© Springer-Verlag 2001 Oecologia 128(4)

Received: 6 October 2000 / Accepted: 12 February 2001 / Published online: 28 April 2001

Matthew R. Henn¹ and Ignacio H. Chapela^2,

Introduction

The natural distribution of stable isotopes is increasingly used to study pathways and rates of exchange between various ecosystem components, from the microscopic to the regional and global scales (Griffiths 1998). The use of this powerful approach relies on several assumptions, some of which have received only cursory attention. In particular, it has been convenient to assume that microbial transformations of organic and inorganic substrates do not significantly alter the natural distribution of stable isotopes. For example, Wedin et al. (1995), Bernoux et al. (1998), and Bashkin and Binkley (1998) used shifts in isotopic values of soils to characterize changes in organic input as ecosystems shift between C₄- and C₃-dominated vegetation. These shifts were, however, uncorrected for potential changes caused by fungi and bacteria during decomposition, although Wedin et al. (1995) did suggest that ¹³C enrichments in the litter might result from immigration of carbon into the decaying organic material from microbes. Recent evidence is mounting to challenge the assumption that microbial isotopic fractionations are not a significant consideration in terrestrial ecology. Studies conducted under field, mesocosm, and laboratory conditions indicate that both C and N isotopic values of the whole microbial organism may not reflect the isotopic composition of the source substrate (Gebauer and Taylor 1999; Gleixner et al. 1993; Handley and Scrimgeour 1997, and references therein; Henn and Chapela 2000; Hobbie et al. 1999b; Högberg et al. 1999b; Kohzu et al. 1999; Will et al. 1986, 1989) and that various known microbial nutrient transformations fractionate stable isotopes (Blair et al. 1985; Handley and Raven 1992, and references therein; Macko and Estep 1984). We provide here an analysis of field data aimed at quantifying and characterizing the fractionation effects due to fungal transformation of N and C compounds of natural substrates in Northern Hemisphere forest ecosystems.

Through their decomposing, mycorrhizal, predatory, scavenging, and pathogenic activities, fungi form a crucial link in terrestrial nutrient cycling. Although their microscopic hyphae are often concealed, it has been estimated that fungi can represent 10-60% of the total biomass in the litter layer (Frey et al. 1999; Newell 1992), and single fungal individuals can cover many hectares (Smith et al. 1992). Fungi are the most abundant microorganisms in aerated soils with typical biomass ranges of 500-5,000 kg ha^-1 (Metting 1993).

Despite their significance, the contribution of fungi to changes in isotopic distributions in terrestrial ecosystems remains poorly understood. Evidence from earlier studies in fungi suggests that specific biochemical pathways do have strong fractionation effects (Abraham et al. 1998; Gleixner et al. 1993; Hollander et al. 1979). Enrichments in the heavy or light isotope in fungal biomass relative to bulk substrates of up to ~7 for ¹³C (Will et al. 1986) and ~4 for ¹⁵N (Gebauer and Taylor 1999; Högberg et al. 1999b; Kohzu et al. 1999) are documented, and a growing number of studies suggest that isotopic measurements of fungal sporocarps in the field can differentiate between ectomycorrhizal (EM) and saprotrophic (SAP) forest fungi (Hobbie et al. 1999b; Högberg et al. 1999b; Kohzu et al. 1999). Högberg et al. (1999b) propose that such differences could be used as a diagnostic tool to elucidate ecological roles of various fungi in the field. However, the ecophysiological basis of this apparently robust fractionation pattern remains obscure. Most importantly, it is unclear whether fractionation patterns are determined by extrinsic environmental properties or intrinsic physiological determinants. We have shown in the laboratory that fungi can indeed produce strong fractionation effects that are taxon- and substrate-specific, as well as influenced by microenvironmental conditions (Henn and Chapela 2000), but the extent that this intrinsic fractionation potential is expressed in the field, as separate from substrate effects, has not been tested. Here, we explore these questions for the case of forest basidiomycetes, with the benefit of a large pool of field data produced over the last few years.

Materials and methods

Field sampling

We collected fungal sporocarps (mushrooms), mycelium and substrates in stands of Monterey pine (Pinus radiata) as well as mixed conifer stands in the Klamath Knot region of Northern California. Sporocarps and mycelium representing 22 genera were collected fresh, immediately frozen to -80°C and subsequently lyophilized. Taxonomic determinations were performed using traditional mycological identification methods, and checked for consistency using DNA-based methods (Gardes and Bruns 1991). For each fungal specimen collected, substrate samples were obtained from the immediate vicinity. Designated substrates for each specimen were chosen on the basis of a direct mycelial connection between fungus and substrate (SAP fungi) or on accepted knowledge of ecological role (EM fungi). Substrate samples were immediately dried at 60°C for 18-24 h. All samples were ground to a fine powder using ceramic mortar and pestle followed by Wiggle-L-Bug micromaceration for 2 min. For SAP fungi, we obtained substrates avoiding visually detectable mycelium. SAP substrates included soil (0-15 cm, excluding litter and F-horizon), decayed litter, fresh litter, fresh and decayed wood. Soil samples had all roots greater than 100 µm in diameter removed under a dissecting microscope. For wood decomposers, sections of unstained and undecayed wood were obtained from the same log where fungi were collected; if such a section was not available a fresh sample from the closest tree of the same species was taken. Because SAP fungi other than wood decomposers are known to potentially access nutrients from several different substrates simultaneously, they were compared against an integrated substrate value, determined by calculating a weighted average of all the substrate materials collected in the immediate vicinity of the fungal sporocarp. Fresh pine needles were chosen as the closest approximation to the main source of C for EM fungi, since it is known that EM predominantly receive recently synthesized carbohydrates from the host tree (Smith and Read 1997 and references therein). Because variation in isotopic composition is also known to occur at various canopy heights and orientations (Livingston et al. 1998; Schleser 1992), we sampled a diversity of needles from the vicinity of fungal samples. Isotopic needle values used to estimate EM fungal C source are weighted averages of these diverse canopy samples for each site. No substrate correction was attempted for ¹⁵N values since actual N sources and their ¹⁵N values are currently not knowable even as an approximation. For this reason, statements regarding enrichment and depletion of ¹⁵N in fungal tissues are relative to the ¹⁵N of atmospheric N, as an arbitrary point of reference. All reported isotopic values correspond to tissues from whole sporocarps and are not differentiated based on pilei (caps), stipes, or bases.

Isotopic analysis

Isotopic composition and percent element were determined using an online continuous flow CN analyzer coupled with an isotope ratio mass spectrometer (ANCA-NT module coupled to a model 20-20 IRMS, Europa Scientific). Isotopic values represent total C or N and are not differentiated based on cellular fractions. Values are reported in the standard notation (¹³C or ¹⁵N; ) relative to Pee-Dee Belemnite for C and Atmospheric N₂ for N, using NIST Peach Leaves no. 1547 as a standard, where X=[(R_sample/R_standard)-1]1,000 and R is the molar ratio ^heavyX/^lightX (Lajtha and Michener 1994). C isotopic discrimination, the change in ¹³C from the substrate to the fungal biomass, is reported in units as ¹³C=¹³C_fungus-¹³C_substrate. Isotopic discrimination for N was not calculated since the source of N for each fungus could not be reliably established in the field. Instrumental precision for solid samples, estimated as the standard deviation of measurements of ground peach leaves (NIST no. 1547) was 0.13 for C and 0.20 N (SD for all isotopic runs combined; n=106). Each sample was run twice and values averaged, with duplicates always within <0.07 of each other. Values were corrected for linearity relative to the beam area of the standard.

Meta-analysis

We performed a meta-analysis of ¹³C and ¹⁵N values for basidiomycete sporocarps from across the Northern Hemisphere using data from our own collections in California, Hobbie et al. (1999b) (Alaska, US), Högberg et al. (1999b) (Sweden) and Kohzu et al. (1999) (Japan and Malaysia). These studies are focused predominantly on conifer forests, but also include broadleaf forests and a lowland tropical forest in Malaysia; 387 datapoints were included for ¹³C measurements and 231 for ¹⁵N, corresponding to 306 identified species and 93 genera. Isotopic values reported by Högberg et al. (1999b) and Hobbie et al. (1999b) as multi-specimen averages for single species were treated as individual measurements in the meta-analysis. A substantial ¹⁵N fungal database published by Taylor et al. (1997) was not included in our meta-analysis since these authors report ¹⁵N values of fungal caps, stipes, and bases independently, while studies included here report whole sporocarp values. Taylor et al. (1997) indicate that differences of several can exist between each anatomical structure in sporocarps.

Two manipulations were applied to the meta-analysis data to explore the existence of a fractionation signal associated specifically with fungal physiological or ecological behavior. First, in order to allow for comparison across studies, data for both C and N were corrected to account for site-specific biases. An overall average was derived for the whole data set and the difference between the overall mean and the mean in each study site was calculated. This difference was assumed to accumulate systematic bias in the data for each site, including instrumental and operator bias, as well as regional or local variation in substrate isotopic composition. To correct, this site-specific difference of means was subtracted from each data-point to produce a "site-corrected" value. Such site-specific correction ranged from -2.35 to 1.35 for C and -4.11 to 3.66 for N. Second, we artificially removed the difference in means between EM and SAP fungi for the C data, based on the suggestion by our own California data set (Fig. 1b) that a large proportion of the difference between these ecological groups could be due to substrate. To achieve this analytical "correction" for presumed substrate effects, we subtracted from each data point the difference of means between the overall mean (after site correction) and each of the site-corrected means for EM or SAP fungi.

Statistical analyses

All statistical analyses were performed using JMP 3.2 (SAS Institute) or SAS 6.12 (SAS Institute). Normality of distributions was tested using the Shapiro-Wilkes test for normality and values are reported as P-values. Measures of skewness and kurtosis were determined using the g₁ and g₂ statistics respectively and reported as P-values were alpha () is two tailed and set to 0.05 and beta () =0.10 (Zar 1999). Nonparametric statistics were used when distributions failed a test of normality.

Results

C values

As reported previously by other authors (Henn and Chapela 2000; Kohzu et al. 1999; Will et al. 1986, 1989), important fractionation effects were observed for C stable isotopes. Our California data show fractionation effects that differentiate between EM and SAP fungi (see Fig. 1), a phenomenon we term here the "EM-SAP Divide". Each of these two ecological groups had average ¹³C values that closely matched values found by other researchers (Hobbie et al. 1999b; Högberg et al. 1999b; Kohzu et al. 1999), with EM fungi ¹³C=-25.29±0.31 (mean±SE) and SAP fungi ¹³C=-22.14±0.26 (mean±SE) (Fig. 1). The range of ¹³C values for SAP fungi in California, -25.19 to -17.95, was larger than that of EM fungi in the same sites, viz. -27.98 to -23.98 (Fig. 1). Although initial perusal of the ¹³C frequency distribution for EM alone, SAP alone, and all measurements combined in our limited California study did not appear to be normal, only the EM distribution was significantly different (=0.05) from normal. While some overlap in frequency distributions of ¹³C values for EM and SAP fungi was observed they can still be distinguished as significantly different populations (Wilcoxon rank sums, P<0.001) (Fig. 1).

However, when ¹³C values for individual specimens are considered relative to the specific substrate used and expressed as ¹³C, frequency distributions of SAP and EM fungi became indistinguishable (Fig. 1b). Mean substrate-corrected values for EM and SAP species were ¹³C=3.35±0.41, and 4.25±0.28 (mean±SE), respectively (P=0.079; Fig. 1). These values are in the same range as those reported by Kohzu et al. (1999).

A meta-analysis of data obtained around the Northern Hemisphere support, on average, the ¹³C EM-SAP divide, with reported ¹³C values of -25.45±0.09 for EM fungi and -22.86±0.12 (mean±SE) for SAP fungi (t-test, P<0.0001; see Fig. 2a). Even when site-specific biases are taken into consideration there is a significant difference between ¹³C values of EM (-25.09±0.08,mean±SE) and SAP fungi (-23.46±0.11; mean±SE) (t-test, P<0.0001; Fig. 2b). An overlap of 5 was observed between EM and SAP ¹³C frequency distributions in the site-specific corrected meta-analysis data (Fig. 2b). Further, ¹³C frequency distributions of these data for EM and SAP fungi suggest a multi-modal signal with maxima at -25.3 and -24.3 for EM, and -24.3 and -23.3 for SAP. The overall site-specific corrected meta-analysis ¹³C frequency distribution was leptokurtic (P<0.005), indicating a significant narrowing of variability than would be expected from a normal distribution.

Extrapolating from our observation in California that substrate-corrected data did not show a difference between SAP and EM fungi, we analytically removed additional substrate effects due to differences in C source access (Fig. 2c; see Materials and methods). The correction for EM fungi was -0.59 and 1.02 for SAP fungi. Such "substrate-corrected" data produce a multi-modal distribution as described above, but the two maxima for each of EM and SAP frequency distributions become practically identical, at an estimated ¹³ C of -25 and -24 (Fig. 2c). The estimated ¹³C substrate value for EM fungi in the Northern Hemisphere is -29.41±0.09 (mean ± SE) and has a range of -35.66 to -26.46; while the corresponding estimated ¹³C substrate value for SAP fungi is -26.02 ± 0.12 and has a range of -32.86 to -21.91.

N values

Fractionation effects of stable N isotopes were very variable in our California study. We observed both depletion and enrichment for ¹⁵N relative to atmospheric N, covering a range of values from ¹⁵N =-4.11 to 15.93 (Fig. 1). On average, the difference between EM (10.14±0.84; mean±SE) and SAP (0.05±0.58; mean±SE) fungi was significant (t-test, P<0.0001, Fig. 2) although these frequency distributions overlapped by 2 (Fig. 1). The overall frequency distribution for our California ¹⁵N data departed significantly from normality (Shapiro-Wilkes, P=0.04) and was apparently multi-modal, although the small sample number makes it difficult to support this statement (Fig. 1).

In the meta-analysis, basidiomycete sporocarps collected from across the Northern Hemisphere ranged from ¹⁵N=-7.10 to 21.8 (Fig. 2a) and EM and SAP fungi, on average were significantly distinguishable (Wilcoxon, P<0.001). The site-corrected data had a range of ¹⁵N=-8.90 to 21.81 (Fig. 2b). EM and SAP fungi were, on average, significantly different (Wilcoxon, P<0.0001) with mean ¹⁵N=6.40±0.42 and 0.83±0.39 (mean±SE), respectively. However, both distributions had wide ranges (Fig. 2b), overlapped by 18 and were not demonstrably normal, with strong suggestions of multimodality (Fig. 2b, c). Because no reliable estimation of N source for individual specimens was available we could not correct ¹⁵N data to take additional EM and SAP substrate effects into account (see Gebauer and Taylor 1999).

Discussion

Significant stable isotopic discrimination for C and a wide variety of values for N isotopes are observed to occur in fungi under field conditions. These results indicate the need to revise the tacit assumption of many ecological studies that fungal interfaces do not alter the natural distribution of C and N isotopes in the field. Field results show that, when averaged over many species and specimens, forest basidiomycetes become enriched in ¹³C relative to their bulk C source, and either enriched or depleted in ¹⁵N relative to atmospheric N (Fig. 1). The magnitude of these isotopic effects is variable from negligible levels to very large fractionations.

One of the strongest statements emerging from recent field studies of fungi is epitomized in the "EM-SAP Divide" phenomenon, where naturally occurring C and N stable isotopic values strongly differentiate between EM and SAP fungi (Figs. 1, 2) (Hobbie et al. 1999b; Högberg et al. 1999b; Kohzu et al. 1999). We have confirmed this phenomenon in our local field experiment as well as in a meta-analysis that included the four comparable major studies (17 sites) performed to date in the Northern Hemisphere. Indeed, a very strong separation between EM and SAP fungi can be observed when N and C isotopic values are considered together in dual-isotope plots (Fig. 2). While frequency distributions overlap when each isotope is considered separately, dual-isotope plots in the individual California site we studied showed practically no overlap of 95% confidence limits for EM and SAP (Fig. 1a). Such a clear separation does not occur when substrate effects are taken into account (Fig. 1b) nor when variability across sites and studies is considered in a meta-analysis (Fig. 2b). Thus the suggestion that isotopic values could be used as diagnostic of fungal SAP or EM ecophysiological roles seems to hold only at a small scale in relatively homogenous sites, and seems most robust only when N and C isotopes are considered together. Studies where this suggestion is supported have been performed in long-established native ecosystems presumably under relatively steady-state conditions, but the diagnostic power of dual N and C isotopic analysis should not be overestimated without additional information, especially in changing ecosystems. The possible existence of such qualitative difference between steady-state and changing ecosystems was also pointed out by Gebauer and Taylor (1999).

Although attempts have been made to determine the ecophysiological basis of the EM-SAP Divide, no comprehensive mechanistic interpretation of this important pattern has emerged to date. However, we observed in the detailed California study that when substrate values are taken into account, such differentiation is removed for C isotopes (Fig. 1b), suggesting that the EM-SAP Divide is determined more by the nature of the substrate being utilized rather than by strong physiological differences between EM and SAP fungi. This observation should temper the growing tendency to use isotopic measurements on sporocarps to provide ecophysiological inferences, in the absence of information that may help discern one controlling factor from the other (Hobbie et al. 1999b; Högberg et al. 1999b; Kohzu et al. 1999). It must be noted that the problem of isotopic fractionation effects by fungi is expected to be most critical in C₃-dominated ecosystems, since laboratory experiments show that strong fractionation effects can be observed when fungi are grown on C₃ substrates but not on their C₄ isomers (Henn and Chapela 2000).

In light of the uncertainty factors discussed here, it seems imperative to collect other ecologically relevant data in addition to isotopic determinations to approach ecological interpretations. Such data should include some measure of taxon-specific isotopic fractionation, ideally determined in the laboratory and the field, as well as consideration of fine-scale differentiation of substrates and mycelia. For example, Chapela et al. 2001 performed careful sampling of sporocarps as well as mycorrhizal root-tips of DNA-identified Suillus luteus in pine plantations in Ecuador, showing that isotopic values in root-tips containing at least 40% fungal biomass were undistinguishable from host values, and significantly different from fungal sporocarp values. In addition, these authors utilized radiocarbon dating to complement stable isotopic measurements before a suggestion could be made implicating a role of S. luteus in removing organic C from plantation soils (Chapela et al. 2001). Inferring fungal ecophysiological behavior on the basis of stable isotopes should consider taxon-specific effects as well as fine-scale spatial and physiological effects derived from substrate and mycelia.

A revised approach to utilize fungal field stable isotopic data for ecological inference, might also include measurements of specific cellular components that may prove useful in a manner analogous to the use of cellulose as a comparable and relatively stable source of isotopic values in plants (MacFarlane et al. 1999). Carefully controlled laboratory experiments are needed to validate the use of such a marker molecule that could help account for physiological discrimination effects in fungi (Abraham et al. 1998; Gleixner et al. 1993; Henn and Chapela 2000; Will et al. 1986, 1989) as well as metabolic isotope scrambling effects (Hollander et al. 1979). In these studies, careful account must be taken of potential discrimination due to stereochemical isotope effects in the substrate (Henn and Chapela 2000; Rossman et al. 1991). Indeed, the non-random stereochemical distribution of heavy isotopes within organic molecules might prove useful to elucidate extracellular breakdown of organic compounds by fungi and their subsequent uptake into microbial biomass (Henn and Chapela 2000). A further possibility is to use the isotope distribution pattern in a molecule as a historical record of the movement of the organic compound through various decomposition stages.

Based on our ability to match specific substrates with each fungal specimen in our California study, we interpret the ¹³C EM-SAP Divide as reflecting two main C substrate pools being tapped preferentially by each of these fungal groups. The two pools could be characterized as (1) simple carbohydrates recently synthesized from atmospheric CO₂ (EM source) and (2) more complex plant, microbial and animal substrates that, on average, have undergone repeated ecosystem re-processing. Plant cellulose biosynthesis (Gleixner et al. 1993) and fungal degradation of simple disaccharides (Henn and Chapela 2000) are known to result in an average enrichment for the heavier C isotope. The reliance by SAP basidiomycetes on relatively enriched cellulose and microbially re-processed substrates under natural conditions could account for the observed relative enrichment of SAP fungi when compared, as a group, with EM species. Such a mechanism for enrichment in ¹³C was also suggested by Gleixner et al. (1993), although we now know that the fractionation mechanism does not involve the respiration into the atmosphere of relatively depleted CO₂, as suggested by Gleixner et al. (1993), but rather a discrimination against the lighter isotope during sugar uptake (Henn and Chapela 2000). We now know that different species can have very different C fractionation effects on a given substrate, but these effects might not be expressed in certain circumstances, such as during degradation of C4-derived substrates (Henn and Chapela 2000; Will et al. 1986, 1989).

A further observation emerged from the meta-analysis, which is not readily noticeable in any of the individual studies. ¹³C values for both EM and SAP fungi suggested a bimodal distribution, once site and presumed substrate effects were removed as noise from the data (Fig. 2c). This bimodal distribution might signal the existence of two distinct C fractionating pathways with different discrimination effects. As a working hypothesis, we would suggest that this dual C processing effect could be related to the dual uptake mechanism recently described in our laboratory (Henn and Chapela 2000).

A mechanistic understanding of the observed differentiation between EM and SAP fungi in terms of N isotopes presents much greater challenges. Although the EM-SAP Divide is clearly observed when ¹⁵N is considered at the ecosystem-scale and also in our hemispheric-scale meta-analysis, we could not reliably assign substrate data to differentiate between substrate and physiological fractionation effects. It appears from the only study where this was attempted (Gebauer and Taylor 1999) that site-specific and other fine-scale effects predominate over simple ecophysiological functionalities. Gebauer and Taylor (1999) found that the ¹⁵N values for a single fungal species could vary over relatively short latitudinal differences in provenance. These authors discount simple mechanisms to explain the EM-SAP Divide, such as the transfer of N from EM fungi to their hosts suggested by Gebauer and Dietrich (1993) and Hobbie et al. (1999a), and suggest that other phenomena might be more relevant, such as versatility in substrate usage, change in net flows across mycelia, and taxon-specific N uptake and metabolic processing (Gebauer and Taylor 1999). Similarly, Högberg et al. (1999a) showed that large variation can be experimentally obtained in EM ¹⁵N values through manipulation of uptake rates. Finally, N sources and their ¹⁵N values in the field are not presently knowable. For these reasons, we did not perform any substrate correction on our ¹⁵N data, and little mechanistic explanation can be added to the descriptive value of observing a strong EM-SAP Divide (Figs. 1, 2), and a multimodal distribution in our meta-analysis (Fig. 2). The latter might possibly represent diverse balances between N processing pathways, for example the glutamine synthase (GS), the glutamate dehydrogenase (GDH), and the GS pathway linked with the 2-oxoglutarate amino transferase pathway (GOGAT) (Jennings 1987, 1995; Smith and Read 1997), which would provide the taxon-specific processing observed in all field studies to date (Genetet et al. 1984; Quoreshi et al. 1995), but this must remain speculative, awaiting more detailed laboratory studies. Results reported by Handley et al. (1996) and Taylor et al.(1997) that indicate differences in the ¹⁵N value of various parts of fungal sporocarps (i.e. caps, stipes, and bases), further suggest that variations in ¹⁵N sporocarp values may result from internal cycling of N as well as varying N sources and other processes such as the uptake of N and N translocation.

The need to provide ecologically relevant statements based on biased or small samples remains one of the limitations of field studies of fungi using stable isotopes. Högberg et al. (1999b) and Hobbie et al. (1999b), for example, had to base their statements on samples that were biased for EM or SAP fungi, while Gebauer and Taylor (1999) and our own California study were limited by small sample numbers. Under these circumstances, some important features of the measurements can be lost, such as the bimodal distribution of ¹³C values observed only with the large database of our meta-analysis. The meta-analysis presented here not only allows the resolution necessary to begin elucidating such physiological determinants of isotopic fractionation, but it also provides a solid statistical measure of expected variability, which can then be used to establish statistical limits to ecological statements based on smaller-scale studies. Means and standard error of means in our meta-analysis can be used as most conservative estimators of variability, determining the statistical power of ecological statements, as well as providing a guide to robust experiment design. Furthermore, our meta-analysis suggests that the current practice of using simple arithmetic means and assuming normal distributions of isotopic values might belie important patterns in isotopic natural distribution. The existence of multimodal frequency distributions for both C and N points to the possibility of identifying distinct physiological pathways involved in fungal isotopic processing in the field which might not fit the EM-SAP Divide, as has also been found by Gebauer and Taylor (1999).

Acknowledgements. We are grateful to Coastal Ways Ranch, Pescadero, Calif. for access to their native stand of Monterey Pine; T. Dawson and M. Firestone for review of the manuscript; M. Garbelotto for helpful discussions; and P. Brooks for spectroscopy support. This work was supported by grants from the Hellman Family Fund; USDA Agricultural Research Station; the College of Natural Resources, University of California-Berkeley; the William Carol Smith Graduate Fellowship, UC Berkeley, and NASA Headquarters under the Earth System Science Fellowship Grant NGTS-30183.