Summer temperature regulation of interannual variability in the carbon balance in an urban forest in the Megalopolis of Beijing, China

Abstract

Afforestation projects in urban areas of northern China have thrived over the past two to three decades. Urban forests play an important role in sequestrating carbon (C) at local to regional scales. However, understanding the current role of C-sink strength in urban forests is generally lacking. We used continuous eddy covariance and hydrometeorological data over the period 2012–2020 from an urban forest in Beijing, to assess the interannual variability (IAV) in C-fluxes and their key controlling factors. Over the study period, the urban forest was mostly a C-sink (mean net ecosystem exchange (NEE) was -195.28 ± 98.66 g C m-2 yr-1). Interannual variability in NEE was largely due to variation in ecosystem respiration (Re), more so than to gross ecosystem production (GEP). Interannual variation in annual GEP and Re were controlled to a large extent by summer soil water content (SWC). Since the effects of SWC on GEP and Re partially offset their individual responses, IAV in annual NEE was seen as being less influenced by summer SWC. Annual NEE was constrained by changes in summer air temperature (Ta). Residuals based on a Re-to-SWC regression exhibited significant decreases with increasing summer Ta, appreciably more than the residuals based on GEP-to-SWC regressions. This pointed to the importance of summer Ta in impacting annual NEE through the suppression of Re, rather than through GEP. Our analysis further revealed that Re was more strongly affected by drought than GEP. Long-lasting drought, greater than one month, was seen to impact GEP and Re more than short-term drought. Under future climate projections of more frequent heatwaves and prolonged periods of soil water shortages, C-sink-source relationships in urban forests are likely to shift. Our results underscore the need to adequately represent the inhibitive characteristics of temperature on stand-level C-fluxes in ecosystem models for improved prediction.