Abstract：Automatic weather station(AWS) is usually configured with several meteorological observing element and gradient observation, according to each meteorological application requirement. But its output data is always saved in fixed format which is defined with all observing element and maximum gradient, so that it contains lots of redundant data. An self-adaptive data processing method is advanced based on flexible definition of data output format from AWS hardware interface. Meteorological elements and corresponding gradient configuration of each AWS are identified automatically, and observation data of each gradient for each element are processed self-adaptively and saved dynamically. It has indicated that data products can be achieved self-adaptively for AWS network whose element and gradient configuration differ from one another. Redundancy for data products is reduced effectively and data processing and transmission efficiency is increased greatly with a lot of storage resources being saved.