Abstract: Accurate prediction of soil respiration (Rs) under climate change requires detailed knowledge of its key control factors. However, interactive effects of multiple biophysical factors on long-term seasonal Rs are uncertain due to a paucity of direct observations. Here, we developed a new Projection to Latent Structures (PLS) approach, i.e., PLS-lag, by representing time-lag and nonlinearity between variables to explore the impacts of potential biophysical variables on Rs in growing and non-growing seasons in a temperate deciduous forest. Results show that biophysical factors explained Rs dynamics better at night than in daytime during growing season, with stronger relationship between nocturnal Rs and temperature (including air and soil temperature). Soil temperature at 10 cm depth (STP10) was the dominant factor controlling Rs during both growing and non-growing seasons. We also find that photosynthetic active radiation (PAR, with about a six-day time-lag), leaf area index (LAI) and soil moisture at 10 cm depth (SWC10) modulated Rs during growing season only. The PLS-lag approach significantly improved the explanatory power by 13–19% for Rs during growing season. Our findings highlight the necessity to account for time-lag and nonlinearity in exploring the mechanisms underlying Rs for a more robust understanding of soil carbon cycling and its response to climate and environmental changes.
