All studied animals belong to the goat experimental herd of the Arid Areas Institute of Médenine Tunisia (33°30' N and 10° 40' E), which is located in southeastern Tunisia, between the mountains of Matmata and the Mediterranean Sea. This region is characterized by an arid continental Mediterranean climate; with irregular precipitations with an average annual rainfall of about 200 mm. The summer is normally the hottest and driest season with a maximum temperature of 47°C ( Ouni et al. , 2007 ).

The Tunisian local goat population is very polymor­phic ( Najari, 2005 ), but it is generally characterized by its small body size with average height of 76 cm for the male and 60cm for the female ( Ouni et al ., 2007 ). This local goat breed is famous for its walking ability, resistance to hydric restrictions and high temperature stress and good productivity in harsh conditions.

Records of 462 kiddings produced from mating of 11 sires to a total number of 185 goats between 1992 and 2014 were used for this study. Animals were mated following a breeding system of one kidding per year. The main mating period was from June to August which corresponds to births in autumn. If a doe was not pregnant during the first mating period, it was transferred to the group that was mated at the next mating period (October-November, which corresponds to births in spring). The female kids were mated for the first time between 12 and 18 months of age, depending on their birth season.

The number of goats mated per sire in a mating season varied from 5 to 17. The season of kidding begins in October and continues until February, with a concentration during November and December. Goats were randomly assigned to bucks. Bucks were changed every 5 years with replacements coming from outside flocks. Bucks were also selected from the experimental flock on the basis of weaning weight and good conformation. Does were mated to the males after weaning their kids.

The identities of newborns and their parents, date of mating, date of kidding, sex of kid, litter size and parity of does were recorded. For each individual under study a record sheet with full details of each parameter along with pedigree information were maintained. New-born kids were allowed to suckle their does and were left with them up to 5 months of age.

Two variables, kidding interval (KI) and litter size at birth (LSB) were considered as measures of goat's fertility since they represent two main components of the reproductive performance.

The significance of fixed effects of year and month of kidding, age of dam, parity and their first order interactions was tested using the ANOVA procedure of the Statistical Package of Social Sciences (SPSS.20).Significant means were separated using the Duncan's multiple range test.

A two-trait animal model, where the significant effects from the ANOVA analyses were included to correct for environmental factors, was used. The model used was:

y _ijklm = YM _i + P _j +b (AGE) + a _k + pe _l + e _ijklm

where, y _ijklm was the observation for each trait; YM _i was the fixed effect of the interaction between year and month of kidding; P _j was the fixed effect of the j ^th parity (j= 2, 3,…, 9); b was a fixed regression coefficient of y on age of the goat in which the measurement was observed (AGE); a _k was the random additive genetic effect of goat; pe _l was the permanent environmental effect of goat; and e _ijklm was the random residual term.

The effects of animal, permanent environmental and residual were assumed to be random factors with zero means and variances A⊗G ₀ , I⊗P ₀ and I⊗R ₀ , where A was the additive relationship matrix, I was identity matrixand G ₀ , P ₀ and R ₀ were co-variance matrices of the two traits for additive genetic, permanent environmental and residual effects, respectively. Variance components and associated genetic parameters, together with solutions for position parameters were estimated under a Bayesian approximation using the Gibbsf90 programme of the BLUPf90 family of programmes ( Misztal et al. , 2002 ). A total of 200,000 samples with 50,000 used as burn-in were obtained to get the posterior distributions of the parameters of interest.

Solutions for genetic effects from the previous model were used to obtain genetic trends. Averages by year of birth of the estimated genetic values were plotted against year of birth and a linear regression line fitted to obtain estimates of genetic progress per year.

Table 2 shows the results of the ANOVA analyses to test the significance of the year, month and parity number effect on K and LSB. Effect of the year-month interaction was significant for the two traits. This effect is especially due to the variations of climate, food nutritional quality and herd management along time. Present findings were also in partial agreement with the finding of Alexandre et al. (2001) in Creole goats.

Fig.1a shows seasonal patterns for KI, indicating that the worst reproductive results are obtained with goats kidding in the winter season (December-January) which would correspond to matings in July-August. This is meaning that the harsh summer conditions affect the fertility of females, both by high temperature peaks and food scarcity. Under our breeding pastoral mode, where kids are not weaned, a late kidding month means a milking period until the following mating period in summer, and these tardive goats may have serious problems to ensure fecundation under non favorable endocrine and body condition profile. Thus, the local breeders tend to reduce the kidding period to the early winter. Overall, apart from the seasonal differences, KI showed a stable trend along years.

The trend of the effect of kidding month on LSB is shown in Fig.1b. The highest values are observed for does kidding in spring, which corresponds to matings occurred in October -November. Our findings were in agreement with those of severa1 authors ( Güney et al. , 2006 ; Mellado et al. , 2006 ) who found that kidding season has a significant effect on reproductive traits. According to some authors ( Mellado & Meza-Herrera, 2002 ; Chemineau et al. , 2004 ), reproductive traits in goats and sheep varied mainly with the photoperiod. As for KI, apart from the seasonal variations, LSB shows a stable trend across the period of 14 years. The lack of a steady trend in the reproductive traits along time indicates that no improvements in the environment have been practiced in this population for reproductive traits.

Fig. 2 shows the effect of parity number on both traits. A large value of KI was observed for the second parity and then, it tended to decrease. Similar results were found by other authors ( Odubote et al. , 1992 ; Greyling, 2000 ; Song, 2003 ; Hamed et al. , 2009 ). An explanation of the larger KI for parity 2 is that goats are still growing and need more energy for this purpose. If resources are scarce, there may be a negative energy balance and production is delayed. For LSB, there seems to be a general increase as parity progresses up until the fifth parity. This may be due to improved efficiency of reproduction as the goat matures ( Song, 2003 ).

Estimated variance components from the two-trait REML analyses are presented in Table 3. According to the results shown in this table; heritability estimates of those traits are rather low. Low heritability estimates of reproductive traits are mainly due to the greater proportional influence of environmental effects, since genetic variation seems to be significant. According to the obtained estimates for the additive genetic variance, the expected values of the progeny of the best and worst animals according to their genetic values (assuming two standard deviations around the mean) could differ in up to 8 months for KI and 0.54 kids for LSB. Thus, enough genetic variability exists to practice selection.

The estimate of KI heritability was 0.12. This result was higher than 0.04 reported by Odubote (1996) and lower than 0.30 obtained by Nahardeka et al. (1995) . Singh et al. (2002) mentioned that the low heritability estimates for KI may be attributed to the low quality of pastures on which the flock was maintained, not allowing for a full expression of the reproductive potential resulting in a high environmental variance.

Heritability estimate for LSB obtained in this study, 0.15, is in agreement with the findings of Devendra (1984) for Alpine goat in France and Black Bengal goat in India, Hamed et al. (2009) for Zaraibi goats in Egypt and Amble et al. (1964) for Black Bengal and Beetal goat. Land (1978) reviewed literature on heritability estimates of LSB in sheep and found an average of 0.1. Bradford (1985) pointed out that heritability of LSB is quite low and, summarizing over 30 estimates for different breeds and methods of estimation, reported a range from 0.15 to 0.35.

Repeatability estimates obtained for LSB was 0.35. A similar result was found by Odubote (1996) for West African Dwarf goat. This moderate repeatability estimate implies that assessment of prolificacy or culling of unproductive animals based on one individual litter size may not yield the desired accurate results. The repeatability estimate of KI was lower, 0.24, indicating a less predictable estimation of individual KI from single records.

Both genetic and phenotypic correlations between KI and LSB were inferred to be low and slightly positive. A positive correlation could derive from the fact that animals that show larger KI with the previous kidding are expected to show better conditions to carry more fetuses in the following pregnancy. On the other hand, a negative correlation might have been expected since animals that have superior capacities to become pregnant in shorter periods might also be those that can carry more fetuses if the success of the reproductive axis fertility-prolificacy depends on similar pathways in the utilization of energy resources or in the endocrine processes that determine overall reproductive success. The large standard errors associated with the estimated co-variances and subsequent correlations indicate that the amount of available information may not be sufficient to estimate accurately these parameters.

Accurate prediction of the breeding values (EBVs) of animals is one of the best tools available to maximize response to a selection plan. The success of a breeding program can be assessed by testing the actual change in breeding value expressed as a proportion of expected theoretical change in the mean of breeding value for the trait under selection.

Fig. 3 shows the trend for the average estimated genetic values per year of birth. As expected from the lack of selection for the reproductive performance, no trend is observed for any of the reproductive traits. This also might indicate that, although harsh, the environmental conditions may be sufficient for a variety of genetic backgrounds to reproduce.

Reproductive performance in this population has remained at a stable level during the past 14 years, showing, however, seasonal variations that point at reduced reproductive results for summer matings. Young females showed somehow worse reproductive performance than older animals, indicating possible conflicts between growth needs and food availability. Estimates of heritability for reproductive traits in this study were moderately low, indicating that accurate selection based on doe's own performance to improve these reproductive traits will require information from a large number of kiddings.

In the next stage, the traits which could be used as selection criteria to indirectly improve doe reproductive traits should be investigated. The low estimates of heritability for reproductive traits indicated the presence of large environmental variances. More studies and larger data bases are needed to identify the true genetic behavior of such genetic resources for these environments, especially to make inference about the genetic correlation between the two reproductive traits.

In order to optimize the reproductive potential of the local goat population, it is essential to adopt an appropriate reproductive management program addressing the most important individual traits which are directly involved in increasing lifetime productivity. The genetic variability detected in the studied population allows to hold expectations for the implementation of selection programmes for fertility traits in this local breed.