Premium
Modeling Lamb Weight Changes on Wheatgrass and Wheatgrass‐Sainfoin Mixtures
Author(s) -
Karnezos T.P.,
Matches A.G.
Publication year - 1992
Publication title -
agronomy journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.752
H-Index - 131
eISSN - 1435-0645
pISSN - 0002-1962
DOI - 10.2134/agronj1992.00021962008400010002x
Subject(s) - grazing , agropyron cristatum , agropyron , zoology , regression analysis , mathematics , limiting , ovis , agronomy , biology , statistics , ecology , mechanical engineering , engineering
Prediction of animal weight change (CUM) with regression models developed from grazing trials typically uses herbage parameters and CUM measured on the same day. We hypothesized that lamb ( Ovis aries L.) CUM recorded at time t was a function of herbage quality and/or quantity measured at a previous harvest t − x (where x = days prior to measurement of CUM). Our objectives were (i) to determine if time series regression analysis (TSR) could be used to model CUM from three irrigated wheatgrasses, ‘Hycrest’ [ Agropyron cristatum (L.) Gaertner × A. desertorum (Fischer ex Link) Shulters], ‘Luna’ [ Thinopyrum intermedium subsp. barbulatum (Schur) Barkw. and D.R. Dewey], and ‘Jose’ [ T. ponticum (Podp.) Barkw. and D.R. Dewey] grown alone and with ‘Renumex’ sainfoin ( Onobrychis viciifolia Scop.), and (ii) to test the models. Replicated pastures grown on a fine, mixed thermic Torrertic Paleustolls were rotationally grazed by Rambouillet × Suffolk wether lambs for an average of 77 d in spring of 1987 and 1988. Herbage quality, quantity, and plant parts were estimated from pregrazing, after 2 and 4 d of grazing, and postgrazing (7 d) harvests and used as variables in TSR. For TSR models, lagged variables ( t − x ) were selected more (67–92% of total) than nonlagged variables ( t ), supporting our hypothesis. Time series regression models described CUM accurately (average R 2 > 0.70), but selected variables were not consistent among treatments, time lags, or years. Model testing indicated poor predictive accuracy ( r 2 = 0.07−0.51), limiting the usefulness of projecting CUM across seasons and demonstrating the necessity of testing regression models.