| 20 Aug 2026
Changes in temperature sensitivity of soil respiration with decadal warming
Abstract. Global warming may enhance soil organic carbon (SOC) turnover and CO2 release, potentially accelerating further warming in a positive feedback loop. The temperature sensitivity (TS) of soil respiration is a key determinant for the climate feedback of soils. However, it is uncertain how TS may change under continued warming due to potential microbial thermal acclimation and changes in substrate quality and availability. We examined TS in an incubation experiment using intact soil cores that had undergone 13 years (decadal) of in situ geothermal warming spanning a broad warming gradient of +0 °C to +75 °C.
The observed loss of SOC stock (medians ± SE, kg C m-2) compared to ambient had increased with decadal warming, from -4.9 ± 2.4 % at +1–3 °C to -19.3 ± 1.5 % at +7–10 °C. Across incubation temperatures from 5 to 35 °C in the lab experiment, area-based TS generally declined with decadal warming intensity while SOC-normalized TS was similar across treatments. This suggests that SOC depletion rather than microbial thermal acclimation was the dominant control for reduced respiration rates with warming. TS showed variation with temperature range and was generally lower between the 5–15 °C and 25–35 °C incubation temperature changes and markedly higher in the 15–25 °C change. Best-fit standard field models for soil respiration (Van’t Hoff and Lloyd & Taylor models, R2 = 0.842–0.929) failed to capture these shifts in TS, while a sigmoidal function closely captured the observed TS dynamics (R² = 0.970–0.998). Our study suggests that using fixed Q10 TS values to upscale C budgets is inadequate. Although area-based respiration rates declined with warming, the observed decline in soil respiration with warming suggests a latent reduction of further SOC loss over time following prolonged exposure to warming.
This manuscript sets out to determine how the temperature sensitivity of soil respiration is affected by decadal warming, using intact soil cores collected along a geothermal gradient in Iceland and incubated at four temperatures. Long-term warming manipulations rarely exceed +6 °C above ambient and rarely run beyond a few years. The present dataset is thus extremely valuable and has the potential to address relevant and timely questions. The text is generally well written and the figures are clear.
My concerns are with the inferences drawn from the data rather than with the data themselves, and I do not think that the interpretations and conclusions stand as written.
The first concern is about roots. The cores contain roots, with root mass declining across the decadal warming gradient (Fig S1), so the measured flux is a mixture of autotrophic and heterotrophic sources in proportions that differ among treatments. At ambient and low warming, the root contribution to both the flux and its temperature sensitivity is potentially substantial, while at high and hotspot levels the flux is likely almost entirely heterotrophic with the observed sensitivity being dominated by heterotrophic sensitivity. Any change across the decadal warming levels could reflect this shift in source composition rather than anything about microbial adaptation or substrate depletion, and the conclusions (for example L434-435 and L486-488) therefore cannot be reached. It is true that three months of cold storage would have perhaps killed the roots, but that does not resolve the problem because dying root material is then a labile carbon input that also varies systematically with treatment. The authors do have the data to address this: root mass per core is reported in Fig. S1 and could be used as a covariate, or to express fluxes on a root-free basis (assuming constant root respiration per root biomass), or at the very least to reflect on the possible autotrophic contribution.
The second concern is about the normalization of respiration rates by SOC. Respiration per unit SOC is the product of microbial biomass per unit SOC and respiration per unit microbial biomass. A decline in respiration per unit biomass, which is what downregulation would imply, could have occurred while being masked by a compensating increase in biomass per unit SOC. The inference implies that microbial biomass per unit SOC would have been conserved across a gradient spanning 13 years and +0 to +75 °C, which seems like a questionable assumption that should explicitly be treated as such. Walker et al. (amongst others) is invoked, but they did measure microbial biomass, and the study concerns warming of at most +6 °C. The absence of microbial measurements is acknowledged in the limitations, but the conclusions elsewhere are written as though it is not a limitation. This concern is not independent of the choice of equations used to describe the temperature response, which is my third concern.
The choice of equation used to describe the temperature response is very consequential for interpretation, notably in relation to shifts in sensitivity along incubation temperature gradients. The current framing of the temperature response literature around Van't Hoff and Lloyd &Taylor omits a substantial body of directly relevant work on thermodynamically interpretable response functions, on how sensitivity changes with short-term temperature changes, on thermal acclimation including at geothermal sites, and also on partitioning the contribution of root and heterotrophic components to sensitivity (many of them from groups in New-Zealand: Kirschbaum, Schipper, Alster, Liang, Moinet).
In the present manuscript, each of the three functions does have a parameter defining the magnitude of fluxes and one defining the shape of the response. Their interpretation, however, vary, and the equations are not interchangeable. Strictly speaking, the relative sensitivity to incubation temperature can be defined as the first derivative divided by R. Q10 can also be determined mathematically from each equation. These formulations impose a relationship between incubation temperature and relative sensitivity (or Q10). Equation 6 leads indeed to a flat Q10 while Lloyd & Taylor leads to declining Q10 with temperature. This confirms the results on the modelled Q10. The surprising result comes from the sigmoidal equation. This equation is able to capture the observed peak in Q10 at intermediate temperatures. However, a mechanistic equation able to capture a sigmoidal response with a temperature optimum, MMRT, imposes a declining Q10 (or relative sensitivity) with temperature, with Q10 values often in excess of 4 at low temperature. The authors insist on the observed peak in Q10 at 15-25 interval, but the exceptional result here is the low Q10 at the low interval (particularly relevant for soils in subarctic regions with low annual temperatures). Q10 of just above 1 over 5-15oC is very unusual. So the question is why these soils appear almost temperature insensitive at low temperature, and why this departs form mechanistically grounded theories describing a sigmoidal response of R to temperature but predicting a monotonous decline in Q10. Two features of the design deserve scrutiny in this respect. All cores went through the same ascending sequence with no reciprocal subset or descending sequence, so incubation temperature is confounded with time. And cores were saturated and drained to field capacity just before a 24 h preincubation at 5 °C after three months of storage. It is possible that cores were pre-adapted to the low temperature which is the first in sequence, or that a flush similar to a birch effect would have occurred, in both cases leading to an inflated rate a 5oC and ultimately leading to lower sensitivity over 5-15oC compared to 15-25. I understand the incubation cannot be run again, but these possibilities need to be raised.
Finally, I have a concern on the statistics. Three sampling locations were available for ambient, low and mid, two for mid-high and high, and one for the hotspot, with five cores collected within 10 cm of each other at each location. The 70 cores are therefore not 70 independent observations of warming level: location is a grouping factor and should be treated as a random effect. Also, each core was measured at all four incubation temperatures, so the flux data are repeated measures on individual cores, and this is not accounted for either. The main analyses should be repeated in mixed effect models.
Smaller points. Throughout the manuscript it is frequently unclear whether "warming" refers to the decadal gradient or to the incubation temperature. Consistent terminology would help. Finally, in addition to missing literature, some papers are (partly) mis-cited. Teh example of Walker et al. above is a minor one. Another more obvious one is Crowther et al. 2016 (L457), who does not mention priming at all. Please check carefully what can be attributed to the cited papers.