The δ13C, δ18O and Δ47 records in biogenic, pedogenic and geogenic carbonate types from paleosol-loess sequence and their paleoenvironmental meaning

Abstract Paleoenvironmental reconstructions are commonly based on isotopic signatures of a variety of carbonate types, including rhizoliths and land-snail shells, present in paleosol-loess sequences. However, various carbonate types are formed through distinct biotic and abiotic processes over various periods, and therefore may record diverging environmental information in the same sedimentological layer. Here, we investigate the effects of carbonate type on δ13C, δ18O, and clumped isotope-derived paleotemperature [T(Δ47)] from the Quaternary Nussloch paleosol-loess sequence (Rhine Valley, SW Germany). δ13C, δ18O, and T(Δ47) values of co-occurring rhizoliths (-8.2‰ to -5.8‰, -6.1‰ to -5.9‰, 12–32°C, respectively), loess dolls (-7.0‰, -5.6‰, 23°C), land-snail shells (-8.1‰ to -3.2‰, -4.0‰ to -2.2‰, 12–38°C), earthworm biospheroliths (-11‰, -4.7‰, 8°C), and “bulk” carbonates (-1.9‰ to -0.5‰, -5.6‰ to -5.3‰, 78–120°C) from three sediment layers depend systematically on the carbonate type, admixture from geogenic carbonate, and the duration of formation periods. Based on these findings, we provide a comprehensive summary for the application of the three isotopic proxies of δ13C, δ18O, and Δ47 in biogenic and pedogenic carbonates present in the same sediment layer to reconstruct paleoenvironments (e.g., local vegetation, evaporative conditions, and temperature). We conclude that bulk carbonates in Nussloch loess should be excluded from paleoenvironmental reconstructions. Instead, pedogenic and biogenic carbonates should be used to provide context for interpreting the isotopic signature for detailed site- and time-specific paleoenvironmental information.

More recently, carbonate clumped isotope (Δ 47 ) thermometry has been used in terrestrial paleoclimate reconstructions (e.g., Affek, 2012;Eiler et al., 2014). Δ 47 thermometry is an indicator of carbonate formation temperature independent of δ 18 O water through measure of the abundance of "clumped" carbonate molecules containing both 13 C and 18 O in excess of the abundance that would be expected if heavy isotopes were distributed randomly (Ghosh et al., 2006;Eiler, 2007). Recent applications of clumped isotope thermometry to study carbonates formed in or associated with paleosols and loess show the potential for using independent (paleo)temperature constraints to extract paleoclimate information from these records (Eagle et al., 2013;VanDeVelde et al., 2013;Garzione et al., 2014;Ji et al., 2018;Lechler et al., 2018;Zhai et al., 2019).
Paleosol-loess sequences contain various carbonate types (e.g., loess dolls, nodules, rhizoliths, land-snail shells, and earthworm biospheroliths) that are formed through a variety of processes that occur over different time periods (Zamanian et al., 2016a). All carbonate types that might be found in paleosol-loess sequences can be classified in three groups: (1) geogenic carbonates (i.e., "bulk" loess carbonates), which account for up to 30% of loess material; (2) biogenic carbonates, such as land-snail shells, earthworm biospheroliths, and micro mammal remains such as bones and teeth; and (3) pedogenic carbonates, such as rhizoliths, carbonate pendants, and nodules (Zamanian et al., 2016a). Formation of each carbonate type is governed by the same equilibrium equations: where calcite activity (α CaCO3 ) varies as a function of the concentration of calcium ions in aqueous solution (m Ca 2+ ), the partial pressure of CO 2 in the soil gas ( pCO 2 ), and temperature-sensitive equilibrium constants for the dissociation of carbonic acid (K 1 ), bicarbonate (K 2 ), and calcite (K cal ), and for the hydration of CO 2 (K CO2 ) (Drever, 1982).
For each carbonate type, Equation 1 is driven to the left to cause CaCO 3 precipitation via distinct processes that operate over different time spans and locations (Zamanian et al., 2016a). In the case of biogenic carbonates, environmental conditions that promote (or restrict) biological activity, and growth must be considered. The mean lifespan of land-snails (depending on the species) is 3-5 years (Kerney and Cameron, 1979), and snails are more active at temperatures close to their growth optimum (Zaarur et al., 2011;Yanes et al., 2012). For example, the number of active Arianta arbustorum at 8°C is about five times higher than at 0°C and about two times higher than at 22°C (Cameron, 1970). Pupilla muscorum is active during the whole year, while the activity of Succinella oblonga is limited to the warm seasons of May to July (Pokryszko, 2001). In contrast, some pedogenic carbonates (e.g., rhizoliths) may form over a few decades (Gocke et al., 2011). Pedogenic carbonate nodule and pendant formation takes place over hundreds to thousands of years (Gile et al., 1966;Birkeland, 1999) and occurs seasonally (Breecker et al., 2009), leading to clumped isotope temperatures that presumably reflect long-term averaged, and in many cases seasonally biased, environmental temperatures (Peters et al., 2013;Huntington and Lechler, 2015;Burgener et al., 2016;Gallagher and Sheldon, 2016;Gallagher et al., 2019;Huth et al., 2019;Kelson et al., 2020). The origin and time-integration of geogenic carbonate isotopic compositions may be more difficult to ascertain. Geogenic carbonates found in paleosol-loess sequences may record formation temperatures of minerals in the deflation area from which dust was sourced, where the minerals may have been formed through a variety of geological processes, including metamorphism (Catoni et al., 2012;Baez-Hernandez et al., 2019). Therefore, various carbonate types-even located in the same stratigraphic level in a paleosol-loess sequencemay record a mixture, but specific and potentially complementary environmental information based on differences in δ 13 C, δ 18 O, and Δ 47 values.
We examine paleoenvironmental information recorded by carbonate types from a well-known paleosol-loess sequence, the Upper Pleistocene Nussloch paleosol-loess in the Rhine Valley of southwestern Germany (Zoller et al., 1988;Hatté et al., 1999;Rousseau et al., 2002Rousseau et al., , 2017aAntoine et al., 2009;Prud'homme et al., 2018). Taken together, the δ 18 O, δ 13 C, and Δ 47 data for multiple carbonate types enable us to investigate the effects of geogenic carbonate input, pedogenesis, and environmental factors on resulting isotopic variations.

Stable isotope analysis and calculations
Carbonate samples were ground with a mortar and pestle into homogenized powders prior to stable isotope (δ 18 O, δ 13 C, Δ 47 ) analyses at the University of Washington IsoLab. Sample preparation and analysis followed methods of Burgener et al. (2016) and Schauer et al. (2016). Briefly, 8-20 mg of powdered carbonate was reacted in a common phosphoric acid bath held at 90°C. The released CO 2 was purified cryogenically through an automated nickel and stainless steel vacuum line using an ethanol-dry ice slush trap and liquid N 2 trap, then chromatographically with a helium carrier through a Porapak Q trap (50/80 mesh, 15 cm long, 4.5 mm ID, 0.635 mm OD, held at -20°C). Preparation was completed when the sample CO 2 was frozen into a 6 mm Pyrex tube and flame sealed. The break seal containing purified CO 2 was loaded on the
autosampler of a Thermo MAT 253 dual inlet mass spectrometer to measure mass to charge ratio (m/z) 44-49. The intralaboratory carbonate standards of C64, C2, and Coral (referenced to VPDB using NBS19 and LSVEC for δ 13 C, and NBS19 and NBS18 for δ 18 O), as well as interlaboratory calcite standards ETH1-4 (Meckler et al., 2014;Bernasconi et al., 2018), were purified and analyzed following the same procedures as Nussloch samples. Two to eight replicates were analyzed per carbonate sample, based on available sample mass. Δ 47 , δ 13 C, and δ 18 O values from mass spectrometer data were calculated based on Huntington et al. (2009) using the pressure baseline correction of He et al. (2012), with data reduction and 17 O correction methods and scripts from Schauer et al. (2016). Δ 47 values were placed in the carbon dioxide equilibrium scale, or absolute reference frame (Dennis et al., 2011), using CO 2 gases heated to 1000°C or equilibrated with water at 4°C and 60°C. Peirce outlier tests (Ross, 2003) were applied to identify replicate outliers (4 of 75 Δ 47 analyses [5%] identified as outliers based on either Δ 47 or δ 13 C value), following which mean Δ 47 and standard error (S.E.) of the mean values were calculated. Standard error of the mean Δ 47 value is calculated as the standard deviation (taken to be the standard deviation of the sample replicates or the long-term standard deviation of carbonate standards, whichever is larger) divided by the square root of the number of replicates. T(Δ 47 ) was calculated for mean Δ 47 using the calibration of Petersen et al. (2019), which is based on data produced using the same or similar analytical methods and the same 17 O correction as our samples. δ 13 C and δ 18 O values were placed on the VPDB scale using C64 and C2 for δ 13 C, and C64 and coral for δ 18 O.
To calculate the oxygen isotopic composition of the water from which the carbonate precipitated (δ 18 O water ), the equations in Kim and O'Neil (1997)

RESULTS
Isotopic results for Upper loess, Paleosol, and Lower loess samples are summarized in Table 1, and full data for samples and standards is provided in Supplementary Table A1.  Fig. 2). While all Upper loess rhizoliths have similar δ 18 O values (-5.9 ± 0.1‰ VPDB, ± 1σ, n = 22), δ 13 C values depend on rhizolith size. Small rhizoliths (Fine-rhizo A and B) have average δ 13 C values (-6.1 ± 0.4‰) that are 2‰ higher than Large-rhizo samples from this interval (-7.0‰ and -8.2‰). δ 13 C values for the shell sample (-6.6‰) and pedogenic loess dolls (∼-7.0‰) from the Upper loess fall within the range of rhizoliths from this horizon. Earthworm biospheroliths from the Upper loess are characterized by the lowest δ 13 C for the interval at -11.5 ‰. Measured δ 13 C for bulk loess in the Upper loess interval (-0.5‰) is significantly higher than all other carbonates from the same horizon, a relationship consistently observed in the other sampled horizons.
The Lower loess and Paleosol samples show similar shifts in δ 13 C and δ 18 O values by carbonate type, with the shells having significantly lower δ 13 C and higher δ 18 O values than the bulk loess or bulk paleosol. The three sampled shell species S. oblonga, P. muscorum, and T. hispidus in the Lower loess and the Paleosol have a narrow and similar range of δ 18 O values (-3.8 ± 0.2‰), which is more than 1‰ lower than values for A. arbustorum (-2.3 ± 0.1‰). The δ 13 C values of S. oblonga, P. muscorum, and T. hispidus sampled in both horizons (-3.1‰ to -6.1‰) are systematically offset relative to δ 13 C values of the bulk paleosol (-1.9‰) and bulk loess from the Lower loess interval (-0.9‰) (Fig. 2). δ 13 C values for A. arbustorum are similar across the Paleosol and Lower loess horizons (-8‰). A similar magnitude δ 13 C offset is observed for bulk loess/paleosol carbonate samples from the Lower loess and Paleosol horizons (Fig. 2).
The calculated value of reconstructed δ 18 O water Calculated apparent δ 18 O water values vary among carbonate types (Fig. 4). Apparent δ 18 O water for bulk loess and paleosol

DISCUSSION
Stable isotopic compositions vary with carbonate type within a single sedimentological layer within the Nussloch loesspaleosol sequence. T(Δ 47 ) in excess of 78°C indicates bulk loess and bulk paleosol carbonates are not suitable for paleoenvironmental reconstructions. In contrast, the pedogenic and biogenic carbonate types record distinct paleoenvironmental information, with sample δ 18 O, δ 13 C, and Δ 47 values likely reflecting specific formation mechanisms and time periods over which the carbonates precipitated.
Geogenic carbonates: bulk loess and bulk paleosol samples δ 13 C, δ 18 O, and T(Δ 47 ) values suggest carbonates in bulk loess and bulk paleosols of the Nussloch sequence are detrital rather than pedogenic in origin. Apparent T(Δ 47 ) values of bulk carbonates (∼78-120°C) far exceed the range of plausible Earth-surface temperatures (Fig. 3). Additionally, calculated apparent δ 18 O water values of bulk carbonates (6.3-12‰; Fig. 4) are inconsistent with formation in situ in the (paleo)soil unit. Instead, the isotopic data point to diagenetic and/or metamorphic sources for these carbonates, which we suggest were likely derived from pre-existing rock through processes of weathering and erosion. Waters in metamorphic environments are commonly characterized by high δ 18 O values (potentially > 0‰ VSMOW) as a result of dehydration processes and water-rock isotopic exchange in deep-crustal environments (Sheppard, 1986), which may explain the positive δ 18 O water values calculated for many bulk carbonates in the Nussloch sequence. Diagenetic and metamorphic carbonates are exposed throughout Western Europe, making it difficult to ascertain the exact source(s) of Nussloch detrital carbonate. That said, prevailing westerly and north-westerly winds into the Nussloch region during the last glacial period (Antoine et al., 2013;Rousseau et al., 2014) point to source regions to the west of the Nussloch Basin, possibly marbles exhumed in the Scottish Highlands and abundant limestone throughout the Devon Basin in England. Additional geochemical data (e.g., Sr-Nd isotopes) could potentially help decipher source(s) for Nussloch detrital carbonate. Nevertheless, high apparent T(Δ 47 ) and δ 18 O water values of bulk carbonates indicate that bulk carbonates do not record the local environmental conditions during or following sedimentation and cannot be used as a direct paleoenvironmental proxy.

Pedogenic carbonates: rhizoliths and loess dolls
Rhizoliths and loess dolls are pedogenic carbonates formed in situ in the soil (e.g., Gocke et al., 2010Gocke et al., , 2011Li et al., 2015;Gao et al., 2020;Sun et al., 2020). Their δ 13 C values should largely reflect vegetation type, their δ 18 O water mirror the values of local soil water, and T(Δ 47 ) is presumably a measure of the soil temperature during soil drying and CO 2 degassing. The pedogenic carbonate isotopic values are generally consistent with formation in soil conditions. In the following paragraphs, we examine the pedogenic carbonate data in the context of the range of δ 13 C, δ 18 O, and T(Δ 47 ) values expected for Nussloch site paleoenvironmental conditions and discuss implications of distinct seasonality and mechanisms of pedogenic carbonate formation for paleoenvironmental reconstruction.
Multiple lines of evidence place constraints on Nussloch paleoenvironmental conditions relevant to the period of pedogenic carbonate formation. We lack direct constraints on local terrestrial δ 18 O water values during this period; however, Sima et al. (2009) assumed that precipitation sources and dominant wind trajectories were similar to modern during the Last Glacial. Such δ 18 O water values of precipitation would differ by only ∼0.5‰ from modern (-11‰ to -7‰) (Field, 2010;Terzer et al., 2013) due to the ice volume effect (Prud'homme et al., 2016). Today, minimum and maximum monthly mean air temperatures in the study area range from ∼0 to ∼25°C, and average cold-and warm-season air temperatures are 1.7 ± 0.9°C (DJF) and 18 ± 2°C (JJA), respectively. The mean annual air temperature is 10°C, and modern mean annual soil temperature is therefore likely to be in the range of 10°C to up to 3-5°C higher due to radiative heating of the soil surface (Roxy et al., 2014). Significant elevation of soil surface temperatures via radiative heating is common when the annual precipitation is < 600 mm (Burgener et al., 2019;Gallagher et al., 2019), as it is estimated to have been in Nussloch (Prud'homme et al., 2018). Recent air and soil temperature estimates from earthworm calcite granules at Nussloch suggest similar-to-modern mean warm season air and soil temperatures of 12 ± 4 and 15 ± 5°C, respectively, during paleosol formation (Prud'homme et al., 2016). These temperature estimates were based on measured carbonate δ 18 O and assumed δ 18 O water values for the Nussloch paleosolloess sequence. Other paleo-temperature estimates suggest comparable-to-modern mean annual air temperatures of 8.5-10°C during paleosol formation in analogous paleosol-loess sequences in Hungary (Schatz et al., 2015).
Within these ranges of temperature, the season over which the rhizoliths are formed and the formation mechanisms of rhizoliths should determine the isotopic values and temperatures recorded. Nussloch rhizolith samples have δ 13 C values that are dependent on rhizolith size: small rhizoliths have higher δ 13 C values than large rhizoliths and loess dolls from the same horizon (Fig. 2). We interpret the dependence of the δ 13 C values with rhizolith size to suggest that these carbonates originate from different plant species or groups, and/ or form through specific mechanisms (e.g., root water uptake or soil wetting-drying cycles), motivating hypotheses that can be tested using the suite of δ 13 C, δ 18 O and T(Δ 47 ) data.
First, it is possible that rhizolith formation mechanisms vary with time. Water uptake by roots and resulting water flow toward the root is accepted as the main formation mechanism for rhizoliths (Klappa, 1980;Lambers et al., 2009;Gocke et al., 2011). Following root death, water uptake by the root and water flow towards the root is stopped. However, non-biogenic mechanisms (e.g., wetting-drying cycles) can still drive further carbonate dissolution and precipitation. In this way, a rhizolith formed in association with water uptake by roots may subsequently act as a framework for further carbonate precipitation to build on such that increased size of the rhizolith records a longer formation period (Sun et al., 2019b, c;Huguet et al., 2020). In this scenario, specific formation mechanisms operating over different time scales would affect the environmental information recorded by rhizoliths larger than a few millimeters compared to smaller rhizoliths that presumably form over a short period of time associated with active root water uptake.
Vegetation type could also cause differences in the isotopic signals recorded by large and small rhizoliths. Indeed, a variety of vegetation species are associated with rhizolith formation (Rodriguez-Aranda and Calvo, 1998;Alonso-Zarza, 1999;Matteucci et al., 2007). If large rhizoliths form along perennial roots that are bigger than the roots of annual grasses, this would lead to continued root-water uptake over many years. In contrast, the morphology of small rhizoliths suggest that they formed along the smaller roots of annual grasses, likely over one growing season. Such a difference in duration of carbonate formation could explain the observed differences in δ 13 C, δ 18 O, and T(Δ 47 ) values among large and small rhizoliths. Alternatively, small rhizoliths might have formed when the plant was under water stress, for example at the end of the winter season through the early spring (Tranquillini, 1982), which is most probably the case in glacial times (Yung et al., 1996;Medeiros and Ward, 2013). Under such conditions the plant will close the stomata to reduce water loss, which leads to 13 C enrichment by C3 photosynthesis (e.g., Dawson et al., 2002). Observed differences in temperature and δ 18 O water values among small versus large rhizoliths could be consistent with small-rhizolith formation under water stress, with fine rhizoliths recording higher formation temperature (Fig. 3) and forming from more evaporative waters (Fig. 4). The δ 13 C values (Fig. 2) do not show obvious evaporation control, potentially reflecting other factors, such as plant type, that may influence δ 13 C values.
We suggest that rhizolith formation temperatures (Fig. 3) are consistent with differences in formation duration and mechanisms. We suggest small rhizoliths likely formed over short periods, potentially as short-lived as during a single soil-drying event due to root-water uptake. In contrast, large rhizoliths may have formed over a few years up to a few decades (Gocke et al., 2011;Sun et al., 2019a) integrating several warm seasons, and therefore recording the mean temperature of those seasons. We therefore interpret the clumped isotope temperature for the large rhizoliths (average T[Δ 47 ] = 14 ± 5°C) to represent warm-season soil conditions averaged over up to a few decades.
The clumped isotope warm-season soil temperature reconstruction is consistent with the average mean warm-season soil temperature estimates (13 ± 4°C and 15 ± 3°C) reported for Nussloch (Prud'homme et al., 2016(Prud'homme et al., , 2018 Loess doll formation may occur over centuries or longer depending on size (Zamanian et al., 2016a), rather than during a single vegetation period like small rhizoliths. However, the loess dolls and one of the small rhizolith samples are similar in both formation temperatures and size (Figs. 1, 3). Thus, it is possible that the small loess dolls might also be formed over short periods, although more evidence than just formation temperature would be necessary to substantiate such an argument.
Taken together with the δ 13 C and δ 18 O data, T(Δ 47 ) values further support interpretation of differences in duration of carbonate growth by carbonate type (Fig. 5). We suggest that small rhizoliths (a few millimeters in diameter) and loess dolls exhibiting large variability in T(Δ 47 ) (19-32°C) record the temperature of one growing season and/or soildrying event. Such transient growth during peak warmseason events can explain why observed T(Δ 47 ) values are significantly warmer than expected for growth during glacial conditions. In contrast, we interpret average T(Δ 47 ) of ∼14°C for large rhizoliths (a few centimeters in diameter) as a more reliable measure of average paleo-temperatures across multiple growing seasons, or over decades (Table 1, Fig. 3).

Biogenic carbonates: land-snail shells and earthworm biospheroliths
The δ 13 C and δ 18 O values of distinct land-snail species (Fig. 2) reflect the specifics in the diet of these organisms (i.e., ingested plants and, to a lesser extent, geogenic carbonate). δ 13 C values in shell carbonate depend mainly on the land-snail diet regime (Metref et al., 2003;Balakrishnan et al., 2005;Zhang et al., 2014). The similar shift (∼ -1‰) to lower δ 13 C values from Lower loess to paleosol in both land-snail shells and bulk carbonates reveals that the land snails S. oblonga, P. muscorum, and T. hispidus were ingesting bulk carbonate (Pigati et al., 2010;Yanes et al., 2012) (Fig. 2). These identified species are all able to ingest carbonates (Fechter et al., 1990;Boschi et al., 2011), which comprise ∼5-25% of their diet (Pigati et al., 2010;Yanes et al., 2012). We can estimate the δ 13 C of vegetation during paleosol formation and Lower loess deposition using the equation of Yanes et al. (2008) based on shell carbonate δ 13 C values. Using this equation, and assuming 0% carbonate ingestion, the range of δ 13 C of vegetation is calculated as -22 to -17‰ VPDB; but considering 5-25% carbonate ingestion, the estimated δ 13 C values of vegetation range from -25‰ to -20‰ VPDB. The latter calculated δ 13 C range of vegetation is consistent with the reported δ 13 C values of organic carbon (-24.9 to -23.7‰) for the Nussloch paleosol-loess sequence (Hatté et al., 1999;Prud'homme et al., 2018), supporting our interpretation of significant carbonate ingestion by analyzed snail species. Similar calculations using A. arbustorum, which we assume did not ingest carbonate due to the similarity of δ 13 C values for co-occurring shell and non-biogenic carbonate in paleosol and Lower loess, also gives a δ 13 C value of -26‰ for vegetation. Lastly, earthworm biospheroliths also can be used to reconstruct plant δ 13 C (Prud-hommme et al., 2018). Using an epsilon calculated from Canti (2009) and typical diet estimation equations (Passey et al., 2005), we calculate a δ 13 C plant of -23‰ from the earthworm biospheroliths δ 13 C values.
The clumped isotope data and calculated δ 18 O water values are broadly consistent with the hypothesis that land-snail individuals record seasonal and local changes in atmospheric temperature (Eagle et al., 2013;Wang et al., 2016;Zhang et al., 2018;Zhai et al., 2019) and the isotopic composition of local precipitation modified by evaporation (Zaarur et al., 2011;Dong et al., 2020). Different species grow in seasons with different temperatures Zhai et al., 2019), precipitation δ 18 O water values, and relative humidity. Furthermore, because individual snails are short lived, considerable variation in the temperatures and δ 18 O water values recorded by land-snail individuals is expected (Eagle et al., 2013). This variability is further exacerbated because the recorded clumped isotope temperatures depend on shell properties such as shell color and morphology, and the ecological behavior of the snail (Zaarur et al., 2011). Accordingly, the observed variability among individual snail-shell T(Δ 47 ) and reconstructed δ 18 O water values within a single horizon is expected; however, this natural variability can mask and complicate interpretation of environmental shifts in the proxy records.
Land-snail shells, following embedding in loess and paleosol, might also have undergone recrystallization and associated re-equilibration of shell carbonate isotopic values due to dissolution and re-precipitation (Zamanian et al., 2016b, c). Hence, it is possible that recrystallization might have biased the recorded temperatures toward warmer, post-burial environmental conditions. However, we are unable to determine the extent of such effects. Large T(Δ 47 ) uncertainties for some analyzed biogenic carbonate samples ( Table 1) further limit our ability to test the hypothesis of land-snail variability rigorously. Future studies, where it is possible to sample a large number of individual shells of each species, could achieve sufficient precision (Eagle et al., 2013).
In contrast to land-snail shells, earthworms seem to not incorporate the ingested carbonates in their biospheroliths (Moine et al., 2017). Thus, the T(Δ 47 ) values (8 ± 2°C) of earthworm biospheroliths likely record the temperature at the soil surface during earthworm activity. Our direct biospheroliths T(Δ 47 ) measurement is within the range, but on the lower side of our average warm-season soil temperature estimate based on large rhizolith clumped isotope measurements (14 ± 5°C), and the warm-season soil (13 ± 4°C to 15 ± 3°C) temperature ranges inferred by Prud´homme et al. (2016) for the Nussloch paleosol-loess sequence. The uncertainties are large enough to permit significant variation in biospherolith-based temperature estimates, which would not be surprising given that the carbonate samples here and in the study of Prud´homme et al. (2016) homogenize a large number of individual earthworm granules, each of which formed in a single day. Soil carbonate formation and earthworm activity were likely both restricted to the warmest periods of year in the Nussloch permafrost environment, so we do not suggest that the low resolution of these thermometry techniques masks a seasonal difference in rhizoliths and biospheroliths formation.

Implications for paleoenvironmental reconstructions from stable isotopes of carbonates in loess and paleosol deposits
Our observations of the δ 13 C, δ 18 O, and clumped isotope temperatures of carbonate types associated with the Nussloch paleosol-loess sequence suggest several broader implications for paleoenvironmental reconstructions from these and similar deposits (Table 2, Fig. 5).
First, the contribution of detrital carbonate to bulk loess and paleosol carbonate makes these materials unsuitable for paleoenvironmental reconstruction (e.g., interpreting mean annual or seasonal air temperatures), as observed in other recent T(Δ 47 ) investigations of loess-paleosol sequences . The contribution of geogenic detrital carbonate was made obvious by the clumped isotope apparent temperatures well in excess of plausible Earth surface conditions. However, it is possible that in the absence of clumped isotope data, the contribution of bulk geogenic carbonate may not be recognized. Geogenic carbonate contribution may be particularly difficult to recognize in deposits where pedogenic grain coatings of carbonate formed in situ also contribute significantly to the isotopic composition of the bulk carbonate samples. In such cases, it may be possible to interpret variations in carbonate δ 13 C and δ 18 O to document temporal variation in the seasonality of precipitation, in vegetation or in other paleoenvironmental variables, when in reality the carbonate δ 13 C and δ 18 O variations may document variations in the degree of pedogenic overprinting of detrital carbonate in bulk loess or paleosol deposits. Clumped isotope data would provide a simple test of the interpretation. In addition, isotopic analysis of bulk carbonate in paleosol-loess deposits can be valuable in interpreting isotopic compositions of land-snail shells, given that many of these organisms ingest this carbonate as a significant proportion of their diet.
Second, our findings contribute to a growing body of literature concerning seasonal bias in pedogenic carbonate formation, and the implications of this bias for paleoenvironmental reconstruction (e.g., Burgener et al., 2016;Kelson et al., 2020). Most previous work on this topic has focused on carbonate pendant (clast coating) and nodule formation in modern and Holocene soils (Breecker et al., 2009;Quade et al., 2011;Peters et al., 2013;Burgener et al., 2016;Gallagher and Sheldon, 2016;Ringham et al., 2016). We complement these approaches by examining isotopic values in carbonates of different types in the same temporal horizon. The variations in carbonate δ 13 C, δ 18 O (Fig. 2), and clumped isotope temperatures (Fig. 3) are consistent with specific mechanisms 266 K. Zamanian et al.
of carbonate formation that integrate environmental information over time periods, which we suggest may range from a single vegetation growth event in small rhizoliths, to decades and centuries of summers in larger rhizoliths. Together, our interpretations provide guidance for sampling strategies, to consider time resolution in future studies. Such studies may target carbonates that grow over short periods to examine seasonal variability and extremes, and carbonates that grow over longer periods to document seasonal average temperatures and representative vegetation and precipitation isotopic values (Burgener et al., 2016;Ringham et al., 2016;Ji et al., 2018;Lechler et al., 2018). Finally, our results and those of other studies in the Nussloch remind us that pedogenic/biogenic processes as well as seasonal biases must be considered in paleoenvironmental interpretations of paleosol archives. In European paleosolloess archives alone, reconstructions of Last Glacial paleotemperatures depend significantly on both the proxy method and the type of deposit to which the method is applied. Specific proxies may provide estimates for a particular month or average warm season air temperature (Rousseau, 1991;Moine et al., 2002), for mean annual air temperature (Renssen and Vandenberghe, 2003;Vandenberghe et al., 2014;Schatz et al., 2015), for winter air temperature (Rousseau, 1991), or for seasonally biased or mean annual soil temperature (Prud 'homme et al., 2016). Mean annual temperature estimates for periods of loess formation and paleosol formation also differ from one another (Schatz et al., 2015), with paleosol formation representing milder climate conditions during interstadial events (Prud'homme et al., 2016). Better understanding of the processes responsible for differences in proxy records as a function of the periods over which proxy materials form is needed to integrate multiple records into more complete paleoenvironmental reconstructions.

CONCLUSIONS
Various carbonate types deposited in the same paleo-horizon or layer of the Nussloch paleosol-loess sequence record specific environmental information because of differences in formation mechanisms and duration. Bulk carbonates in the Nussloch paleosol-loess sequence are detrital materials that record the diagenetic conditions of carbonates in the deflation area instead of local environmental conditions. Therefore, bulk carbonates should be excluded from samples taken to reconstruct the local paleoenvironment. We interpret small rhizoliths, loess dolls, land-snail shells, and earthworm biospheroliths as records of temperatures and soil water content that reflect environmental conditions over the short time periods over which they form. In contrast, large rhizoliths (centimeters in thickness) record mainly summer temperatures averaged over decades. This finding suggests that rhizolith size should be considered for paleoenvironmental reconstructions, with larger rhizoliths being more reliable indicators of average warm season conditions. Earthworm biospheroliths and land-snail shells are suitable materials for paleoenvironment and paleovegetation reconstructions. In the case of land snails, their ecological habit, including the proportion of ingested bulk/geogenic carbonate and the isotopic composition of bulk soil carbonates should be considered. This study highlights that with proper The δ 13 C, δ 18 O and Δ 47 records in biogenic, pedogenic and geogenic carbonate types from paleosol-loess sequence consideration of proxy carbonate type and associated formation mechanisms, integrated stable (δ 13 C, δ 18 O) and clumped (Δ 47 ) isotopic values of pedogenic and biogenic carbonates can yield robust paleoenvironment reconstructions with decadal and even seasonal temporal resolution.

ACKNOWLEDGMENTS
Our sincere thanks go to Manfred Löscher for kind discussion in the field and recognizing the sampled sediment layers. Special thanks to Landon Burgener for X-ray analyses of land-snail shells and Casey Saenger for suggestions regarding calculation of δ 18 O water from shell carbonate δ 18 O. We highly appreciate comments and suggestions of associate editor Jeff Pigati and reviewers Denis-Didier Rousseau and Christophe Lécuyer. The study was supported by funding from the German Academic Exchange Service (

SUPPLEMENTARY MATERIAL
The supplementary material for this article can be found at https:// doi.org/10.1017/qua.2020.109