Construction of rapid prediction models for TN and TP in dairy farms slurry under different seasons by near infrared spectroscopy

Abstract

Rapidly and stably measuring the total nitrogen (TN) and total phosphorus (TP) contents of dairy slurry was a critical step before the field application, which was a key component of sustainable agriculture and slurry resource utilization. In present study, near-infrared spectroscopy (NIRS) models were established for swiftly determining TN and TP contents in the slurry from diverse dairy farms under various conditions. A total of 828 samples were gathered from 33 intensive dairy farms in Tianjin City throughout spring, summer, autumn and winter. The analysis initially explored the seasonal influences on the distribution characteristics of TN and TP in slurry and the effects on NIRS. Subsequently, the study employed partial least squares (PLS) regression to establish the single-season models for predicting slurry TN and TP contents. The results demonstrated that the TN models were excellent in all the seasons, with Rp2 values ranging from 0.94 to 0.95 and RPD values between 3.88 and 4.29. The TP models were superior during spring, summer, and winter compared to autumn. Lastly, the global models of four seasons (GMFS) were developed to predict TN and TP contents. For TN, the Rp2 and RPD values were 0.85 and 2.38, which were deemed very good. For TP, the Rp2 and RPD values were 0.76 and 1.87, indicating good performance. Consequently, the GMFS incorporating data from diverse seasons exhibited a broader applicability scope, albeit with reduced precision and accuracy compared to the single-season models. These models accommodated various situational needs during slurry reapplication to farmland, providing a theoretical basis for the precise and scientifically-informed reuse of dairy farm slurry in agricultural fields.