Soil respiration was measured under the canopy of 4 replicate shrub and 4 inter-shrub patches, dominated by invasive Annuals. Shrubs were paired with an inter-shrub patch within 2m and replicate Shrub and Annual patches were 4-5m apart to avoid spatial auto-correlation and ensure random sampling of the ecosystem (Strahm 2005). In each Shrub and Annual patch total respiration (Rs) was partitioned into heterotrophic (Rh) and autotrophic (Ra) with paired root exclusion and non-exclusion treatments immediately adjacent to each other (total of 16 collars). Root exclusion cores consisted of 20cm diameter PVC pipe installed to 30cm depth. Total respiration was measured in 20cm diameter collars installed to 5cm depth to minimize disturbance of vegetation and root structure. All collars were installed following the first rainfall in December 2009 and soils allowed to acclimate before measurements were initiated on 9 February 2010. For each pair of intact and root-exclusion chambers Ra contributions could be estimated as the difference between Rs and Rh (Ra = Rs – Rh).

Soil respiration was measured continuously with an LI-8100 Automated Soil CO₂ flux system (LI-COR) on cycle that re-measured a chamber every four hours. Every flux measurement lasted two minutes with 10 second logging interval. Soil moisture (percent volumetric water content (% VWC), EC-5, Decagon Devices Inc.) and soil temperature (°C, LI-8100 thermistors) at 0-5cm depth inside each collar, and chamber air temperature (°C, LI-8100 thermistors) were measured concurrent with respiration measurements. Soil moisture probes were calibrated using a linear regression with monthly gravimetric water samples (0-15cm) collected near each collar (relationship between gravimetric and volumetric samples, R² = 0.472) (Czarnomski et al. 2005).

Soil respiration data was recorded for one-and-a-half growing seasons, February 2010-June 2010 and September 2010 – July 2011. In September 2010 measurements were resumed slightly before the first rainfall event. Measurements in both years terminated at the end of the growing season after invasive annuals senesced, midday temperatures approached 50°C, further rain was unlikely, and all respiration rates were close to 1μmol CO₂ m^-2s^-1.

Fluxes were estimated with exponential curves fit to each two-minute flux measurement using Licor 8100 FileViewer software (Version 2.0.0, LI-COR Inc. 2004-2006). Flux values with an exponential R² fit less than 0.97 or a coefficient of variation (CV) > 2 were removed (approximately 15% of all measurements). Further quality control was conducted using Matlab (Version 7.4.0 R2007a, The MathWorks Inc. 1984-2007) and fluxes were removed if values in one chamber deviated substantially from correlation with all the other chambers (Savage et al. 2008). For most chambers less than 4% of the data was removed this way, but for one chamber up to 20% of the data was removed. 

Soil moisture, soil temperature and chamber temperature data were also quality controlled and erratic values removed. When soil moisture or temperature was missing due to prolonged sensor failure, data was filled using averages from the 3 other chambers in the same treatment. Gap-filled soil temperature and moisture data were used in statistical analyses.

Soil respiration measurements depend on accurate chamber temperature measurements. In chambers with failed thermistors flux rates were therefore corrected with a correction formula from Licor using the average chamber temperature from the three other chambers in the same vegetation and root exclusion treatment. In one chamber the thermistor never recorded accurate temperatures and all fluxes were corrected this way. Most other chambers only had episodic thermistor failure and approximately 16% of all flux measurements were corrected for bad chamber temperatures.

Daily average fluxes were calculated for Rs and Rh and expressed as µmol CO₂ m^-2s^-1. Ra was calculated as an absolute flux and as a percent of Rs. In July 2011 some chambers developed leaks due to UV degradation of tubing connections, particularly in the Annual patches. Ra was only calculated when there was at least one Rs-Rh pair.