1Resource pulses generally result in a burst of biological activity at multiple scales. For plants, the increased activity is generally considered positive due to an overall up-regulation of physiological activity during the pulse. Longer-term effects remain an understudied aspect of resource pulses.
2We monitored the short- and long-term effects of nitrogen (N) pulse to the long-lived desert perennial, Cryptantha flava. One group of plants were treated with a one-time application of N in the spring of 1999, a second group received two N pulses (one in the spring of 1999 and one in the spring of 2000), and a third group received ambient N (controls).
3In the short-term, N-pulse treated plants rapidly increased leaf N concentrations, which in turn increased physiological activity and growth. But these responses were mediated by the availability of precipitation.
4In a year with above-average precipitation, all plants increased in size, but N-treated plants grew more and had higher reproductive outputs than control plants. However, when the N pulse was followed by below-average precipitation in the next year, plants with the highest growth rates due to N pulses experienced greater reduction in size and reproduction coupled with increased mortality rates relative to controls.
5At the population level a matrix model showed higher population growth rates in wetter years for N-treated plants compared to controls, but in drier years, N-treated plants showed lower growth rates. Size hierarchies were restructured as a result of the combination of variability in precipitation and N pulses creating more even size distributions.
6Synthesis. The high degree of spatial heterogeneity of N pulses offers opportunities for enhanced growth and reproduction to individuals of C. flava within the larger population. Small plants with access to high soil N were able to maintain high survival, growth rates and reproduction as long as precipitation was adequate. However, increased N inputs probably resulted in a trade-off between reproduction and survival in small plants when precipitation was limiting. The high degree of unpredictability of resources in time and space ultimately contributes to the size hierarchies in this population and the variability in population growth rates.
In natural systems, resources are rarely supplied uniformly. Arid systems in particular are noted for the spatial and temporal variability in soil resources, such as nitrogen (N) and water. Arid land soil nutrient concentrations may vary widely at spatial scales of only a few centimetres (Burke 1989; Jackson & Caldwell 1993). Local disturbances such as earthworm casts (Zaller & Arnone 1999), large mammal deaths and animal excretions (Afzal & Adams 1988) can increase the spatial heterogeneity of soil nutrients. Contributing to this spatial heterogeneity, coefficients of variability in seasonal precipitation often exceed 100% of mean monthly precipitation (Ehleringer 1985). In cold deserts, seasonal variability in precipitation combines with snowmelt patterns to generate nutrient pulses through transient increases in rates of mineralization and nutrient release from over-winter storage (Burke 1989; Bowman 1992; Gallardo & Schlesinger 1992).
In habitats with low average soil nutrient concentrations, nutrient pulses may comprise the dominant fraction of annual nutrient supply for plants (Campbell & Grime 1989). Therefore, plants in these habitats with high temporal and spatial heterogeneity in soil nutrient availability have evolved mechanisms to exploit pulses of soil nutrients. Mechanisms include increased fine root growth (Jackson & Caldwell 1989; Caldwell et al. 1992; Larigauderie & Richards 1994), and/or increased rates of nutrient uptake for a given root area in response to microsite nutrient enrichment (Jackson et al. 1990; Caldwell et al. 1992) in perennial shrubs of the Great Basin Desert, USA.
Just as plants have different capabilities for N uptake, they also have differential capabilities for N utilization for growth and reproduction (Yoder & Caldwell 2002; Peek & Forseth 2003a). Although a few studies have examined the use of short-term supplies of N, fewer studies have measured the longer term consequences of resource-pulse exploitation on plant performance. This may be due to the general assumption that acquisition and uptake of nutrient pulses are universally beneficial to plants. However, research on N saturation of ecosystems due to anthropogenic emissions of N has demonstrated that plant species adapted to low soil nutrients may be at a competitive disadvantage under conditions of greater N availability (Vitousek et al. 1997). Although research on N saturation is concentrated on the effects of constant, higher levels of N availability, the timing of N pulses in relation to plant phenology and the supply of other resources may also have unforeseen consequences for plant population processes. For example, Schwinning et al. (2005) demonstrated that spring and summer N pulses did not result in increases in physiology or growth in some plant species, probably because of simultaneous water stress.
In order to explore the population-level responses of plants in a resource pulse-dominated system, we used matrix models to examine the consequences for population size and stage distribution that occur when individuals of a short-lived desert perennial are treated with nutrient pulses. Matrix projection models use a matrix of transition probabilities among different recognizable plant stages to project the change in population numbers over time (Caswell 2001). These models identify important life-history components that influence the growth or decline of a population (Silvertown et al. 1993; Caswell 2001). Since herbaceous perennials go through several recognizable stages during their lifetime, matrix models may be particularly useful in elucidating critical factors in their population persistence and dynamics.
The presence of nutrient hot spots within root uptake zones increases the growth and reproduction of individual plants of Cryptantha flava (Peek & Forseth 2003a). Pre-reproductive plants not only grow faster in response to pulsed N, but they are also induced to flower earlier and produce more floral stalks than plants not exposed to enhanced N (Peek & Forseth 2003a). These initial studies were restricted to small, pre-reproductive plants to minimize complications due to plant size. However, reproduction in C. flava is positively related to plant size (Casper 1996). We hypothesized that the response to N pulses will differ between size classes, with larger, older plants having less of a response than smaller, younger plants. We further hypothesized that the aggregate responses of individuals would result in higher rates of population growth relative to a population that did not experience pulsed N enhancement. To test these hypotheses, we exposed individual plants of differing sizes to N pulses over a 2-year span. We continued observations of plant growth and reproduction over 3 years, during which monthly precipitation varied year-to-year (Peek & Forseth 2003b). Therefore, we tested the post hoc hypothesis that N uptake, and therefore benefits to the plant, would differ in years with different amounts of precipitation. The study addressed the following questions: How does the presence of N hot spots affect individual plant growth and reproduction in small and large plants within the population? What effects do these individual plant responses have on population stage structure and growth in future years? How does year-to-year variability in rainfall affect population structure and dynamics with and without N pulses?
study species and system
Crypthantha flava (A. Nels.) Payson (Boraginaceae) is a herbaceous perennial that is the most widespread species within the genus Cryptantha. The centre of diversification of this genus is in the Colorado Plateau region, where mountainous dispersal barriers combined with glacial and interglacial events cause cyclic expansion and contraction of desert regions, producing adaptive radiation of numerous taxa (Axelrod 1950; McLaughlin 1986). Approximately 39 perennial species of Cryptantha inhabit the intermountain west region between the Colorado Rockies to the east and the Sierra Nevada and Cascade mountains to the west with a number classified as rare and protected (Higgins 1971; Cronquist et al. 1984). Many of the recognized Cryptantha species are endemic and are restricted to specific substrate types (Cronquist et al. 1984). Because of its herbaceous life-history properties and widespread distribution, we chose C. flava as a study system with which to examine the population-level effects of spatial N heterogeneity. This species was also attractive as a model system because of previous work on its breeding system (Casper & Wiens 1981; Casper 1982), physiology (Forseth et al. 2001; Peek & Forseth 2003a,b; Peek & Forseth 2005), and population response to drought (Casper 1996).
Cryptantha flava occurs on sandy, low-nitrogen soils throughout the Colorado Plateau region. Its growth habit consists of a woody, below-ground caudex that supports rosettes of nearly vertical leaves. Extending from the woody caudex is a single taproot that may grow up to 1 m in length (personal observation). Numerous lateral roots extend from the single taproot in the upper 20–30 cm of the soil and can reach a radius of one meter or more from the above-ground canopy, even in small plants (Peek & Forseth 2005). A leaf rosette dies after producing a single flowering stalk. New rosettes are produced from axillary meristems on expanded rosettes. Above-ground vegetative growth begins in late March, flowering in mid-May and seeds are set by mid-July. Senescence of most rosettes occurs during the hot, dry summers that characterize the Colorado Plateau, followed by some vegetative growth and seedling germination in the fall.
This research was conducted at a site in Uintah County in northeastern Utah (1730 m a.s.l., 40°30′N, 109°22′30″E). Vegetation is dominated by sagebrush, Artemisia tridentata Nutt., rabbitbrush, Chrysothamnus nauseosus (Pallas) Britt., and Utah juniper, Juniperus osteosperma (Torr.) Little. The study area is characterized by substantial environmental variation, both seasonally and spatially. Annual precipitation averages 208 mm and is highly variable (coefficient of variation of monthly rainfall ranges from 50% to 125% of the mean). During this study, above-average precipitation fell in 1999, while 2000 was a year of below-average precipitation (Peek & Forseth 2003b). A majority of the annual precipitation in 1999 fell during the spring growing season, while the final three months in 1999 received almost no precipitation, followed by a dry spring in 2000. Precipitation in 2001 followed similar patterns as in 2000, with less than average annual precipitation and a drier than average winter and spring. Thus, winter recharge of soil-moisture profiles and spring precipitation essential to this system (Dobrowolski et al. 1990) were both well below normal for the growing season of 2000 and 2001.
Experimental plants were chosen based upon their size, reproductive status and spatial location. In early spring 1999, one thousand plants were marked and their vegetative and reproductive status was recorded from 1999 to 2001. Plants were divided into three treatment groups: one group receiving a one-time application of N in the spring of 1999, one group receiving two N pulses (in spring 1999 and spring 2000, respectively), and the third group receiving ambient N (controls). All N-pulse treated plants received a 1-L application of 5.3 g urea N dissolved in distilled water, along with all fecal pellets from a nearby single mule deer excretion. Pellets were constrained to an area within a 5 cm radius of the plant's above-ground rosettes, while the 1-L N solution was poured directly over the pellet group in the same 10 cm diameter circle. This N-pulse treatment was designed to simulate a mule deer (Odocoileus hemionus) excretion event, which are common in this habitat (Peek & Forseth 2003a). Background N concentrations in the soil range from 1 to 20 µg N g−1 soil with mineralization rates supplying similar concentrations per day, but these rates are highly variable seasonally and spatially (Peek & Forseth 2003b). In 1999, 670 plants (control) did not receive N pulse, whereas 330 plants were treated once with the N pulse described above. In the following year (2000), half of the N-pulse treated plants from 1999 received an additional N pulse (2 × N, n = 165), whereas the other half were left untreated (1 × N, n = 165). A subset of the control plants (n = 330) were treated with deionized water to identify any water addition effects from adding 1L of the urea–water mix, but these treatments were combined in the results due to the absence of any water addition effects (Peek & Forseth 2003a,b). The experimental design allowed us to examine plant performance in response to N pulses in the year following treatment as well as 2 years after the initial, one-time treatment. It also provided us with a comparison between plants receiving multiple N pulses compared to those with a single pulse event.
Leaf tissue was collected for measurement of N concentrations at various times during the 1999 and 2000 seasons from different subsets of plants to minimize destructive harvesting. Leaf tissue was dried to a constant mass in an 80 °C oven, and ground in liquid N2 to a fine powder with a mortar and pestle. Nitrogen concentration was measured on a percentage dry weight basis via combustion in a Perkin-Elmer 2400 Series CHN analyzer (Norwalk, CT).
Demographic measurements were collected on all treatment plants for the 3 years of the study. We counted the total number of rosettes, which is positively correlated with above-ground biomass and number of flowering stalks on each individual. An annual census was performed once in June of each year, as rosette numbers do not change within a growing season (Casper et al. 2001). We calculated size-specific growth rates as: ([Rosette numbert – Rosette numbert–1]/Rosette numbert–1).
We also randomly selected individual flowering stalks from the control (n = 15) and N-pulse treatments (n = 15) and counted the number of flowers per stalk and the number of mature nutlets per flower in 1999 and 2000. Unlike growth responses, we found that seed development was a function of fertilization in the current year; we therefore grouped plants into two treatments (those that received a N-pulse in the current year (N) and those that received no supplemental N) when analysing seed set for each treatment every year. We also noted mortality rates between 1999–2000, and 2000–2001 for all plants.
Plant stages were assigned using reproductive traits and plant size. We defined three stages of the life cycle for C. flava: a non-reproductive seedling stage (SN) from germination to first age of reproduction (typically 1–4 years, Casper 1994), a small reproductive stage (SR), and a large reproductive (LR) stage. We divided reproductive plants into two size categories based on the number of expanded rosettes and to maximize sample size for each transition element. Small plants were designated as having < 12 rosettes, large plants as having 12 or more rosettes. Our matrix model is highly connected, with transitions possible between every stage. We assumed no seed bank by incorporating seed production and seedling recruitment into all transitions generating small non-reproductive (SN) plants (Table 1).
Table 1. Individual elements in projection matrices representing the life cycle of Cryptantha flava from 1999 to 2001 for control plants (CONT), N-pulse treated plants in 1999 and 2000 (2 × N), and N-pulse treated plants in 1999 only (1 × N). Stage abbreviations are SN, small non-reproductive; SR, small reproductive; LR, large reproductive
2 × N
1 × N
SN – SN
SN – SR
SN – LR
SR – SN
SR – SR
SR – LR
LR – SN
LR – SR
LR – LR
Over the three years of the experiment all transition elements were measured. We hypothesized that seedling survivorship was low and set the probability of surviving to 0.001 to obtain a population growth rate not significantly different from stasis for the control population during the 1999–2000 transition (Table 2). We did this to examine the effect of the N-pulse treatments relative to controls, not to obtain true population growth rates per se. We used the same seedling recruitment estimate for all treatment matrices, as germination percentages were not affected by N-pulse treatment (M. S. Peek, unpublished data). We considered this to be a conservative approach, since higher N availability may lead to higher quality or larger seeds where greater rates of establishment and growth are associated with larger seed size in many plant species (Baker 1972; Mazer 1989). Differential yearly mortality in the seedling stage may also be dependent upon temperatures and precipitation (Lundholm & Larson 2004) which were unaccounted for in this study. Therefore, we analysed the control matrix by solving for the minimum seedling recruitment values needed to keep population numbers constant. We then compared this minimal transition value among treatments and years.
Table 2. Cryptantha flava's population growth rates and 95% confidence intervals for each transition matrix from 1999 to 2000 and 2000 to 2001. Confidence intervals were obtained with 1000 bootstrapped randomizations
95% Confidence limits
95% Confidence limits
95% Confidence limits
Demographic data were incorporated into a stage-based matrix population model with no seed bank. Transitions (aij's) between stages of the life cycle were summarized in the form: n(t + 1) =A·n(t), where n(t) is a vector of stage abundance at time t, and A is a matrix of aij's that describes how each stage contributes to the number of individuals in all other stages at the next time step. We calculated the dominant eigenvalue (λ) of A, which represents the asymptotic finite rate of increase at the stable-stage distribution for each treatment in each year. We then calculated average survival rates for each treatment by multiplying the stable-stage distribution by the stage specific survival rate, thus creating weighted average survival rates where the weights are the relative proportion of individuals in each size category.
Size, growth rates, flowering and N data were analysed using analysis of variance techniques, where life-history stage and treatment were considered fixed effects. A posteriori contrasts were made using a Tukey adjustment. Mortality counts were analysed using log-linear techniques. Because of the small sample size, 1000 randomizations were conducted, preserving sample sizes within treatments, to compare observed χ2-values to that of the random distribution (Manley 1991). Confidence limits for λ were obtained via bootstrapping methods. The bootstrapping procedure involved generating a re-sampled matrix for each treatment in every year and computing each element and calculating λ, repeating this 1000 times (Kalisz & McPeek 1992; Bierzychudek 1999; Caswell 2001).
A single pulse of urea N produced significant increases in leaf nitrogen (N) concentrations of individual plants (F2,888 = 96.38 P < 0.0001; Fig. 1). In 1999, leaf N remained elevated for much of the growing season for N-pulse treated plants relative to controls. Plants receiving N amendments for both years also showed elevated leaf N concentrations in 2000 compared to control plants. Plants that were treated in 1999 and left untreated in 2000 (1 × N) had a significant treatment by size interaction (F4,888 = 6.42 P < 0.0001). Small 1 × N plants were similar to small 2 × N plants (i.e. higher leaf N concentrations), while large 1 × N plants had lower leaf N concentrations that were not statistically different from control plants.
No treatment effects on plant size were found for small plants in all three years (Fig. 2). Large plants, on the other hand, showed a bias in the initial plant selection, with 1 × N plants 17.8% larger than control plants. However, this difference was magnified to 33.2% in the following year due to the N-pulse treatment (Fig. 2). All large plants shrank in size between 2000 and 2001, with no significant differences between treatment categories. All treatments had positive relative growth rates from 1999 to 2000 and negative rates from 2000 to 2001 (Fig. 3). There were significant main effects of treatment (F1,329 = 20.8, P < 0.0001) and life stage (F2,329 = 3.9, P < 0.03) for relative growth rates from 1999 to 2000. The N-pulse treatment produced significantly greater growth rates for all life cycle categories. Small non-reproductive plants had significantly greater growth rates than large reproductive plants (t = 2.6, P < 0.01). There was also a significant treatment effect on relative growth rate from 2000 to 2001 (F2,353 = 3.5, P < 0.04), although the direction of the effect was reversed from that of the prior year. Both N-pulse treatments had significantly more negative growth rates than control plants for each life cycle category.
The average number of flowering stalks per individual in 1999 was not significantly different between treatments for each size category (Fig. 4). Despite reduced soil moisture contents in 2000 (Peek & Forseth 2003b), all plants produced more flowering stalks than in 1999. Large N-pulsed treated plants produced significantly more flowering stalks in 2000 (9.9 ± 0.6) than did control plants (6.2 ± 0.7). There were no significant treatment effects on flower stalk production for small or large plants in 2001. The N-pulse treatment had a significant effect on the number of seeds produced per flowering stalk (Fig. 5). On average, control plants produced 41 ± 1.7 seeds per flowering stalk, compared to 68 ± 2.5 seeds per flowering stalk for N-pulse treated plants. This difference was due primarily to more seeds per flower and not to the development of more flowers. Multiple seeds within each flower were found for N-pulse treated plants, compared to a preponderance of single-seeded flowers in controls. For seed production in the matrix model in 2000, we assumed that the seed production from the 1 × N treatment would be similar to controls, since embryo production was a function of fertilization in the current year.
Because we assigned a value of 0.001 for seedling recruitment, the control population intrinsic rate of increase was not significantly different from 1.0 in either year (Table 2). The N-pulse treatment, using the same recruitment rate, produced a positive growth rate of 1.18 for the year of above average precipitation. The year of below average precipitation, 2000, produced a λ significantly below 1.0 for both N-pulse treatment populations.
Few plants died from 1999 to 2000, and mortality was not significantly different between stages or treatments (P = 0.78, P = 0.65; Fig. 6). However, mortality from 2000 to 2001 was significantly higher than that from 1999 to 2000, and there was a significant stage × treatment interaction (P < 0.001). The interaction was driven by similar mortality rates for small, non-reproductive plants in the N-pulse treatments, while the 2 × N treatment had higher mortality for both SR and LR stages. Control plants had the lowest overall mortality rates.
The stable stage distribution derived from the transitions from 1999 to 2000 showed that the control population had more plants in the LR category (44%) than either of the two small categories (SN = 36%, SR = 20%) (Fig. 7). The N-pulse treatment had slightly more reproductive plants (SR + LR), 71% compared to 64% in the controls, but was more biased toward large reproductive plants (LR = 58%). The transition from 2000 to 2001 shifted size hierarchies with the control population having the largest proportion of plants in the small reproductive category (47%). The 2 × N plants had a more even distribution, while the 1 × N plants showed the largest proportion in the small reproductive category. Overall, there was a reduction in the proportion of plants in the LR category from the wet to the drier year, whereas the SR category showed increased proportions of plants (Fig. 7). When calculating the average adjusted vital rate (weighting by the stable-stage distribution), in the 1999–2000 transition, the average survival rate of the large reproductive plants, at nearly 60%, was more than twice that of the small categories, SN = 29% and SR = 25%, and 20% higher than that of the large reproductive control plants. In the drier year transition when vital rates were declining, average adjusted survival rates were highest for the small categories in both N-treated plants, with none above 35%. The largest average survival rate was in the control small reproductive category at 42%.
Resource pulses generally result in a pulse of biological activity at multiple scales. The increased activity is generally considered positive for plants by increasing N availability (Cui & Caldwell 1997), water uptake (Sala et al. 1982) and an overall up-regulation of physiological activity (Peek & Forseth 2003a; Schwinning et al. 2003; Huxman et al. 2004). The long-term (> 1 year when the pulse is on the order of days) response to a positive N pulse, however, remains an important under-explored question, particularly when climate change predicts episodic events to change either in frequency and/or magnitude. In an attempt to understand the long-term consequences of an N pulse on plant performance, we found an initial positive individual response through increased leaf N concentrations, but more complex effects due to variations in the timing of precipitation. A one-time application of N increased leaf nitrogen concentrations for both large and small plants in 1999 and 2000, regardless of precipitation. However, the greatest enhancement of growth and reproduction was seen in large plants from 1999 to 2000, a year of above-average precipitation. From 2000 to 2001, a year of below-average precipitation, a reduction in plant size and higher mortality resulted for all plants, and these negative effects were greatest for plants that had received a supplemental N pulse in previous years.
Plant productivity and population growth are often positively correlated with precipitation in arid regions (Archer et al. 1988; Golubov et al. 1999). Our results suggest similar patterns with greater relative growth rates in the wetter year and reductions in plant size in a drier year (Fig. 3). Population growth rates showed similar patterns with larger vital rates in the year with above-average precipitation. Although not explicitly interested in precipitation effects, vital rates and growth showed predictable patterns with these precipitation patterns. However, different patterns may occur if precipitation patterns differ, such as consecutive wet or dry years.
Nevertheless, adding a pulse of N magnified the positive effects of precipitation in the wet year by increasing growth and seed production (Figs 3 and 5) leading to a 20% higher population growth rate (Table 2). This was largely due to the greater proportion of large reproductive plants. In the dry year, population growth decreased 10–15% more relative to controls if previously treated with an N pulse. In part, the reduced growth rates were a result of size reductions as most large plants entered the two small categories. These N-pulse effects highlight the complexity of patterns that can result when looking at two limiting resources.
The analysis in Table 1 assumed that all members of the population are affected equally in the N-pulse treatments, or are not affected at all in the case of the control population. However, Peek and Forseth (unpublished data) found that only 50% of the individuals in a natural population fall within the lateral root spread of naturally occurring N hot spots. The large size variability present in natural populations of C. flava (Peek, personal observation, Casper 1996) could be due in part to this spatial heterogeneity in soil resources. Not only does precipitation change size hierarchies (Casper 1996), but having access to soil N hot spots also produces differences in stage distributions. Kadmon (1993) found similar population-level consequences of environmental heterogeneity in the desert annual Stipa capensis. Although precipitation contributed to the largest changes in plant performance, Kadmon (1993) concluded that microhabitat location was also important, presumably due to the availability of other resources.
While the positive effects of N and precipitation were expected, the negative effects of adding N pulses were surprising. The most striking were the reduction in relative growth rates and increases in mortality. The higher N supplies may have led to higher allocation to flowering and above-ground growth, reducing allocation to maintenance and/or storage. Evans & Black (1993) showed a similar response in Artemesia tridentata, where increased watering did not affect vegetative growth, but did increase current reproductive allocation. Cryptantha flava is restricted in its distribution to sandy, low-N soils. The N-pulse treatments may have cued a premature flowering response before the build-up of the needed carbon resources that normally accompany N uptake. In seasonal environments, such as this one, storage reserves in perennial plants are often high to help mitigate competing demands in times of low resource availability (Chapin et al. 1990). Large plants, with adequate storage reserves of N and carbon may bear the cost of reproduction more easily than small plants (see lower mortality for LR plants in 2001). Small plants may have less flexible allocation patterns than large plants, being seemingly constrained to allocating to reproduction once flowering is initiated. Therefore, high reproductive efforts could come at the expense of below-ground maintenance or storage, since a unit of resource cannot be allocated to all competing functions simultaneously (Bloom et al. 1985).
Alternatively, preferential allocation of resources to roots, to enable better anchoring, water uptake, and nutrient foraging may be very important for seedling establishment, as is the case for the woody perennial, Chrysothamnus nauseousus (Donovan & Ehleringer 1993). Delayed reproduction (1–3 years) in C. flava while roots and storage reserves are built may enhance the survival of young seedlings. Cryptantha flava is an obligate perennial, and no reports of first-year flowering have been made. Therefore, developmental decisions on allocation to reproduction vs. survival are postponed for at least the first year of growth. However, seed set is positively related to plant size in C. flava (Casper 1996). Thus, small plants face a conundrum, where early flowering and fast vegetative growth are both selectively advantageous. The rapid growth and flowering response to increased nutrients and water may be a response to a highly unpredictable resource environment. Therefore, current shoot growth and reproduction may be favoured when resources are available, at the cost of reduced allocation to root growth and storage. We saw no difference in the number of flowers produced for small plants regardless of fertilization treatment, suggesting that once flowering is initiated, a premium is placed on maturing flowers in small plants, rather than aborting them, resulting in the increased mortality in 2001 compared to large plants.
Root : shoot ratio in desert plants is an important factor determining plant size (Donovan & Ehleringer 1994). Larger plants have greater transpiration demands than smaller plants because increased leaf area causes greater evaporation. Therefore, large plants may be constrained in the number of above-ground rosettes (leaf area) that they expand for a given amount of resources. An increase in nutrients allocated to reproduction would lead to a strong increase in flowering in large plants. But in a dry year, the ability of below-ground structures to supply water to numerous vegetative rosettes and floral stalks may be outstripped. Height limitations due to water availability have been hypothesized in large perennial trees (Ryan & Yoder 1997), but may also apply to size in small herbaceous perennials. The greater negative relative growth rates for large plants support the contention that rapid growth under high-N and high-water conditions incurs a later cost during low-water conditions.
In conclusion, the high degree of spatial heterogeneity of N pulses in arid lands offers opportunities for enhanced growth and reproduction to individuals of C. flava within the larger population. The positive or negative consequences of exploitation of soil N patches are mediated by precipitation. We suggest that there is a trade-off between current allocation patterns and future growth and survival. Clear trade-offs are often difficult to demonstrate for several reasons. For example, Reekie and Bazzaz (1987) showed that plants can physiologically compensate for the costs of reproduction because reproductive structures can be photosynthetic. In addition, mobile nutrients (e.g. N) may be used for different functions at different times repeatedly throughout the lifetime of an individual. Also, costs may not be evident due to non-uniform resource supply. Thus, increased resource inputs or favourable microsites may mask any trade-offs as a result of a particular allocation pattern (Tuomi et al. 1982, 1983). We found evidence of this when small plants with access to high soil N were able to maintain high survival, growth rates and reproduction as long as precipitation was adequate. However, increased N inputs most likely resulted in a trade-off between reproduction and survival in small plants when precipitation was limiting. Although these results may be due to the observed precipitation pattern, that is, above-average followed by below-average, the high degree of unpredictability of resources in time and space ultimately contributes to the size hierarchies in this population and the variability in population growth rates.
The authors thank D.A. Wait and H. Kempenich for field assistance. We acknowledge J. Sinclear of the Vernal district BLM for cooperation using the field site as well as Dr L. Squires of Utah State University Vernal Branch Campus for laboratory space. We also thank Dr B. Casper for discussions on the ecology of C. flava. This manuscript was significantly improved by the comments of our editors and two anonymous referees. This work was supported by NSF award IBN95-27833-00.