This paper describes a kind of mesoscale analysis field by use of LAPS （Local Analysis and Prediction System） ingesting surface observation, radiosonde observation, radar reflectivity, and analyzes the LAPS field compared with observed precipitation, radar echo, and TBB. The results indicate that LAPS analysis field has more detailed description of mesoscale system. Various kinds of experiments are designed to ingest various observed data and investigate the effect of these data in LAPS, then the RMS errors of these experiments are calculated. It is shown that the LAPS analysis field which ingested all of the observed data is the best. The model AREM is used to forecast precipitation respectively by two kinds of initial fields, and the TS scores of two forecast fields show that LAPS has good ability to improve the initial field, and then improve the forecast ability of model.