The study was conducted in three agroecological zones: Congo-Nile/Western Agroecological Zone (WAZ), Central plateau/Central Agroecological Zone (CAZ), and Eastern plateau/East Agroecological Zone (EAZ) (Fig. 1). In WAZ, Karongi, Nyabihu and Rubavu districts were involved in the study. The zone is characterized by high altitude (1800–3000 m) and high annual rainfall (1200–1600 mm). The average annual temperature ranges between 15 and 17oC. Livestock is dominated by crossbreds and exotic cattle and production system is mainly extensive in Nyabihu while in Karongi and Rubavu is predominantly zero grazing. Districts selected from CAZ were Huye and Kamonyi. The altitude ranges from 1100 to 1700 m, the annual rainfall varies from 1000 to 1500 mm and the annual average temperature ranges from 18 to 20°C. Livestock is dominated by cattle, goats and pigs in a predominantly zero grazing practice. In EAZ, the study was conducted in the districts of Kayonza, Rwamagana and Gasabo. In this zone, cattle and goats are the dominating livestock. Both exotic and local cattle are found in the area. The farming is characterized by mixture of zero and free grazing. The pasture is fenced by euphorbia spp. In zero grazing, the cows are mainly feed on napier planted along the road, on contour hedge rows bordering planting for soil erosion control and around the marshland. In this land use system, there is variation of altitude with gentle slopes between 1200 to 1500 m above sea level. The rainfall varies between 800 and 1000 mm annually and the temperature ranges between 20 and 22 oC.
The cows were managed differently depending on feed availability under both intensive and semi-intensive management system. The feed included natural pasture (cut-and-carry), hay, milling by-products, concentrate mix and non-conventional feeds. Cows were hand-milked twice per day. Animals were watered from piped water and mineral licks were provided ad libitum. Natural mating and artificial insemination were used for breeding cows. There was regular vaccination against notifiable diseases such as anthrax, black leg, lumpy skin disease and foot and mouth disease. Spraying or dipping against ticks was regular and farmers called veterinarians to treat their animals whenever diseases occurred.
Data recording and statistical analysis
This study focused on specific dairy cattle farms within different agroecological zones of Rwanda. Within these zones, the survey was conducted in districts known to have a large dairy cattle population. Ease of access and size of the farm were among the factors considered in purposely selecting farm and the list of farmers was acquired in collaboration with district and sector veterinary officers. The survey was conducted by University of Rwanda students as interviews in the year 2013. The respondents were either owners or managers of each dairy farm. The survey technique was of single-visit multi-subject type (ILRI, 1990). The information collected covered several production parameters including breed of the cow, dam weight, age and puberty weight, age at first calving, current age, calving interval, milk yield, lactation length and longevity, but for this study only milk yield per day (in L) was considered. Milk produced at the farms was recorded for both morning and afternoon for that particular day and summed to daily milk yield. Breed groups were classified as Ankole x Holstein Friesian (AF), other Ankole crossbreds (AX, mainly Jersey, Sahiwal and Brown Swiss) and pure Holstein Friesian (F). Combining several crosses into AX became necessary because each of these crossbreds occurred with low frequencies. Unfortunately, purebred Ankole were found almost only in CAZ and were thus not included in the analysis.
The data were entered and organized in Excel spreadsheets and statistical analyses were carried out using SAS software (SAS, 2012). The linear model included fixed effects of age of cow (in years, 2–8+), breed group (AF, AX, or F), agroecological zone (WAZ, CAZ, or EAZ), and the interaction between breed group and agroecological zone. Differences between Least Squares Means (LSM) were considered to be significant at the level p < 0.05.