Prediction Of Fecal Nitrogen And Phosphorus Excretion For Lactating Dairy Cows In Large-scale Dairy Farms

As the growing demand for milk and meat products,intensive dairy farming has developed rapidly,especially in China which has been the third largest milk production country after America and India in2014.However,environmental issues caused by nitrogen and phosphorus excretion also become increasingly prominent due to lacking of reasonable development plan and sufficient land to utilize manure.Therefore,there is a need for prediction equations of fecal nitrogen(FN,g·d^-1·head^-1)and fecal phosphorus(FP,g·d^-1·head^-1)excretion in order to facilitate its utilization as organic fertilizer and reduce any associated potential environmental problems.In this study,an analysis of predicting FN and FP excretion for lactating dairy cows was conducted using a data set from 18 dairy farms in the north of China,including Heilongjiang Province,Liaoning Province,Inner Mongolia,Shandong Province,Hebei Province and Tianjin,which are major districts of dairy farming.The independent variables data set,obtained by the questionnaire,consisted of 131 sets of diet nutrient composition,including dry matter intake(DMI,kg·d^-1·head^-1),crude protein?CP,%of DM?,organic matter intake(OMI,kg·d^-1·head^-1),nitrogen intake(NI,kg·d^-1·head^-1)and phosphorus intake(PI,g·d^-1·head^-1),and animal characteristics,including days in milk?DIM,d?,milk yield(MY,kg·d^-1·head^-1)and body weight(BW,kg·head^-1).Besides,131 fecal samples were collected to analyze FN and FP excretion,which were as dependent variables of prediction models.The whole data set was divided into training data set?N=110?and testing data set?N=21?in proportion to 5:1.In the training data set,correlation between diet and animal variables were examined and several variable subpools were derived,which were used to develop equations to predict FN and FP excretion based on stepwise regression analysis.The predictive power of equations was assessed using 3-fold cross-validation and external validation,which was conducted using the testing data set.The main results were listed below:?1?Prediction equations for FN excretion based on CP?FN=71.176+4.707×CP?or MY?FN=91.243+1.876×MY?alone and the equation based on CP and PI?FN=21.564+4.101×CP+0.746×PI?were not recommended to predict FN excretion with low determination coefficients of calibration?R_C²<0.5?and large root mean square error of calibration(RMSEC>15 g·d^-1·head^-1).?2?Prediction equations for FN excretion using NI?FN=40.956+195.383×NI?or OMI?FN=-4.134+7.823×OMI?alone as the independent variable and using PI and MY?FN=57.880+0.573×PI+1.455×MY?or NI and DIM?FN=27.131+207.64×NI+0.036×DIM?as independent variables can be used to predict FN excretion.The determination coefficients of calibration?R_C²?of these equations were higher than 0.5 but less than 0.7,and the stability and predictability of equations need to be further improved.?3?Prediction equations based on OMI and BW?FN=33.050+7.809×OMI-0.06×BW?or DMI and CP?FN=-33.480+6.885×DMI+2.296×CP?showed accurate prediction for FN excretion with R_C²>0.7 and small RMSEC,which were recommended to predict FN excretion accurately.?4?Prediction equations using MY?FP=41.451+0.237×MY?or NI?FP=31.263+31.430×NI?or OMI?FP=19.618+1.488×OMI?or DMI?FP=19.533+1.382×DMI?alone as the independent variable were not recommended to predict FP excretion with low R_C²?R_C²<0.5?and large RMSEC.The prediction equation based on PI?FP=21.087+0.340×PI?was recommended to predict FP excretion accurately with R_C²=0.67 and RMSEC=4.99 g·d^-1·head^-1.And the validation results,assessed using 3-fold cross-validation and external validation,showed that stability and predictability of the equation were good.