4.2.1. How Does Soil Redox Affect P?

The relationships between soil redox factors (precipitation, soil moisture, and soil drainage) and soil P contents suggest that moisture is not a dominant control on soil P. While there is a weak possible correlation between MAP and P (Supplemental Figure S3A), the range of P concentrations for a given MAP is large (ca. 1500 ppm) and is not particularly predictive. Adding more climate data (e.g., correlating continental climate data with the USGS sample locations) could improve this potential relationship. As with other possible relationships in soils, however, it could be mediated by an additional soil-forming factor (i.e., precipitation is related to weathering intensity, and possibly to the presence of vegetation).

There exist only weak patterns between P and soil moisture or drainage. Udic soils have higher mean and ranges of P than other moisture regimes (Figure 8D), and poorly-drained soils can have higher P than well- or excessively-drained soils (Figure 8B). However, the bin sizes for these soil moisture regime types are relatively small; more data could clarify these trends. A relationship between soil P concentration and moisture regime would not necessarily be expected because soil moisture regimes are defined by the number of saturated days per year [83], but not necessarily if those days are linked by season (i.e., they do not need to be consecutive). While soil redox state would affect the (short-term) phase of P adsorption sites (such as Fe oxides) and possibly the accumulation of organic matter, which can mediate P-oxide interactions [8,9], as well as sequester P in organic compounds, only precipitation and its links to weathering intensity would be expected to exert long-term control on P concentrations.

#### 4.2.2. How Does Soil Redox Affect Fe Concentrations and Speciation?

Our primary hypothesis was that precipitation would exert control on soil moisture and therefore on redox conditions in the soil porewaters, controlling Fe speciation. However, mean annual precipitation did not correlate with total Fe and showed none of the expected correlations with the four Fe species. There was perhaps a maximum precipitation (ca. 2000 mm yr<sup>−</sup>1) for most Fe in carbonates (Figure 10B), but little else. Turning to a more direct approach of constraining soil moisture—soil moisture regime—revealed one interesting correlation: although variability in total Fe was low across soil moisture regimes (Figure 9C,D), there are some patterns of trade-off between species within a moisture regime (Aquic, Udic). Similarly, variability between soil drainage was low, but a general pattern of moderately-drained soils having higher Fe in oxides and carbonate, whereas labile Fe (reduced) showed no trend. Based on these data, soil moisture, redox state, and associated chemistry are highly heterogeneous even within soil orders, moisture regimes, and drainage ability, and broad generalizations cannot be made.

Iron oxide mineralogy and soil magnetic properties have been linked to regional precipitation [84–87] and soil moisture [42,43,88,89], but relatively few studies explore the links between soil iron oxide mineralogy and soil order [90,91]. Cervi et al. [90] found parent material and climate to be primary controls on soil magnetic susceptibility. A recent survey of Australian soils explored 471 unaltered topsoils' magnetic properties [92], which was the only study we could find of a similar scale to the work presented here. Hu et al. [92] found parent material to be the strongest control on soil iron mineralogy, with climate (precipitation, temperature) showing weaker influence and vegetation effects on soil iron being regional.

In addition to using the large soil datasets that are increasingly available, more targeted studies on the influence of soil moisture and drainage on redox conditions—using larger datasets and including more detail of soil texture—could strengthen these trends and reveal additional patterns. Many of the soils included in this study were well- or very-well drained, with a smaller subset being poorly- or very poorly-drained. It is possible that relationships between Fe species and moisture are stronger in soil drainage regimes for which we have effectively too small datasets in this study. Additionally, highly localized slopes (along with the context of grain size composition) likely exert strong controls on soil moisture and redox state, but we were not able to expand this study to include those data. Doing so would help clarify potential relationships between soil moisture and Fe species. Using high-resolution slope data (e.g., fine-scale LiDAR/digital elevation models) could be useful for determining these relationships between local soil redox conditions and geochemistry.

#### *4.3. How Do Fe Species Associate with P?*

We expected that Fe oxides, having been shown to efficiently sorb P onto their surfaces in other soils (e.g., [8,9,44,93]), would have a more robust correlation than between either labile Fe and P (in reducing microenvironments) or Fe in carbonates and P. However, no species showed a strong correlation with P (Figure 9). Overall, Fetot and P in all soils showed a positive correlation, suggesting that Fe species is not a dominant control on P concentration, and that no single species is more likely to be associated with P, at least on the regional scales examined in this work. Because of the weak relationships between Fe species and assorted soil moisture proxies, as well as a surprising lack of correlation with clay content, there is no simple rule connecting Fe species, soil redox state, and P. However, an increase in any Fe-bearing species should correspond to an increase in P due to the general positive relationship (Supplemental Figure S4). Fe (and Al) oxides in the clay fraction specifically may be most important for terrestrial P transport (both in soils and fluvial systems) (e.g., [39,44,94]). We expect that repeated analyses within an individual profile or on a small local scale (i.e., meters) would show stronger expected relationships, but they cannot be generalized to regional or continental scales.

#### *4.4. How Does Vegetation A*ff*ect P and Fe Concentrations?*

Biology—here, generalized as dominant vegetation on the soil surface (not considering microbial life and mesoscale organisms)—plays an important role in terrestrial P cycling. We found that Barren soils' B horizons had more P accumulation than other, vegetated soils' B horizons, reflecting plants' effectiveness at moving P from regolith to higher in the soil profile ([15,16]; see Section 4.7.4). The other vegetation types showed surprisingly little variability, but P does tend to increase up-profile, suggesting that once a landscape is vegetated—regardless of specific ecosystem type—P is efficiently mobilized by plants to the upper horizons and recycled and stored there. Additionally, vegetated soils had far higher ranges/maximum P concentrations than unvegetated soils, supporting the importance of plants' role in P mobilization and accumulation. Barren landscapes lack the rooting systems to mobilize and transport P, hence the increased accumulation in B horizons. Halsted and Lynch [95] found essentially no discrimination in P uptake between C3 and C4 plants, which our results further support.

Aside from 'Other' vegetation (which was biased by Fe-rich parent material), Cultivated soils' B horizons were the only ones that showed depletion, suggesting that ecosystem variability is less important than parent material and soil mineral accumulation rates for Fe in soils. Deep tilling in agricultural practice could be an additional factor in lower-soil-profile nutrient depletion, but it is less common than it used to be. Similarly, vegetation type does not appear to be a direct control on Fe species within soils; Forests did have greater Feacet, which is likely linked to the defined property of Alfisols of high cation exchange capacity and cation mobility, as well as their tendency to be well-drained.

Overall, the presence of vegetation—rather than differences in the type of vegetation—exerts the most control on P accumulation in soils (Figure 2). Although the mean P values were similar between vegetation types within a soil horizon (with the exception of B horizons in Barren soils), non-Barren (vegetated) soils had higher ranges of P than Barren soils. Forests also have higher CIA values than grasslands and shrublands (Figure 7), suggesting that the evolution and spread of different ecosystems could have changed the distribution of weathering intensity. This has interesting implications for the ongoing debate on land plants' influence on continental P fluxes (see Section 4.7.4).

Vegetation changes with latitude; the soils and ecosystems represented in this work span tropical and temperate forests, continental grasslands, dry shrublands, and high-latitude, little-vegetated landscapes. This inherently contributes to the observed latitudinal trend in weathering because different plants have different interactions with soil and bedrock. While there are ground-level differences between plants within a plant type (e.g., grasses), the functional difference between plant functional types (e.g., grasses vs. trees) are far larger and more meaningful on continental scales. While at local or regional scales, high-resolution, species-specific studies on weathering and plant-soil relationships can be done, that level of analysis is neither possible nor necessary for continental-scale analysis of vegetation-weathering relationships. Global climate, biogeochemical, and environmental change models rely on continent-scale data; therefore, while locally, species variability within a plant functional type may lead to nuanced plant-soil geochemical relationships, the continental scale and umbrella vegetation type is appropriate for this style of inquiry.

#### *4.5. Is Soil Order Predictive of P and Fe Concentrations?*

Soil order is not quantitatively predictive of P and Fe concentrations. Soil order, Fe, and P are linked indirectly through the correlation between Fe/P and CIA; this relationship is stronger for Fe, which shows a clear Fe accumulation increasing with weathering (Figure 6A,C), which is also linked to the Inceptisol to Ultisol soil progression common in chronosequences. Soil order and weathering show a very well-behaved and expected relationship, with CIA values increasing from Entisols to Ultisols/Oxisols (Figure 5). An exception to this trend is high-latitude Gelisols, which have longer formation times and lower weathering rates. These results align with existing soil formation frameworks and support proxies that rely upon their use, such as the CIA-based mean annual precipitation proxy, the paleosol weathering index for mean annual temperature, and Bt thickness ([79] and refs. therein). However, because soil order depends on a variety of factors (i.e., weathering stage, vegetation), quantitative relationships between soil order and either Fetot or P concentrations cannot be generalized, and samples within one soil order should not necessarily be used to predict Fe or P concentrations quantitatively for other samples of the same order. This is important to remember when using limited numbers of samples of fossil soils (see Section 4.7.1).

#### *4.6. Implications for P in Modern Soils, Climate Change, and Soil Fertility*/*Food Security*

#### 4.6.1. Soil P, Erosion and Transport, and Human Activity

Aside from the geologically-recent uptick in anthropogenic influence on landscapes, continental scale-vegetation was relatively stable over the past millennium in North America (e.g., [96]). Concern for changing P fluxes in terrestrial ecosystems largely stems from the occurrence of harmful algal blooms in rivers, lakes, and coastal areas, often linked to mass-production agriculture and nutrient-rich runoff. The implications for these fluxes from this work center on land use change and soil degradation, as well as climate change-driven shifts in vegetation (e.g., grassland to barren). P would also be lost in that shift because vegetation loss exacerbates erosion and landscape degradation [97].

Human-induced changes (e.g., industrial agriculture, extensive monoculture, unsustainable farming practices, etc.) can lead to natural P being lost through subsidence and erosion, P being added through fertilizer, and reduced plant P retention if natural, diverse ecosystems are replaced with monoculture (e.g., [98]). Land-use changes typically result in the removal of the upper portion of a soil profile (various, e.g., [99]). Because P tends to accumulate in the upper horizons (O,A), our results suggest that this could lead to a noticeable loss in P and in agricultural soils, increased need for fertilizer (see Section 4.6.3). Land-use changes in vegetated soils could increase the P flux from land to lakes and coastal waters, leading to eutrophication. The spread of Aridisols through desertification could mean more regions experience erosion due to vegetation loss, but perhaps in turn lower weathering due to that loss of vegetation and to drier climatic conditions [100]. As desert regions expand, global dust volumes and distributions will change, which would change which areas see P removal (through increased erosion and P loss through dust) and which see P accumulation through dust deposition. Changes in atmospheric and soil CO2 concentrations are likely to affect both plant and soil productivity (as well as mycorrhizal efficiency) under climate change, although the extent and duration of changes to productivity are debated [100–103].

#### 4.6.2. Weathering, Climate Change, and Soil P

In addition to changes in vegetation (ecosystem), climate change will impact weathering intensity and erosion, affecting global soil P reservoirs. As a consequence, soil fertility and food security risk will shift and vary depending on how different regions and biomes respond to climate change [104]. Currently, there is a general trend of intense, rapid weathering at lower latitudes as a consequence of high heat and precipitation. Higher latitudes tend to be cooler and drier, and therefore, weathering intensity and rates are lower [14,24,105]. While low-latitude areas are expected to undergo desertification (decreased precipitation, loss of vegetation), higher latitudes may experience an increase in precipitation and temperature. Therefore, the style and intensity of weathering will change regionally. Based on our results, where the range in P concentrations peaks at mid-intensity weathering (i.e., CIA~60; Figure 5) and decreases in older and/or more strongly-weathered soils (as expected), the climate change expected

for higher-latitude regions may result in decreased soil fertility (limiting agricultural production, see Section 4.6.3) and increased risk of eutrophication in surrounding waters. The former effect could potentially be mediated by vegetation, which holds P and slows erosion.

Except for the increase in range at mid-intensity weathering, P concentrations were relatively consistent across weathering intensities, which was unexpected. We had hypothesized that as older soils tend to be depleted in P, a higher weathering value would correspond to lower P. This was not demonstrated by our results. We interpret the lack of correlation as highlighting the significance of soil age—rather than weathering intensity (climate)—in controlling soil P, as some previous literature has suggested (e.g., [19,106]). A nuance that this large dataset could be missing is P speciation, which could still show P phase-specific variability with weathering intensity. The implication for P transport from soils remains unchanged from previous studies, where older soils would likely have smaller pools of P for mobilization. Mapping soil P with soil age on a large scale (such as the USGS dataset used here) would be a valuable new addition to the field and could elucidate the P-soil age-weathering puzzle. Additionally, exploring relationships between soils and weathering with species-level variability (on smaller, local scales) could help refine interpretations based on dominant regional plant functional type.

The difference in weathering intensity/CIA values between Forests and other natural vegetation types (Grasslands, Shrublands), as well as the high CIA values in Altered/Cultivated soils suggests that if more land is converted to farming (cultivated) from natural grasslands (e.g., the Great Plains) or if agricultural intensity increases, weathering rates would increase, depleting the soils' nutrients more quickly and perhaps increasing P fluxes. To our knowledge, this is the first data-based demonstration of the principle on the continental scale rather than as a model result. Understanding how human activity and agricultural degradation of soils affects weathering and resulting P is crucial for predicting changes in terrestrial P.

#### 4.6.3. Soil P, Plants, and Agriculture

Climate change is also likely to affect plant physiology and biological processes, such as P uptake (mediated by mycorrhizal fungi) and C storage, as a result of changing atmospheric carbon dioxide concentrations (*p*CO2). Plants rely heavily on symbiotic relationships with mycorrhizae to mobilize and uptake recalcitrant mineral P and fix nitrogen [102,107–109]. Mycorrhizal activity and P uptake can also be affected by soil moisture [110,111], which depends on a number of factors (discussed in Section 2.2) that could change in response to altered land use and climate change. P uptake efficiency also varies between species of symbiont [112], which may respond differently to climate stressors such as *p*CO2 and temperature [110,113,114]. Some work suggests that increased atmospheric *p*CO2 will be advantageous for plant growth, leading to greater terrestrial biomass and C storage [115,116], though potentially with an upper limit on *p*CO2 vs. growth returns [117]. However, observations have shown that increases in *p*CO2 during and after industrialization did not always increase plant C richness or biomass growth [118–120], and the biomass increase effect can still be mediated by P (and N) limitation in soils [121–123] and mycorrhizal species [102,114].

Ecosystem-specific C:N:P (Redfield) ratios vary, but throughout natural marine and terrestrial systems, P is the limiting nutrient [4,5]. For example, while the original marine phytoplankton-based Redfield ratio is 106:16:1, in soils, it is naturally higher at 186:13:1 [124]. The C:P ratios in both A and C horizons (16.7:1 and 14:1, respectively) were far lower than what is observed for O horizons (186:1 [124]) or inland streams/rivers (167:1 [125]), and are closer to ratios for microbial biomass in soils (~50:1 [124]). However, this ratio was derived with total P rather than organic P, and is therefore likely an overestimation. In these samples, the Corg:Porg may be closer to the expected value for soils, but further speciation work is needed. Additionally, it is reasonable to expect that mineral soil horizons (rather than organic-rich horizons, like O) would have lower Corg and nutrient concentrations. For developed soils (mean Corg:P = 15.25), this discrepancy could be linked to deep tilling removing nutrients from subsurface horizons and potentially increasing nutrient fluxes as

compared to undisturbed or shallow-tilled soils. For natural soils with the potential for cultivation/deep tilling (e.g., grasslands, mean Corg:P = 20.5), the lower C:P ratio in subsurface mineral horizons could be inferred as diminishing returns for deeper tilling. There were no strong correlations between Corg, Cinorg, or Ctot and variables of interest (CIA, clay content, P, Fe) (Supplemental Figures S6 and S7).

Complicating this response is the fact that increases in atmospheric *p*CO2 could increase the rate of bedrock weathering (e.g., [126]), which mediates P availability and biolimitation. Moderate increases in *p*CO2 and weathering could lead to higher P availability and greater plant biomass, but if weathering increases too much, P in soils could be depleted while plant growth continues to increase. If high weathering rates lead to faster or premature depletion of P, especially exacerbated by anthropogenic activity (agriculture), crop stresses and regional food shortages could occur. Additionally, demand for P as fertilizer could increase; fertilizer P is ultimately sourced from bedrock, which, on human timescales, is a nonrenewable material [127]. Without developing efficient P recycling methods or adding sources such as bone char (e.g., [128,129]), P could run out in a matter of decades [130–132]. Therefore, when considering how climate change will affect agricultural yields, it is important to consider plant physiology and P uptake mechanisms together with mineral/geochemical changes in the soil driven by natural and anthropogenic forces.

Overall, this work aids in predicting changes in terrestrial P fluxes primarily by linking soil P, weathering intensity, and the presence of vegetation. However, in this study, "vegetation" remains something of a black box; the mechanism by which the presence of vegetation affects soil P could be through slowing erosion, increasing biotic P cycling, retaining P via plant/microbial biomass, increasing bedrock apatite weathering, or—most likely—a combination of those factors. Exploring each of these factors at large spatial scales, and to the extent possible pairing those analyses with smaller-scale species-specific soil-plant relationships, is a critical next step in elucidating the potential for P fluxes in different terrestrial regions.

#### *4.7. Implications for P in the Fossil Record of Soils and Its Geologic Use*
