The influence of soil age on ecosystem structure and function across biomes

Ecosystem development over centuries to millennia is traditionally attributed to five state factors that also control pedogenesis: time (soil age), climate, topography, parent material, and organisms (vegetation). While soil age is hypothesized to be a major driver at local scales, it is not the only determinant of ecosystem structure and function, as ecosystems with the same soil age can differ markedly. A comprehensive, cross-biome assessment of the relative importance of soil age compared with other state factors has been lacking because most studies have examined spatial (climate, vegetation) or temporal (age) gradients separately. This study asks how soil age compares to parent material, climate, vegetation type, and topography in explaining variation in multiple ecosystem structural and functional properties across global biomes, and how environmental context influences the absolute values, trajectories, and rates of development in these properties.

Previous work using long-term soil chronosequences has documented changes in biogeochemical cycles and biotic communities with ecosystem development, including shifts in C stocks, C:N:P stoichiometry, and soil pH. Foundational studies suggest phosphorus declines and potential ecosystem retrogression with age, primarily in mesic systems, whereas drier environments may show weaker development trends. Regional analyses across climatic gradients (e.g., Western Australia) suggest climatic context modulates chronosequence patterns. Despite these advances, a global, quantitative evaluation partitioning the contributions of the five state factors to ecosystem properties across biomes has been missing.

The study combined two approaches. (1) Cross-biome field survey: New standardized field data were collected (2016–2017) from 16 globally distributed soil chronosequences spanning hundreds to millions of years and diverse climates (tropical, temperate, continental, Mediterranean, polar, arid), vegetation types (grasslands, shrublands, forests, croplands), and parent materials (volcanic, sedimentary, dunes, glacier). Sampling focused on 0–10 cm topsoil. Each chronosequence stage included a 50 m × 50 m plot; within each, five composite topsoil samples were taken under dominant vegetation. In total: 16 chronosequences, 87 plots, 261 transects, and 435 soil samples. Thirty predictors described the five state factors: soil age (quantitative and categorical metrics), climate (WorldClim: MAT, MAP, seasonality), vegetation type, parent material (substrate origin, lithology, USDA soil order), and topography (elevation, slope, aspect). Ecosystem properties measured (32) spanned soil properties (CIA, TBR, texture, pH, total P by H2SO4 and HF), stoichiometry (C:N:P ratios), water resources (water-holding capacity, potential infiltration), nutrient cycling (available N, Olsen P; enzyme activities for β-glucosidase, N-acetylglucosaminidase, phosphatase; lignin and glucose respiration), carbon cycling (soil organic C concentration and stocks to 10 cm via bulk density; soil respiration incubations), plant production (MODIS NDVI), vegetation composition (cover of trees, shrubs, grasses, forbs, open), microbial structure (PLFA-derived total, bacterial, fungal biomass; fungi:bacteria ratio; amplicon-based relative abundance of ectomycorrhizal and arbuscular mycorrhizal fungi from 18S rRNA sequencing and FUNGUILD assignments). (2) Meta-analysis: A literature synthesis added 48 independent chronosequences worldwide (excluding the 16 new sites) with comparable surface soil data (top 10 cm) for soil C stocks, total P, pH, texture, and C:N ratio. Data were extracted via SCOPUS (keywords including chronosequence and carbon, nitrogen, phosphorus, biomass, diversity). Analyses: Variation Partitioning Modeling (vegan::varpart in R) quantified unique and shared variance explained by soil age, climate, vegetation, parent material, and topography for each ecosystem property; negative adjusted R2 components were set to zero. Partial Spearman correlations (ppcor::pcor) assessed associations between environmental variables and properties controlling for soil age and geography. Within-chronosequence trends were evaluated using Spearman rank correlations between chronosequence stage and properties; PERMANOVA tested changes across stages for microbial biomass, plant cover, and mycorrhizal groups. Current climate was used as the climate surrogate and cross-validated against paleoclimate, showing relative stability over long periods for many chronosequences.

Across biomes, soil age was significant but relatively weak. Collectively, parent material, climate, vegetation type, and topography predicted on average 24 times more variation in ecosystem properties than soil age alone. On average (across 32 properties), unique variance explained: soil age 2.1% (shared with environment 3.5%); parent material 14.5% (about 7× soil age), climate 8.0% (∼4×), vegetation 6.1% (∼3×), topography 1.5%. Parent material, climate, and vegetation were especially predictive of bacterial biomass (33.9%), soil pH (21.2%), and texture (16.2%), respectively. Despite lesser global importance, soil age uniquely explained notable fractions for soil N:P (7.8%) and C:P (6.3%) ratios, soil pH (6.3%), total base cation reserve (TBR; 7.3%), plant productivity (5.4%), and soil C stocks (5.1%). The meta-analysis of 48 additional chronosequences corroborated that soil age explains a smaller but significant portion compared to other state factors combined. Environmental context strongly determined absolute values and trajectories: sandy substrates had lower total and available P, C stocks, respiration, microbial biomass, and CIA than volcanic or sedimentary substrates; sandy soils showed steeper increases in N:P and C:P with time but flat/negative development in respiration, microbial biomass, C stocks, and available P. Drier and non-forested ecosystems had more alkaline soils, lower C stocks, lower plant productivity, lower C:P and N:P, lower microbial biomass, a smaller proportion of ectomycorrhizal fungi, and weaker increases in productivity over time than mesic forests, which showed more acidic soils, higher weathering (CIA), greater microbial biomass, and higher tree cover. Comparisons at equal soil ages highlighted context effects: 1000-year soils from temperate Mexican forests and Australian shrublands had ∼2× C and microbial biomass and ∼4× higher N:P than arid U.S. sites; 20,000-year Hawaiian tropical forest soils had ~13× C, ~71× N:P, and ~13× microbial biomass than similarly aged arid soils (New Mexico, Bolivia); 3–4 Myr Hawaiian volcanic forest soils had ~17× microbial biomass and ~12× N:P than Arizona soils of similar age. Within-chronosequence trends were consistent for over two-thirds of properties: generally increasing with age for soil N:P (13/16 ecosystems) and C:P (11/16), total microbial (10/16), bacterial (9/16) and fungal biomass (9/16), C stocks (9/16), available P (8/16), tree cover (7/16), CIA (7/16), and P mineralization (7/16); generally decreasing for soil total P (P-HF) and pH (8/16); generally unchanged for lignin degradation (15/16), glucose respiration (15/16), soil respiration (14/16), wood resources (14/16), arbuscular mycorrhizal fungi (14/16), plant functional group cover and potential infiltration (>12/16). Some properties showed biome- or site-dependent patterns (e.g., sugar/chitin degradation, fungi:bacteria ratios, plant productivity, plant composition).

The findings demonstrate that, although soil age contributes to ecosystem development, environmental context—parent material, climate, vegetation, and topography—dominates both the distribution and long-term trajectories of ecosystem properties across biomes. Sandy, dry, and non-forested contexts limit development of key functions (C storage, nutrient availability, microbial biomass), whereas mesic forests foster greater weathering, acidity, microbial biomass, and productivity gains over time. Thus, ecosystems of similar age may lie at different positions along development due to context, challenging assumptions that time alone governs structure and function. These insights imply that shifts associated with global change (warming, drying, deforestation, land-use) will alter the conditions under which ecosystems develop, potentially slowing or redirecting long-term trajectories. Nevertheless, including soil age can improve models for specific properties (e.g., stoichiometric ratios, pH, P pools, microbial biomass, C stocks) that show consistent age-related patterns.

Soil age is a significant but relatively weak driver of ecosystem structure and function across biomes compared to parent material, climate, vegetation, and topography, which collectively explain vastly more variation. Environmental context governs both absolute levels and trajectories of ecosystem development, with drier, sandy, and non-forested systems showing reduced development relative to mesic forests. Despite this, soil age helps refine predictions for key slowly changing properties (e.g., N:P, C:P, pH, C stocks, microbial biomass). The study quantifies the relative roles of state factors globally and highlights that ongoing climatic and land-use changes may slow or alter long-term ecosystem development. Future work should expand chronosequence coverage in underrepresented regions (e.g., continental Africa), integrate deeper soil horizons, and incorporate dynamic climate histories and disturbance regimes to further resolve mechanisms shaping development trajectories.

Geographic coverage is uneven (notably limited representation of some regions such as continental Africa). Analyses focused on topsoil (0–10 cm), potentially missing deeper profile dynamics. Current climate was used as a surrogate for climate history, although cross-validation suggested relative stability in many sites. Some ecosystem properties exhibited biome- or site-specific or inconsistent relationships with age, indicating influences of local idiosyncratic conditions (e.g., dust inputs, erosion, disturbance). The meta-analysis was restricted to variables commonly reported with comparable methods. Variation partitioning can yield small or zero unique fractions and treats negative adjusted R2 as zero. Seasonal timing of sampling varied but was assumed to have minimal influence given the cross-biome design.