An attempt of using public ambient temperature data in swine genetic evaluation for litter size traits at birth in Japan

To obtain the fundamental information on using ambient temperature information in developing the model for routine swine genetic evaluation in Japan, we analyzed total number born (TNB), number born alive (NBA), and number stillborn (NSB) collected at a Japanese farm, together with off-farm ambient temperature measured at a nearest Automated Meteorological Data Acquisition System station. Five repeatability animal models were exploited, considering the effects of farrowing season (model 1), farrowing month (model 2), quadratic regressions of daily maximum ambient temperature of farrowing day (model 3), season and temperature (model 4), or month and temperature (model 5). Patterns of the effects of daily maximum temperature of farrowing day estimated using model 3 was similar to those of farrowing season by model 1 and those of farrowing month by model 2. Adding the effect of daily maximum temperature of farrowing day (models 4 and 5) could explain phenotypic variability greater than only considering either of farrowing season and month (models 1 and 2). Estimated heritability was stable among the models and the rank correlation of predicted breeding values between models was >0.98 for all traits. The results indicate the possibility that using public ambient temperature can capture a large part of the phenotypic variability in litter size traits at birth caused by the seasonality in Japan and do not harm, at least, the performance of genetic evaluation. This study could support the availability of public meteorological data in flexible developing operational models for future swine genetic evaluation in Japan.

season and temperature (model 4), or month and temperature (model 5). Patterns of the effects 23 of daily maximum temperature of farrowing day estimated using model 3 was similar to those In Japan, pork production traits, including average daily gain, longissimus muscle area, and 38 intramuscular fat content with middle to high heritabilities, have been genetically improved by 39 selection (e.g., Suzuki et al. 2005;Kadowaki et al. 2012;Ohnishi and Satoh 2018). Now,40 improving sow lifetime productivity is a pressing challenge to efficient pork production, 41 although the heritabilities of litter size traits at birth have been estimated to be low (e.g., values that have higher accuracies by using more phenotypic information obtained from 50 relatives reared on different farms and that can be directly compared between individuals on 51 different farms. Therefore, it is important to provide an operational model suitable for a large-52 scale routine genetic evaluation by simultaneously using data collected from around Japan. 53 Japan is an island country that has four distinct seasons with a climate ranging from 54 subarctic in the north to subtropical in the south, and the conditions are different between the 55 Pacific side and the Sea of Japan side 56 (https://www.data.jma.go.jp/gmd/cpd/longfcst/en/tourist.html). Japanese pig farms are widely 57 distributed in Japan (e.g., Koike et al. 2018;Ogawa et al. 2019c;Fujimoto et al. 2021).

4
Previous studies have reported that seasons affect the meat production and reproductive 59 performance of pigs reared in Japan (e.g., Harada et al. 1992 The following single-trait linear animal model was used: of the random effects were as follows:  Table 3. Fig. 5 shows the relationship between mating and farrowing dates and that 201 between off-farm daily maximum temperatures of mating and farrowing days.

202
Values of the effects of spring and summer at farrowing on TNB estimated using 203 model 1 were similar to each other, that of autumn was slightly lower than those of spring and 204 summer, and that of winter was the lowest (Table 2). TNB would be largely determined by 205 ovulation rate, early embryonic mortality, and early fetal death (e.g., Edwards et al. 1968;Wildt 206 et al. 1975;Nardone et al. 2006). Considering that the average value of gestation length in our 207 population was 115.9 days, nearly 4 months, and that >90% of the farrowing records exhibited 208 the gestation length ranging from 114 to 118 days (Fig. S1), most litters farrowed in winter had 209 been artificially inseminated from August to October, hotter months in this study (Fig. 2) Values of the estimated effects of spring and autumn at farrowing on NSB were similar, 218 that of summer was slightly higher than those of spring and autumn, and that of winter was the 219 highest (Table 2). It has been reported that heat stress in later pregnancy increased the number 220 of stillborn piglets (e.g., Edwards et al. 1968;Omtvedt et al. 1971;Wegner et al. 2016).

221
Therefore, the slightly lower value for summer at farrowing might be due to heat stress in dams.

222
On the other hand, the lowest value for winter might be caused by cold stress in not only dam 223 but also piglet. For example, previous studies reported the range of comfortable temperature of 224 18℃ to 23℃ for lactating sow (Yan and Yamamoto 2000;Brown-Brandl et al. 2001)  structure about these factors (Fig. 5), due to seasonal variation in ambient temperature and less

Genetic parameter estimation and breeding value prediction 299
Results of estimating genetic parameters are listed in Table 4. Values of Spearman's rank 300 correlation coefficients of predicted breeding values the 437 sows with their own records 301 between the models are shown in Table 5 data collected around Japan. In this study, we used public meteorological data as a source of 311 climate information to analyze phenotypic records of TNB, NBA, and NSB. This is the first 312 study to assess the performance of using public ambient temperature data in swine genetic 313 evaluation in Japan. We revealed that adding the effect of temperature could explain additional 314 variations that did not explain by considering only the effect of season (Table 3). Possible 315 reasons for our results would be that high one-to-one correspondence of off-farm temperature 316 data with on-farm temperature (Fig. 1) and that the values of temperature varied within each 317 season (Fig. 2). On the other hand, it should be noted that such correspondence might not be Restrictions apply to the availability of these data, which were used under license for this study.

359
The authors declare that they have no competing interests.    Note: Values were adjusted so that the estimated values of winter were 0 for Models 1 and 4 and those of January were 0 for Models 2 and 5.