Ecosystems

, Volume 12, Issue 8, pp 1352–1368

A Quantitative Model of Soil Organic Matter Accumulation During Floodplain Primary Succession

Authors

    • School of Aquatic and Fishery SciencesUniversity of Washington
  • Robert J. Naiman
    • School of Aquatic and Fishery SciencesUniversity of Washington
Article

DOI: 10.1007/s10021-009-9294-9

Cite this article as:
Bechtold, J.S. & Naiman, R.J. Ecosystems (2009) 12: 1352. doi:10.1007/s10021-009-9294-9

Abstract

Texture is an important influence on organic matter (SOM) dynamics in upland soils but little is known about its role in riverine soils. We hypothesized that texture might be especially important to SOM accumulation in young alluvial soils. We combined the soil component of the CENTURY ecosystem model, which uses sand, silt, and clay concentration as primary variables, with a simple simulation model of fluvial deposition, and forest production to predict changes in soil carbon (C) and nitrogen (N) during primary succession on floodplains and terraces of the Queets River, Washington. Simulated soil C accumulated to a plateau of about 4000 g m−2 at 110 years, closely matching observed patterns in an empirical chronosequence. Although direct fluvial OM deposition had only a small and short-lived influence on soil C, fluvial silt and clay deposition were an important influence on equilibrium C. The model underestimated soil N by about 35%, which appears to be due to failure of the model to account for N enrichment of an OM pool after its initial formation. These results suggest that basic influences on SOM retention in these young soils are not functionally different than those that apply to upland soils, but occur within highly dynamic physical contexts. Overbank deposition of silt and clay establishes a basic capacity for SOM retention. SOM, in turn, facilitates N retention. In this way, silt and clay are instrumental in propagating N forward from N-fixing red alder (Alnus rubra) stands to mature conifer forests that are frequently N-limited.

Keywords

floodplain soilssedimentationCENTURY modelQueets Riverprimary succession

Introduction

The accumulation of soil organic matter (SOM) to levels capable of sustaining a reliable supply of nutrients and water to biota is an important achievement in the early development of ecosystems. During primary succession, soils undergo a period of SOM increase as the plant community develops, leveling off to a plateau as production of organic debris comes into equilibrium with decomposition. The time scales over which these changes occur vary greatly among ecosystems (Walker and del Moral 2003). Floodplain soils develop particularly rapidly, commonly evolving from bare sediments seemingly incapable of supporting plant life to highly productive soils within a few years (for example, Van Cleve and others 1996; Robertson and Augspurger 1999; Corenblit and others 2007). Floodplains are also unique among terrestrial ecosystem types to experience recurrent primary succession over time scales relevant to development and succession of plant communities. For many unaltered rivers, especially in piedmont areas, channel migration rates are sufficient to maintain larges areas in relatively immature stages of development (Salo and others 1986; O’Connor and others 2003).

Factors affecting the balance of SOM are well described in conceptual (Jenny 1941) and quantitative (Smith and others 1997) models. In addition to climate (temperature and moisture) and the amounts and chemical composition of organic inputs, soil texture is recognized as a primary influence on soil organic matter (SOM) and nutrient cycling (Jenny 1941; Chapin and others 2002). Adsorption of SOM to silt and clay particles and complexation within aggregates renders it less susceptible to leaching and can inhibit decomposition. These influences are reflected in SOM turnover times, which range from 10s to 100s of years for OM encapsulated in aggregates and from 100s to 1000s of years for clay-adsorbed OM in temperate ecosystems (Trumbore 1993; Gaudinski and others 2000). Clay content, or some other measure of soil texture, is among the driving variables included in most soil organic matter cycling models (Smith and others 1997).

However, most of the research literature treats soils as closed systems. Although this may be acceptable in mature, stable ecosystems, floodplains exchange both sediments and organic matter with rivers, and sorting of these materials during transport and deposition can lead to spatial and temporal heterogeneity in their size distributions within alluvial landforms. Deposition of fertile alluvium is popularly seen as playing a basic role in contributing nutrients and OM to floodplains (Carter and Dale 1974). Although ecologists also recognize that natural patterns of disturbance and flow variability are essential to maintaining the integrity of rivers, there has been relatively little formal study of fluvial influence on alluvial soil organic matter dynamics. Such knowledge would aid in understanding the influence of natural disturbance patterns on riverine ecosystems and how that process may be altered by human activity.

The CENTURY model (Version 4.0; Parton and others 1987; Metherell and others 1993) has been applied to a diversity of mature soils to predict changes in soil organic matter pools and fluxes in response to cropping practices, timber harvest, changing climate, and other perturbations. However, no study has considered its suitability for examining organic matter dynamics during primary succession on riverine floodplains. We adapted the CENTURY model to examine soil C and N dynamics along a 330-year chronosequence of floodplains and alluvial terraces of the Queets River, Washington. The soil components of the CENTURY model were supplemented with equations derived from field collected data on fluvial deposition, litterfall, and tree production. Our objective was to provide a quantitative basis for examining the interaction of fluvial deposition and in situ processes on alluvial SOM and nutrient accumulation.

Study Area

This research was conducted in the Queets River watershed (1157 km2), on the western edge of Washington State’s Olympic Peninsula, USA (Figure 1). The Queets flows west from its headwaters on the southwestern flank of Mount Olympus through a steep confined channel for the uppermost 22 km, and then changes to low gradient alluvial character for the remaining 42 km to the Pacific Ocean. The study area and almost all the upstream watershed lies within Olympic National Park, and is largely free of human impacts. Mean annual precipitation is 3.87 m y−1, with most precipitation occurring between November and March. The Queets River has a highly dynamic but predictable hydrologic regime. Although discharge may be as low as 20 m3 s−1 in late summer, it regularly exceeds 2000 m3 s−1 during autumn storms. The highest recorded discharge during a 75-year event in 2001 was less than twice the mean annual peak flow (U.S. Geological Service 2006).
https://static-content.springer.com/image/art%3A10.1007%2Fs10021-009-9294-9/MediaObjects/10021_2009_9294_Fig1_HTML.gif
Figure 1

Queets River study area in western Washington, USA. Floodplains (shown in black) and fluvial terraces (shown in light gray) comprise 3.4% of watershed area. Study sites are indicated by triangles. Study area is incised 100–150 m into surrounding glacial outwash terraces.

The parent material is uplifted marine sediments, consisting primarily of sandstone, shale, and various conglomerates (Tabor and Cady 1978). The most recent of the Pleistocene mountain glaciers to reach the lower valley retreated approximately 12,000 years ago (Thackray 2001). Despite the absence of extreme floods, a highly dynamic channel migration regime is maintained by undermining of outwash terraces during fall and winter storms. The lower Queets valley is incised approximately 100 m into outwash terraces; continued downcutting is attested to by the stair-stepping of younger terraces below older terraces. Channel migration has resulted in a complex network of side and abandoned river channels separating forest patches at different stages of soil and vegetative development (Latterell and others 2006).

Newly formed bars evolve into floodplains experiencing periodic inundation and then to terraces isolated from direct fluvial influence through the combined effects of sediment deposition and channel downcutting (Figure 2). This usually occurs within 40 years. In most cases, initial colonization of river bars by some combination of mosses, grasses, forbs, and willows (Salix spp.) is followed by rapid floodplain elevation, and a shift from bedload dominated deposition to overbank deposit of fine sand, silt, and clay. Red alder (Alnus rubra), a vigorous N-fixer (Binkley and others 1994), shades out willow within a few years and forms a single species canopy for the following 50–70 years. The mature Sitka spruce dominated forest gradually transitions into a mixed forest of spruce, black cottonwood (Populus balsamifera trichocarpa), bigleaf maple (Acer macrophyllum), and western hemlock (Tsuga heterophylla). Although hemlock is considered the climax species (Franklin and Dyrness 1973), channel migration limits its dominance to older, higher terraces on the valley margins.
https://static-content.springer.com/image/art%3A10.1007%2Fs10021-009-9294-9/MediaObjects/10021_2009_9294_Fig2_HTML.gif
Figure 2

Schematic diagram of the Queets chronosequence showing approximate flood return intervals of different-aged landforms at bottom. Alluvial landforms are constructed primarily from sandy sediments deposited over river cobbles. Suspended sediments are the source of a thin (usually <20 cm thick) veneer of silt and clay deposited toward the end of the 40-year period during which landforms are subject to flooding.

The study area is located between 17.5 and 25 km from the Pacific Ocean. The valley bottom is between 1.0 and 1.8 km wide in the study area. Average channel slope is 0.006%. Although localized wetlands are common in swales and abandoned channels, the valley bottom is dominated by well-drained loamy sand soil caps surmounting porous cobble beds. More-developed soils on older terraces are classified as Entisols, belonging to the Huel and Tealwhit series, with a weakly developed A and one or more C horizons. The oldest alluvial terraces are classified as Andisols (Queets series) despite the absence of tephra (McCreary 1975).

Methods

The model simulates soil C and N cycling during the evolution from bare sediments to mature forest, and consists of three interacting components (Figure 3): a sedimentary submodel simulating fluvial deposition of sediments and OM, and subsequent changes to soil texture as a result of sediment weathering; a forest production submodel simulating autochthonous C and N inputs from litterfall and tree mortality, biological N fixation, and N uptake by forest vegetation; and a soil submodel that simulates decomposition, with gaseous (respiration/denitrification) and leaching C and N outputs. Equations for model inputs were derived from data collected for related Queets River studies. Fluvial sediment and OM data (Latterell and Naiman 2007; Bechtold and Naiman, unpublished) were collected along stream margins of the lower Queets valley (Figure 1). Data for autochthonous soil inputs were from an adjacent chronosequence of 25 sites ranging in age from 3 to 330 years relative to initial plant colonization (O’Keefe and Naiman 2006; Van Pelt and others 2006). These data were supplemented with literature values for N fixation and lignin concentrations. Equations for soil C cycling, N cycling, and soil hydrology were from the CENTURY model (Version 4.0; Metherell and others 1993). Soil C and N data from the chronosequence were used to validate the model (Bechtold and Naiman, unpublished). A 330-year simulation period was selected to encompass the time period for which forest and soil data were available. The model was implemented in a series of large MS Excel (Microsoft 2003) workbooks.
https://static-content.springer.com/image/art%3A10.1007%2Fs10021-009-9294-9/MediaObjects/10021_2009_9294_Fig3_HTML.gif
Figure 3

Organic matter simulation model showing fluvial inputs, autochthonous inputs, and soil cycling as described by the CENTURY model. Arrow thickness distinguishes major from minor fluxes. Dashed arrows indicate gaseous CO2 outputs due to respiration. Letters indicate fluxes influenced by soil texture: A, silt and clay inhibit decomposition of active soil OM; B, silt and clay reduce leaching by adsorbing OM and reducing hydrologic flux; and C, passive OM is formed by OM association with clays.

Sedimentary Submodel

The sedimentary submodel simulates deposition and burial of fluvial sediments and OM during the period that soils are subject to flooding, and in situ changes in soil particle size distribution occurring over the 330 year chronosequence (Table 1). Deposition was simulated as occurring during the first 40 years of site development, based on visual evidence of overbank flooding at the study sites following several storms, including an 80 year event in 1999. Total deposition depth of 75 cm represents the average soil depth (above river cobbles) of sites older than 40 years. Soil depths were highly variable among the 12 chronosequence sites with ages between 3 and 39 years; logarithmic functions were used to approximate the transition from intense bedload deposition of sand-dominated sediments during the first decade to less intense deposition of suspended sediments with greater silt and clay during the latter stages of floodplain construction. Initial simulated sediment size distribution was set to reflect measured proportions in stream margin flood deposits (92% sand, 7% silt, and 1% clay), increasing to 40% silt and 5% clay in 40-year-old soils as previously deposited coarse sediments are buried below the 20 cm simulated soil depth.
Table 1

Inputs to Sedimentary Submodel

 

Annual input

Period

Total

Sand deposition

6.6 * (e0.08*age (silt + clay deposition) cm y1

0–40 years

55.9 cm

Silt deposition

6.6 * (e0.08*age) * (0.0805*LN(age) + 0.07) cm y1

0–40 years

17.7 cm

Clay deposition

6.6 * (e0.08*age) * (0.0108 * LN(age) + 0.01) cm y1

0–40 years

2.4 cm

Silt weathering

0.016/bulk density cm y1

0–330 years

4.9 cm

Clay weathering

0.0054/bulk density cm y1

0–330 years

1.6 cm

Bulk density

0.4 * eage*0.15 + 1.2 g cm3

0–330 years

 

Sand-adsorbed C

2.1 * sand deposition mg g1

0–40 years

1602 g m2

Silt-adsorbed C

6.6 * silt deposition mg g1

0–40 years

1531 g m2

Clay-adsorbed C

20.1 * clay deposition mg g1

0–40 years

638 g m2

Sand-adsorbed N

0.2 * sand deposition mg g1

0–40 years

149 g m2

Silt-adsorbed N

0.9 * silt deposition mg g1

0–40 years

225 g m2

Clay-adsorbed N

2.2 * clay deposition mg g1

0–40 years

70 g m2

POC

0.12 * sediment-adsorbed C mg mg1

0–40 years

452.52 g m2

PON

0.10 * sediment-adsorbed N mg mg1

0–40 years

44.4 g m2

LWD C

17.08*(1 + *e0.13*age) g m2

0–20 years

113.9 g m2

LWD N

0.005* 17.08*(1 + *e0.13*age) g m2

0–20 years

0.6 g m2

Soil particle size distribution also decreases after sediments have been deposited. Simulated increases of 0.8 mg silt g soil−1 y−1 and 0.27 mg clay g soil−1 y−1 are based on observed changes in silt and clay contents over the 330 year chronosequence (Bechtold and Naiman, unpublished). Windblown dust may contribute some silt and clay. However, increases in clay and silt were observed at all depths suggesting that weathering was primarily responsible. This is consistent with weathering rates observed during formation of pro-glacial soils (Burt and Alexander 1996; Egli and others 2006). Notably, weathering is a more important source of clay than fluvial deposition.

Estimation of fluvially deposited OM was based on the observation that most particulate OM transported by rivers (Naiman and Sedell 1979; Ittekkot and Laane 1991) and deposited on floodplains (Steiger and others 2001; Morozova and Smith 2003) consists of predictable concentrations of amorphous sediment-associated OM. Measured organic loadings on fluvial sediments, supplemented with measurement of particulate OM with no sediment association (including <10 cm diameter detritus—POM, and large woody debris—LWD), were thus used to estimate fluvial organic inputs (Table 1). Overall, 87% of simulated C deposition and 91% of simulated N deposition was sediment-associated (Bechtold and Naiman, unpublished).

Flood deposits from eight locations along channel margins were size/density fractionated to isolate four fractions: sediment-associated sand (54–2000 μm), silt (2–540 μm), and clay (<2 μm), and non-sediment-associated POM smaller than 2 mm. Deposition of sediment-associated OM was simulated by multiplying mean organic loadings of mineral sediments by the mass of fluvially deposited mineral sediment in each size class. Deposition of 2–20 mm POM was estimated relative to the total mass of deposited sediments by sieving flood deposits. POM between 2 and 10 cm was estimated from organic debris in soil pits at 3-, 4-, and 10-year-old chronosequence sites. Inputs of fluvial large organic debris are based on estimates of stemwood deposition from a related study (Latterell and Naiman 2007), with allometric equations (Means and others 1994) used to account for coarse roots and branches. POM deposition was simulated as occurring proportionally to silt deposition. Large organic debris deposition was simulated as occurring during the first 20 years of floodplain development, reflecting the observed pattern on the Queets floodplain (Latterell and Naiman 2007). Fluvially deposited OM was assigned to functional pools (described in forest and soil submodels below) based on low C mineralization rates observed during long-term laboratory incubations (Bechtold and Naiman, unpublished): 85% of sediment-associated OM was assigned to the slow soil pool and 15% to the passive pool. Large woody debris was transferred to the dead wood pool upon deposition and particulate OM was assigned to surface structural OM.

Forest Submodel

The forest production submodel simulates ongoing C and N inputs from living vegetation, inputs of C and N in woody debris following tree mortality, and N outputs due uptake by vegetation. Aboveground litterfall (Tables 2 and 3) was modeled as moving averages of C and N in leaf (by species), epiphyte, and small branch litterfall fractions collected from litter baskets at 15 sites (O’Keefe and Naiman 2006). Studies measuring regrowth in clipped plots indicated that the grass/forb was inconsequential as a litter source, and no estimate of grass/forb inputs were included (T. O’Keefe, personal communication). Belowground litterfall of fine root C and N were estimated from an equation relating fine root production to tree stem biomass (fine root biomass = 0.02 + 0.28(−stem biomass/40); (Nalder and Wein 2006), using Queets stem biomass data (Van Pelt and others 2006). Fine root turnover was estimated at 2 years (Gill and Jackson 2000). N uptake by forest vegetation was estimated as equaling the sum of N in leaf litter, fine branch litter, fine root turnover, and growth of woody (stem, coarse root, and large branch) tissues.
Table 2

Aboveground Litterfall Biomass Inputs to Forest Submodel (g m−2)

Site age

Willow leaf

Red alder leaf

Other decid leaf

Conifer leaf

Epiphyte

Small branch

8

17.3

24.1

0

0.1

2.4

31.1

11

266.5

242.5

0

1.7

0.2

40.3

12

76.2

94

0

0

0

29.7

16

119.8

62.5

0

0.1

0

27.7

22

21.2

443.5

0

6.1

1.1

66.9

39

0

492.9

0

0.1

3.2

41.5

39

10.2

497.7

0

0

0.1

52.2

51

0

467.9

3.2

0.1

1.5

175.4

51

0

445.5

0

4

9.7

39.8

60

0

173.1

1.4

436.8

6.2

112.8

84

0

211.8

0.1

363.2

65

253.7

165

0

4.7

132.5

220.6

21

92.2

200

0

1.4

3.6

400.8

50.7

222.4

265

0

12

117.2

365

24.8

80.9

330

0

0.7

45.1

297.9

22.8

44.4

Table 3

C, N and Lignin in Detrital Inputs to Forest Submodel

 

% C

C/N

Reference

% lignin

Reference

Alder leaf

0.44

22

O’Keefe and Naiman (2006)

0.14

Domenach and others (1994)

Alder fine root

0.45

25

Chen and others (2002)

0.15

Chen and others (2002)

Other deciduous leaf

0.46

31

O’Keefe and Naiman (2006)

0.14

Hessl and others (2004)

Deciduous fine root

0.45

25

Chen and others (2002)

0.15

Chen and others (2002)

Decid. small branch

0.50

300

Nalder and Wein (2006)

0.16

Edmonds and others (1986)

Decid. coarse wood

0.50

300

Nalder and Wein (2006)

0.16

Edmonds and others (1986)

Conifer leaf

0.47

55

O’Keefe and Naiman (2006)

0.22

Hessl and others (2004)

Conifer small branch

0.50

300

Nalder and Wein (2006)

0.29

Costa E Silva and others (1998)

Conifer coarse wood

0.50

300

Nalder and Wein (2006)

0.29

Costa E Silva and others (1998)

Epiphyte/misc.

0.44

37

O’Keefe and Naiman (2006)

0.15

Assumed

Because most tree recruitment occurs in discrete cohorts in these forests, growth and mortality can effectively be estimated from changes in biomass and stem density over time. Biomass and stem density data for red alder and Sitka spruce collected at 21 sites (Van Pelt and others 2006) were fit to Gompertz (1825) and logarithmic models, respectively (Table 4). The time derivative of biomass was used to estimate growth and the derivative of the stem density was used to estimate mortality. Growth and mortality of minor canopy species were estimated as the most reasonable fit to available data. Large branch and coarse root biomass was estimated as a function of stem biomass according to literature values (Means and others 1994). Inputs of stem wood, large branch, and coarse roots to soils are all assumed to result from tree mortality. Upon tree death, woody debris first enters a dead wood pool, and then is gradually transferred to structural SOM. The transfer of stem wood from dead to litter pools is modeled linearly over 50 years for conifers and 10 years for hardwoods (Edmonds and others 1986).
Table 4

Forest Submodel Equations for Tree Mortality and Tree Growth

Species

Years

Annual input/output (g m−2 y−1)

Mortality

  

 Red alder

11–131

117071 * SiteAge * e−2.325 * 0.137 * \( {\text{e}}^{{ - 7. 4 3*{\text{e}}^{{ - 0.0 6 2 *{\text{ SiteAge}}}} }} \)

 Sitka spruce

1–330

1.405 * e−0.005 * SiteAge * 3.9 * \( {\text{e}}^{{ - 5. 7 0 3*{\text{e}}^{{ - 0.0 0 8 *{\text{ SiteAge}}}} }} \)

 Willow

6–36

7029.7 *SiteAge−1 * 0.0003 * SiteAge − 0.0018

 Bigleaf maple

202–330

(0.004 * SiteAge + 0.0193) * SiteAge − 0.7972

 Black cottonwood

5–200

97.544 * SiteAge−1 * (0.002 * SiteAge − 0.0075)

Stem growth

  

 Red alder

2–105

88362*siteage−1.325 *0.0680 * \( {\text{e}}^{{ - 0.062*{\text{SiteAge}} - 7.431*{\text{e}}^{{ - 0.0 6 2 1 *{\text{ SiteAge}}}} }} \)

 Sitka spruce

1–330

292.75 * e−0.0048 * SiteAge * 0.153 * \( {\text{e}}^{{ - 0.008*{\text{SiteAge}} - 5.703*{\text{e}}^{{ - 0.0 0 9 *{\text{ SiteAge}}}} }} \)

 Willow

6–36

(−7029.7 * LN(SiteAge) + 25062) * 0.0003

 Bigleaf maple

1–200

(−0.0014 * SiteAge2 + 0.6193 * SiteAge − 28.445) * 0.009

 Black cottonwood

1–200

(−95.759 * LN(SiteAge) + 524.69) * 0.002

 Western hemlock

39–330

(99.301 * LN(SiteAge) − 424.02) * 0.0008

Large branch growth1

  

 Red alder

2–105

(0.039 * stem biomass − 6.87) * stem biomass−1 * stem growth

 Sitka spruce

1–330

(0.0435 * stem biomass − 16.4) * stem biomass−1 * stem growth

 Willow

6–36

0.0132 * e0.217 * stem biomass * stem biomass−1 * stem growth

 Bigleaf maple

1–200

(0.0435 * stem biomass − 16.4) * stem biomass−1 * stem growth

 Black cottonwood

1–200

(0.00145 * stem biomass + 0.0735) * stem biomass−1 * stem growth

 Western hemlock

39–330

(0.145 * stem biomass − 8.10) * stem biomass−1 * stem growth

Coarse root growth1

  

 All species

1–330

(0.12 + 0.58*e−stem biomass/20 − 0.7*e−stem biomass/100) * stem growth

1Adapted from Means and others (1994).

Nitrogen fixation was estimated from literature values. Reported N-fixation rates for red alder stands range from 2.8 to over 30 g m−2 (Binkley and others 1994). However, most of these estimates are based on soil N accretion, making no account of leaching losses which are high from the Queets soils (Bechtold and others 2003). We estimated average alder N-fixation over the first 70 years of the successional sequence at 8.5 g m−2 y−1, assuming stream export of about 4 g m−2 y−1 (Compton and others 2003). N-fixation was scaled to alder stem biomass (0.003 * stem biomass), with a maximum of 12 g m−2 y−1 at age 39. Alder-fixed N is first input to leaf and fine root litter and cycles through soils as the litter is decomposed.

In the older conifer forest, N fixation is primarily by the lichen Lobaria oregana. We estimated canopy N fixation at the high end of the range measured in Oregon forests (Antoine 2004) due to the exceptionally high canopy epiphyte biomass measured in these forests (R. Van Pelt and R.J. Naiman, unpublished data). Epiphyte biomass is assumed to have a turnover time of 5 years (Pike 1978) and is assumed to decompose and accumulate in soils in the same manner as leaf litter. Canopy N-fixation is scaled to stem biomass (0.0002 * stem biomass) beginning at 100 years, increasing to a maximum of 1.4 g m−2 y−1 at 330 years. Atmospheric N deposition, based on measurements at the nearby Hoh River (Edmonds and Blew 1997), is estimated at 0.325 g m−2 y−1 and is weighted to a constant concentration in precipitation.

Model validation was performed with soil data from three soil pits in each of 25 sites (Bechtold and Naiman, unpublished). Carbon and N were measured in the organic layer whereas C, N, particle size distribution and bulk density were measured in 10 cm depth increments in the soil profile. Concentrations of sand, silt, and clay-associated C and N were measured by first separating size classes via sieving and centrifugation of ultrasonically dispersed sediments and measuring then C and N in the individual size fractions.

CENTURY Soil Submodels

The soil component of the CENTURY model simulates organic matter dynamics in surface and root litter and in the 0–20 cm mineral soil depth. The soil C decomposition rate is determined using the initial chemistry of organic inputs, soil temperature, soil moisture, and sand, silt and clay concentrations. In the model, plant residues first enter metabolic (labile or rapidly decomposing) or structural (decay resistant) litter pools depending on their lignin contents, and are then cycled through active surface and three mineral soil pools (active, slow, and passive) differing in decomposition rates. The active surface and mineral soil pools are comprised of microbes and products of recent microbial decomposition and have turnovers of months to a few years. The slow pool consists of decomposed structural materials and stabilized microbial products and has a turnover of 20–50 years. The passive pool consists of highly resistant OM with a turnover time of 400–2000 years.

N is cycled through analogous pools, with gross N mineralization inversely proportional to the mass C:N ratio of OM for a given pool. The C:N ratio of material entering structural and metabolic pools reflects the N concentrations of the senesced plant material. Structural OM is fixed at a C:N ratio of 150, reflecting the low N content of stem wood and other structural plant materials. The C:N of the surface microbial pool increases from 10 to 20 as the N concentration in the material entering the pool decreases from 2.0 to 0.01%. The C:N ratios of OM in active, slow, and passive pools are determined by mineral N concentrations at the time that the material was formed, decreasing from 15 to 3, 20 to 12, and 10 to 7, respectively, as inorganic N increases from 0 to 2 g m−2. N output due to denitrification is simulated as a constant percentage of mineral N remaining after plant uptake using the default value of 2%.

Texture affects C and N retention at three steps in the soil OM cycle (identified by letters in Figure 3). First, adsorption to silt and clay inhibits decomposition of labile OM following initial breakdown of organic debris, presumably by incorporation in aggregates. Second, clay content controls the amount of SOM entering the passive pool, the mechanism for long-term sequestration. Third, silt and clay influence C and N leaching. Leaching of organic C and N solubilized from the active pool is influenced by clay content which provides surfaces for adsorption, and by the flux of water to below the 30 cm soil depth, which is also influenced by texture. Inorganic N in excess of biological demand is subject to leaching, which is determined by the flux of water below the 30 cm soil depth. Inorganic N is not influenced by adsorption to sediments; the model assumes nitrification of excess N and reflects the high mobility typical of nitrate in soils.

Temperature and rainfall were simulated as monthly averages, based on a 30-year record from a NOAA weather station (National Climate Data Center 2005), located approximately 48 km northwest of the study site. No attempt was made to account for interannual variation in precipitation or temperature.

Uncertainty/Sensitivity Analysis

Monte Carlo analysis was conducted to assess the sensitivity of model outputs to uncertainties in fluvial and autochthonous inputs using the Simlab software program (Joint Research Centre 2006). Two outputs were considered: average soil C during 100–330 years when soil OM is in approximate equilibrium with inputs, and OM accumulation rates during the first 100 years. Input variables considered include C and N inputs from (1) leaf and small branch litter, (2) fine root turnover, (3) tree mortality, and (4) fluvial OM deposition, and (5) increases in silt and clay content due to fluvial deposition and weathering. The method of Sobol’ (1993), was used to generate 1536 samples for each input variable and generate global sensitivity indices for inputs. Overall model uncertainty was estimated as the standard deviation of model outputs.

Statistical distributions for each of the five input variables (Table 5), expressed as percent deviation from the default model inputs, were estimated as described below. Uniform distributions were assumed for all variables due to insufficient data to determine distributions for most inputs. Potential error in leaf litter and mortality inputs were estimated from related data collected in related Queets studies (O’Keefe and Naiman 2006; Van Pelt and others 2006). Although fine root turnover is generally weakly correlated with litterfall, most estimates of fine root turnover in mesic environments are between 50 and 150% of aboveground litterfall (Nadelhoffer and Raich 1992). To gauge potential error in estimated fluvial OM deposition, we identified two other studies in which both fluvial OM and sediment size were measured. Sediment size and OM were correlated in both studies. On the Middle Severn River, UK, floodplain deposits with 65% silt plus clay contained about 4% OM (Steiger and others 2001). OM was linearly correlated with sediment size in flood deposits along the Saskatchewan River, Canada (Morozova and Smith 2003), with about 4% OM in elevated floodplain sediments with a median clay-size, and higher OM concentrations in lacustrine and wetland deposits. We thus estimate a potential for 2–3 times higher OM deposition (assuming 0.45 g C g−1 OM) than was observed on the Queets River.
Table 5

Estimated Ranges of Input Variables as a Percentage of Default Simulation Values

Input

Low

High

References

Leaf and small branch litter

70%

130%

O’Keefe and Naiman (2006)

Fine root turnover

50%

150%

Nadelhoffer and Raich (1992); Uselman and others (2007)

Tree mortality

70%

130%

Van Pelt and others (2006)

Fluvial OM deposition

100%

300%

Steiger and others (2001); Morozova and Smith (2003); Bechtold and Naiman (unpublished)

Silt and clay

50%

150%

Bechtold and Naiman (unpublished)

Results

C Cycling

Simulated soil C accumulation closely approximated measured values, increasing rapidly to near-plateau concentrations of about 4000 mg ha−1 at about 100 years, and remained near that level for the length of the simulation (Figure 4A). Equilibrium soil C was relatively robust relative to uncertainties in inputs, with a standard deviation of 562 g m−2 (CV = 13%) estimated from uncertainty analysis within the range of likely inputs. The 34 g m−2 y−1 rate of C increase during the first 100 years is about the average observed in forest soils following agricultural abandonment (Post and Kwon 2000). Most of the modeled C increment was in the slow pool which accounted for 71% of soil C in soils more than 100 years old, and was also robust to input uncertainties (CV = 17%). Active C and the surface litter similarly reached maxima at about 100 years. Modeled surface litter C was somewhat higher (13%) than measured values (10%), and declined slightly over the following 230 years. Passive C accumulation increased from 8 to 11% of SOM by the end of the simulation. The transition to plateau C at 100 years occurred as annual inputs of leaf, fine root, and small branch litter came into equilibrium with respiration outputs (Figure 5). Fluvial deposition was the greatest source of OM to soils during the first 10 years, but was quickly eclipsed by autochthonous inputs and accounted for 33% of accumulated soil C in age 25 soils.
https://static-content.springer.com/image/art%3A10.1007%2Fs10021-009-9294-9/MediaObjects/10021_2009_9294_Fig4_HTML.gif
Figure 4

Soil C (A) and N (B) simulated over 330 years of floodplain development. Total C and N accumulation is indicated by sum of shaded areas. Shaded areas indicate sizes of individual soil pools: dark gray—surface and root litter; black—active pool; light gray—slow pool; diagonal bars—passive pool. Triangles indicate C and N measured in field studies. Dashed line in B indicates total simulated N when the model was altered to allow N-enrichment of structural litter and slow pool N after initial formation.

https://static-content.springer.com/image/art%3A10.1007%2Fs10021-009-9294-9/MediaObjects/10021_2009_9294_Fig5_HTML.gif
Figure 5

Simulated soil C relative to source illustrates the dominance of annual inputs of leaf, fine root, and small branch inputs (dashed line) relative to fluvially derived SOM (including particulate, sediment-associated, and fluvial large woody debris; thin line) and SOM resulting from tree mortality (thick line).

Total simulated soil respiration increased over the first 100 years to about 500 g m−2 y−1, at the approximate transition between alder and spruce forests, and then declined gradually to about 400 g m−2 y−1 at year 330 (Figure 6A). Respiration per unit soil C increased over the first 14 years as rapidly decomposing annual litter from the developing forest reached a maximum, and then declined as soil OM increasingly became dominated by the slow and passive OM pools, resulting in soil C turnover of about 10 years in the mature conifer forest. Litter, active OM, and slow OM were all important sources of respiration, contributing 44, 34, and 22% of respired C over the simulation period. Simulated C leaching peaked at about 2 g m−2 y−1 in 100-year-old soils (Figure 6B). This rate is at the high end of the range reported for soils (Neff and Asner 2001) but is not unexpected for sandy soils and still accounts for less than 0.4% of C outputs.
https://static-content.springer.com/image/art%3A10.1007%2Fs10021-009-9294-9/MediaObjects/10021_2009_9294_Fig6_HTML.gif
Figure 6

Simulated C and N gaseous and leaching outputs: A respiration; B DOC leaching; C denitrification; and D inorganic N leaching.

N Cycling

The CENTURY model was less successful at predicting changes in soil N. The model simulated increasing soil N to a peak at 100 years, and correctly predicted maintenance of soil N at equilibrium levels despite much reduced N fixation in the conifer forest. However, equilibrium soil N was underestimated by more than a third (Figure 4B). Inorganic N leaching increased to a maximum of 8 g m−2 y−1 at 39 years, with an average soil export of 7.8 g m−2 y−1 during the first 70 years when alder was dominant. Inorganic N leaching (Figure 6C) decreased to about 1.9 g m−2 y−1 in the older conifer forest, and was in approximate equilibrium with canopy N fixation by the end of the simulated period. These leaching rates are slightly higher than what would be expected from the results of previous studies of nitrate fluxes through red alder forests. Ritzenthaler (1998) found that about 30% of leached nitrate was removed from hyporheic water passing through Queets River red alder stands, thus yielding estimated stream export of 5.5 g m−2 y−1, compared with measured export of 4.0 g m−2 y−1 in Oregon red alder stands (Compton and others 2003). Simulated denitrification (Figure 6D) followed a similar pattern to inorganic N leaching, reaching a maximum of 1.35 g m−2 y−1 at 39 years, decreasing to 0.47 g m−2 y−1 in mature forests. Simulated organic leaching was a minor output, reaching a maximum of 0.21 g m−2 y−1 in 84- to 129-year-old soils and declining to 0.14 g m−2 y−1 by 330 years. Inorganic N leaching accounted for 82%, denitrification accounted for 14%, and organic N leaching accounted for 4% of total N outputs.

The discrepancy between measured and simulated soil N (Figure 4B) was due to underestimation of soil N immobilization rather than insufficient N inputs. The CENTURY model predicts an average bulk C:N ratio of 12.3 whereas the Queets soils had an average measured C:N ratio of 9.4. However, increasing the simulated N inputs had little effect on soil N concentrations; doubling N fixation rates increased soil N by only 11% but accelerated inorganic N leaching by 72%. Increasing soil N retention by restricting simulated water drainage by half increased organic N retention by 12% but also resulted in inorganic N accumulation to unrealistic levels (>10 mg g soil−1). In the CENTURY model, organic N concentrations are set at the time of entry into a given OM pool with no subsequent alteration until that OM is decomposed into another pool. Altering the CENTURY model to (1) simulate N enrichment of slow and structural OM after its initial formation to the maximum defined for that pool by the criteria governing its initial formation from active SOM (C/N ratio decreases from 20 to 10 as inorganic N increases from 0 to 2 g m−2 soil), and (2) simulate N enrichment of structural OM (C/N ratio decreases from 150 to 30 as inorganic N increases from 0 to 2 g m−2 soil) removes most of the discrepancy between observed and simulated N accumulation (dashed line in Figure 4B).

Relative Importance of Fluvial and Autochthonous Influences

Equilibrium (>100 years) OM was most sensitive to estimation of autochthonous OM inputs from leaf litter and fine root inputs, and fluvial deposition of silt and clay, and insensitive to tree mortality and fluvial OM deposition (Table 6). Simulation at the upper and lower limits of estimated litterfall and fine root inputs resulted in 23 and 24% differences, respectively, in equilibrium C, whereas there was a 19% difference in equilibrium C between upper and lower estimates of sediment deposition and subsequent weathering. The influence of sediment deposition on equilibrium soil OM is further appreciated by comparing simulation results at the extremes of silt and clay content textures encountered in 40- to 330-year-old Queets soils (16–60% silt, 1–14% clay), with about 30% greater equilibrium C in the fine soil than the coarse one. The effect of sediment deposition on N retention was even greater due to the concentration of N-rich OM associated with silt and clay, with 40% higher N in soils experiencing high silt/clay deposition (Figure 7).
Table 6

Sensitivities of Equilibrium C, Equilibrium N, and 0–100 Year C Accumulation Rate to Uncertainties in Inputs Variables.

Input

Equilibrium C (%)

Equilibrium N (%)

0–100 Year C rate (%)

Leaf and small branch litter

28

15

32

Fine root turnover

47

51

15

Tree mortality

2

1

4

Fluvial OM deposition

1

2

36

Silt and clay

29

38

15

https://static-content.springer.com/image/art%3A10.1007%2Fs10021-009-9294-9/MediaObjects/10021_2009_9294_Fig7_HTML.gif
Figure 7

Simulated soil C within the range of silt and clay concentrations found in age 40 and greater sites (thin solid lines) indicates a moderate influence of texture on C retention. Texture was a more important influence on simulated N (dashed lines) due to the concentration of N-rich OM on clay and silt particles.

Soil OM accumulation during the first 100 years was more sensitive to estimation of fluvial OM deposition and less sensitive to estimation of autochthonous and sediment inputs (Table 6). At the default estimate of OM deposition, fluvial sources were an important source of OM to soils only during the first few decades of floodplain development, accounting for only 12% of accumulated C by 50 years (Figure 8). However, at the upper limit of estimated OM deposition, fluvial sources accounted for 26% of accumulated OM at 50 years and 12% of accumulated OM at 100 years. Inputs from tree mortality were greatest during the transition from alder to conifer forest from about 60 to 110 years but not an important influence on either 0–100 year C accumulation or equilibrium C (data not shown). It is also important to keep in mind that the model only considers the uppermost 20 cm of the soil profile. Approximately 22% (3287 g m−2) of simulated fluvial and autochthonous C inputs and 26% (258 g m−2) of N inputs were buried beneath the 20 cm depth by fluvial deposition during the first 40 years. Most (65%) of the buried OM derived from the slow pool. Shallow burial in well-drained soils such as these would be expected to minimally affect decomposition, and most would be expected to be decomposed within the first hundred years.
https://static-content.springer.com/image/art%3A10.1007%2Fs10021-009-9294-9/MediaObjects/10021_2009_9294_Fig8_HTML.gif
Figure 8

Soil C under alternative fluvial C deposition scenarios: thick solid line indicates default estimated OM deposition; dashed line indicates no OM deposition; dotted line indicates 200% of default OM deposition; thin solid line indicates 300% of default OM deposition.

Discussion

The accumulation of SOM to levels capable of providing a stable source of nutrients and water to biota is an important event in the evolution of terrestrial ecosystems, and is directly related to soil texture. As with the upland soils to which CENTURY is usually applied, mineral sediments play a fundamental role in constraining floodplain SOM cycling. We found that SOM accumulated to equilibrium concentrations similar to those predicted for similarly textured mature soils. Fluvial deposition was an important influence on SOM accumulation but not for the reasons commonly supposed. Fluvially derived OM was small and a short-lived influence on SOM relative to autochthonous sources. Fluvial sediments, on the other hand, were an enduring influence on SOM retention. The model was less successful at predicting N accumulation due to underestimation of immobilization. Although the mechanism behind the discrepancy between modeled and measured soil N accumulation cannot be determined from this study, high inorganic N and amorphous Fe concentrations create conditions favorable for abiotic N immobilization.

How Does Fluvial Deposition Influence Floodplain SOM Cycling?

The simulation results lead to the somewhat surprising conclusion that the texture of deposited sediments played a fundamental enabling role in floodplain soil OM accumulation, but that direct OM deposition was of little consequence. This was clearly the case in soils older than 100 years old, where texture played an important part in defining near-equilibrium OM concentrations and even the highest estimates of fluvial OM deposition were of little consequence. However, soil OM accumulation was relatively sensitive to estimation of fluvial OM deposition and insensitive to texture and autochthonous inputs during early site history (<100-year-old soils), when fluvial OM deposition might be especially important to soil moisture and nutrient regimes.

Our estimates of fluvial OM deposition suggest that fluvial OM composes a significant fraction of SOM only during the first few decades of floodplain development, and are consistent with measured patterns of soil OM accumulation. Higher estimates of fluvial OM estimated in other studies (Steiger and others 2001; Morozova and Smith 2003) would place somewhat greater importance on fluvial OM deposition, but are less consistent with observed OM accumulation rates. In part, differences in estimation of fluvial OM deposition in these studies may be due to greater deposition of non-sediment-associated organic debris in the other rivers, which lie within low gradient drainages and have long hydroperiods. However, the relatively unweathered state of Queets sediments probably accounts for most of the differences. Queets sediments derive largely from glacial outwash deposits, with silt and clay-sized particles consisting mostly of feldspar fragments (S. Bechtold, unpublished data), which have a low reactive surface area relative to the secondary minerals that are more abundant in geologically older environments.

The results indicate that observed patterns of rapid OM accumulation during floodplain succession are consistent with stabilization mechanisms implemented in the CENTURY model. Sediment characteristics also affect soil OM and nutrient dynamics indirectly by influencing primary productivity. In the CENTURY model, plant productivity decreases under conditions of less than optimal moisture, N or phosphorus availability. However, moisture and inorganic N levels never decreased to levels where they would restrict primary productivity in Queets soil simulations. Measured indices of phosphorus availability (Bechtold and Naiman, unpublished) indicated abundant plant-available phosphorus throughout the chronosequence. However, in practice it is difficult to distinguish the influence of sediments on decomposition rates from their influence on net primary production. The two effects are not independent; OM stabilization by sediments contributes to many of the factors potentially affecting primary productivity. C isotope data (for example, Gaudinski and others 2000) showing the great age of OM adsorbed to sediments provides compelling evidence for a sediment influence on decomposition but, like incubation studies conducted for a limited period under controlled temperature and moisture regimes, leaves large uncertainties as to the kinetics.

Silt and clay concentrations also influence solute leaching rates. Dissolved organic C (DOC) leaching was small relative to respiration losses, and was unimportant for the mass balance of soil C. Such low DOC leaching is impressive but similar to rates observed in older soils (Neff and Asner 2001). However, even low DOC leaching rates have important implications for DOC loads in surface and ground waters. Correlations between river DOC and soil texture of the surrounding watershed have been observed in several studies (Nelson and others 1993; McClain and others 1997). This influence might be especially pronounced in hyporheic/ground water sediments, where denitrification is often limited by C availability (Hedin and others 1998). Overlying soils are a major C source to hyporheic biota in the Queets (Clinton and others 2002) even where the hydrology is dominated by infiltrating river water. The large differences in simulated DOC leaching between coarse and fine soils (Figure 6A) suggest that spatial variation in soil texture could be an important source of spatial heterogeneity in hyporheic microbial respiration.

How Does Floodplain SOM Cycling Compare with Upland Forest Soils?

The 10-year SOM turnover reported here is slightly faster than the 12-year turnover reported during secondary succession of forests on abandoned agricultural land, and much faster than the greater than 25 year turnover times typical of mature forests (Harrison and others 1995). Rapid SOM turnover is a key characteristic of these soils. They are highly dynamic, with large proportions of SOM recycled at short to intermediate time scales and gross nutrient mineralization rates several times higher than in mature forest soils of similar nutrient capital. Only small amounts of C and N were sequestered in the passive pool, which is dominated by N-rich OM associated with clay particles. Although the capacity for passive OM storage increased quite rapidly as clays were weathered from coarser sediments, data from approximately 1000-year-old soils (Bechtold and Naiman, unpublished) suggest that the period of rapid weathering is fairly short-lived. Also lacking in these soils are the thick fibrous organic layers characteristic of old growth coniferous forests which can immobilize large amounts of nutrients for long periods. Formation of a thick organic layer may be greatly facilitated by senescence of the original cohort of spruce trees still present as live biomass at the end of the simulated period.

What Factors Explain Discrepancies Between Modeled and Measured N Accumulation?

Poor performance of the CENTURY model in predicting forest soil N immobilization was also observed by Kirschbaum and Paul (2002), who found that immobilization was over-estimated in forest litter. All immobilization in the CENTURY model occurs during microbially mediated transformations between soil pools. Neither is there provision for N-enrichment of OM in a given soil pool after its initial formation nor is there a mechanism for immobilization in structural OM. However, immobilization can also occur abiotically (Davidson and others 2003; Fitzhugh and others 2003). Conditions favoring abiotic immobilization include high N status (Berntson and Aber 2000) and by the presence of non-crystalline iron, which tends to be abundant in these (Bechtold and Naiman, unpublished) and other riverine soils (Darke and Walbridge 2000; Poulton and Raiswell 2002). Abiotic mechanisms were responsible for 78% of N immobilization in other Washington red alder stands (Johnson and others 2000). Plausible pathways for abiotic immobilization in the CENTURY model include the transfer of surface structural OM to the slow pool and immobilization of N percolating through the forest floor into structural OM. In both cases, N concentrations are determined exclusively by the C/N ratio of the OM entering the pool, even in the presence of high inorganic N concentrations. Modifying the model to simulate the ongoing influence of inorganic N concentrations (dashed line in Figure 4B), such as might result from an abiotic mechanism, removed most of the discrepancy between simulated and observed soil N concentrations.

The model estimated denitrification to account for only about 14% of N outputs. However, Century 4.0 performs only rudimentary simulation of denitrification. Models incorporating high spatial and temporal resolution are considered necessary to simulate the interfaces where most denitrification occurs (Boyer and others 2006). The coarse, well-drained soils with little aggregation simulated in this study rarely experience anoxic conditions and would probably not be expected to have high denitrification rates. In a synthetic study of European floodplains, denitrification was found to be a significant flux in soils above a threshold of 60% silt and clay (Pinay and others 2007). All but one of the Queets soils fall below this threshold. Not simulated in this study are the abundant swales, side channels, and extensive hyporheic area characteristic of the Queets valley. Denitrification in these areas may account for the 30% N removal rate observed in another Queets study (Ritzenthaler 1998).

Conclusions

Redistribution of sediments plays a pivotal role during the early history of terrestrial ecosystems. Sediment size influences initial plant colonization (Kalliola and others 1991; Hupp and Osterkamp 1996; Robertson and Augspurger 1999), and colonizing plants trap sediments, shaping alluvial landforms (Bennett and Simon 2004; Gurnell and Petts 2006). The results presented here suggest the complex mosaics of alluvial landforms characteristic of natural floodplains may confer substantial ecological heterogeneity in soil processes as well as providing habitats for a diversity of plant species. In addition, the high amount of edge between soil volumes differing in resource (C and N) availability and hydrological conductivities creates conditions favorable for redox-driven reactions (McClain and others 2003), contributing to the capacity of riparian areas to buffer nutrient fluxes (Correll 1997).

Human activity has substantially altered the sedimentary regime of most rivers through dams, levees, road construction, and other activities affecting the supply and routing of sediments (Syvitski and others 2005). Although human alteration of flow regimes is believed to result in loss of ecological heterogeneity (Poff and others 1997; Tockner and others 2007), the degree to which altered sediment distribution contributes to loss of ecological function is unknown. Spatially explicit modeling offers promise for a better understanding of how the underlying sedimentary template influences watershed biogeochemistry (Boyer and others 2006; Groffman and others 2009). Such an analysis could inform the choice between passive restoration methods intended to allow a river to assert it natural variability and more engineered solutions to re-creating desirable river characteristics.

Of course, river valleys exhibit different styles of sedimentation depending on geomorphic and climatic context (Nanson and Croke 1992). The results presented here are most directly applicable to piedmont rivers, where reduced stream energies result in deposition of coarse sediments eroded from mountains, which can be especially pronounced in areas where glacial history or tectonic activity contributes to sediment supply. Less extreme expressions of the same processes influence ecological patterns in other medium to high energy rivers, which comprise a large proportion of the length of most river systems in temperate climates (Schumm 1977; Nanson and Croke 1992). The model may be less applicable to large lowland floodplains where prolonged seasonal inundation and the predominance of suspended sediment deposition creates a very different template for soil processes (Junk and others 1989; Spink and others 1998). The model also makes no account of riverine wetlands, where restricted gas exchange may result in very different patterns of OM cycling.

Acknowledgments

We thank Robert Edmonds and Kristiina Vogt for useful comments on early drafts. Research support was received from the Andrew W. Mellon Foundation, the National Science Foundation and the UW School of Aquatic and Fishery Sciences.

Copyright information

© Springer Science+Business Media, LLC 2009