基于中红外光谱的规模化奶牛场粪水总氮快速预测方法

为建立适用于规模化奶牛场粪水中总氮含量的快速预测方法,以天津市23家种养结合型规模化奶牛场粪水治理全过程环节的样品为研究对象,探讨了基于中红外衰减全反射光谱技术快速预测粪水总氮含量的可行性。以偏最小二乘法分别建立了同一奶牛场粪水总氮含量的全程定标模型和不同奶牛场粪水总氮含量的全局定标模型,并采用独立的预测集验证了模型有效性。结果表明:全程模型总氮预测含量与实测含量的线性拟合相关系数为0.98,预测均方根误差RMSEP为130.18 mg/L,剩余预测偏差为4.97,可用于某一奶牛场粪水运移全过程环节总氮含量的快速预测;全局模型总氮预测含量与实测含量的线性拟合相关系数为0.97,预测均方根误差RMSEP为191.66 mg/L,剩余预测偏差为3.83,可用于不同奶牛场多元因素条件下总氮含量的快速预测。研究表明,基于中红外衰减全反射光谱可以实现对不同类型规模化奶牛场粪水治理全过程环节样品总氮含量的即时监测和跟踪。

Rapid prediction method of total nitrogen in slurry of large-scale dairy farm by mid-infrared spectroscopy

How to treat the high amount and concentration of slurry has been the unprecedented challenge for the intensive dairy farms in China for now. Recycling to the farmland is the fundamental way out in line with the long-term practical experiences from many developed countries. But the nitrogen content in the slurry was hard to rapidly and accurately predict on spot that caused the difficulty of recycling. Many characteristics, such as the breeding scale, layout of dairy barns, breeding modes, approaches of manure collection and treatment that influence on the variation of nitrogen content in the links of slurry movement route between China and western countries were mostly different. And the conventional regular monitoring process was normally time-consuming and costly that throughout the sample collection, transfer, preservation, pre-treatment, measurement and so forth. Therefore, it was extremely urgent and meaningful to develop rapid quantitative analysis method which was appropriate for the complicated on-spot factors and conditions. In recent, Ministry of Agriculture and Rural Affairs of China has intensively issued a series of action plans to clearly indicate the importance of improving the testing method and criteria system for recycling the slurry to the farmland. So, 23 typical large-scale dairy farms from 5 predominant dairy industry areas of Tianjin with the farming-breeding combination mode were selected, the whole process analysis of the total nitrogen (TN) in one farm, encompassing the whole chain of slurry management, was carried out. Meanwhile, the overall analysis of TN in 23 different types of dairy farms was implemented, as well that integrating with comprehensive factors including the district, scale, manure collection and treatment ways and so forth. Main objective of the research was to establish the mathematical models available to rapidly predict the TN content under the conditions of the whole process of slurry management together with the on-spot complex situations, and to provide the practical technology for criteria setting in order to help recycling the slurry to the farmland. The feasibility of fast and accurately measurement of the TN contents by means of the mid-infrared attenuated total reflectance (ATR) technology was studied in this research. Calibration model of whole process for TN contents of slurry from the identical dairy farm and calibration model of overall situation for TN contents of slurry from different dairy farms were respectively established using the partial least squares (PLS). The model availability was verified by the independent prediction set. And the principal component analysis (PCA) clustering towards the mid-infrared ATR was also used in this study. The results showed that the characteristics of slurry samples from different dairy farms were different. Linear fitting correlation coefficient between the predicted TN contents in the whole process model and measured contents was 0.98, while the root mean square error of prediction (RMSEP) and residual predictive deviation (RPD) was 130.18 mg/L and 4.97, respectively. In the global model, linear fitting correlation coefficient was 0.97, while the RMSEP and RPD was 191.66 mg/L and 3.83, respectively. Prediction results with extensive application and better stability would be achieved via the established models. Instantaneous monitoring and tracing on the TN contents of samples from the whole management course and sections in different types of large-scale dairy farms based on the mid-infrared ATR could be realized. The research would provide a reference for the development of generalized rapid and accurate prediction technology and equipment in TN content for large scale farm management.