Quantifying middle-atmosphere turbulence from observational data remains a major challenge. While ground-based lidars have accumulated decades of high-resolution wind and temperature profiles, extracting a quantitative turbulence metric from them has not been possible, forcing observers to rely on ambiguous proxies like the Richardson number (Ri). However, using the UA-ICON model, we first demonstrate that Ri alone is an insufficient indicator of turbulence intensity. To overcome this limitation, we present an algorithm that extracts turbulent kinetic energy (TKE) directly from observational data by adapting the prognostic TKE budget of the ICOsahedral Non-hydrostatic (ICON) general circulation model to work exclusively with observable variables. Validated against UA-ICON output, our algorithm reconstructs TKE from wind and temperature fields without tuning, matching the model TKE with a correlation of 0.98. Applying this method to 187 hours of ALOMAR RMR lidar measurements, we obtain the first quantitative estimate of TKE in the 35 to 60 km altitude range derived entirely from lidar observations. The results reveal turbulence as highly structured, intermittent patches with a minimum of 0.09 m² s^−2 and a maximum of 2.8 m² s^−2 (1st–99th percentile) and a median of 0.54 m² s^−2. By aligning observations with model parameterisations, this approach unlocks the potential of the long-term observational record for the quantitative study of middle-atmosphere turbulence.