data provided from meteorological networks. In order to reconstitution of climatic data, we use the explicate variables, longitude, latitude and altitude, the variables to explain being the rainfall and temperatures. To define the best approach, the network calibration has been activated on climatic parameters taken globally or solely, for the whole of study zone, and by geographical sector. The results of the interpolation are expressed through a climatic parameter cartography, released automatically by the MapInfo software. The reliability results obtained by this method can be appreciated by elaboration of errors maps comparing to reference data. (C) 2010 Academie des sciences. Published by Elsevier Masson SAS. All rights reserved.