We examined soil organic C (OC) turnover and transport across the rainfall transition from a biotic, arid site to a largely abiotic, hyperarid site. With this transition, OC concentrations decrease, and C cycling slows precipitously, both in surface horizons and below ground. The concentration and isotopic character of soil OC across this transition reflect decreasing rates of inputs, decomposition, and downward transport. OC concentrations in the arid soil increase slightly with depth in the upper meter, but are generally low and variable (∼0.05%; total inventory of 1.82 kg m−2); OC-Δ14C values decrease from modern (+7‰) to very 14C-depleted (−966‰) with depth; and OC-δ13C values are variable (−23.7‰ to –14.1‰). Using a transport model, we show that these trends reflect relatively rapid cycling in the upper few centimeters, and spatially variable preservation of belowground OC from root inputs, possibly during a previous, wetter climate supporting higher soil OC concentrations. In the driest soil, the OC inventory is the lowest among the sites (0.19 kg m−2), and radiocarbon values are 14C-depleted (−365‰ to –696‰) but show no trend with depth, indicating belowground OC inputs and long OC residence times throughout the upper meter (104 y). A distinct depth trend in δ13C values and OC/ON values within the upper 40 cm at the driest site may reflect photochemical alteration of organic matter at the soil surface, combined with limited subsurface decomposition and downward transport. We argue that while root inputs are preserved at the wetter sites, C cycling in the most hyperarid soil occurs through infrequent, rapid dissolved transport of highly photodegraded organic matter during rare rain events, each followed by a pulse of decomposition and subsequent prolonged drought. These belowground inputs are likely a primary control on the character, activity, and depth distribution of small microbial populations. While the lack of water is the dominant control on C cycling, very low C/N ratios of organic matter suggest that when rainfall occurs, hyperarid soils are effectively C limited. The preservation of fossil root fragments in the sediment beneath the driest soil indicates that wetter climate conditions preceded formation of this soil, and that vadose zone microbial activity has been extremely limited for the past 2 My.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 In arid and semiarid ecosystems, water availability constrains primary production and thus soil C cycling [Austin et al., 2004]. In the Atacama Desert of northern Chile, a natural rainfall gradient allows us to directly observe the effects of even drier conditions: the transition from arid soils supporting plants to hyperarid soils that are largely abiotic [Ewing et al., 2006, 2007]. With this transition to extreme aridity, there is a transformation from biologically mediated soil nutrient cycling to the accumulation of potential nutrients through deposition of atmospheric materials. At the hyperarid extreme, nitrate and organic N (ON) accumulate with time, but slow C cycling is sufficient to limit accumulation of organic C (OC) from atmospheric deposition [Ewing et al., 2006, 2007]. To understand the nature and rates of C cycling across the arid-hyperarid transition, we measured the stable and radiogenic isotopic trends in soil OC with depth at three representative sites. We compare the results of these observations with model predictions in order to characterize fundamental changes in OC cycling that occur as water and life approach zero.
 Much is known about the effect of rainfall on soil C cycling in non-hyperarid settings. Globally, total soil OC decreases with mean annual precipitation as a result of the balance between inputs and losses (MAP [Amundson, 2001; Post et al., 1982]). In most places, plants dominate OC inputs to soil with both aboveground and belowground additions, but inputs also include a small amount of direct deposition of OC from the atmosphere (Figure 1). Total inputs are balanced by losses due to the release of CO2 produced by microbial decomposition, and to a lesser extent the dissolved flux out of soils. With depth in many soils, OC concentrations decrease, OC-Δ14C values decrease, and OC-δ13C values increase. The decrease in soil OC-Δ14C values reflects an increasing mean residence time of OC with depth [Baisden et al., 2002; Nadelhoffer and Fry, 1988; Wang et al., 1996a, 1996b], while an increase in OC-δ13C values reflects increasingly microbially altered OC [Baisden et al., 2002; Feng et al., 1999; Santruckova et al., 2000]. The inverse correlation of OC-δ13C values and OC-Δ14C values [Baisden et al., 2002; Nadelhoffer and Fry, 1988; Torn et al., 2002] reflects the intimate relationship between the mean residence time of OC and the degree of microbial alteration in most soils.
 Isotopic trends in OC with depth are strongly dependent on soil hydrology and biological mixing, and do not always develop in desert soils [Amundson, 2001]. In arid environments, plant cover and root distribution are spatially variable, and thus so are carbon inputs to soils [e.g., Amundson et al., 1988; Austin et al., 2004]. These spatially variable inputs are then subject to temporally variable in-soil transport and decomposition, as a result of episodic wetting events of highly variable magnitude [Austin et al., 2004]. This disrupts depth-dependent trends that are a function of relatively continuous water availability. In addition, aridity limits mineral weathering, restricting production of secondary clay minerals that may strongly adsorb OC. A potential consequence of limited decomposition and transport in arid soils is the preservation of organic matter formed under previous climate conditions. In general, high spatial variability in biological processes is inherent to arid systems.
 Little is known about how OC cycling changes when aridity leads to the absence of vegetation. In the southern Atacama Desert where plants are present, annual rainfall is a few cm-scale events a year. In the hyperarid region to the north, cm-scale events occur approximately once per decade, while larger events occur even more infrequently [Warren-Rhodes et al., 2006]. Fog and dew events that wet the upper few mm of soil are frequent, but on average soil relative humidity values in the top few cm are 20% to 44% [Warren-Rhodes et al., 2006]. The potential benefit of fog and dew events for very near surface biological activity is counteracted by high UV intensity, restricting extremely limited surface populations to UV tolerant organisms and lithophilic cyanobacteria that rely on the UV protection offered by translucent clasts [Warren-Rhodes et al., 2006]. Thus we hypothesized that across the arid-hyperarid transition, soil OC should reflect (1) increasingly limited biological inputs, in-soil transport and decomposition; and (2) the increased relative importance of inputs by atmospheric deposition of OC. In addition, we suspected that OC in these soils would also indicate legacy effects, given substantial emerging evidence of previously wetter climate episodes [e.g., Grosjean et al., 2003; Lamy et al., 2000; Latorre et al., 2003, 2006; Maldonado et al., 2005; Stuut and Lamy, 2004; Stuut et al., 2006]. While previous observations of biological limitation in the Atacama have been interpreted as a lack of biogeochemical cycling [Ehleringer et al., 2000], these limitations can be more accurately understood as a transformation of N and C cycling caused by hyperaridity and extremely limited biological activity [Ewing et al., 2007]. This paper explores how hyperaridity changes the nature and rates of soil C cycling.
2.1. Field Sites
 Rainfall declines from south to north (decreasing latitude) in northern Chile. We selected three sites at latitudes ranging from 27 to 24°S (Figure 2) that comprise a rainfall gradient from 21 to <2 mm y−1 mean annual precipitation (MAP) [Warren-Rhodes et al., 2006]. Potential evapotranspiration (PET) levels for the region are ∼400–700 mm y−1, making rainfall levels less than ∼20–35 mm y−1 “hyperarid” (MAP/PET < 0.05). Thus these sites span the transition from marginally arid to hyperarid climate conditions. Sites were chosen to minimize differences in fog frequency (similar elevations and distances from the coast), parent material (dominantly Mesozoic granitic alluvium), and landform age (Pliocene terraces or alluvial fans) [Ewing et al., 2006]. Plant distribution corresponds to rainfall in the region [Latorre et al., 2002]. Three unidentified species of small (<20 cm) shrubs form a sparse (∼5%) but continuous cover across terraces and hillslopes at the arid site (Copiapó, 21 mm rain y−1). A single type of fog-adapted shrub (Cistanthe spp.) occurs only in hillslope positions and on terrace surfaces (∼5% cover) that intercept fog at the intermediate site (Altamira, ∼10 mm rain y−1). Plants are completely absent at the driest site (Yungay, <2 mm rain y−1).
 Soils were excavated to depths of ∼2 m, and the profiles were sampled by horizon. Samples were also collected from plant-associated coppice dunes and soil adhering to rocks visibly colonized by lithophytic cyanobacteria. Soil morphology and geochemistry at these sites are discussed by Ewing et al. , while trends in N cycling with increasing aridity are discussed by Ewing et al. . The age of the driest soil (Yungay) is between 1.9 and 2.1 My; the two southern sites are >2 My [Ewing et al., 2006].
2.2. Analytical Methods
 Soil samples were homogenized by sieving to separate the <2 mm fraction from the gravel fraction. A 100-g subsample was ground by hand using a corundum mortar and pestle. The total soil OC concentration and the Δ14C value of OC were determined using sealed tube combustion of about 1 g soil, following acidification to remove carbonate, as described previously [Ewing et al., 2006]. Samples with high chloride or very high sulfate were rinsed with deionized water in order to prevent tube explosions during combustion and to eliminate unwanted gases during post combustion sample processing. Graphitization of purified CO2 and subsequent analysis was done at the Lawrence Livermore National Laboratory Center for Accelerator Mass Spectrometry for most samples. The δ13C value of the CO2 in all samples, and the Δ14C value in 13 samples, were determined at UC Irvine's Keck Carbon Cycle Accelerator Mass Spectrometry Laboratory facility (analytical precision of 0.1‰). For radiocarbon, the long-term accuracy and precision (1 σ) of this technique on modern C (Δ14C > 0‰) is better than 9‰ [Vogel et al., 1984]. Laboratory blanks yielded a Δ14C value of −996‰.
 Total soil N (TN) was determined by automated combustion and gas chromatography (Carlo Erba), and inorganic N (IN) was measured in aqueous soil extracts by ion chromatography (NO3−) and automated colorimetry (NO3−, NH4+). Organic N (ON) was calculated as TN-IN, with uncertainty propagated using standard methods.
 Surface horizons were separately sampled for polar lipid fatty acid (PLFA) analysis. Samples were frozen (−20°C) upon collection in Chile, kept on dry ice during transport back to the lab, kept frozen (−20°C) prior to analysis, and processed within one week. Ten to 15 replicate 30-g extractions were made for each sample and pooled, so that the final pooled mass of soil analyzed was 300 to 500 g. PLFAs were extracted and purified as described previously [Macalady et al., 2002], and analyzed by gas chromatography-mass spectrometry (GC-MS) in the lab of T. Torok at Lawrence Berkeley National Laboratory. Peaks were identified using 33 bacterial FAME standards and MIDI peak identification software (Microbial ID Inc., Newark, DE) integrated with the GC-MS system. Microbial biomass was estimated by assuming 2.2 × 107 PLFA molecules/cell.
2.3. Model of C Transport and Isotopic Fractionation
 A model of OC transport and isotopic fractionation provides a means of evaluating hypotheses regarding major controls on OC in arid to hyperarid soils. The inventory and isotopic composition of soil OC is a function of inputs and outputs. Aboveground inputs result from biological activity (plant litter, photosynthetic soil microorganisms) and atmospheric deposition. In most settings, plant litter dominates aboveground inputs. Belowground inputs are primarily from roots. Losses include decomposition and dissolved transport.
 Mathematically, these processes are described by a general mass balance equation for the change in the soil OC inventory (Ctotal; mol m−2) with time (t; y):
where FAB is aboveground biological inputs (mol m−2 y−1), Fatm is atmospheric OC deposition (mol m−2 y−1), FB,total is total belowground inputs, k is the first order constant of decomposition (y−1), and Fout is dissolved flux out of the soil. Equation (1a) represents a simplification of the physical setting because (1) the fluxes are assumed to be constant with time, and (2) decomposition is assumed to be a first order loss from a single C pool, with a rate constant that is constant with time. The fate of OC entering the soil is dependent on net downward transport that can be broadly categorized as diffusive (Dd2C/dz2) and/or advective (νdC/dz), and that serves to spatially redistribute OC and make it either more or less available for loss processes. Incorporation of transport into the model is discussed below.
 The driest soil in this study represents a special case in which there is no dissolved flux out (Fout = 0). If the absence of plants means that the only OC input is atmospheric deposition (Fatm), and the starting OC concentration of the alluvial deposit is zero, the overall OC inventory for the soil profile reflects only the accumulation of OC from the atmosphere and losses due to decomposition:
In the driest soil in this study, a quantitative accumulation of atmospheric salts over the last ∼2 My provides direct evidence of the maximum depth of water movement [Ewing et al., 2006] and therefore an absence of dissolved OC losses below this depth. Thus in this soil, evidence of OC loss must reflect OC turnover (gas phase loss as CO2).
2.3.1. Steady State Values for the Driest Soil
 We assume that the inventory of OC to the maximum depth of water movement in the driest soil reflects the net accumulation of atmospheric OC over the last 2 My. If the soil OC inventory is at steady state, with constant inputs, then:
where 14C is an amount of radioisotope, 14Fatm is the input rate of 14C, and λ is the 14C decay constant (1.2 × 10−4 y−1). Combining equations (2) and (3), and solving for k gives [Wang et al., 1996a, 1996b]:
where R is the 14C/12C ratio in soil (Rs) or production (RFatm) and 12C ∼ C. The reciprocal of k is a steady state mean residence time for soil organic C, which can be calculated if the radiocarbon values of both the soil inventory and the inputs are known. In the driest soil, the calculated steady state k value can be combined with the soil OC inventory to calculate the mean atmospheric input rate using equation (2). We use a constant 1950 ratio for inputs (RFatm = 1). This is reasonable where measured soil Δ14C values are ≤ −300‰, which implies residence times of 1000 y or longer. The calculation and interpretation of residence times less than 1000 y requires accounting for production and incorporation of radiocarbon by atmospheric testing of nuclear weapons since 1950 (“bomb” carbon) [Trumbore, 1993].
 If radioactive decay of 14C is the only loss process (no microbial decomposition or other loss) and deposition remains constant, then the soil C inventory accumulates as an essentially linear function of the deposition rate (because the amount of 14C present is small and decay is slow):
If the radiocarbon input and decay rates are approximately equal (i.e., inputs are slow enough to achieve radiocarbon steady state in the timeframe of interest, as is the case in the driest soil), then:
Inserting equation (6) into equation (5), substituting the modern value for the radiocarbon content of inputs (Ratm = 1), and solving for the soil value gives:
For million-year timescales and slow OC inputs (“bomb” C influence negligible), equation (7) indicates that soil OC would be radiocarbon “dead” (>50 ky; Δ14C< −996). Thus for the driest soil, an OC inventory that is not radiocarbon dead must indicate that there is OC turnover that may be biological.
 In the Atacama Desert, extended dry periods help to lengthen the residence time of OC in soils. At the driest site, rainfall events of ∼1 cm occur approximately once per decade, and the only soil moisture in the intervening period is wetting of the top ∼1 cm of soil by fog. Larger events of ∼10 cm occur less frequently, perhaps once per century. If decomposition occurs only while the soil is still moist (t*/t = ∼7 days per decade, or 0.002 y/y at the driest site), then the observed value of k based on soil and input radiocarbon values (kobs) will represent the average rather than the actual rate constant of decomposition (k*) during the time when microorganisms are active.
2.3.2. Incorporating C Transport
 Accurate interpretation of depth profiles in soil OC requires consideration of OC transport. If Fz represents the net downward OC flux at soil depth z (mol m−2 y−1), then:
This assumes that the rate constant of decomposition (k) does not vary with depth. This is a reasonable simplification in this environment, where conditions are more likely to be similar with depth than in wetter environments [e.g., Feng et al., 1999]. Assuming that the concentration-dependent portion of downward flux is advective [Baisden et al., 2002; Feng et al., 1999], and further assuming both a constant flux by belowground inputs (FB = FB,total/w where w is the depth of wetting) and a constant velocity (v; cm y−1), the transport equation is:
Due to salt and dust accumulation, the driest soil has undergone nearly 100% physical expansion (by volume) over 2 My [Ewing et al., 2006], amounting to an annual increase in the soil height (h) by dh/dt = 0.00025 cm y−1. Averaged over the whole soil, this increasing height amounts to an effective downward velocity, x. The transport equation becomes (Figure 3):
Without belowground inputs (FB = 0), the simplified equation is:
2.3.3. Steady State With Transport
 If belowground inputs are assumed to be invariant with depth, the steady state concentration at depth z is (equation (10a) = 0):
The concentration will approach FB/k (belowground inputs relative to decomposition) with depth. In addition to the relative magnitude of belowground inputs and decomposition, the OC concentration with depth reflects aboveground inputs (as C(0)), and the relative magnitude of decomposition versus transport (k/(v+x)). Without belowground inputs, the steady state concentration at any depth (z) is:
Thus when decomposition and belowground inputs are negligible (k∼0, FB = 0), the steady state OC concentration is constant with depth, and equal to the surface concentration, C(0). With decomposition, an exponential decrease in OC with depth is observed. If belowground inputs are large enough, they will determine the nature of the depth trend; thus OC may increase with depth if belowground inputs are substantial and do not decrease with depth.
 Without belowground inputs, the steady state 13C concentration is:
This can be combined with the steady state C concentration (equation (11b)) to derive the steady state 13C/12C ratio:
This solution of the advective model is identical to the Rayleigh equation, applied to OC decomposition as a function of depth [e.g., Wynn et al., 2005]. It is essentially a special case for soil OC, applicable only to unique environments that have (1) no belowground OC inputs from roots or preferential flow, and (2) whose first order decomposition constant does not vary with soil depth (or for different OC pools). This equation shows that (1) without decomposition and belowground inputs, the 13C/12C ratio is invariant with depth; and (2) with downward transport and decomposition but without belowground inputs, δ13C values increase with depth (α < 1). Belowground inputs can change this trend depending on their magnitude and depth distribution. If they are constant with depth, the steady state 13R value is,
where RB is the ratio of belowground inputs. This value will approach RB/α at depth. Based on equation (11b) (FB = 0), the steady state radiocarbon value at depth z is:
where FM is the fraction modern C (1950 oxalate standard) as defined by Stuiver and Polach (1977). This equation shows that with downward transport of continuous surficial OC deposition, an exponential decrease in radiocarbon values with depth occurs, even with no decomposition. With belowground inputs, the steady state radiocarbon value at depth z is:
where FM(FB) is the fraction modern carbon in the belowground inputs. This value approaches [FM(FB)k]/(k+λ) with increasing depth. Thus the isotopic trends with depth depend on the relative magnitude of aboveground inputs (affecting C(0)), belowground inputs, decomposition (k), fractionation (α or λ), and transport (v+x).
2.3.4. Constraining Downward Transport and Belowground Inputs in the Driest Soil
 In the driest soil, the overall steady state value of k is known from the radiocarbon value of the soil inventory (equation (4)) and is assumed to be relatively invariant with depth. As discussed below, the radiocarbon data also indicate inputs of young OC to all depths in the driest soil. This could reflect rapid translocation of relatively unaltered OC along preferential flowpaths, or belowground primary production. We argue for the former. This belowground OC flux by preferential flow (FP, total) most plausibly originates as atmospheric deposition (Fatm). If the surface horizon (or any depth interval from the surface) is considered as a discrete compartment of thickness h (cm), its OC concentration and radiocarbon content can be used to calculate the magnitude of downward transport. The change in concentration with time in this surface compartment is (Figure 3):
At steady state with respect to time (which is expected to occur within the shortest time period in the surface horizon),
The transport term, T, describes the downward movement of OC due to advective flux and physical expansion of the soil. Solving for T,
Solving for v/h,
This term allows the rate of advective flux (vC(0)/h) to be quantified. Thus in the driest soil, Fp,total can be calculated from radiocarbon values (giving k, Fatm), mass balance (for x [Ewing et al., 2006]), and the time interval of wetting. The true value of v reflects the time interval of wetting (t*/t = 0.002y/y).
 For 13R with or without belowground inputs by preferential flow,
where α = Rlost/Rsubstrate. The 13R of the surface horizon will vary from that of the atmosphere only if T is less than or comparable to k.
 For radiocarbon,
Equation (19b) shows that the radiocarbon value of the surface horizon will vary from that of the atmosphere only if transport and decomposition (T+k) occur at rates comparable to radioactive decay (λ; 1.24 × 10−4 y−1). Combining equation (19a) with equation (16) and solving for Fp,total,
Equation (20) indicates that in the driest soil, the total amount of preferential flow bypassing the surface horizon can be calculated from the radiocarbon values and the inventory of the surface horizon, compared to those of the whole soil.
3. Results and Discussion
 In this section, we first evaluate depth trends and their implications at each site. We then discuss the three sites together, and consider inheritance effects on soil OC.
3.1. Results of Soil Analyses
3.1.1. Vegetated Arid Site (Copiapó, 21 mm rain y−1)
 The arid soil is located in a desert ecosystem broadly analogous to the driest portions of North America's Mojave Desert [e.g., Amundson et al., 1989a, 1996b]. There is sparse (∼5%) plant cover at this site, with three shrub species (not identified) up to 20 cm in height, and an ecosystem supporting an array of drought-adapted fauna. OC does not decrease with soil depth (Figure 4a), but is highly spatially variable at the soil surface (Table 1a). Compared to the surface horizon where plants were absent (0.051% OC), OC is elevated in plant associated coppice dunes (0.35%) and soil containing hypolith communities (0.48%), reflecting the patchy distribution of biological activity and soil nutrients typical of desert environments [Schlesinger et al., 1990]. The surface horizon δ13C value (−21.1‰; Figure 4b) is slightly higher than both the value for hypolith communities found under translucent quartz clasts (−25.6‰), and the value for coppice dune OC (−22.3‰), suggesting that this OC is relatively more decomposed, or that it includes a component of more 13C-enriched inputs. Radiocarbon values in the surface horizon (7 ± 44‰, n = 5; Figure 4c) are equal within error to hypolith OC values (5 ± 12‰, n = 2), indicating similar cycling rates (∼decadal, k∼0.1−1). A coppice dune was more enriched in 14C (Δ14C = 97 ± 2‰, n = 1), indicating more rapid cycling (values are closer to current atmospheric CO2, k∼1). Microbial biomass is two orders of magnitude higher in coppice dunes (1 × 106 cells g−1) and hypolith communities (3 × 106 cells g−1) than in the surface horizon (2 × 104 cells g−1) (Table 1a). Thus surficial OC cycling is spatially variable, but relatively rapid, at this site.
Table 1a. Measured OC, OC-δ13C, and OC-Δ14C Values at the Arid Sitea
Sample numbers: H indicates hypolith sample; P indicates plant sample; C indicates coppice dune; numbers indicate horizon designation; “total” indicates concentration-weighted, depth-integrated value.
One standard deviation of the mean.
15 (mol m−2)
 OC concentrations increase with depth in the upper meter, and are higher in the upper meter (0.034 to 0.203%) than in the lower meter (0.0068 to 0.014%), with the exception of the horizon at 200 cm depth (0.049%; Figure 4a). Roots were observed where OC is highest in the upper 107 cm, particularly below 15 cm depth. In the upper meter, the total OC inventory is 1.82 kg m−2, consistent with previous observations for desert soils [Amundson, 2001]. Higher concentrations in the upper meter indicate retention of OC and possible occlusion within pedogenic carbonate (IC), which is highest in this zone (0.37 to 0.83% IC between 22 and 112 cm depth [Ewing et al., 2006]).
 Δ14C values decrease rapidly with depth from the surface horizon (7 ± 44‰, n = 5) to a minimum of −966‰ at 77 cm. Contrary to typical trends, OC-Δ14C values are inversely correlated with OC concentrations (logarithmic; R2 = 0.87) below 10 cm depth, suggesting that OC is protected from decomposition, possibly by residing below regions of more frequent wetting and microbial processing. Notably, the zone where most roots were observed (15–107 cm) is also the zone of highest OC concentrations and low Δ14C values (Figures 4a and 4c). This indicates long-term preservation of root inputs and possible preservation of OC from a previously wetter climate. A fit of the observed data using equation (11a) (Figure 5a) and equation (14b) (Figure 5c) with estimated parameters (k = 0.1 to 0.01, v/h = 0.2 to 0.7, FB = 10 to 70 mmol m−2 y−1; Table 2) matched the radiocarbon data only when the fraction modern value of belowground inputs was ∼0.05 (Δ14C = ∼−950‰). Since inputs are unlikely to be this old, the depth trends in OC (increasing with depth) and radiocarbon (decreasing with depth) likely reflect current OC dynamics superimposed on relict OC from a previously wetter climate. This is consistent with growing evidence of wetter episodes that have occurred periodically during the past 2 My at this latitude [e.g., Lamy et al., 2000; Latorre et al., 2002; Stuut et al., 2006].
Table 2. Parameters Used in Modeled Depth Distributions of Soil Organic C (Figures 5–7)
mmol m−2 y−1
mmol m−2 y−1
δ13C Fb (max)
δ13C Fb (min)
δ13C z = 0
FM z = 0
 The OC-δ13C values generally indicate variable inputs and decomposition over time. They are weakly inversely correlated with Δ14C values (increasing δ13C values with decreasing Δ14C values; R2 = 0.14; Figures 4b and 4c). These δ13C values may have been increased in the past by more prevalent C4 plants, although long-term (40 ky) records indicate very limited C4 influence [Latorre et al., 2002]. Pulsed wetting may also move small amounts of OC downward and increase residual OC-δ13C values via decomposition. When these values are modeled using present decomposition rates and estimated transport parameters (Table 2), the result shows only a one or two ‰ effect on δ13C values (Figure 5b). This is consistent with previous results for semiarid soils [Feng et al., 1999], but the effect of slow decomposition by longer-term episodic wetting of relict belowground OC is unknown. In the lower meter where δ13C values are highly variable (δ13C = −23.7 to –14.1‰; Figure 4b), Δ14C values are somewhat more elevated (Δ14C = –462 to −781‰; Figure 4c). This suggests occasional transport of younger OC from the surface, as no roots were observed in the lower meter. Overall, results indicate that OC depth trends in this soil reflect long-term climate variability, with conditions that are currently more arid than in the past, and present belowground inputs that are small compared to inventories preserved in the soil over ky timescales.
3.1.2. Intermediate Site (Altamira, ∼10 mm rain y−1)
 At the intermediate site, frequent fog supports widely scattered stands of a single type of highly fog-adapted plant (Cistanthe spp.). As observed at the arid site, coppice dunes at the intermediate site also have somewhat higher OC than the surface horizon (∼0.20%; C isotopes not measured), as does hypolith-associated soil (0.25 ± 0.04%) (Table 1b). Plant material had a mean Δ14C value of 119 ± 15‰ (n = 2), while slower cycling is indicated by values for hypolith-associated OC (−21 ± 5‰, n = 1), and the surface horizon (−39 ± 69‰, n = 3) (Table 1b). Microbial biomass is low but variable–two orders of magnitude higher in coppice dunes (9 × 105 cells g−1) than in the surface horizon (6 × 103 cells g−1) (Table 1b). Spatially variable biotic activity persists at this site, but is less active and pronounced than at the arid site, in terms of both cycling rates and biomass.
 Maximum radiocarbon values at the surface generally indicate surface focused C cycling, but at slower mean rates (1/k∼102 y mean residence time) than at the arid site (1/k∼1–10 y mean residence time). These k values can be compared to belowground OC concentrations (C(z) = FB/k at steady state with respect to depth; equation (11a)) to estimate the magnitude of belowground inputs (FB) based on the assumptions of the transport model. Since belowground OC concentrations are low (∼102 mmol m−2 cm−1; Table 1b), these k values indicate belowground inputs (FB ∼1 mmol m−2 cm−1) that are an order of magnitude less than at the arid site (FB = ∼50 mmol m−2 cm−1) (Table 2).
 With increasing depth in the soil, OC is consistently low in concentration (0.014 to 0.023%, Figure 4a). The total soil inventory is 0.25 kg m−2 – an order of magnitude less than that of the arid soil. Radiocarbon values decrease with depth from a maximum at the surface (∼39 ± 69‰, n = 3) to a minimum of –806‰ at 89 cm (Figure 4c). Plant material is slightly more enriched (Δ14C = +119 ± 15‰, n = 2) than suggested by the coppice dune soil value at the arid site (Δ14C = +97 ± 2‰, n = 1), suggesting a ∼decadal residence time in plants. While OC concentrations are generally low, they are inversely correlated with OC-Δ14C values (logarithmic; R2 = 0.72). A modeled fit to the OC and radiocarbon data (Figures 6a and 6c) using estimated parameters (k = 0.01 to 0.001, FB = 0.3 to 1.5 mmol m−2 cm−1; Table 2) in equations (11a) and (14b) required a radiocarbon value for belowground inputs of FM = 0.200 (Δ14C = ∼−800‰). This suggests preservation of older OC from root inputs, as observed at the arid site, but to a more limited degree.
 At the same time, this soil includes large (∼30 cm wide) subsurface polygonal sulfate blocks and deep (∼1 m) cracks rich in carbonate [Ewing et al., 2006], suggesting that belowground inputs via preferential flow have also occurred in this soil. The pitted, undulating tops of these blocks indicates alteration by significant rainfall events. These blocks are overlain by smaller, low-density gypsum prisms that indicate more recent hyperarid conditions. Thus oscillating wet/dry conditions have contributed to the observed OC trends, leading to a belowground pool that may reflect both past root inputs, and variable inputs by preferential flow.
 OC-δ13C values that are variable with depth (−14.3 to –26.6‰; Figure 4b) and uncorrelated with soil Δ14C suggest variable OC-δ13C input values over time (Figure 6b). The OC-δ13C value of basal stem material from plants at this site is −19.3 ± 0.1‰ (n = 2), while the value for soil containing hypolith communities is –23.7 ± 0.2‰ (n = 1). This range in stable isotope values approximates that of the soil, with a lower maximum value. Atmospheric deposition likely contributes to the soil values given the moderate turnover rates implied by the surface horizon (∼102 y). With this residence time, the surface horizon inventory implies a net input rate of 200 mmol m−2 y−1, an order of magnitude lower than the atmospheric deposition rate implied by the driest soil (103 mmol m−2 y−1; see next section). This suggests that downward dissolved transport of OC below the surface horizon has occurred, and that the total soil inventory values reflect a mix of plant and atmospheric inputs with limited decomposition. The highest OC-δ13C values are at the surface (−14.3‰) and at 89 cm depth (−14.9‰) (Figure 4b). The 13C enriched surface value may reflect isotopic enrichment through photochemical degradation of plant material, or isotopic enrichment of leaves and other material entering the soil, compared to the stem material analyzed. The high value at depth, where the Δ14C value is lowest (−806‰) and the carbonate concentration is highest (0.98%), suggests 13C-enriched OC is occluded within the carbonate-sulfate matrix, and is similar (with respect to stable C isotopes) to material at the soil surface. Sporadically higher values at greater depths likely reflect sporadic influxes of surface OC via preferential flow along cracks. The modeled effect of isotopically variable belowground inputs is shown in Figure 6b. In summary, OC cycling is slowed in this soil relative to the arid site, belowground inputs occur via preferential flow and limited roots, and the presence of a few plants has focused OC cycling at the soil surface.
3.1.3. Driest Site (Yungay, <2 mm rain y−1)
 The most hyperarid soil is located in a broad region virtually devoid of plants. Exceedingly rare hypolithic cyanobacteria are present in widely dispersed locations [Warren-Rhodes et al., 2006], indicating that primary production is quite limited and that atmospheric deposition must dominate OC inputs [Ewing et al., 2007]. OC is low throughout the upper meter (0.012 to 0.023%; Figure 4a), resulting in a total inventory of only 0.19 kg m−2 – equal within error to that of the intermediate site and lower than previously observed soil inventories on Earth [Amundson, 2001]. In the upper meter, Δ14C values are well in excess of the detection limit (−996‰; “radiocarbon dead”; minimum age >40 ky), ranging between –365‰ and −696‰ (Figure 4c).
 The well-cemented halite layer at 134 cm depth indicates the maximum depth of dissolved transport during the last 2 My, and is evidence that the underlying zone is effectively locked out of modern surficial processes. In the underlying layers, radiocarbon concentrations are below the detection limit, and higher OC (0.094 to 0.270%, Figure 4a) likely reflects OC accumulation during wetter climate conditions prior to deposition and stabilization of the present alluvial surface (2My [Ewing et al., 2006]). Radiocarbon values for root fragments in these layers also were below detection.
184.108.40.206. Decomposition Based on Radiocarbon
 To interpret the radiocarbon data, we assume that the soil is at steady state. A steady state assumption for OC with time is reasonable in this hyperarid location because small rainfall events have occurred repeatedly and relatively frequently (e.g., approximately decadally) over the exposure time of the landform (2 My). Climate variation during this period is thought to have been limited, although exact variations are unknown. Assuming steady state and modern inputs (equation (4)), the OC inventory (46 mol m−2) and total radiocarbon value (−777‰) to the base ofthe cemented halite layer indicate a mean residence time (1/k) of 29 ky (k = 3.6 × 10−5 y−1; equation (4)) and an atmospheric deposition rate of 1.64 mmol m−2 y−1 (equation (2)). If we assume decomposition only occurs during times of water availability, then the effective time for decomposition (t*) is much shorter than the total time of soil formation (t), and we can calculate an effective rate constant (k* = kt/t*) is 0.018 y−1 (t*/t = 0.002 y/y as discussed previously). Thus while apparent surface cycling rates are exceedingly slow, the actual rate of activity during infrequent wetting events is more comparable to rates found at wetter locations.
 Because there are no dissolved OC losses from the driest soil, the radiocarbon values indicate that OC turnover must occur in this soil (see discussion of equation (7)). This is consistent with the presence of a small microbial biomass (103 cells/g) in the surface horizon. A higher radiocarbon value was obtained for surface horizon soil containing a cyanobacterial community in this region (−281 ± 8‰, n = 1), indicating that more rapid soil C cycling occurs in the very rare locations where these communities are found. [Warren-Rhodes et al., 2006]. Higher rates of cycling in these communities are expected due to the ability of cyanobacteria to fix CO2, and their ability to access condensed fog waters around stones, facilitating both fixation and decomposition and driving a more vigorous OC cycle [Warren-Rhodes et al., 2006]. Yet even at locations devoid of these communities, we find that OC cycling slows but does not stop in this hyperarid environment [Ewing et al., 2007].
220.127.116.11. Nature of Downward Transport Based on Radiocarbon
 In this soil, unlike the other two soils examined in this study, the radiocarbon values of OC do not decrease with depth. This indicates belowground inputs of relatively young OC. Higher Δ14C values in horizons at 55 cm (−480 ± 6‰, n = 1) and 92 cm (−356 ± 8‰, n = 1) (Figure 4c) suggest stochastic downward water movement that transports younger OC deeper in the soil.
 We can calculate the flux of downward moving C using the value of Fatm, and the radiocarbon value and inventory of the surface horizon, in equation (20). The calculated flux Fp,total is 1.62 mmol m−2 y−1. This is essentially equal to modeled atmospheric deposition (1.64 mmol m−2 y−1), indicating that most deposited OC is transported to greater depth relatively rapidly in this soil. The long-term average magnitude of downward transport from the surface horizon is two orders of magnitude lower, with rates of 1.0 × 10−2 mmol m−2 y−1 (v/h = 4.3 × 10−5; equation (17b)). During infrequent wetting events (t*/t = 0.002 y/y), the rates of downward advective transport and decomposition loss from the surface horizon are similar at ∼5 mmol m−2 y−1, while preferential flow moves 180 mmol m−2 y−1. Thus our model indicates that most atmospheric inputs are incorporated belowground by repeated, infrequent events of rapid preferential flow, each followed by a pulse of decomposition. These events occur at intervals of decades to centuries, and thus elude observations made by most researchers in this region. Calculation of preferential flow rates with increasing depth (Table 1c and equation (20) for discrete compartments to each depth) suggests that inputs from this source are relatively uniform with depth, and exceed losses to decomposition and advective downward transport. Modeled depth trends based on k values between 1.2 × 10−4 and 1 × 10−5y−1, and slightly higher transport values (v/h = 0.0001 to 0.0004 y−1) indicate that small changes in the decomposition constant may explain the observed variation in concentrations (Figure 7a) and radiocarbon values (Figure 7c) with depth.
18.104.22.168. OC-δ13C Values With Depth
 In this extremely arid environment, the soil OC-δ13C values vary less than in the more humid soils–between −22.7‰ and −27.8‰ (Figure 4b). The highest and lowest δ13C values occur in the first (1–2 cm depth) and second (2–3 cm depth) horizons, respectively. The δ13C value for the surface horizon (−22.7‰) is consistent with the expected value for OC derived from fixation by marine phytoplankton at this latitude [Rau et al., 1989], and thus could reflect sea-salt associated OC derived from the productive upwelling environment off the coast of northern Chile [Ewing et al., 2007]. The large decrease in isotope values over a short vertical distance is followed by an increase to –24.8‰ at 33 cm depth. The pronounced difference in isotope values of the upper two horizons has also been observed in the stable isotope composition of nitrate-N, Ca, sulfate-S, sulfate-O and nitrate-O in this soil [Ewing et al., 2007, 2008]. For Ca, S and O, the trend is driven by isotopic fractionation with small, repeated precipitation of CaSO4 minerals. For OC, fractionation of stable isotopes with dissolution and adsorption is unlikely (J. Sanderman and R. Amundson, manuscript in preparation, 2008). Thus the mechanism of mass dependent isotopic fractionation of OC must be somewhat different, but also must result from water-mediated downward transport. It must occur repeatedly over short timescales with infrequent rainfall events, since the radiocarbon data do not indicate increasing OC residence time with depth in this zone.
 In the second horizon and below, the increasing OC-δ13C values with depth implies preferential loss of 12C (e.g., through decomposition). Preferential loss of 12C through decomposition is consistent with detectable radiocarbon in the upper meter, as discussed above, as well as observation of low but detectable microbial biomass in the surface horizon of this soil [Ewing et al., 2007] (Figure 8d) and in the shallow subsurface of soils in the region. However, the loss of 12C with decomposition cannot explain the decrease in δ13C between the first and second horizons. One plausible explanatory hypothesis for this trend is photochemical decomposition of OC in the surface horizon [Austin and Vivanco, 2006]. If photodegradation favors the mobilization of light C isotopes into the dissolved phase [Opsahl and Zepp, 2001], shallow downward transport (1 cm) of a small amount of OC during fog events could produce the sharp contrast in δ13C values between the surface and second horizons. An increase with depth could occur through mixing of this photochemically mobilized pool (∼−28‰) with belowground inputs resembling inputs at the surface (∼−25‰), as well as be subsequent 13C enrichment of the photochemically altered pool by microbial decomposition.
 Modeled depth trends in OC−δ13C values for this soil using equation (13b) with a surface input value of −28‰, and a belowground input value of −25 to −27‰, are illustrated in Figure 7b. Isotopic fractionation with decomposition is estimated to be 1‰ [Baisden et al., 2002] (other parameters shown in Table 2). Increased advective transport (k/v ∼ 0.1 rather than calculated value of 1) reproduced the observed depth trend in the upper horizons (Figure 7b). Thus photochemical mobilization of isotopically light OC, and belowground decomposition and mixing with downward transport, could lead to the observed depth trend in OC-δ13C values.
22.214.171.124. OC/ON Ratios With Depth
 OC/ON ratios vary from 0.1 (92 cm) to 3.4 (2.5 cm) in the driest soil (Table 1c and Figure 9). The highest values (2.5 to 3.4) are in the upper horizons where the distinctive depth trend in OC-δ13C values also occurs (Figure 9). The lowest value (0.06 ± 0.02) is in the horizon at 92 cm, where the highest radiocarbon value was observed (Δ14C = −356‰, Figure 4c), indicating that this layer reflects downward transport of relatively young OC. The OC/ON ratio in the surface horizon is also low (1.0 ± 0.4) relative to typical miniumum soil organic matter values (∼5), which are generally interpreted to be equal to the C/N ratio in microbial biomass.
 The measured values in this soil suggests that photolysis at the soil surface may favor removal (by downward transport) of OC over ON. Irradiation of natural waters by sunlight leads to production of low molecular weight carbonyl compounds lacking N (which may be mineralized to NH4+) that are then highly biologically labile [Moran and Zepp, 1997]. In the upper subsurface horizons, a periodically active microbial population may maintain higher OC/ON values by selectively accessing mineral N and carbonyl-C. As a result, a residual pool of photodegraded organic matter may be transported to greater depth, leading to the exceedingly low overall OC/ON ratio (0.54 ± 0.06) in this soil.
3.2. Biogeochemical Threshold and Climate Legacy
 Soil OC inventories and turnover in surface horizons and in the upper meter decrease precipitously with the transition from arid to hyperarid conditions (Figures 8a and 8c). Overall OC-δ13C values decrease between the arid (−18.5‰) and hyperarid (−25.3‰) sites (Figure 8b), and OC/ON ratios decrease by two orders of magnitude (from ∼20 to 0.5), reflecting greatly reduced photosynthetic activity by plants. In the driest soil, the low OC-δ13C values and OC/ON ratios may also reflect retention of isotopically depleted, N-rich photodegradation products. The decline in C cycling corresponds to decreases in soil microbial biomass by an order of magnitude (Figure 8d), with very low cell concentrations in the surface horizon of the driest soil (1000 g−1) compared to most soils (109 g−1). This decrease also limits the production of CO2 through decomposition, and consequently the inventory of IC also decreases with decreasing rainfall (Figure 8a).
 While all of these soils exhibit C turnover – albeit with broad variation in rates – all sites arguably bear the legacy of environmental change in their OC inventories and isotope values. Belowground radiocarbon values are depleted at the wetter sites as a result of preservation of soil C (Figures 4c and 8c). This belowground OC is likely the result of root inputs during past intervals of increased rainfall. Radiocarbon depleted OC in subsurface horizons has been observed in semiarid desert soils subject to climate change during their formation [Wang et al., 1996a, 1996b], though the processes responsible for C isotope trends in these other soils are not well understood.
 At the hyperarid extreme, this imprint of past climate has been buried in accumulating atmospheric salts and dust, and a transformation of the soil C cycle is observed. OC supply is no longer a function of in situ photosynthesis, but rather atmospheric deposition reflecting external biological activity. OC cycling persists belowground as a result of delivery of dissolved and photodegraded OC to belowground microbial populations that are likely highly adapted to the unique circumstances of high salt concentrations, prolonged drought, and limited availability of OC. Hyperarid soil C cycling is driven by decadal to millennial rainfall events—hydrologic conditions that are extremely unusual and difficult to observe at the timescale of a human lifetime.
 Despite pronounced within-site heterogeneity of OC inputs, and the long time period of hydrological history reflected in these soils, striking isotopic patterns with depth are apparent, as are gradients in microbial biomass, OC/ON, and the isotopic character of the soil OC inventory. These results point to the long-term stability of arid conditions in the region, compared to other arid regions on Earth, and the persistence of the north-south climate gradient over time.
 Our examination of soils along a rainfall gradient in Chile begins at the dry end of previous studies of soil C cycling, and reveals unusual and complex changes of the soil C cycle (Figure 10). The amount and isotopic composition of soil OC at our arid site resembles soils found in North American deserts capable of supporting plants [e.g., Amundson et al., 1989a]. Rapid cycling occurs in spatially segregated patches at the soil surface, and plants control the supply of OC to soils. Belowground, root inputs combine with limited downward transport and decomposition to preserve aged OC in the shallow subsurface, even as climate conditions change.
 The driest soil we examined in the Atacama Desert supports a novel C cycle that presently appears to operate via slow delivery of atmospherically derived OC, surficial biotic and abiotic degradation, and infrequent downward transport and microbial decomposition of this influx during rare precipitation events. Over millions of years, these events have produced unique depth profiles that indicate the importance of rare rainfall events. A distinctive variation in stable isotope ratios of many elements with depth is observed in the uppermost soil horizons [see Ewing et al., 2007, 2008], including 13C/12C. In the case of OC, this distinctive variation with depth may indicate photochemical degradation at the soil surface, and limited transport and decomposition of labile OC with ephemeral wetting events, each step involving isotopic fractionation. Larger rainfall events may very occasionally transport some decomposition products to relatively greater depths. Despite extreme aridity, the soil C/N ratios indicate that when water is present, a small soil microbial biomass is active, but limited by the low OC substrate availability. Thus, in this hyperarid environment, biogeochemical cycling is not only limited by water; it is transformed by a combination of highly adapted biological systems and abiotic processes. Ancient hyperarid soils contain a record of distinctive biogeochemical processes over time, and can provide critical evidence for understanding both climate history and biological persistence at the arid extreme.
 Funding for this work was provided by grants from NASA-Ames and NSF, and the NASA Graduate Student Research Program and UC Dissertation Year fellowships. We thank G. Dos Santos and staff at UC Irvine's Keck AMS facility for C-14 and C-13 analysis, and M. Kashgarian and staff at LLNL's AMS facility for C-14 analysis. Thanks also to C. Latorre for plant identification at the Altamira site, B. Sutter and B. Gomez-Silva for field assistance, J. Wu for laboratory assistance, C. Lewis for modeling discussion and encouragement, and two anonymous reviewers for useful comments on the manuscript.