Soil Rh:Rs rose further in 193 SRDB records added after 2017 (mean 0.72 in 2015–2019), but temperature relations in the same data explain about 8% of the 1989–2019 rise

Question: Do soil respiration records that were not available to Bond-Lamberty et al. (2018, Nature) show the same rise in the heterotrophic share of soil respiration (Rh:Rs), and can site warming explain it? Approach: We applied the 2018 filters to the current Soil Respiration Database (SRDB, data update 2025-12-30). This gave 476 records from 183 studies: 283 records that were in the 2017 input file and 193 records entered in 2018-2025, all from studies that were not in that file. We reproduced the 2018 model (n = 318, P = 0.0080; published 0.0087). We then did a replication test (new records for the same years), an out-of-sample test (old-record models predict new records), covariate adjustment with study-clustered errors, within-site change, and a warming attribution with CRU TS 4.10 temperature. Findings: The 2018 inference was optimistic. With one row per record and site-clustered errors, the original trend has P = 0.17. But the new records confirm the rise. For the same years, new records agree with the old trend (mean residual +0.004; 95% study bootstrap -0.064 to +0.065). New records from 2015-2019 have mean Rh:Rs 0.72, which is 0.13 (0.08 to 0.20) above a no-trend prediction. In all records, Rh:Rs increased by 0.068 per decade (95% CI 0.015 to 0.122) after adjustment for partitioning method, land cover, stand stage, mean climate and a China indicator. Rh increased by 3.0% per year, but Ra did not change. Local warming (0.23 C per decade) explains little of the rise: the spatial temperature coefficient implies -1%, and the year-anomaly coefficient implies 8% (-5% to 54%). Sites that warmed faster did not show a clearly steeper rise. Limitations: The SRDB is a compilation of published studies, and its composition changes with study year. Chinese studies are 16% of old and 49-57% of new records, and the trend outside China is uncertain (0.050 per decade, -0.009 to 0.109). Repeated site records are short (median span 2 years). Unrecorded drift in measurement methods cannot be excluded. The rise is a reproducible feature of the database, but its cause remains open.

soil respiration · heterotrophic respiration · Rh:Rs ratio · SRDB · out-of-sample test · replication · carbon-climate feedback · partitioning · CRU TS · meta-analysis database