Background <p>Forest ecosystems are in the spotlight for their potential to mitigate anthropogenic carbon dioxide (CO₂) emissions through net photosynthesis. However, this mitigation potential can be counteracted by respiratory losses, e.g., from soils and the forest floor. With global warming, soil respiration (SR) rates are expected to increase, unless acclimation occurs. Using manual and automated chambers as well as a below-canopy eddy-covariance system, we quantified SR and forest floor net CO₂ exchange (NEE_ff) for 13&#xa0;years throughout an 18-year study period (2006–2010, 2015–2016, 2018–2023) in a mixed deciduous forest ecosystem in Switzerland. We identified the contribution of environmental drivers for SR and NEE_ff using Extreme Gradient Boosting models and Shapley additive explanations (SHAP) analyses and assessed the long-term temperature sensitivity of SR and NEE_ff.</p> Results <p>Over the 18-year study period, soil temperature increased significantly and was the main driver of both SR and NEE_ff, explaining over 50% of their temporal variability. Differences in drivers and magnitudes of SR vs. NEE_ff were only found in early spring, when the forest floor vegetation showed net CO₂ uptake. Finally, we found no evidence that SR or NEE_ff (at mean annual temperatures) had increased between 2006 and 2023. Similarly, no significant change in the temperature sensitivity of SR and NEE_ff was observed.</p> Conclusions <p>Combining multiple techniques to assess long-term responses of CO₂ fluxes to environmental conditions with machine learning approaches enhanced our understanding of forest responses to climate change. Moreover, our findings suggest that soil and forest floor respiration already acclimated to warmer conditions, highly relevant for predicting future mitigation potentials of forest ecosystems.</p>