Soil temperature vs air temperature: temporospatial variance in estimating soil respiration

Background/Question/Methods

Global soil respiration models use air temperature to estimate soil temperature, a key factor which determines soil respiration. However, air temperature may be a poor proxy for soil temperature in different regions and at different times of the year, thus biasing respiration estimates.  Our objective was to compare the air temperature with soil temperature cross different times and spatial scales for the continental United States.  We utilized publically available data from Ameriflux, a soil respiration database (SRDB), and related published papers and applied single linear regression (SLR), as well as  segmented and multiple linear regression (MLR) to analyze the temporospatial variance of soil temperature vs air temperature relationship.   

Results/Conclusions

SLR analysis result showed that over all the sites, daily average air temperature can explain 83.57% variance of daily average soil temperature (R²=0.8357), slope  =  0.76. More interesting, however, a segmented regression improved the explained variation from 0.8357 to 0.8724.  We observed a significant breakpoint when air temperature = 0.54 ℃. When air temperature is less than 0.54 ℃, the relationship between soil temperature and air temperature is much weaker (R²=0.12), slope=0.188. When air temperature is above 0.54 ℃, air temperature explains more soil temperature variance (R²=0.8183), and slope=0.903. In addition, MLR analysis shows that vegetation type, biome type and climate type also affect the relationship between soil temperature and air temperature.  In sum, our results indicate that using air temperature as a predictor will underestimate soil respiration during cold periods in colder regions and may lead to an underestimation of soil respirations contribution to global carbon cycles.