Hauser_FinalProjectManuscript.Rmd

---
title: "Rooting system strategies for acquiring organic vs. mineral nutrients "
author: "Emma Hauser"
params:
  today: !r.Sys.Date
bibliography: RootsRefs.bib
output: html_document
---
##Abstract
 Plant rooting systems may have different strategies for obtaining nutrients from organic vs. inorganic nutrient sources. Such varied strategies have implications for how plants allocate carbon (C) in soil ecosystems. In this study, we track correlations between root abundance and pools of both mineral- and organic matter- bound nutrients at the Calhoun Critical Zone Observatory, SC. Results suggest that roots may track mineral-bound nutrient pools where those nutrient elements are relatively more available. Further analyses of rooting system C distribution are needed to clarify the energetic drivers behind plant nutrient acquisition mechanisms. These findings, in conjunction with future C analyses, will help determine how plants achieve sufficient nutrition, particularly as nutrient forms shift due to anthropogenic land use change. 
 
##Introduction
<!--A chunk to set up data and Rmd settings for prelim data analyses--> 
```{r setup, include=F}
source("RootsPackages.R")
NewPack <- SoilPack[!SoilPack %in% rownames(installed.packages())]
if(length(NewPack)) install.packages(NewPack, repos = "http://cran.us.r-project.org")
devtools::install_github("emhauser/RPackageHW_Hauser/SoilMatrix2")
library(magick)
library(SoilMatrix2)
```

Carbon (C) is the currency through which plants meet life’s demands. After CO~2~ uptake, plants redistribute C for growth, using it either to build new cells or obtain nutrients from soils [@AberMelillo2001]. Although scientists have focused on plant C allocation in terms of aboveground productivity, up to 60% of phosynthetically fixed C can be redistributed to rooting systems that serve as plants’ primary nutrient conduits [@NewmannRomheld2001]. Belowground C serves numerous purposes including fine root proliferation, which increases absorptive root surface area for nutrient uptake, and production of C-compounds, called exudates, that break down nutrient-containing material after release into the surrounding soil matrix [@Roviera1969]. Although exudates can come either directly from roots or from root-associated mycorrhizal fungi, both sources represent an investment of plant C. As such, exudates are costly: the plant can no longer use that C for growth [@Roviera1969, @Bunemann2011]. Thus, plants tend to allocate more C belowground when nutrients are limited [@Bunemann2011, @JiBever2016].  

We know root systems and their associated biogeochemical processes propagate meters deep in some ecosystems, highlighting the extensive role of rooting system processes in belowground C storage and turnover [@Maeght2013, @Billings2018]. Yet our knowledge of soil C-cycling remains limited because most studies focus on the upper 30cm of soils, leaving a great deal of uncertainty about rooting system function, particularly where roots grow deep [@RichterBillings2015]. Conventional viewpoints suggest that deep roots are primarily for water uptake, but more recent studies indicate that many plants obtain their water from the uppermost 60cm of a soil profile [@Gaines2016], leading to inferences that plant C allocation to deep roots may be directed at nutrient acquisition [@Hasenmueller2017]. Although biologic activity and organic matter are concentrated in uppermost soil horizons [@RichterMarkewitz2001, @Billings2018], these surface conditions may be challenging for root nutrient acquisition. Not only is there greater competition from neighboring plants and microbes, but nutrients are present primarily as large organic molecules requiring decay and subsequent mineralization before mobilization for root uptake [@Zhu2016]. In contrast, deep soils may be rich in relatively unweathered soil mineral nutrients, but roots must reach sufficient depth to access them.  

The capacity for rooting system strategies to enhance nutrient availability may depend on which of the above nutrient forms dominates different depths within the soil system. Different rooting system exudates—from either roots or their mycorrhizae—can mobilize nutrients through distinct chemical reactions with mineral vs. organic nutrient forms [@Bunemann2011]. For example, enzyme exudates release nutrient ions by cleaving monomers from organic matter [@MarschnerRengel2007, @Bunemann2011], while organic acid-generating anions induce inorganic nutrient ion solubility through ligand exchange and desorption reactions with soil minerals [@SatoComerford2006]. Roots and their mycorrhizal symbionts may leverage enzymes, organic anions, or some combination of the two for nutrient acquisition [@MarschnerRengel2007], although the different processes by which exudates mobilize nutrients suggest that plants may receive greater nutrient benefits from organic anions where mineral nutrients are present and from enzymes in the presence of organic matter. In addition, the C-cost of exudate production further complicates the benefits of these rooting strategies, as enzyme production may require considerably more plant C than organic anion production, while reaching deep mineral-nutrient pools may necessitate considerable C for root growth [@Leake2008, @MarschnerRengel2007, @LynchHo2005]. Thus, the depth-associated exchange between plant C-resources belowground and nutrient uptake may determine the fate of fixed C in ecosystems, as well as the means of plant nutrient acquisition.  

Any depth-dependent tradeoff between C allocation for accessing deep mineral vs. shallow organic nutrient pools may be altered or even reversed with land use change. For instance, tillage-driven erosion exposes deep, relatively mineral-rich soil horizons, resulting in soil profiles containing more mineral nutrients than otherwise expected [@RichterMarkewitz2001, @Yoo2015]. In this case, shallow soil may harbor more reactive mineral surface area than might be anticipated, and thus may be more reactive to chemical weathering agents like root-derived organic acids, similar to deep soils [@Fortner2012, @Yoo2015]. In addition, many previously farmed lands received mineral fertilizer, modifying nutrient stocks even decades post-abandonment [@Richter2006]. Such shifts in the depths and relative abundances of mineral vs. organic nutrients imply that plants change their strategies to maximize nutrient acquisition, potentially lessening deep C cycling [@Brantley2017]. However, we do not know what compounds comprise deep root carbon exudates and whether roots receive considerable nutrient benefits from deep C investments [@Billings2018]. Thus, the function of root C remains unclear, preventing projections about how and why soil C depth distributions will shift with land use change. An analysis of root exudate composition and root distribution relative to the distribution of nutrient forms in soil profiles would illuminate changes in root C allocation and nutrient acquisition strategies as the abundance of organic- vs. mineral-derived nutrients changes with soil depth and land use history.  

In this study, I will attempt to examine the extent to which roots track mineral and/or organic nutrient pools in forest soil profiles up to 2m deep. This will allow me to test that hypothesis that:  

_rooting systems exhibit plasticity in nutrient acquisition strategies, such that roots track mineral nutrient pools where those are relatively more available due to the lower C-cost of organic acid exudate production. Conversely, roots will track organic matter derived nutrient pools in soil profiles where few fresh mineral surfaces remain.Where humans have altered the landscape in ways that deplete surface horizons’ organic matter stocks (i.e. tillage, fertilization, and deforestation) I expect that root distributions will be shallower and will change nutrient acquisition mechanisms to produce more organic acid exudates at shallower depths, thus leveraging more shallow mineral-rich material._  
     
If this hypothesis is accepted, it suggests an under-appreciated mechanism by which human impacts—here historic land use—can alter deep soil C inputs. Deviation from this observation would suggest that rooting system patterns are genetically-based, species-specific traits, or that exudates behave synergistically and may be exuded in mixtures that enhance nutrient availability more than a single exudate alone [@Vance2003].  

To examine this hypothesis I will focus on phosphorous (P), because it is available from both organic matter- and mineral-derived sources [@Vitousek1997]. In addition, rooting systems are thought to have P-specific exudate responses including phosphatase enzyme exudation in the presence of organic P (P~o~), and production of organic acids like oxalic, malic and citric acid that have been shown to mobilize mineral-bound P (P~g~) [@MarschnerRengel2007]. Thus, using P as a model nutrient allows us to examine variations in rooting system distributions as organic vs. mineral pools of the same nutrient change across soil type and depth, clarifying how plants allocate C for a vital resource. 


##Methods  
###Study site
I am conducting this study at the Calhoun Critical Zone Observatory, SC, which is home to both 200-year-old hardwood forests and ~80-year-old post-agricultural pine forests. The pine forests grow on land previously plowed for >100 y, and frequently fertilized with P-rich mineral fertilizer. As in most soils that do not contain buried O or A horizons, surface soils in both forest types are relatively P~o~-rich and harbor minerals with low P content; with increasing depth, relative abundances of mineral P~g~ increase and P~o~ decreases. In pine forests, P~g~ remains relatively high in surface horizons due to historic erosion and fertilization (Richter et al. 2006). 

```{r, echo=FALSE, fig.cap="Figure 1. Image of a hardwood forest soil pit sampled at the Calhoun Critical Zone Observatory.", out.height="50%", out.width="50%"}
#soilPit1 <- "https://nicholas.duke.edu/cczo/soilpits/R2H1-PROFILE-PA200226.JPG"
pic <- image_read("https://nicholas.duke.edu/cczo/soilpits/R2H1-PROFILE-PA200226.JPG")
pic %>%
  image_rotate(90)
#knitr::include_graphics(soilPit1)
```

###Data Collection
Researchers collected soil samples from 5 forest replicates, each containing a paired un-altered hardwood and post-agricultural pine forest soil pit. Soils were analyzed for extractable organic C (EOC)--a proxy for organic acid concentration--in every soil horizon and root density in every square centimeter of soil depth. Bulk elemental concentrations were analyzed using ICP-OES in every soil horizon up to ~2m deep. As well, scientists measured several fractions of organic and inorganic P for 7 fixed soil depths between 0 and 2 meters in each soil profile. These P fractions are estimated to have varying degrees of plant availability, as described by the Hedley [@Hedley1982] fractionation method. For this analysis, I will restrict my computation to data from the first 2 meters of soil depth, as not all datasets include data beyond this depth.

###Computational analyses
  This analysis relies on compiling three datasets--phosphorus fractions, rooting system C distribution, and bulk elemental data--each generated in a different collaborating lab. Therefore, not all data points from one dataset will have an exactly analogous point in another set. To make more powerful comparsions between these datasets, we modeled expected values at every 1cm soil increment using the `extrapolate()` function of the SoilMatrix2 R package. This function performs linear interoplation between observed sample points in each set of data, such that all datasets contain the same number of points at each soil profile depth.
  
  After data interpolation and organization, we performed linear regression analyses using the `SoilCorMat()` function of the SoilMatrix2 package to examine relationships  between root density distribution, P-fractions, nutrient elements Ca, Mg, and K, and soil organic carbon (SOC). SOC serves as a proxy for soil organic matter (SOM)--the major source of organically-derived plant nutrients--while inorganic P, Ca, Mg, and K represent mineral-derived nutrient pools. The strength of correlation between the fraction of roots found at a given soil depth and the mineral or organic nutrient form of interest is an indicator as to whether roots may be tracking organic vs. inorganic nutrient pools. 

  We finally visualized these relationship using the `SoilRainbowStats()` function of the SoilMatrix2 package, which not only illuminates the strongest correlations in these datasets, but also visualizes depth-dependent relationships embedded within correlations via its graphing color scheme. Thus, we were able to tell what resources roots in these systems track most strongly and how those responses change with depth and land use history. 
  
##Results

 Correlation matrices of all root distribution, root C, P-fraction, and soil elemental variables demonstrate strong positive correlations between root distribution, SOC, and Ca for both hardwood and pine forests (Fig. 2). Depth-indicator colors suggest that relationships between root distributions and P-fractions may be depth-dependent, requiring closer analysis. 

<!--A chunk to import datasets, which have been interpolated and organized in an R script-->
```{r datasource, include=F}
source("CPRootData.R")
```


<!--A chunk to make first correlation matrix plot-->
```{r hwMat, echo=F, fig.cap= "A)"}
SoilRainbowStats(HWSoilDat, 1, 175, 17, 25)
```

<!--A chunk to make second correlation matrix plot-->
```{r prelim, echo=FALSE, include=T, fig.cap="Figure 2. A) Correlation matrix showing relationships between root distributions, P-fractions, Ca, Mg, and K in Hardwood forests at the Calhoun Critical Zone Observatory, including Pearson's correlation coefficient. B) Plots of those same data relationships for Pine forests at Calhoun."}
SoilRainbowStats(PineSoilDat, 1, 175, 17, 25)
```



Our observation of strong correlation between roots and SOC encouraged closer analysis of these plots (Fig. 3). 

<!--A chunk to source new graphing code, made after original draft for better modularity and cleaner code -->
```{r PRFunc, echo=F}
source("PR_Fun.R")
```

<!--A chunk to make SOC plots-->
```{r SOC, echo=F, fig.cap="Figure 3. Plots of percent root fraction vs. soil organic carbon for a complete soil profile, lower half of the soil profile, and deepest sampling horizon in hardwood forest soil samples at the Calhoun CZO."}
HWDeep<-HWSoilDat[175:100,]
HWMid<-HWSoilDat[175:50,]
HWTotal<-HWSoilDat[175:1,]
depth_colTot <- rev(colorspace::rainbow_hcl(nrow(HWTotal)))
par(mfrow = c(1,3), mar = c(4.5,5,3,3))
PR(y=HWTotal$Roots, x=HWTotal$SOC, main = "Deep (0-175cm)", xlab = "")
PR(y=HWMid$Roots, x=HWMid$SOC, main = "Mid (50-175cm)", xlab = "SOC (gC gsoil-1)")
PR(y=HWDeep$Roots, x=HWDeep$SOC, main = "Total (100-175cm)", xlab = "")
```


When shallower depths are included, there is a strong positive correlation between roots and SOC (R^2^= `r round(summary(lm(HWTotal$Roots~HWTotal$SOC))$r.squared, digits = 3)`; p= `r round(summary(lm(HWTotal$Roots~HWTotal$SOC))$coefficients[2,4])`)
As shallower horizons are removed the correlation between roots and SOC weakens, such that roots appear to have a lower correlation with SOC in deeper soils (mid-depth: R^2^= `r round(summary(lm(HWMid$Roots~HWMid$SOC))$r.squared, digits = 3)`; p= `r round(summary(lm(HWMid$Roots~HWMid$SOC))$coefficients[2,4])`, deepest horizons: R^2^= `r round(summary(lm(HWDeep$Roots~HWDeep$SOC))$r.squared, digits = 3)`; p= `r round(summary(lm(HWDeep$Roots~HWDeep$SOC))$coefficients[2,4])`). 



Upon closer analysis, we observed opposing trends between roots and available P~o~, measured in HCO~3~ extracts compared to roots and available P~g~ measured in HCL extracts in shallow vs. deep hardwood soils (Fig. 4). 

<!--A chunk to make P Figure-->
```{r P figs, echo=F, fig.cap="Figure 4. Root distributions relative to organic phosphorus (HCO3 Po) fractions and inorganic, mineral-bound phosphorus fractions (HCl Pi)  in deep (150-175cm) soil horizons and shallow (0-50cm) horizons at the Calhoun CZO."}

HWDeep<-HWSoilDat[100:175,]
HWShallow <-HWSoilDat[50:1,]
par(mfrow = c(2,2))

#Make plots with R2 and p values.

PR(y=HWShallow$Roots, x= HWShallow$HCO3.Po., cex.main = 1.5, cex.axis = 1, cex.lab = 1.25, main = "Shallow", xlab = "Mineralizable Po (HCO3 Po mg/kg)")
PR(y=HWDeep$Roots, x=HWDeep$HCO3.Po.,cex.main = 1.5, cex.axis = 1, cex.lab = 1.25, main = "Deep", xlab = "Mineralizable Po (HCO3 Po mg/kg)")
PR(y=HWShallow$Roots, x=HWShallow$Con.HCL.Pi., cex.axis = 1, cex.lab = 1.25,main = "", xlab = "HCl Pi (mg/kg)")
PR(y=HWDeep$Roots, x=HWDeep$Con.HCL.Pi., cex.axis = 1, cex.lab = 1.25, main = "", xlab = "HCl Pi (mg/kg)")
```


We also observed depth-dependent relationships between roots and Ca in deep hardwood soils (Fig. 5). Although Ca data are bulk elemental measurements, deep Ca stocks are likely derived from soil minerals rather than organic matter. 


<!--A chunk to make Ca figure -->
```{r CaFigs, echo=F, fig.cap="Figure 5. Fraction of roots in deep (100-175cm) hardwood forests, where nutrients elements are expected to be mineral-bound, compared to calcium weight percent of soil."}

HWDeep<-HWSoilDat[100:175,]
par(mfrow = c(1,1))
PR(y=HWDeep$Roots, x=HWDeep$Ca, cex.axis = 1.25, cex.lab = 1.5, main = "", xlab = "Ca (%)")
``` 


In pine forests, we observed correlations between root distribution and P fractions that were similar to those same correlations in much deeper hardwood forests. Shown here for residual P, deep roots in hardwood forests track residual P to a similar degree as shallower roots in pine forests (Fig. 6).

<!--A chunk for depth-translation of P relationships between HW and Pine -->
```{r PdepthShift, echo=F, fig.cap="Figure 6. Fraction of fine roots in deep soils compared to residual phosphorus (mg P per kg soil) in deep hardwood forest (150-175cm) and mid-depth pine forest (50-80cm) soil profiles."}
HWDeep<-HWSoilDat[175:150,]
PineDeep<-PineSoilDat[80:50,]
par(mfrow = c(1,2))

PR(y=HWDeep$Roots, x=HWDeep$Residual, cex.axis = 1.25, cex.lab = 1.5, main = "Hardwood", xlab = "Residual P (mg/kg)")
PR(y=PineDeep$Roots, x=PineDeep$Residual, cex.axis = 1.25, cex.lab = 1.5, main = "Pine", xlab = "Residual P (mg/kg)")
```


#Discussion
Recent studies suggest that plants allocate their C resources to nutrient acquisition strategies that maximize the nutrient return on C investment [@LynchHo2005]. In shallow soil horizons, the relatively high abundance of soil organic matter suggests that SOM decomposition may be a readily available nutrient source, while weathering of mineral-bound nutrients may be the least C-expensive strategy if roots can reach sufficient depth [@Goransson2005, @Lambers2008, @Zemunik2015]. Assuming that nutrient acquisition is a primary driver of root growth, we might expect to see roots tracking SOM pools in upper soil horizons and mineral pools in deeper horizons where unweathered minerals are more readily available.  

Results of this study suggest that deep roots may indeed be tracking mineral-bound nutrients where they those nutrients are relatively more available. Throughout the soil profile, we observed strong correlations between roots and soil organic carbon (SOC), which is a proxy for organic matter (SOM). Although such strong correlations are expected because roots themselves comprise part of SOC [@Lal2016], the correlation is tightest in shallow soils (Fig. 3), where roots have the most access to organically-derived nutrient pools. This correlation weakens deeper in the soil profile, suggesting that the relationship between root distribution and SOC dissolves where roots are growing deep. Therefore, shallow roots may track readily available SOM-derived nutrients, but deep roots may be growing to obtain a different resource.

Phosphorus is a nutrient element of frequently low availability that is derived from either organic matter or mineral weathering, making it an ideal model to examine the nutrient pools roots track [@Vitousek1997]. Our P findings corroborate results from SOM. The relationship between root abundance and mineralizable P~o~ measured using HCO~3~ extractions also shows a strong positive correlation with root abundance (Fig. 4), furthering evidence that shallow roots may leverage organically-derived nutrient pools. However, deep in the soil profile (100-200cm), the relationship between roots and P~o~ reverses, suggesing that deep roots are actually _less_ prevalent surrounding P~o~ (Fig. 4). In contrast, we found a significant positive correlation between root abundance and mineral-bound P~i~ in deep soils--a relationship that was negative in shallow soil horizons (Fig. 4). This result suggests that roots track inorganic, mineral-bound P to obtain sufficient P nutrition in deep soils. 

We can further strengthen the argument that deep roots track mineral-bound nutrients by examining the relationship between root abundance and other mineral-bound nutrients. Calcium (Ca) is a nutrient element that, like P, becomes incorporated into organic matter recycling processes in shallow soils but is derived primarily from mineral weathering deep in the soil profile [@DijkstraSmits2002]. We therefore expect that most Ca in deep soils is mineral-bound. Root abundance shows a strong positive correlation to Ca in deep (100-175cm) hardwood forest soils, providing additional evidence that roots track mineral-bound nutrients where they are relatively more abundant (Fig. 6). Consistent evidence of positive correlation between mineral nutrients and root abundance (e.g. Ca and P~i~), as well as reduced and negative correlation between deep roots and organic nutrients (e.g. SOC and P~o~) implies that plants obtain nutrition from mineral-bound sources where those sources are relatively more available. This root response may be due to a lower photosynthate C requirement for rhizosphere-induced weathering processes compared to the C-cost for SOM decomposition processes [@LynchHo2005]. Differences in the nutrient pools roots track at different depths within the same soil profile may also suggest that rooting systems exhibit plasticity in nutrient acquisition strategies when varied resources are available. 

Evidence of such plasticity may also be observed via differences in root nutrient-tracking strategies between hardwood and pine forests at Calhoun. The history of agricultural land use in pine forests has left the soils more eroded, meaning that much SOM originally present at the soil surface has been lost [@Richter2006]. This, combined with chemical fertilization, has enriched shallower soils in mineral-bound nutrients, such that shallower soil horizons may appear mineralogically more similar to deep, undisturbed hardwood forest soils. Indeed, in shallow pine forest soils, we observed correlations between root abundance and mineral P forms that are similar to correlations with those same mineral-bound P forms in deep hardwood soils (Fig. 6). Therefore, roots may change their nutrient acquisition strategy to leverage the nutrient pools available [@Brantley2017]--an ecosystem characteristic that may be a function of shifting nutrient forms due to land use change.


Probing these questions further will address multiple uncertainties surrounding belowground elemental cycling. Foremost, an analysis of root C exudate distribution in these same soil profiles will clarify the nutrient return on plant root investments from distinct mineral vs. organic pools, a considerable knowledge gap surrounding key plant resources [@Gilbert2009]. This research also will determine the extent to which nutrient form governs the depth distribution of fixed C, which has numerous implications for downstream C fates including soil organic C formation, priming, and retention[@Callesen2016]. Furthermore, the compilation of these findings will illuminate the extent to which land use history inadvertently alters deep C cycling via little-considered rooting system mechanisms responding to erosion-generated changes nutrient distributions. This improved understanding of plant C allocation for nutrient acquisition will help discern whether plants exhibit plasticity in response to organic vs. mineral forms of the same nutrient—-an important advance in understanding how the soil developmental changes that follow from land use change will impact plant productivity and ecosystem services.

#References