A METHODOLOGY FOR DETERMINING CATTLE'S DUNG POSITION IN GRAZED HILL PASTURE

摘要

Livestock excrement is one of the major sources of greenhouse gas (GHG) emission in grazed pasture. This study compared several modeling approaches in estimating spatial distribution of cattle's dung from the animal activity and geographical data. Animal activities (grazing [active] or resting [inactive]) and their GPS locations were obtained in our previous results (Yoshitoshi et al. 2011, ACRS). The study was conducted in a mixed sown pasture plot (0.85 ha) located on a northeast slope ranging from 115 to 135 m above sea level. 20 cows were grazed there for four days (June 14 to June 18, 2010), and six cows in them were fitted with GPS-accelerometer (LCEX) collars. We observed the behaviors of the six cows for each 15 hours during the grazing period. After the four days grazing treatment, we set 10 m ×10m grid cell in the plot and counted the number of dung in each cell. We also estimated the grazing time in each cell from the LCEX data. Of several modeling (Geographically Weighted Regression [GWR], k-Nearest Neighbour Regression [kNNR], Random Forest Regression [RFR] and the Generalized Additive Model [GAM]) approaches developed, the RFR model showed the best prediction (R~2 = 0.92) about the excrement distribution, using independent variables the animal activity, grass quantity and quality, slope, distance from water trough and fence, northings and eastings in 100 m~2 grid cells. This model will be revised as new data become available and by inclusion of farm features such as trees, shelter belts and gateways around which animals typically congregate.