The joint statistical distribution of principal meteorological variables (temperature of air and soil, humidity of air and soil,precipitation, pressure, shortwave radiation, cloudiness) are investigated. Ten years data sets of one hour temporal resolution for several meteorological stations of Russian North-West region are used. The data set of simultaneous observations ( for all variables and stations ) let us to determine main features of joint diurnal distributions by means of the fuzzy set constructions. Known and novel interrelationships between various meteorological and connected soil variables are reviewed. Implementation of neural networks (back propagation algorithm) let us to perform several modelling experiments. Two stationary and two transitionary diurnal modes were introduced to simulate simplest durnal cycles. They describe the conservation of warm and cold weather patterns and possible transitions from one to another. Fuzzy logic model was costructed for weather pattern recognition, based on several meteorological variable observational data of one hour resolution. Determination of neural network hidden layers and nodes are investigated. Problem of optimum neural network achitecture ( number of nodes in hidden layers ) is discussed. For example, it is possible to reconstruct the diurnal cycles of some meteorological variables with satisfactory ( at the observational error levels ) precisions for each of them,by using 1, 2 or 3 instant meteorological observations. Adapted nonlinear mapping of one group variables (and its diurnal distributions) on another one let us to investigate the possibilities of this approach for diagnostical aims.
Potential application ranges of the developed approach :
- downscaling in global models;
- spatial field reconstructions in case of missing observational data ( for some variable(s) );
- short-term weather forecasting;
- parametrization of energy exchange processes at the surface between Earth surface and atmosphere;
- micrometeorology and microclimatology .