Undisturbed mires are wetland biomes where peat accumulates, typically at rates of ∼1 mm year⁻¹ over centuries (Parish et al., 2008), making these ecosystems one of the largest global organic carbon (C) reserves with substantial impact on atmospheric carbon dioxide (CO₂) concentrations (Moomaw et al., 2018). In Denmark, wetlands with >6% organic C cover about 291,000 ha, of which 59% are used for agriculture (Greve et al., 2021). The massive losses of C from these agroecosystems are controlled by the balance between current net C inputs and peat mineralization, which is substantial and largely depends on the drainage conditions (Straková et al., 2012). National emission factors for drained organic agricultural soils in Denmark were established by empirical gas flux measurements in 2008–2009 and averaged 35 Mg CO₂ ha⁻¹ year⁻¹ across eight sites in crop rotation and with permanent grass (Elsgaard et al., 2012). Climate change mitigation by rewetting of agricultural soils with >6% organic C is currently supported by national governmental incentives. Following this, it is envisaged that an area of 88,500 ha potentially can be rewetted and converted to permanent natural grassland (Ministry of Food, Agriculture and Fisheries of Denmark, 2021). Whereas reductions in CO₂ emissions from slower peat mineralization are well documented in relation to increasing groundwater tables (Renou-Wilson et al., 2014), there is an unmet challenge in documenting the net C sequestration from new plant biomass on wet organic soils. This is in particular true for wetlands with >12% organic C and cultivated with perennial grasses, also known as paludiculture, which may contribute to greenhouse gas (GHG) mitigation (Tanneberger et al., 2020) and nutrient retention (Giannini et al., 2017; Vroom et al., 2018).

The input and cycling of organic C in soil ecosystems is highly affected by plant mechanisms regulating the development of aboveground (AGB) and belowground (BGB) biomass (i.e., shoots and roots, respectively) and consequently the quantity of litter input (Kumar et al., 2017), while decomposition of soil organic matter (SOM) is affected by soil nutrient stoichiometry (Kumar et al., 2021). Factors controlling AGB production of perennial grasses are well studied, but little is known about BGB, in particular for flood-tolerant perennial grasses. Roots play a significant role in the soil C cycle (Puget and Drinkwater, 2001; Moore et al., 2019; Dijkstra et al., 2020), indicate productivity (Thakur et al., 2021) and are crucial for the buildup of SOM on both mineral soils and peatlands (Klingenfuß et al., 2014; Leifeld et al., 2020). Not only root biomass but also in particular root exudates, secretions, lysates, cap cells, and mucilages (Carminati and Vetterlein, 2013; van Veelen et al., 2018) are important C inputs affecting the soil status of being either a source or a sink of C. For the estimation and modeling of changes in soil C stocks, a fixed default root:shoot ratio (R/S) is widely used to account for total biomass C. However, R/S is known to vary as a result of multiple environmental and climatic factors as well as management (Kibet et al., 2016; Sainju et al., 2017a; Hu et al., 2018). The optimal partitioning theory (OPT) of plant biomass allocation between AGB and BGB proposes that environmental factors will force plants to allocate new biomass to those parts needed to secure the most deficient resources for optimal plant growth (Fraser et al., 2015; Yang et al., 2018). In contrast, the isometric allocation hypothesis (IA) states that BGB scales linearly with AGB, independent of abiotic factors. Further, it has been stated that defoliation of AGB by harvest or grazing will decrease total BGB (Reid et al., 2015). However, due to the high on-site variability and the challenge of root extraction, in particular for perennial grasses, accurate estimations of R/S under different conditions are rare (Bolinder et al., 2002). Instead, and notably for grassland ecosystems, the allometric approach, using a fixed R/S (Bolinder et al., 2007), is used for modeling of BGB soil C inputs. Nevertheless, recent research highlighted the potential overestimation as well as uncertainty of this modeling approach (e.g., Mokany et al., 2006; Taghizadeh-Toosi et al., 2016; Keel et al., 2017).

While currently an effort is made to review R/S for different biomes and climate zones (e.g., Qi et al., 2019), an assessment of the R/S of grasses under different harvest frequencies is still lacking. This is particularly true for flood-tolerant grass species, which are increasingly introduced on both wetland and upland soils for both climate change mitigation and added-value products, such as grass protein as a substitute for soy (Nielsen et al., 2021). Hence, there is a need for consolidated estimates of R/S for commonly used paludiculture crops under different harvest frequencies (Karki et al., 2014). In the present study, we addressed this need in an annual trial and hypothesized that different R/S would be observed in flood-tolerant perennial grasses by manipulating the harvest frequency during the growth season under provision of adequate nutrient availability. The specific aims of the study were 1) to determine the effect of harvest frequency on the root and shoot development in the first year of cultivation of the perennial grasses reed canary grass (RCG; Phalaris arundinacea L.), tall fescue (TF; Festuca arundinacea Schreb.) and festulolium (FL; Festuca spp. × Lolium spp.), 2) to assess species-specific differences in R/S biomass ratios, 3) to determine the effect of harvest frequency on the C and nitrogen content in above- and belowground biomass, and 4) to assess the implications of R/S for soil C modeling.

In this study, we highlight that AGB and BGB as well as R/S differed greatly among the various harvest frequencies, with frequent cuts resulting in reduced BGB yields and lower R/S for all assessed species. However, while the effects of water saturation and nutrient availability on biomass development and the R/S are relatively well known (e.g. Guo et al., 2016), there are only little comparable data available regarding R/S for RCG, TF, and FL under various annual cuts within the first year of establishment. Mander et al. (2012) reported an R/S of 0.91 (unfertilized) and 0.67 (fertilized) for RCG on an abandoned peat extraction site in Estonia without harvest, while Klimesová (1994) found a R/S of between 1.9 and 2.1 for RCG in a pot experiment under similar soil and water conditions and for the same timeframe as in this study. The latter values are similar to the R/S of 2.3 for the RCG treatment without harvest in our study. Bolinder et al. (2002) reported for RCG and TF in the second year after cultivation R/S values of 1.0 and 0.6 for a treatment with two annual cuts. These values, 0.2 and 0.4 lower than the corresponding R/S from RCG and TF under two annual cuts observed in our study, are within a similar range. However, the higher R/S observed in our study probably results from a younger sward age, indicating the plant’s need for optimal biomass allocation under the establishing growth period. Cougnon et al. (2013) reported 1 kg DM m⁻² more BGB for TF with five annual cuts than in our experiment for an already established sward, receiving 100 kg N ha⁻¹ and 100 kg K ha⁻¹ more than in this study, and 300 (FL)–700 (TF) g DM m⁻² more BGB in 3-year-old swards with five annual cuts, receiving similar fertilization amounts as in our study (Cougnon et al., 2017). We also found such differences in R/S for treatments without any harvest. For instance, Xiong et al. (2009) determined an R/S of 6.5 for RCG after a full year of growth, which differs significantly to our observed value of 2.3 after 210 days. Since the water table has been permanently controlled to –20 cm, a depth indicated as optimal for AGB development of flooding-tolerant perennial grasses (Miller and Zedler, 2003; Ustak et al., 2019), and adequate nutrients were provided, we interpret the observed differences in R/S for all species regarding harvest frequencies as a response of the plant’s biomass allocation. This is in accordance with the OPT (Kobe et al., 2010), where, as a consequence of more frequent harvest and removal of biomass involved in light energy capturing, more biomass is allocated to AGB organs in order to maximize photosynthesis. Further, the IA was supported by significant linear relationships between AGB and BGB, which is in line with other studies for temperate grasslands (e.g., Wang et al., 2010; Yang et al., 2018), indicating a coexistence of the OPT and IA theories. However, even though we were able to confirm our hypothesis, that different ratios between AGB and BGB will be observed in flood-tolerant perennial grasses under different harvest frequencies during the growth season under provision of adequate nutrient availability, there might remain a limitation of our observation for wet organic soils: that the R/S analysis was not performed on samples grown on peat soil cores. However, since R/S is rather affected by nutrient availability than soil types (Lambert et al., 2014; Pinno et al., 2014; Lehtonen et al., 2016), we assume a reliable validity of the indicated R/S for similar growth conditions, including fertilizer management, also on peat soils. This is supported by other research, comparing differences in R/S for certain species and treatments between cultivation on mineral and organic soil types. Xiong et al. (2009) for example found similar R/S for RCG under different fertilization rates on both mineral and organic soils. Björk et al. (2007) reported similar findings for root mass in Swedish heath and meadow tundra ecosystems, and Lambert et al. (2014) stated that R/S of Arundo donax was stable across soil types. Nonetheless, we propose that future research not only has to emphasize an assessment of cultivation treatments on R/S on mineral and organic soils but also needs to monitor TC and TN in AGB and BGB, as well as their development with increasing production years, due to expected variations (Sainju et al., 2017b). The C/N ratio can aid as an indicator for litter quality, with easier decomposable substrates having low C/N ratios (Rydin and Jeglum, 2013). Recalcitrant plant litter, indicated by high C/N ratios (Poirier et al., 2018), hence has the potential to increase soil C input-either directly to the pool of particulate organic matter (POM) or as microbial necromass following decomposition by microorganisms in the acrotelm (Worrall et al., 2017; Rossi et al., 2020). The potential for long-term C storage in the POM pool is in particular high for wet organic soils due to anoxic conditions in the catotelm, indicated by higher contents of lignin with increasing peat depth (Williams and Yavitt, 2003). Further, in soils with <12% SOC, additional C storage through the POM pool has the potential to overcome soil C saturation (Cotrufo et al., 2019). However, for drained soils where microbial activity is found in deeper layers, there is no consensus whether the C/N in BGB can be used as a predictor for the decomposability of root litter, and hence the C storage potential, as indicated by interspecific variation (Bonanomi et al., 2021). However, while R/S and C in biomass have been assessed for RCG on mineral and organic soils (Xiong et al., 2009; Xiong and Kätterer, 2010), this study was to our knowledge among the first to, besides R/S, also assess C/N in AGB and BGB parts.

The input of carbon into soil is a critical component of the global C cycle, thus significantly contributing to various aspects of ecosystem functioning. Modeling existing and changing soil organic carbon (SOC) stocks hence is a critical task in decision support related to optimal climate and land management (Taghizadeh-Toosi et al., 2016). However, while it is recognized that C allocation in belowground plant organs depends on vegetation type (Keller and Phillips, 2019), growing conditions (Whitehead, 2020), and plant development stages and management (Pausch and Kuzyakov, 2018), only few of these complex relationships (Cheng et al., 2014) are set into context with soil geochemistry and accounted for during SOC modeling (Finke et al., 2019). For instance, previously reported R/S for perennial grasses showed a broad distribution of median values, which has been shown in reviews by Bolinder et al. (2007) and Pausch and Kuzyakov (2018). Given the critical role of BGB carbon input, our study highlights the importance of acquiring more extensive knowledge regarding the development of perennial grass root systems depending on AGB manipulation by harvest. We only found marginally significant differences in R/S for the various treatments and species, apart for the, in agriculturally used grasslands, uncommon zero-cut strategy, which could indicate that a fixed R/S for perennial grasses might be reasonable to use if accurately defined.

However, our calculation of soil C input for perennial grasses under different harvest frequencies revealed significant differences for the various management options and applied R/S. This highlighted a potential risk for over- and underestimation of the C sink functioning of wetlands and grassland ecosystems. While, for instance, using the IPCC default R/S of 2.8 resulted in an estimated annual carbon input into soil of 9.6 t C ha⁻¹ year⁻¹ for a RCG yield of 10 t DM ha⁻¹ year⁻¹, our results ranged between 2.8 and 7.3 t C ha⁻¹ year⁻¹ for the same AGB yield, depending on annual harvest frequencies. Since, for instance, the default IPCC (2006) estimate of R/S of 2.8 is applied in the Danish National Inventory Report (2020), the discrepancy of TC input from BGB resulting from varying R/S might have far-reaching consequences for policymaking. For instance, depending on whether management measures are extensive, e.g., designated nature areas without any biomass manipulation, or intensive, e.g., biomass harvest up to five times annually, the choice of R/S for the quantification of an organic soil C sink function must be made carefully and adapted to the ecosystem in question. This is in particular true for the designation of rewetting measures on wetland areas, including the choice of land use and land cover, for climate considerations. In the context of optimum grassland management for agricultural production, we showed that a strategy with two annual cuts has the highest potential to contribute to SOC buildup. We hence advocate an inclusion of more accurate R/S for modeling, taking differences resulting from grass management, as well as site-specific climatic and biogeochemical conditions (Sahoo et al., 2021), into consideration to reduce uncertainties for policymaking on agroecosystems.

However, a limitation of this study is the assessment of harvest frequency on R/S during the establishment year only, which emphasizes the need for more long-term data on root growth of managed and unmanaged perennial grasses in wet environments. Hence, we suggest a future multi-annual study, with at least 2 years of trial, to increase the validity of results by accounting for the plants long-term response to management, taking interannual climatic variability into consideration. Previous studies demonstrated that R/S varies with ley age (Bolinder et al., 2002; Acharya et al., 2012; Huang et al., 2021), the associated interannual climatic variability (Poorter et al., 2012; Li et al., 2021), and nutrient availability (Cong et al., 2019); hence, further years of experimental analysis have to determine whether our findings regarding the effect of harvest frequency on R/S and the associated implications of management for SOC input are applicable on the long term. Further, it is yet unclear how the determined R/S for the assessed grasses can be applied in the evaluation of SOC input by plant biomass in wet or rewetted agricultural wetlands. For wet organic soils and in the context of paludiculture, we suggest that not only root growth but also rooting depth and specific root turnover rates have to be assessed. Only few studies assessed the maximum potential rooting depth by flood-tolerant perennial grasses in correlation with the WTD profile on organic soils, as compiled by Fan et al. (2017). Houde et al. (2020) highlighted the significance of turnover of perennial grass roots for the C storage potential, and Schwieger et al. (2020) emphasized the significance of roots as the main peat-forming component in grass-covered fen peatlands. However, while wetland ecosystems potentially have the second-highest root turnover rates of all ecosystems (Gill and Jackson, 2000), the complexity between rooting depth, WTD, C/N, and litter recalcitrance (Shurpali et al., 2010; Straková et al., 2012; Leifeld et al., 2015; D’Imperio et al., 2018) as well as AGB manipulation and root turnover rates for soil C input still needs to be defined.

With an increasing policy focus on wetland restoration, including paludiculture, for GHG mitigation and nutrient retention, as well as the concomitant need to point out sites with the highest mitigation potential, not only assessments of GHG emissions and optimum management (e.g., Geurts et al., 2019; Tanneberger et al., 2020; Nielsen et al., 2021) but also site- and plant community-specific BGB NPP should be taken into consideration for an accurate evaluation of the C sink potential.

While it is known that C inputs in soils are affected by development of AGB and BGB, as well as a variety of biotic and abiotic factors, little has been known so far on the effect of harvest frequency on the R/S of perennial grasses. In conclusion, this study found significant differences in the R/S for the flood-tolerant perennial grasses Phalaris ar., Festuca ar., and Festuca × Lolium, affected by annual harvest frequencies of AGB, with less biomass allocated to belowground parts with increasing number of cuts. No species-specific differences in the ratio were observed for any treatments. In addition, our results showed that both the OPT of plant biomass allocation and the IA hypothesis seemed to coexisted. Further, we demonstrated the importance to accurately define R/S for the calculation of carbon input into soils to avoid significant over- or underestimation. Our results showed that there are significant differences regarding the annual carbon input into soils, depending on the R/S applied. We found that using the IPCC default factor for R/S of 2.8, applied for both managed and unmanaged grasslands, resulted in 55%–71% higher carbon input rates as compared to our scenarios with two and five annual cuts, commonly applied in agricultural systems. This discrepancy indicates a significant inaccuracy for modeling and quantification of the C sink or source function of wet organic grassland areas, which might have far-reaching consequences for policymaking and carbon accounting. Further, we not only demonstrated how measurements of AGB and BGB provided a more accurate baseline for estimation of soil carbon input but also indicated the need for further assessment of R/S and C/N of perennial grasses, particularly for those cultivated on wet organic soils, to define the soil C sink capacity.