Applying a regression model to estimate honey production from weather variables

Beekeeping is a special agricultural field that provides essential resource to other agriculture through pollination services; while it creates value with honey production. Beekeeping is also one of the most weather-sensitive agricultural field. Weather influences honey production in three ways: (1) with seasonal scale weather patterns influencing phenological growth; (2) short-term (daily) weather influence on nectar production and (3) direct weather influence on bee flight.

In this study, a linear time series model is presented to predict honey production describing weather as external variable. Honey production time series was derived from hourly values of total bee and honey weight per hive, measured with hive scales located in apiaries in Debrecen, Hungary in the black locust blooming periods of 2021–2022. Data was pre-processed to remove the seasonal blooming pattern and the typical two-peak diurnal pattern of gathering by bees. Meteorological data was obtained from nearby monitoring sites of the Hungarian Meteorological Service. Mean and maximum temperatures, as well as global radiation was used as external variable in the time series regression model. Consecutive 1-day hindcasts of honey production were evaluated against observation data. Model RMSE was found to be approximately 0.6 sigma (in standardized units) over a range of model settings. However, the fitted regression coefficients had large error and were barely significant.

The method presents a statistical approach to create hive-specific daily forecasts of honey production for hives equipped with scales. The definition of the seasonal pattern (i.e. phenology) is a crucial step to create a stationary time series for predictive modeling.