Simple Summary
Sheep may provide food and complementary income to Andean households, but farms differ in their land, flock size, feeding, reproduction, animal-health practices, infrastructure, and market participation. A survey of 79 sheep farms in 18 localities of Luya and Chachapoyas, Amazonas, Peru, described this diversity. Most farms were small, used extensive management and natural grazing, and combined sheep with other rural activities. Multivariate analysis produced three exploratory profiles. Profile I combined smaller flocks, fewer reported health-management components, very limited supplementation, and more frequent reproductive control. Profile II had larger flocks, allocated a high proportion of farm land to sheep, and rarely controlled reproduction. Profile III included six farms with more total land and a smaller proportion allocated to sheep. The profiles overlapped, and their membership changed when management components or land variables were represented differently. The six-farm configuration was particularly dependent on land-related variables and remained provisional. These findings describe combinations observed in the surveyed farms and provide starting points for discussions with producers. Possible feeding, health, and reproductive-management options require prospective evaluation before being used as support recommendations.
Abstract
Sheep farming in the Peruvian Andes is heterogeneous, but farm-level evidence from Amazonas is limited. This cross-sectional exploratory study characterized 79 convenience-sampled farms in 18 localities of Luya and Chachapoyas. Quantitative and categorical survey data were integrated through factor analysis of mixed data (FAMD) and Ward.D2 clustering. Median flock size was eight sheep; 96.2% of farms used extensive management and 92.4% relied principally on natural grazing. Four FAMD dimensions explained 47.20% of inertia and produced three exploratory profiles (37, 36, and 6 farms), differing in land allocation, flock size, supplementation, and management components. Separation was modest (mean silhouette = 0.215), with subsampling Jaccard stabilities of 0.751, 0.677, and 0.683. Three- and five-dimension alternatives showed adjusted Rand indices (ARIs) of 0.733 and 0.744 against the reference. Agreement was much lower when infrastructure components were represented individually (ARI = 0.023) or land variables were excluded (ARI = 0.155–0.174). Profile III therefore remained provisional. The profiles describe configurations conditional on the sample and analytical representation. Management implications require prospective validation, and non-probability sampling precludes population-level estimates of profile frequencies.