Kouloumbris Farm Data

Overview-------------- Daily herd-management records from a commercial dairy cattle farm in Cyprus (Kouloumbris Farm) covering the period July 2001–May 2026. The dataset contains milk production records, animal information and daily environmental heat-stress (HS) measurements (maximum temperature-humidity index, THI). The analyses presented in the accompanying publication were conducted on a filtered subset generated from this dataset using the preprocessing pipeline provided in the accompanying source code. File----- kouloumbris_farm_data.csv The CSV file is encoded in UTF-8 and includes a header row. Variables------------- Animal Number Farm-assigned animal identifier. Animal Life No. Official livestock identification number. Lactation Number Current parity. Date Observation date (DD-MM-YYYY). Mother Life No. Dam identification number. Lactation Days Days in milk (DIM). Group Number Farm management group identifier. Group Farm management group. Location Barn location. Gestation Gestation status. Day Production Daily milk yield (kg). THI_max Daily maximum THI. Raw dataset------------------ Date range: July 2001 – December 2025Records: 930,130Variables: 12 Preprocessing for the published analyses-------------------------------------------------------- The analyses reported in the accompanying publication were performed on a filtered subset of the raw dataset. The following inclusion criteria were applied: Observations between January 2022 and December 2025. Lactation number between 1 and 4. Daily milk production greater than 15 kg. Lactation days between 1 and 300 days. Barn location restricted to MILKING AREA A and MILKING AREA B. Subsequently, the filtered records were used to construct lagged and cumulative HS exposure variables, seasonal covariates and the analysis dataset used for causal machine learning. Data Collection--------------------- The data were routinely recorded by the farm herd-management system during commercial dairy operation. Daily maximum THI values were linked to each production record and used to quantify environmental heat exposure. Weather data were obtained via the Cyprus National Open Data Portal. Associated publication------------------------------- This dataset accompanies the manuscript: Causal machine learning reveals heterogeneous and dynamic effects of heat stress on dairy production.