Modelling the flight activity of Apis mellifera at the hive entrance

Hives equipped with electronic bee counters were installed near sunflower fields located in the agricultural district of Birieux (France). Counter data were collected during about 50 consecutive days from about mid-July to the beginning of September 2001. Temperature, relative humidity, rainfall, wind, and global radiation were also recorded. A palinological analysis was performed to help in the interpretation of the results. Principal component analysis (PCA) was first used to structure the different data matrices. Co-inertia analysis (CIA) was then employed to find co-structures between the outgoing bee data and climatic parameters. Lastly, quantitative models were derived from partial least squares (PLS) regression analysis. The obtained PLS regression results were confirmed by means of artificial neural network models. While promising results were obtained, they clearly showed the necessity to include additional variables in the modelling process. (C) 2003 Elsevier B.V. All rights reserved.