Abstract: In order to explore the pollution and distribution of residual mulching plastic film in soil in cotton field in recent 30 years, the sample plots in 121st State Farm of Xinjiang Corps of Production and Construction were selected as the experimental areas, where the mulching film was used for 5, 9, 11, 13, 15 and 19 years respectively. The Matlab program was used to develop the BP neural network model for hierarchical prediction and overall prediction of sample data. The results showed that the residual film area and quality data predicted with the model could be used to perfectly reflect the distribution of residual film in soil in cotton field in recent 30 years. The residual film in topsoil was broken year by year with the time increase of mulching plastic film and moved down to deep soil (30-40 cm). The predicted result revealed that the residual film density in cotton field would reach 419.19 kg·hm^-2 after mulching plastic film was used for 30 years, which would be nearly 6 times of the national standard limit (75.0 kg·hm^-2). In order to solve this problem, it is suggested to use the plastic film thicker than 0.010 mm, increase the recovery of mulching plastic film, and ensure the sustainable development of cotton field.

Key words: typical oasis, BP Neural Network, cotton field, residual film pollution, residual film density, sustainable development, Shihezi