Small Ruminant Research

Transcription

1 Small Ruminant Research 17 (1) Contents lists available at SciVerse ScienceDirect Small Ruminant Research j our na l ho me p age: ww w.elsevier.com/locate/smallrumres Characterization of the lactation curve in Murciano-Granadina dairy goats J.M. León a, N.P.P. Macciotta b, L.T. Gama c,, C. Barba a, J.V. Delgado a a Universidad de, Campus de Rabanales, Edf, Gregor Mendel, 1471, Spain b Dipartimento di Scienze Zootecniche, Universitá di Sassari, Via De Nicola 9, Sassari 71, Italy c CIISA Faculdade de Medicina Veterinária, Universidade Técnica de Lisboa, Lisboa, Portugal a r t i c l e i n f o Article history: Received 4 November 11 Received in revised form 6 March 1 Accepted 16 May 1 Available online 1 June 1 Keywords: Dairy goat Environmental effects Lactation curve Spline a b s t r a c t The lactation curve of Murciano-Granadina goats was studied by using records collected between 4 and 1 under its Official Dairy Recording program, which includes three milking control programs (, and ). A total of 518,557 test-day records corresponding to 69, lactations by 8,9 does in 1 herds were included in the analyses. Different mathematical models were fitted, and average curves for region of production, parity, kidding season and type of kidding were estimated. A quadratic spline function gave the best fitting performance and provided a good description of the Murciano- Granadina lactation curve. All the factors studied affected both the scale of the lactation curve and its shape, with more distinct curves in first lactations, goats kidding in the summer and those producing singletons. When compared with later parities, first-parity does had a lower initial level of production, later and lower peak yield, and a smaller decline in milk production after the peak. Goats kidding in the Summer had a steep increase in milk yield up to the peak, which occurred earlier and with a higher yield than in other seasons, but seasonal effects differed slightly among regions. The lactation curve was flatter and with a later peak in goats producing single kids, and a steeper increase in yield up to the peak was observed as the number of kids increased. 1 Elsevier B.V. All rights reserved. 1. Introduction The lactation curve is the mathematical representation of the physiological response of milk production throughout the milking period (Masselin et al., 1987). The study of lactation curves has several practical applications in the dairy industry, both for breeding and management purposes. The prediction of future records of lactating animals allows an early evaluation of selection candidates in progeny test programs, and the assessment of total milk yield and knowledge of lactation curve traits, including production peak and persistency of lactation, is essential for a proper estimation of animal requirements for Corresponding author. Tel.: ; fax: address: ltgama@fmv.utl.pt (L.T. Gama). diet formulation. Furthermore, the analysis of lactation curves also allows the early detection of animals to be culled for low production or with sub-clinical pathologies (Dematawewa et al., 7). In these circumstances, a crucial point is the choice of a suitable mathematical function able to efficiently describe the evolution of milk yield throughout lactation. An appropriate choice should be based on data coming from different populations and herds, and the effects of different environmental factors on function parameters should be evaluated (Wood, 198). The pattern of milk yield throughout the lactation in dairy species is usually characterized by two different phases, i.e., the ascending phase, from parturition to peak production; and the descending phase, from this maximum point until drying-off, with the slope during this phase representing the persistency of lactation (Masselin et al., 1987). The lactation curve has been deeply studied in dairy /$ see front matter 1 Elsevier B.V. All rights reserved.

2 J.M. León et al. / Small Ruminant Research 17 (1) cattle (Pollot, ), and several mathematical models have been proposed to describe it (Wood, 1967; Wilmink, 1987; Cappio-Borlino et al., 1995; Serchand et al., 1995; Vargas et al., ). There are fewer references regarding lactation curves in dairy goats (Gipson and Grossman, 1989, 199; Montaldo et al., 1997; Macciotta et al., 7; Menéndez- Buxadera et al., 1), but it is known that the shape of the lactation curve in this species is greatly affected by environmental conditions, especially in local breeds under extensive or semi-extensive farming systems (Fresno et al., 199; Rota et al., 199; Peña et al., 1999). The major factors affecting lactation curve shape include breed, kidding season, age of the doe, litter size, herd and feeding practices (Morand-Fehr and Sauvant, 198; Gipson and Grossman, 199; Wahome et al., 1994; Ruvuna et al., 1995; Montaldo et al., 1997; Macciotta et al., 7). Thus, studies on the lactation curve of a breed, herd or population may express some peculiarities of goat productive behavior in their own context, which must be studied in further detail. Currently, breeding values for production traits in dairy cattle selection programs are often predicted with random regression models (Jensen, 1; Schaeffer and Jamrozik, 8), which take into account the trajectory of milk yield throughout lactation. Nevertheless, the use of such models in selection of dairy goats has been adopted in limited situations, mostly in goats farmed in intensive conditions (Andonov et al., 7; Zumbach et al., 8), but it is not clear whether those functions could also be suitable for breeds kept in semi-extensive production systems. Therefore, preliminary work with a large amount of information is needed to assess the usefulness of different lactation curve models in extensively or semi-extensively produced breeds, before a test-day model can be widely adopted in their genetic evaluations. Murciano-Granadina is the most cosmopolitan Spanish goat breed (Camacho-Vallejo et al., 1), and it is farmed in several countries in Europe, Africa and South America. The breed is generally kept in semi-extensive systems under different climatic conditions, grazing on natural pastures and shrubs throughout the year, with supplementation in critical periods, based on either by-products or commercial feed supplements. Under these conditions, the average total yield for a standardized lactation length of 1 days is 416. ± kg for milk, ± 7.1 kg for fat, 1.5 ± 5.1 kg for protein and 5. ± kg for dry matter (unpublished data). Notwithstanding, the productive potential of Murciano-Granadina goats should be much higher, as indicated by record yields above 15 kg (Delgado et al., 1). The breed does not show seasonal reproductive behaviour and bucks are often kept in permanence with goats, so milk is produced throughout the whole year, which could explain their short lactation length. An active national breeding program of the Murciano- Granadina breed aimed at improving milk yield and composition has been in place for some years (Analla et al., 1996), and a large data set has been generated through its milk recording program. This information is useful to gather baseline information which can be used to further develop the genetic improvement program of Murciano-Granadina dairy goats. Furthermore, the vast amount of data currently available in this program can Table 1 Number of test-day records by level of the factors considered. Factor Number of records Total 518,557 Region, ,6 171,54 Kidding season Spring 159,66 Summer 94,59 Autumn 11,7 Winter 151,66 Kidding type Single 4,589 Twin 6,84 Triplet or more,584 Lactation number First 16,44 Second 144,8 Third 91,145 Fourth 55,58 Fifth or upper 67,46 provide information which may serve as a model, extending the conclusions to other goat breeds maintained under different management conditions. In the present paper, the lactation curve of Murciano- Granadina goats farmed under semi-extensive conditions is modeled with standard mathematical functions to: (1) assess the suitability of the different models to describe the lactation curve in this breed; () characterize the evolution of milk yield throughout lactation; () investigate how different environmental and biological factors affect the shape of the lactation curve.. Materials and methods.1. Data The historical archives of the official milk recording program carried out by the National Breeders Association of the Murciano-Granadina Goat Breed were used in this research. Milk recording was carried out according to A4 methods of ICAR (199), and the mean lactation length was 1 days. Data collected between 4 and 1 were edited to exclude lactations with less than 6 records, daily yields greater than 1 kg or below. kg, and yields recorded at days or after 8 days from parturition. The edited file contained 518,557 test-day records corresponding to 69, lactations of 8,9 does in 1 herds located in three dairy recording nuclei (, and ). Animals were grouped into five parity classes (1,,, 4 and 5), four kidding seasons (Winter, Spring, Summer, Fall) and three classes of kidding type (single, twin, triplet or more). The distribution of records by the different levels of the above mentioned classification factors is reported in Table 1. Overall, the information used in this study is perhaps one of the more complete data sets available for dairy goats, and could thus provide insight which can be useful for other goat breeds... Mathematical functions Daily milk yield was first averaged by day of lactation for the full data set and for each level of classification factor (region, kidding season and type, lactation number). Then, average curves were fitted with six mathematical functions, to model the lactation curve as a function of days in milk (t). The first four functions were chosen because they have been widely used in dairy cattle and, to a lesser extent, in sheep and goats. These functions were:

3 78 J.M. León et al. / Small Ruminant Research 17 (1) ) Wood (1967) incomplete gamma function (WO) y t = at b e ct ) Modification of WO introduced by Cappio-Borlino et al. (1995; CB) y t = at bect ) Cobby and Le Du (1978) function (CL) y t = a(1 e ct ) bt 4) Wilmink (1987) function (WI) y t = a + be kt + ct In all these models, y t represents the mean daily milk production recorded in day t, whereas a, b, c and k are function parameters related to curve shape. Specifically, a is a scale parameter related to milk yield at the beginning of lactation, while b and c are parameters related to the shape of the lactation curve in its ascending and descending phase, respectively. In the WI model, k is a parameter related to the time of occurrence of peak yield. The other two considered models have been chosen because, even though they are more general functions not specifically conceived to fit the shape of the lactation curve, they found recent application in lactation curve modeling, especially for their relevant flexibility (Macciotta et al., 5b). These were: 5) Quadratic spline function (SP) with one knot y t = ˇ + ˇ1t + ˇt for t X and y t = ˇ + ˇ1t + ˇt + ˇ(t X) for t > X where X is treated as an additional parameter to be estimated, and represents the day of lactation where the knot-point occurs, i.e., where the two polynomial functions are linked. 6) Legendre polynomials of order three (LG) y t = ˇP + ˇ1P 1 + ˇP + ˇP In both functions, ˇ, ˇ1, ˇ and ˇ are function parameters, while P, P 1, P and P in LG are (Schaeffer, 4): P =.771w ; P 1 = 1.47w 1 ; P =.796w +.717w ; P =.86w w with w representing a standardized time unit calculated as w = ((t 1)/(8 1)) 1 For the CL, SP, WI and WO models, time at peak occurrence was calculated analytically from estimated parameter values, while for the other functions it was assessed empirically by finding the day of lactation with the highest predicted yield. Peak yield was assessed with the estimated function, except for WO where it was calculated from parameter values. Table Coefficient of determination (R ), mean squared error (MSE), Durbin Watson statistic (DW), estimated day of peak (DP), peak yield (PY) and total milk yield (TY) for the full data set analyzed with the models of Cappio-Borlino et al. (CB), Cobby and LeDu (CL), Legendre polynomials (LG), Spline function (SP), Wilmink (WI) and Wood (WO). Parameter Model CB CL LG SP WI WO R MSE DW DP PY TY Finally, total milk yield was predicted by summing the predicted daily milk yield throughout the lactation... Lactation curve fitting and evaluation The six mathematical models were fitted both to the overall curve, averaged by day of lactation, and to average curves of different regions, kidding seasons, kidding type and lactation number, to assess their suitability in describing lactation patterns obtained in different scenarios of extensive farming. The Legendre polynomials were fitted to the lactation data using linear regression in SAS (SAS Institute Inc, 7) whereas the other five models were fitted using the NLIN Procedure of SAS (SAS Institute Inc, 7) with the method of Marquardt as estimation procedure. The convergence criterion, defined as the change in the error sum of squares between successive iterations, was set to 1 6. For the SP function, the knot position was searched between 1 and 1 days in milk. The goodness of fit achieved with each lactation curve model was evaluated by standard procedures, by obtaining the coefficient of determination (R ) and mean squared error (MSE) for each model by environmental factor combination. In addition, the possible existence of autocorrelation among residuals was tested by computing the Durbin Watson (DW) statistic, which measures the significance of the correlation between successive residuals, and should have a value close to if there is no correlation (Silvestre et al., 6). In addition to the curves fitted by level of the different factors considered, the best model was also fitted to lactation data for combinations of region-season of kidding, to investigate possible differences in the seasonal effect among the three regions considered.. Results The results of fitting the six mathematical models to the overall test-day data are summarized in Table. The R for the different models ranged between.89 (CL) and.97 (SP). The DW statistic was below 1 for all models except SP, for which it was about 1.8, indicating the existence of some autocorrelation among residuals. The estimated total milk yield was very similar across the different models, but a large variation in some shape traits was observed, with the most relevant difference observed for time at peak yield (Table ). The CL model estimated a much earlier peak (9 days of lactation) whereas the latest was for the LG model. The other four models showed intermediate values for peak occurrence. Moreover, the CL and LG models differed considerably from the other functions especially for a higher level of production at the beginning of lactation. Throughout the lactation, WO and WI models followed a very similar pattern, and the same was true for the SP and CB functions. The better fit of the SP function observed for the overall curve was confirmed also when the six different

4 J.M. León et al. / Small Ruminant Research 17 (1) mathematical models were used to analyze the data grouped by region, type of kidding, number of lactation and season of kidding (Table S1). The SP model resulted in the highest R and the lowest MSE in 89 out of 9 modelfactor combinations. Also, with only two exceptions, a DW statistic closer to was obtained for SP when compared with the other models. In any case, all the models except CL performed well, with R near or above.9, confirming their suitability, even under different management circumstances. On the other hand, the CL and LG models resulted in estimates of the DW statistic often below 1, indicating the existence of a positive autocorrelation among residuals resulting from fitting those models. The above results regarding goodness of fit indicate that the SP function is the best among the different mathematical functions tested to describe the lactation curve in Murciano-Granadina goats, so it was chosen for further analyses. The estimated curve parameters for the SP function applied to the full data set (Table ) resulted in a mean lactation curve with an initial level of milk production of 1.9 kg, a peak yield of.4 kg occurring at day 45, and with a steady decline in milk production afterwards. The cumulative milk production up to the peak and to 15 days of lactation corresponded to.4% and 54.8%, respectively, of the total lactation yield, which was estimated to be 4.6 kg at 1 days. Parameter estimates obtained by fitting the SP function to the different levels of factors of variation considered in this study are reported in Table, and the corresponding curves are graphically presented in Fig. 1a d. When the SP function was fitted by geographical region (, and ), the shape of the lactation curve differed markedly between the three groups (Fig. 1a), even though the initial level of milk production was similar (P >.5, Table ). Both the linear and quadratic components of the SP function differed among and the other regions (P <.5), but not between and (P >.5). Overall, a steeper increase in milk yield up to the peak was observed in, with the lowest increase observed in. The peak occurred at around 9 days of lactation in and, and nearly 1 days later in. Peak yield was highest in and lowest in, but the decline in milk yield after the peak was nearly parallel in the two regions. Overall, total and peak milk yield were highest in and lowest in. The quadratic SP function was then fitted to test day records grouped by type of kidding (Fig. 1b). Initial milk yield differed among types of kidding (P <.5), with an increase of about.5 kg for each additional kid produced. The linear and quadratic components of the SP function differed in single-kid parities (P <.5), but were similar among twin- and triplet-parities (P >.5). The major feature observed was that the lactation curve was flatter for goats producing single kids, and a steeper increase in yield up to the peak was observed as the number of kids increased. Mean peak yield occurred at about 4 days in goats producing twins and triplets or more, and nearly 1 days later in those producing singles. At the peak, milk yield was highest in goats producing triplets or more, by about.8 and.8 kg when compared with those producing singles and twins, respectively. For total milk yield, these differences were 1 and 56 kg, respectively. Curves estimated by number of lactation (Fig. 1c) showed a major difference between first and later parities, mainly in initial production level but also in the shape of lactation. When compared with older animals, first parity goats had a lower initial level of production (P <.5), by about.5 kg when compared with second parity and about.6.7 kg when compared with other parities. Furthermore, the quadratic component of the SP function differed for first parities (P <.5), which had a later and lower peak yield and a smaller decline in milk production after the peak, when compared with other parities. Differences between the other orders of lactation were much smaller, and mostly significant for second lactations, which had slightly lower initial and peak yields, when compared with later parities. On the other hand, the decline in production after the peak was more pronounced in fifth and later parities. Overall, milk yield throughout the lactation was lowest in first parities and highest in third and fourth parities. The application of the SP function to lactations grouped by season of kidding (Fig. 1d) allowed the detection of major differences in the shape of the lactation curve depending on the time of the year when kidding takes place. Goats kidding in the Summer have an initial level of production similar to those kidding in the Fall and Winter, but lower than Spring-kidding does (P <.5). However, their lactation curve shape differs widely (P <.5) from the other seasons, with a steep increase to the peak, which occurs earlier (nearly days) and with a higher yield than in other seasons, while after the peak the decline in milk production is similar to that observed in Spring-kidding goats. Winter kiddings resulted in lactations showing a flatter curve, with a lower peak yield and higher persistency when compared to other seasons. Lactations initiated in the Fall had a steeper increase in milk production than those started in the Spring, even though they had a similar peak yield, reached at about the same time. Nevertheless, the decline in milk production after the peak was highest in Fall-kidding does. Total milk yield was higher in Springand Summer-kiddings by about kg when compared with lactations started in the Fall and Winter. The effect of season of kidding on the shape of the lactation curve differed between the three regions considered in our study (Fig. ), with more pronounced differences among seasons than among regions. In, the lactations initiated in the summer resulted in a steeper increase in milk yield up to the peak when compared with the other regions, while those initiated in the Spring had a higher yield throughout the lactation. In lactations started in the Fall, persistency was higher for goats, but differences among regions were minor in Winter-kidding does. 4. Discussion The shape of the lactation curve supplies very important information on aspects related to milk production, such as the level and temporality of maximum production (peak), enabling farmers to plan parturition groups in order to maintain stable production levels throughout the year, responding to market demands, and also to organize

5 Table Estimated curve parameters ± standard error for milk yield (kg/day), and predicted day of peak (DP), peak yield (PY) and total milk yield (TY) for the quadratic spline function fitted to the full data set and to each level of the factors analyzed. Factor Curve parameters DP PY TY ˇ ˇ1 ˇ ˇ X Full data set 1.96 ± ± ± ± ± Region 1.87 a ± a ± a ± a ± a ± a ±.1.44 b ± b ±.6.59 b ± b ± a ± b ± b ±..5 b ±. 5. a ± Type of kidding a ± a ± a ± a ± a ± b ± b ± b ±.7. b ± b ± c ±.44. b ±..6 b ±.47.8 b ± b ± Lactation number a ± a ± a ± a ± a ± b ± bc ± bc ±..96 bc ± bc ± c ±.6.4 ab ±.9.69 bc ±.5.8 b ± c ± c ±.48.1 c ±.9.99 b ± c ± c ± c ±.16. ab ±.14.6 c ±..5 b ± ab ± Kidding season Spring.11 a ±.4.16 a ±..17 a ± a ± a ± Summer b ± b ± b ±..111 b ±. 1.7 b ± Fall 1.8 b ±.4.5 c ± c ±.9.7 c ± a ± Winter 1.87 b ± abc ± abc ±..184 a ± a ± a,b,c For a given factor, parameter estimates not sharing a common superscript are different (P <.5). 8 J.M. León et al. / Small Ruminant Research 17 (1) 76 84

6 J.M. León et al. / Small Ruminant Research 17 (1) (a) (b) (c).8 (d) Spring Summer Fall Winter Fig. 1. Quadratic spline function fitted to lactations of goats: (a) in the, and regions; (b) with single (1), twin () and triplet plus quadruplet () type of kidding; (c) in first (1), second (), third (), fourth (4) and fifth and upper lactation (5); (d) kidding in the Spring, Summer, Fall and Winter. (a) (b) (c) (d) Fig.. Quadratic spline function fitted to lactations recorded in the, and regions, for goats kidding in the (a) Spring; (b) Summer; (c) Fall and (d) Winter.

7 8 J.M. León et al. / Small Ruminant Research 17 (1) feeding groups according to nutritional demands, as a function of availability of resources and physiological needs. This is particularly important in non-seasonal breeds, such as the Murciano-Granadina. Furthermore, the accuracy of genetic evaluation for dairy traits can be increased if the shape of the lactation curve lactation is taken into account, e.g., by using random regressions (Jensen, 1; Menéndez-Buxadera et al., 1) which are based on the inclusion of mathematical models describing the trajectory of milk yield throughout the lactation. One of the aims of our study was to investigate the suitability of different mathematical models to describe the lactation curve in Murciano-Granadina, which is the major dairy goat breed in Spain, where it is raised mostly under extensive or semi-intensive conditions. The selection program now existent for Murciano-Granadina uses a genetic evaluation based on a repeatability model (Delgado et al., 6), but could advantageously incorporate a testday model for the prediction of breeding values (Wiggans and Hubbard, 1; Menéndez-Buxadera et al., 1). We used a data set including more than half-million test-day records, covering a broad geographical area and a wide range of management conditions, which should provide a good basis for the characterization of the lactation curve in Murciano-Granadina dairy goats. Moreover, our results give insight into how different factors affect the evolution of milk yield in the goat species in general, as it is not common to have such a large and well-structured field data set available, especially under extensive and semi-intensive environmental conditions. Different mathematical models have been proposed to describe the lactation curve in dairy cows, and some of them have been incorporated in genetic evaluations with random regression models (Schaeffer and Jamrozik, 8). Among the more popular in cattle are the models of Wood (1967) and Wilmink (1987), which have also been used in dairy goats (Montaldo et al., 1997), and some modifications to Wood s model have been introduced to address the specificities of small ruminants (Cappio-Borlino et al., 1995). In dairy goats, the model proposed by Cobby and Le Du (1978) was also shown to be useful in some circumstances, such as the Canary Goat Group (Fresno et al., 199) and Saanen and crosbred goats in Turkey (Takma et al., 9). In recent years, random regression models in dairy cattle have progressively adopted the use of Legendre polynomials and Spline functions to describe the evolution of milk yield throughout lactation (Silvestre et al., 6; Bohmanova et al., 8; Schaeffer and Jamrozik, 8). In our study, we investigated the usefulness of the above-mentioned models to the milk recording information collected over 7 years in the Murciano-Granadina recording scheme. When applied to the full data set, all the models studied, with the exception of LD, performed well in terms of goodness of fit, with R near or above.95, and a MSE equal or below.5. The DW statistic was below 1 for all models except SP, suggesting that some degree of autocorrelation among residuals may exist when those models are used. However, for the SP function, the Durbin Watson statistic was nearly 1.8, indicating that with this model the residuals tend to be independent. In our analyses, the suitability of the SP function as well as its flexibility were further confirmed by its superior results in nearly all the sub-sets of data where it was applied, including records grouped by region, type of kidding, lactation number and season of kidding. With the exception of the region, where nearly all models had a poorer performance, in all cases the SP function resulted in the highest R and the lowest MSE. On the other hand, the CL and LG models had the lowest R in nearly all cases. To our knowledge, LG polynomials have seldom been used to model the lactation curve in dairy goats (Zumbach et al., 8) and SP functions have not been applied in this species, even though they have been widely adopted in dairy cattle analyses over the last few years (Schaeffer and Jamrozik, 8). Overall, our results indicate that the SP quadratic function is the most adequate to describe the lactation curve in Murciano-Granadina dairy goats, as has also been shown in dairy cattle (Silvestre et al., 6; Bohmanova et al., 8; Schaeffer and Jamrozik, 8). In addition, we have successfully applied an iterative procedure to estimate the best knot-point where the two polynomial functions are linked, as opposed to the use of a pre-defined knot-point which is commonly used in dairy cattle analyses (Silvestre et al., 6; Macciotta et al., 1). The results for Murciano-Granadina indicate that peak yield occurs on average at 45 days of lactation, which is in the range found for other dairy goat populations, where the peak usually takes place between the fifth and the eighth week of lactation (Morand-Fehr and Sauvant, 198; Kala and Prakash, 199). In Spanish goats, peaks in the second and third week have been reported in the Majorera breed (Fresno et al., 1994), in the fourth week in the Malagueña (Herrera et al., 1984) and Tinerfeña breeds (Fresno, 199) and in the fifth week in Murciano-Granadina goats studied in an experimental farm (Peris, 1994). Important differences were found in the shape of the lactation curve among the three regions considered, such that goats in the region had a steeper increase to the peak, with a higher milk yield, while those in had a later peak and goats from showed a flatter lactation curve. These differences probably reflect the diversity of environmental conditions (weather, management systems, tradition, production level, etc.) associated with the extensive and semi-extensive production of Murciano-Granadina goats in those areas. The increase in the number of offspring born was associated with a higher level of initial yield and a steeper increase in milk production to reach a progressively higher peak yield. This pattern may be a consequence of the hormonal impact of the number of kids born or of the stimulus provided by a higher number of kids suckling (Fresno, 199). On the other hand, persistency of lactation was higher when the number of offspring was smaller. Herrera et al. (1984), working with Malagueña goats in southern Spain, found lactation shapes and the influence of kidding type to be very similar to those found in our study, suggesting that probably the effect of type of kidding is similar in goat breeds kept under analogous management systems. Goats in their first lactation had a lower initial level of production and a later and much lower peak yield, but with

8 J.M. León et al. / Small Ruminant Research 17 (1) a higher persistency, when compared to other lactations. A scaling effect was found in the second lactation, which increased to reach milk levels close to later lactations, even though it showed somewhat lower initial, peak and total milk yield relative to the third and later parities. Goats in the third and fourth lactation had a similar lactation curve, while those in the fifth parity increased similarly up to the peak, but declined more pronouncedly afterwards. The lower and later peak yield in first lactation dairy goats has been reported in several different breeds (Gipson and Grossman, 1989; Akpa et al., 1; Fernández et al., ; Macciotta et al., 5a), but their higher persistency has not been consistently found. Overall, the general pattern in our and other studies is that lactation number affects notably the scale and shape of the lactation curve, particularly in first-parity goats, with an increase in starting production and peak yield in later lactations (Gipson and Grossman, 199). A major seasonal effect on the shape of the lactation curve was found in our analyses, reflecting the environmental constraints associated with the extensive conditions under which Murciano-Granadina dairy goats are raised. Does kidding in the Summer had a very steep increase in milk yield in early lactation, probably reflecting the mobilization of body reserves accumulated with the abundance of pastures in the Spring. On the other hand, goats producing kids in the winter had a very modest increase in milk yield up to the peak, probably reflecting the scarcity of feed resources and the adverse influence of weather conditions in this period. Spring-kidding goats had the highest level of initial milk production, and remained with good persistency after the peak. However, does kidding in the Fall had poor lactation persistency, probably reflecting the detrimental effects of low temperatures and feed shortage in late lactation. Using a smaller data set and obtaining means per month of lactation for different kidding seasons, Díaz et al. (1999) reported a similar pattern for the effect of season on the evolution of lactation in Murciano-Granadina goats. In the United States, Gipson and Grossman (199) reported that lactations starting in the Spring reached peak yield at 9 1 weeks after parturition, whereas lactations started in June and January had a higher initial production, showed earlier peaks (5 8 weeks) and had the best persistency. Nevertheless, comparisons of seasonal effects reported for other regions and breeds must be taken with caution, as they are highly dependent on the variability among seasons and the degree to which management systems are able to minimize the exposure of goats to the impact of seasonal environmental effects. The lactation data set used in our study covered different geographical areas in southern Spain, and it was found that seasonal effects on the shape of the lactation curve were more pronounced than regional differences. However, the effects of season of kidding differed among the three regions considered, probably reflecting the severity of climatic conditions and management constraints in each region. For example, it was found that Spring and Summer kiddings had a more favorable response in early lactation in goats from than from the regions of and. This probably reflects the effects of a more intensified management practiced in region as well as its milder climate, given its proximity to the ocean, resulting in lower summer temperatures than in the and regions, which often have maximum temperatures exceeding 4 C. Also, the decline in milk yield after the peak in fall-kidding does was less pronounced in the region, probably as a result of less severe winter conditions, which would not be as detrimental in late lactation as the cold weather typical of this season in and. 5. Conclusions The results of our study indicate that a quadratic spline function was able to adequately fit the milk production pattern of the Murciano-Granadina dairy goat breed, both considering the overall curve or some classification criteria such as region, type of kidding, lactation number and season of kidding. In our research, most of the factors studied affect both the scale of the lactation curve as well as its shape, with important differences in the initial level of production, steepness to reach peak yield and persistency of lactation, especially in first lactations, goats kidding in the summer and those producing singletons. A similar pattern could be expected to occur in other goat breeds managed under extensive and semi-intensive systems. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at References Akpa, G.N., Asiribo, O., Oni, O.O., Alawa, J.P., 1. The influence of nongenetic factors on the shape of lactation curves in Red Sokoto Goats. J. Anim. Sci. 7, 9. Analla, M., Jiménez-Gamero, I., Muñoz-Serrano, A., Serradilla, J.M., Falagán, A., Estimation of genetic parameters for milk yield and fat and protein contents of milk from Murciano-Granadina goats. J. Dairy Sci. 79, Andonov, S., Ødegård, J., Boman, I.A., Svendsen, M., Holme, I.J., Adnøy, T., Vukovic, V., Klemetsdal, G., 7. Validation of test-day models for genetic evaluation of dairy goats in Norway. J. Dairy Sci. 1, Bohmanova, J., Miglior, F., Jamrozik, J., Misztal, I., Sullivan, P.G., 8. Comparison of random regression models with Legendre polynomials and linear splines for production traits and somatic cell score of Canadian Holstein cows. J. Dairy Sci. 91, Camacho-Vallejo, M.E., Barba, C.J., León, J.M., Vallecillo, A., Pleguezuelos, J., 1. Murciano-Granadina. In: Fernández, M., Gómez, M., Delgado, J.V., Adán, S., Jiménez, M. (Eds.), Guia de campo de las razas autóctonas Españolas. Ministerio de Medio Ambiente, y Medio Rural y Marino, Spain, pp.. Cappio-Borlino, A., Pulina, Rossi, G., A non-linear modification of Wood s equation fitted to lactation curves of Sardinian dairy ewes. Small Rumin. Res. 18, Cobby, J.M., Le Du, Y.L.P., On fitting curves to lactation data. Anim. Prod. 6, Delgado, J.V., León, J.M., Gama, L.T., Lozano, J., Quiroz, J., Camacho, M.E., 6. Genetic parameters for milk yield traits in Murciano Granadina goats in the high lands. In: 8th World Congress on Genetics Applied to Livestock Production, Belo Horizonte, Brazil. Delgado, J.V., Gómez, M., Pleguezuelos, J., 1. Catálogo de Sementales 11 de la Asociación Nacional de Criadores de Caprino de Raza Murciano-Granadina. Córdoba, Spain, 8 pp. Dematawewa, C.M.B., Pearson, R.E., van Raden, P.M., 7. Modeling extended lactations of Holsteins. J. Dairy Sci. 9,

9 84 J.M. León et al. / Small Ruminant Research 17 (1) Díaz, E., Analla, M., Muñoz-Serrano, A., Alonso, A., Serradilla, J.M., Variation of milk yield and contents of total casein and casein fractions in Murciano-Granadina goats. Small Rumin. Res. 4, Fernández, C., Sánchez, A., Garcés, C.,. Modeling the lactation curve for test-day milk yield in Murciano-Granadina goats. Small Rumin. Res. 46, Fresno, M., 199. Study of Milk Production in the Canary Goat Group. Ph.D. Thesis. University of, Spain. Fresno, M.R., Delgado, J.V., Rodero, J.M., 199. First lactation curve in Canary goats. Arch. Zootec. 41, Fresno, M.R., Gómez, J., Molina, A., Darmanin, N., Capote, J.F., Delgado, J.V., Preliminary study of the Majorera milk goat productive performance. Arch. Zootec. 4, Gipson, T.A., Grossman, M., Diphasic analysis of lactation curves in dairy goats. J. Dairy Sci. 7, Gipson, T.A., Grossman, M., 199. Lactation curves in dairy goats: a review. Small Rumin. Res., Herrera, M., Peña, F., Aparicio, J.B., Subires, J., Lactation curve of Malagueña goat breed. In: 9th Scientific Conferences of the Spanish Society on Sheep and Goat Production, Málaga, Spain, 6 9 September, pp ICAR, 199. International Committee for Animal Recording. International Regulations for Milk Recording in Goats. ICAR, Rome, Italy, p. 15. Jensen, J., 1. Genetic evaluation of dairy cattle using test-day models. J. Dairy Sci. 84, Kala, S.N., Prakash, B., 199. Genetic and phenotypic parameters of milk yield and milk composition in two Indian goat breeds. Small Rumin. Res., Macciotta, N.P.P., Fresi, P., Usai, G., Cappio-Borlino, A., 5a. Lactation curves of Sarda breed goats estimated with test-day models. J. Dairy Res. 7, Macciotta, N.P.P., Vicario, D., Cappio-Borlino, A., 5b. Detection of different shapes of lactation curve for milk yield in dairy cattle by empirical mathematical models. J. Dairy Sci. 88, Macciotta, N.P.P., Dimauro, C., Steri, R., Cappio-Borlino, A., 7. Mathematical modelling of goat lactation curves. In: Cannas, A., Pulina, G. (Eds.), Dairy Goats Feeding and Nutrition. CAB International, United Kingdom, pp Macciotta, N.P.P., Miglior, F., Dimauro, C., Schaeffer, L.R., 1. Comparison of parametric, orthogonal, and spline functions to model individual lactation curves for milk yield in Canadian Holsteins. Italian J. Anim. Sci. 9, Masselin, S., Sauvant, D., Chapoutot, P., Milan, D., Les modèles d ajustement des courbes de lactation. Ann. Zootech. 6, Menéndez-Buxadera, A., Molina, A., Arrebola, F., Gil, M.J., Serradilla, J.M., 1. Random regression analysis of milk yield and milk composition in the first and second lactations of Murciano-Granadina goats. J. Dairy Sci. 9, Montaldo, H., Almanza, A., Juárez, A., Genetic group, age and season effects on lactation curve shape in goats. Small Rumin. Res. 4, 195. Morand-Fehr, P., Sauvant, D., 198. Composition and yield of goat milk as affected by nutritional manipulation. J. Dairy Sci. 6, Peña, F., Vega, J., Sánchez, M., Martos, J., García, A., Doménech, V., Milk production and adjustment of the lactation curve in Florida goat breed. Arch. Zootec. 48, Peris, S., Characteristics of Lactation Curve and Machine Milkability of Murciano-Granadina Dairy Goats. Ph.D. Dissertation, Univ. Autónoma of Barcelona, Bellaterra, Spain. Pollot, G.E.,. A biological approach to lactation curve analysis for milk yield. J. Dairy Sci. 8, Rota, A.M., Gonzalo, C., Rodríguez, P.L., Rojas, A.I., Martín, L., Tovar, J.J., 199. Effects of stage of lactation and parity on somatic cell counts in milk of Verata goats and algebraic models of their lactation curves. Small Rumin. Res. 1, Ruvuna, F., Kogi, J.K., Taylor, J.F., Mkuu, S.M., Lactation curves among crosses of Galla and East African with Toggenburg and Anglo-Nubian goats. Small Rumin. Res. 16, 1 6. SAS Institute Inc., 7. SAS User s Guide: Statistics version. Version 9.1. SAS Institute, Inc., Cary, NC. Schaeffer, L.R., 4. Applications of random regression models in animal breeding. Livest. Prod. Sci. 86, Schaeffer, L.R., Jamrozik, J., 8. Random regression models: a longitudinal perspective. J. Anim. Breed. Genet. 15, Serchand, L., McNew, R.W., Kellogg, D.W., Johnson, Z.B., Selection of a mathematical model to generate lactation curves using daily milk yields of Holstein cows. J. Dairy Sci. 78, Silvestre, A.M., Petim-Batista, F., Colaç o, J., 6. The accuracy of seven mathematical functions in modeling dairy cattle lactation curves based on test-day records from varying sample schemes. J. Dairy Sci. 89, Takma, C., Akbaş, Y., Taskin, T., 9. Modelling lactation curves of Turkish Saanen and Bornova goats. Asian J. Anim. Vet. Adv. 4, Vargas, B., Koops, W.J., Herrero, M., van Arendonk, J.A.M.,. Modelling extended lactations of dairy cows. J. Dairy Sci. 8, Wahome, R.G., Carles, A.B., Schwartz, H.J., An analysis of the variation of the lactation curve of small East African goats. Small Rumin. Res. 15, 1 7. Wiggans, G., Hubbard, S.M., 1. Genetic evaluation of yield and type traits of dairy goats in the United States. J. Dairy Sci. 84 (E. Suppl.), E69 E7. Wilmink, J.B.M., Adjustment of test-day milk, fat and protein yield for age, season and stage of lactation. Livest. Prod. Sci. 16, Wood, P.D.P., Algebraic model of the lactation curve in cattle. Nature 16, Wood, P.D.P., 198. Breed variation in the shape of the lactation curve of cattle and their implications for efficiency. Anim. Prod. 4, Zumbach, B., Tsuruta, S., Misztal, I., Peters, K.J., 8. Use of a test day model for dairy goat milk yield across lactations in Germany. J. Anim. Breed. Genet. 15,