Decentralized electric heat pumps are an important component in many future energy scenarios. However, their dispatch and efficiency are highly weather-dependent, which is often inadequately considered in the scenarios. This contribution aims to address this shortcoming by analyzing a large-scale empirical dataset of monitored heat pumps from the UK. More specifically, we propose to model heat pump behavior as a function of temperature equivalents, which do not only capture ambient temperature but also wind-driven infiltration and solar heat gains as determinants of heat demand. Based on these temperature equivalents, we derive regression models for the (normalized) heat supply and for the coefficient of performance (COP) of the heat pumps. These models can be used to better disentangle and represent the variability of heat pumps heat pumps in future energy system analyses.