Prediction of the Nutrient Content in Dairy Manure Using Artificial Neural Network Modeling

• Published in: Journal of dairy science 2008-12, Vol.91 (12), p.4822-4829
• Main Authors: Chen, L.J., Cui, L.Y., Xing, L., Han, L.J.
• Format: Article
• Language: English
• Subjects: ammonium nitrogen; Animal productions; Animals; artificial neural network; Biological and medical sciences; Cattle - growth & development; Cattle - metabolism; dairy manure; Dairying - methods; electrical conductivity; Female; Food industries; Fundamental and applied biological sciences. Psychology; Linear Models; Manure - analysis; Milk and cheese industries. Ice creams; neural networks; Neural Networks (Computer); nitrogen content; nutrient content; organic fertilizers; phosphorus; physicochemical properties; physicochemical property; potassium; prediction; regression analysis; simulation models; specific gravity; Terrestrial animal productions; Vertebrates
• ISSN: 0022-0302; 1525-3198
• DOI: 10.3168/jds.2007-0978

Nutrients in animal manure are valuable inputs in agronomic crop production. Timely and reliable information on animal manure nutrient content will facilitate the utilization of manure as organic fertilizer and reduce any associated potential environmental problems. The objective of this study was to investigate the feasibility of using multiple linear regression (MLR), polynomial regression, and artificial neural network (ANN) models to determine nutrient content in dairy manure. Fresh manure samples (n=86) from Holstein dairy cattle were collected from 34 dairy farms located in Beijing city, China. All samples were analyzed for nutrient content (ammonium nitrogen, total potassium, total nitrogen, and total phosphorus) by standard laboratory methods. The physicochemical properties (specific gravity, electrical conductivity, and pH) of dairy manure samples were measured. Relationships between nutrient content and physicochemical properties were explored by MLR, polynomial regression, and ANN models. Several parameters (R2, modeling efficiency statistic, mean squared error of prediction, mean bias, linear bias, and maximum bias) were calculated to evaluate model performance. The residual analysis results indicated that all MLR models for the testing data set had significant mean and linear bias. When compared with MLR and polynomial regression models, the ANN model for all nutrient contents had better performance with higher R2 and modeling efficiency statistics and lower mean squared error of prediction, mean bias, linear bias, and maximum bias. These findings demonstrated that the ANN model may be an appropriate tool to predict dairy manure nutrient content.