Abstract
Soil quality assessments are needed to improve a professional’s ability to manage urban soils and trees. This research was conducted to identify which soil properties are most useful for relating information on urban tree performance. In total, 48 soil properties were measured at 84 sites from five urban landscapes in the western suburbs of Chicago, Illinois, U.S. Key physical, chemical, and biological properties to be included in a minimum data set (MDS) for assessing urban soil quality were identified using statistical approaches and practical considerations. The MDS included: texture, bulk density, wet-aggregate stability, pH, electrical conductivity, soil organic matter (SOM), and particulate organic matter. The MDS was used to establish an urban soil quality index (USQI). The MDS and USQI were highly correlated with tree size attributes of height, trunk diameter, crown area, and age. Correlations between the MDS and USQI with trunk diameter growth rate, height growth rate, foliar N, and chlorophyll content were often significant, but less strong. Among the MDS parameters, SOM, pH, and texture appear to be the most informative measures for soil quality relating to urban tree performance. Soil quality and tree performance increased logarithmically following site disturbance, with a plateau after 50 years.
Soil quality is defined as the capacity of soil to function (Karlen et al. 1997). Some important soil functions include: water and solute flow and retention, physical stability and support, retention and cycling of nutrients, buffering and filtering of toxic materials, and maintenance of biodiversity and habitat (Larson and Pierce 1994; Doran et al. 1996). The term soil quality also refers to the effects of human use and management on these soil functions (Doran and Jones 1996; Seybold et al. 1999).
As a consequence of anthropogenic influences and management practices, the quality of urban soils is commonly impaired. Urban soils often have high bulk densities and low porosities, poor soil structure, altered water status and redoximorphic features, elevated pH and salinity, environmental contaminants, reduced organic matter contents, and altered microbial populations (e.g., Short et al. 1986; Craul 1999; Scharenbroch et al. 2005; Scheyer and Hipple 2005; Pouyat et al. 2007). Degraded soil conditions constrain urban tree growth and health (Craul 1992; Watson and Neely 1994; Neely and Watson 1998; Watson et al. 2008). Assessment and improvement of urban soil quality is imperative for the establishment, growth, and longevity of urban trees.
Urban tree performance (e.g., establishment, growth, longevity) is influenced by interactions of edaphic, genetic, climatic, and anthropogenic factors (Kozlowski 1971a; Kozlowski 1971b; Harris et al. 1999). Foresters commonly express the quality of the site in terms of site index or the average height of dominant trees at age 50. Techniques for estimating site quality have emphasized analysis of soil properties because site quality within moderately broad geographic areas appear to be controlled more by soil characteristics than by climatic factors (Colie and Schumacher 1953). Soil quality indicators are commonly used to predict or assess forest stand productivity (Burger and Kelting 1999; Woolery et al. 2002). To date, no such approaches have been developed for urban trees and soils.
Indices of forest soil quality are most useful if they incorporate soil physical, chemical, and biological properties, and are sensitive to management-induced changes, easily measured, inexpensive, relevant across sites and over time, and adaptable for specific ecosystems (Schoenholtz et al. 2000). Soil physical properties, especially those with direct impact on soil moisture, have been shown to have large impacts on tree growth (e.g., Zahner and Stage 1966; Schoenholtz et al. 2000; Woolery et al. 2002; Galvez et al. 2004; de Castilho et al. 2006). The importance of soil chemical properties to tree growth is represented by the voluminous literature showing the stimulating effects of fertilizers on growth of trees (e.g., Himelick et al. 1965; Schoenholtz et al. 2000; Udawatta and Henderson 2003; Galvez et al. 2004; Hamel et al. 2004). Soil biological factors influence both soil tilth and fertility (Knoepp et al. 2000), and need be considered for their impacts on urban tree growth.
Minimum data sets (MDS) of soil parameters and methods to use for assessing soil quality have been identified (e.g., Arshad and Coen 1992; Doran and Parkin 1994). These indicators should: correlate well with ecosystem processes, integrate soil properties and processes, be accessible to many users, be sensitive to management and climate, and when possible, be components of existing databases (Doran and Parkin 1994). Minimum data sets should include soil physical, chemical, and biological properties (Gregorich et al. 1994). Relatively easy to perform and cheap field diagnostic techniques are preferred to expensive laboratory measurements (Halvorson et al. 1996).
Assessment of soil quality requires evaluation of the current state of an indicator in comparison with known or desired values (Karlen et al. 1997; Burger and Kelting 1999). However, established norms for urban soil properties that accurately reflect a soil’s inherent productive or environmental filtering potential do not exist. Currently, it is very difficult to evaluate urban soil quality as related to soil function criteria. Multivariate statistical approaches such as principal component analyses (PCA) provide non-subjective means to extract and weight information in complex univariate data sets and are appropriate first steps towards soil quality assessments (Halvorson et al. 1995; Wander and Bollero 1999).
In this study, physical, chemical, and biological properties contributing to urban soil quality were evaluated in relation to tree performance. The specific objectives were to 1) sample 84 plots in five locations in the western suburban area of Chicago, Illinois, U.S.; 2) characterize soils (17 physical, 17 chemical, and 14 biological responses) and trees (9 performance responses); 3) establish MDS and an urban soil quality index (USQI) for assessing urban soil quality; 4) test the MDS and USQI for predicting urban tree performance; and 5) discuss the mechanism(s) driving soil quality and tree performance in these urban landscapes.
MATERIALS AND METHODS
Site Description
Sampling occurred at four residential homeowner’s associations (sites): Arboretum Estates in Glen Ellyn, IL; Baker Hill in Glen Ellyn, IL; River Oaks in Warrenville, IL; and Stonebridge in Aurora, IL; and at The Morton Arboretum in Lisle, IL; all located approximately 20 km west of Chicago, IL (Appendix 1). The four residential homeowner associations were selected from a larger group based on similarities in human impacts, soil management, topography, parent material, and age. Surveys of all trees in roadside planting strips were performed at each of the four homeowner’s associations. Street trees were located in the space between the sidewalk and the road. In a few plots without sidewalks or with sidewalks up to road, street trees were in the front yard within 10 m of the street. On each site, approximately 25% of the street tree population was sampled. The major species representing each site were identified and at least nine and up to 26 trees per species were then randomly selected from that population for sampling. The Morton Arboretum was included as a fifth site to provide information on older, less-disturbed urban landscapes of similar parent material, topography, and management. Twelve arboretum trees were randomly selected from the collections list to match Fraxinus and Quercus genera found in the homeowner’s associations.
Soils at the residential locations are classified as urban land, built up areas and deep, gently rolling to nearly level, moderately to poorly drained soils that have clayey subsoil and formed in glacial till (Mapes 1979). Typical native soils in these areas and at the arboretum include forest (e.g., Markham series) to prairie (e.g., Ashkum series) soils. The Markham series are fine, illitic, mesic Mollic Oxyaquic Hapludalfs, consisting of very deep, moderately well drained soils on Wisconsin, U.S., till plains formed in thin layers of loess or silty material and in the underlying silty clay loam till. Ashkum series are fine, mixed, superactive, mesic Typic Endoaquolls, consisting of very deep, poorly drained soils on till plains formed in colluvial sediments and in the underlying silty clay loam till. Mean annual precipitation is about 890 mm, and mean annual air temperature is about 10°C (Mapes 1979).
Soil management and human impacts at the residential sites were attained from interviews with managers from each homeowner’s association. All sites were developed in a similar fashion, which is typical of the region. At site development, topsoil (c.a., 0 to 25 cm) was removed, and the subsoil was graded and compacted to standard densities to support the infrastructure (c.a., 1.5 to 1.7 Mg m−3). Nominal depths (3 to 6 cm) of topsoil were replaced on the sites after construction. The ages of urban landscapes were inferred from manager interviews and tree cores from the sites. Site disturbance at the arboretum was relatively minimal and trees were planted in soils that were not truncated, buried, graded, or compacted. Tree planting and care on all sites was performed by qualified arborists. Tree fertilization and irrigation was performed according to standards and best management practices during the tree establishment period (two to three years). Tree fertilization was not performed on any of the trees in the three years leading up to sampling. Granular NPK fertilizer was applied annually in the spring, following manufacturer’s label, to turf areas in the residential associations, but not in the arboretum. No trees received irrigation, herbicides, or pesticides in any of the sites, including the arboretum trees, over at least the past five years. Trees in all sites were pruned as needed to develop ideal form and remove dead wood, typically at three- to five-year pruning cycles. Tree ages ranged from 5- to 86-years-old, trunk diameters from 7 to 80 cm, and heights from 4 to 27 m (Appendix 1).
At each individual tree plot the following was measured and recorded: address, GPS coordinates, slope, aspect, distance to nearest building, distance to nearest hardspace, estimation of plantable space, and the percent of that space covered by cement, tar, rock, mulch, grass, and herbaceous/shrub plants (Appendix 1). In estimating plantable space, it was assumed that tree roots were able to grow beneath sidewalks, but researchers did not extend this beyond buildings, driveways, and roads. Maximum plantable space was set at 3,848 m2 (35 m radius), which is approximately 10 times the maximum canopy surface area.
Soil Assessment
The eighty-four plots were sampled between July and August in 2010. All soil sampling was performed at random locations under the tree drip line. Many of the field assessments of soil quality were adapted from Doran’s soil test kit procedure (Sarrantonio et al. 1996) and the field book for describing and sampling soils (Schoeneberger et al. 2002). Given the uncertainty and heterogeneity of urban soils, the study includes a wide and exhaustive range of soil properties from which a small subset is selected for relating tree performance. Certain soil properties are known to be highly variable (e.g., moisture and temperature) or influenced by other soil properties (e.g., penetration resistance); however, these properties were included in the initial characterization given reoccurrence in the literature to assess to soil quality, ease of measurement, and potential ability to correlate with more time-consuming or expensive soil assessments.
Soil physical observations were performed on a 25 cm × 25 cm × 20 cm deep excavation on each plot. Soil color (hue, value, and chroma) was determined using Munsell soil color charts on wet and air dried soils. Color determinations were made by three individuals per soil and means were calculated. Soil structure type (platy-massive = 0, angular blocky = 1, subangular blocky = 2, granular = 3), structure grade (structureless = 0, weak = 1, moderate = 2, strong = 3), and structure size (very fine = 0, fine = 1, medium = 2, coarse = 3) were described and scored. Numbers of fine (1 to 2 mm diameter) plant roots were estimated at ten, one cm2 points. Coarse (2 to 10 mm in diameter) roots were estimated at four faces, each 100 cm2. The areas of redoximorphic concentrations and depletions were estimated on the exposed profile face.
The soil from the 25 cm × 25 cm × 20 cm hole was sorted for earthworms. A hot mustard powder solution (50 g L−1) was poured in the excavated hole to extract deeper earthworms (Lawrence and Bowers 2002). The number of adult and juvenile earthworms was tallied for each excavation, and reported as individuals m−3. Earthworms were stored on ice in a cooler in petri dishes with damp towels and returned to the laboratory. Earthworms were identified using a dissecting microscope and dichotomous key for the Great Lakes, U.S., region (Hale 2010). For each plot, adult earthworm biomass (ash-free dry mass) was determined for each species (Hale et al. 2004). Lumbricus terrestris was the only species encountered in the earthworm sampling.
Three penetration resistance profiles (0 to 45 cm) were measured per plot using a cone penetrometer (FieldScout SC 900 Soil Penetrometer, Spectrum Technologies, Inc., Plainfield, IL, U.S.). Mean penetration resistance over the 0 to 20 cm depth was calculated. The root restriction depth was calculated by computing the mean penetration resistance profiles for each plot and then identifying the depth at which penetration resistance exceeded 2.3 MPa. According to Day and Bassuk (1994), the critical soil strength above which woody plant root elongation is restricted is in the vicinity of 2.3 MPa, depending on soil type and plant species.
Infiltration rate was measured at two locations per plot using a double-ring infiltrometer (Turf-Tec International, Tallahassee, Florida, U.S.). Sample locations were pre-saturated with 1 L of water. Infiltration rates were measured twice at each sample point and means computed. Volumetric water content of the 0 to 20 cm depth was measured at ten points per plot using a time-domain reflectrometry probe (Field-Scout TDR 300 Soil Moisture Meter, Spectrum Technologies, Inc., Plainfield, Illinois, U.S.). Soil temperature was measured by inserting a 10 cm thermocouple thermometer into the soil (Fluke 52 K/J Thermometer, Everett, Washington, U.S.).
Bulk density (ρb) was measured on undisturbed soil core samples (70 mm wide × 70 mm deep) collected from each plot. Only samples that completely filled the entire core volume were used. The core samples were kept shaded and on ice for transport to the lab. Soil was sieved, homogenized, and dried in an oven for 48 hours at 105°C. Material (roots, rock, etc.) greater than 2 mm was removed, and its volume and oven-dry weight determined for bulk density corrections for non-soil material.
Ten, 2.5 cm wide × 20 cm deep cores were taken from random plot locations for soil characterization in the laboratory. A uniform sampling depth of 20 cm was adopted since this depth is likely to include mainly the A horizon across all sample plots. The ten soil cores were composited per plot and kept shaded and on ice in a cooler for transport to the lab. In the laboratory, soil sub-samples were weighed, dried for 24 hours at 105°C, and reweighed to calculate gravimetric soil moisture (Topp et al. 2008). Sand, silt, and clay (%) were calculated using the modified pipette method of Kettler et al. (2001). Stability of aggregates (1 to 2 mm) was measured by oscillation of the sample through a height of 37 mm height, 29 times per minute for ten minutes in water (Angers et al. 2008). The oven-dry weight of water-stable aggregates (WAS) per total oven-dry soil was expressed as a percentage.
Soil pH and electrical conductivity (EC) in dS cm−1 were measured in 1:1 (soil:deionized) water pastes (Model Orion 5-Star, Thermo Fisher Scientific Inc., Waltham, Massachusetts, U.S.). Total C and N (%) were determined by automated dry combustion analyzer (Elementar Vario EL III CHNOS, Elementar, Hanau, Germany). Loss on ignition at 360°C for six hours was used to determine the soil organic matter (SOM) (Nelson and Sommers 1996). Soil sub-samples were extracted with 1 M NH4OAc (pH 7.0) and mg kg−1 of potassium (K), calcium (Ca), magnesium (Mg), and sodium (Na) were determined with atomic adsorption spectroscopy (Model A5000, Perkin Elmer Inc., Waltham, MA, U.S.) (Schollenberger and Simon 1945). The sum of these exchangeable bases was expressed as effective cation exchange capacity (eCEC) (Sumner and Miller 1996). Sodium adsorption ratio was computed as the milliequivalent weight of Na divided by the square root of the milliequivalent weight of Ca and Mg divided by two. Soil phosphorus (P) was determined with the Olsen extraction and extracts were analyzed colorimetrically at 882 nm on a spectrophotometer (Model UV mini 1240, Shimadzu Inc., Kyoto, Japan) (Olsen and Sommers 1982). Soils were extracted with 0.5 M K2SO4 and NH4+ measured using a modified indophenol blue method for microplate analyses at absorbance readings of 650 nm (Model ELx 800, Biotek Instruments Inc., Winooski, Vermont, U.S.) (Sims et al. 1995). With Devarda’s alloy, NO3− was reduced to NH4+, which was then quantified using Sims et al. (1995). Dissolved organic N was reduced to NH4+ with persulfate and Devarda’s alloy and also measured following Sims et al. (1995). Inorganic N was the sum of extracted NH4+and NO3−.
Particulate organic matter (POM) was measured after shaking 25 g subsamples for 15 hours with sodium hexametaphosphate (NaPO3)6 and then collecting litter organic matter on a 2000 µm sieve, coarse POM on a 250 µm sieve, and fine POM on a 53 µm sieve (Gregorich and Beare 2008). Loss on ignition at 360°C for six hours was used to determine the OM content of the litter SOM, fine POM, and coarse POM fractions (Nelson and Sommers 1996). The soil fumigation-extraction method (Brookes et al. 1985) was used to determine microbial biomass N in mg kg−1. Soil subsamples were fumigated with ethanol-free chloroform for five days and extracted with 0.5 M K2SO4. Microbial biomass N was the difference in dissolved organic N between the fumigated and unfumigated baseline samples, using an extraction efficiency factor ofkEN = 0.54 (Joergensen and Mueller 1996). Nitrogen mineralization and microbial respiration were measured using 20-day soil incubations in the dark, at 25°C and with soils adjusted to 60% water-filled-pore space. Carbon dioxide in 0.25 M NaOH traps was precipitated with BaCl2, followed by 0.25 M HCl (standardized) titration to a phenolphthalein endpoint (Parkin et al. 1996), expressed as soil respiration (mg CO2 kg−1 d−1). Concentrations of NH4+ and NO3− in incubated soils were determined colorimetrically as previously described (Sims et al. 1995). Nitrogen mineralization was determined by subtracting inorganic N (NH4+ and NO3−) in base extracts from the extracts of the incubated soils divided by the incubation period (mg NH4+/NO3− kg−1 d−1). Microbial biomass C (MBC) was calculated from microbial biomass N values, assuming 8/1 microbial C/N (Inubushi et al. 1991), and used to calculate two microbial efficiency indices. The metabolic quotient (qCO2) mg CO2 evolved h−1 g−1 / mg MBC g−1 and the ratio of MBC to soil C (MBC/TOC).
Tree Performance
Tree age (years) was determined by collecting increment cores (one per tree at 1 m from ground) with a tree increment borer. Increment cores were mounted on grooved wood blocks, sanded using progressively finer sandpaper to help distinguish rings, and analyzed using a Velmex stage micrometer and a Metronics Quick-Chek 1000 (Heidenhain, Schaumburg, Illinois, U.S.). The program Measure J2X v4.2 (VoorTech Consulting, Holderness, New Hampshire, U.S.) was used to record annual ring-widths to the nearest 0.001 mm. The mean diameter growth rate (mm yr−1) was computed as the mean annual increment growth throughout the tree’s lifespan. Tree height (m) was measured using a Suunto clinometer (Suunto, Ogden, Utah, U.S.). Tree height growth rate was tree height divided by tree age (m yr−1). Tree trunk diameter (cm) was measured at 1.38 m with a Lufkin diameter tape (Lufkin, Lufkin, Texas, U.S.). The short and long diameters of the crown were measured from the trunk to dripline, with crown area (CA) (m2) = crown width long * crown width short * π / 400 (Uzoh and Ritchie 1996).
Leaf chlorophyll content (leaf greenness) was measured on ten random leaves per tree using a SPAD-502 Plus Chlorophyll meter (Konica Minolta, Tokyo, Japan). After measurement the ten leaves were collected and composited per tree, stored on ice, ground with a mortar and pestle in the laboratory, and analyzed for total N (%) by automated dry combustion (Elementar Vario EL III CHNOS, Elementar, Hanau, Germany). A qualitative tree condition index (TCI) value was calculated based on Webster (1979). The TCI was a summation of trunk, crown, root, structure, growth, pest, and life expectancy factors (Table 1).
Trunk, crown, root, structure, growth, pest, and life expectancy factors comprising the urban tree condition index. Tree condition index is the summation of the seven scores. Adapted from Webster (1979).
Statistical Analyses
Statistical analyses were conducted using SAS JMP 7.0 software (SAS Institute Inc., Cary, North Carolina, U.S.). Data distributions were checked for normality using the Shapiro-Wilk W test. All soil responses aside from clay, WAS, ρb, water content, NH4+, dissolved organic N, Mg, and microbial biomass N, required data transformations or non-parametric tests. Variables were grouped into physical, chemical, and biological categories. Multivariate statistical analysis was conducted in two steps as suggested by Hatcher and Stepanski (1996). Multivariate analysis of variance (MANOVA) was used to detect significant location effects on at least one physical, chemical, or biological variable assessed. After meeting the criteria, analysis of variance (ANOVA) of individual parameters was run on all parameters. The obtained F statistic was used to test the null hypothesis of no location effect. Those variables for which the F statistic was significant (P ≤ 0.05) and variance was low (CV ≤ 60) (Hatcher and Stepanski 1996), were retained for further analyses. Treatment mean separations were interpreted using Tukey-Kramer’s HSD test to protect for the overall error rate.
Other studies have found principal component analyses (PCA) to be practical and effective tools in selection of appropriate soil quality indicators for predicting plant performance (Maddonni et al. 1999; Brejda et al. 2000a; Brejda et al. 2000b; Shukla et al. 2006; Rodrigues de Lima et al. 2008; Bautista-Cruz et al. 2011). Retained parameters from the ANOVA were used in PCA for further screening. Eigenvalues are the amount of variance explained by each factor. Factors with eigenvalues greater than one were retained for interpretation, because factors with eigenvalues less than one explained less variance than individual soil attributes (Kaiser 1960). The retained factors were subjected to varimax rotation, which redistributes the variance of significant factors to maximize the relationship between interdependent soil variables. All meaningful loadings (i.e., >0.40) were included in the interpretation of the PCA. Principal components that explained more than 5% of the total variance were considered significant. PCA was also used to identify single values for an USQI and also to synthesize tree attributes. Multivariate regression was used to identify relationships in the data sets. Step-wise regression modeling with mixed direction and probability to enter or leave at P ≤ 0.05 was used to develop predictive models among soil and tree properties.
RESULTS AND DISCUSSION
Establishing the MDS for Assessing Urban Soil Quality
Using ANOVA, PCA, and regression analyses the data were screened to identify soil parameters to include in the MDS for assessing urban soil quality. The ANOVA revealed the general location effects on individual soil parameters. The PCA identified which variables differed most and which were most informative and unique. Regression analyses detected redundancy among the soil properties, and were used to select the practical and informative measures in the final MDS.
The results from the ANOVA analyses are summarized in Table 2. Location effects were evident for many factors. Sixteen parameters were excluded from further consideration because they failed to meet the screening criteria (P < 0.05 for location effect and CV < 60%). Thirty-two parameters met the criteria and were retained for further screening.
Analysis of variance (ANOVA)zy for physical, chemical, and biological soil propertiesx in 84 plots in western suburban Chicago, IL.
The relative significance of the data set parameters was assessed using PCA of the 32 retained variables from the ANOVA (Table 3). There were five significant principal components (PC) that explained 65% of the variance. The first PC explained 26% of the total variance, contrasted ρb, and was positively related to measures of SOM. The seven parameters with significant positive loading on PC1were: SOM, total N, total C, fine POM, total POM, K, and WAS (Table 3). Bulk density was a significant negative loading on PC1. Higher PC1 scores appear to relate increases in soil quality. Principal component 2 explained 16% of the total variance and included five positive significant variables: Ca, pH, eCEC, EC, and C/N (Table 3). Higher PC2 scores show relative decreases in soil quality. Significant PC3, PC4, and PC5 loadings explained 10%, 7%, and 6% of the total variance (Table 2). Higher PC3, PC4, and PC5 scores were interpreted as increases in soil quality. The PC3 scores were positively loaded with root restriction depth, and negatively loaded with water content, penetration resistance, and chroma. The PC4 scores were positively loaded with microbial biomass N, MBC/TOC, and silt and negatively loaded with sand. Microbial respiration was a positive, and the qCO2 was a negative loading on PC5.
Principal component scores based on 32 variableszyx. Data from 84 plots in western suburban Chicago, IL.
The 24 soil properties (listed in order of importance) passing the ANOVA and PCA screening included: SOM, N, C, fine POM, total POM, K, ρb, WAS, Ca, pH, eCEC, EC, C/N, gravimetric soil moisture, volumetric water content, penetration resistance, root restriction depth, microbial biomass N, MBC/TOC, silt, sand, clay, respiration, and qCO2. Further screening with regression analyses were used to identify redundancy among the 24 remaining soil parameters (Table 4). In the following section, justification is provided for reducing these 24 parameters to the nine MDS parameters of: sand, silt, clay, ρb, WAS, pH, EC, SOM, and POM.
Multivariate regression and r-values with 24 soil parametersz passing ANOVA and PCA screening for urban soil MDS. Data from 84 plots in western suburban Chicago, IL.
Measures of organic matter (SOM, C, N, and C/N) were heavily weighted in the PCA. These responses had relatively low CV values and high R2 values for the ANOVA site differences. Higher SOM, C, and N contents indicate increases in soil quality (Doran and Parkin 1994; Knoepp et al. 2000). Increased C/N ratios indicate lower decomposition rates and relatively lower N mineralization potentials (Bengtsson et al. 2003). Loss on ignition is the least costly of these analyses and was significantly correlated with the other responses (C = 1.57 + 0.0329 * SOM, R2 = 0.43, P < 0.0001), (N = −0.0458 + 0.00479 * SOM, R2 = 0.86, P < 0.0001), (C/N = 24.0 – 0.147 * SOM + 24.0, R2 = 0.42, P < 0.0001), and color (SOM = 114 – 14.1 * dry value, R2 = 0.24, P < 0.0001) (Table 4). Overestimation errors may occur with loss on ignition for soils with high clay contents and carbonate materials (Nelson and Sommers 1996). Soil C, N, or C/N ratio are more costly, but often preferred to loss on ignition due to greater accuracy. Measurements of C, N, C/N ratio, and SOM relate similar information, and any one of these measurements may be suitable. Loss on ignition was chosen in the MDS due to its lower cost of analysis.
Particulate organic matter appears to be a sensitive indicator for assessing urban soil quality. Particulate organic matter is positively related to nutrient supply and soil physical condition (Wander et al. 1994; Six et al. 2000; Scharenbroch and Lloyd 2006). Particulate organic matter was identified as the primary indicator of soil quality for assessing the impact of tillage in Illinois, U.S. (Wander and Bollero 1999). The ANOVA showed strong location effects for POM, and POM was a highly weighted variable in the PCA. In this data set, POM was significantly correlated with 27 of 48 total parameters (7 of 17 physical responses, 15 of 17 chemical responses, and 5 of 14 biological responses) (data not shown). Indices of microbial respiration, microbial biomass, and N mineralization are good estimates of potential nutrient availability, gross microbial functioning, and soil quality (Knoepp et al. 2000). Particulate organic matter was significantly correlated with these measures (microbial biomass N = 87.4 + 14.1 * POM, R2 = 0.26, P < 0.0001), (soil respiration = 53.0 + 6.85 * POM, R2 = 0.05, P = 0.0328), and (N mineralization = 1.18 + 0.184 * POM, R2 = 0.13, P = 0.0010) (Table 4). The POM assessment requires substantially less time, money, and expertise to measure compared to those microbial assessments. These findings suggest POM as a necessary inclusion in a MDS to assess urban soil quality.
Soil pH influences many soil properties and is often included in assessments of soil quality (Schoenholtz et al. 2000). Higher soil pH values (>8.0) are associated with decreases in soil quality (Gale et al. 1991). Acidity is also known to inhibit biological activity, so the relationship with soil reaction and tree growth is likely not linear. Soil pH was heavily weighted in the PCA. Location effects were largely significant and variation was low for pH. Soil pH is relatively easy to measure and cost of analysis is cheap. For all of these reasons soil pH should be included in a MDS to assess urban soil quality.
Increased Na and EC indicate greater salinity and are interpreted as deleterious to soil quality (Doran and Parkin 1994; Karlen and Stott 1994). Increases in exchangeable bases can indicate greater nutrient availability and retention, but are also associated with higher soil pH values (Doran and Parkin 1994). The eCEC is a value that corresponds to total CEC, but only represents the base contribution (i.e., sum of exchangeable bases). Measures of soil salinity and exchangeable bases were heavily weighted in the PCA and site differences were apparent with the ANOVA. Significant relationships were detected for EC and these other measures (Na = −59.4 + 1.10 * EC, R2 = 0.40, P < 0.0001), (Ca = 802 + 2.43 * EC, R2 = 0.33, P < 0.0001), (K = 206 – 0.332 * EC, R2 = 0.09, P = 0.0065), (sodium adsorption ratio = −0.349 + 0.00631 * EC, R2 = 0.40, P < 0.0001), and (eCEC = 10.3 + 0.0153 * EC, R2 = 0.19, P < 0.0001) (Table 4). Conductivity measurements are relatively easy and have lower costs compared to measurements of exchangeable bases. Consequently, EC is the preferred inclusion in the MDS to assess urban soil quality. Exchangeable Na, Ca, Mg, and K can also be included in the MDS, but these parameters are secondary inclusions given their potential redundancy, greater costs, and need for laboratory instrumentation.
Soil texture is often included in other MDS for assessing soil quality (Doran and Parkin 1994). Loam-textured soils are preferable for plants compared to soils with higher proportions of clay or sand (Larson and Pierce 1994). Clay contents are relatively high in these urban soils, so greater amounts of silt and sand indicate higher soil quality. Location effects were significant and variation was low for percentages of sand, silt, and clay. Percentages of sand and silt were heavily weighted in PCA. For these reasons, soil texture is included in the MDS for assessing urban soil quality.
Soil compaction is a major problem in urban soils (Gregory et al. 2006). Soil quality decreases with increasing ρb and penetration resistance (Doran and Parkin 1994; Larson and Pierce 1994). Soil ρb and penetration resistance were highly loaded in the PCA, and the ANOVA showed strong site effects for these measurements. Comparisons of penetration resistance across sites and time-frames are susceptible to interferences associated with soil texture and moisture content. Soil ρb measurements involve field collection and minimal laboratory work (Larson and Pierce 1994) and do not have inherent spatial or temporal bias. For these reasons, ρb is suggested as a primary inclusion in the MDS for assessing urban soil quality. Soil penetration resistance was significantly correlated with ρb (ρb = 1.05 + 0.0000789 * penetration resistance, R2 = 0.19, P = 0.0004) (Table 4). Soil penetration resistance may be included, but given its inaccuracies, it is only recommended as a secondary inclusion.
Wet-aggregate stability (WAS) increases with tilth and is suggested as a necessary parameter in a soil quality MDS (Arshad and Coen 1992). Aggregate stability was heavily weighted in the overall PCA, and strong location effects were detected for WAS in the ANOVA. Aggregate stability is a measure that integrates physical, chemical, and biological properties. Aggregate stability was correlated with 21 of 48 total parameters (5 of 17 physical responses, 10 of 17 chemical responses, and 6 of 14 biological responses) (data not shown). It is a low-cost measure that can be performed in the field with minimal equipment requirements, thus WAS is included in the MDS for urban soil quality.
Both gravimetric and volumetric soil moisture contents revealed significant location effects with the ANOVA, but soil moisture was weakly loaded in the PCA. Soil water changes rapidly, and repeated measurements are often needed to make inference on soil quality. Furthermore, information on both water content and tension is needed to provide accurate assessments of plant available water status. Soil water content appears too elusive and not practical enough to be included in the MDS.
Compared to other studies (Wander and Bollero 1999; Brejda et al. 2000a; Brejda et al. 2000b), researchers did not detect just one soil property with the greatest potential for relating soil quality. Similar to other studies, the MDS included physical, chemical, and biological soil properties (Shukla et al. 2006; Rodrigues et al. 2008). The MDS in this research included seven soil properties: three physical (texture, ρb, and WAS), two chemical (pH and EC), and two biological (SOM and POM).
Establishing an Urban Soil Quality Index
A PCA was performed with the MDS soil properties to create an urban soil quality index (USQI) (Table 5). The first principal component explained 44% of the MDS soil parameters and was mostly highly loaded by SOM and pH, followed by WAS, POM, ρb, and EC. The second and third components explained an additional 16% and 13% of the MDS variance and were loaded by soil texture (clay, silt, and sand). The first PC explained most variance, and these values were selected for use as the USQI scores.
Principal component scores from nine soil MDS parameterszyx. Data from 84 plots in western suburban Chicago, IL.
Evaluating the MDS Parameters and USQI for Predicting Urban Tree Performance
A PCA was performed on the nine tree response parameters to identify which variables were most important in explaining variance of the measured tree responses (Table 6). The first principal component explained 50% of the tree performance variation and was positively loaded by tree size parameters (age, trunk diameter, tree height, and crown area), and to a lesser degree, leaf N content. The second principal component explained an additional 20% and was related to tree growth parameters (trunk diameter growth rate and height growth rate). The principal components derived from tree responses (PC1 ≈ tree size) and (PC2 ≈ tree growth) were assessed in relation to the soil properties identified in the MDS and the USQI (Table 7).
Principal component scores based on nine tree response variables.zy Data from 84 trees in western suburban Chicago, IL.
Correlations (r-values) between the urban soil MDS parametersz, urban soil quality index, individual tree responses, and principal components derived from nine tree response variables. Data from 84 plots in western suburban Chicago, IL.
The majority of the individual soil MDS parameters were well correlated to the tree response parameters (Table 7). Significant correlations were detected for silt (7 of 9), clay (8 of 9), WAS (7 of 9), ρb (6 of 9), pH (9 of 9), EC (6 of 9), SOM (6 of 9), and POM (5 of 9) with the individual tree response parameters. Soil pH (5 of 9), EC (5 of 9), clay (4 of 9), SOM (4 of 9), POM (4 of 9), and silt (3 of 9) were highly correlated (r-values > 0.4 and P < 0.0001) with many of the tree response parameters. Soil pH, clay, EC, and SOM were well correlated with PC1 ≈ tree size variable and explained 54, 32, 32, and 29% of its variance, respectively. The second principal component relating to tree growth was only correlated with WAS. The USQI values (loaded by SOM and pH) were significantly correlated with all tree responses aside from the tree condition index. Step-wise regression produced significant models for all tree response parameters, including PC1 ≈ tree size and PC2 ≈ tree growth (Table 7). Soil parameters that appeared most often in the step-wise models were SOM, pH, and texture (clay and silt). Relationships between the tree responses and the USQI and also the step-wise models were tighter than for the individual MDS parameters, suggesting multiple parameters are better predictors of urban tree performance compared to any single soil measurement.
These analyses suggest that SOM, pH, and texture are capable and most useful in explaining urban tree performance attributes. To date, no available studies have assessed urban soil quality in relation to tree performance, so it is not possible to relate these findings to an existing knowledge base. Similar approaches to this study have examined soil quality in relation to land use or agricultural plant performance. Most of these studies report SOM or C (Brejda et al. 2000a; Brejda et al. 2000b; Shukla et al. 2006; Rodrigues de Lima et al. 2008; Bautista-Cruz et al. 2011) as primary indicators of soil quality. Some of these studies also report pH and texture (Shukla et al. 2006; Bautista-Cruz et al. 2011) along with other properties (e.g., available water, porosity, bulk density, aggregate stability, earthworms, and micronutrients) to be important secondary inclusions to indicate differences in soil quality. Despite differences in parent material, climate, organisms, relief, time, and human influences, the MDS derived for these urban soils appears to be similar to MDS from other systems.
Urban Tree Size, Soil Quality, and Landscape Age
Tree size increased with urban landscape age and also across the urban soil quality gradient (Figure 1). It is reasonable to expect tree size to increase with age and also improved soil quality with time; however, this study was unable to distinguish if either of these two mechanisms were more important. Urban soil quality is linked with urban landscape age (Figure 1). The authors propose two mechanisms for the increase in soil quality with time: 1) advances in construction equipment and compaction technology increasing the soil impact on more recently developed sites, and 2) biogeochemical processes increasing soil quality with time.
Relative tree size, urban soil quality index, and urban landscape age for 84 plots in western suburban Chicago, IL,U.S.
Major advances in earthmoving and soil compaction equipment in the last century include the standardization of the internal combustion engine, the sheep’s foot roller (c.a., 1920s), and the vibratory compactor (c.a., 1960s) (Harris 2006). It is likely that the progression in construction technology over the past century has influenced the extent and degree of urban site disturbance and impact on soil quality. However, the significant linear relationship with soil quality and age of site disturbance in the most recently disturbed sites (within the last 30 years) (USQI = −5.66 + 0.318 * age; R2 = 0.43, P < 0.0001) (data not shown) suggests that technological advances may not play the only role in explaining the observations.
Time is one of the five soil formation factors (Jenny 1945). There are many biogeochemical processes that may increase soil quality over time (Buol et al. 2003). The predominant examples relevant to urban soils include: littering (organic accumulation on soil surface), desalinization (removal of soluble salts), dealkalization (removal of sodium carbonate), lessivage (migration of mineral particles from A to B horizons), pedoturbation (biological or physical churning of soil materials), decomposition (breakdown of mineral and organic materials), synthesis (formation of new mineral and organic species), humification (transformation of raw organic materials to humus), mineralization (release of oxide solids through organic matter decomposition), and loosening (increase in void volumes through biological and physical processes or by leaching). A number of processes may also contribute to decreases in urban soil quality with time: salinization (accumulation of soluble salts), alkalization (accumulation of sodium carbonate), and hardening (decrease in void volume by collapse, compaction, and in-filling).
Others studies have found improvements in urban soil quality with landscape age. An urban soil study in the U.S. Pacific Northwest by Scharenbroch et al. (2005) found increased SOM contents, increased nutrient availability, increased biological activity, increased microbial efficiency, and decreased bulk density with urban landscape age. Smetak et al. (2007) found greater earthworm biomass and abundances in older urban soils compared to younger ones. Beyer et al. (1995) report increased SOM and microbial efficiencies with landscape age in urban soils in Kiel, Germany. Studies in non-urban systems confirm these studies showing soil recovery after site disturbance progresses via organic matter accumulations, increases in microbial activity and nutrient availability (e.g., Pastor et al. 1987; Zak et al. 1990; Diquelou et al. 1999).
Soil recovery following disturbance may be delayed initially, but there is a tendency for the soil to rapidly increase after 10 to 20 years, often in association with woody vegetation establishment and differentiation of the A horizon (Zak et al. 1990; Diquelou et al. 1999). Rates of A horizon formation are variable and range from 10 year cm−1 in an Oregon, U.S., Mollisol soil (Forcella 1978); 12 year cm−1 in an Iowa, U.S., Mollisol soil (Simonson 1959); to 38 year cm−1 in a Wisconsin, U.S., Alfisol soil (Nielson and Hole 1964). The results of the current study demonstrate linear improvements in urban soil quality in urban landscapes that are <30-years-old, with a potential progression towards a flattening of this curve near 100 years. Scharenbroch et al. (2005) supports this finding of an approximate urban soil recovery time of >50 to 100 years.
CONCLUSION
The urban soil quality MDS for predicting urban tree performance included texture, ρb, WAS, pH, EC, POM, and total SOM. The authors suspect the MDS developed from these urban soils in western Chicago, IL, will be applicable to other urban soils. The MDS includes physical, chemical, and biological properties. The MDS includes parameters commonly included in standard soil assessments and other MDS databases (e.g., SOM and pH). The MDS includes soil properties that are responsive to soil management practices and disturbances at local scales (e.g., POM, WAS, and ρb). The MDS includes relatively permanent soil properties that are used in soil classification at global scales (e.g., texture and EC). A model for assessing urban soil quality for trees must be accurate, conceptually simple, cheap, and easy to apply. The proposed MDS meets these criteria and provides a framework for further testing. Future research should apply this MDS and test its predictive ability for tree performance in other urban systems. In order to expand to other urban areas climate (mean annual precipitation and temperature, degree days, etc.), parent material and site factors (e.g., proximity to infrastructure and plantable space) may also need to be considered. However, adding complexity to the MDS to improve accuracy must be balanced with practical considerations of taking those additional measurements.
Acknowledgments
The authors gratefully acknowledge financial support from the Tree Research & Education Endowment Fund John Z. Duling grant and The Morton Arboretum Endowment. We thank The Morton Arboretum Soil Science laboratory volunteers: Mark McKinney, Marlene Hahn, Kathy Davidson, and other volunteers for their assistance. Special thanks to the residents and homeowners associations and (their managers): Arboretum Estates Townhomes in Glen Ellyn, IL, (Lisa Cherry); Baker Hill in Glen Ellyn, IL, (Peggy Drescher and Daniel Richardt); River Oaks in Warrenville, IL (Ruth Brackmann); and Stonebridge Country Club in Aurora, IL (Roy Logan). We thank Edith Makra and Beth Corrigan for their assistance in working with the homeowners’ associations.
- © 2012, International Society of Arboriculture. All rights reserved.
LITERATURE CITED
Jump to section
Related Articles
Cited By...
- No citing articles found.