1 Introduction

Urbanization is homogenizing areas across the moisture gradient of the United States [1] and residential soils are specifically homogenizing [2]. Land development has an impact on ecosystems, and in particular, soil quality. In residential developments, soil quality is often degraded by removal of predevelopment topsoil, regrading with low quality fill dirt, and compaction by heavy machinery traffic during construction [3, 4]. Compaction and the low organic matter content in urban soils reduces plant-available water storage, reduces infiltration, and increases surface runoff, compared to soils which are undisturbed [5,6,7,8]. Additionally, sandy soils common to many regions are especially susceptible to reduced infiltration due to compaction [6, 8].

Residential landscapes are commonly composed of some combination of turfgrass and landscape beds. Milesi et al. [9] estimated turfgrass covers 1.9% of the United States land areas, as it is effective for erosion control and costs less than landscape bed plantings per unit area installed. Florida residents use over half of domestic freshwater consumption outside the home for landscape irrigation [10]. Despite receiving a statewide average of 137 cm precipitation yr−1 of precipitation [11], Florida uses the fourth most water (11,500 Mgal d−1) of any state in the United States, and the most of any state east of the Mississippi River [12]. Florida’s population and number of visitors have increased over time to 21.5 million residents in 2020 [13] and to 131 million out-of-state visitors in 2019 [14], respectively. As a result, in 2015, Florida’s domestic freshwater use (2,215 Mgal d−1) surpassed agricultural use (2,089 Mgal d−1 [10]) for the first time in recorded history.

Soil quality may be defined as "the capacity of a specific kind of soil to function, within natural or managed ecosystem boundaries, to sustain plant and animal productivity, maintain or enhance water and air quality, and support human health and habitation” [15]. Past studies have shown that soil quality tends to improve over 20–55 years in urban, turfgrass-based systems as organic matter accumulates [16,17,18,19]. Accumulation of organic matter and changing soil structure can affect soil quality and resources needed to support landscape vegetation. These studies took place across a variety of climates and locations within the United States, including Colorado and Wyoming (putting greens) [16], Maryland (lawns) [17], Ohio (golf courses) [18], and Alabama (untreed lawns) [19]. However, these studies were either from golf courses or from lawns across municipal areas with varying land development and landscape management practices. No previous studies have investigated the evolution of soil quality over time under both lawns and landscape beds within the constraints of individual residential developments, to minimize confounding factors of soils, development practices, and landscape management practices.

Further, soil textures were variable among previous studies, with most sites characterized by loamy soils and only one site (putting greens [16]) exhibiting sandy soils. Sandy soils are common throughout the state of Florida (Fig. 1), and the Candler series (hyperthermic, uncoated Lamellic Quartzipsamments) is the most widespread soil in the state [20]. Well-drained conditions combined with temperature and rainfall accelerate mineralization, resulting in low soil organic carbon content (< 1%) for many of Florida’s sandy soils [21, 22]. Undisturbed and non-agricultural A horizon soils in Florida, that are moderately- to excessively well-drained, average less than 1.5% organic carbon (22). By comparison, previous studies from other regions of the US, where conditions are less favorable for mineralization, show organic matter increases with time.

Fig. 1
figure 1

The gravimetric concentration of sand-sized particles in 1615 surface soils collected and analyzed during state soil survey of Florida [22]

Full size image

Given Florida’s expected population growth, and the growth of urban areas worldwide, it is important to understand the impact of development on urban soils and their future evolution. A scarcity of evidence for improvement to soil quality over time in urban developments with sandy soils or a sub-tropical climate warranted study of this phenomenon within the state of Florida where accumulation of soil organic matter in undisturbed soils is limited in well drained soils [20, 21]. The overall goal of this study was to assess soil quality (0 to 30 cm) change over time since initial disturbance in two residential developments under lawns and landscape beds, located in Central and Southwest Florida. in regions. Specifically, the objectives were to: (1) identify trends in soil properties (bulk density, organic matter content, and organic carbon accumulation) over time for three land cover types (landscape beds, lawns, and common areas); (2) compare and contrast analyzed relationships based on sample collection sites (Central or Southwest FL), location (landscape beds, lawns, common areas), and sample depth to determine similarities and differences.

In Sect. 2, the approaches are presented for achieving the objectives via soil sampling from residential lots and the subsequent analyses. Data and analytical results are presented in Sect. 3 and then put into context with previous findings from literature in Sect. 4. Overall conclusions pertaining to soil development on sandy soils and considerations for future research on urban soil trajectories are included in Sect. 5.

2 Materials and methods

2.1 Site descriptions

Lots were sampled from two urban residential developments in Florida (Fig. 2): On Top of the World (OTOW) in Central Florida (latitude 29.1°N longitude 82.3°W), and Lakewood Ranch (LR) in Southwest Florida (latitude 27.4°N longitude 82.4°W). Mean annual rainfall for OTOW is 1318 mm and mean annual temperatures is 21 °C; for LR, mean annual rainfall is 1377 mm, and mean annual temperature is 23 °C [23] Prior to development most areas within OTOW were primarily mapped as the Candler soil series (hyperthermic, uncoated Lamellic Quartzipsamments; [20]), the most common soil series in Florida. Prior to development in LR areas were mapped as either the Myakka series (sandy, siliceous, hyperthermic Aeric Alaquods) or the Eau Gallie series (sandy, siliceous, hyperthermic Alfic Alaquods) [20]. However, published soil surveys do not reflect potential additions of fill or other landscape modifications that may have occurred during home construction [8] as soil surveys were published for Marion County (OTOW) and Manatee County (LR) in 1978 and 1980, respectively; before either development began construction.

Fig. 2
figure 2

Location of the On Top of the World and Lakewood Ranch communities in the state of Florida, USA (left). Aerial views of On Top of the World and Lakewood Ranch communities (right) above and below, respectively

Full size image

Residential lots in LR (LR-R; n = 50) were sampled in 2020 and 2017 for OTOW (n = 45), while the LR common areas (LR-CA; n = 16) were sampled in 2015 (Table 1). Lot age and common area age were calculated based on the year the property was developed and the year the property was sampled. The development date of each residential lot was verified using the Marion and Manatee County Property Appraiser’s databases, respectively, for OTOW and LR [24, 25]. Lot ages ranged from 0 to 30 years since development (1986–2016) for LR-R locations, 0 to 20 years (1995–2015) for LR-CA, and 0 to 22 years for OTOW (1998–2020).

Table 1 The number of lots sampled is within each lot age range for two residential developments in Florida. The sampling year and the age of the oldest lot at the time of sampling were used to calculate the lot age
Full size table

For OTOW and LR-R, between six and 19 lots were selected within five age group categories: 0–5 yrs., 6–10 yrs., 11–15 yrs., 16–20 yrs., ≥ 21 yrs. OTOW had a total of 45 sites while LR-R had 50 sites. For LR-CA, three to four locations were sampled from within areas 0, 1, 5, 13, and 20 years after development, for a total of 111 sites.

2.2 Soil sampling

For LR-CA, a soil core sampler was used to collect samples from 0 to 6 cm and 6–30 cm depths. For the 6–30 cm depth, the sampling started at the 15 or 16 cm depth to capture the midpoint of this subsurface layer. A minimum of two subsamples per depth were taken within each common area.

For OTOW and LR-R, survey flags were placed in random locations throughout the lawn and landscape beds to mark sampling locations (Fig. 3-A). In each lot, subsamples were collected from five locations in the front lawn and five locations in the front landscape beds, excluding any areas between the sidewalk and the street. Landscape beds containing utility boxes, mature trees, gravel, or landscaping fabric were avoided. Mulch in landscape beds was pulled aside to allow for penetration of the soil sampler. For OTOW only, soil samples were collected from new lots (Age = 0 yrs.), just before installation of turfgrass or landscape beds. Nine total sampling locations were randomly selected within the new lots, and the mean of the nine subsamples was used to represent one lot.

Fig. 3
figure 3

Soil samples collection process for lawns and landscape beds. White circles (a) show a partial distribution of survey flags for the selection of sampling locations. (b) shows the use of a punch tube to collect samples, and (c) shows the process of measuring and cutting samples into the top 0–15 cm and lower 15–30 cm portions

Full size image

Using a soil core sampler to determine bulk density at multiple depths is destructive [26], and was therefore impractical for collecting samples in the LR-R and OTOW settings. Instead, soil samples were collected using narrow, tubular soil samplers, or punch tubes. For LR-R, a 56 cm punch tube with an internal cutting diameter of 1.9 cm was used to collect samples. In OTOW, a 121 cm punch tube with an internal cutting diameter of 1.7 cm was used for soil sample collection (18 from pre landscaping and 144 from aged lots). In both OTOW and LR residential areas, the punch tube was pushed into the ground to a depth of 30 cm or until refusal, and the depth to refusal was recorded (Fig. 3-B).

Occasionally, high levels of compaction, tree roots, rocks, or other obstructions were prohibitive to collecting a sample deeper than 15 cm. If a sample at a depth of at least 15 cm could not be obtained, a new sampling point within 1–2 m of that was selected for resampling. Total sampling depth was recorded to account for samples ranging in total depth from 15 to 30 cm.

Each soil sample was divided into an upper 15 cm section and a lower section of up to 15 cm. Any surface vegetation or turf thatch were cut from the lawn-collected samples before dividing into upper and lower depths (Fig. 3-C). The five subsamples taken at each location (lawn or landscape bed) and depth (0–15 cm or 15–30 cm) were composited by sample location and depth (e.g., lawn & 0–15 cm depth) for each lot, which yielded four composite samples per lot.

2.3 Soil sample analyses

2.3.1 Bulk density

Soil samples were passed through a #10 sieve (< 2 mm) to remove oversized particles, including roots, gravel, and other debris. All soil samples were oven dried at 105 °C for 48 h or until weights no longer declined. Dry bulk density was calculated using the sieved oven-dry mass and cylindrical volume of either (1) soil plugs (cutting diameter and length) collected with the punch tube or (2) soil core collected with soil core sampler. While gravels and other rock fragments in the soil were infrequent, > 2 mm-sized particles, including roots, were removed from the bulk density determination.

2.3.2 Soil texture

Soil particle size distributions for soil samples from OTOW and LR-R sites were determined by laser diffraction analysis, a volumetric determination of particle size, using a Beckman Coulter LS‐13,320 multi‐wave particle size analyzer at the University of Florida Environmental Pedology and Land Use Laboratory. Subsamples from composite samples were analyzed without dispersion or pre-treatment [27]. Particle size distributions are summarized here as the volumetric concentrations of sand- (2.0–0.05 mm), silt- (0.05–0.002 mm), and clay-sized (< 0.002 mm) particles. For the LR common areas, subsamples were analyzed for sand, silt, and clay-sized particles using the hydrometer method [28]. The textural class for each sample was determined using the relative amounts of each particle size [29].

2.3.3 Soil organic matter

Soil organic matter (SOM) concentration was determined by loss-on-ignition (LOI) of a 5 g sample, with a balance accuracy of 0.001 g (0.02%), an ignition temperature of 550 °C, and a total ignition time of 4 h. While LOI can over-estimate SOM due to crystalline and hygroscopic water, and CO2 loss from carbonate, soils were very low in clay and CaCO3 content, minimizing the potential for most common sources of error. Previous studies have found that LOI has comparable coefficients of variation to dry combustion methods, 5.6% versus 4.9%, respectively [30]. Soil organic carbon (SOC) was estimated from SOM as 0.50*SOM [31, 32]. Carbon density was then calculated by multiplying SOC by bulk density and depth of the soil sample. A carbon accumulation rate was estimated by the difference in mean carbon density between 0- and 22-year-old lots for LR-R and 0- and 30-year-old lots for OTOW.

2.4 Data analysis

Linear regression models were developed using R [33] to test for associations between lot age and soil organic matter, and bulk density. Separate linear models were generated for soil samples collected from lawns and landscape beds, as well as for the upper and lower portions of the collected soil profile. Assumptions of normality were assessed by visual inspection of a normal Q-Q plot and histogram of residual values, and by the Shapiro–Wilk test. Homoscedasticity was evaluated visually by plotting model residuals versus predicted values and by the Breusch–Pagan test.

In several cases, the statistical significance of the linear regression model was evaluated on transformations of the dependent variable to satisfy linear model assumptions. The Box-Cox method [34] was used to select the most appropriate power transformation (e.g. log, square root, cubic, etc.). Simple linear regression assumptions could not be satisfied by transformation in the analysis of soil texture, so the Thiel-Sen estimator method, a nonparametric assessment of linear trends, was used instead of a simple linear regression [35]. Base R [33] and packages ‘car’ [36], ‘ggplot2’ [37], ‘ggpubr’ [38], ‘lmtest’ [39], ‘MASS’ [40], and ‘mblm’ [41] were used for data organization, analysis, modeling, and visualization.

3 Results

3.1 Soil particle size

All soil samples from LR-R lots were dominated by sand-sized particles. While texture is strictly based on particle mass, rather than volume, textures were estimated to put soils into context. For most soil size fractions (sand, silt, clay), results of the Thiel-Sen estimator method indicated a significant association (P < 0.05) between lot age (years) and particle size. However, any differences in soil texture observed over the range of lot ages represented in this study (0–22 years) were not large enough to change textural classifications. All soils collected from LR-CA had a sand texture with approximately 1% clay.

Soil samples from OTOW and LR typify the coarse-textured nature of FL soils (Table 2). Volumetric concentration of sand-sized particles averaged more than 95% from 0 to 15 cm in both lawns and landscape beds at both residential developments, and always exceeding 84%. From 15 to 30 cm, the concentration of sand-sized particles averaged 94% in lawns and landscape beds of both residential developments, and never less than 78%. These volumetric concentration estimates correspond to the USDA texture classes of sand and loamy sand as defined gravimetrically on the soil texture triangle.

Table 2 Volumetric concentration (%) of sand-, silt-, and clay-sized particles of residential lots sampled in On Top of the World and Lakewood Ranch. Values are shown as means ± standard deviations with the range listed below as minimum – maximum
Full size table

3.2 Bulk density

3.2.1 OTOW

For OTOW, bulk densities ranged from 1.02 to 1.84 g cm−3 at depths of 0–15 cm and from 1.19 to 2.31 g cm−3 at depths of 15–30 cm. Increased lot age was associated with a decreased soil bulk density for all sampling depths and locations (p < 0.05), and changes in bulk density with lot age had a higher adjusted R2 compared to trends observed in LR-R (Fig. 4). For lawn-collected soil samples (Fig. 4a), lot age significantly predicted soil bulk density at a depth of 0–15 cm (inverse cube root transformation; F(1, 43) = 98, P = 1.1*10–12, R2adj = 0.69) and at a depth of 15–30 cm (F(1, 42) = 54, P = 4.7*10–9, R2adj = 0.55). For soil samples collected from landscape beds (Fig. 4c), lot age significantly predicted soil bulk density at a depth of 0–15 cm (F(1, 43) = 49, P = 1.1*10–8, R2adj = 0.52) and at a depth of 15–30 cm (F(1, 42) = 37, P = 3.4*10–7, R2adj = 0.45). Additionally, the regression models suggest that over 30 years, mean bulk density at a depth of 0–15 cm decreased by 0.41 and 0.39 g cm−3 (34% and 32% change) in lawns and landscape beds, respectively. Bulk density decreased to a lesser extent at a depth of 15–30 cm, with a mean decrease of 0.38 and 0.39 g cm−3 (19 and 20% decrease) for lawns and landscape beds, respectively.

Fig. 4
figure 4

Bulk density in relation to residential lot age for soil samples collected at depths of 0–15 and 15–30 cm for OTOW (a and c) and LR-R (b and d). Lawns (a and b) and landscape beds (c and d) are represented. The shaded area represents a 95% confidence interval for the regression. *P < 0.05; ***P < 0.001. For OTOW, 0-year-old lots were sampled before establishing lawns or landscape beds

Full size image

3.2.2 LR-R

For LR-R, bulk densities ranged from 0.84 to 1.95 g cm−3 at depths of 0–15 cm and from 1.04 to 2.31 g cm−3 at depths of 15–30 cm. Increased lot age was associated with a decrease in soil bulk density for all sampling depths and locations (P < 0.05), with weaker associations for samples collected at a depth of 15–30 cm. For lawn-collected soil samples (Fig. 4b), lot age significantly predicted soil bulk density at a depth of 0–15 cm (F (1, 48) = 41.48, P < 0.0001, R2adj = 0.45) and at a depth of 15–30 cm (cubic transformation; F (1, 48) = 16.95, P < 0.0001, R2adj = 0.24). For soil samples collected from landscaping beds (Fig. 4d), lot age significantly predicted soil bulk density at a depth of 0–15 cm (F(1, 48) = 22.64, P < 0.0001, R2adj = 0.31) and at a depth of 15–30 cm (cubic transformation; F (1, 48) = 5.314, P = 0.03, R2adj = 0.08). Additionally, the regression models suggest that over 22 years, mean bulk density at a depth of 0–15 cm decreased by 0.54 and 0.35 g cm−3 (31 and 24% decrease) in lawns and landscape beds, respectively. Bulk density decreased to a lesser extent at a depth of 15–30 cm, with a mean decrease of 0.33 and 0.19 g cm−3 (15 and 9% decrease) for lawns and landscape beds, respectively.

3.2.3 LR-CA

In LR common areas, bulk densities ranged from 1.20 to 1.94 g cm−3 at depths of 0–6 cm and from 1.31 to 2.37 g cm−3 at depths of 6–30 cm. Decreased bulk density was associated with an increase in development age (Fig. 5a) at a depth of 0–6 cm (F (1, 35) = 53.78, P = 1.4*10–8, R2adj = 0.59) and a depth of 6–30 cm (reciprocal square transformation; F (1, 25) = 25.71, P = 3.1*10–5, R2adj = 0.49). Mean bulk density decreased from 1.67 to 1.30 g cm−3 (22% decrease) between 0- and 20-year-old common areas.

Fig. 5
figure 5

Bulk density (a) and soil organic matter concentration (b) in relation to development age for soil samples collected at depths of 0–6 and 6–30 cm for LR-CA. The shaded area represents a 95% confidence interval for the regression. ns = nonsignificant; ***P < 0.001

Full size image

3.3 Soil organic matter

3.3.1 OTOW

For lawn-collected soil samples (Fig. 6a) at a depth of 0–15 cm, lot age significantly predicted a reciprocal cube root transformation of SOM (F(1, 43) = 88, P = 6.3*10–12, R2adj = 0.66), but the relationship between lot age and SOM was not significant at a depth of 15–30 cm (P = 0.995). For samples collected from landscape beds (Fig. 6c), lot age significantly predicted a reciprocal square root transformation of SOM at a depth 0–15 cm (F(1, 43) = 33, P = 7.5*10–7, R2adj = 0.42), but similarly to lawn-collected samples, the relationship between lot age and SOM was not significant at a depth of 15–30 cm (P = 0.524). These results also suggest that in the upper 15 cm of the soil profile, SOM increased by approximately 5.7% and 2.5% (384% and 194% increase) in lawns and beds, respectively, between 0 and 30 years after lot development. By comparison, samples collected from 0 to 15 cm (n = 80) depth of a mature pine forest adjacent to OTOW, and analyzed using the same methods, had an average SOM of 1.23% and a standard deviation of 0.26%.

Fig. 6
figure 6

Soil organic matter concentration in relation to residential lot age for soil samples collected at depths of 0–15 and 15–30 cm for OTOW (a and c) and LR-R (b and d). Lawns (a and b) and landscape beds (c and d) are represented. The shaded area represents a 95% confidence interval for the regression. ns = nonsignificant; *P < 0.05; **P < 0.01; ***P < 0.001

Full size image

3.3.2 LR-R

Overall, increased lot age was associated with a significant (α = 0.05) increase in SOM for all sampling depths and locations, with weak associations for samples collected at a depth of 15–30 cm. For lawn-collected soil samples at a depth of 0–15 cm (Fig. 6b), lot age significantly predicted SOM (log transformation; F (1, 48) = 57.08, P < 0.0001, R2adj = 0.53), and an inverse cube root transformation of SOM at a depth of 15–30 cm (F(1, 48) = 7.602, p = 0.008, R2adj = 0.12). For soil samples collected from landscape beds (Fig. 6d), lot age significantly predicted a log transformation of SOM at a depth 0–15 cm (F (1, 48) = 50.8, p < 0.0001, R2adj = 0.50) and 15–30 cm (F (1, 48) = 4.27, p = 0.04, R2adj = 0.06). These results also suggest that in the upper 15 cm of the soil profile, SOM increased by approximately 2.7% and 3.2% (163% and 158% increase) in lawns and beds, respectively, between 0 and 22 years after lot development. In contrast, an increase in SOM of only 0.53 and 0.44% (53% and 36% change) was respectively observed for lawns and beds at a depth of 15–30 cm. By comparison, undisturbed and non-agricultural A horizon soils that were moderately-well to excessively-well drained (n = 222) had on average organic carbon content of 1.41% or less [22].

3.3.3 LR-CA

An increase in SOM was significantly associated with increased development age at a depth of 0–6 cm (F (1, 27) = 75.33, P = 2.5*10–11, R2adj = 0.61), but not at a depth of 6–30 cm (P = 0.86; Fig. 5b). Mean SOM increased by 3.4% (334% increase) between 0- and 20-year-old common areas.

3.4 Soil carbon accumulation rates

For OTOW the estimated carbon accumulation at a depth of 0–15 cm was 0.156 and 0.064 kg C m−2 yr−1 for lawns and landscape beds, respectively (Table 3). Changes in SOM were not significantly associated with increased lot age at a depth of 15–30 cm for OTOW, and thus a carbon accumulation rate was not estimated. For LR-R, an estimation of carbon sequestration at a depth of 0–15 cm yielded 0.084 and 0.100 kg C m−2 yr−1, respectively, for lawns and landscape beds. Compared to surficial soil, carbon sequestration rates at a depth of 15–30 cm were four to five times less. For LR-CA, the change in mean SOM between 0- and 20-year-old common areas suggested a carbon accumulation rate of 0.060 kg C m−2 yr−1, lower than other surface soils (OTOW & LR-R, 0–15 cm), but greater than LR-R rates at 15–30 cm.

Table 3 Soil carbon accumulation rates in lawns and landscape beds at depths of 0–15 and 15–30 cm. Changes in SOM were not significantly associated with lot age at a depth of 15–30 cm for OTOW, so a carbon accumulation rate was not estimated
Full size table

4 Discussion

4.1 Bulk density

Bulk densities over 1.80 g cm−3 for sandy soils restrict root growth (USDA-NRCS, 2008). While bulk densities from deeper soils significantly decreased over time, they generally remained above or near 1.70 g cm−3. We found a significant negative association between bulk density and lot age across all sampling depths, sampling sites, and sampling locations. In LR-R, bulk density decreased most rapidly with increasing lot age (0.024 g cm−3 yr−1) in lawns at a depth of 0–15 cm, and the least change in bulk density was observed in landscape beds (0.008 g cm−3 yr−1) a depth of 15–30 cm. In contrast, the relationship between bulk density and lot age was more consistent across sampling sites and depths for OTOW (0.013–0.014 g cm−3 yr−1). In LR-CA, changes in bulk density were similar between sampling depths, with mean decreases of 0.017 and 0.021 g cm−3 yr−1 at depths of 0–6 and 6–30 cm, respectively.

Past chronosequence studies have shown mixed results when analyzing the relationship between lot age and soil bulk density. Scharenbroch et al. [43] found that in the Northwestern US, a sample of 48 urban and residential sites with a silt loam soil texture showed significantly reduced bulk density (0.006 g cm−3 yr−1, 18% difference) between 9- and 64-year-old sites. However, a 55-year chronosequence study by Raciti et al. [17] of 32 residential sites in the Baltimore, Maryland, area found that lot age was not a significant predictor of bulk density, though residential bulk densities were not significantly different from forested soils.

The range of bulk densities observed in this study suggests a linear relationship to time. However, linear extrapolation is unsustainable; it is more likely that soils would reach an equilibrium density, a steady structural state. As vegetation is the main driver, though not the only contributor, one equilibrium point would be when porosity cannot be increased further without decreasing water or nutrient availability.

It is important to account for the initial condition or compaction level, in these types of studies when substituting space for time. By including landscapes from residential developments within the same communities, construction and landscape management methods are as uniform as possible over the lifetimes of these landscapes, therefore supporting the assumption that soil development processes are similar from lot to lot.

4.2 Soil organic matter

Amounts of SOM were similar for landscape beds and turfgrass for LR-R and OTOW, with higher accumulation at 0–15 cm than 15–30 cm. Significant increases in SOM were most prominent in surficial soils within depths of 0–15 and 0–6 cm across sampling sites. At a depth of 15–30 cm, a significant positive relationship between SOM and lot age was only observed in LR-R, and between 0- and 22-year-old lots, deeper SOM increased by a mean of 44%, compared to 160% increase for surficial soil depths of 0–15 cm. In LR-CA between 0- and 20-year-old, SOM increased by 336% within the surficial soil depth of 0–6 cm, which was the highest increase in SOM among study sites. There was very limited accumulation of SOM within LR-CA soils from the 15–30 cm depths. Managed urban turfgrass can accumulate higher levels of carbon compared to native soils [44].

Although SOM increased significantly (p < 0.05) over time within 15–30 cm depths under lawns and beds of LR-R, accumulation rates were much slower (0.02 kg C m−2 yr−1) than shallower soils. The accumulation rates within 15–30 cm depths were not significant for OTOW, nor for the 6–30 cm depth for LR-CA. Even after 20–30 years, SOM values are similar to that of new fill material. However, within the top 15 cm, SOM increased significantly (p < 0.001) under each vegetation coverage for both developments, with accumulation rates being greater in beds than lawns for LR-R and in lawns than beds for OTOW. Pooled mean for lawns and beds in LR-R was 0.092 kg C m−2 yr−1 and 0.110 kg C m−2 yr−1 for OTOW. This is likely due to soil being more compact at depth (15–30 cm) than near the surface (0–15 cm).

4.3 Soil carbon accumulation

Past chronosequence studies of carbon accumulation in urban soils of residential and recreational sites (i.e., golf courses) have found C accumulation rates between 0.026 and 0.355 kg C m−2 yr−1 at depths of up to 100 cm [16,17,18,19]. Huyler et al. [19] reported a relatively low carbon sequestration rate of 0.026 kg C m−2 yr−1 at a depth of 0–15 cm for pure lawns in Auburn, Alabama. The rates of carbon accumulation at 0–15 cm depth found in this study (0.064 to 0.156 kg C m−2 yr−1) are more akin to results of a study by Qian and Follett [16], which found C accumulation rates of between 0.082 and 0.091 kg C m−2 yr−1 for soil samples collected at a depth of 0–11.4 cm in Midwestern US golf courses. Carbon accumulation in LR-CA turfgrass areas was less than that of LR-R and OTOW turfgrass sites, perhaps because of the differing turfgrass management strategies. It was surprising to find a relatively high C accumulation rate in our study given the preconception that coarse soil textures and warmer climates tend to relatively lessen C accumulation [32]. Managed turfgrass can accumulate higher level of carbon compared to native soils [44]. For managed landscapes, this is likely more dependent on nitrogen inputs to landscapes from fertilization than from irrigation. Zirkle et al.[45] estimated that humification of plant biomass accounted for 0.046–0.127 kg C m−2 yr−1, fertilization 0.078–0.098 kg C m−2 yr−1, and irrigation 0.005–0.010 kg C m−2 yr−1, for a total of 0.129–0.235 kg C m−2 yr−1 soil organic C accumulation. Thus, it is likely that fertilization and humification are the drivers of C accumulation in urban lawns, with mulch being the main source in landscape beds and mulched grass clippings in turfgrass areas.

The lack of SOC accumulation at depth (15–30 cm) seems counter to bulk densities decreasing over time and it does not reflect the increasing trend observed in 0–15 cm soil layers. The root zone for turfgrasses commonly grown in Florida is about 20 cm, with the majority of root mass within the top 15 cm. Horizontal distribution of root mass under lawns is more homogenous than within ornamental plant beds. As bulk densities increase with depth, it becomes more difficult for roots to penetrate deeper and deeper into the soil profile. Therefore, the tradeoff is between greater access to water and possibly nutrients vs. overcoming the soil structure.

Standardization of home building and land development practices suggests that these characteristics are representative of broader patterns in residential land use. While this study was conducted in Florida, urban landscapes are converging [1]. As land development in the United States moves towards planned communities, the approach is similar: to clear and grade the development site. The inferior quality of fill material and compaction creates a less than optimal growing condition for landscape vegetation. The underlying soil condition at construction sets the trajectory of the underlying soils for the future. However, turfgrass is the most common vegetation in many urban settings because it is the lowest cost option per unit area to purchase and install.

In addition, the annual cycle of root penetration and subsequent dieback is an important contributor to SOC accumulation. However, while bulk densities at 15–30 cm decreased over more than 20 years, the soil at this depth is still restrictive to turfgrass rooting. Further, frequent and shallow irrigation would promote shallower rooting depth, which may have also contributed to the lack of deeper SOC accumulation.

This is the first study to compare carbon accumulation rates under lawns to those under landscape beds. No distinct differences between lawns and landscape beds were found for bulk densities and soil organic matter. While soils underneath turfgrass accumulated soil carbon, these rates were similar to those within landscape beds. This suggests that equivalent or similar processes are occurring at similar rates within both parts of the landscape. From a management perspective, current irrigation and fertilization recommendations are lower for landscape beds than lawns in Florida. In both LR-R and OTOW, supplemental irrigation and fertilization are standard practices. As such, it can be concluded that landscape beds were likely more efficient for increasing SOC due to reduced inputs. This could have implications for efforts focused on increasing carbon sequestration and crediting.

To better understand the evolution of urban soil characteristics, future research should investigate whether the same processes are occurring within both lawns and beds, or whether similar but different processes are responsible for similar changes.

5 Conclusions

This study sought to assess soil quality change over time within the top 30 cm of two residential developments in Florida. The primary research questions were whether temporal trends in soil properties could be identified and whether there were differences between developments, soil depth, or locations.

Results of this study provides evidence for a positive relationship between soil quality and lot age in Central (OTOW) and Southwest Florida (LR) residential lawns and landscape beds with sandy soil textures, providing a framework for characterizing the evolution of residential soil characteristics in Florida and similar climatic and soil settings. Increased soil organic matter (0.060–0.156 kg C m−2 yr−1) and decreased bulk density (22–34%) were associated with increased lot age at a depth of 0–15 cm, and 0–6 cm for common areas. No broad differences were observed between the lawns and beds. While bulk densities at 15–30 cm decreased similarly to shallower soils, SOM accumulation was only significant at the LR sites, although the magnitude was negligible by comparison. This decoupling between SOM accumulation and decreasing bulk densities at depth is likely due to remnant compaction limiting processes that facilitate SOM accumulation, such as root penetration and cycling, and invertebrate activity. Estimated carbon sequestration rates were representative of past chronosequence studies that have largely been conducted in cooler climates with finer soil texture. This suggests that sandy soil texture and warm climate in Florida do not govern carbon sequestration rates as anticipated, but those conditions are overcome by vegetation type and landscape management practices (e.g., fertilization, irrigation, mulching, and mowing). However, the age of lots in this study did not exceed 31 year and it is unknown whether the trends from this study continue beyond this, particularly as SOM accumulation rates cannot increase indefinitely.

This study in combination with previous studies provides evidence that these processes likely occur over a much broader scale than the localized survey areas in this and previous studies. Future research should seek to better understand the spatiotemporal evolution of urban soils, driving factors, and the ecohydrological impacts. The costs and benefits of interventions, such as amending soil with compost via tillage, should be quantified to better inform policies pertaining to urban landscapes. While chronosequence studies have characterized temporal trends in urban soils, future research should focus on identifying distinct factors that govern urban soil development and linking them to temporal trends.

This study is also the first known to examine soil development characteristics under both lawns and landscape beds. While turfgrass is recognized to carbon accumulation, this study provides evidence that landscape beds provide similar carbon sequestration, while typically requiring fewer inputs and maintenance. Better data and understanding of urban carbon cycles and evolution are needed to further understand the ecological states and trajectories of urban landscapes.