The short-term pulse of CO2 following the re-wetting of dried soils (Robertson et al., 1999; Franzluebbers et al., 2000) has been widely used as an indicator of soil health because of its strong relationship with soil organic C, particulate organic C, microbial biomass C, and cumulative nitrogen and C mineralization over longer periods (e.g., 24 d; Franzluebbers et al., 2000). This method is derived from the “Birch effect”, whereby re-wetted dry soils release a pulse of CO2 resulting from increased microbial activity (Birch, 1958). Drought stress drives microbial communities to dormancy or death (Borken and Matzner, 2009), and re-wetting stimulates C mineralization (Kim et al., 2012).

Several mechanisms could explain the Birch effect, reviewed by Jarvis et al. (2007), among others. Briefly, these include the following: (1) drying and wetting destroy soil aggregates, thus making previously inaccessible organic substrates available (Denef et al., 2001; Homyak et al., 2018); (2) microbes killed after drying are decomposed upon re-wetting (Blazewicz et al., 2014, 2020); (3) microbes release solutes to avoid bursting in response to osmotic stress caused by re-wetting (Schimel et al., 2007); and (4) microbial populations and their activity rebound in response to re-wetting (e.g., Barnard et al., 2013). Recent studies suggest that both cellular and extracellular C are likely to contribute to the larger CO2 pulse following re-wetting (Slessarev et al., 2020; Warren, 2020), implying that multiple mechanisms could be important.

Although the short-term pulse of CO2 following the re-wetting of dry soils is a widely used method in soil health assessments (e.g., Culman et al., 2013; Ladoni et al., 2016; Morrow et al., 2016; Sprunger and Robertson, 2018), there may be potential bias introduced by assaying soils of different moisture contents at the time of sampling. Soils that differ in moisture will dry down at different rates, potentially losing different amounts of C prior to the start of the assay. If sufficiently large, differential pre-assay losses could complicate soil health comparisons across field treatments, landscape catenas, or different time points within the same soil.

Here we investigate the influence of initial soil moisture levels on pre-assay CO2 release during drying for an Alfisol soil in the upper Midwest, USA. We test the hypothesis that moister soil will have higher pre-assay CO2 loss because a longer dry-down period results in more time for such losses to occur.

As expected, soil in the 50 % and 70 % WFPS treatments took longer to dry than did soil in the 30 % WFPS treatment (Fig. 1a). A day after wetting, soil from the 30 % WFPS treatment was completely air-dry, but soil in the 50 % and 70 % WFPS treatments had lost only 79 % and 68 % of their initial moisture, respectively. All soil was air-dry by 3 d after wetting. Pre-assay CO2 losses mirrored soil moisture loss, reaching zero for all WFPS treatments by day 3 (Fig. 1b). Both GSM at the time of sampling and day had effects on pre-assay CO2 loss rates (P<0.0001), but initial WFPS treatment did not (P=0.28), probably because GSM captures more variation in soil moisture than WFPS treatment as the soil dries. However, there was an interaction between treatment and Day (P=0.0005). Soil of even the lowest initial WFPS treatment lost C as CO2 over 3 d of drying (26 µg CO2-C g-1 soil for 30 % WFPS), but losses were disproportionately higher from wetter soil (62 and 71 µg CO2-C g-1 soil for 50 % and 70 % WFPS, respectively).

Initial soil moisture (i.e., WFPS treatment) had a significant effect on 24 h CO2 pulses after re-wetting air-dried soil (P=0.007; Fig. 2). While final CO2 pulses were lower for the 50 % and 70 % WFPS treatments relative to 30 % WFPS (Fig. 2), the 50 % and 70 % WFPS treatments also tended to have greater pre-assay CO2 losses during 3 d of dry-down, which represented 77 % and 95 % of their 24 h CO2 pulses, respectively. After accounting for these losses with correction factors, the 24 h CO2 pulses were similar across initial WFPS treatments (P=0.28; Fig. 3).

Initial soil moisture levels significantly affected the results of the conventional 24 h CO2 pulse assay, thus calling into question its reliability as a soil health indicator (Fig. 2). Wetter soil lost more C during dry-down, presumably because soil microbes remained active for a longer period of time. Without knowledge of pre-assay CO2 losses, one might erroneously conclude that soil from the 30 % WFPS treatment had about a 35 % greater CO2 pulse than the others (Fig. 2), but this trend is instead due to higher pre-assay CO2 losses during the dry-down period for wetter soil (Fig. 3). It is striking that even short drying intervals (i.e., 1 versus 3 d) can affect soil health interpretations as deduced from the 24 h CO2 pulse after re-wetting air-dried soil. However, we were able to account for the pre-assay CO2 losses for our soil with a correction factor that made 24 h CO2 pulses approximately equivalent across all initial WFPS treatments. It is unlikely that any mechanism other than dry-down loss in wetter soil is responsible for the treatment differences we observed. First, soil samples were treated identically except for initial water content. Were potential mechanisms behind the Birch effect responsible, wetter soils should have released higher CO2 pulses upon re-wetting due to greater microbial activity (e.g., Linn and Doran, 1984), more microbial biomass (e.g., Franzluebbers, 1999), or a greater release of osmolytes due to the increased risk of lysis. However, the 30 % WFPS treatment had the largest CO2 pulse upon re-wetting (Fig. 2), so dry-down loss in wetter soils is the most plausible explanation (Fig. 1).

These trends suggest that efforts to characterize soil health via short-term CO2 pulses following the re-wetting of dry soil should exercise caution if comparisons involve soils with a range of initial soil moistures. This includes soils compared across seasons; across drought, precipitation, or irrigation gradients; across landscape catenas; across crop, grazing, or forest management practices; and as well in cross-site comparisons and meta-analyses that include soils collected at different initial soil moistures.

A correction factor that accounts for pre-assay CO2 losses may help to normalize such comparisons. In our soil, pre-assay CO2 losses led to a C mineralization bias as high as 32 %, for which we could confidently correct by applying a correction factor based on measured rates of pre-assay CO2 loss (Eq. 3). However, we acknowledge that our use of a correction factor is not intended to imply that there is a fixed, available C pool, but rather to demonstrate the way antecedent moisture conditions can affect soil health tests and to offer a potential solution. Other soils with moisture contents sufficient for microbes to oxidize available C during dry-down will require different correction factors as the C quality and microbial communities could change by soil type as well as within and across seasons (Groffman et al., 1996; Wuest, 2014). A soil-specific correction factor can be calculated by measuring CO2 loss during dry-down on a subset of samples, as we described above (Eq. 3).

An alternate solution is to minimize the dry-down period such that little C is lost prior to the assay. Strategies to minimize pre-assay CO2 loss include exposing soils to temperatures high enough to speed evaporation but low enough to avoid sterilization (Jager, 1968) or otherwise artificially disrupt the microbial community (Evans and Wallenstein, 2012). This could be performed in a closed vented chamber such as a soil incubator. Alternatively, faster and more even drying might be achieved with a steady flow of air (i.e., a fan or vented system) over exposed soil samples.

Overall, our results demonstrate that using the 24 h CO2 pulse following the re-wetting of a dried soil to evaluate soil health can be misleading for soils with different moisture contents at time of sampling. For such soils, rapid drying methods and/or a correction factor based on pre-assay CO2 losses should be considered.