Accurate projections of climate change impacts on the vast carbon stores of northern peatlands require detailed knowledge of ecosystem respiration (ER) and its heterotrophic (Rh) and autotrophic (Ra) components. Currently, however, standard flux measurement techniques, i.e. eddy covariance and manual chambers, generate empirical ER data during only night- or daytime, respectively, which are extrapolated to the daily scale based on the paradigm that assumes a uniform diel temperature response. Here, using continuous autochamber measurements, we demonstrate a distinct bimodal pattern in diel peatland ER which contrasts the unimodal pattern inherent to the classical assumption. This feature results from divergent temperature dependencies of day- and nighttime ER due to varying contributions from Rh and Ra. We further find that disregarding these bimodal dynamics causes significant bias in ER estimates across multiple temporal scales. This calls for improved process-based understanding of ER to advance our ability to simulate peatland carbon cycle-climate feedbacks.