Improving ecosystem respiration estimates for CO2 flux partitioning by discriminating water and temperature controls on above- and below-ground sources

• The choice of driving temperature does bias ecosystem respiration (ER) estimates. • A dual-source module with parameter α is proposed to reduce this uncertainty. • The module significantly improves the performance of selected ER models. • Variation of α is largely controlled by biomass and moisture conditions. Empirical models for estimating ecosystem respiration (ER) are widely used in CO 2 flux partitioning algorithms that partition net ecosystem CO 2 exchange (NEE) into gross primary productivity (GPP) and ER due to advantages of simple structures. However, empirical ER models remain limited due to single-source conceptualization that doesn’t discriminate different responses of aboveground respiration (AGR) and belowground respiration (BGR) to environmental factors (i.e., temperature and/or soil moisture). In this study, a dual-source module with only one parameter α was proposed and incorporated into six widely used ER models to enhance model capabilities. Long-term flux measurements of six typical terrestrial ecosystems and soil chamber respiration data at two sites were collected to evaluate models. Results showed that integration of the dual-source module can significantly improve the performance of empirical models in selected ecosystems with mean R 2 improvement of 0.10 ± 0.16. The site years with relative increased R 2 (Δ R 2 ) larger than 10 % range from 6 % to 79 % amongst different models. Further validation between soil respiration and estimated BGR showed good correlations ( r > 0.7) and demonstrated that proposed method can provide robust estimate of above/belowground respiration. Calibrated α varies amongst ecosystem types. Further analysis indicates variation of α is largely influenced by ratio of above/belowground biomass and annual average moisture conditions. Our findings highlight the critical need for partitioning ER models into dual-source for developing CO 2 flux partitioning algorithms and support the approach as an effective means to enhance the understanding of global carbon cycles with changing climate.