Soil water availability strongly modulates soil CO2 efflux in different Mediterranean ecosystems: Model calibration using the Bayesian approach

Soil respiration in drought prone regions is highly dependent on the precipitation regime and soil moisture conditions, which are expected to change in a global warming context. In the present study we used an extensive collection of field chamber measurements of soil respiration (R-s) from forest and grassland sites of centre and south of Portugal distributed over a 10 year period. This data were summarized and analysed with the objective to describe seasonal variability of R-s as affected by soil moisture (H-s) and soil temperature (T-s). A Bayesian framework was used to test the effectiveness of soil bioclimatic models in estimating R-s on a daily and monthly time step. R-s seasonality was similar between sites, reaching a maximum in spring and autumn and a minimum in the dry season (July-September). No differences were observed for R-s between sites with different standing biomass or soil carbon stocks either on an annual or seasonal timescale. H-s, and not T-s, was the driving factor of R-s during most of the year. T-s drove R-s response only above certain H-s limits: 10% for forest sites and 15% for grassland sites leading to a Q(10) of 2.01, 1.61 and 1.31 for closed forests, open forests and grasslands, respectively. The Bayesian analysis showed that models using H-s as an independent variable performed better than models driven by T-s alone. Monthly estimates of R-s in grasslands can be predicted by simple climatic models based on H-s but none of them was suitable for forest ecosystems, stressing the need for a process-based approach. This study adds to the evidence that H-s controls R-s fluxes for Mediterranean ecosystems and should always be taken into account for extrapolation purposes