A precise prediction of the response in soil respiration (R_s) to climate warming requires the temperature-driven portion of the response to be characterized separately from the global, undecomposed response. Eliminating non-temperature-driven effects, which tend to confound the response in R_s, is possible with Singular Spectrum Analysis (SSA), a non-parametric approach to timeseries analysis. R_s and ancillary data were collected at three microsites in a desert-shrubland ecosystem in northwest China over a three-year period (i.e., 2013–2015). The microsites include desert-sites with non-biocrusted, lichen-crusted, and moss-crusted soil surfaces, or NC, LC, and MC, respectively. Results revealed that temperature sensitivity in R_s (i.e., Q₁₀) based on the original, unprocessed R_s-data and a seasonal equation relating R_s to temperature were similar among microsites, with similar amplitudes in R_s. High-frequency bins of R_s, with sub-signals with a period < 90 days, obtained by subtracting the low-frequency bins from timeseries of R_s, were positively correlated with soil moisture (θ) at all microsites. Whereas, the same high-frequency bins were individually decoupled from soil temperature at LC and MC. This suggests that the influence of θ on the response in R_s to temperature dominated the R_s-measurements over the shorter term. After eliminating the non-temperature effects on R_s with SSA, Q₁₀ tended to produce lower values, i.e., 1.40, 1.28, and 1.34 for NC, LC and MC, respectively, when compared to cases where non-temperature effects were not removed, yielding a mean of 2.28. Q₁₀-values estimated with SSA-processed R_s-data were consistent with those published at the ecosystem level (i.e., R_e ~ 1.40). The research suggests that SSA-adjusted Q₁₀-values tend to be similar with those at the ecosystem scale. Future theoretical work is anticipated using a convergent temperature-only-based value of Q₁₀ between R_s and R_e across ecosystems.