Modeling rennet coagulation time and curd firmness of milk

Milk coagulation properties (MCP) are traditionally expressed using rennet coagulation time (RCT), time to curd firmness (CF) of 20 mm (k(20)), and CF 30 min after enzyme addition (a(30)) values, all of which are single-point measures taken from the output of computerized renneting meters, such as the Formagraph. Thus, traditional MCP use only some of the available information. Moreover, because of the worldwide spreading of breeds such as the Holstein-Friesian, characterized by late-coagulating milk, it happens often that some samples do not coagulate at all, that an is strongly and negatively related to RCT, and that k20 is not measurable. The aim of the present work was to model CF as a function of time (CFt, min) over a 30-min interval. The model tested was CFt = cF(p) x (1 - e(-kCFx(t-RCT))), where CFp (mm) is the potential asymptotical CF at an infinite time, K-CF (min(-1)) is the curd firming rate constant, and RCT is measured in minutes. The CFt model was initially applied to data of milk of each of 105 Brown Swiss cows from 7 herds, each sampled once (trial 1). Four samples did not coagulate within 30 min. Eighty-seven of the 101 individual equations obtained fit the CF data of milk samples very well, even though the samples differed in composition, and were produced by cows of different ages and days in milk, reared on different farms (coefficient of determination >0.99; average residual standard deviation = 0.21 mm). Samples with a very late RCT (slowly coagulating samples) yielded so few observational data points that curve parameters could not be precisely estimated. The repeatability of CFt equation parameters was estimated using data obtained from 5 replicates of each of 2 samples of bulk milk from 5 Holstein-Friesian cows analyzed every day for 5 consecutive days (trial 2). Repeatability of RCT was better than that of the other 2 parameters. Moreover, traditional MCP values (RCT, a(30), and k(20)) can be obtained from the individual CFt equations, using all available information. The MCP estimated from equations were very similar to the single-point measures yielded by the computerized renneting meter (coefficient of determination >0.97), but repeatability was slightly better. The model allowed the estimation of k(20) for samples with a very late coagulation or with very slow curd firming. Finally, the 3 novel parameters used to assess different milk samples were less interdependent than are the traditional measures, and their practical and scientific utility requires further study.