Soil moisture affects microbial decay of SOM and rhizosphere respiration (RR) in temperate forest soils, but isolating the response of soil respiration (SR) to summer drought and subsequent wetting is difficult because moisture changes are often confounded with temperature variation. We distinguished between temperature and moisture effects by simulation of prolonged soil droughts in a mixed deciduous forest at the Harvard Forest, Massachusetts. Roofs constructed over triplicate 5 × 5 m2 plots excluded throughfall water during the summers of 2001 (168 mm) and 2002 (344 mm), while adjacent control plots received ambient throughfall and the same natural temperature regime. In 2003, throughfall was not excluded to assess the response of SR under natural weather conditions after two prolonged summer droughts. Throughfall exclusion significantly decreased mean SR rate by 53 mg C m−2 h−1 over 84 days in 2001, and by 68 mg C m−2 h−1 over 126 days in 2002, representing 10–30% of annual SR in this forest and 35–75% of annual net ecosystem exchange (NEE) of C. The differences in SR were best explained by differences in gravimetric water content in the Oi horizon (r2=0.69) and the Oe/Oa horizon (r2=0.60). Volumetric water content of the A horizon was not significantly affected by throughfall exclusion. The radiocarbon signature of soil CO2 efflux and of CO2 respired during incubations of O horizon, A horizon and living roots allowed partitioning of SR into contributions from young C substrate (including RR) and from decomposition of older SOM. RR (root respiration and microbial respiration of young substrates in the rhizosphere) made up 43–71% of the total C respired in the control plots and 41–80% in the exclusion plots, and tended to increase with drought. An exception to this trend was an interesting increase in CO2 efflux of radiocarbon-rich substrates during a period of abundant growth of mushrooms.
Our results suggest that prolonged summer droughts decrease primarily heterotrophic respiration in the O horizon, which could cause increases in the storage of soil organic carbon in this forest. However, the C stored during two summers of simulated drought was only partly released as increased respiration during the following summer of natural throughfall. We do not know if this soil C sink during drought is transient or long lasting. In any case, differential decomposition of the O horizon caused by interannual variation of precipitation probably contributes significantly to observed interannual variation of NEE in temperate forests.