Author for correspondence: Michael S. Peek Tel: +1 973 7202245 Fax: +1 973 7202338 Email: firstname.lastname@example.org
• Fine roots of an annual grass, a perennial grass and a perennial shrub were examined. Based on life histories and tissue composition, we expected the greatest root persistence for the shrub and shortest for the annual grass.
• Roots were observed with minirhizotrons over 2 yr for number, length and diameter changes. A Cox proportional hazard regression correlated root persistence with soil water, depth, diameter and date of production.
• In 2001, grass roots had similar persistence times, but shrub roots had the shortest. In 2002, the annual had the longest median root persistence, the perennial grass intermediate and the perennial shrub had the shortest. All species responded similarly to the magnitude of seasonal precipitation; root numbers increased with favorable soil moisture and disappeared with drying; fewer, thinner roots at greater soil depths were found in the drier year (2001). Root persistence increased with soil moisture, diameter and earlier appearance in the spring.
• Plasticity in root morphology and placement was influenced by water availability, yet persistence was surprisingly contrary to expectations.
Fine root production and turnover are important components of the carbon budget of plants. In arid and semi-arid ecosystems, root biomass and annual production can greatly exceed that of the shoot system (Caldwell, 1975, 1979; Caldwell & Richards, 1986; Dobrowolski et al., 1990). Since below-ground resources are usually low and variable in these environments, fine root dynamics have important implications for individual plant growth, competitive interactions, carbon turnover and nutrient cycling. Yet, we understand few of the factors affecting fine root production and turnover (Eissenstat et al., 2000; Anderson et al., 2003).
Some of the variability in root lifespans in the soil can also be associated with environmental factors. There are broad-scale patterns, such as the association between longer-lived roots and resource-poor environments with the notion that longer life is necessary in order to ensure that construction costs are met by the return of water and nutrients for the plant (Eissenstat & Yanai, 1997). Local conditions such as temperature, moisture and nutrients are linked with root length growth, root mortality (Hendrick & Pregitzer, 1993) and decomposition (Silver & Miya, 2001). Experimental manipulations that increased water and nitrogen indicated stimulation of fine root production (Pregitzer et al., 1993), but the effect on lifespan was quite variable (Pärtel & Wilson, 2001). It is likely, however, that a combination of several factors influences fine root dynamics (Anderson et al., 2003).
In semi-arid systems, a variety of life forms occur from short-lived annuals to perennial shrubs and trees, and these can often occur in nearly monospecific stands (Billings, 1949; Smith & Nowak, 1990). The perennial shrub Artemisia tridentata dominates significant portions of the Great Basin semi-arid steppe lands. In large areas, the shrubs have been removed and exotic perennial grasses, such as Agropyron desertorum have been sown for forage. Disturbance of these areas has also led to the widespread invasion by exotic annual grasses, such as Bromus tectorum (Smith & Nowak, 1990). The long-term persistence of these three life forms in extensive stands suggests that fine root characteristics of these three species mesh with the environmental conditions, yet may differ among them. Our primary objective was to compare the fine root distribution in the profile, diameter and residence times of these three contrasting life forms: annual grass, perennial grass, and shrub in nearly monotypic communities growing on the same soil type in west central Utah. Based on the differences in life form and above-ground growth patterns, we posited that: (1) Rooting depth would correspond with relative plant size (Artemisia > Agropyron > Bromus), but roots should be concentrated in upper soil layers for all three species since this is where nutrients are most abundant; (2) structural differences should exist between species such that diameters of roots are hypothesized to be greatest for the slow-growing shrub and least for the fast-growing annual grass (Artemisia > Agropyron > Bromus); (3) fine root residence times would reflect the lifespan of leaves and duration of above-ground physiological activity (Artemisia > Agropyron > Bromus); (4) since water is the most limiting resource in this environment, we hypothesize that soil moisture significantly influences root distribution and persistence; (5) finally, we examined a post hoc hypothesis that root distribution and persistence would change similarly among species to maximize resource capture in response to differences in precipitation. Large differences in winter/spring precipitation in 2001 compared with 2002 allowed us to examine this hypothesis.
Minirhizotrons are a useful tool to study root system dynamics because repeated measurements of individual root segments can be made nondestructively over multiple time intervals. Nevertheless, certain limitations exist with this technique. For example, functionality of roots often cannot be gleaned. However, it is certainly assumed that thinner-diameter fine roots primarily serve in nutrient and water uptake while thicker roots are primarily for anchorage and transport (Eissenstat et al., 2000). Root death sometimes can be related to visible color changes in roots (Comas et al., 2000). In the study of Comas et al. (2000), white roots were clearly alive and black roots were demonstrated to be dead, while roots that had disappeared indicated complete decay. However, roots exhibiting various shades of brown appeared to be in a continuum between reduced metabolic activity and death when tested with vital stains and respiration measurements (Comas et al., 2000). In arid systems, senescing or dead roots can persist for long periods of time since the dry soils are not conducive to rapid decomposition (Moretto & Distel, 2003). In our study, roots were never observed to turn black, rather just various shades of brown. Thus, a clear designation of root death was elusive until roots disappeared, which owing to slow decomposition, may have been long after death of the roots. Thus, we speak of persistence, rather than longevity. This designation does not change our life form and physiologically based hypotheses about root persistence because fast-growing species often invest fewer resources into structural tissue (Ryser & Lambers, 1995) and therefore these roots should both have the shortest life spans and also decompose most rapidly.
Materials and Methods
We examined roots in nearly monotypic stands of Artemisia tridentata ssp. vaseyana (Rybd.) Beetle (big sagebrush), Agropyron desertorum (Fish. Ex. Link) Schult. (crested wheatgrass) and Bromus tectorum L. (cheatgrass) in Rush Valley, Utah, USA (40°17′ N, 112°28′ W, elevation 1660 m). This Great Basin area is characterized by hot, dry summers with sporadic, but generally small, rain events and cold winters with most of the significant annual precipitation falling as snow in the winter and rains in the spring (Dobrowolski et al., 1990). This winter and spring precipitation is primarily responsible for the annual soil moisture recharge of the soil profile, which leads to most of the annual plant growth when temperatures permit in the spring and early summer (Caldwell, 1985). These plants showed only very small responses to sporadic summer rainfall events (Ivans et al., 2003). Soil in the study area is a fine, loamy, mixed mesic Aridic Calcixeroll of the Taylorsflat series to at least 3 m depth, with low shrink/swell potential (Tooele County, Utah Soil Survey, 2 January 1993).
Roots were examined using 24 minirhizotron clear cellulose acetate butyrate (CAB) tubes inserted at an angle of 30° from vertical in the three monospecific vegetation sites (eight tubes per site). These tubes were installed in 1999 to a vertical depth of 120 cm. The portion of the minirhizotron tube extending above the ground was painted and capped to prevent rainfall and to prevent light from entering the tubes. Tubes were anchored to prevent horizontal or vertical movement and a 5-cm diameter (outside diameter) rubber ring was placed around the tube flush with the ground to prevent preferential water flow down the outside of the tube following rain events. Roots were examined in 1999 and 2000, but root colonization during those years was insufficient for analysis. Therefore, only 2001 and 2002 data are presented and disturbance effects from tube installation should have been minimal. Images were taken every 2 wk throughout the spring (March–June), and less frequently during the summer months (July–August, every 3–4 wk) and then again every 2 wk during the autumn months (September–October) with a camera and digitizing software at depths of 5, 10, 15, 30, 45, 60, 90 and 120 cm (Bartz Technology Co., Santa Barbara, CA, USA).
Digital images were processed using a Windows-based root tracing software (rootracker; Duke University, Durham, NC, USA). Root length and diameter were determined by cursor tracings calibrated against a standard image at each date. We evaluated individual observation frames of 226 mm2 in triplicate for each depth for a total of 192 frames per sampling date. In total, we identified 1142 Bromus, 399 Agropyron and 196 Artemisia root segments, or branches from segments, although actual branching was seldom observed in these small frames. Each visible root segment was classified as present or absent, the latter designating disappearance of the root segment. In our study, root color changes were minimal, only progressing from white to brown. None of the roots turned black, although this color transition was reported by Comas et al. (2000) and Anderson et al. (2003) in other systems.
Precipitation was measured adjacent to the three sites (< 50 m) using an automated rain gauge associated with a weather station (Campbell Scientific, Logan, UT, USA). Soil water potential was also recorded hourly throughout this study using individually calibrated screen-cage thermocouple psychrometers (JRD Merrill Specialty Equipment, Logan, UT, USA; Wescor, Logan, UT, USA) connected to a Campbell Scientific data logger (model CR7; Logan, UT, USA). Psychrometer arrays were established in each of the three vegetation types in three trenches per site containing two to three psychrometers at depths of 30, 45, 60, 90 and 120 cm in each trench. Psychrometers were installed horizontally into the trench wall (c. 15 cm) and refilled with soil. Monitoring depths less than 30 cm was not possible because of diurnal temperature fluctuations (Rundel & Jarrell, 1989). For simplicity, a weighted mean soil water potential from the measurements at the five depths is presented for each sampling date for each species. The soil water potential value at each depth was weighted by the average number of roots across all tubes at the corresponding depth.
Individual root characteristics A mixed model analysis was used to examine root segment length per frame area and root segment diameter. Species was used as the fixed effect while year, tube and depth nested within each tube, were random effects to account for non-independence using the compound symmetry covariance matrix. Root-length distribution is presented as a relative increment of root length with increasing depth. All pairwise comparisons were adjusted for experiment-wise error using Tukey's adjustment.
Survival (persistence) analysis Owing to the limitation of root classification of dead or alive roots, as discussed earlier, we conducted a survival analysis on root persistence rather than actual root life spans. The two analyses we conducted on root persistence were (1) to examine median persistence time of root segments and (2) to determine the effects of continuous variables that may influence root persistence in the soil. For the first analysis, we modeled the median root persistence times (in days) of all roots pooled across all tubes using the Kaplan–Meier product-limit method (Kaplan & Meier, 1958). Since some root segments were present at the start of measurement, at the end, or both, the actual residence times of some roots were unknown. These data are considered to be censored (Allison, 1995; Black et al., 1998). For these root segments, those that were present for the duration of the study were designated as ‘interval censored’, those root segments that were present at the start of the season, but disappeared before the end of the season were designated as ‘left censored’ and roots that appeared after the start of the season, but were still present during the last measurement were ‘right censored’. We used maximum likelihood estimation for its appropriate treatment of censored data (Lee, 1992).
In our second persistence analysis, we used Cox proportional hazard regression (Cox, 1972) to model persistence as a function of depth, root diameter, date of first appearance and the change in soil water potential between the current and preceding sampling interval for roots pooled across all tubes. We used change in soil water potential since, unlike the other parameters, water potential is comparatively dynamic and tends to change between sampling intervals. Individual root segments were followed for the duration of the study. Roots were marked as present or absent at each time-period and a persistence time calculated for each root for each time period. Each root ‘birth’ was marked as time t0 and root persistence was calculated for each time period. If a root did not disappear by the end of our measurement period, it was considered ‘right censored’ in our model (as already described). Since all roots were adjusted to time t0, in effect ignoring when the root was produced, we used the day of the year when the root first appeared as a covariate effect in our model. In addition, since a large number of roots were already present in our rhizotron images at the start of each measurement period, real persistence times could not be calculated for these roots. However, we conducted our analysis with and without these roots and found no difference in median persistence times or our time-dependent covariate parameter estimates.
A complete discussion of survival (persistence, in this study) analysis using time-dependent covariates can be found in Allison (1995), and biological examples of their use in Wells and Eissenstat (2001) and Anderson et al. (2003). To facilitate interpretation, the hazard risk in our Cox proportional hazard regression refers to the instantaneous risk of root disappearance, or mortality in survival analysis, at time t. We used proc phreg version 8.02 (SAS Cary, NC, USA) to estimate the parameter β, which corresponds to the magnitude of the covariate effect on the risk of disappearance (Allison, 1995). Negative parameter estimates indicate that increasing values of the covariate are associated with a decreasing risk of disappearance. Alternatively, a positive parameter estimate indicates that increasing values of the covariate correspond to an increasing risk of disappearance. Another useful statistic obtained in this analysis is the risk or hazard ratio, which can quantify the proportional risk as a function of a one-unit increase in the covariate. For quantitative covariates, we subtracted 1 from each risk ratio and multiplied by 100 to get the per cent change in the hazard for each one-unit increase or decrease in water potential change (MPa), diameter (mm) and depth (cm) (Allison, 1995). In addition, to illustrate the effects of the covariate graphically, we modeled the response of 45-d-old root for covariates that were significant across all species within the range of observed measurements (change in soil water potential and date of first appearance) using the baseline command in proc phreg. This illustrates the effects of the changes in the continuous covariates for an even-age root for each species. For example, Bromus roots might experience a drop in soil water potential of −1 MPa between two measurement periods; we therefore modeled the probability of survival from a −1 to +1 MPa change in water potential for a 45-d-old root holding all other covariates at their respective mean. For the timing of root production (day of year), roots appeared continually for the perennials, but no new roots were produced for Bromus after day 155, therefore Fig. 5c is truncated at that time for Bromus.
Our study site received below-average precipitation in both 2001 (223 mm) and 2002 (187 mm) compared with the long-term mean of 264 mm from the Vernon (UT, USA) weather station (40°05′ N, 112°27′ W, 1671.8 m elevation), approx. 30 km south of our field site. However, the temporal distribution within these years was significantly different. In 2000–01, winter–spring precipitation (November–May) was 105 mm, considerably below the 40-yr mean of 127.6 mm (t39 = 7.5, P < 0.001). While in 2001–02, winter–spring precipitation was 160 mm, significantly greater than 127.6 mm (t39 = 5.9 P < 0.001). Summer precipitation (June–August) in both years was less than that of the long-term mean of 65 mm leading to the low annual values mentioned above. Yet, even summer precipitation was different between the two study years, with only 13 mm falling after the wet winter–spring in 2002, but 52 mm in 2001.
Mean root segment length per frame differed among species (Fig. 1, F2,3806 = 469.22, P < 0.0001). Bromus had the greatest root length followed by Agropyron and Artemisia. Within all species, mean root length per frame was greater in 2002 than in 2001 (Fig. 1, F5,3806 = 26.84, P < 0.0001). Root diameters among the species were opposite in pattern compared with that of root lengths, with Artemisia having the largest diameter roots, and Bromus having the smallest (Fig. 1, F2,3806 = 469.22, P < 0.0001). Within species, Agropyron and Artemisia had significantly greater root diameters in 2002 than in 2001, while Bromus root diameters were not significantly different between years. Root length distribution with depth, expressed as a proportion of the total, differed among species (F14,19000 = 6.52, P < 0.0001). The relative root distribution in the upper soil layers was similar among the three species, but relative root length differed among species in deeper soils (Fig. 2). All three species had 50% of their root length above 20 cm. At greater depths, however, Artemisia had close to 10% of its total root length below 100 cm, whereas the two grasses had less than 5% of total root length below 100 cm. Root length distribution with depth also varied between years among the species (F37,19000 = 10.6, P < 0.0001). For all three species, roots penetrated to a greater depth in 2001 than they did in 2002. For example, 75% of the total root length in Bromus was above 50 cm in 2001, but above 30 cm in 2002. Within each year, however, Bromus showed no change in root depth distribution (F8,13000 = 1.3, P > 0.05), while both Agropyron and Artemisia fine roots were present at greater depths as the growing season progressed (Agropyron F8,4573 = 2.2, P < 0.02; Artemisia F8,1353 = 3.1, P < 0.001).
The standing crop of root segments pooled across all depths varied greatly among species (Fig. 3). Bromus had in excess of 50% more root segments than Agropyron and approx. 80% more root segments than Artemisia. The standing crop of root segments for all species increased dramatically early in the season when soil water potentials were the greatest. As the soil dried throughout the summer, the new root production decreased. Agropyron and Artemisia showed almost no new fine roots in the autumn when water potentials began to slowly rise following the dry late summer period. For the two intervals between the last three measurement periods (2001: 8 and 16 August and 26 October; 2002: 14 and 26 September and 19 October), Agropyron produced 13 and nine new root segments and Artemisia only four and five new roots in 2001 and 2002, respectively. Bromus had 67 and 65 new root segments for those same dates in 2001 and 2002, respectively, owing to germination of seeds and establishment of seedlings. However, these new roots contributed little to the overall root length, as changes in cumulative root length were similar for all dates (Fig. 2). The number of disappeared roots remained highest during the second and third measurement periods of 4 May and 22 May and decreased as the 2001 season progressed, whereas root disappearance in 2002 remained low in the spring, and increased as soil moisture declined through late spring and summer.
The patterns of fine root persistence in the soil also differed among species. Median persistence times in 2001 for Bromus and Agropyron were not statistically different, 51 d and 45 d, respectively, while Artemisia median fine root persistence time was only 33 d, significantly shorter than that of Bromus and Agropyron. Median persistence times were approx. 20 d longer in 2002 for all species, and all species were significantly different from each other (Bromus 77 d; Agropyron 65 d; Artemisia 55 d).
When incorporating the covariate effects on root persistence, the grasses had similar persistence curves in both years. However, persistence of Artemisia roots was markedly lower than that of the grasses in both years (Fig. 4), consistent with the shorter median residence times. Root diameter, change in soil water potential and date of first appearance all had significant effects on fine root persistence (Table 1). Root diameter was significantly related to increased persistence for larger diameter roots of Bromus and Agropyron but not Artemisia. The relative increase in persistence for an 1-mm increase in root diameter ranged from 68% for Bromus in 2002–96% for Agropyron in 2002 (Table 1). Therefore, a 0.1 mm increase in root diameter would decrease the risk of disappearance by 6.8% and 9.6% for Bromus and Agropyron, respectively. Changes in soil water potential significantly affected root persistence each year for all species with the exception of Artemisia in 2002. The decrease in risk of disappearance for a 1-MPa increase in water potential ranged from 45 to 99%. For each day later in the season that a root appeared, the risk of disappearance increased for all species, from 0.8% to 3.1% per day. For example, the risk of disappearance increased in our 2-wk sampling interval by 11.2% and 42% (e.g. 14 d × 0.8% and 14 d × 3.1%).
Table 1. Factors correlating with fine root persistence in the soil for individual roots using a proportional hazards regression
Species, year and effect
Bold indicates significance at probability less than 0.05.
Root diameter (mm)
Water potential change (MPa)
Date of first appearance
Root diameter (mm)
Water potential change (MPa)
Date of first appearance
Root diameter (mm)
Water potential change (MPa)
Date of first appearance
Root diameter (mm)
Water potential change (MPa)
Date of first appearance
Root diameter (mm)
Water potential change (MPa)
Date of first appearance
Root diameter (mm)
Water potential change (MPa)
Date of first appearance
To illustrate covariate effects more clearly, we modeled the response of a 95-d-old root for the significant diameter effects and a 45-d-old root for the change in soil water potential and date of first appearance covariate effects (Fig. 5). We modeled a 95-d-old root rather than a 45-d-old root for the diameter effects to illustrate the diameter effects more clearly because the likelihood of persistence remained above 90% for roots of this age. The diameter effects were significant for the two grasses and showed an increase of up to 30% in the likelihood of persistence for the range of diameter measured (c. 0.1–1 mm), although overall the likelihood of persistence for a 95-d-old root remained above 75% for both grasses (Fig. 5a). The largest effect on persistence for 45-d-old roots came from changes in soil water potential and date of first appearance (Fig. 5b,c). Bromus, in both years, and Artemisia in 2001 were strongly influenced by changes in soil water potential. Increases in water potential resulted in more than 80% increase in likelihood of persistence, but with a decrease in water potential, persistence dropped precipitously. All three species were strongly influenced by date of first appearance, with high probabilities of root disappearance during the summer months.
Root persistence in the soil is determined by its active lifespan and, once dead, the decomposition rate. As posited at the outset of this work, based on plant life history or individual fine root characteristics (i.e. diameter), we expected root persistence to follow the sequence of Artemisia > Agropyron > Bromus. Other studies under more mesic conditions have shown that larger-diameter roots persist longer (Eissenstat et al., 2000; Anderson et al., 2003; Matamala et al., 2003) and evergreen species have longer-lived roots than co-occurring deciduous species (Black et al., 1998; Matamala et al., 2003). Surprisingly, our results were opposite for both the diameter and life-history predictions under the prevailing soil conditions. Therefore, direct comparisons of physical root characteristics across life forms should be made with caution. Nevertheless, within a species, larger-diameter roots were generally more persistent (Table 1, Fig. 5; Wells & Eissenstat, 2001).
Fine root longevities, assessed with minirhizotrons, range from days to years (Hendrick & Pregitzer, 1992; Eissenstat et al., 2000). Although there are several hypotheses that relate to root tissue lifespan (Ryser, 1996), very few published studies have examined the factors that may be directly correlated with mortality or disappearance of individual roots (Wells & Eissenstat, 2001; Anderson et al., 2003). By using a Cox proportional hazards model, we were able to identify several covariates that correlate with fine root persistence while controlling for the effects of all other covariates (Table 1). In all species, a change in soil water potential and date of first appearance were significantly related to fine root persistence, such that negative changes in soil moisture were coupled with disappearance of roots and earlier production of roots contributed to higher probability of persistence (Fig. 5). However, separation of soil water potential and date of first appearance effects may be difficult primarily because plants experienced lower soil water potentials at progressively later dates in the year. Agropyron was least influenced by changes in soil moisture, suggesting other factors may affect root death and decomposition including, but not limited to, nutrients (Schlapfer & Ryser, 1996), temperature (Pregitzer et al., 2000) and/or tissue composition (Ryser, 1996). Nevertheless, the patterns were consistent for all species in all years with the exception of Artemisia in 2002, for which soil water potential effects were not significant.
In this water-limited environment, we predicted soil water to influence fine roots such that roots would be most abundant during favorable soil moisture, and roots to disappear during declining soil moisture. In general, this pattern was realized here (Fig. 3 and Fig. 5b) and demonstrated in other arid systems using minirhizotrons where a positive relationship between fine root length and soil moisture was evident for several shrub species (Reynolds et al., 1999; Wilcox et al., 2004). However, a considerable number of roots remained in the soil throughout the summer months (Fig. 3). The persistent roots during the late summer in all three species might be due, in part, to slow decomposition in the dry soils at that time of year. Since the perennial species, and particularly Artemisia, depleted soil moisture much more effectively than the annual grass (Fig. 3), greater root persistence might be expected in the perennials. Yet, roots of the annual grass were clearly most persistent. Trumbore & Gaudinski (2003) suggest that many fine roots are likely to die quickly and disappear, while others persist for much longer periods of time. In our system, large numbers of roots in all species disappeared in the late spring, when soils were still comparatively moist, lending to the observed median fine root residence times ranging from 1 to 2.5 months.
Fine root morphology is strongly affected by soil characteristics (Pregitzer et al., 1993; Eissenstat et al., 2000). In cold desert systems, the availability of water to plants largely depends on the placement and activity of their root systems (Fernandez & Caldwell, 1975; Schenk & Jackson, 2002). All three species in our study showed fine roots at greater depths in 2001 than in 2002. We saw no differences in the depth of infiltration of soil moisture in our soil water potential data; however, the greater duration of soil moisture in the spring and early summer of 2002 may have been responsible for the differences in root distribution with depth between years. In both years, the perennials showed a progressively deeper penetration of roots as the soil dried. For mature aridland plants in the field, Wan et al. (2002) demonstrated an increase in root length density at greater depths for Gutierrezia sarothrae as upper soil layers dried and Fernandez & Caldwell (1975) showed progressively deeper root growth of A. tridentata ssp. wyomingensis and two other shrub species as water was removed from the upper soils layers with progression of the season.
Aridland species differ in their ability to utilize temporally variable soil moisture. The most reliable source of moisture in cold deserts is the winter–spring recharge, which coincides with the most active period of above-ground growth for these plants (Caldwell, 1985). This period also coincides with the greatest below-ground activity for all three species we studied. For differences in soil moisture resulting from winter and spring precipitation in the 2 yr, we found clear morphological differences in these root systems. Fine root length was greater in 2002 for all three species, and mean root diameter was greater for the two perennials than in 2001 (Fig. 1). We did not expect to see any response of roots to the differences in summer precipitation in our study even though the magnitude of summer rains varied greatly between the 2 year (52 mm in 2001 vs 13 mm in 2002). Most summer precipitation events were small (< 15 mm) and rarely penetrated the first few cm of soil (Sala & Laurenroth, 1982), and the phenology of the species is such that Bromus endures this part of the season in seed form, while Agropyon remains largely dormant (Bilbrough & Caldwell, 1997) and Artemisia has much reduced physiological activity (Evans & Black, 1993). Ivans et al. (2003) found no morphological responses to either small (5 mm) or moderate (15 mm) simulated rainfall events in the field for Agropyron and Artemisia.
Over the 2 yr of this study, the difference in timing of annual precipitation and associated soil moisture was likely the major factor causing differences in rooting depth distributions, changes in root absorptive area (length and diameter) and persistence between years for these species. What we do not know is how large a role the differences in soil moisture generated by the different plant species themselves played in determining differences in fine root residence times among the species. Even though these three nearly monospecific stands received the same precipitation, the root systems did not experience the same soil water potentials during much of the warm season. This is because of large differences in water extraction by these species (Fig. 3). Artemisia continues to extract water during the summer and can draw soil moisture potentials down to −5 MPa (DeLucia & Heckathorn, 1989; Ryel et al., 2003) whereas Bromus is apparently unable to withdraw moisture below c. −1 MPa and then rapidly senesces when this threshold is reached in the upper soil layers in the late spring (Klemmedson & Smith, 1964; Rice et al., 1992). In turn, the fine roots of these three species experienced different water potentials during the late spring and summer, which may have a bearing on their residence times. Had we artificially controlled soil moisture in these experiments so that all species experienced the same conditions, the residence times observed for the three species might well have followed different patterns. Nevertheless, as mentioned earlier, surprisingly, the annual grass had the most persistent roots, even though the perennials deplete moisture to a greater degree in the warm season, which should have contributed to slower decomposition rates. The goal of our study was to observe differences in root persistence in the natural environment, including that created by the root systems.
Our initial hypotheses hinged on the premise that life history differences would determine the differences in fine root morphology and persistence. Indeed, we did see differences in morphology (length and diameter) that followed our expectations. However, the pattern of root persistence was contrary to our predictions, since the annual species had the longest root residence times in the wet year, 2002. In the drier year, median root persistence of the two grass species was not significantly different, but the rank order remained consistent between the 2 yr. The two perennials had larger diameter roots, as expected, but they exhibited shorter persistence times than the annual. The balance between root production during favorable soil moisture and disappearance/decomposition during declining soil moisture relates directly to carbon turnover and ecosystem water use. Since arid systems are primarily limited by moisture, it was not surprising to see that both root production and disappearance were apparently related to fluctuations in soil moisture. Despite differences among species in morphology and persistence, we found remarkable similarities for all species in response to differences in annual precipitation. Root persistence, placement at different depths and morphology (length and diameter) all changed similarly among the species in response to the magnitude of winter/spring precipitation. The overall root plasticity in response to water availability likely plays a key role in the success of each of these species; however, in this arid system, predicting fine root persistence across species of different plant life forms based on life form or individual root characteristics (e.g. diameter) remains difficult.
This work was funded by the National Science Foundation (DEB-9807097) and the Utah Agricultural Experiment Station. We thank Ann Mull, Polly Squires, Jamila Squires, Anna Vedina for excellent field and laboratory assistance and Darrell Johnson for allowing us to establish permanent plots on his private rangeland in Rush Valley, UT, USA. Thanks also to Eddy Berry for discussions on and help with the survival analysis. Dave Eissenstat and two anonymous reviewers provided very helpful comments to improve this manuscript.